V. 115, NO.

1
JANUARY 2018

ACI
STRUCTURAL J O U R N A L

A JOURNAL OF THE AMERICAN CONCRETE INSTITUTE

 CONTENTS
STATEMENT OF OWNERSHIP
Board of Direction ACI Structural Journal
President
Khaled Awad January 2018, V. 115, No. 1
Vice Presidents a journal of the american concrete institute
David A. Lange an international technical society
Randall W. Poston

Directors 3 Editors-in-Chief of ACI Journals Named

JoAnn P. Browning
Cesar A. Constantino 4 Letter from the Editor-in-Chief
Frances T. Griffith
H. R. Trey Hamilton
R. Doug Hooton 5 Effect of Geometric Scaling on Shear Strength of Reinforced
Joe Hug Concrete Beams without Stirrups, by Derek Daluga, Kaylor McCain,
Kimberly Kayler Matthew Murray, and Santiago Pujol
William M. Klorman
Neven Krstulovic-Opara 15 Three-Dimensional Grid Strut-and-Tie Model Approach in Struc-
Tracy D. Marcotte tural Concrete Design, by Young Mook Yun, Byunghun Kim, and
Antonio Nanni Julio A. Ramirez
Roberto Stark

Past President Board Members 27 Verification of Three-Dimensional Grid Strut-and-Tie Model Approach
Michael J. Schneider in Structural Concrete, by Young Mook Yun, Hyun Soo Chae, Byunghun
Sharon L. Wood Kim, and Julio A. Ramirez
William E. Rushing Jr.
41 Time-Dependent Buckling Testing of Eccentrically Loaded Slender
Executive Vice President High-Strength Concrete Panels, by Yue Huang, Ehab Hamed, Zhen-
Ron Burg Tian Chang, and Stephen J. Foster
Editorial Board 53 Nonlinear Backbone Modeling of Concrete Columns Retrofitted
Robert J. Frosch, Editor-in-Chief
  Purdue University with Fiber-Reinforced Polymer or Steel Jackets, by José C. Alvarez,
Catherine French Sergio F. Breña, and Sanjay R. Arwade
  University of Minnesota
Michael Kreger 65 
Shear-Friction Strength of Low-Rise Walls with 550 MPa
  University of Alabama (80 ksi) Reinforcing Bars under Cyclic Loading, by Jang-Woon Baek,
David Sanders Hong-Gun Park, Byung-Soo Lee, and Hyun-Mock Shin
  University of Nevada, Reno
James Wight
  University of Michigan 79 Behavior of Straight and T-Headed ASTM A1035/A1035M Bar
Splices in Flexural Members, by Sergio F. Breña, Jeffrey Messier, and
Staff Sean W. Peterfreund
Publisher
John C. Glumb 91 End-Region Behavior of Pretensioned I-Girders Employing 0.7 in.
(17.8 mm) Strands, by J. Salazar, H. Yousefpour, R. Alirezaei Abyaneh,
Engineering H. Kim, A. Katz, T. Hrynyk, and O. Bayrak
Managing Director
Michael L. Tholen
103 Stepwise Bond Model Including Unconfined and Partially Confined
Managing Editor Hooks, by Armin Erfanian and Alaa E. Elwi
Jerzy Z. Zemajtis
113 Seismic Performance of Innovative Reinforced Concrete Coupling
Staff Engineers
Katie A. Amelio
Beam—Double-Beam Coupling Beam, by Youngjae Choi, Poorya Hajy-
Robert M. Howell alikhani, and Shih-Ho Chao
Khaled Nahlawi
Marc M. Rached 127 A New Approach to Modeling Tension Stiffening in Reinforced
Matthew R. Senecal Concrete, by Angus Murray, Raymond Ian Gilbert, and Arnaud Castel
Gregory M. Zeisler
Publishing Services
Manager
Barry M. Bergin Contents cont. on next page

Editors
Carl R. Bischof Discussion is welcomed for all materials published in this issue and will appear ten months from
this journal’s date if the discussion is received within four months of the paper’s print publication.
Kaitlyn J. Dobberteen Discussion of material received after specified dates will be considered individually for publication or
Tiesha Elam private response. ACI Standards published in ACI Journals for public comment have discussion due
Kelli R. Slayden dates printed with the Standard.
Editorial Assistant ACI Structural Journal
Angela R. Matthews Copyright © 2018 American Concrete Institute. Printed in the United States of America.
The ACI Structural Journal (ISSN 0889-3241) is published bimonthly by the American Concrete Institute. Publication
office: 38800 Country Club Drive, Farmington Hills, MI 48331. Periodicals postage paid at Farmington, MI, and at
additional mailing offices. Subscription rates: $172 per year (U.S. and possessions), $181 (elsewhere), payable in
advance. POSTMASTER: Send address changes to: ACI Structural Journal, 38800 Country Club Drive, Farmington
Hills, MI 48331.
Canadian GST: R 1226213149.
Direct correspondence to 38800 Country Club Drive, Farmington Hills, MI 48331. Telephone: +1.248.848.3700.
Facsimile (FAX): +1.248.848.3701. Website: http://www.concrete.org.

ACI Structural Journal/January 20181

 Contributions to
CONTENTS ACI Structural Journal
The ACI Structural Journal is an open
forum on concrete technology and papers
139 Punching Shear Strength of Slabs and Influence of Low Reinforcement related to this field are always welcome.
Ratio, by Susanto Teng, Khatthanam Chanthabouala, Darren T. Y. Lim, and All material submitted for possible publi-
Rhahmadatul Hidayat cation must meet the requirements of
the “American Concrete Institute Publi-
cation Policy” and “Author Guidelines
151 Biaxial Interaction and Load Contour Method of Reinforced Concrete and Submission Procedures.” Prospective
T-Shaped Structural Walls, by Tae-Sung Eom, Hye-Sung Nam, and authors should request a copy of the Policy
Su-Min Kang and Guidelines from ACI or visit ACI’s
website at www.concrete.org prior to
163 Eccentric Punching Shear of Waffle Slab, by Ahmed Faleh Al-Bayati, submitting contributions.
Lau Teck Leong, and Leslie A. Clark Papers reporting research must include
a statement indicating the significance of
175 Modeling Time-Dependent Deformations: Application for Reinforced the research.
Concrete Beams with Recycled Concrete Aggregates, by Adam M.
The Institute reserves the right to return,
Knaack and Yahya C. Kurama
without review, contributions not meeting
the requirements of the Publication Policy.
191 Carbon Fiber-Reinforced Polymer-Strengthened Reinforced Concrete
Beams Subjected to Differential Settlement, by Yail J. Kim and All materials conforming to the Policy
Aiham Al-Kubaisi requirements will be reviewed for editorial
quality and technical content, and every
effort will be made to put all acceptable
203 Tests of Short Headed Bars with Anchor Reinforcement Used in Beam- papers into the information channel.
to-Column Joints, by Ján Bujn 
ák and Matúš Farbák However, potentially good papers may be
returned to authors when it is not possible
211 Evaluation of Flexural and Shear Stiffness of Concrete Squat Walls to publish them in a reasonable time.
Reinforced with Glass Fiber-Reinforced Polymer Bars, by Ahmed Arafa,
Ahmed Sabry Farghaly, and Brahim Benmokrane Discussion
All technical material appearing in the
223 Long-Term Multipliers and Deformability of Fiber-Reinforced Polymer ACI Structural Journal may be discussed.
Prestressed Concrete, by Yail J. Kim and Raymon W. Nickle If the deadline indicated on the contents
page is observed, discussion can appear
235 Seismic Performance of Reinforced Concrete Columns with Lap in the designated issue. Discussion should
Splices in Plastic Hinge Region, by Chul-Goo Kim, Hong-Gun Park, and be complete and ready for publication,
Tae-Sung Eom including finished, reproducible illustra-
tions. Discussion must be confined to the
247 Conventional and High-Strength Steel Hooked Bars: Detailing Effects, scope of the paper and meet the ACI Publi-
by J. Sperry, D. Darwin, M. O’Reilly, A. Lepage, R. D. Lequesne, A. Matam- cation Policy.
oros, L. R. Feldman, S. Yasso, N. Searle, M. DeRubeis, and A. Ajaam Follow the style of the current issue.
Be brief—1800 words of double spaced,
259 
Thermomechanical Relaxation of Reinforced Concrete Beams typewritten copy, including illustrations
Strengthened with Carbon Fiber-Reinforced Polymer, by Yail J. Kim and and tables, is maximum. Count illustrations
Abdulaziz Alqurashi and tables as 300 words each and submit
them on individual sheets. As an approxi-
269 Practicability of Large-Scale Reinforced Concrete Beams Using Grade mation, 1 page of text is about 300 words.
Submit one original typescript on 8-1/2 x
80 Stirrups, by Jung-Yoon Lee, Jae-Hoon Lee, Do Hyung Lee, Seong-Jun
11 plain white paper, use 1 in. margins,
Hong, and Ho-Young Kim and include two good quality copies of the
entire discussion. References should be
281 Use of Anchored Carbon Fiber-Reinforced Polymer Strips for Shear complete. Do not repeat references cited
Strengthening of Large Girders, by William A. Shekarchi, Wassim M. in original paper; cite them by original
Ghannoum, and James O. Jirsa number. Closures responding to a single
discussion should not exceed 1800-word
equivalents in length, and to multiple
discussions, approximately one half of
the combined lengths of all discussions.
Closures are published together with
the discussions.
ON COVER: 115-S22, p. 271, Fig. 1—Test setup and measurement. Discuss the paper, not some new or
outside work on the same subject. Use
references wherever possible instead of
repeating available information.
Discussion offered for publication should
offer some benefit to the general reader.
Discussion which does not meet this
Permission is granted by the American Concrete Institute for libraries and other users registered with the Copyright requirement will be returned or referred to
Clearance Center (CCC) to photocopy any article contained herein for a fee of $3.00 per copy of the article. Payments
should be sent directly to the Copyright Clearance Center, 21 Congress Street, Salem, MA 01970. ISSN 0889-3241/98 the author for private reply.
$3.00. Copying done for other than personal or internal reference use without the express written permission of the
American Concrete Institute is prohibited. Requests for special permission or bulk copying should be addressed to the
Managing Editor, ACI Structural Journal, American Concrete Institute. Send manuscripts to:
The Institute is not responsible for statements or opinions expressed in its publications. Institute publications are not able http://mc.manuscriptcentral.com/aci
to, nor intend to, supplant individual training, responsibility, or judgment of the user, or the supplier, of the information
presented.
Send discussions to:
Papers appearing in the ACI Structural Journal are reviewed according to the Institute’s Publication Policy by individuals
expert in the subject area of the papers. Journals.Manuscripts@concrete.org

2 ACI Structural Journal/January 2018

 EDITORS-IN-CHIEF OF ACI JOURNALS NAMED
ACI has formed Editorial Boards to oversee the ACI
Materials Journal and the ACI Structural Journal. Each
Editorial Board will comprise an editor-in-chief and four
research expert members. The focus of the Editorial Boards
will be on making and keeping ACI’s serial publications the
preferred publishing venue for academic researchers in the
field of concrete. The Boards’ initial missions will include
identifying new topics for commissions and special editions,
advising on the direction for the journals, making sugges-
tions for both subject matter and potential authors, providing
content by writing occasional editorials and other short arti-
cles, and overseeing journal quality.
Jason Weiss, Oregon State University, serves as Editor-
in-Chief of the ACI Materials Journal Editorial Board,
with Board members Zachary Grasley, Texas A&M Univer-
sity; Maria Juenger, University of Texas at Austin; Kamal
Khayat, Missouri S&T; and Michael Thomas, University of
New Brunswick.
Robert Frosch, Purdue University, serves as Editor-in-
Chief of the ACI Structural Journal Editorial Board, with
Board members Catherine French, University of Minnesota;
Michael Kreger, University of Alabama; David Sanders,
University of Nevada, Reno; and James Wight, University
of Michigan.
“The formation of the Editorial Boards for the ACI
Materials Journal and ACI Structural Journal will further
support the advancement of concrete technology and
ACI Structural Journal/January 2018
concrete construction worldwide,” said Khaled W. Awad,
ACI President. “By selecting and publishing the most
current research with the greatest impact on concrete design
and construction, ACI Editorial Boards will provide access
to professionals around the world to the latest information
on concrete and its uses. I would like to take the opportunity
to thank the world-renowned experts who accepted to serve
on the Boards for their passion and dedication to support
ACI in advancing concrete knowledge globally.”
The ACI Materials Journal and ACI Structural Journal
have been publishing research on properties of concrete
materials and design of structural concrete since ACI began
publishing journals back in December 1912. The goal has
always been to ensure that those who design and build
with concrete have access to the most current and cutting-
edge research and information on concrete. In many cases,
research published in ACI’s journals directly impacts the
codes, specifications, and practices published by the Insti-
tute and other standards developing organizations.
To date, more than 4700 papers have been published in the
print and digital editions of the ACI Materials Journal and
ACI Structural Journal. Subscriptions to the digital editions
of both journals are included with ACI membership.
Reprinted from the November 2017 issue of Concrete
International, V. 39, No. 11, p. 12.
3
 LETTER FROM THE EDITOR-IN-CHIEF
This issue (Volume 115, No. 1) marks a new era in the evolu-
tion of the ACI Structural Journal. ACI began publishing in
1905 with the Proceedings of the First Convention of the
National Association of Cement Users, which was held in
Indianapolis, IN. For the next century, ACI has continued
to advance the profession through dissemination of concrete
knowledge and research. This is no small feat and shows the
dedication of ACI to advance technical knowledge in the field
of concrete. Having such a long history is quite remarkable,
considering the changes that have occurred in technology
and society over the 115 volumes of the Journal. To date,
ACI has published more than 9000 papers and discussions
between the ACI Structural Journal, ACI Materials Journal,
and the ACI Journal Proceedings (which predated the sepa-
rate Structural and Materials Journals). These publications
have influenced building codes, design specifications, and
construction practice and resulted in a more economic and
safer built environment. The goal has always been to ensure
that those who design and build with concrete have access to
the most current and cutting-edge research and information
on concrete.
As time has progressed, the Journal has seen numerous
changes from preparation (manuscripts prepared by type-
writers being submitted through the mail to electronic prepa-
ration and submission) to dissemination (hard-copy mailings
to electronic distribution). Management of the Journal has
also evolved over the years. Starting with this issue, both
the ACI Structural and Materials Journals will be overseen
by an Editorial Board. The Editorial Board is comprised of
an editor-in-chief and four research expert members. As
Editor-in Chief of the Structural Journal, I am honored to
be joined with an outstanding Editorial Board that includes
Catherine French, Michael Kreger, David Sanders, and
James Wight. I am sure that that these individuals are known
to you based on their excellent research and contributions to
both the profession and the Institute.
The Editorial Boards will focus on the mission of making
and keeping ACI’s serial publications the premier place to
4 ACI Structural Journal/January 2018
publish for academic researchers in the field of concrete. The
Boards’ initial missions include identifying new topics for
commissions; special editions; and advising on the direction
for the journals, making suggestions for both subject matter
and potential authors, providing content by writing occa-
sional editorials and other short articles, overseeing journal
quality, and more.
I am very excited about the opportunities that this new
management structure will provide. This is the first time in
the history of the Institute that an Editorial Board has been
formed. As Editor-in-Chief, I plan to explore ways that
we can further strengthen and enhance the reputation of
the ACI Structural Journal true to ACI’s tagline “Always
Advancing.” It is often said that “change is the constant,”
and that is no different with scholarly publications. Changes
are continually occurring in the means and methods of publi-
cation and the manner of dissemination due to electronic and
social media. The ACI Journals will continue to evolve to
remain relevant in this ever-changing world. I believe this is
an exciting time to explore new approaches, allowing us to
continue to advance the state of knowledge and improve the
built environment.
Robert J. Frosch, PhD, PE, FACI
Editor-in-Chief
Purdue University
p.s. A couple of interesting notes:
1. Amazingly, ACI has scanned every publication
going back to Volume 1 in 1905. ACI members can
download for free any of the more than 9000 papers
and discussions through the ACI International Concrete
Abstracts Portal: https://www.concrete.org/publications/
internationalconcreteabstractsportal.
2. Information regarding submitting manuscripts is
covered in detail through the Publications Portal: https://
www.concrete.org/publications/acistructuraljournal.
 ACI STRUCTURAL JOURNAL TECHNICAL PAPER
Title No. 114-S113
Note: Paper 114-S113 of the November-December 2017 ACI Structural Journal was printed with the erroneous title: “Effect of
Geometric Scaling on Shear Strength of Reinforced Concrete Beams with Stirrups.” A reprint including the corrected title is included
herein. Please reference the January 2018 version of this paper only.

Effect of Geometric Scaling on Shear Strength of

Reinforced Concrete Beams without Stirrups
by Derek Daluga, Kaylor McCain, Matthew Murray, and Santiago Pujol

The experimental work and studies reported in this article were

undertaken to study the effects of dimensional changes on the unit
shear strength of reinforced concrete beams. The tests included the
following variables: beam depth; maximum aggregate size; and
bar cover, size, and spacing.
Studies of the experimental results suggested that if changes
in the considered variables are controlled so that the variables
change in the same proportion, the variation in the experimental
results is not more than the expected scatter in results from nomi-
nally identical beams.

Keywords: maximum aggregate size; reinforced concrete; scaling; shear;

size effect.

INTRODUCTION
Although the work of ACI-ASCE Committee 326 has
helped the industry with respect to shear and diagonal
tension, there are cases, albeit infrequent, where beams
and slabs without stirrups and depths much larger than the
depths of the beams considered by Committee 326 (depths
exceeding 36 in. [914 mm]) are built, and the reliability of
such elements has been questioned.1-3 There are many alter-
natives to theories explaining size effect4-6; however, in this
paper, the focus was placed on experimental results.
Current design methods related to shear are based predom-
inantly on results from tests on 194 beams with depths
smaller than 24 in. (610 mm).7-9 Eighty-eight percent of these
beams had depths less than 16 in. (406 mm). Later experi-
ments included beams with larger depths.1,10-15 The results of
these experiments have been interpreted to suggest that, as
beam depth increases from 12 to 36 in. (305 to 914 mm), the
unit shear strength can decrease by as much as 68%.14 This
phenomenon, often termed “size effect”, is more noticeable
in beams without shear reinforcement,12 and is illustrated in
Fig. 1 using data collected by Reineck et al.16 The data in this
figure have parameters in these ranges
2.3 ≤ a/d ≤ 4.0
3/8 in. (10 mm) ≤ ag ≤ 1.2 in. (30 mm)
0.01 ≤ ag/d ≤ 0.2
3400 psi (23 MPa) ≤ fc′ ≤ 10,800 psi (74 MPa)
6 in. (152 mm) ≤ h ≤ 49 in. (1245 mm)
0.6% ≤ ρ ≤ 2.9%
Fig. 1—Shear test results.16
The majority of previous investigations on the relation-
ship between shear strength and depth focused on beams
with similar shear span-to-effective depth and reinforcement
ratios. However, ratios of depth to maximum aggregate size
and depth to reinforcement diameter, spacing, and cover
differed. The ratio of beam depth to width is not mentioned
because previous investigations suggest that beam width does
not have a perceptible effect on unit shear strength.1,17,18 On
the other hand, tests have shown that decreasing aggregate
size has a negative effect on shear capacity, bar diameter is
likely to affect dowel force, and crack width and spacing are
sensitive to bar cover and spacing.13,19-22 These three issues
are discussed briefly in the next sections.
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2015-318.R2, doi: 10.14359/51700947, received August 16, 2016, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.

ACI Structural Journal/January 2018 5

Aggregate size
The sensitivity of shear strength to maximum aggregate
size is illustrated in Fig. 2 using the same data from Reineck
et al.16 in Fig. 1. Maximum aggregate size is defined herein
as the size of the smallest sieve opening through which
100% of particles pass. The data illustrated in Fig. 2 suggest
that, for normalweight aggregate (such that the intrinsic
aggregate strength is capable of transferring shear across the
cracks), shear strength increases with an increase in the ratio
of reported maximum aggregate size to depth. This trend
suggests that, to avoid large reductions in shear strength
with increases in depth, aggregate size should be increased
in large beams. This is often not feasible in the field. For
instance, if a mixture with common 3/4 in. (19 mm) aggregate
is used for a 12 in. (305 mm) deep beam, then a comparable
48 in. (1219 mm) beam would require a 3 in. (76 mm) aggre-
gate mixture. To achieve a workable mixture with aggregate
of this size is difficult (especially for mixtures that need to
Fig. 2—Variation of shear strength with relative aggregate size.
Fig. 3—Crack patterns.23
6 ACI Structural Journal/January 2018
be pumped to high stories). Aggregate is scaled in this inves-
tigation for academic purposes.
The trend in Fig. 2 also suggests that aggregate size can
affect aggregate interlock, which affects shear strength. If
all variables remain equal, as depth increases, crack width
increases as well. If aggregate size does not increase at
the same rate as depth (and subsequently crack width), its
capacity to transfer shear across a crack decreases.
Bar diameter
Shear force resisted by longitudinal reinforcement, or
dowel force, is affected by bar diameter. If dowel force is
at least in part caused by bending of the longitudinal rein-
forcement as it spans an inclined crack, a larger number of
smaller bars with the same total cross-sectional area as a
smaller number of larger bars would be expected to transfer
less shear across an inclined crack (at least before a splitting
crack forms). In beams of equal size, equal reinforcement
ratio, and different bar size, the number of bars is propor-
tional to the inverse of the area (and the square of bar diam-
eter). The bending stiffness of a single bar is proportional to
the fourth power of its diameter. For the group of bars, this
results in a total lateral stiffness proportional to the square
of bar diameter. For example, if instead of ten No. 4 bars
(with a total area of 2 in.2 [1290 mm2]) one uses two No. 9
bars (resulting in the same area), the bending stiffness of
the group of No. 4 bars is one-fifth (or 20% = 42/92) of the
bending stiffness of the No. 9 bars. In addition, if the bars
are arranged in a single layer, the beam with smaller bars
has less area of concrete resisting tension across the poten-
tial plane of splitting. The lateral stiffness of the longitudinal
bars may also affect crack width, combining the effects of
dowel force and aggregate size.
Bar cover and spacing
To illustrate the relevance of bar cover c and bar spacing s
on cracking, consider tests reported by Sneed and Ramirez.14
They tested beams where concrete cover and bar spacing
were not scaled in proportion to depth. Figure 3 shows a
comparison of crack patterns in beams with different effec-
tive depths d but similar shear span-to-effective depth and
reinforcement ratios.23 These patterns were observed at
loads producing comparable bending stresses at any given
cross section. The observed crack patterns were different,
with the smaller beam (c/d = 0.3) having fewer cracks than
the larger beam (c/d = 0.11). Figure 4 shows how the number
of cracks increased with increases in depth in the tests by
Sneed.23 It is plausible that differences in crack patterns and
crack spacing result in differences in relative crack width,
stress distribution, load path, and shear strength.
Taylor24 hypothesized that the ratios of aggregate size and
bar cover, size, and spacing to depth affect shear strength.
To test this idea, Taylor tested 16 beams in which aggregate
size, width, and bar cover, size, and spacing were scaled in
proportion to depth. All beams had a reinforcement ratio
of 1.35% and a shear span-to-effective depth ratio of 3.0.
His focus on scaling yielded a 20% decrease in unit shear
strength between beams with depths of 9.4 and 39.4 in. (250
and 1000 mm). In his test, however, a direct comparison
between beams of different size but similar nondimensional
parameters could only be made between six beams.
The purpose of the tests reported in this paper was to test
Taylor’s hypothesis that if dimensions including maximum
aggregate size and bar cover, size, and spacing are scaled in
proportion to depth, an increase in depth causes a negligible
decrease in unit shear strength. For the reasons presented,
ratios of bar spacing, diameter, and cover to depth may be
important, but in the reported experiments, these ratios were
kept nearly constant to concentrate on the effects of aggre-
gate size-to-depth ratios.
RESEARCH SIGNIFICANCE
Shear strength of beams has been reported to decrease
with increases in depth. This phenomenon is commonly
referred to as size effect. From 1900 to 1980, ACI published
four articles on the subject (keyword search: size effect25).
In the following 35 years, 130 articles related to size effect
have appeared in ACI publications. More than 20 of these
papers were published in the last 5 years, indicating it is an
important matter in the profession.
Taylor23 observed that the reported reduction in shear
strength decreased dramatically when maximum aggre-
gate size, and bar cover, size, and spacing were scaled in
proportion to depth. This observation shed new light onto
the problem of size effect. Nevertheless, Taylor’s conclu-
sion seems to have gone mostly unnoticed by the profession,
perhaps because it was based on a small number of tests. The
study reported herein was conducted to test his conclusion,
and to separate the plausible effects of increases in size from
effects related to nondimensional variables.

EXPERIMENTAL PROCEDURE
Thirty reinforced concrete simply supported beams
without shear reinforcement were tested by applying a
concentrated load at midspan until shear failure occurred.
For 13 specimens, external stirrups were used to reinforce
the half of the beam where failure first occurred. Load was
applied after the external stirrups were installed to test to
failure the other half of the beam. Here, results from oppo-
site halves of a test beam are distinguished from one another
by using the suffix -N or -S after the beam ID. The nominal
dimensions and parameters of the beams are listed in Table 1.
The larger beam dimensions were 2.5 and 4 times the dimen-
Fig. 4—Number of flexural cracks.23
Table 1—Beam properties
Effective Area of tension Maximum Bar Length of bearing
Depth h, Width, b, depth, d, in. reinforcement, L, in. s, in. aggregate size diameter db, Cover, c, plates, lb,
Series in. (mm) in. (mm) (mm) As, in.2 (mm2) (mm) (mm) ag, in. (mm) in. (mm) in. (mm) in. (mm)
A 30 (762) 22.5 (572) 26.25 (667) 4.68 (3018) 120 (3048) 7.5 (191) 1 (25) 1.41 (35.8) 3.75 (95) 10 (254)
B 12 (305) 9 (229) 10.5 (267) 0.93 (600) 48 (1219) 3 (76) 1 (25) 0.63 (15.9) 1.50 (38) 4 (102)
C 12 (305) 9 (229) 10.5 (267) 0.60 (387) 48 (1219) 3 (76) 1 (25) 0.5 (12.7) 1.50 (38) 4 (102)
D 12 (305) 9 (229) 10.5 (267) 0.60 (387) 48 (1219) 3 (76) 0.5 (13) 0.5 (12.7) 1.50 (38) 4 (102)
E 30 (762) 22.5 (572) 26.25 (667) 4.68 (3018) 150 (3810) 7.5 (191) 1 (25) 1.41 (35.8) 3.75 (95) 10 (254)
F-1, F-2 12 (305) 9 (229) 10.5 (267) 0.93 (600) 60 (1524) 3 (76) 0.5 (13) 0.63 (15.9) 1.50 (38) 4 (102)
G 48 (1219) 24 (610) 42 (1067) 8.00 (5162) 240 (6096) 12 (305) 2 (51) 2.26 (57.3) 6.00 (152) 10 /12† (254/305)
*

H 12 (305) 9 (229) 10.5 (267) 0.93 (597) 60 (1524) 3 (76) 0.375 (10) 0.63 (15.9) 1.50 (38) 4 (102)
F-3, F-4, I, J 12 (305) 9 (229) 10.5 (267) 0.93 (597) 60 (1524) 3 (76) 0.5 (13) 0.63 (15.9) 1.50 (38) 4 (102)
K 12 (305) 9 (229) 10.5 (267) 0.93 (597) 60 (1524) 3 (76) 1 (25) 0.63 (15.9) 1.50 (38) 4 (102)
*
At support.
†
At load point.

ACI Structural Journal/January 2018 7

sions of the smallest beams. Table 2 shows the ratios of key diameter-to-depth ratio. Longitudinal reinforcement ratio
beam dimensions (shear span a, bar diameter db, concrete ranged between 0.63% and 0.98%. Cross-sectional dimen-
cover c, reinforcement spacing s, and maximum aggregate sions are shown in Fig. 5. The ratio of width to effective
size ag) to effective depth d. Changes in shear span-to-effec- depth was smaller in the 48 in. (1219 mm) deep beams
tive depth ratio a/d were introduced to study the sensitivity (Series G) so the entire beam could be cast from the same
of results to this parameter. Two different shear span-to-ef- batch. Fewer bars were used in the 48 in. (1219 mm) deep
fective depth ratios were used: 2.3 and 2.9. Other ratios were beams (two instead of three) to keep all nondimensional
nearly constant among all test beams except for beams in ratios within the ranges of comparable beams (Table 2).
Test Series B, C, and K. In Series B, C, and K, the ratio For this reason, the reader is encouraged to compare results
of maximum aggregate size to depth ag/d was larger than in terms of stress (accounting for differences in concrete
in other series (0.095 versus 0.036 to 0.048) to study the strength) instead of force. It is also helpful to realize that the
effects maximum aggregate size. Other differences in ag/d effects of increases in aggregate size and bar size, cover, and
are consequences of limitations in material availability. The spacing, although not always viable in practice, help reduce
same is true for differences in reinforcement ratio and bar differences in aggregate interlock and dowel force (again in
terms of stress, not force).
Table 2—Ratio of beam parameters to effective Testing was conducted at the Robert L. and Terry L. Bowen
depth Laboratory for Large-Scale Civil Engineering Research at
Shear Bar Reinforce- Cover Maximum Purdue University. Figure 6 shows the test setup used for the
span to diameter ment spacing to aggregate 30 and 48 in. (762 and 1219 mm) deep beams. The test setup
depth to depth to depth depth size to depth was similar for the 12 in. (305 mm) deep beams. Rollers
Series a/d db/d s/d c/d ag/d supported the beams at each end of the test span. The sizes
A 2.3 0.054 0.286 0.143 0.038 of plates used to distribute reactions and loads are listed in
Table 3.
B 2.3 0.060 0.286 0.143 0.095
Beams were cast with bars on the bottom. Each individual
C 2.3 0.048 0.286 0.143 0.095 beam was cast with concrete from the same batch. After the
D 2.3 0.048 0.286 0.143 0.048 concrete set, it cured under a layer of burlap and plastic.
E 2.9 0.054 0.286 0.143 0.038 Cure time differed among beams, but all specimen dried for
at least 2 weeks before testing.
F-1, F-2 2.9 0.060 0.286 0.143 0.036
Concrete was made using Type I portland cement and
G 2.9 0.054 0.286 0.143 0.048 aggregate mined in Indiana. Series G was made with crushed
H 2.9 0.048 0.286 0.143 0.036 dolomitic limestone, and the other series were made using
F-3, F-4, I, J 2.9 0.048 0.286 0.143 0.048 gravel from alluvial deposits. High-strength longitudinal
reinforcement (yield stress fy ≥ 100 ksi [690 MPa]) was used
K 2.9 0.048 0.286 0.143 0.095
as needed to prevent yielding before shear failure. Other

Fig. 5—Specimen design and test setup.

8 ACI Structural Journal/January 2018


Fig. 6—Loading setup.
details, including aggregate gradations and mixture propor-
tions, are reported by Murray,9 McCain,26 and Daluga.27
Daluga also examined the possibility that tensile concrete
strength varied consistently with aggregate size in the
mixtures used. Such variation would have introduced an
additional (unintended) variable to the reported test program.
No clear correlation between tensile concrete strength and
aggregate size was detected for the mixtures used. Tensile
concrete strength obtained from split cylinder tests ranged
between 310 and 560 psi (2.1 and 3.9 MPa).
EXPERIMENTAL RESULTS AND DISCUSSION
In each beam, flexural cracks began to form near the
midspan. As load increased, additional flexural cracks began
to form closer to the supports. These cracks became inclined
toward the load point with increasing loads. All beams failed
because of the formation of an inclined crack, resulting in
shear failure.
Experimental results are summarized in Table 4. Listed
values of concrete compressive strength fc′ are the means
of the strengths of three 6 x 12 in. (152 x 305 mm) cylin-
ders tested on the same day as the corresponding beam test.
Figure 7 shows all the measured load-deflection curves for
beams of different size organized by relative aggregate size
and aspect ratio.
Adequate scaling is deemed to have been achieved
because similar crack patterns were observed in small and
larger beams (Fig. 8). Two crack patterns are superim-
ACI Structural Journal/January 2018
Table 3—Plate dimensions
Group Bearing plates, in. (mm) Location
B, C, D, F,
9 x 4 x 3/4 (229 x 102 x 19) Supports and load point
H, I, J, K
A, E 22.5 x 10 x 2 (572 x 254 x 51) Supports and load point
24 x 10 x 2 (610 x 254 x 51) Supports
G
36 x 12 x 3 (914 x 305 x 76) Load point
posed in Fig. 8. One was observed in a 12 in. (305 mm)
deep beam and the other in a 48 in. (1219 mm) deep beam.
Both beams had nine to ten flexure cracks at loads producing
similar bending stresses. The number of flexural cracks was
not always the same in the tests reported herein, but on
average, the number of cracks did not change consistently
with changes in depth, as shown in Fig. 9. If there is a trend
in number of cracks, however, it is both difficult to distin-
guish and a much weaker trend than that in Fig. 4. Figure 9
was produced by counting cracks at loads producing similar
bending stresses at load stages close to failure.27
The peak load P measured in the laboratory was used
to calculate nominal peak shear stress V/(bd), both at the
center of the support and at a distance of d (effective depth)
from the center of the support. The calculated shear stress
includes self-weight and weight of the loading equipment
(Table 4). As the specimens had different concrete compres-
sive strengths, unit shear stress is normalized with respect
to the square root of concrete strength (v/√fc′) in Table 4 and
the following comparisons. The square root of fc′ is used to
conform to the conventions adopted in design codes in the
United States.28
Peak unit shear strength at d from support center for beams
with a shear span-to-effective depth ratio of 2.3 and 2.9 and
ag/d within the narrow range of 0.036 to 0.048 are shown in
Fig. 10(a) and 10(b), respectively. Beams in Series B, C, and
K (which had a large ratio of aggregate size to depth) are
excluded from Figure 10 to try to isolate the effects of depth
increases by keeping other relevant variables within these
narrow ranges
0.63% ≤ ρ ≤ 0.98%
0.036 ≤ ag/d ≤ 0.048
s/d = 0.29
c/d = 0.14
2700 psi (19 MPa) ≤ fc′ ≤ 5000 psi (34 MPa)
The data in Fig. 10(a) and 10(b) show a reduction in mean
shear strength of 5% and 14%, respectively. This reduction
between beams with depths of 12 and 48 in. (305 and 1219 mm)
is smaller than both the spread of the data and reductions
reported in other studies (up to 68%).1,14,29 It is plausible that
this reduction in strength is related to differences in casting
and curing conditions between large and small beams. It is
also plausible that in larger beams, with larger aggregate
particles and reinforcement bars, more bleed water gets
trapped under the aggregate particles and bars weakening
the concrete around them.
9
Table 4—Summary of experimental program
v, psi* v/√fc′ (fc′ in psi)†
Reinforcement
Specimen fc′, psi ft, psi ratio,% Peak load P, kip At support At d from support At support At d from support
A-1 3600 520 0.79 146 130 127 2.2 2.1
A-2 4700 560 0.79 132 118 116 1.7 1.7
A-3 4000 430 0.79 190 167 165 2.6 2.6
A-4 4500 425 0.79 183 161 159 2.4 2.4
Average 2.2 2.2
B-1 3600 520 0.98 34 183 182 3.0 3.0
B-2 3600 520 0.98 40 215 214 3.6 3.6
Average 3.3 3.3
C-1 4900 560 0.63 44 236 235 3.4 3.4
C-2 4700 560 0.63 39 209 208 3.1 3.0
Average 3.2 3.2
D-1 2700 400 0.63 22 119 118 2.3 2.3
D-2 2800 400 0.63 23 125 124 2.4 2.3
D-3 4400 440 0.63 27 146 145 2.2 2.2
D-4 4500 480 0.63 31 167 166 2.5 2.5
D-5 4500 480 0.63 31 167 166 2.5 2.5
D-6 4600 470 0.63 29 156 155 2.3 2.3
Average 2.4 2.3
E-1 4300 420 0.79 130 118 115 1.8 1.8
E-2 4300 470 0.79 136 123 120 1.9 1.8
E-2-S 4400 470 0.79 142 128 125 1.9 1.9
Average 1.9 1.8
F-1 4900 490 0.98 27 146 145 2.1 2.1
F-2 5000 450 0.98 29 157 156 2.2 2.2
Average 2.2 2.2
F-3-S 4900 510 0.63 27 147 146 2.1 2.1
F-3-N 4900 510 0.63 27 147 146 2.1 2.1
F-4-S 4900 510 0.63 26.6 145 144 2.1 2.1
F-4N 4900 510 0.63 31.2 169 168 2.4 2.4
Average 2.2 2.2
G-1-N 4500 450 0.79 204 114 110 1.7 1.6
G-1-S 4500 450 0.79 233 128 124 1.9 1.9
G-2-N 3400 410 0.79 205 115 110 2.0 1.9
G-2-S 3400 410 0.79 205 115 110 2.0 1.9
Average 1.9 1.8
H-1-N 4100 310 0.63 19.9 110 109 1.7 1.7
H-1-S 4100 310 0.63 24.9 136 135 2.1 2.1
H-2-S 4100 310 0.63 21.6 119 118 1.9 1.8
H-2-N 4100 310 0.63 23.3 128 127 2.0 2.0
Average 1.9 1.9
I-1-S 4400 440 0.63 24.8 136 135 2.0 2.0
I-1-N 4400 440 0.63 30 163 162 2.5 2.4

10 ACI Structural Journal/January 2018

Table 4 (cont.)—Summary of experimental program
I-1-S 4200 440 0.63 23.8 130 129 2.0 2.0
I-1-N 4200 440 0.63 27.2 148 147 2.3 2.3
Average 2.2 2.2
J-1-S 3300 330 0.63 22.9 126 125 2.2 2.2
J-1-N 3300 330 0.63 23.8 130 129 2.3 2.3
J-2-S 3300 330 0.63 22.7 125 123 2.2 2.1
J-2-N 3300 330 0.63 26.9 147 146 2.6 2.5
Average 2.3 2.3
K-1-N 4000 360 0.63 27.5 150 149 2.4 2.4
K-1-S 4000 360 0.63 28.9 157 156 2.5 2.5
K-2-S 4300 450 0.63 29.7 162 161 2.5 2.4
K-2-N 4300 450 0.63 32 174 173 2.6 2.6
Average 2.5 2.5
*
Includes self-weight and weight of loading equipment (500 lb [2.2 kN] for A and E; 100 lb [0.4 kN] for D, F-1, and F-2; 270 lb [1.2 kN] for F-3, F-4, H, I, J, and K; and 2000 lb
[8.9 kN] for G).
†
To convert this value into terms of MPa, divide it by 12.
Notes: 1 psi = 0.0069 MPa; 1 kip = 4.45 kN.

Fig. 7—Load-deflection curves (x-axis = ∆/L and y-axis = v/√fc′).

To try to isolate the effect of size, the ratio of maximum
aggregate size to effective depth was controlled. This is
seldom feasible in the field, but it does help to understand
the causes of the reduction in unit shear strength that has
often been reported as size effect. As previously explained,
maximum aggregate size can affect aggregate interlock.
The effect of aggregate size is illustrated in Fig. 11, which
compares results from Table 4 for beams where the only
nominal difference is maximum aggregate size. Increases
in maximum aggregate size from 3/8 to 1 in. (10 to 25 mm)
yielded increases in shear strength larger than 23% in beams
with shear span-to-effective depth ratios of 2.3 (Fig. 11(a)) and

ACI Structural Journal/January 2018 11

Fig. 8—Crack patterns.27

Fig. 9—Number of flexural cracks.

2.9 (Fig. 11(b)). This observation suggests that increasing
depth without increasing aggregate size can result in a
reduction in unit shear strength.

CONCLUSIONS
Thirty beams were tested to study the sensitivity of unit
shear strength to increases in cross-sectional depth. Beams
with two shear span-to-effective depth ratios were tested:
a/d = 2.3 and 2.9. In all test series except Series B, C, and
K, maximum aggregate size and bar cover, size, and spacing
were increased in similar proportion to beam depth. Beam
depth ranged from 12 to 48 in. (305 to 1219 mm).
For both shear span-to-effective depth ratios a/d, and
for beams with similar ratios of aggregate size to effective
depth, a reduction in unit shear strength V/bd was observed
with an increase in cross-sectional depth. For beams with
depths of 12 and 48 in. (305 and 1219 mm), mean unit shear
strength decreased approximately 14%. This reduction is
smaller than: 1) the spread of the data; and 2) the reduc-
tion observed in previous studies (up to 68%) (Fig. 12).1,14,29
The smaller reduction in shear strength with depth reported
herein was observed in beams within these ranges:

2.3 ≤ a/d ≤ 2.9

0.63% ≤ ρ ≤ 0.98%
0.036 ≤ ag/d ≤ 0.048
s/d = 0.29 Fig. 10—Variation of unit shear strength with beam depth
(mean unit shear strength marked in solid line).
c/d = 0.14

12 ACI Structural Journal/January 2018

 Fig. 12—Comparison of data from Fig. 1 and data from this
investigation.
Matthew Murray is a Project Manager at CE Solutions, Inc., Carmel, IN.
He received his BS and MS in civil engineering from Purdue University
in 2009 and 2010, respectively. His research interests include structural
design of new construction; additions; and repair and restoration of indus-
trial, environmental, commercial, and residential structures.

Santiago Pujol, FACI, is a Professor of civil engineering at Purdue Univer-

sity. He is a member of ACI Committees 133, Disaster Reconnaissance;
314, Simplified Design of Concrete Buildings; and 318, Structural Concrete
Building Code; and Joint ACI-ASCE Committees 441, Reinforced Concrete
Columns; and 445, Shear and Torsion.

ACKNOWLEDGMENTS
The authors express their gratitude to M. Sozen and J. Ramírez for their
help with the project and critical reviews of the results, Bowen Labora-
tory staff (K. Brower and H. Tidrick) and research assistants for their help
with tests, and MMFX for its generous donations. The constructive criti-
cism kindly provided by A. B. Acevedo at EAFIT, Colombia, is gratefully
acknowledged.

Fig. 11—Variation of shear strength with aggregate size

NOTATION
(d = 10.5 in. [267 mm]). As = cross-sectional area of reinforcing bars
a = shear span
2700 psi (19 MPa) ≤ fc′ ≤ 5000 psi (34 MPa) ag = maximum aggregate size
b = cross-sectional width
12 in. (305 mm) ≤ h ≤ 48 in. (1219 mm) c = distance from nearest concrete face to centroid of longitudinal
reinforcement
d = effective depth
The reported observations support those made by Taylor24 db = bar diameter
and suggest that the problem commonly referred to as size fc = shear stress at comparable loading stage (P/bd)
effect is an effect more closely related to scaling, particu- fc′ = compressive strength of standard 6 x 12 in. (152 x 305 mm)
concrete cylinder
larly of the maximum aggregate size, than to size itself. h = cross-sectional height
L = length of beam between centers of supports
AUTHOR BIOS lb = length of bearing plate
ACI member Derek Daluga received his BS and MS in civil engineering m = length of overhanging portion of beam
from Purdue University, West Lafayette, IN, in 2013 and 2015, respectively. P = applied concentrated load
His research interests include the effects of aggregate size on the shear s = longitudinal reinforcement spacing, center-to-center
strength of reinforced concrete. V = shear force
v = unit shear stress (V/bd)
ACI member Kaylor McCain is a Design Engineer at Structural Design ∆ = deflection
Group, Nashville, TN. He received his BS in civil engineering from Auburn ρ = longitudinal reinforcement ratio (As/bd)
University, Auburn, AL, in 2010, and MS in civil engineering from Purdue
University in 2012. His research interests include size effect and the design
and strengthening of reinforced concrete structures.
REFERENCES
1. Kani, G. N. J., “How Safe Are Our Large Reinforced Concrete
Beams?” ACI Journal Proceedings, V. 64, No. 3, Mar. 1967, pp. 128-141.

ACI Structural Journal/January 2018 13

 2. Collins, M. P., and Kuchma, D. A., “How Safe Are Our Large, Lightly
Reinforced Concrete Beams, Slabs, and Footings?” ACI Structural Journal,
V. 96, No. 4, July-Aug. 1999, pp. 482-490.
3. Collins, M. P.; Bentz, E. C.; Quach, P. T.; and Proestos, G. T., “The
Challenges of Predicting the Shear Strength of Very Thick Slabs,” Concrete
International, V. 37, No. 11, Nov. 2015, pp. 29-37.
4. Bažant, Z. P., and Kim, J., “Size Effect in Shear Failure of Longi-
tudinally Reinforced Beams,” ACI Journal Proceedings, V. 81, No. 5,
Sept.-Oct. 1984, pp. 456-468.
5. Vecchio, F. J., and Collins, M. P., “The Modified Compression-Field
Theory for Reinforced Concrete Elements Subjected to Shear,” ACI Journal
Proceedings, V. 83, No. 2, Mar.-Apr. 1986, pp. 219-231.
6. Bažant, Z. P., and Yu, Q., “Design against Size Effect on Shear
Strength of Reinforced Concrete Beams without Stirrups: II. Verification
and Calibration,” Journal of Structural Engineering, ASCE, V. 131, No. 12,
2005, pp. 1886-1897. doi: 10.1061/(ASCE)0733-9445(2005)131:12(1886)
7. Joint ACI-ASCE Committee 326, “Shear and Diagonal Tension,” ACI
Journal Proceedings, V. 59, No. 2, Feb. 1962, pp. 277-333.
8. ACI Committee 318, “Building Code Requirements for Reinforced
Concrete (ACI 318-63),” American Concrete Institute, Farmington Hills,
MI, 1963, 144 pp.
9. Murray, M. R., “An Investigation of the Unit Shear Strength of
Geometrically Scaled Reinforced Concrete Beams,” thesis, Purdue Univer-
sity, West Lafayette, IN, 2010, 106 pp.
10. Leonhardt, F., and Walther, R., “Contribution to the Treatment of
Shear in Reinforced Concrete,” Technical Translation 1172, J. P. Verschuren
and J. G. MacGregor, translators, 1962, 183 pp.
11. Iguro, M.; Shioya, T.; Nojiri, Y.; and Akiyama, H., “Experimental
Studies on Shear Strength of Large Reinforced Concrete Beams under
Uniformly Distributed Load,” Concrete Library of JSCE, translation from
Proceeding of JSCE, V. 1, No. 348, Aug. 1984, 18 pp.
12. Frosch, R. J., “Behavior of Large-Scale Reinforced Concrete Beams
with Minimum Shear Reinforcement,” ACI Structural Journal, V. 97,
No. 6, Nov.-Dec. 2000, pp. 814-820.
13. Sherwood, E. G.; Bentz, E. C.; and Collins, M. P., “Effect of Aggre-
gate Size on Beam-Shear Strength of Thick Slabs,” ACI Structural Journal,
V. 104, No. 2, Mar.-Apr. 2007, pp. 180-190.
14. Sneed, L. H., and Ramirez, J. A., “Effect of Depth on the Shear
Strength of Concrete Beams without Shear Reinforcement—Experimental
Study,” PCA R&D SN2921 Report, Portland Cement Association, Skokie,
IL, 2008, 182 pp.
15. Yu, Q., and Bažant, Z. P., “Can Stirrups Suppress Size Effect on Shear
Strength of RC Beams?” Journal of Structural Engineering, ASCE, V. 137,
No. 5, 2011, pp. 607-617. doi: 10.1061/(ASCE)ST.1943-541X.0000295
14
16. Reineck, K.; Kuchma, D. A.; Kim, K. S.; and Marx, S., “Shear Data-
base for Reinforced Concrete Members without Shear Reinforcement,” ACI
Structural Journal, V. 100, No. 2, Mar.-Apr. 2003, pp. 240-249.
17. Lubell, A.; Sherwood, E.; Bentz, E. C.; and Collins, M. P., “Safe
Shear Design of Large, Wide Beams,” Concrete International, V. 26,
No. 1, Jan. 2004, pp. 66-78.
18. Sherwood, E. G.; Lubell, A.; Bentz, E. C.; and Collins, M. P., “One-
Way Shear Strength of Thick Slabs,” ACI Structural Journal, V. 103, No. 6,
Nov.-Dec. 2006, pp. 794-802.
19. Broms, B. B., “Crack Width and Crack Spacing in Reinforced
Concrete Members,” ACI Structural Journal, V. 62, No. 10, Oct. 1965,
pp. 1237-1255.
20. Shioya, T., “Shear Properties of Large Reinforced Concrete
Member,” Special Report, Institute of Technology, Shimizu Corp., No. 25,
1989, 198 pp.
21. Shioya, T.; Iguro, M.; Nojiri, Y.; Akiyama, H.; and Okada, T., “Shear
Strength of Large Reinforced Concrete Beams,” Fracture Mechanics:
Application to Concrete, SP-118, V. C. Li and Z. P. Bažant, eds., American
Concrete Institute, Farmington Hills, MI, 1989, pp. 259-279.
22. Bentz, E. C., and Collins, M. P., “Development of the 2004 CSA
A23.3 Shear Provisions for Reinforced Concrete,” Canadian Journal of
Civil Engineering, V. 33, No. 5, 2006, pp. 521-534. doi: 10.1139/l06-005
23. Sneed, L. H., 2007. “Influence of Member Depth on the Shear
Strength of Concrete Beams,” PhD thesis, Purdue University, West Lafay-
ette, IN, 2007, 259 pp.
24. Taylor, H. P. J., “Shear Strength of Large Beams,” Journal of the
Structural Division, ASCE, V. 98, Nov. 1972, pp. 2473-2490.
25. American Concrete Institute, http://www.concrete.org. (last accessed
Jan. 30, 2015)
26. McCain, K., “The Effect of Scale on the Resistance of Reinforced
Concrete Beams to Shear,” thesis, Purdue University, West Lafayette, IN,
2012, 155 pp.
27. Daluga, D. R., “The Effect of Maximum Aggregate Size on the Shear
Strength of Geometrically Scaled Reinforced Concrete Beams,” thesis,
Purdue University, West Lafayette, IN, 2015, 200 pp.
28. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 2014, 519 pp.
29. Bhal, N. S., “The Effect of Beam Depth on the Shear Capacity of
Single Span Reinforced Concrete Beams with and without Shear Reinforce-
ment,” PhD thesis, University of Stuttgart, Stuttgart, Germany, 1968, 124 pp.
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S01
Three-Dimensional Grid Strut-and-Tie Model Approach in
Structural Concrete Design
by Young Mook Yun, Byunghun Kim, and Julio A. Ramirez
The strut-and-tie model approach has now been incorporated
in current U.S. design codes and guidelines for the design of
disturbed regions in structural concrete elements. However, more
work is needed to extend the approach to design of three-dimen-
sional (3-D) structural concrete. It is also important to consider
its verification with experimental evidence. The application to 3-D
design situations of this approach brings more uncertainties with
its proper application and limitations. To reduce uncertainty and
assist designers in the application of the strut-and-tie model to 3-D
situations, the authors present in this paper a 3-D grid strut-and-tie
model approach consisting of three key steps: 1) grid elements to
construct a 3-D strut-and-tie model; 2) triaxial failure model of
concrete to determine effective strengths of concrete struts and
nodal zones; and 3) iterative technique to evaluate the axial stiff-
ness of struts and ties. In this paper, the authors also incorporate in
the strut-and-tie model a new concept of maximum cross-sectional
areas of struts and ties to examine the strut-and-tie model’s geomet-
rical compatibility. The approach is illustrated with the redesign
of a deep pile cap with tension piles available in the literature. In
a subsequent paper, the authors will evaluate the approach with
test results of 157 specimens tested to failure. The tests include 78
reinforced concrete pile caps, 19 slab-column joints, and 60 beams
subjected to torsion.
Keywords: disturbed region; effective strength; grid strut-and-tie model;
three-dimensional structural concrete.
INTRODUCTION
The strut-and-tie model approach for disturbed regions
(D-regions) in structural concrete promotes a better under-
standing of load-transfer mechanisms and structural
behavior, and it improves the designers’ ability to handle the
often complex and unusual design of these regions. These
advantages have resulted in the implementation of the strut-
and-tie model approach in major design codes and guidelines
around the globe (CSA 2004; Comite Euro-International du
Beton 2010; American Association of State Highway and
Transportation Officials 2010; ACI Committee 318 2014).
Strut-and-tie model code approaches have been established
by considering the behavior of two-dimensional (2-D) struc-
tural concrete. Consequentially, their use in the analysis and
design of three-dimensional (3-D) structural concrete with
disturbed stress regions (for example, thick pile caps, bridge
piers subject to multi-axial loads, and anchorage zones) is
often confusing.
The current practice of arranging the struts and ties based
on the load path method in accordance to the elastic stress
trajectories can lead to several plausible models for the
same design situation. This situation can lead to potentially
unsafe outcomes, as it relies on the designer’s experience,
ACI Structural Journal/January 2018
TECHNICAL PAPER
and multiple strut-and-tie models may be required in design
situations involving multiple load combinations. Thus, even
in the case of 2-D design situations, the approach incorpo-
rating stress trajectories presents significant challenges for
efficient use in practice. Further, considering the case of
a 3-D disturbed region with significant nonlinearity in the
distribution of stresses, the visualization of the stress trajec-
tories needed to deploy the strut-and-tie model is almost
unattainable. Thus, in combination, the items described in
this paragraph could cause the improper placing of struts and
ties and unsafe designs.
To overcome this problem, studies on the construction of
2-D and 3-D strut-and-tie models using performance-based
and topology-based optimization techniques have been
conducted (Liang et al. 2001; Ali and White 2001; Liang et
al. 2002; Kwak and Noh 2006; Leu et al. 2006; Lee 2007).
However, the construction of strut-and-tie models using
optimization techniques has the following shortcomings: 1)
it can result in statically determinate strut-and-tie models
that induce difficulties in terms of distribution patterns of
reinforcing bars (which are different than practical hori-
zontal or vertical distributions of reinforcing bars, or both)
and congestion of reinforcing bars due to large concen-
trated tensile forces in the steel ties; 2) a fine finite element
(FE) mesh generation is required for initial FE modeling,
demanding significant time and effort; 3) different strut-
and-tie models are constructed according to the types of FE
models and elimination criteria of FEs; and 4) FE models of
composite concrete members (concrete and reinforcing bars)
are required to consider the proper structural behavior.
While use of statically determinate strut-and-tie models
is advantageous to expedite structural analysis and subse-
quent design, they may be limited in representing the load-
transfer mechanisms of various types of 2-D and 3-D struc-
tural concrete disturbed regions. In addition, these models
may limit the effective deployment of reinforcement due
to large tensile forces concentrated in the steel ties. These
limitations have led to studies on the use of statically inde-
terminate strut-and-tie models for the analysis and design of
2-D structural concrete (Alshegeir 1992; Foster and Gilbert
1998; Hwang et al. 2000; Yun 2000; Reineck 2002; Tjhin
and Kuchma 2002; Bakir and Boduroglu 2005; MacGregor
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2015-193.R2, doi: 10.14359/51700791, received February 28, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
15
Fig. 1—Basic and auxiliary grid elements for construction of 3-D grid strut-and-tie model.
and Wight 2005; Park et al. 2005; Comite Euro-International
du Beton 2010; Kim and Yun 2011).
Studies on the effective strength have been conducted based
on the theory of plasticity, FE nonlinear analyses of unrein-
forced structural concrete, and failure mechanism of concrete
(Yun and Ramirez 1996; Yun 2000, 2005, 2006; Ha and Hong
2005). However, those studies were based on a proper place-
ment of concrete struts to describe the load transfer mecha-
nism accurately and consider only 2-D situations. Thus, the
uncertainty relating to the construction of a strut-and-tie
model in 3-D situations has not been accounted for.
In this paper, a 3-D grid strut-and-tie model approach for
the analysis and design of structural concrete members with
D-regions is proposed. The proposed approach is deemed
sufficient to overcome the uncertainties and limitations
of existing strut-and-tie model approaches related to the
construction of the strut-and-tie model geometry, the selec-
tion of the structural types of strut-and-tie model (determi-
nate or indeterminate), and the determination of the effective
strengths of struts and nodal zones.
RESEARCH SIGNIFICANCE
The strut-and-tie model approach has now been incor-
porated in current U.S. design codes and guidelines for the
design of disturbed regions in structural concrete elements.
However, more work is needed to extend the approach to
the design of 3-D structural concrete. In this paper, a new
approach using the novel concept of grid strut-and-tie
models for application in 3-D design of structural concrete
is presented. Its use is illustrated in the design of a pile cap,
and the approach is verified with experimental evidence
in a subsequent paper. To reduce uncertainty and assist
designers in the application of the strut-and-tie model to 3-D
16
situations, the authors present in this paper a 3-D grid strut-
and-tie model approach.
THREE-DIMENSIONAL GRID STRUT-AND-TIE
MODEL
Construction of grid strut-and-tie model
A 3-D grid strut-and-tie model is composed of numerous
basic and auxiliary grid elements. It is constructed by
considering the loading and geometrical conditions of the
structural concrete. A basic grid element with 12 horizontal/
vertical components, 12 inclined plane components, and
four inclined space components is shown in Fig. 1(a). The
auxiliary grid elements, which are parts of the basic grid
element, are shown in Fig. 1(b). The basic and auxiliary
grid elements allow for load transfers in all directions at a
node. The auxiliary grid elements are used for structural
members that cannot be modeled properly by using basic
grid elements only, such as inclined pier caps, the top flanges
of box girders subjected to lateral tendon forces, and flanges
in T-shaped beams.
In the modeling of a load-transfer mechanism in structural
concrete with strut-and-tie models, the node of the model
underneath a loading or bearing plate can be subjected to
a large concentrated force and possible crushing of the
component (concrete strut) in a grid strut-and-tie model.
Thus, to prevent crushing of the concrete strut, the size of
a grid element must be selected by considering the sizes of
loading and bearing plates and the effective strength of the
concrete strut. Further, the ratio of the vertical to horizontal
length of a grid element in two directions must be limited to
0.75 ≤ lh/lvx ≤ 1.33 and 0.75 ≤ lh/lvy ≤ 1.33. In these relation-
ships, lh, lvx, and lvx are the lengths of the boundary compo-
nents of a grid element (refer to Fig. 1). The background for
ACI Structural Journal/January 2018
Fig. 2—Procedure for determination of effective strengths of 3-D concrete struts.
this requirement is given in the “Geometrical compatibility
of grid strut-and-tie model” section of this paper. The length
ratios limit the slopes of inclined struts between 28 and
62 degrees.
Effective strengths of struts and ties
A grid strut-and-tie model is composed of struts, ties, and
nodal zones. In the strut-and-tie model design of structural
concrete, the strength of struts and ties needs to be properly
determined. The design strength of a concrete strut, Pr,cs,
may be expressed as
Pr,cs = ϕcs fcsAcs (1)
where ϕcs, fcs, and Acs are the strength reduction factor, the
effective strength, and the cross-sectional area of a concrete
strut, respectively. The effective strength of a concrete
strut is chosen as a fraction of the compressive strength of
concrete, fc′
fcs = νfc′ (2)
where ν is the coefficient of the effective strength of the
concrete strut. The coefficient is introduced to reflect the
strength reduction of a concrete strut due to cracks parallel
to the longitudinal axis of the strut.
As previously discussed, current stress limits warrant
further study for application to 3-D cases. In this paper,
the method proposed by Yun (2005) for 2-D concrete struts
will be extended to 3-D applications. This method can be
used to determine the effective strengths of 3-D concrete
struts consistently because it considers the state of stress,
the extent of cracking, the length of the struts, the deviation
angle between the strut axis and compressive principal stress
flow, and the degree of confinement provided by reinforcing
ACI Structural Journal/January 2018
bars. The procedure is illustrated using the example of a pile
cap shown in Fig. 2. The order of necessary steps is:
Step 1: From an FE model of 3-D unreinforced structural
concrete (Fig. 2(a)), select the FEs that are located closely
along the longitudinal axis of a concrete strut (Fig. 2(b)).
Step 2: Determine the principal stresses σk (k = 1 to 3,
σ1 ≤ σ2 ≤ σ3) and the principal directions (θ1, θ2) of each
FE after conducting a linear elastic FE analysis of the 3-D
unreinforced structural concrete (Fig. 2(d)). By associating
the principal stresses σk with the failure criteria of 3-D unre-
inforced concrete described by the capped five-parameter
failure model (Willam and Warnke 1974), determine the
failure principal stresses σkf (k = 1 to 3, σ1f ≤ σ2f ≤ σ3f) of
each finite element (Fig. 2(e)).
Step 3: Take σ1f as the effective strength fcs,e of a finite
element if the principal direction of σ1f and the longitu-
dinal axis of the concrete strut coincide. Reduce the effec-
tive strength of the FE if its principal direction does not
coincide (Fig. 2(f)). This is accomplished by using a 3-D
Mohr’s circle (Fig. 2(g)) or by using the approach of axis
transformation. In the same manner, determine the effective
strengths of the other FEs located along the longitudinal axis
of the concrete strut.
Step 4: Calculate the standard deviation of the effective
strengths of the FEs located closely along the concrete strut.
Then, determine the effective strength of concrete strut,
fcs, by averaging the effective strengths of the FEs that are
within the standard deviation range.
Step 5: Evaluate the confinement force using the proce-
dure shown in Fig. 3 to reflect the effect of concrete confine-
ment by reinforcing bars. Apply the confinement force to the
FE model of the 3-D unreinforced structural concrete along
with external loads, and return to Step 2. Iterate Steps 2 to 5
until the difference of the confinement forces in previous and
current steps is within a specified tolerance.
17
 respectively, are expressed as functions of their effective
strengths

Pr,st = ϕst fstAst; Pr,ct = ϕct fctAct (3)

where ϕst, fst, and Ast are the strength reduction factor of
the steel tie, the effective strength of the steel tie, and the
cross-sectional area of the steel tie, respectively. The vari-
able ϕst, fct (= 0.62√fc′), and Act are the strength reduction
factor of the concrete tie (which is the same as the concrete
strut’s strength reduction factor), the effective strength of
the concrete tie, and the cross-sectional area of the concrete
tie, respectively.

Effective strength of nodal zone

The cross-sectional forces of struts and ties must be trans-
ferred to neighboring struts and ties through nodal zones.
Accordingly, the load-carrying capacity of a strut-and-tie
model is controlled by the capacities of struts and ties and
by the capacities of nodal zones, requiring the verification
of the strengths of nodal zones. The strength of a nodal zone
Fig. 3—Algorithm for considering confinement effect by depends on factors such as: a) the confinement of the nodal
reinforcing bar. zone by reactions, compression struts, anchorage plates for
prestressing, reinforcement from the adjoining members,
Step 6: The effective strength of a concrete strut is influ-
and hoop reinforcing bars; b) the effects of strain discon-
enced by the compressive strength of the concrete (Berg-
tinuities within the nodal zone; and c) the splitting stresses
meister et al. 1993; MacGregor 1997). Thus, the effective
and hook-bearing stresses that result from the anchorage
strength fcs, determined in the previous step, is modified by
of the reinforcing bars in or immediately behind the nodal
multiplying the coefficients 1.1 – 0.25fc′/70, 1.0, and 0.85 for
zone. To consider the effects of the aforementioned factors
28 < fc′ < 70 MPa, fc′ ≤ 28 MPa, and fc′ ≤ 70 MPa,  respectively.
in the strength verification of nodal zones, the methods
In the strut-and-tie model analysis and design, the concrete
conducting FE material nonlinear analyses of unreinforced
struts are placed parallel to the directions of corresponding
concrete nodal zones (Yun 2006) or strut-and-tie models for
cracks or compressive principal stress trajectories. However,
nodal zones (Hong 2000) were introduced. Although the
it is difficult to satisfy this requirement. A generally accepted
two methods seem to be appropriate, they are not efficient
approach is to place a concrete strut as parallel as possible to
to implement in practice. Moreover, the methods are only
the elastically obtained compressive principal stress trajec-
applicable to the nodal zones in 2-D structural concrete.
tories. Here, angles of inclined struts and ties are controlled
In this paper, the approach in current design codes is used,
by the shapes of grid elements used in the modeling. This
verifying the strength of a nodal zone by comparing the
constraint may cause a difference in the directions between
maximum surface areas at the nodal zone boundaries with
the concrete struts and corresponding cracks, thus violating
the required surface areas. The maximum surface areas are
the requirement on the placement of concrete struts. In this
obtained by using the maximum cross-sectional areas of
study, the difference in the directions between the concrete
the struts and ties framing the nodal zone. The maximum
struts and corresponding cracks is compensated by adjusting
cross-sectional areas of struts and ties are defined in the
the effective strengths of concrete struts. If the direction of a
“Geometrical compatibility of grid strut-and-tie model”
concrete strut is parallel to the direction of a corresponding
section. The required surface areas at the nodal zone bound-
crack or the elastic compressive principal stress trajecto-
aries are obtained by dividing the cross-sectional forces of
ries, the concrete strut may carry the compressive force up
the struts and ties framing the nodal zone by the effective
to its effective strength. Otherwise, the effective strength of
strength of the nodal zone.
the concrete strut is reduced by considering the deviation
To examine the strengths of 3-D nodal zones, the effective
angles between the strut and the compressive principal stress
strengths of the nodal zones need to be established. Herein,
trajectories, as illustrated in Step 3 of Fig. 2. In addition, the
the effective strengths of the nodal zones were determined
value of 0.34fc′, suggested in ACI 318-14 for a concrete strut
by incorporating the capped five-parameter failure model
located at a tensile stress region, is taken as the minimum
of 3-D unreinforced concrete. The procedure is illustrated
effective strength in this study.
in Fig. 4. Because the strength of the 3-D nodal zone is
In the proposed approach, if tensile regions without rein-
generally greater than the strength of the 2-D nodal zone,
forcing bars exist in 3-D structural concrete, concrete ties
the effective strength of the ACI 318-14 Compression-Com-
are used to describe an appropriate load-transfer mechanism.
pression-Tension (C-C-T) nodal zone, 0.51fc′, was taken as
The design strengths of steel and concrete ties, Pr,st and Pr,ct,
the minimum strength in this study.

18 ACI Structural Journal/January 2018

Fig. 4—Procedure for determination of effective strength of 3-D nodal zones. (Note: 1 MPa = 145 psi.)

Fig. 5—Algorithm for determining axial rigidities of struts and ties.

Stiffness of struts and ties
The required cross-sectional areas of struts and ties in a
grid strut-and-tie model under design loads were determined
using an iterative technique. The approach identifies the
main struts and ties that play an important role in transfer-
ring design loads to supports or neighboring members, and
assigns appropriate stiffness of struts and ties to enable the
structural analysis of an indeterminate grid strut-and-tie
model. The algorithm for determining the cross-sectional
areas of struts and ties is given in Fig. 5.
Once the stiffness of concrete struts and steel ties is
obtained multiplying the cross-sectional areas of the struts
and ties by their modulus of elasticity, the indeterminate
grid strut-and-tie model must be analyzed to determine the
concrete struts and steel ties. In this analysis, an inclined
plane component crossing an inclined plane concrete strut

ACI Structural Journal/January 2018 19

Fig. 6—Maximum cross-sectional areas of components of basic grid element.
placed almost parallel to the compressive stress trajectory
by transferring a large force would generally be consid-
ered a steel tie. In this case, the cross-sectional force of the
steel tie can be replaced with 2-D horizontal and vertical
forces. The same rule applies to cases of irregularly-shaped
grid elements.
Geometrical compatibility of grid strut-and-tie
model
The geometrical compatibility of a 3-D grid strut-and-tie
model must be examined as part of the strut-and-tie model
design. The required cross-sectional areas of struts and
ties, determined using an iterative approach, must be less
than their maximum cross-sectional areas. The maximum
cross-sectional areas of struts and steel ties in the strut-
and-tie model application in current design codes are deter-
mined by using initial design conditions such as the geom-
etry of the structural concrete, the sizes of bearing and
loading plates, and the pattern of reinforcement details. In
the proposed approach, the method employed in the 2-D
grid strut-and-tie model approach proposed by Yun and Kim
(2008) is extended. Namely, in addition to the strut-and-tie
model approaches of current design codes, the maximum
cross-sectional areas of struts and ties are defined by the
20
maximum areas that struts and ties can have without over-
laps in a grid element (refer to Fig. 6).
The maximum cross-sectional area of a component of a
grid element is determined by multiplying two maximum
widths of the component evaluated on the component’s two
orthogonal planes. For example, as shown in Fig. 6(c), the
maximum cross-sectional area of the x-direction component
of length lx1 is determined by multiplying the z-direction
maximum width wHE and the y-direction maximum width
wVE. Here, the maximum width wHE is taken as 0.2lz1 + 0.2lz2,
with lz1 and lz2, corresponding respectively to the lengths
wH1 and wH1 shown in Fig. 6(a). These are the lengths of
the z-direction components connected to the ends of the
x-direction component. The reason for taking 20% of the
lengths of the neighboring elements is based on the rule
that the maximum widths of all elements in a grid element
should be the same initially. A mathematical derivation for
the 20% rule is provided in the Appendix of this paper. Simi-
larly, the maximum width wVE is defined as 0.2ly1 + 0.2ly2,
where ly1 and ly2 are the lengths of the y-direction compo-
nents connected to the ends of the x-direction component.
The maximum cross-sectional areas of the y- and z-direction
components are determined following a similar approach.
The maximum cross-sectional area of an inclined plane
component of a given grid element is determined in the same
ACI Structural Journal/January 2018
manner. As shown in Fig. 6(d), the maximum cross-sectional
area of the inclined plane component of x-direction length lx1
in the xy-plane is determined by multiplying the maximum
width wIPE of the xy-plane and the maximum width wHEZ of
the plane orthogonal to the xy-plane. Here, the maximum
width wIPE of a given plane is determined following the
method illustrated in Fig. 6(b). The maximum width wIPE
of the xz-plane is taken as the length (the smaller of wa +
wb and wc + wd) of the perpendicular lines connecting the
points (a, b, c, d), which are determined by considering the
maximum widths of the neighboring horizontal and vertical
elements. Following the methods illustrated in Fig. 6(a) and
6(b), the maximum widths wIPE and wHEZ of the xy-plane and
its orthogonal planes, respectively, are expressed as
0.6lx1l y 2
wIPE = , wHEZ = 0.2ls1 + 0.2ls 2 (4)
lx21 + l y22
where lx1, ly2, lz1, and lz2, as shown in Fig. 6(d), are the lengths
of the grid element components. The maximum cross-sec-
tional areas of the inclined plane components in the xz-plane
and yz-plane are determined following a similar approach.
The cross-sectional area of an inclined space component
that exists only in a 3-D grid element is defined as the area
excluding all the cross-sectional areas of horizontal compo-
nents, vertical components, and inclined plane components
in a grid element. For example, as shown in Fig. 6(e), the
maximum cross-sectional area of the space component
is determined by multiplying the maximum width wIPE of
the xz-plane and the maximum width wISE of the yz-plane
rotated about the y-axis. The maximum widths wIPE and wISE
are expressed as
0.6lx1lz 2 0.6 lx21 + lz22 l y 2
wIPE = , wISE =
lx21 + lz22 lx21 + l y22 + lz22
The maximum cross-sectional areas of the other three
space components are determined in the same way. Fig. 6(f)
shows an example where the entire available volume of
concrete is required when the maximum cross-sectional areas
of all components in a 3-D grid strut-and-tie model are used
without overlaps.
In the proposed approach, a component of a grid element
is assumed to fail if its maximum cross-sectional area is less
than that required for carrying the cross-sectional force. To
reflect more precisely the characteristics of the load-transfer
mechanism in a grid element, the modified maximum
cross-sectional area of a component that is determined by
considering the required cross-sectional areas of neigh-
boring components (instead of the maximum cross-sectional
area defined above) is used for examining the geometrical
compatibility. Here, the modified maximum cross-sectional
area of a component is determined by multiplying two modi-
fied maximum widths of the component in two orthogonal
planes. An example of determining the modified maximum
area of an inclined plane component of a grid element is
shown in Fig. 7. If the required cross-sectional area of compo-
ACI Structural Journal/January 2018
Fig. 7—Modified maximum cross-sectional areas of compo-
nents of basic grid element.
nent CA exceeds the maximum cross-sectional area defined
above, the maximum cross-sectional area of the component
AIPE (= wIPE × wHEZ) increases to AIPE (= wIPE × wHEZ ) if the
required widths of the neighboring components CF~CI of the
xy-plane and CB~CE of the plane orthogonal to the xy-plane
are less than their maximum widths. In Fig. 7, w′HEZ is the
limit of the modified maximum width of component CA that
can be maximized when the required widths of the compo-
nents CB~CE are equal to zero. Also, w′IPE is the limit of
the modified maximum width of component CA that can be
maximized in the xy-plane. The concept employed herein
originates from the rule stating that the overlaps of areas
(or widths) of neighboring components are not allowed,
as introduced in Fig. 6. The modified maximum width of a
component of a grid element in any plane of a 3-D grid strut-
and-tie model is obtained in a similar manner.
(5) IMPLEMENTATION
The analysis or design of D-Regions in structural concrete
using the proposed approach incorporating basic grid
elements to construct a 3-D strut-and-tie model is conducted
using the procedure shown in Fig. 8. A detailed explanation
of the procedure follows.
Step 1: Set up the initial conditions, including loading
and boundary conditions, material strengths, and sizes of
the loading and bearing plates. For a strut-and-tie model anal-
ysis using the proposed approach, information regarding
reinforcement details is required to calculate the cross-
sectional areas of the steel ties.
Step 2: Conduct a linear elastic FE analysis. The concrete
element or structure is modeled as a plain (unreinforced)
isotropic 3-D problem. Construct a 3-D grid strut-and-tie
model using basic and auxiliary grid elements. To facilitate
a proper load transfer within the concrete structure, the size
of grid elements must be selected considering the sugges-
tions given in the section, “Construction of grid strut-and-tie
model.”
Step 3: Determine the effective strengths of the concrete
struts by using the linear elastic FE analysis results.
Step 4: Determine the axial stiffness and cross-sectional
forces of the struts and ties using the iterative technique
21
Fig. 8—Procedure for applying 3-D grid strut-and-tie approach.
shown in Fig. 5. For strut-and-tie model analysis, the stiff-
ness of the steel ties must be determined using the informa-
tion on reinforcement details given in Step 1. Determine the
effective strengths of the 3-D nodal zones using the linear
elastic FE analysis results of the grid strut-and-tie model.
Step 5: To consider the effects of the confinement of
concrete by reinforcing bars on the effective strengths of
concrete struts, apply the cross-sectional forces of the hori-
zontal and vertical steel ties additionally to the FE model of
the structural concrete. Iterate Steps 2 to 5 until the cross-sec-
tional forces of the horizontal and vertical steel ties converge.
Step 6: Determine the maximum cross-sectional areas of
the struts and ties to verify the conditions of strut-and-tie
model’s geometrical compatibility and nodal zone strength.
If necessary, the maximum cross-sectional areas of the struts
and ties should be adjusted.
22
Step 7: In this linear analysis, the ultimate strength of the
structural concrete is estimated by examining the load-car-
rying capacities of the struts, ties, and nodal zones of the
constructed 3-D grid strut-and-tie model under its ultimate
loads while satisfying strut-and-tie model’s geometrical
compatibility condition. The ultimate strength is obtained by
examining the conditions given in Table 1 for compressive
failure of the concrete struts, tensile failure of horizontal (or
vertical) concrete ties, tensile failure of steel ties, and local
failure of nodal zones. Here, failures of the inclined plane
and space concrete ties of a grid element were excluded
because the cross-sectional forces are distributed to the
nearby horizontal and vertical ties. In the nonlinear strut-
and-tie model analysis, the ultimate strength of the struc-
tural concrete is estimated by examining the condition of
structural stability of the constructed 3-D grid strut-and-tie
ACI Structural Journal/January 2018
Table 1—Ultimate strength defined by conditions of 3-D grid strut-and-tie model’s components
Component Ultimate strength (= Pu) States of components at Pmax
Concrete strut Pmax Acs req’d = Acs pro’d, Act req’d < Act pro’d, fs < fy, σnz < fnz
Concrete tie Pmax Acs req’d < Acs pro’d, Act req’d = Act pro’d, fs < fy, σnz < fnz
Steel tie Pmax × fy ÷ fs Acs req’d ≤ Acs pro’d, Act req’d ≤ Act pro’d, fs > fy, σnz < fnz
Nodal zone Pmax × fnz ÷ σnz Acs req’d ≤ Acs pro’d, Act req’d ≤ Act pro’d, fs < fy, σnz > fnz

Notes: Pmax is maximum load that 3-D grid strut-and-tie model can carry by satisfying condition of strut-and-tie model’s geometrical compatibility (Acs req’d ≤ Acs pro’d and Act
req’d ≤ Act pro’d); Acs req’d and Act req’d are required cross-sectional areas of concrete strut and concrete tie at Pmax, respectively; Acs pro’d and Act pro’d are maximum cross-sectional areas of
concrete strut and concrete tie, respectively; fy and fs are yield strength of steel and stress of steel tie at Pmax, respectively; fnz is effective strength of nodal zone; σnz is compressive
stress of nodal zone boundary at Pmax.

Fig. 9—Geometry and load condition of pile cap. (Note: 1 mm = 0.0394 in.; 1 MPa = 0.145 ksi; 1 kN = 0.225 kip.)
model, as shown in Fig. 8. The conditions of the strut-and-tie
model’s geometrical compatibility and nodal zone strength
must be satisfied. If this is not the case, it will be necessary
to iterate, by changing the initial conditions, or increasing
the strength of the selected model, to satisfy the original
conditions. Once the initial conditions are satisfied, the rein-
forcement is selected to satisfy the calculated tie forces at the
proper locations.
DESIGN OF DEEP PILE CAP WITH TENSION PILES
The deep pile cap with tension piles, used in the illus-
tration of the ACI 318 strut-and-tie model method in ACI
SP-273 (Reineck and Novak 2010), is selected to illustrate
the grid strut-and-tie model design procedure and compare
the results with those obtained in the original example. The
geometry of pile cap, material properties, loading, and geom-
etry of a 3-D strut-and-tie model constructed in ACI SP-273
are shown in Fig. 9. Details regarding the load condition,
construction of strut-and-tie model geometry by conven-
tional strut-and-tie model approach, strength verifications of
struts and nodes, and anchorage and reinforcing bar details
are described in the reference.
To design the pile cap by the proposed approach, a linear
elastic 3-D solid finite element analysis about the pile cap
is conducted first. In the analysis, pile reactions are applied
as external forces as in the case of ACI SP-273. A hinge
and multiple rollers at the interface of column and pile cap
are imposed to describe the structural behavior of the pile
cap properly and stabilize the finite element model. Next,
a 3-D grid strut-and-tie model is constructed. To compare
the design results, the top and bottom nodes of the grid
strut-and-tie model are located at 75 and 200 mm (2.95 and
7.87 in.) away from the top and bottom surfaces of the pile
cap, as in the case of the ACI SP-273 strut-and-tie model.
The x- and z-coordinates of the nodes underneath the column
and at pile supports in the grid strut-and-tie model are the
same as those of the ACI SP-273 strut-and-tie model. The
number of grids is decided to satisfy the strength require-
ment at critical nodes (such as the nodes at supports and
underneath column) and describe the principal compressive
stress flows. The shapes of grid elements are constructed to

ACI Structural Journal/January 2018 23


Fig. 10—Dimensioned shapes of 3-D grid strut-and-tie
model for pile cap. (Note: Full-color rendering of figure can
be viewed online at www.concrete.org.)
Fig. 11—Cross-sectional forces of ties in pile cap strut-and-tie models. (Note: 1 kN = 0.225 kip.)
24
fulfill the suggested requirement on the vertical to horizontal
length ratios of a grid element in x- and z-directions.
The effective strengths of concrete struts and nodal zones
are determined by following the procedures explained in
Fig. 2, 3, and 4. The cross-sectional areas of struts and ties
are determined by using the iterative technique shown in
Fig. 5. Dimensioned shapes of the grid strut-and-tie model
are shown in Fig. 10. Based on the dimensioned shapes satis-
fying the geometrical compatibility of strut-and-tie model, it
is known that the strength conditions of concrete struts and
nodes are met.
The cross-sectional forces of ties in the ACI SP-273 and
grid strut-and-tie models are determined and shown in Fig.
11. Using the tie forces, the required areas of steel are deter-
mined. The required areas of reinforcement by the two types
of strut-and-tie models are in reasonably good agreement
at every region of the pile cap. However, the total required
areas of reinforcing bars by the proposed approach are 9%
less compared with that of the ACI SP-273 approach, indi-
cating that the strut-and-tie model constructed using the
proposed approach is more efficient because it undergoes
less strain energy. This is achieved by placing the struts and
ties according to principal stress trajectories.
ACI Structural Journal/January 2018
 SUMMARY
The current strut-and-tie model approach in U.S. guide-
lines is implemented using statically determinate models.
While in many cases, determinate strut-and-tie models are
sufficient, they have limits in representing the load-transfer
mechanisms in various types of structural concrete with
disturbed regions. The work presented herein is a refinement
of the strut-and-tie model approach in the construction of
the model itself, the selection of the model’s structural type,
and in the determination of the effective strengths of struts
and nodal zones. The proposed 3-D grid strut-and-tie model
approach is intended to extend the application of the strut-
and-tie approach by illustrating the implementation of 3-D
strut-and-tie models to complex design situation that are more
realistically represented by 3-D load-transfer mechanisms in
structural concrete. It also illustrates the implementation of
a computer-based method aimed at eliminating numerous
hand calculations and simplifying the strength verification
in the strength analysis and design of the entire 3-D member.
The implementation of the proposed approach requires first
a FE linear analysis of the member using a strut-and-tie
model of the member without reinforcement. The proposed
approach using a 3-D grid strut-and-tie model is presented as
an effective and practical design technique offering a more
effective alternative to the nonlinear finite element analysis
for the design, and particularly the detailing, of structural
concrete members and subassemblies.
In a subsequent paper, the ultimate strength of 78 rein-
forced concrete pile caps, 19 slab-column joints, and 60
torsional beams have been evaluated to verify the proposed
approach in this paper. In addition, the design of a pier cap
subjected to multiple load combinations with longitudinal
and lateral loads is also conducted to illustrate its application.
AUTHOR BIOS
Young Mook Yun is a Professor in the Department of Civil Engineering
at Kyungpook National University, Daegu, Korea. He received his PhD
from Purdue University, West Lafayette, IN, in 1995. His research inter-
ests include strut-and-tie model analysis and design of concrete members,
computational mechanics, and computer graphics.
Byunghun Kim is a Lead Engineer in the Infrastructure Department
of Hyundai Engineering, Seoul, South Korea. He received his PhD from
Kyungpook National University in 2004. His research interests include
strut-and-tie model design of concrete structures and development of
numerical analysis methods for steel and reinforced concrete structures.
Julio A. Ramirez, FACI, is a Professor in the Lyles School of Civil Engi-
neering at Purdue University. He is a member of Joint ACI-ASCE Commit-
tees 408, Bond and Development of Reinforcement; and 445, Shear and
Torsion. He received the ACI Delmar L. Bloem Distinguished Service
Award and the ACI Joe W. Kelly Award in 2000 and 2006, respectively.
REFERENCES
ACI Committee 318, 2014, “Building Code Requirements for Struc-
tural Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 520 pp.
Ali, M. A., and White, R. N., 2001, “Automatic Generation of Truss
Model for Optimal Design of Reinforced Concrete Structures,” ACI Struc-
tural Journal, V. 98, No. 4, July-Aug., pp. 431-442.
Alshegeir, A., 1992, “Analysis and Design of Disturbed Regions with
Strut-and-tie Models,” PhD dissertation, School of Civil Engineering,
Purdue University, West Lafayette, IN.
American Association of State Highway and Transportation Officials,
2010, “AASHTO LRFD Bridge Design Specifications,” fifth edition, Wash-
ington, DC.
ACI Structural Journal/January 2018
Bakir, P. G., and Boduroglu, H. M., 2005, “Mechanical Behavior and
Non-linear Analysis of Short Beams using Softened Truss and Direct Strut
& Tie Models,” Engineering Structures, V. 27, No. 4, pp. 639-651. doi:
10.1016/j.engstruct.2004.12.003
Bergmeister, K.; Breen, J. E.; Jirsa, J. O.; and Kreger, M. E., 1993,
“Detailing in Structural Concrete,” Research Report 1127-3F, University of
Texas at Austin, Austin, TX.
Canadian Standards Association, 2004, “Design of Concrete Structures
for Buildings (A23.3-M04),” Rexdale, ON, Canada.
Comite Euro-International du Beton, 2010, “CEP-FIP Model Code
2010,” International Federation for Structural Concrete (fib), Lausanne,
Switzerland.
Foster, S. J., and Gilbert, R. I., 1998, “Experimental Studies on High-
Strength Concrete Deep Beams,” ACI Structural Journal, V. 95, No. 4,
July-Aug., pp. 382-390.
Ha, T. H., and Hong, S. G., 2005, “Effect of Diagonal Cracking on
the Strength of Concrete Strut in RC Members,” Journal of the Korean
Concrete Institute, V. 17, No. 1, pp. 383-386.
Hong, S. G., 2000, “Strut-and-Tie Models and Failure Mechanisms for
Bar Development in Tension-Tension-Compression Nodal Zone,” ACI
Structural Journal, V. 97, No. 1, Jan.-Feb., pp. 111-121.
Hwang, S. J.; Lu, W. Y.; and Lee, H. J., 2000, “Shear Strength Predic-
tion for Deep Beams,” ACI Structural Journal, V. 97, No. 3, May-June,
pp. 367-376.
Kim, B. H., and Yun, Y. M., 2011, “An Indeterminate Strut-and-tie
Model and Load Distribution Ratio for RC Deep Beams - (I) Model & Load
Distribution Ratio,” Advances in Structural Engineering, V. 14, No. 6,
pp. 1031-1041. doi: 10.1260/1369-4332.14.6.1031
Kwak, H. G., and Noh, S. H., 2006, “Determination of Strut-and-Tie
Models Using Evolutionary Structural Optimization,” Engineering Struc-
tures, V. 28, No. 10, pp. 1440-1449. doi: 10.1016/j.engstruct.2006.01.013
Lee, W. S., 2007, “Strut-and-Tie Model Approach Using Evolutionary
Structural Optimization for Reinforced Concrete Structural Members,”
PhD dissertation, Kyungpook National University, Daegu, Korea, 210 pp.
Leu, L. J.; Huang, H. W.; Chen, C. S.; and Liao, Y. P., 2006, “Strut-
and-Tie Design Methodology for Three-Dimensional Reinforced Concrete
Structures,” Journal of Structural Engineering, ASCE, V. 132, No. 6, pp.
929-938. doi: 10.1061/(ASCE)0733-9445(2006)132:6(929)
Liang, Q. Q.; Uy, B.; and Steven, G. P., 2002, “Performance-based Opti-
mization for Strut-and-Tie Modeling of Structural Concrete,” Journal of
Structural Engineering, ASCE, V. 128, No. 6, pp. 815-823. doi: 10.1061/
(ASCE)0733-9445(2002)128:6(815)
Liang, Q. Q.; Xie, Y. M.; and Steven, G. P., 2001, “Generating Optimal
Strut-and-Tie Models in Prestressed Concrete Beams by Performance-based
Optimization,” ACI Structural Journal, V. 98, No. 2, Mar.-Apr., pp. 226-232.
MacGregor, J. G., 1997, Reinforced Concrete: Mechanics and Design,
third edition, Prentice Hall, Englewood Cliffs, NJ.
MacGregor, J. G., and Wight, J. K., 2005, Reinforced Concrete:
Mechanics and Design, fourth edition, Pearson Prentice Hall, Upper Saddle
River, NJ.
Park, H. G.; Kim, Y. G.; and Eom, T. S., 2005, “Direct Inelastic Strut-
and-tie Model Using Secant Stiffness,” Journal of the Korean Concrete
Institute, V. 17, No. 2, pp. 201-212. doi: 10.4334/JKCI.2005.17.2.201
Reineck, K. H., ed., 2002, Examples for the Design for Structural
Concrete with Strut-and-Tie Models, SP-208, American Concrete Institute,
Farmington Hills, MI.
Reineck, K. H., and Novak, L. C., eds., 2010, Further Examples for the
Design of Structural Concrete with Strut-and-Tie Models, SP-273, Amer-
ican Concrete Institute, Farmington Hills, MI, 288 pp.
Tjhin, T. N., and Kuchma, D. A., 2002, “Computer-Based Tools for
Design by Strut-and-Tie Method: Advances and Challenges,” ACI Struc-
tural Journal, V. 99, No. 5, Sept.-Oct., pp. 586-594.
Willam, K. J., and Warnke, E. P., 1974, “Constitutive Model for the
Triaxial Behavior of Concrete,” Proceedings of the International Associa-
tion of Bridge Structures, V. 19, pp. 1-30.
Yun, Y. M., 2000, “Nonlinear Strut-and-tie Model Approach for Structural
Concrete,” ACI Structural Journal, V. 97, No. 4, July-Aug., pp. 581-590.
Yun, Y. M., 2005, “Effective Strength of Concrete Strut in Strut-and-tie
Model (I): Methods for Determining Effective Strength of Concrete Strut,”
Journal of the Korean Society of Civil Engineers, V. 25, pp. 49-59.
Yun, Y. M., 2006, “Strength of Two-Dimensional Nodal Zones in Strut-
and-tie Models,” Journal of Structural Engineering, ASCE, V. 132, No. 11,
pp. 1764-1783. doi: 10.1061/(ASCE)0733-9445(2006)132:11(1764)
Yun, Y. M., and Kim, B. H., 2008, “Two-Dimensional Grid
Strut-Tie Model Approach for Structural Concrete,” Journal of Struc-
tural Engineering, ASCE, V. 134, No. 7, pp. 1199-1214. doi: 10.1061/
(ASCE)0733-9445(2008)134:7(1199)
25
 Yun, Y. M., and Ramirez, J. A.1996 , “Strength of Struts and Nodes in
Strut-and-Tie Model,” Journal of Structural Engineering, ASCE, V. 122,
No. 1, pp. 20-29. doi: 10.1061/(ASCE)0733-9445(1996)122:1(20)

APPENDIX
Mathematical proof for taking 20% of the length of
neighboring horizontal (vertical) components in the
determination of the maximum width of vertical (horizontal)
components (Fig. A1).
By using the equation of perpendicular, the half of the
maximum width of the inclined plane component of the
square grid element shown previously is described as follows

ax3 + by3 + c
wb = (A1)
(a 2 + b 2 )
Fig. A1.
where the coefficients a, b, and c are a = y1 – y2 = –1.0, b = x2
– x1 = 1.0, and c = x1y2 – y1x2 = 0, respectively. Substituting maximum widths of the vertical and inclined components
the coefficients into Eq. (A1), we have are the same, that is, by equating wa to wb, we have

− wa + 0.5 (A2) 0.5

wb = wa = = 0.2071 ≈ 0.2 (A3)
2 2 +1

As the value 0.5 is the maximum half width that the
vertical component can take, that is, as wa = 0.5, |–wa + Therefore, the maximum width of the inclined plane
0.5| becomes –wa + 0.5. By assuming that the halves of the component equals to those of neighboring vertical
(horizontal) components if we take 20% of the length of
neighboring horizontal (vertical) components.

26 ACI Structural Journal/January 2018

 ACI STRUCTURAL JOURNAL
Title No. 115-S02
Verification of Three-Dimensional Grid Strut-and-Tie Model
Approach in Structural Concrete
by Young Mook Yun, Hyun Soo Chae, Byunghun Kim, and Julio A. Ramirez
In this study, a proposed three-dimensional (3-D) grid strut-and-tie
model approach is verified using the experimental capacity from
78 reinforced concrete pile caps, 19 slab-column joints, and 60
torsional beams obtained by others. The analysis results were
compared with those obtained using the BS 8110, EC2, CRSI,
fib, AASHTO-LRFD, and ACI 318 design provisions. In addition,
the design of a pier cap subjected to multiple load combinations
with longitudinal and lateral loads was conducted to illustrate the
proposed approach. The design results from the pier cap example
were also compared with those obtained using ACI 318-14. The
analysis results agreed well with experimental results.
Keywords: grid strut-and-tie model; pier cap; pile cap; slab-column joint;
torsional beam.
INTRODUCTION
In the paper by Yun et al. (2017), the limitations of the
strut-and-tie model approach in current U.S. design codes
and guidelines with respect to the construction of the strut-
and-tie model, selection of the strut-and-tie model’s struc-
tural type, and determination of the effective strengths of
concrete struts and nodal zones for the designs of three-
dimensional (3-D) structural concrete with D-regions were
examined. To overcome the limitations of the strut-and-tie
model approach in U.S. codes of practice, a 3-D grid strut-
and-tie model approach that incorporates grid elements
to construct a 3-D strut-and-tie model was proposed. The
approach described in the paper by Yun et al. (2017) includes
a five-parameter failure model of unreinforced concrete to
determine the effective strengths of concrete struts and nodal
zones and an iterative technique to evaluate the axial stiff-
ness of struts and ties. The maximum cross-sectional area of
struts and ties approach is used to examine the strut-and-tie
model’s geometrical compatibility with the strut-and-tie
model approaches of current design codes.
The proposed approach was validated using the experi-
mental capacity of 78 reinforced concrete pile caps, 19 slab-
column joints, and 60 torsional beams obtained by others.
The ultimate strength of 78 pile caps estimated using the
proposed approach was compared with ultimate strength
based on provisions in ACI 318 sectional design methods
(1999, 2014), CRSI sectional design method (2008), and
ACI 318-14 strut-and-tie model approach (2014). The ulti-
mate strengths of the slab-column joints predicted using
the proposed approach were compared with those from the
two-way slab provisions of BS 8110 (1997), EC2 (2004), fib
(2010), and ACI 318-14 (2014). The ultimate strengths of
the torsional beams were compared with those of AASHTO-
LRFD (2010), ACI 318-14 (2014), Hsu and Mo (1983),
and Rahal and Collins (1995). The proposed approach was
ACI Structural Journal/January 2018
TECHNICAL PAPER
also examined through the design of a pier cap subjected
to multiple load combinations with longitudinal and lateral
loads. The design results of the pier cap were compared with
the results using the ACI 318 sectional design methods. The
reliability of the design results were verified using the strut-
and-tie models suggested in current design codes.
RESEARCH SIGNIFICANCE
To overcome limitations and improve clarity in the design
of structural concrete using 3-D strut-and-tie model approach
in current design codes, the authors have proposed a 3-D
grid strut-and-tie model approach (Yun et al. 2017). The
approach consists of three key elements: 1) use of basic grid
elements to construct a 3-D strut-and-tie model; 2) a triaxial
failure model of concrete to determine effective strengths of
concrete struts and nodal zones; and 3) a simple iterative
technique to evaluate the axial stiffness of struts and ties. In
this paper, the approach is validated using an extensive data-
base of relevant tests including reinforced concrete pile caps,
slab-column joints, and torsional beams. In this verification
of the approach the provisions of BS 8110, EC 2, CRSI, fib,
AASHTO-LRFD, and ACI 318-14 were used. Finally, the
approach is illustrated using an example of a bridge pier cap
under multiple loading conditions.
ANALYSIS OF PILE CAP TESTS
To verify the 3-D grid strut-and-tie model approach, the
ultimate strengths of 78 reinforced concrete pile caps, tested
to failure by Clarke (1973), Sabnis and Gogate (1984),
and Suzuki et al. (1998, 2000) were estimated using the
approach. In addition, capacities were calculated using ACI
318 sectional design methods (1999, 2014), CRSI sectional
design method (2008), and ACI 318-14 strut-and-tie model
approach (2014). The pile cap tests were conducted to inves-
tigate the effects of patterns of flexural reinforcing bars on
their structural behaviors and strengths (Clarke 1973; Sabnis
and Gogate 1984; and Suzuki et al. 1998, 2000). The mate-
rial properties and geometries of the pile caps are summa-
rized in Table 1. Reinforcement details and geometric shapes
of representative pile caps tested by Suzuki et al. (1998)
are shown in Fig. 1. The pile caps were failed by bending
(flexure) and/or shear (one-way, two-way, and punching).
Detailed information, including the failure patterns of the
pile caps, is given in the aforementioned references.
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2015-192.R2, doi: 10.14359/51700946, received July 17, 2015, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
27
Table 1—Specification of pile caps
Reinforcing
Investigators (A) w/d d, mm Column size, mm Pile size, mm fc′, MPa fy, MPa ρ, % bar layout
Clarke (1973) 12 0.25 to 0.49 109 to 117 200, S 200, R 17.9 to 41.0 510 0.188 to 0.238 B and G
Sabnis and Gogate (1984) 8 0.54 to 0.58 400 76, R 76, R 17.4 to 36.3 414 0.136 to 1.287 G
Suzuki et al. (1998) 28 0.40 to 0.80 150 to 250 300, S 150, R 18.9 to 31.3 356 to 413 0.228 to 0.546 B and G
Suzuki et al. (2000) 30 0.29 to 0.67 150 to 350 200 to 300, S 150, R 24.6 to 28.8 358 to 383 0.181 to 0.272 G
Total 78 0.25 to 0.80 109 to 400 76 to 300 76 to 200, R 17.4 to 41.0 356 to 510 0.136 to 1.287 B, G

Notes: (A) is number of pile caps; w is distance from face of column to center of nearest pile; d is effective depth; ρ is flexural reinforcement ratio; R, S, B, and G is round, square,
bunched, and grid; 1 mm = 0.0394 in.; 1 MPa = 145 psi.

Fig. 1—Reinforcement details and geometrical shapes of pile caps (Suzuki et al. 1998). (Note: 1 mm = 0.0394 in.)
The ultimate strength of the pile caps based on ACI 318
sectional design methods was determined by applying the
flexural design provisions of beams and slabs (ACI 318-14,
Sections 22.2 through 22.4), the bearing strength provisions
(ACI 318-14, Section 10.14), the shear design provisions of
one-way and two-way slabs (ACI 318-14, Sections 22.5 and
22.6), and the shear design provisions of deep beams with
ln/d ≤ 5 (where ln is the length of clear span, and d is the
distance from the extreme compression fiber to the center
of longitudinal reinforcement) (ACI 318-99, Section 11.8).
In the one-way shear design of the pile cap, the nominal
shear strength provided by the concrete at the critical section
was determined from ACI 318-14 Eq. (22.5.5.1) and Eq.(a)
of ACI 318-14 Table 22.5.5.1. For the two-way shear design,
the section along the circumference separated from the face
of the column by a distance of d/2 was regarded as the crit-
ical section, and the smallest value determined from Eq.(a)
to (c) of ACI 318-14 Table 22.6.5.2 was taken as the nominal
shear strength at the critical section. When the ratio of the
clear span to the effective depth of pile cap was less than 5,
the nominal shear strength was determined from ACI 318-99
Eq. (11.29).
The detailed procedure for evaluating the pile cap’s
ultimate strength is shown in Tables 2(a) and (b) with an
example of the specimen BPC-30-25-1 (Suzuki et al. 1998).
In an application of the ACI 318 sectional design method,
the minimum value among Pu, f, Pu,bc, Pu,bp, Pu,1s, and Pu,2s
evaluated, respectively, from the flexural strength Mn, f, the
bearing strength Pn,bc at column, the bearing strength Pn,bp
at the pile, the shear strength of one-way slab Vn,1s, and the
shear strength of two-way slab shear Vn,2s, was taken as the
ultimate strength of the pile cap Pu. The ultimate strengths of
the other pile caps were determined in the same way.
The CRSI sectional design method for the pile caps is
quite similar to the ACI 318 sectional design method. For the
shear design of a pile cap that can be regarded as a one-way
slab, the section at the face of the column is considered to be
the critical section, and the nominal shear strength Vc at the
critical section is determined from the following equations
Vc = VcACI ≥ 0.17 f c′bd for w/d ≥ 1.0 (1a)

28 ACI Structural Journal/January 2018

Table 2(a)—Evaluation of ultimate strength of pile cap BPC-30-25-1 by current design codes:
sectional methods
Bearing
Methods Flexure Pu,f, kN Pu ,bc, kN Pu ,bp, kN One-way shear Pu,1s, kN Two-way shear Pu,2s, kN Ultimate load Pu, kN
ACI 318-99 (1999) 902.5 3091.9 3496.8 864.1 890.1 864.1
ACI 318-14 (2014) 902.5 3091.9 3496.8 364.6 890.1 364.6
CRSI (2008) 902.5 3091.9 3496.8 1728.1 890.1 890.1

Notes: w is 125 mm (4.92 in.); d is 250 mm (9.84 in.); Pu,f = 2Mn,f /w, Mn,f is nominal flexural strength; Pu,bc = Pn,bc, Pn,bc is nominal bearing strength of column; Pu,bp = 4Pn,bp, Pn,bp
is nominal bearing strength of pile; Pu,1s = 2Vn,1s, Vn,1s is nominal shear strength of beam (ACI 318-99 Eq. (11.29), Eq. (a) of ACI 318-14 Table 22.5.5.1, Eq. (1)); Pu,2s = Vn,2s, Vn,2s
is nominal shear strength of two-way slab (Eq. (a) to (c) of ACI 318-14 Table 22.6.5.2, Eq. (2)); Pu = min(Pu,f, Pu,bc, Pu,bp, Pu,1s, Pu,2s); 1 kN = 0.2248 kip.

Table 2(b)—Evaluation of ultimate strength of pile cap BPC-30-25-1 by current design codes: ACI 318-14
strut-and-tie model approach
Concrete strut βs fc′, MPa fcu, MPa Fu, kN Aprov, mm2 Areq, mm2 Aprov /Areq Pu, kN
S1 0.51 29.1 14.84 321.5 13,717 21,663 0.63
Steel tie βt fy, MPa fcu, MPa Fu, kN Aprov, mm 2
Areq, mm 2
Aprov /Areq
T1 1.00 405 405 169.3 285 418 0.68
Nodal zone βn fc′, MPa fcu, MPa Fu, kN Aprov, mm2 Areq, mm2 Aprov /Areq
P/4 135.8 15,625 5491 2.85 543.3
CCC 0.85 29.1 24.74
S1 203.6 13,717 8230 1.67
R 135.8 17,671 9152 1.93
CTT 0.51 29.1 14.84 S1 203.6 22,963 13,717 1.67
T1 107.2 15,000 7225 2.08

Notes: βs, βt, and βn are coefficients of effective strength of strut, tie, and node; fcu is effective strength (= βs fc′ for strut, = βtfy for tie, = βnfc′ for node); FU is cross-sectional force
under experimental failure load; Aprov is maximum available area (refer to Fig. 2(b)); Areq is required area (= Fu/fcu); Pu is ultimate load (= minimum of Aprov/Areq × experimental
failure load); 1 kN = 0.2248 kip; 1 mm2 = 0.00155 in.2; 1 MPa = 145 psi.

d shown in Fig. 2(a) was used in this study. The load-carrying

Vc = Vc ≤ 0.83 f c′bd for w/d < 1.0 (1b)
w ACI
where fc′ is the compressive strength of concrete, in MPa;
w is the distance from the face of column to the center
of the nearest pile; b is the width of compression face of
the member; Vc ACI in Eq. (1a) is the same as Eq. (a) of
ACI 318-14, Table 22.5.5.1; and Vc ACI in Eq. (1b) is the
same as ACI 318-99 Eq. (11.29). For the shear design of a
pile cap that can be regarded as a two-way slab, the section
at a distance d/2 from the face of column is considered to
be the critical section and the nominal shear strength Vc
at the critical section is determined from Eq. (a) to (c) of
ACI 318-14, Table 2.6.5.2 and Eq. (2) for w/d ≥ 0.5 and w/d
< 0.5, respectively
d  d
Vc = 1 +  × 0.17 f c′bs d ≤ 2.67 f c′bo d (2)
w c the nodal zone. The procedure for evaluating the pile cap’s
where bo (mm) is the perimeter of critical section; and
bs equals 4 × c for a square column of dimension c. The
procedure for evaluating the ultimate strength of the pile
cap is shown in Table 2(a) with an example using specimen
BPC-30-25-1. The ultimate strengths of the other 77 pile
caps were determined in the same way.
The ultimate strength using the ACI 318-14 strut-and-tie
model approach was determined by examining the load-
carrying capacities of the concrete struts, steel and concrete
ties, and nodal zones of a selected strut-and-tie model. For the
pile caps, the ACI 445 3-D strut-and-tie model (Klein 2002)
capacities of the elements were examined by comparing the
maximum available cross-sectional areas of the elements
with the required cross-sectional areas. The available areas
of concrete struts and nodal zones of the strut-and-tie model
were determined using the ACI 445 (Klein 2002) approach,
which considers the geometrical shape of the selected strut-
and-tie model and the sizes of the loading and bearing plates.
The available areas of the concrete struts and nodal zones
are shown in Fig. 2(b). The total cross-sectional area of the
reinforcing bars placed within the effective width of steel
tie T1 was taken as the maximum available area of the steel
tie. The required areas of the struts and ties were determined
by dividing their cross-sectional forces by their effective
strengths. The required areas of nodal zone boundaries were
determined by dividing the cross-sectional forces of the struts
and ties framing the nodal zone by the effective strength of
ultimate strength is shown in Table 2(b) with an example of
the specimen BPC-30-25-1, wherein the minimum ratio of
the maximum available area to the required area was taken
to determine its ultimate strength.
The ultimate strength of the pile caps using the 3-D grid
strut-and-tie model approach was determined by the linear
analysis procedure presented in Yun et al. (2017). In the
linear analysis procedure, the ultimate strength of a concrete
member is obtained by estimating the maximum load that a
3-D grid strut-and-tie model for the concrete member can
carry by satisfying the condition of the strut-and-tie model’s

ACI Structural Journal/January 2018 29

Fig. 2—ACI 445 3-D strut-and-tie model for Pile Cap
BPC-30-25-1. (Note: 1 kN = 0.2248 kip; 1 mm = 0.0394 in.;
1 mm2 = 0.00155 in.2; 1 MPa = 145 psi.)
geometrical compatibility and by evaluating the tensile
stress of steel tie and the compressive stress of nodal zone
boundary at the maximum load. For the strength analysis of
Specimen BPC-30-25-1, the 3-D finite element (FE) model
shown in Fig. 3(a) was constructed. The model is composed
of eight-node unreinforced concrete elements. The external
load acting on the column was distributed to the nodes of the
finite elements of the column by considering tributary areas
of the nodes. The x- and z-directional horizontal rollers were
30
Fig. 3—3-D grid strut-and-tie model for Pile Cap BPC-30-
25-1. (Note: 1 kN = 0.2248 kip; 1 mm = 0.0394 in.)
placed at the FE nodes located at the interfaces of the piles
and the pile cap, and the y-directional vertical roller was
placed at the FE node located at the bottom center of the pile
cap. A FE linear elastic analysis of the model was conducted
and compressive principal stress flows were obtained. By
considering the stress flows, the position of the flexural
reinforcing bars, the clear cover, the locations of piles, and the
ratio of vertical to horizontal length of a grid element in two
directions, a 3-D grid strut-and-tie model composed of four
basic grid elements was constructed as shown in Fig. 3(b).
ACI Structural Journal/January 2018
Table 3(a)—Evaluation of ultimate strength of Pile Cap BPC-30-25-1 by 3-D grid strut-and-tie model
approach: ratio of cross-sectional areas under maximum load Pmax (=1158 kN)
Elements βs (βt) fc′, MPa fcu, MPa Aprov, mm2 Fu, kN Areq, mm2 Aprov /Areq
S1 1.15 29.1 33.57 16146 542.0 16146 1.000
T1 0.12 29.1 3.36 7784 25.4 7550 1.031
Notes: S1,T1: refer to Fig. 3(b); fcu = βs (βt) × fc′; Aprov is maximum available area in 3-D grid strut-tie model; Fu is cross-sectional force under maximum load Pmax; Areq is
required area (= Fu/fcu); 1 kN = 0.2248 kip; 1 mm2 = 0.00155 in.2; 1 MPa = 145 psi.

Table 3(b)—Evaluation of ultimate strength of Pile Cap BPC-30-25-1 by 3-D grid strut-and-tie model
approach: strength verification of steel tie under Pmax
Elements βt fy, MPa fcu, MPa Aprov, mm2 Fu, kN fs, MPa fy/fs Pfail, kN
T1 1.00 405 405 285.2 132.5 464.6 0.872 1009.7
Notes: fcu = βtfy; Fu is cross-sectional force under maximum load Pmax; fs = Fu/Aprov; Pfail = Pmax × fy/fs; 1 kN = 0.2248 kip; 1 mm = 0.00155 in. ; 1 MPa = 145 psi.
2 2

Table 3(c)—Evaluation of ultimate strength of Pile Cap BPC-30-25-1 by 3-D grid strut-and-tie model
approach: strength verification of nodal zone under Pmax
Connected
Nodal zone βn fc′, MPa fnz, MPa element σnz, MPa fnz/σnz Pfail, kN
CCC 3.45 29.1 100.5 S1 45.5 2.208 2559.2
CTT 0.60 29.1 17.5 S1 22.3 0.784 907.9

Notes: CCC, CTT: refer to Fig. 2(a); fnz = βnfc′; σnz is compressive stress of nodal zone boundary at Pmax; Pfail = Pmax × fnz/σnz; 1 kN = 0.2248 kip; 1 MPa = 145 psi.

The horizontal ties were placed at the center of the flexural
reinforcing bars. The external load acting on the column was
distributed to the nodes (of the grid elements) located at the
interface of the column and pile cap by considering tributary
areas of the nodes. In the 3-D grid strut-and-tie model, the
x- and z-directional horizontal rollers were placed at the FE
nodes located at the centers of the piles. The y-directional
vertical roller was placed at the FE node located at the bottom
center of the pile cap. Following the procedure presented in
Yun et al. (2017), the effective strengths of concrete struts
were determined. The yield strength of the reinforcing bar
and the tensile strength of the concrete were taken as the
effective strengths of the steel and concrete ties, respectively.
The required cross-sectional areas and the axial stiffness of
the struts and ties under the external load were determined
by using the simple iterative technique presented in Yun et
al. (2017). A structural analysis of the 3-D grid strut-and-tie
model was conducted and the effective strengths of the
nodal zones were determined by incorporating the structural
analysis results. After carrying out structural analyses of
the 3-D grid strut-and-tie model a few times according to
the linear analysis procedure presented in Yun et al. (2017),
the maximum load that the grid strut-and-tie model can
resist by satisfying the conditions of nodal zone strength
and the strut-and-tie model’s geometrical compatibility
was determined. The dimensioned grid strut-and-tie model
under the maximum load is shown in Fig. 3(c). In the figure,
the cross-sectional areas of struts and ties are depicted as
circular shapes for the purpose of easier visual verification
of the strut-and-tie model’s geometrical compatibility. The
maximum load that Specimen BPC-30-25-1 could resist
was 1158 kN (260.3 kip) (135% of the experimental failure
load). At the maximum load, the stress fs (464.6 MPa [67.4
ksi]) of the steel tie located at the bottom of the grid strut-
and-tie model was greater than the yield strength fy (405
MPa [58.7 ksi]) of the reinforcing bar with a ratio of fy/fs
= 0.872, and the compressive stress σnz of the nodal zone
boundary framed by the concrete strut S1 at the CTT node
was greater than the effective strength fnz of the nodal zone
with a ratio of fnz/σnz = 0.784. Therefore, according to the
definitions of the ultimate state of the strut-and-tie model’s
components given in Yun et al. (2017), 907.9 kN (204.1 kip)
(1158 kN [260.3 kip] × fnz/σnz, 105.8% of the experimental
failure load) was determined to be the ultimate strength of
Specimen BPC-30-25-1. The procedure for evaluating the
ultimate strength of the specimen is shown in Tables 3(a),
3(b), and 3(c). The ultimate strengths of the other pile caps
were determined in the same way.
The calculated ultimate strengths, coefficients of vari-
ance, dispersions, and correlation coefficients of 78 pile
caps are compared for several approaches in Fig. 4. As
shown in Fig. 4, the ultimate strengths were evaluated quite
conservatively with respect to test values using the ACI 318
sectional design method with the provisions of flexure,
bearing strength, and one-way slab shear. This indicates
that the use of a one-way slab shear provision is inappro-
priate for pile caps with w/d ≤ 5. The average ratios of the
experimental failure strength to the calculated strength (and
the percentages of failure mode agreement) based on three
approaches: 1) ACI 318 provisions for flexure, bearing
strength, and two-way slab shear; 2) ACI 318 provisions
for flexure, bearing strength, and deep beam and two-way
slab shear; and 3) the CRSI sectional design method; were
1.13 (92.3%), 1.17 (69.2%), and 1.12 (92.3%), respectively.
The calculated strengths of more than 40% of the pile caps
resulted in values greater than the experimental capacities
by the three approaches. The average ratio of the experi-
mental strength to calculated strength (and the percentage of

ACI Structural Journal/January 2018 31

Fig. 4—Ultimate strengths of pile caps. (Note: 1 kN = 0.2248 kip.)
failure mode agreement) using the ACI 318-14 strut-and-tie
model approach was 2.29 (85.9%); therefore, resulting in
excessive conservatism with respect to the experimental
failure strength. However, the 3-D grid strut-and-tie model
approach estimated the experimental failure strengths with
reasonable accuracy when compared with the experimental
capacity. The test to calculated capacity average ratio
(and the percentage of failure mode agreement) was 1.12
(91.0%), overcoming the significant conservatism observed
with strut-and-tie model approaches and sectional design
methods in the United States.
ANALYSIS OF SLAB-COLUMN JOINTS TESTS
The ultimate strengths of 19 slab-column joints, tested by
Yamada et al. (1992) and Ghannoum (1998) to investigate
the effects of tensile and compressive flexural reinforcement
32
ratios, patterns of shear reinforcing bars, and the shear
reinforcement ratio are estimated using the 3-D grid strut-
and-tie model approach and the two-way slab design
provisions of BS 8110 (1997), EC2 (2004), fib (2010), and
ACI 318-14 (2014). Details of the slab-column joint tests are
summarized in Table 4. The geometry of the specimens and
detailing of the flexural reinforcing bars are shown in Fig. 5
and 6. All the test specimens were failed by punching shear.
The ultimate strengths of the slab-column joints based on
BS 8110 sectional design method were determined using
the equation given in Table 3.8 of BS 8110 for the nominal
shear strength of a two-way slab. The ultimate strengths
of the slab-column joints based on EC2 and fib sectional
design methods were determined using EC2 Eq. (6.47), EC2
Eq. (6.52), and fib Eq. (7.3.40) and (7.3.42) for the nominal
shear strength of a two-way slab. The ultimate strengths of
ACI Structural Journal/January 2018
Table 4—Specification of slab-column joints
No. of
Investigators Specimens specimens Slab size, mm Column size, mm fc′, MPa d, mm ρf ρv Ptest, kN
T-series 6 2000 x 2000 300 x 300 21.6 to 24.4 180 1.23 0.0 to 1.53 441 to 762
Yamada et al. (1992)
K-series 7 2000 x 2000 300 x 300 25.9 to 27.8 180 1.53 0.0 to 1.98 658 to 1498
U-type 3 2300 x 2300 225 x 225 37.2 to 67.1 125 0.96 — 301 to 443
Ghannoum (1998)
B-type 3 2300 x 2300 225 x 225 37.2 to 67.1 125 1.92 — 317 to 485

Notes: d is effective depth of slab; ρf, ρv are flexural and shear reinforcement ratios; Ptest is experimental failure load; 1 kN = 0.2248 kip; 1 mm = 0.0394 in.; 1 MPa = 145 psi.

Fig. 5—Geometries and details of flexural reinforcing bars of slab-column joints (Yamada et al. 1992). (Note: 1 mm = 0.0394 in.)
the slab-column joints using the ACI 318-14 sectional design
method were determined using Eq. (a) to (c) of ACI 318-14
Table 22.6.5.2 and Eq. (a) of ACI 318-14 Table 22.6.6.2.
The ultimate strengths of the slab-column joints using
the 3-D grid strut-and-tie model approach were deter-
mined using the linear analysis procedure shown in Yun
et al. (2017). A detailed procedure for evaluating the ulti-
mate strengths is illustrated using Ghannoum’s (1998)
S1-U specimen. For the strength analysis of the specimen,
a 3-D FE model composed of eight-node unreinforced
concrete elements was constructed, as shown in Fig. 7(a).
The external loads acting on the loading plates were distrib-
uted to the nodes of the finite elements located at the loading
plates by considering tributary areas of the nodes. A y-
directional vertical roller was placed at the FE node located
at the bottom center of the column, and x- and z-directional
horizontal rollers were placed at the other nodes located at
the bottom of the column. A FE linear elastic analysis of the
model was conducted and the compressive principal stress
flows were obtained. A 3-D grid strut-and-tie model was
constructed by considering the compressive principal stress
flows, the position of the flexural reinforcing bars, the clear
cover and the ratios of the vertical-to-horizontal lengths of
a grid element in two directions, as shown in Fig. 7(b). The
horizontal tensile components of the grid elements were
placed at the center of flexural reinforcing bars. The external
loads acting on the loading plates were distributed to the FE
nodes (of the grid elements) located at the loading plates by
considering the tributary areas of the nodes. As boundary
conditions, the y-directional vertical roller at the FE node
located at the bottom center of the column and the x- and
z-directional horizontal rollers at the other nodes located at
the bottom of the column, were placed. Following the proce-
dure presented in Yun et al. (2017), the effective strengths
of concrete struts were determined by reflecting the state
of stresses of the 3-D FE model shown in Fig. 7(a). The

ACI Structural Journal/January 2018 33

Fig. 6—Geometry and detailing of flexural reinforcing bars of slab-column joints (Ghannoum 1998). (Note: 1 mm = 0.0394 in.)
yield strength of reinforcing bar and the tensile strength of
concrete were taken as the effective strengths of the steel
and concrete ties, respectively. The required cross-sectional
areas and axial stiffness of the struts and ties under the
external loads were determined by using the simple iterative
technique presented in Yun et al. (2017). After analyzing the
3-D grid strut-and-tie model a few times, the maximum load
that the grid strut-and-tie model can resist by satisfying the
strength conditions of the concrete struts was determined to
be 292 kN (65.6 kip) (97.0% of experimental failure load). At
the maximum load, the required cross-sectional areas of the
diagonal plane struts connected to the column in the bottom
x-z plane reached their maximum cross-sectional areas. The
dimensioned grid strut-and-tie model under the maximum
load is shown in Fig. 7(c). At the maximum load, the stress
at the four nodal zones located symmetrically at the bottom
of the grid strut-and-tie model, σcn (= 97.5 MPa [14.1 ksi]),
was greater than the effective strength of the nodal zones,
fcn (2.49fc′ = 92.2 MPa [13.4 ksi]). Therefore, according to
the linear analysis procedure of the approach with the defi-
nitions of ultimate states of strut-and-tie model components
(these are explained in Table 1 of Yun et al. 2017), 277.5 kN
(62.4 kip) (292 kN [65.6 kip] × fcn/σcn, 92.2% of the exper-
imental failure load) was determined to be the ultimate
strength of Specimen S1-U. Similarly, the ultimate strength
of the other slab-column joints was determined.
The average ratios of the experimental failure strength
to the evaluated strength based on the sectional design
methods of BS 8110, EC2, fib, and ACI 318-14 are 1.24,
0.98, 1.46, and 1.72, respectively, and the coefficients of
variation (COV) of each method are 35.8%, 35.2%, 34.3%,
and 52.9%, respectively. The 3-D grid strut-and-tie model
approach estimated the experimental failure strengths fairly
well and consistently with an experimental failure-to-
evaluated strength ratio and coefficient of variation of 1.16 Fig. 7—3-D grid strut-and-tie model for Slab-Column Joint
and 23.0%, respectively. All the test specimens by the 3-D S1-U. (Note: 1 mm = 0.0394 in.; 1 kN = 0.2248 kip.)

34 ACI Structural Journal/January 2018

Table 5—Specification of torsional beams
Investigators Specimens (A) fc′, MPa fly, MPa fty, MPa ρl ρt s, mm Ttest, kN-m
Series B 10 26.0 to 29.0 313.7 to 334.4 318.5 to 342.7 0.534 to 2.67 0.537 to 2.61 57 to 181 22.3 to 61.7
Series M 6 26.5 to 30.5 317.8 to 335.1 326.1 to 357.2 0.827 to 3.16 0.549 to 2.13 70 to 149 30.4 to 60.1
Series I 5 44.7 to 45.8 310.2 to 343.4 325.4 to 348.9 0.827 to 2.67 0.832 to 2.61 57 to 127 36.0 to 76.7
Hsu (1968)
Series J 4 14.3 to 16.9 319.9 to 338.5 331.6 to 346.1 0.534 to 1.60 0.537 to 1.61 92 to 152 21.5 to 40.7
Series G 8 26.8 to 31.0 319.2 to 338.5 321.3 to 349.6 0.400 to 1.58 0.402 to 1.60 86 to 187 26.8 to 73.5
Series N 6 27.3 to 30.4 330.9 to 352.3 337.8 to 360.6 0.611 to 1.42 0.622 to 1.42 51 to 92 9.0 to 15.7
Series B30 3 36.3 to 41.7 604.7 to 637.8 665.3 to 672.2 3.521 1.5 90 15.3 to 16.6

Rasmussen and Series B50 3 57.1 to 61.8 612.3 to 614.3 665.3 3.521 1.5 90 18.5 to 20.0
Baker (1995) Series B70 3 76.2 to 77.3 614.3 to 617.1 655.7 to 663.3 3.521 1.5 90 20.1 to 21.0
Series B110 3 105.0 to 109.8 617.8 to 634.3 655.0 to 659.8 3.521 1.5 90 23.6 to 24.8
Koutchoukali and
Series B 9 39.6 to 93.9 373.0 to 386.1 373.0 to 399.2 0.500 to 0.625 0.009 108 18.4 to 24.0
Belarbi (2001)

Notes: A is number of specimens; fly, fty are yield strengths of longitudinal and transverse reinforcing bars; ρl, ρt are longitudinal and transverse reinforcement ratios; s is spacing of
transverse reinforcing bars; Ttest is experimental failure load; 1 kN-m = 8.85 kip-in.; 1 mm = 0.0394 in.; 1 MPa = 145 psi.

Table 6—Ultimate strengths of torsional beams

No. of Rahal and Collins AASHTO-LRFD
Investigators specimens Hsu and Mo (1983) (1995) (2010) ACI 318-14 (2014) Present approach
Hsu (1968) 39 0.98 1.00 0.66 1.02 1.08
Rasmussen and
12 0.55 0.59 0.34 0.64 1.05
Baker (1995)
Koutchoukali and
9 0.82 0.80 0.67 1.01 1.02
Belarbi (2001)
Mean of Ttest/Tpredicted 0.87 0.88 0.60 0.94 1.06
Coefficient of variation, % 21.1 22.7 30.2 27.3 20.4

grid strut-and-tie model approach were failed by punching 2 Ao At f yt

shear (failure of inclined concrete struts encasing columns).
This indicates that the approach can be used effectively for
the strength analysis and design of slab-column joints by
overcoming the inaccuracies and inconsistencies of current
sectional design methods.
ANALYSIS OF TORSIONAL BEAMS
The verification of the 3-D grid strut-and-tie model
approach was carried out using the ultimate strength of 60
reinforced concrete beams tested to pure torsion to failure by
Hsu (1968), Rasmussen and Baker (1995), and Koutchou-
kali and Belarbi (2001). For the purposes of comparison
with current codes of practice, the ultimate strength of the
beams was also compared with AASHTO-LRFD (2010) and
ACI 318-14 (2014) design provisions, Hsu and Mo (1983)
method, and Rahal and Collins (1995) method. The data for
the torsional beams are listed in Table 5, and the geometry
and details of the longitudinal reinforcement are shown in
Fig. 8. All the torsional beams were failed by yielding of
transverse reinforcing bar(s) and diagonal cracks.
The ultimate strength of the torsional beams based on the
design provisions of ACI 318-14 and AASHTO-LRFD was
determined using the following equation for the nominal
torsional strength Tn (kN)
Tn = cot θ (3)
s
where Ao is the gross area enclosed by the shear flow path
(≈0.85Aoh , where Aoh is the area enclosed by the centerline
of the outermost closed transverse torsional reinforcement);
At is the area of one leg of a closed stirrup resisting torsion
within spacing s; fyt is the yield strength of the transverse
reinforcing bar; and θ is the angle between the axis of the
strut and the tension chord of the torsional beam. As the
specimens tested were not prestressed, 45 degrees was taken
as angle θ in applying the provisions of ACI 318-14. In
applying the provisions of AASHTO-LRFD, the angle was
determined by the method described in Section 5.8.3 of the
LRFD Specifications, incorporating the nominal shear stress
and the longitudinal strain of the torsional beam. The details
of the methods by Hsu and Mo (1983) and Rahal and Collins
(1995) can be found in the references provided and is not
discussed herein.
The ultimate strength of each of the specimens tested was
determined using the nonlinear analysis procedure of the
3-D grid strut-and-tie model approach. The detailed proce-
dure for evaluating the ultimate strength is illustrated with
an example of Beam B110-1 tested by Rasmussen and Baker
(1995). For the strength analysis of the beam, a 3-D FE
model for one-half of the effective span was developed, as

ACI Structural Journal/January 2018 35

Fig. 8—Geometries and details of flexural reinforcing bars of torsional beams. (Note: 1 mm = 0.0394 in.)

Fig. 9—3-D grid strut-and-tie model for Torsional Beam B110-1. (Note: 1 mm = 0.0394 in.)
shown in Fig. 9(a). The torsional force was linearly distrib-
uted to the nodes of the finite elements modeling the end
region of the beam. 3-D hinges were imposed at the nodes
located at the mid-section of the beam. A FE linear elastic
analysis of the model was performed, and the compressive
principal stress flows were obtained as shown in Fig. 9(b).
As shown in Fig. 9(c), a 3-D grid strut-and-tie model was
constructed by considering the compressive principal stress
flows, the positions of transverse and longitudinal rein-
forcing bars, the clear cover of the concrete, and the ratios
of the vertical-to-horizontal lengths of a grid element in two
directions. The torsional force was distributed to the nodes
(of the basic grid elements) located at the end region of the
beam by considering tributary areas of the nodes. 3-D hinges
were imposed at the nodes of the grid elements located at the
midsection of the beam. Following the procedure presented
in Yun et al. (2017), the effective strengths of the concrete
struts were determined. The yield strength of reinforcing bar

36 ACI Structural Journal/January 2018

Fig. 10—Geometry, material properties, and load cases of pier cap. (Note: 1 mm = 0.0394 in.; 1 MPa = 145 psi.)
and the tensile strength of concrete were taken as the effec-
tive strengths of the steel and concrete ties, respectively. The
required cross-sectional areas and axial stiffness of the struts
and ties under the external loads were determined by using
the simple iterative technique presented in Yun et al. (2017).
After analyzing the 3-D grid strut-and-tie model a few times,
the maximum torsional force that the grid strut-and-tie model
can transfer by satisfying the strength conditions of concrete
struts was determined to be 95.2 kN-m (843.2 kip-in.) (385%
of experimental failure force). At the maximum torsional
force, the required cross-sectional areas of the inclined
plane struts (Struts 742 and 743 in Fig. 9(c)) reached their
maximum cross-sectional areas. The dimensioned grid strut-
and-tie model under the maximum torsional force is shown
in Fig. 9(d). To determine the ultimate strength of Spec-
imen B110-1 using the nonlinear analysis procedure of the
approach, a FE material nonlinear analysis of the 3-D grid
strut-and-tie model was performed. The maximum torsional
force was applied incrementally with 40 load steps. At the
10th step, the grid strut-and-tie model became unstable with
sudden increases of nodal displacements. Therefore, by the
definitions of ultimate states given in Table 1 of Yun et al.
(2017), 23.8 kN-m (210.8 kip-in.) (95.2 × 10/40, 96.3% of
experimental failure force) was determined to be the ulti-
mate strength of Specimen B110-1. Following a similar
approach, the ultimate strength of the other specimens tested
was determined.
The ultimate strength of each of the 60 torsional beams
evaluated with the existing and proposed methods are
compared in Table 6. The average ratios of the experimental
strength to the calculated strength by the methods or design
provisions of Hsu and Mo (1983), Rahal and Collins (1995),
AASHTO-LRFD, and ACI 318-14 are 0.87, 0.88, 0.60, and
0.94, respectively. It must be noted that all the methods over-
estimated the experimental capacities. On the other hand,
the ultimate strengths were evaluated satisfactorily by the
proposed 3-D grid strut-and-tie model approach. Thus, the
ACI Structural Journal/January 2018
proposed approach can be conservatively used for members
under torsion. The failure of all the beams predicted by the
proposed approach occurred by diagonal cracks (failure of
inclined concrete struts).
DESIGN OF PIER CAP
To verify the capability and effectiveness of the 3-D grid
strut-and-tie model approach in the design of a concrete
member with multiple load combinations, we selected a
bridge pier cap subjected to three load cases with 3-D loads.
The shear span-to-effective depth ratio of the cap is 1.25.
The design conditions including geometry, material proper-
ties, and load cases are given in Fig. 10.
Following the linear analysis procedure presented in
Yun et al. (2017), the design of the pier cap was conducted
using the 3-D grid strut-and-tie model shown in Fig. 11(a).
The grid model was constructed using basic grid elements
with vertical-to-horizontal length ratios of 0.96 and 1.09
in two directions. The auxiliary grid elements were also
used to model the inclined regions of the pier cap and the
outer regions of the round column. The AASHTO-LRFD
(2010) strength-reduction factors of 0.75 and 0.90 were used
for  the concrete struts and steel ties. The dimensioned strut-
and-tie model for load case one is shown in Fig. 11(b). In Load
Case 1 with the dead and live loads acting vertically, most of
the design loads at loading plates 2, 3, and 4 were transferred
to the pier directly by the concrete struts connecting the
loading plates and column. On the other hand, most of the
design loads at loading plates 1 and 5 were transferred to
the pier by the curved-shaped paths described by multiple
concrete struts. The pier cap subjected to load cases 2 and
3 with dead and live loads acting in three directions were
designed with the same 3-D grid strut-and-tie model shown
in Fig. 11(b). This illustrates the versatility of the approach
to handle multiple loading, including 3-D loads, by using a
single type of 3-D grid strut-and-tie model. The conditions
of the strut-and-tie model’s geometrical compatibility under
37

Fig. 11—3-D grid strut-and-tie model for pier cap. (Note:
1 kN = 0.2248 kip.)
all three load cases were satisfied, and the strengths of all
nodal zones were enough to transfer the strut and tie forces.
Thus, the maximum cross-sectional area of a steel tie at a
certain position under the three load cases was taken as the
required area of the reinforcing bar that should be distributed
with the effective area of the steel tie. The required areas
of reinforcing bars corresponding to the other steel ties
were obtained and distributed in the same way. To complete
the design, the code provisions for auxiliary or minimum
reinforcement must be observed.
To compare the design results of the 3-D grid strut-and-tie
model approach, the designs of the pier cap were carried out
using the ACI 318-14 flexural design method, ACI 318-14
bracket provisions, and ACI 318-99 deep beam provisions.
The pier cap was also designed based on the strut-and-tie
model approaches of current design codes, with the
two-dimensional (2-D) strut-and-tie models shown in Fig.
11(a) and 11(b) representing arch and truss load-transfer
mechanisms, respectively. As the 2-D strut-and-tie models
are valid for 2-D load cases, only Load Case 1 (Fig. 11)
was considered. Nodes F and G in Fig. 12 were placed at
a vertical distance of 200 mm (7.87 in.) from the plane in
38
which the column ends. The vertical positions of both nodes
are the same as those of corresponding nodes of the 3-D
grid strut-and-tie model shown in Fig. 11. The nodes were
also placed at a horizontal distance of 620 mm (24.4 in.)
from the boundary of the column with a transformed square
cross section. The horizontal positions of both nodes can be
moved toward the boundary of the pier by considering the
required cross-sectional areas of the concrete struts located
in the pier.
The required cross-sectional areas of reinforcing bars
obtained using each method are compared in Fig. 13. From
the design results, two features can be observed. First, more
flexural reinforcing steel is required by the strut-and-tie
model approach than by the other methods. This is mainly
due to the positions of nodes F and G shown in Fig. 12 and
the corresponding nodes shown in Fig. 11. If these nodes
are moved toward points F and G by considering the critical
section (the S-S line in Fig. 12) defined by ACI 318-14 flex-
ural design and bracket provisions, similar flexural design
results will be obtained. Second, more vertical shear rein-
forcing bars are required by the strut-and-tie model of truss
mechanism than by the 3-D grid strut-and-tie model. This
is because the design loads at loading plates 1 and 5 are
transferred to the column mainly by truss action in the strut-
and-tie model of the truss mechanism. On the other hand, the
design loads are transferred to the pier by the combined arch
and truss mechanisms in the 3-D grid strut-and-tie model.
In other words, a portion of the design loads are transferred
directly to the pier through arch-type concrete struts in the
3-D grid strut-and-tie model. These design outcomes of
the pier cap (with a shear span-to-effective depth ratio of
1.25) based on the 3-D grid strut-and-tie model approach
are supported by the statically indeterminate strut-and-tie
models recommended by Foster and Gilbert (1998), fib
(2010), and Kim and Yun (2011). The recommended models,
consisting of the arch and truss load transfer mechanisms,
are known to be most appropriate for deep beams with shear
span to effective depth ratios of 0.45 < a/d < 1.80, 0.90 < a/d
< 1.56, and 0.50 < a/d < 3.00, respectively.
SUMMARY AND CONCLUSIONS
The strut-and-tie model approach in current design codes
incorporated with simple truss-type 2-D strut-and-tie models
for the design of structural concrete with disturbed regions
in most cases is not sufficient to properly represent the
complex load transfer mechanisms encountered in some 3-D
disturbed regions. To overcome this limitation, the authors
have proposed a 3-D grid strut-and-tie model approach in
the authors’ previous paper. The approach employs grid
strut-and-tie models to represent the complex load transfer
mechanisms, and to determine the effective strengths of 3-D
concrete struts and nodal zones. The approach is envisioned
for use in practice and not only as a research tool. The same
basic concepts have been adopted of the strut-and-tie model
approaches used in current design codes in terms of the
strength verification of struts, ties, and nodal zones.
The approach was used to estimate the experimental capac-
ities of pile caps, slab-column joints, and torsional beams.
The average ratios of the experimental strength to calculated
ACI Structural Journal/January 2018
Fig. 12—Two-dimensional strut-and-tie model for pier cap. (Note: 1 kN = 0.2248 kip; 1 mm = 0.0394 in.)

Fig. 13—Design results of pier cap. (Note: 1 cm2 = 0.155 in.2)

ACI Structural Journal/January 2018 39

strength (and COV) by ACI 318-14 sectional design methods
for pile caps, slab-column joints, and torsional beams were
1.13 (29.3%), 1.72 (52.9%), and 0.94 (27.3%), respectively.
The average ratios (and COV) by EC 2 and fib for slab-
column joints were 0.98 (35.2%) and 1.46 (34.3%). The
average ratios (and COV) by AASHTO-LRFD and Rahal
and Collins (1995) method for torsional beams were 0.60
(30.2%) and 0.88 (22.7%). The average ratios (and COV) by
the proposed approach for pile caps, slab-column joints, and
torsional beams were 1.12 (12.8%), 1.16 (23.0%), and 1.06
(20.4%), respectively. The approach evaluated the experi-
mental capacities fairly well and consistently, and enabled
the design of the pier cap by representing its load transfer
mechanisms properly. By considering the strength analysis
and design results of this study, the 3-D grid strut-and-tie
model approach can help overcome the limitations and
uncertainties of the strut-and-tie model approaches used in
current design codes.
AUTHOR BIOS
Young Mook Yun is a Professor in the Department of Civil Engineering
at Kyungpook National University, Daegu, Korea. He received his PhD
from Purdue University, West Lafayette, IN, in 1995. His research inter-
ests include strut-and-tie model analysis and design of concrete members,
computational mechanics, and computer graphics.
Hyun Soo Chae is a Deputy Head in Program Development Department
of Hangil IT, Seoul, Korea. He received his PhD from Kyungpook National
University in 2012. His research interests include numerical and strut-
and-tie model analysis/design of disturbed regions in concrete members.
Byunghun Kim is a Lead Engineer in the Infrastructure Department of
Hyundai Engineering, Seoul, Korea. He received his PhD from Kyungpook
National University in 2004. His research interests include strut-and-tie
model design of concrete structures and development of numerical analysis
methods for steel and reinforced concrete structures.
Julio A. Ramirez, FACI, is a Professor in the Lyles School of Civil Engi-
neering at Purdue University. He is a member of Joint ACI-ASCE Commit-
tees 408, Bond and Development of Steel Reinforcement, and 445, Shear
and Torsion. He has received the 2000 ACI Delmar Bloem Award and the
2006 ACI Joe W. Kelly Award.
ACKNOWLEDGMENTS
This research was supported by Basic Science Research Program through
the National Research Foundation of Korea (NRF) funded by the Ministry
of Education, Science and Technology (NRF-2015R1D1A1A01061333).
REFERENCES
American Association of State Highway and Transportation Officials,
2010, AASHTO LRFD Bridge Design Specifications, fifth edition, Wash-
ington, DC.
40
ACI Committee 318, 1999, “Building Code Requirements for Struc-
tural Concrete (ACI 318-99) and Commentary (ACI 318R-99),” American
Concrete Institute, Farmington Hills, MI, 369 pp.
ACI Committee 318, 2014, “Building Code Requirements for Struc-
tural Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 519 pp.
British Standards Institution, 1997, “Structural Use of Concrete: BS
8110,” Milton Keynes, UK.
Clarke, J. L., 1973, “Behavior and Design of Pile Caps with Four Piles,”
Report No. 42.489, Cement and Concrete Association, London, UK.
CEB-FIP, 2010, “CEP-FIP Model Code,” Comité Euro-International
du Béton, International Federation for Structural Concrete (fib), Lausanne,
Switzerland.
Concrete Reinforcing Steel Institute, 2008, CRSI Handbook, Chicago, IL.
CEN, 2004, “Eurocode 2: Design of Concrete Structures,” European
Committee for Standardization, Brussels, Belgium, 229 pp.
Foster, S. J., and Gilbert, R. I., 1998, “Experimental Studies on High-
Strength Concrete Deep Beams,” ACI Structural Journal, V. 95, No. 4,
July-Aug., pp. 382-390.
Ghannoum, G. M., 1998, “Effect of High-Strength Concrete on the
Performance of Slab-Column Specimens,” MS thesis, Department of
Civil Engineering and Applied Mechanics, McGill University, Montreal,
QC, Canada.
Hsu, T. T. C., 1968, “Torsion of Structural Concrete—Interaction Surface
for Combined Torsion, Shear, and Bending in Beams Without Stirrups,”
ACI Journal Proceedings, V. 65, No. 1, Jan., pp. 51-60.
Hsu, T. T. C., and Mo, Y. L., 1983, “Softening of Concrete in Torsional
Members,” Research Report No. ST-TH-001-83, Department of Civil Engi-
neering, University of Houston, Houston, TX.
Kim, B. H., and Yun, Y. M., 2011, “An Indeterminate Strut-Tie Model
and Load Distribution Ratio for RC Deep Beams—(I) Model & Load
Distribution Ratio,” Advances in Structural Engineering, V. 14, No. 6,
pp. 1031-1042.
Klein, G. J., 2002, “Example 9: Pile Cap,” Examples for the Design for
Structural Concrete with Strut-and-Tie Models, SP-208, K.-H. Reineck, ed.,
American Concrete Institute, Farmington Hills, MI, pp. 213-223.
Koutchoukali, N., and Belarbi, A., 2001, “Torsion of High-Strength
Reinforced Concrete Beams and Minimum Reinforcement Requirement,”
ACI Structural Journal, V. 98, No. 4, July-Aug., pp. 462-469.
Rahal, K. N., and Collins, M. P., 1995, “Analysis of Sections Subjected
to Combined Shear and Torsion—A Theoretical Model,” ACI Structural
Journal, V. 92, No. 4, July-Aug., pp. 459-469.
Rasmussen, L. J., and Baker, G., 1995, “Torsion in Reinforced Normal
and High-Strength Concrete Beams—Part 1: Experimental Test Series,”
ACI Structural Journal, V. 92, No. 1, Jan.-Feb., pp. 56-62.
Sabnis, G. M., and Gogate, A. B., 1984, “Investigation of Thick Slab
(Pile Cap) Behavior,” ACI Journal Proceedings, V. 81, No. 1., Jan.-Feb.,
pp. 35-39.
Suzuki, K.; Otsuki, K.; and Tsubata, T., 1998, “Influence of Bar Arrange-
ment on Ultimate Strength of Four-Pile Caps,” Transactions of the Japan
Concrete Institute, V. 20, pp. 195-202.
Suzuki, K.; Otsuki, K.; and Tsuchiya, T., 2000, “Influence of Edge
Distance on Failure Mechanism of Pile Caps,” Transactions of the Japan
Concrete Institute, V. 22, pp. 361-368.
Yamada, T.; Nanni, A.; and Endo, K., 1992, “Punching Shear Resistance
of Flat Slabs: Influence of Reinforcement Type and Ratio,” ACI Structural
Journal, V. 88, No. 4, July-Aug., pp. 555-563.
Yun, Y. M.; Kim, B. H.; and Ramirez, J. A., 2017, “Three-Dimensional-
Grid Strut-and-Tie Model Approach in Structural Concrete Design,” ACI
Structural Journal, V. 115, No. 1, Jan., pp. 15-26.
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S03
Time-Dependent Buckling Testing of Eccentrically Loaded
Slender High-Strength Concrete Panels
by Yue Huang, Ehab Hamed, Zhen-Tian Chang, and Stephen J. Foster
The results of nine full-scale high-strength concrete panels tested
under eccentric sustained loads are presented in this paper. The
panels were uniaxially loaded for up to 4 months. Four panels
failed due to creep buckling at different times under sustained loads
that are lower than their short-term capacities. The other panels
exhibited a long-term stable behavior; they were then loaded to
failure without the release of the existing sustained loads and
exhibited a reduction in their residual strength due to creep. The
study investigates the influences of aging of concrete, magnitude
of the applied load, and the slenderness ratio. The experimental
results are compared to theoretical predictions generated from a
theoretical analysis that was previously developed by the authors.
The formulation is based on the rheological generalized Maxwell
model and it accounts for large deformation kinematics in the
structural level. A close correlation is achieved between the exper-
imental and theoretical results.
Keywords: buckling; cracking; creep; high-strength concrete; panels.
INTRODUCTION
The use of high-strength concrete (HSC), whose compres-
sive strength is over 55 MPa (7975 psi), has been rapidly
growing over the past decades due to its superior prop-
erties such as higher strength, improved durability, and
lighter-weight construction in comparison to conventional
normal-strength concrete (NSC). HSC panels have been
widely used in engineering structures such as load-bearing
walls, core walls in high-rise buildings, box girders in
bridges, and hulls of offshore structures. Usually, HSC
panels are more slender than typical NSC panels because
of their higher stress capacity. The increased slenderness
of the panel highlights the need for investigating the buck-
ling capacity of the panel, and its degradation with time due
to creep, shrinkage, and time-dependent cracking of the
concrete. This is the focus of this study.
Vertical HSC wall panels are subjected to in-plane
compressive loads and out-of-plane bending moments. The
moments can be a result of vertical load eccentricity, initial
imperfection, construction inaccuracy, and moments due
to dynamic lateral actions such as wind gusts and seismic
inputs. Such loading combinations may lead to buck-
ling failures of slender panels or material failures in short
panels that are characterized by crushing of the concrete
and/or yielding of the steel reinforcement. In moderately
slender walls, which is the most common case in practice,
such material failures can be triggered or accelerated by
the geometric nonlinearity (P-Δ effect), and they can be
observed at load levels that are much lower than what is
predicted from a geometrically linear analysis that does not
ACI Structural Journal/January 2018
TECHNICAL PAPER
consider the P-Δ effect. Therefore, there is a need to look
at the combined geometric and material nonlinearity of the
problem. However, in most cases, a considerable portion of
the load can be classified as a sustained load that leads to
an increase of the axial and out-of-plane deformations with
time due to creep and the geometric nonlinearity, which
increase the bending moments due to the increased P-Δ
effect as a result of increasing Δ with time. This may lead to
cracking of the wall and, consequently, may lead to loss of
stability at certain time—so-called “creep buckling.” Such
behavior was observed in reinforced concrete (RC) columns
and shells.1-3 The creep and shrinkage deformations may not
necessarily lead to buckling failure, but they may increase
the internal stresses and decrease the residual strength and
the factor of safety of the wall.
The instantaneous buckling behavior of NSC and HSC
panels has been extensively investigated both theoretically
and experimentally by numerous studies.4-7 Nevertheless,
no experimental study could be found in the literature
regarding the influence of creep on the buckling capacity of
HSC panels, which is the contribution of this paper. Among
the studies that focused on the buckling response of panels
without creep, Oberlender and Everard4 tested 54 one-way,
simply supported NSC panels to failure with varying slen-
derness ratios, concrete strength, and reinforcement ratios.
Saheb and Desayi5 studied the influence of vertical and hori-
zontal reinforcement ratios, as well as the slenderness and
aspect ratio (height/width) on the behavior of one-way NSC
wall panels. The ultimate strength of the panel was found to
vary linearly with respect to the vertical reinforcement ratio
and to the aspect ratio. Fragomeni and Mendis6 numerically
studied the stability of NSC and HSC wall panels. Huang
et al.7 tested eight full-scale HSC panels under short-term
eccentric loading with different load eccentricities, reinforce-
ment ratios, and slenderness ratios. A theoretical model
that considered material nonlinearity, cracking, and the
tension-stiffening effect was also developed and validated.
While no experimental studies regarding the time-dependent
buckling behavior of wall panels could be found in the litera-
ture, a number of experimental works have been undertaken
to study the geometrically nonlinear response (P-Δ effect)
of RC columns under sustained loads.8-13 Despite the simi-
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2015-382.R2, doi: 10.14359/51700913, received February 22, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
41
larities in the modeling and behavior between RC columns
and one-way panels, there are some differences that require
attention and justify the need for creep testing of HSC
panels. The concrete in panels is unconfined due to the
absence of stirrups, and panels have significantly smaller
reinforcement ratios. Due to these differences, the role and
importance of the tension-stiffening effect is magnified, but
most importantly, there is less restraint to the creep defor-
mations of the concrete. Among the studies that focused on
the time-dependent creep buckling response of RC columns,
Green and Breen8 tested 10 identical NSC columns under
eccentric sustained loads. Two columns failed under the
sustained loads within 2 months. The remaining columns
continued to deform with time with no indication of failure
for 1.5 years. Goyal and Jackson9 reported experimental Fig. 1—Schematic description of tested panels and
results of 20 NSC columns tested under sustained loading. typical cross section. (Note: All dimensions in mm; 1 mm
The residual strength of the long-term loaded columns = 0.0394 in.)
ranged from 82% to 102% of the short-term capacity.
ison aims to validate the model, which can then be used to
Claeson and Gylltoft12 tested one NSC column and one HSC
generate comprehensive parametric studies.
column under eccentric sustained loading. The NSC column
failed after 22 days under a sustained load level of 80% of
RESEARCH SIGNIFICANCE
the short-term failure load, while the HSC column exhibited
Slender HSC wall panels may experience excessive out-of-
little increase in the deflection after 6 months and thereby
plane deflections and cracking over time under combined
was loaded to failure at 190 days. The load level at failure
axial and transverse loads due to creep and shrinkage. More-
was 85 to 89% of its accompanying column that was tested
over, they may fail by creep buckling. A significant number
on the same day under short-term loading, which highlighted
of studies have been conducted to study the instantaneous
the influence of creep on the load-carrying capacity.
behavior of RC panels and the time-dependent responses
A limited number of studies have been undertaken to
of RC columns. However, no studies have focused on the
investigate the time-dependent behavior of RC panels in
time-dependent buckling behavior of HSC panels, which are
the literature. However, they focused on the serviceability
more slender than typical NSC panels. In this study, nine HSC
response rather than the ultimate buckling capacity of
panels were tested under sustained axial loads, which showed
the panel under sustained loads. Moreover, the panels in
that creep can lead to a significant reduction in the load-car-
these studies were made of NSC rather than HSC. Tatsa14
rying capacity of HSC wall panels. The test results were also
tested seven concrete panels under low levels of eccentric
used to validate a theoretical model that was previously devel-
sustained load at one end and concentric sustained load at
oped by the authors,16 and which can be further used in the
the other end. It was observed that the neutral axis shifts
future for conducting more comprehensive parametric studies
toward the compression face with time. It was also found
that are beyond the specific geometry, materials, and loadings
that there was a 23% reduction in the residual strength of
used in this experimental program. The theoretical model of
the panel when tested for failure after 4 months of creep.
Huang and Hamed16 was selected because it has many advan-
Lee et al.15 investigated the time-dependent behavior of NSC
tages over existing models and it can describe the structural
panels, where the sustained load was applied concentrically
behavior at different scales. The experimental results also
on part of the entire width of the wall. Following the test
provide a benchmark database to the literature.
results, and in conjunction with a finite element analysis, an
effective width coefficient was proposed to be included in
EXPERIMENTAL INVESTIGATION
the structural analysis and design of concrete panels.
Test specimens
This paper reports the results of nine panels that were
The experimental investigation includes testing of nine
tested to investigate the long-term behavior of one-way HSC
slender, high-strength reinforced concrete wall panels under
panels under eccentric sustained in-plane loads. The panels
eccentric sustained in-plane compressive loads along their
were simply supported along the two short edges. The influ-
short edges. Panels LT1 to LT8 had the same dimensions,
ences of the loading age, level of the in-plane load, and slen-
where the length (L), width (b), and thickness (h) were 2700,
derness ratio on the time-dependent behavior were studied.
460, and 100 mm (106.3, 18.1, and 3.94 in.), respectively
The test findings represent a benchmark database that can be
(Fig. 1). Panel LT9 had the same length and width as LT1
used for further long-term stability studies of HSC panels.
to LT8 but with a thickness of 130 mm (5.12 in.). All panels
Moreover, the experimental results are compared with the
were equally reinforced at the top and bottom layers and
estimations generated by a theoretical model proposed by
had a total reinforcement ratio of 0.22% for LT1 to LT8
the first two authors (Huang and Hamed16). The theoret-
and 0.28% for LT9. Welded wire reinforcement SL52, with
ical model considers geometric nonlinearity in conjunction
bar diameter of 4.77 mm (0.188 in.) at spacing of 200 mm
with concrete creep, shrinkage, and cracking. The compar-
(7.87 in.) in two orthogonal directions, was used as the

42 ACI Structural Journal/January 2018

Fig. 2—Test setup.

Table 1—Tested panels

Concrete age upon Sustained load Predicted short-term edesign*, mm
Panel No. Batch No. L/r loading (t0), days level, kN failure load Pu, kN Sustained load/Pu [edesign/h] eL†, mm
2
eR‡, mm
3

LT1 1 94 22 591 638 93% 16.7 [1/6] 22.5 21.9

LT2 94 146 320 436 73% 33.3 [1/3] 31.1 37.5
2
LT3 94 99 615 820 75% 16.7 [1/6] 19.1 18.8
LT4 94 22 369 499 74% 25 [1/4] 29 31
3
LT5 94 22 548 764 72% 16.7 [1/6] 19.6 21.4
LT6 94 22 490 747 66% 16.7 [1/6] 21.7 20
4
LT7 94 22 808 1526 53% 8.3 [1/12] 7.2 7.1
LT8 94 22 927 1301 71% 8.3 [1/12] 9.5 8.6
5
LT9 72 22 1020 1075 95% 21.7 [1/6] 32.6 28.1
*
Designed eccentricity.
†
Measured eccentricity at left end of tested panel.
‡
Measured eccentricity at right end of tested panel.
Notes: 1 mm = 0.0394 in.; 1 kN = 0.225 kip.

reinforcement for Panels LT1 to LT8, and SL72 mesh with
bar diameter of 6 mm (0.236 in.) and spacing of 200 mm
(7.87 in.) was used for Panel LT9. The concrete cover was
20 mm (0.787 in.) for all panels. The test setup is schematically
depicted in Fig. 2, along with a typical 100 mm (3.94 in.) thick
cross section. Although the panels are intended to work verti-
cally, they were tested horizontally for safety and simplicity
reasons. This is justified because the level of bending caused
by the self-weight and any other imperfections is negligible
compared to the bending caused by the load eccentricity.
Three parameters were examined in the tests, including the
age of loading; the ratio of the sustained load to the failure
load; and the effective slenderness ratio, defined as L/r, where
r is the radius of gyration = r = I eff /Aeff , and Ieff and Aeff are
the effective moment of inertia and cross section area, respec-
tively. The details of the loading conditions are reported in
Table 1. Although the reinforcement ratio is a parameter that
needs to be considered in general, a previous study conducted
by the authors regarding the short-term buckling response of
HSC slender panels (Huang et al.7) has shown that the behavior
is greatly influenced by the load eccentricity and slenderness
of the panel, and much less influenced by the reinforcement
ratio. Based on these findings and due to time limitations, it
was decided not to investigate this parameter. The sustained
load ratio for each panel is determined as the ratio of the
measured sustained load divided by the corresponding short-
term failure load, which was predicted by a theoretical model
developed in Huang et al.7 In Huang et al.,7 the model was
compared with test results generated by the authors for panels
possessing the same dimensions as the ones tested herein, and
with other test results from the literature.5,17 For the predic-
tion of the failure loads, the measured material properties
and actual eccentricities were used. As the sustained loading
testing normally takes a significantly long time, the load ratios
for the specimens in the present study were designed to be
higher than conventional sustained loading tests so that the
panels might fail within a short time since loading (that is, up
to 6 months).

ACI Structural Journal/January 2018 43

Table 2—Material properties of concrete at start
and end of panel testing
Elastic Tensile
Batch Concrete Compressive modulus Ec, strength ft,
No. age, days strength fc′, MPa GPa MPa
22 80.9 39.4 5.3
1
38 89.5 40.3 6.4
99 91.7 40.25 6.9
2
203 95 40.3 7
22 101 39.5 5.9
3 91 103.5 39.8 8.9
126 103 40.5 8.9
22 89.1 39.5 6.2
4
134 100 43.4 7.7
22 88 39.4 5.5
5 30 90.5 39.8 6.0
104 98 41.1 6.5

Notes: 1 MPa = 0.145 ksi; 1 GPa = 145 ksi.

The eccentricities reported herein include the design

eccentricities (edesign) as well as the actual eccentricities at
the left (eL) and right edges (eR) of the panel obtained from
the test. The actual eccentricities were determined as follows

Ec I eff ε top − ε bot

etest = (1)
N test h

where etest is the actual eccentricity in the test at the left edge
(eL) or the right edge (eR) (refer to Table 1); Ntest, εtop, and εbot
are the experimental axial load, and the measured concrete
strain at the top and bottom faces of the panel, respectively.
The effective moment of inertia is obtained using the trans-
formed section method. Equation (1) is based on a linear
strain distribution through the thickness of the panel and
a linear material behavior. As the out-of-plane deflections
near the panel ends are small, the geometric nonlinearity is
neglected in Eq. (1). The mean values of the experimental
eccentricities are used, which were determined at a short-
term load level equal to 30% of the estimated failure load.
Panels LT2 and LT3 were loaded at a concrete age of
146 and 99 days, respectively, whereas the other panels
were all loaded at the same concrete age of 22 days. The
testing age (22 days) was selected from the considerations
of having sufficient strength while allowing significant creep
so that some critical aspects of the structural behavior can be
observed or revealed within a reasonable time period. The
effect of loading age is investigated by comparing the test
results of Panel LT3 to those of LT5. The influence of the
in-plane load level is studied by comparing the results of
Panels LT1, LT5, and LT6. The influence of the slenderness
ratio is examined by comparing the results of Panel LT1 and
LT9. Panel LT4 was initially loaded to 93% of its predicted
short-term failure load, followed by an unloading stage until
the load was dropped to 74% of the failure load, which was
then kept constant with time.
Fig. 3—Details of test setup.
Preparation of test specimens
The formwork of the panel specimens was built using
structural-grade plywood and laid horizontally on the ground
in the laboratory. The SL52 mesh was cut to the required
dimension and placed at the top and bottom layers that were
held in place by steel bar chairs. The panels were cast using
a commercial high-strength concrete. They were made from
five concrete batches with the same concrete mixture design
(refer to Tables 1 and 2). After casting, the panels were
covered with wet hessian and with plastic sheets over the
hessian. They were kept moist in the formwork for 14 days
before being stripped; they then remained in the laboratory
at ambient conditions until the day of loading.
Test setup and instrumentation
The panels were tested in a testing frame with a hydraulic
jack mounted at one end and two load cells mounted at the
other end. The load was applied and monitored to remain
constant with time through the hydraulic jack. The test setup
is shown in Fig. 2 and 3. Two groups of disc springs with
high load-carrying capacity were inserted between the jack
and the end support plate to minimize the reduction of the

44 ACI Structural Journal/January 2018

Fig. 4—Location of strain gauges. (Note: All dimensions in mm; 1 mm = 0.0394 in.)
axial sustained load with time due to creep and relaxation.
The loading and supporting mechanisms at each panel end
consisted of a rotatable knife-hinge to provide a simply
supported boundary condition (refer to Fig. 3). The load
eccentricity was set up in a way that the panel would deform
upwards under loading. The out-of-plane displacements
were measured by laser displacement sensors at three points
that were symmetrically positioned along the top face of the
panel. Two sensors were located at 100 mm (3.94 in.) away
from each edge, and one was placed at the center length.
Eighteen strain gauges were placed on the surfaces of steel
and concrete at critical locations as shown in Fig. 4. The
displacements of the hinge plates at the two ends were moni-
tored using a linear strain conversion transducer (LSCT), as
shown in Fig. 3.
Test procedure
Each panel was loaded up to the desired sustained load
level and the load was then held constant during the test.
The duration of applying the loads was less than 30 minutes.
The sustained load dropped as a result of concrete relax-
ation. The dropping was remarkable, especially at the first
few hours after the sustained load was applied. Therefore,
the load was topped up from time to time, as needed, so
that it stayed at a nearly constant level throughout the test
to simulate creep behavior. The procedure of adjusting the
load was made using a hydraulic pump that was connected
to the loading jack. If the tested panel was found to show
very little increase in the deflection over a certain period
after sustained loading, it was considered to have long-term
stability and the test would be terminated. The panel was
then loaded to failure without releasing the existing sustained
load following the same procedure used for adjusting the
load level. All tests were conducted under typical ambient
indoor conditions within the laboratory. Unfortunately, no
measurements were made to monitor the temperature and
ACI Structural Journal/January 2018
humidity, but the temperature would typically be at room
temperature, varying between 20 and 25°C (68 and 77°F),
and the humidity would be varying between 40 and 60%.
Material properties
Concrete cylinders 100 mm (3.94 in.) in diameter and
200 mm (7.87 in.) high, and concrete prisms of 100 x 100
x 500 mm (3.94 x 3.94 x 19.7 in.) were cast and cured from
each batch during the preparation of the panels. The control
cylinders and prisms were kept in the same environment
as the panels and were similarly cured. The concrete cylin-
ders were tested in compression to measure the compres-
sive strength, the elastic modulus, and the full stress-strain
behavior. The prisms were tested under four-point bending
to determine the flexural tensile strength. The concrete
mechanical properties were measured at the commencement
and completion of each panel test. The material test results
are reported in Table 2.
Two concrete prisms of 75 x 75 x 280 mm (2.95 x 2.95 x
11 in.) with gauge studs mounted on both ends were cast and
cured with each batch in order to measure the shrinkage strain
in accordance with Australian Standard AS1012.13.18 The
prisms were stripped off at the same day as the panels and then
kept in the same environment as the panels. The measurement
of the shrinkage strain started immediately after the panels
were demolded at 14 days by using a vertical comparator, and
stopped after failure of the panel. The measured shrinkage
data for each concrete batch is shown in Fig. 5.
Three cylinders from each batch were placed in a creep
test rig under a sustained stress of 30 MPa (4.35 ksi) when
each group of panels that are made from the same batch were
tested. The stress was applied hydraulically on the same day
of undertaking the long-term test of the corresponding panel,
and was adjusted to maintain constant over time. The total
strains—which included the elastic strain, the shrinkage
and thermal strain, and the creep strain—were measured
45
from the three cylinders using Demec strain gauges. The
shrinkage and thermal strains were measured from two
unloaded companion cylinders. The creep strains were
then determined by subtracting the sum of the measured
shrinkage and thermal strain and instantaneous elastic strain
from the total strains. The creep coefficient, defined as the
ratio of measured average creep strain to the measured
average elastic strain, for each concrete batch is shown in
Fig. 6. Batch 2 exhibited noticeably smaller creep over time
than the other batches because it was tested at older ages
upon testing Panels LT2 and LT3. The curves of the concrete
made from Batches 1 and 5 have a similar trend, as so do the
curves of concrete Batches 3 and 4. However, the creep coef-
ficients of Batches 3 and 4 appear to be smaller compared
to those of Batches 1 and 5, although all concrete cylinders
were loaded at the same age of 22 days. The reason for that
could be mainly attributed to the fact that the batches were
tested at different seasons. Batches 1 and 5 were tested in
winter, whereas Batches 3 and 4 were tested around summer
with much higher relative humidity, which resulted in less
creep. The differences in the temperature between the two
seasons in indoor conditions are too small to significantly
influence the creep response.
Samples of the reinforcing steel were tested to determine
the stress-strain properties. The elastic modulus and yield
strain were measured as 206 GPa (29,878 ksi) and 2 × 10–3.
The measured compressive strains of concrete corresponding
to the compressive strength in the stress-strain curves ranged
from 3.5 × 10–3 to 3.8 × 10–3.

EXPERIMENTAL RESULTS AND DISCUSSIONS

Fig. 5—Shrinkage strains measured in test. Both the experimental results and the predictions by the
theoretical model of Huang and Hamed16 are summarized
in Table 3. In the model,16 the concrete is assumed to be
linear viscoelastic in compression and nonlinear in tension,
where cracking and tension stiffening are included following
Fields and Bischoff.19 The measured material properties
are incorporated into the long-term theoretical model. The
aging effect of concrete, which includes time variation of its
elastic modulus and strength, is modeled by linearly interpo-
lating the experimental data appearing in Table 2 with time.
Because Panels LT1 and LT4 to LT9 were cast using the same
concrete mixture design, cured under similar conditions and
loaded at the same age, their material properties, including
strength, elastic modulus, and shrinkage are considered
the same and the average values of the experimental data
reported in Fig. 5 and Table 2 are used in the model. The
creep properties of concrete, however, were not averaged in
Fig. 6—Creep coefficients measured in test.

Table 3—Comparison of test results to predicted results by theoretical model

Test results Model predictions
Concrete age upon Test duration, Residual failure load Critical time for creep Short-term failure
Panel No. loading (t0), days days Creep buckling after creep PR (kN) buckling*, days load Pu†, kN PR/Pu
LT1 22 13 √ 0 18 638 0
LT2 146 57 — 373 N/A 436 86%
LT3 99 42 — 794 N/A 820 97%
LT4 22 104 — 466 N/A 499 93%
LT5 22 69 √ 0 157 764 0
LT6 22 112 — 665 N/A 747 89%
LT7 22 112 — 1377 N/A 1526 90%
LT8 22 82 √ 0 56 1301 0
LT9 22 8 √ 0 11 1075 0
*
Prediction is based on long-term analysis presented in Huang and Hamed.16
†
Prediction is based on short-term analysis presented in Huang et al.7
Note: 1 kN = 0.225 kip.

46 ACI Structural Journal/January 2018

Table 4–Moduli of springs for Maxwell model of each panel
Panel No. E1, MPa E2, MPa E3, MPa E4, MPa E5, MPa E6, MPa
LT1 9313 4573 3982 3097 2889 15,200
LT2 3494 3661 2560 4849 5493 20,058
LT3 3494 3661 2560 4849 5493 20,058
LT4 8247 4428 3890 3043 2845 16,662
LT5 8247 4428 3890 3043 2845 16,662
LT6 4588 1593 4463 5157 1820 21,882
LT7 4588 1593 4463 5157 1820 21,882
LT8 8224 6047 4398 3717 3628 12,866
LT9 8224 6047 4398 3717 3628 12,866

Note: 1 MPa = 0.145 ksi.

the model due to the notable differences obtained from the

control tests of different batches.
A generalized Maxwell model20 was used in Hamed and
Huang16 to model the concrete following the expansion
of the relaxation function into a Dirichlet series. It was
assumed that the relaxation function is approximately equal
to the inverse of the compliance function of concrete—that
is, Ec/(1 + φc), where Ec is the elastic modulus of concrete
and φc is the creep coefficient. Thus, the experimental relax-
ation function can be obtained by substituting the measured
elastic modulus and creep coefficient into the formula at
different times. By using the least-squares method to fit
the experimental relaxation function, the spring moduli in
the Maxwell chain are determined. Five Maxwell units are
used to model the viscoelastic behavior of concrete with τμ
= 5μ–1 (days), where τμ is the relaxation time of the μ-th unit.
The spring constants that are obtained by the least-squares
methods, which were used for the analysis of the panels, are
given in Table 4.
Cracking occurred in all panels during initial loading
or within a half-day after the imposed instantaneous load
reached the desired sustained load level. The cracks first
started in the middle span and then propagated toward the
ends. The cracked regions were generally concentrated
within the middle one-third span of the panels at the time of
failure. Panels LT1, LT5, LT8, and LT9 failed by creep buck-
ling under the sustained loads at different times, as shown
in Table 3. The other panels exhibited stable responses
with time and were then further loaded to failure. The residual
failure loads of Panels LT2 to LT4 and Panels LT6 to LT7 at the
end of the sustained loading tests are also given in Table 3. As
observed from the strain gauges readings, all panel speci-
mens collapsed in a buckling failure mode before strain
softening of the concrete. A typical buckling failure mode is
shown in Fig. 7.
Deflection versus time
The variation of the measured and predicted central
deflections with time for all Panels except LT4 are presented
in Fig. 8. Panels LT1, LT5, and LT6 were loaded at the same
age (22 days) but under different load levels, which were
93%, 72%, and 66% of the corresponding short-term failure
loads, respectively. It can be seen in Fig. 8 that the three
Fig. 7—Representative buckling failure mode.
panels exhibited distinct time-dependent behaviors under
the influence of creep and shrinkage. Panel LT6 exhibited
a gradually decreasing rate of increase of deformations over
time with a stable time-dependent response. Panels LT1 and
LT5, on the other hand, experienced three stages of increase
in the deflections and failed by creep buckling at 13 days and
69 days since initial loading, respectively. At the initial stage
after the sustained load was applied to Panels LT1 and LT5,
the deflections increased at a decreasing rate. Following this,
a transitional stage occurred where the rate of increase in
the deflections remained nearly constant. In the third stage,
the deflections of the panels increased at an accelerating rate
(so-called tertiary creep response), which ultimately led to
buckling failure of the panels.
It is clear that a higher sustained load level leads to a more
rapid increase of deflection as well as a higher tendency
of creep buckling for slender HSC panels. It can be seen
from Fig. 8 that the theoretical results for all three panels
correlate with the test results reasonably well. The esti-
mated critical time for creep buckling failure is 18 days and
157 days for Panels LT1 and LT5, respectively. The discrep-
ancy between the experimental and predicted critical times
for Panel LT5 is believed to be due to the large sensitivity of
the panel response to uncertainties in the actual load level.
To strengthen this argument, Fig. 9 shows that increasing the
load level by 5% only, which is within the typical tolerance
of an experimental test, can dramatically change the critical
time that leads to creep buckling, and yields a better agree-
ment between the theoretical predictions and the test results.

ACI Structural Journal/January 2018 47

Fig. 8—Time variation of out-of-plane deflections of center points of Panels LT1 to LT3 and LT5 to LT9. (Note: 1 mm = 0.0394 in.)
Fig. 9—Illustration of sensitivity of panel behavior to load
level for Panel LT5. (Note: 1 mm = 0.0394 in.)
Panel LT2 was tested at the age of 146 days with a designed
eccentricity of h/3 and a sustained load level of 73% of its
short-term load capacity. Panel LT3 was tested at the age
of 99 days under a sustained load of 75% of its short-term
load capacity with an eccentricity of h/6. It can be seen in
Fig. 8 that both panels displayed increased deflections over
time at decreasing rates. The deflection-versus-time curves
asymptotically approached constant values, showing a stable
state. The ratios of the deflections at the end of the test over
the instantaneous deflections for Panels LT2 and LT3 were
1.66 and 1.35, respectively, whereas the creep coefficients
of the corresponding concrete cylinders were 0.41 and 0.39,
respectively. Thus, the deflections in the panels underwent
a more rapid increase than the creep coefficients, owing to
the contributions of the geometric nonlinearity (P-Δ effect)
and the cracking of concrete. It can be seen in Fig. 8 that the
predictions by the theoretical model agree well with the test
results for Panels LT2 and LT3.
The effect of aging can be examined by comparing the
results of Panel LT3 and Panel LT5, which had similar load
levels and eccentricities but were loaded at 99 days and
22 days, respectively. It can be observed in Fig. 8 that due
to the strong dependency of the creep and shrinkage on the
48
age of concrete, the early-loaded Panel LT5 showed a larger
increase with time in the deflection and exhibited a sudden
buckling failure after 69 days, whereas Panel LT3 exhibited
a stable response. Hence, to reduce the risk of creep buckling
failures in slender HSC panels, it is desirable to avoid an
early-age loading.
Panels LT7 and LT8 were both tested at 22 days with
different sustained load levels that equal 53% and 71% of
their corresponding short-term load capacities, respectively.
The designed eccentricities were the same and equal to h/12.
The deflection of Panel LT7 reached a limit, which revealed
a stable response, whereas Panel LT8 failed at 82 days by
creep buckling. The model results correlated well with the
test results. One other reason that could have contributed to
the differences in the behavior is that Panel LT8 was made
from concrete Batch 5, which exhibits more creep than
Batch 4, from which Panel LT7 was made (refer to Fig. 6).
Panel LT9 had a thickness of 130 mm (5.12 in.) (slender-
ness ratio equal to 72) in contrast to the other panels whose
thickness was 100 mm (3.94 in.) (slenderness ratio equal to
94). It was loaded to 95% of the short-term failure load with
an eccentricity of h/6 at the age of 22 days. The panel expe-
rienced three stages of increase in the deflection and failed
quickly by creep buckling at 8 days after initial loading. It
can be seen in Fig. 8 that the theoretical model describes the
time-dependent behavior very well, with a predicted critical
time of 11 days.
The effect of the slenderness ratio can be clarified by
comparing the results of Panels LT1 and LT9, as they were
both loaded under similar load levels and load eccentricities.
In addition, they were made from concrete Batches (1 and
5, respectively) that exhibited more or less the same creep
response (refer to Fig. 6). It is seen in Fig. 8 that Panel LT9,
which is less slender than Panel LT1, buckled more quickly.
This observation is contrary to the expectation where the
more slender panel should buckle more rapidly. This is
explained by the difference in the sustained load ratio: LT9
had slightly 2% larger load level than LT1. Although it is
a very small difference, as mentioned previously, the creep
ACI Structural Journal/January 2018
buckling response of HSC panels is very sensitive to the
magnitude of the applied sustained loading, especially at
high load levels. Therefore, an additional case is analyzed
where the load level in Panel LT9 is reduced by 2% to 93%
of its short-term load-carrying capacity. The response is
given in Fig. 10. The behavior of LT9 under a load level
of 93% significantly differs from that under a load level of
95%. The increase in the out-of-plane deflection is much
slower when the load level is 93% and the panel buckles at
292 days. Thus, if the two panels (LT1 and LT9) were loaded
by exactly the same load levels, then the more slender panel
(LT1) would fail first.
The theoretical model, which is validated herein, predicts
that the minimum sustained load that will lead to creep
Fig. 10—Illustration of sensitivity of panel behavior to load
level for Panel LT9. (Note: 1 mm = 0.0394 in.)
Fig. 11—Time variation of strains in concrete and steel reinforcement at bottom face of midlength section of Panels LT1, LT5,
LT8, and LT9.
ACI Structural Journal/January 2018
buckling failure for Panel LT1 is approximately 60% of the
instantaneous failure load after 3087 days since loading at the
age of 22 days. Thus, the buckling capacity of HSC panels
can be reduced by 40% because of the combined effects of
creep, shrinkage, and cracking. The reductions may vary
for different boundary conditions, slenderness ratios, and
reinforcement ratios, which should be carefully considered
in the design and analysis of slender HSC panels.
Strains versus time
The measured and theoretical strains of concrete and steel
reinforcement at the bottom face of the midlength section for
the panels that failed by creep buckling (Panels LT1, LT5,
LT8 and LT9) are plotted against time in Fig. 11. Similar
to the deflection, the compressive strains of the concrete
and steel also increase with time as a result of the combined
effects of creep, shrinkage, and cracking of concrete, as well
as the geometric nonlinearity. Moreover, the concrete strains
are far below the measured peak strain of 3500 µε that corre-
sponds to the compressive strength throughout the entire
creeping period. This means no strain softening of concrete
occurred until failure. The strains in the steel reinforcement
are also smaller than the yielding strain of 2000 µε during
most of the creeping period. These measurements, along
with the development of tertiary creep deformations (Fig. 8),
emphasize that buckling is the failure mode and not typical
flexural failure. In some measurements, it can be seen that
the strains suddenly increased to exceed the limit strains of
the steel only at the moment of failure (LT9, for example).
49
Fig. 12—In-plane load versus out-of-plane center deflection of Panels LT2, LT3, LT4, LT6, and LT7. (Note: 1 kN = 0.225 kip;
1 mm = 0.0394 in.)
Figure 11 shows that the predicted strains of concrete and
steel have good correlation with the test results for Panels
LT1 and LT5. For Panels LT8 and LT9, a reasonably good
agreement is achieved for concrete, but there exist some
discrepancies for the strains in the steel. The reasons for the
discrepancies are not quite clear, but they might be related to
local dislocation of the steel mesh or unusual strain drifting
due to poor installation of the strain gauges on the steel
reinforcement.
Load versus deflection
As seen in Fig. 8, Panels LT2, LT3, LT4, LT6, and LT7
were stable during the sustained loading tests. There-
fore, these panels were loaded to failure at the end of the
sustained loading tests by increasing the load without
releasing the sustained load. The experimental loads versus
the out-of-plane deflections of center points of the Panels are
plotted in Fig. 12. The figure also shows the theoretical load-
versus-deflection curves produced by the short-term model.7
The theoretical results used the material properties of tested
specimens at the age of initial loading and excluded the
effects of creep and shrinkage. The deflections of all panels
increased almost linearly with the loads at the initial stage
of loading. At the sustained load levels, the loads were kept
nearly constant over time and the deflections increased due
to the effects of creep and shrinkage. Figure 12 shows that
the magnitudes of the deflections after increasing the load
to failure were larger than those predicted under short-
term loading. The ratios of the residual failure loads over
the predicted instantaneous failure loads are given in Table
3. It can be seen that the residual capacities of Panels LT2,
50
LT6, and LT7 are approximately 10 to 14% smaller than the
predicted short-term capacities, but only a 3% reduction in
the load-carrying capacity is obtained for Panel LT3. Based
on these results, it is concluded that the long-term effects of
creep and shrinkage could lead to reductions in the residual
strength of HSC panels to varying degrees, which need to be
considered in their design.
As mentioned previously, Panel LT4 was first loaded to
93% of its predicted short-term failure load. The load was
then dropped to 74% of its failure load, and then it was kept
constant with time. The panel exhibited a stable behavior
during the duration of the sustained loading test, and exhib-
ited a small reduction of approximately 7% in the strength
when loaded to failure after creep.
CONCLUSIONS
The outcomes of an experimental study that consisted
of testing nine slender one-way HSC wall panels under
sustained loading is reported in this paper. The panels were
subjected to in-plane eccentric compression loads and were
simply supported along their short edges. Four panels failed
by creep buckling under sustained loads at certain times,
while the other five panels exhibited stable response over
time and were then loaded to failure to determine their
residual capacities. Three parameters that can substantially
affect the response of the panels are considered: 1) the
loading age; 2) the magnitude of the in-plane load; and 3)
the slenderness ratio.
While the loading age is known to influence the creep of
concrete, this parameter seems to have a crucial influence in
geometrically nonlinear buckling problems, as it may domi-
ACI Structural Journal/January 2018
nate the stability of HSC panels. The panel that was tested at
99 days under the same loading conditions as its counterpart
panel, which was tested at 22 days, showed a stable, time-
dependent response. On the other hand, creep buckling
failure occurred to the panel that was loaded at 22 days.
Therefore, it is desirable to avoid loading at early ages to
reduce the risk of creep buckling failures of slender HSC
panels. Furthermore, the results reveal that the long-term
behavior of HSC panels is very sensitive to the in-plane load
level. It is found that a small difference in the load level such
as 2% can give rise to notable variation in the critical time
that leads to creep buckling.
The time-dependent effects of creep and shrinkage, coupled
with the geometric nonlinearity (P-Δ effect), have caused
reductions in the load-carrying capacity of HSC panels to
varying degrees. Some panels failed with time by creep buck-
ling under sustained loads that were approximately 71% of
their short-term load-carrying capacity. Panels that exhibited
a stable response with time were loaded to failure after creep
and showed a reduction of up to 14% in their residual capacity.
Predictions based on the theoretical model of Huang and
Hamed16 that was validated in this study show that sustained
loads as low as 60% of the short-term capacity can lead
to creep buckling failures. Such reductions in the load-car-
rying capacity caused by the long-term effects of creep and
shrinkage should be carefully considered and treated in the
design of slender HSC wall panels.
AUTHOR BIOS
Yue Huang is a Research Associate at the Center for Infrastructure Engi-
neering and Safety, School of Civil and Environmental Engineering, the
University of New South Wales, Kensington, NSW, Australia. He received
his BEng in structural engineering from City University of Hong Kong,
Hong Kong, and his PhD from the University of New South Wales. His
research interests include time-dependent behavior of reinforced concrete
structures.
ACI member Ehab Hamed is a Senior Lecturer at the Center for Infra-
structure Engineering and Safety, School of Civil and Environmental Engi-
neering, the University of New South Wales. He received his PhD in struc-
tural engineering from the Technion – Israel Institute of Technology, Haifa,
Israel. His research interests include the time-dependent behavior of materials
and structures as well as the repair of steel and concrete structures using
composite materials.
Zhen-Tian Chang is a Senior Research Fellow at the Center for Infra-
structure Engineering and Safety, School of Civil and Environmental Engi-
neering, the University of New South Wales. He received his BEng and
MEng in structural engineering from Hunan University, Changsha, China,
and his PhD from the University of New South Wales. His research inter-
ests include steel and reinforced concrete structures and durability of infra-
structures in aggressive environments.
ACI member Stephen J. Foster is Professor and Head of the School of
Civil and Environmental Engineering at the University of New South Wales,
where he received his PhD in 1993. His research interests include the struc-
tural use of high-strength and ultra-high-strength concretes, fiber-reinforced
concrete, nonlinear finite element analysis of concrete structures, and reha-
bilitation of concrete structures.
ACKNOWLEDGMENTS
The work reported in this paper was supported by the Australian Research
Council (ARC) through a Discovery Project (DP120102762).
ACI Structural Journal/January 2018
REFERENCES
1. Bažant, Z. P., “Creep Stability and Buckling Strength of Concrete
Columns,” Magazine of Concrete Research, V. 20, No. 63, 1968, pp. 85-94.
doi: 10.1680/macr.1968.20.63.85
2. Bockhold, J., and Petryna, Y. S., “Creep Influence on Buckling Resis-
tance of Reinforced Concrete Shells,” Computers & Structures, V. 86,
No.  7-8, 2008, pp. 702-713. doi: 10.1016/j.compstruc.2007.07.004
3. Hamed, E.; Bradford, M. A.; Gilbert, R. I.; and Chang, Z. T., “Analytical
Model and Experimental Study of Failure Behavior of Thin-Walled Shallow
Concrete Domes,” Journal of Structural Engineering, ASCE, V. 137, No. 1,
2011, pp. 88-99. doi: 10.1061/(ASCE)ST.1943-541X.0000274
4. Oberlender, G. D., and Everard, N. J., “Investigation of Reinforced
Concrete Walls,” ACI Journal Proceedings, V. 74, No. 6, June 1977,
pp. 256-263.
5. Saheb, S. M., and Desayi, P., “Ultimate Strength of RC Wall
Panels in One-Way In-Plane Action,” Journal of Structural Engi-
neering, ASCE, V. 115, No. 10, 1989, pp. 2617-2630. doi: 10.1061/
(ASCE)0733-9445(1989)115:10(2617)
6. Fragomeni, S., and Mendis, P. A., “Instability Analysis of Normal-
and High-Strength Reinforced Concrete Walls,” Journal of Structural
Engineering, ASCE, V. 123, No. 5, 1997, pp. 680-684. doi: 10.1061/
(ASCE)0733-9445(1997)123:5(680)
7. Huang, Y.; Hamed, E.; Chang, Z. T.; and Foster, S. J., “Theoretical and
Experimental Investigation of Failure Behavior of One-Way High-Strength
Concrete Wall Panels,” Journal of Structural Engineering, ASCE, V. 141,
No. 5, 2015, 4014143. doi: 10.1061/(ASCE)ST.1943-541X.0001072
8. Green, R., and Breen, J. E., “Eccentrically Loaded Concrete Columns
under Sustained Load,” ACI Journal Proceedings, V. 66, No. 11, Nov. 1969,
pp. 866-874.
9. Goyal, B. B., and Jackson, N., “Slender Concrete Columns under
Sustained Load,” Journal of the Structural Division, ASCE, V. 97, 1971,
pp. 2729-2750.
10. Behan, J. E., and O’Connor, C., “Creep Buckling of Reinforced
Concrete Columns,” Journal of the Structural Division, ASCE, V. 108,
1982, pp. 2799-2818.
11. Mickleborough, N. C., and Gilbert, R. I., “Creep Buckling of Uniax-
ially Loaded Reinforced Concrete Columns,” Computer Analysis of Effects
of Creep, Shrinkage, and Temperature Changes on Concrete Structures,
SP-129, C. C. Fu and M. D. Daye, eds., American Concrete Institute, Farm-
ington Hills, MI, 1991, pp. 39-55.
12. Claeson, C., and Gylltoft, K., “Slender Concrete Columns Subjected
to Sustained and Short-Term Eccentric Loading,” ACI Structural Journal,
V. 97, No. 1, Jan.-Feb. 2000, pp. 45-53.
13. Bradford, M. A., “Shrinkage and Creep Response of Slender Rein-
forced Concrete Columns under Moment Gradient: Theory and Test
Results,” Magazine of Concrete Research, V. 57, No. 4, 2005, pp. 235-246.
doi: 10.1680/macr.2005.57.4.235
14. Tatsa, E. Z., “Load Carrying of Eccentrically Loaded Reinforced
Concrete Panels under Sustained Load,” ACI Structural Journal, V. 86,
No. 2, Mar.-Apr. 1989, pp. 150-155.
15. Lee, Y.; Lee, B. Y.; Kwon, S. H.; Kim, Y. Y.; and Kim, J. K.,
“Long-Term Behavior of a Reinforced Concrete Wall under Compressive
Stress Applied to Part of the Wall’s Entire Width,” Magazine of Concrete
Research, V. 60, No. 10, 2008, pp. 759-768. doi: 10.1680/macr.2008.00090
16. Huang, Y., and Hamed, E., “Buckling of One-Way High-Strength
Concrete Panels: Creep and Shrinkage Effects,” Journal of Engineering
Mechanics, ASCE, V. 139, No. 12, 2013, pp. 1856-1867. doi: 10.1061/
(ASCE)EM.1943-7889.0000629
17. Fragomeni, S., and Mendis, P. A., “Applicability of Current
ACI 318 Wall Design Formula for High Strength Concrete Walls,”
Advances in Structural Engineering, V. 2, No. 2, 1999, pp. 103-108. doi:
10.1177/136943329900200203
18. Standards Association of Australia, “Methods of Testing Concrete—
Method 13: Determination of the Drying Shrinkage of Concrete for
Samples Prepared in the Field or in the Laboratory,” AS 1012.13, Sydney,
Australia, 1992.
19. Fields, K., and Bischoff, P. H., “Tension Stiffening and Cracking
of High-Strength Reinforced Concrete Tension Members,” ACI Structural
Journal, V. 101, No. 4, July-Aug. 2004, pp. 447-456.
20. Bažant, Z. P., and Wu, S. T., “Rate-Type Creep Law of Aging
Concrete Based on Maxwell Chain,” Materials and Structures, V. 7,
No. 1, 1974, pp. 45-60.
51
JOIN AN
ACIChapter!
The American Concrete Institute has Chapters and Student Chapters located
throughout the world. Participation in a local chapter can be extremely
rewarding in terms of gaining greater technical knowledge and networking
with leaders in the concrete community.
Because chapters are distinct and independent legal entities, membership includes both ACI
members and non-ACI members and is made up of a diverse blend of architects, engineers,
consultants, contractors, educators, material suppliers, equipment suppliers, owners, and
students—basically anyone interested in concrete. Many active ACI members initially became
involved in ACI through their local chapter. In addition to technical programs and publications,
many chapters sponsor ACI Certification programs, ACI educational seminars, project award
recognition programs, and social events with the goal of advancing concrete knowledge.

Check out the Chapters Special Section from the October 2015 Concrete International:
www.concrete.org/publications/concreteinternational/cibackissues.aspx?m=456

Student Chapters
Join or form an ACI Student Chapter to maximize your influence, knowledge
sharing, and camaraderie! ACI has over 109 student chapters located
throughout the world, each providing opportunities for students to:
• Connect with their peers and participate in concrete-related activities such as: student
competitions, ACI Conventions, ACI Certification Programs, ACI Educational Seminars, local
chapter meetings, social events, and community service projects;
• Network with members of local chapters, many of whom have been in the industry for
decades and can help to develop professional relationships and offer career advice;
• Win recognition for their universities through the University Award; and
• Learn about the many scholarships and fellowships offered by the ACI Foundation and by
ACI’s local chapters.

ACI Student Chapters worldwide

   
www.concrete.org/chapters
 ACI STRUCTURAL JOURNAL
Title No. 115-S04
Nonlinear Backbone Modeling of Concrete Columns
Retrofitted with Fiber-Reinforced Polymer or Steel Jackets
by José C. Alvarez, Sergio F. Breña, and Sanjay R. Arwade
The use of nonlinear analysis procedures in seismic retrofitting of
existing concrete structures has become commonplace in practice.
Backbone curves are often used to capture the nonlinear response
of components in a simplified but sufficiently accurate way. Proce-
dures to construct curves for existing components of frames
(beams, joints, and columns) have received considerable attention
and have been modified over the years. In contrast, recommenda-
tions to construct these backbone curves for retrofitted components
are largely lacking. This paper presents recommendations that
can be used to construct backbone curves of circular and rect-
angular retrofitted columns using jacketing materials within the
context of ASCE/SEI 41-13 and ACI 369R-11. The recommenda-
tions are developed using a sectional model for force parameters
and a statistical study of load-deformation results from a database
of retrofitted columns for drift parameters when sufficient experi-
mental data are available. Key points in the backbone response of
jacketed columns are summarized in tabular form consistent with
ASCE/SEI 41-13 to facilitate possible adoption in future updates of
the standard. A probabilistic model is introduced to allow selection
of drift values that correspond to selected exceedance probabilities.
Keywords: backbone curves; column retrofitting; jacketed columns;
nonlinear modeling parameters.
INTRODUCTION
The potential for failure of nonductile columns with
details that do not conform to modern building codes has
been recognized for many years. With local retrofitting such
as external jacketing, the behavior of nonductile columns
can be greatly improved, providing benefits to the global
performance of building frames and reducing the potential
for collapse. Jacketing materials are selected such that they
can be easily applied to frame elements without disruption
of building operations. The most common types of external
jackets are concrete, steel, and fiber-reinforced polymer
(FRP) materials.
In any retrofitting project, it is paramount to be able to
accurately assess the performance of the retrofitted structure
to ensure adequacy in its performance. ASCE/SEI 41-13,
“Standard on Seismic Evaluation and Retrofit of Existing
Buildings,” provides detailed guidance for the evaluation of
structures in their existing condition. In contrast, very little
information is provided for engineers to verify the response
of the structure in its retrofitted condition. Guidance on
procedures to estimate strength and deformation capacity of
retrofitted components is needed to adequately evaluate if
the structure satisfies the intended performance objectives
for the expected demands.
Jacketing deficient columns can change the sequence
of inelastic hinge formation of an existing frame such that
ACI Structural Journal/January 2018
TECHNICAL PAPER
hinges form at the ends of beams first instead of at the ends
of columns. This causes the collapse mechanism to change
from a failure involving limited lateral deformation capacity
to one involving a more ductile mechanism. As a local
retrofitting technique, jacketing can be selectively used in
cases where other components of frames that make up the
lateral load system have sufficient strength and deformation
capacity. It is critical, however, to be able to calculate the
expected sequence of hinge formation by developing reli-
able nonlinear modeling techniques of retrofitted elements
of potentially vulnerable reinforced frames.
This paper focuses on developing a methodology to model
reinforced concrete columns that are retrofitted using steel
or FRP jackets applied locally in the region where inelastic
hinges are anticipated to form. The methodology focuses
on a recommended procedure to construct the nonlinear
force-deformation backbone response of jacketed columns.
The procedures used within this methodology result in
backbone curves that compare favorably with the backbone
response measured in available tests. Recommendations to
construct backbone curves are provided in a similar format
to the current approach for existing elements contained in
ASCE/SEI 41-13 and ACI 369R-11 by defining key points
in the nonlinear backbone response of reinforced concrete
frame elements. The recommendations were developed in
a way that they may provide the basis for future updates of
these two documents.
The backbone curves are defined in this paper by six
parameters: the force at yield, the peak force, the residual
capacity, the drift at yield, the drift at 80% of peak force,
and the drift at which the force equals the residual capacity.
A sectional model, validated against a database of experi-
mental results, is used to calculate the force at yield and the
peak force. Validated sectional models that can predict drift
of jacketed columns are not currently available and, there-
fore, a statistical model for the drift at 80% of peak force
is developed and calibrated to a database of experimental
results assembled from the open literature for cases when
the axial force ratio is low. Unfortunately, insufficient data
are available in the literature to develop a statistical model
for the drift at 80% of peak load when the axial force ratio
is high or for drift corresponding to the residual capacity of
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2016-121.R3, doi: 10.14359/51700779, received March 10, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
53

Fig. 1—Simplified backbone curve.
jacketed columns. Those parameters are therefore conser-
vatively estimated as for unjacketed columns containing
details similar to those in modern codes.
RESEARCH SIGNIFICANCE
ASCE/SEI 41-13 and ACI 369R-11 contain requirements
to model existing reinforced concrete building frames for
retrofitting purposes. These documents focus on modeling
of existing components but do not provide details on ways
to treat retrofitted components. This paper is intended to fill
this gap in information by developing a procedure that can
be used to construct nonlinear backbone curves for jacketed
columns, a technique that is frequently considered in retro-
fitting projects.
GENERAL BACKBONE FORCE-DEFORMATION
RELATIONSHIP
Both ASCE/SEI 41-13 and ACI 369R-11 use a simpli-
fied backbone model to describe the nonlinear lateral load-
drift response of existing reinforced concrete elements
(Fig. 1). In this model, the nonlinear response is primarily
characterized through the definition of two parameters (a
and b) that are used to represent the plastic drift capacity
of existing columns for given transverse reinforcement
contents and axial load level. These documents tabulate
values of nonlinear drift parameters for columns and other
reinforced concrete components based on experience and
the observed response during laboratory testing of compo-
nents. One of the drawbacks in the currently used simplified
backbone model in comparison with response of compo-
nents measured during tests is that the strength drop after
reaching peak strength is not as sudden, as indicated by the
solid line in the figure, but it is rather gradual in most cases
(dashed line). The simplified backbone response curves for
jacketed columns proposed later in this paper will follow the
general shape as indicated by the dashed line to better reflect
observed response.
Because of the widespread use of simplified back-
bone curves in nonlinear analysis of existing structures,
the approach followed in this paper is to characterize the
nonlinear lateral deformation response of jacketed columns
through parameters a and b. Parameter a can be defined as
the magnitude of plastic drift that corresponds to a 20%
54
reduction in lateral strength from peak. Parameter b is taken
as the plastic drift that corresponds to the point where the
lateral strength has degraded to the residual strength, gener-
ally assumed equal to 0.2Vpeak. These two definitions are
followed to construct simplified backbone curves using a
database of existing tests of jacketed columns as discussed
in the following section.
DATABASE OF JACKETED COLUMNS
An experimental database of available laboratory exper-
iments of circular and rectangular jacketed columns tested
under cyclic loading was assembled for two main purposes.
First, lateral load data were used to evaluate the accuracy
of proposed sectional models described later in this paper
to compute yield and peak strength of jacketed columns.
Second, the data were also used to study the statistical
distribution of drifts at key points of the force-deformation
response of jacketed columns. The information obtained
from the database, therefore, was used to evaluate the accu-
racy of the methods proposed to construct the backbone
response of jacketed columns.
The database consists of a total of 116 columns: 84 and
32 columns jacketed using fiber-reinforced polymer (FRP)
or steel materials, respectively. Several challenges were
encountered when assembling the database, given the vast
variety of details in the columns tested. The specimens were
designed and constructed to capture common deficient prop-
erties typically found in columns of older existing build-
ings such as low shear strength, short lap splices within
the plastic hinge zone, and/or low volume of transverse
reinforcement. Column jacket retrofitting using different
materials and configurations was used to overcome these
existing deficiencies that result in poor column performance.
The laboratory specimens include columns with rectan-
gular or circular cross sections, various levels of applied
axial force (typically axial load levels below the balanced
point in the column interaction diagram), and two different
test setup configurations that generate single- or double-
curvature bending of the columns during lateral loading.
Relevant properties of jacketed columns in the database are
given in Table 1 for circular columns and Table 2 for rect-
angular columns, respectively. Many (not all) of the tests on
retrofitted columns were compared with tests of an unretro-
fitted reference column, but only the retrofitted specimens
are provided in the tables. It can be seen that the highest
number of specimens correspond to FRP-jacketed columns
with rectangular cross sections and the fewest number of
specimens obtained correspond to steel-jacketed circular
columns. The range of values for key column properties in
the database can also be found within these tables (material
strength, reinforcement content, dimensions).
The original (unretrofitted) columns contained one of
the primary three deficiencies that result in nonductile
lateral-load behavior of columns: short lap splices in the
plastic hinge region (LS); insufficient amounts of trans-
verse reinforcement for concrete confinement (C); or low
shear strength (S). The measured lateral load-deformation
response of the column specimens was reported in the orig-
inal reference through either hysteresis curves or backbone
ACI Structural Journal/January 2018
Table 1—Properties of jacketed columns in database: circular sections
Height, Deficient Δy/H, Δp/H, Δmax/H, Par.
Specimen Reference Dc, in. in. fc′, psi P/Ag fc′ ρ, % ρv, % property* Vy, kip Vpeak/Vy % % % a
FRP-jacketed circular columns
CFRP-05 9.5 37.5 3467 0.05 2.54 0.333 LS 9.4 1.4 1.9 6.0 9.2 6.1
KFRP-05 Breña and 9.5 37.5 3467 0.05 2.54 0.333 LS 9.6 1.4 1.4 6.3 9.5 6.8
Schlick
CFRP-15 (2007) 9.5 37.5 3467 0.15 2.54 0.333 LS 9.1 1.8 1.3 6.0 9.2 7.9
KFRP-15 9.5 37.5 3467 0.15 2.54 0.333 LS 9.5 1.4 1.1 6.2 10.0 8.8
CF-R1 24.0 135.0 5223 0.06 1.95 0.103 LS 25.4 1.4 0.4 5.3 6.4 6.0
CF-R2 24.0 135.0 5353 0.06 1.95 0.103 LS 28.3 1.4 0.4 4.4 5.9 5.5
CF-R3 24.0 135.0 4758 0.07 1.95 0.103 LS 31.2 1.4 0.5 3.8 5.6 5.1
CF-R4 24.0 135.0 5469 0.06 1.95 0.103 LS 30.7 1.4 0.4 3.9 5.9 5.5
CF-R5 24.0 135.0 5759 0.06 1.95 0.103 LS 31.1 1.3 0.5 3.9 5.9 5.4
Haroun and
CF-R6 Elsanadedy 24.0 135.0 4802 0.07 1.95 0.103 LS 30.3 1.4 0.4 4.8 7.0 6.6
(2005a,b)
CS-R1 24.0 96.0 5919 0.05 1.95 0.103 S 80.2 1.4 0.2 3.1 3.6 3.4
CS-R2 24.0 96.0 5687 0.05 1.95 0.103 S 83.2 1.4 0.2 3.0 3.7 3.5
CS-R3 24.0 96.0 4961 0.05 1.95 0.103 S 116.1 1.4 0.5 3.8 4.3 3.8
CS-R4 24.0 96.0 5455 0.06 1.95 0.103 S 112.1 1.4 0.5 2.1 3.1 2.5
CS-P1 24.0 96.0 5179 0.06 1.95 0.103 S 119.8 1.4 0.3 3.9 4.1 3.7
Priestley et
Lap Splice R 24.0 144.0 4998 0.18 2.53 0.103 LS 42.7 1.4 0.1 2.4 3.8 3.5
al. (1994a,b)
C2-RT4 24.0 104.0 6501 0.05 1.94 0.103 LS 53.2 1.3 0.4 1.4 5.1 3.2
Xiao and
C3-RT5 24.0 104.0 6501 0.05 1.94 0.103 LS 52.7 1.4 0.4 2.5 5.0 3.7
Ma (1997)
C4-RP4 24.0 104.0 6501 0.05 1.94 0.103 LS 36.5 1.4 0.6 2.0 5.4 2.7
CAF1-2N 14.0 57.9 3611 0.05 1.71 0.275 LS 17.0 1.3 1.0 1.2 4.9 1.1
CAF1-5N Ghosh and 14.0 57.9 3640 0.27 1.71 0.275 LS 12.2 1.4 0.8 4.3 12.3 5.1
Sheikh
CBF1-6N (2007) 14.0 57.9 3843 0.05 1.71 1.032 LS 13.6 1.3 1.0 5.5 8.7 6.2
ST-4NT 14.0 57.9 6497 0.27 1.71 1.032 LS 25.7 1.3 1.8 3.9 9.0 7.2
Steel-jacketed circular columns
2 24.0 144.0 5601 0.16 2.53 0.174 C 38.8 1.4 0.8 2.5 2.5 1.7
4 24.0 144.0 5521 0.16 2.53 0.174 C 48.0 1.4 0.4 6.0 †
5.6
Chai et al.
5 24.0 144.0 5095 0.17 2.53 0.174 C 36.8 1.3 0.4 1.1 6.0 2.9
(1991)
6 24.0 144.0 5426 0.16 2.53 0.174 C 49.2 1.4 0.5 4.6 6.1 5.4
1-R 24.0 144.0 5541 0.16 2.53 0.174 C 38.0 1.4 0.9 2.7 5.1 3.9
SC1 Li et al. 29.9 128.0 3699 0.11 1.32 0.072 C 54.9 1.4 0.4 3.8 5.8 4.5
SC2 (2005) 29.9 128.0 3699 0.11 1.15 0.067 C 51.3 1.4 0.3 3.8 5.8 5.5
C2R 24.0 96.0 4931 0.06 2.53 0.082 S 115.4 1.4 0.3 4.4 ‡ 4.1
C4R Priestley et 24.0 96.0 5101 0.17 2.53 0.082 S 150.4 1.4 0.3 4.1 ‡ 3.8
C6R al. (1994a,b) 24.0 96.0 5801 0.05 2.53 0.082 S 160.9 1.4 0.4 5.5 ‡ 5.1
C8R 24.0 72.0 4521 0.06 2.53 0.082 S 193.0 1.4 0.3 5.2 ‡ 4.9
*
S is shear-deficient; C is inadequate confinement; LS is short lap splice.
†
Test was stopped at peak load.
‡
Test stopped at maximum displacement capacity of actuator.
Notes: 1 in. = 25.4 mm, 1 kip = 4.45 kN, 1000 psi = 6.89 MPa.

curves as a measure of column performance. When only
hysteresis curves were published, the backbone curve was
constructed using a series of lines passing through selected
points in the load-deformation response curve. These back-
bone curves were subsequently simplified to be consistent
with those available in retrofitting guides and standards such
as ACI 369R-11 and ASCE/SEI 41-13. The curves were
constructed by connecting the shear force and drift coor-
dinates obtained during the tests corresponding to column
yielding, lateral strength corresponding to a 20% reduction

ACI Structural Journal/January 2018 55

Table 2—Properties of jacketed columns in database: rectangular columns
bc, Deficient Δy/H, Δp/H, Δmax/H,
Specimen Reference in. hc, in. H, in. fc′, psi P/Agfc′ ρ, % ρv, % property* Vy, kip Vpeak/Vy % % % Par. a
FRP-jacketed rectangular columns
S2 15.8 7.9 70.8 1450 0.15 2.84 0.448 C 8.6 1.2 0.2 3.0 7.2 4.0
S3 Ozcan et al. 15.8 7.9 70.8 1523 0.15 2.84 0.448 C 9.4 1.3 0.3 2.0 4.8 2.7
S4 (2010) 15.8 7.9 70.8 1305 0.15 2.84 0.448 C 7.7 1.4 0.2 2.4 4.6 2.9
S5 15.8 7.9 70.8 2249 0.15 2.84 0.448 C 14.1 1.2 0.3 1.0 6.0 2.5
Confinement R Seible et al. 28.7 19.3 144.0 4998 0.14 4.65 0.128 C 107.2 1.4 0.4 2.7 3.3 2.8
Shear R (1997) 24.0 16.0 96.0 4998 0.06 2.52 0.154 S 97.9 1.2 0.1 1.5 2.3 2.1
C3 7.9 7.9 52.0 6775 0.24 2.00 0.000 C 10.8 1.3 1.0 2.1 7.2 6.2
C4 Wu et al. 7.9 7.9 52.0 6804 0.23 2.00 0.000 C 11.8 1.3 1.1 2.4 8.0 6.9
C5 (2008) 7.9 7.9 52.0 5281 0.30 2.00 0.000 C 11.1 1.3 1.0 4.0 7.2 6.2
C6 7.9 7.9 52.0 5368 0.30 2.00 0.000 C 11.0 1.3 1.0 2.0 7.3 6.3
ASG-2NSS 12.0 12.0 58.0 6165 0.15 2.44 0.316 C 24.6 1.4 0.7 4.6 23.7 2.6
ASG-3NSS 12.0 12.0 58.0 6195 0.15 2.44 0.316 C 24.9 1.4 0.2 6.0 12.0 3.5
ASG-4NSS 12.0 12.0 58.0 6282 0.15 2.44 0.316 C 24.2 1.4 0.3 2.4 11.4 2.0
Memon
ASG-5NSS and Sheikh 12.0 12.0 58.0 6340 0.15 2.44 0.316 C 24.4 1.4 0.6 4.3 11.2 2.0
(2005)
ASG-6NSS 12.0 12.0 58.0 6412 0.15 2.44 0.316 C 29.8 1.4 0.9 11.8 20.7 5.4
ASGR-7NSS 12.0 12.0 58.0 6412 0.15 2.44 0.316 C 24.2 1.4 0.5 5.8 14.9 2.6
ASGR-8NSS 12.0 12.0 58.0 6412 0.15 2.44 0.316 C 25.8 1.4 1.0 4.7 12.2 3.4
ASC-2NS 12.0 12.0 58.0 5295 0.15 2.44 0.321 C 30.4 1.3 0.3 0.7 2.3 1.3
ASC-3NS 12.0 12.0 58.0 5353 0.15 2.44 0.321 C 31.8 1.3 0.2 0.6 2.2 2.0
ASC-4NS 12.0 12.0 58.0 5353 0.15 2.44 0.321 C 26.7 1.3 0.1 0.3 2.2 0.7
Iacobucci
ASC-5NS et al. 12.0 12.0 58.0 5368 0.15 2.44 0.321 C 33.7 1.3 0.4 2.2 4.5 2.3
(2003)
ASC-6NS 12.0 12.0 58.0 5368 0.15 2.44 0.321 C 31.0 1.3 0.2 0.8 3.8 1.7
ASCR-7NS 12.0 12.0 58.0 5368 0.15 2.44 0.321 C 29.3 1.3 0.4 0.7 5.2 2.5
ASCR-8NS 12.0 12.0 58.0 6136 0.15 2.44 0.321 C 24.5 1.3 0.6 2.0 3.0 2.3
F2 18.0 18.0 70.1 3597 0.22 1.48 0.180 C 35.0 1.4 1.1 7.7 13.4 6.2
Harries et
L1 18.0 18.0 70.1 4162 0.22 1.48 0.180 LS 36.2 1.4 1.0 5.6 11.4 6.0
al. (2006)
L2 18.0 18.0 70.1 4162 0.22 1.48 0.180 LS 36.4 1.4 1.2 5.8 8.6 5.4
RF-R1 24.0 24.0 135.0 5135 0.06 2.14 0.103 LS 55.9 1.3 0.7 2.4 2.8 2.1
RF-R2 24.0 24.0 135.0 6078 0.05 2.14 0.103 LS 56.2 1.3 0.7 1.1 4.5 3.8
RF-R3 24.0 24.0 135.0 6122 0.05 2.14 0.103 LS 60.3 1.4 0.7 2.0 3.6 2.8
RF-R4 24.0 24.0 135.0 6122 0.05 2.14 0.103 LS 59.2 1.3 0.8 1.7 3.5 2.7
RS-R1 Haroun and 24.0 18.0 96.0 5527 0.06 2.04 0.137 S 107.0 1.3 0.2 1.7 4.7 3.5
Elsanadedy
RS-R2 (2005a,b) 24.0 18.0 96.0 5701 0.06 2.04 0.137 S 104.6 1.3 0.3 1.1 4.7 3.5
RS-R3 24.0 18.0 96.0 6383 0.06 2.04 0.137 S 105.1 1.3 0.2 1.5 4.4 4.1
RS-R4 24.0 18.0 96.0 6383 0.06 2.04 0.137 S 95.8 1.4 0.1 2.1 4.6 4.5
RS-R5 24.0 18.0 96.0 6383 0.06 2.04 0.137 S 97.2 1.4 0.2 2.5 4.2 4.0
RS-R6 24.0 18.0 96.0 6180 0.06 2.04 0.137 S 107.0 1.3 0.2 1.8 4.9 4.7
C1FP1 11.8 5.9 39.4 3061 0.23 1.72 0.354 LS 10.9 1.4 0.9 3.0 6.0 3.4
C1FP2 11.8 5.9 39.4 3148 0.22 1.72 0.354 LS 5.5 3.9 0.6 3.0 5.1 3.3
C1F1 11.8 5.9 39.4 3177 0.22 1.72 0.354 LS 12.5 1.3 0.9 2.0 5.0 2.5
C1F2 Harajli 11.8 5.9 39.4 3163 0.22 1.72 0.354 LS 11.2 1.4 1.0 3.0 5.1 3.4
and Rteil
C2FP1 (2004) 11.8 5.9 39.4 3061 0.27 3.56 0.354 LS 15.2 1.4 0.8 3.0 5.1 3.0
C2FP2 11.8 5.9 39.4 3148 0.26 3.56 0.354 LS 15.1 1.4 0.7 3.0 5.1 3.2
C2F1 11.8 5.9 39.4 3177 0.26 3.56 0.354 LS 15.6 1.4 0.8 3.1 5.0 3.4
C2F2 11.8 5.9 39.4 3163 0.26 3.56 0.354 LS 15.3 1.4 0.8 3.0 5.1 3.4

56 ACI Structural Journal/January 2018

Table 2(cont.)—Properties of jacketed columns in database: rectangular columns
bc, Deficient Δy/H, Δp/H, Δmax/H,
Specimen Reference in. hc, in. H, in. fc′, psi P/Agfc′ ρ, % ρv, % property* Vy, kip Vpeak/Vy % % % Par. a
SAF1-10N 12.0 12.0 57.9 3887 0.33 2.44 0.321 LS 19.9 1.3 0.8 2.5 7.8 3.2
SBF1-11N Ghosh and 12.0 12.0 57.9 3916 0.05 2.44 1.205 LS 14.3 1.3 0.8 2.3 3.8 2.1
Sheikh
SBRF1-12N (2007) 12.0 12.0 57.9 3945 0.05 2.44 1.205 LS 10.2 1.4 1.9 3.9 8.1 3.5
ASC-2NS 12.0 12.0 57.9 5294 0.33 2.44 1.205 LS 24.3 1.3 0.8 1.7 7.5 2.8
C14FP1 15.7 7.9 59.1 5656 0.00 1.29 0.735 LS 15.4 1.3 0.6 2.1 6.4 2.8
C14FP2 15.7 7.9 59.1 5656 0.00 1.29 0.735 LS 16.3 1.3 0.5 3.2 6.4 3.8
C16FP1 Harajli and 15.7 7.9 59.1 7107 0.00 2.00 0.735 LS 20.0 1.3 0.9 2.1 6.5 2.3
Dagher
C16FP2 (2008) 15.7 7.9 59.1 7107 0.00 2.00 0.735 LS 19.3 1.3 0.9 2.2 6.5 3.8
C20FP1 15.7 7.9 59.1 4641 0.00 2.13 0.735 LS 22.8 1.3 1.0 2.1 6.5 2.2
C20FP2 15.7 7.9 59.1 4641 0.00 2.13 0.735 LS 25.5 1.3 1.0 3.2 6.4 2.9
SC2 12.0 12.0 36.0 5658 0.14 6.11 0.904 S 76.6 1.3 0.1 1.8 5.7 5.6
SC1R 12.0 12.0 36.0 4932 0.16 6.11 0.904 S 72.8 1.2 0.3 1.8 5.8 3.8
SC2R Galal et al. 12.0 12.0 36.0 4932 0.16 6.11 0.904 S 76.4 1.3 0.3 1.8 4.2 2.4
SC1U (2005) 12.0 12.0 36.0 6238 0.12 6.11 0.904 S 45.4 2.2 0.1 0.9 5.6 3.2
SC3 12.0 12.0 36.0 5658 0.14 6.11 0.904 S 78.4 1.3 0.2 1.8 5.7 4.0
SC3R 12.0 12.0 36.0 4932 0.16 6.11 0.904 S 54.6 1.3 0.3 0.6 3.6 1.3
Steel-jacketed rectangular columns
C-66-R Alcocer and 19.7 19.7 78.7 4047 0.15 2.44 0.142 C 42.0 1.4 1.0 2.5 † 1.5
Durán-
C-66-S Hernández 19.7 19.7 78.7 4047 0.15 2.44 0.142 C 64.7 1.4 1.0 2.7 † 1.7
(2002)
RC-2R 10.0 10.0 40.0 8269 0.30 2.48 0.220 C 45.8 1.4 0.4 2.1 6.0 3.1
RC-3R Xiao and 10.0 10.0 40.0 8269 0.30 2.48 0.220 C 51.4 1.4 0.5 3.0 8.0 7.5
RC-4R Wu (2003) 10.0 10.0 40.0 8269 0.30 2.48 0.220 C 50.9 1.4 0.4 3.1 8.0 6.8
RC-5R 10.0 10.0 40.0 8704 0.30 2.48 0.220 C 52.7 1.4 0.4 3.1 8.5 8.1
FC9 18.0 36.0 144.0 2906 0.00 1.95 0.095 LS 38.4 1.4 0.5 2.7 4.9 3.2
FC11 Aboutaha et 18.0 36.0 144.0 2851 0.00 1.95 0.095 LS 45.4 1.4 0.8 2.4 5.5 2.5
FC12 al. (1996) 18.0 36.0 144.0 3266 0.00 1.95 0.095 LS 43.1 1.4 0.6 2.7 5.5 3.7
FC17 18.0 18.0 144.0 2636 0.00 1.95 0.076 LS 45.6 1.4 0.2 2.4 5.4 5.2
FC6 18.0 36.0 144.0 2851 0.00 1.95 0.095 LS 34.9 1.4 1.0 2.5 4.6 2.2
FC7 Aboutaha et 18.0 36.0 144.0 2981 0.00 1.95 0.153 LS 52.0 1.4 1.5 3.9 † 2.4
FC10 al. (1999b) 18.0 36.0 144.0 2596 0.00 1.95 0.095 LS 37.2 1.4 0.8 2.4 3.4 2.6
FC13 18.0 36.0 144.0 3266 0.00 1.95 0.095 LS 50.0 1.4 0.4 3.6 5.0 4.6
SC6 18.0 36.0 48.0 2256 0.00 1.95 0.095 S 111.1 1.4 1.0 3.3 5.1 3.2
SC7 Aboutaha et 18.0 36.0 48.0 2941 0.00 1.95 0.095 S 101.2 1.4 0.5 4.2 6.3 5.8
SC8 al. (1999a) 18.0 36.0 48.0 2786 0.00 1.95 0.095 S 109.7 1.4 0.7 3.9 7.0 5.8
SC10 36.0 18.0 48.0 2391 0.00 1.95 0.191 S 205.2 1.4 0.6 4.0 5.3 4.7
R2R 16.0 24.0 96.0 5601 0.05 2.52 0.163 S 104.4 1.4 0.3 3.6 ‡
Priestley
R4R et al. 16.0 24.0 96.0 5201 0.06 2.52 0.082 S 154.7 1.4 0.3 3.8 ‡
(1994a,b)
R6R 16.0 24.0 72.0 4801 0.06 2.52 0.082 S 205.6 1.4 0.4 3.7 ‡
*
S is shear deficient; C is inadequate confinement; LS is short lap splice.
†
Test stopped when capacity of actuator was reached.
‡
Test stopped at maximum displacement capacity of actuator.
Notes: 1 in. = 25.4 mm; 1 kip = 4.45 kN; 1000 psi = 6.89 MPa.

from peak strength, maximum drift, and, if available, loss after loss of axial capacity and maximum drift were assumed
of axial load-carrying capacity. It should be noted that the based on experience with unjacketed columns because of
results reported in the literature were used directly without lack of experimental data reporting these values.
modification to include P-∆ effects. The residual strength

ACI Structural Journal/January 2018 57

Fig. 2—Simplified backbone curve obtained from measured hysteretic response: (a) hysteresis curve with superimposed back-
bone; and (b) simplified nonlinear backbone curve and nonlinear parameters.
The procedure used to consistently define the key points
in the force-deformation response of jacketed columns is
illustrated schematically in Fig. 2. Shear at yield and defor-
mation at yield (Vy, Δy) were determined as the coordinates
of the point of intersection of two lines drawn on the hyster-
esis curve. One line was parallel to the initial slope of the
measured force-deformation response of the specimens and
the second line was drawn horizontally at a shear force equal
to either 0.8Vpeak or 0.7Vpeak, depending on the shape of the
hysteresis curve. A shear force of 0.8Vpeak was chosen if the
drift at peak was equal or less than 2% followed by rapid
strength loss at increasing drift amplitudes. Otherwise, the
shear force at yield was assumed equal to 0.7Vpeak (Alvarez
and Breña 2014). The peak shear force Vpeak was simply taken
as the highest applied lateral load during testing. Almost
all the columns in the database were not tested beyond a
strength degradation corresponding to a 20% drop from
Vpeak. Therefore, the residual strength of the columns was
assumed equal to 0.2Vpeak. The force values thus determined
were used to evaluate whether the simple sectional models
described later in this paper can be used to estimate yield and
peak force of retrofitted columns. Drifts were studied statis-
tically to propose values corresponding to different exceed-
ance probabilities that can be used in practice.
NONLINEAR MODELING OF JACKETED
COLUMNS
Nonlinear modeling of reinforced concrete frames has
become widespread in structural engineering practice with
the availability of software capable to perform these types of
analyses efficiently. Although significant efforts have been
made to develop modeling procedures for existing reinforced
concrete components, recommendations on how to approach
modeling of retrofitted frame components are lacking. One
retrofitting approach is to jacket frame components to
improve performance, particularly in regions anticipated to
undergo inelastic deformations. Typical reinforcing details
found in columns in older reinforced concrete frames have
been identified as the cause for nonductile column perfor-
mance in past earthquakes and related experimental testing.
The performance of these columns can be improved by jack-
58
eting the region of anticipated inelastic action, but the effect
of this local retrofitting technique on the global response
of a frame is not well understood. This section focuses on
developing an approach to construct the nonlinear backbone
response of jacketed columns for use in nonlinear analyses
of retrofitted frames.
The response of jacketed columns has been studied exten-
sively in laboratory tests conducted over the last 30 years.
These studies were initiated after dramatic failures of bridges
containing columns with deficient details occurred during
earthquakes in the two decades that followed the 1970s
(for example, 1971 San Fernando Earthquake; and Hanshin
Expressway in 1994 Kobe Earthquake). Although failures of
columns in frames have not been pervasive, the reinforcing
details that these elements contain are known to result in
unsatisfactory performance under large earthquakes. The
potentially high consequence caused by failures of defi-
cient columns in framed buildings can be mitigated using
local retrofitting—for example, by using jacketing in plastic
hinge zones near the end of columns. Development of local
retrofitting techniques by jacketing of columns was accom-
plished through laboratory testing, but modeling guidelines
aimed at capturing the nonlinear response of these jacketed
components were not developed accordingly.
The data from experiments of jacketed columns presented
in the previous section of this paper were used to compare
the calculated values with measured response to provide
confidence in the proposed recommendations developed
to construct the nonlinear backbone response of jacketed
columns. Constructing the nonlinear backbone response
of jacketed columns requires knowledge of the force
developed and its corresponding deformation at various
lateral load demands. The procedures followed to determine
these two groups of parameters (force and deformation) at
various demands were different. Internal forces at different
demand levels (yield strength, peak strength, and residual
strength) were determined using well-established sectional
models calibrated to closely capture the measured response
of jacketed columns in the experiments. The deformations
(lateral drift) corresponding to each force level (yield, peak,
drift at residual strength) were determined through statis-
ACI Structural Journal/January 2018
tical analysis of available data given the large variation of
drifts observed during the tests and the lack of a robust drift
model that captures drift values reliably. The procedures to
determine these force and deformation parameters required
to construct nonlinear backbone curves of jacketed columns
are presented in more detail in the following sections. These
methods can be easily applied in practice, as their use only
requires knowledge of the geometric and material properties
of a given jacketed column.
Lateral load strength of jacketed column sections
The sectional model used to determine yield and peak
force of jacketed columns assumed that behavior is governed
by flexure (no shear failure) after retrofitting and that
shear-moment interaction can be neglected. These assump-
tions were verified by calculating the shear strength of retro-
fitted columns using the models proposed by Priestley et al.
(1994a,b) and Seible et al. (1997) for steel jacketed and FRP
jacketed columns, respectively. At the maximum imposed
displacement in the tests, the ratio of applied load to calcu-
lated shear strength did not exceed 0.86 for FRP-jacketed and
0.89 for steel-jacketed columns, respectively. These results
imply that the jacketed columns included in this study would
be expected to develop flexural hinges prior to shear failure,
at the largest displacement demands experienced during
the tests. Flexural yield and flexural strength were there-
fore calculated using a moment-curvature analysis of the
retrofitted cross sections. Moment-curvature analysis was
chosen because of its computational efficiency and accuracy
for sections controlled by flexure. After the moment associ-
ated with each of these two conditions was computed, the
shear force corresponding to yield (Vy) and peak strength
(VPeak) was determined based on the relationship between
moment and shear in each test column. The yield and peak
shear values were compared with the database of jacketed
column test results and found to perform well for undam-
aged columns prior to retrofitting and for columns with an
axial force ratio (P/(Agfc′)) equal or less than 0.10.
To determine flexural yield and peak moment with the
sectional model, the column section was discretized using
fibers to represent concrete, reinforcing steel, and jacket
materials. The behavior of each fiber was assumed to be
governed by the uniaxial stress-strain response of each
material, as is commonly done when using fiber section
models. It is worth noting that the behavior of concrete was
represented by a confined concrete model for concrete in
compression, assuming that the externally applied jacket
was effective in confining the concrete throughout the
cross section. The decrease in confinement efficiency of the
jacket, particularly in the case of rectangular cross sections,
was considered as will be described as follows.
The uniaxial stress-strain behavior of concrete was
modeled including the effects of confinement for concrete
in compression and a linear elastic model for concrete
in tension. The strength of concrete confined by steel or
FRP jackets was calculated using the models proposed by
Priestley et al. (1994a,b) or Lam and Teng (2003a,b), respec-
tively. The use of these models is suitable for the range of
concrete strengths found in the jacketed column database
ACI Structural Journal/January 2018
(maximum of 8300 psi [56 MPa]). Once the confined
concrete strength was determined, the model developed by
Mander et al. (1988) was used to construct the complete
uniaxial stress-strain relation for confined concrete, therefore
neglecting the further increase in confined concrete strength
that results from FRP confinement, particularly at large axial
strains. It was considered that, for the load levels applied
during testing of the columns included in the database, the
axial strains generated in the compression zone from flexure
would not result in a significant further increase in confined
concrete strength. The effects of confinement from steel or
FRP jackets was represented as confinement by an equivalent
amount of transverse reinforcement by equating the lateral
confining stress from each jacket type to that developed by
transverse reinforcement at a given spacing (Alvarez and
Breña 2014). Confinement efficiency is reduced by arching
that develops between layers of transverse steel and between
longitudinal bars laterally restrained by the corners of hoops
or ties. These two effects were considered in the model by
decreasing the confining stress efficiency of jackets as appli-
cable (refer to Alvarez and Breña [2014] for further details).
Longitudinal reinforcing steel was modeled using an elas-
toplastic material model with strain hardening. The contri-
bution of the steel or FRP jackets to flexural strength of the
cross section was only considered for their effect on concrete
confinement. For steel jackets, their contribution to increase
the total tension force in the cross section was neglected
because of the potential for slip between the jacket and grout
used to fill the concrete-steel gap. FRP jackets are typically
applied with fibers oriented perpendicular to the column axis
so their contribution to increased flexural strength was also
neglected. The confinement provided by jacketing is passive,
activated when the concrete expands due to microcracking.
Similar to confinement provided by closely spaced hoops,
jacket confinement efficiency depends on the properties of
the jacket (strength and stiffness) and cross-sectional geom-
etry (circular or rectangular column).
The sectional model provided reasonably accurate results
when the original column had not been damaged prior to
jacketing, and the existing longitudinal reinforcement was
continuous. The yield and peak strength of columns that
were damaged before jacketing could not be accurately
calculated using a moment-curvature approach as described
in this paper. In columns containing lapped longitudinal
reinforcement within the plastic hinge region, the maximum
stress in the longitudinal reinforcement was calculated using
a modified version of the equation proposed by Cho and
Pincheira (2006), which estimates bar stress as a nonlinear
function of splice length lb in accordance with
fs = 1.25(lb/ld)2/3fy ≤ fy (1)
where ld is the calculated development length of the bars
in accordance with ACI 318-14, and fy is the nominal
yield stress of the bar. Figure 3 compares yield strength
(Vy) and peak strength (Vpeak) values determined using the
moment-curvature analysis of the column cross section and
the experimentally determined values of jacketed columns in
the database. For most columns in the database, the discrep-
59
Fig. 3—Comparison between calculated and laboratory jacketed column strength: (a) and (b) shear at yield; and (c) and (d)
shear at peak strength. (Note: 1 kip = 4.45 kN.)
ancy between calculated and measured values at yield and
peak do not exceed 10%, as shown in the figure. Given
the simple sectional model used, a difference of 10% in
calculated and measured shear strength at yield and peak is
considered adequate, given the large variations in jacketing
configurations between testing programs and the diverse
deficiencies that the original columns contained. A sectional
model that incorporates the assumptions described in this
paper is therefore recommended in practice to estimate yield
and peak force of jacketed columns.
Lateral deformation (drift) of jacketed columns
Several deformation components contributed to the total
drift in jacketed columns in the database. These components,
also common in other reinforced concrete elements, include
flexural deformation, shear deformation, and rigid body
rotation induced by bar slippage. Any mechanistic model
that would capture these effects would be too complex for
use in design (in contrast to the relatively straightforward
sectional model for force levels presented in the previous
section). Therefore, the database of experimental results was
used to directly calibrate an empirical model for drift at 80%
of peak force for columns with low axial force ratios. These
values could then be used to find nonlinear drift param-
eter a to construct the backbone nonlinear response curves
of jacketed columns. When the axial force ratio was high
(>0.60), insufficient data were available and, therefore, the
value of parameter a for well-designed unjacketed columns
was adopted. For drift parameter b, the drift at which only
residual shear capacity remains, values from well-designed
unjacketed columns were adopted for both low and high
axial force ratios because insufficient test data on jacketed
60
columns are available. The statistical analysis procedure
employed in this process is described in this section.
Statistical analysis of drift data for jacketed columns in
database—Histograms that show plastic drift data (param-
eter a) of columns in the database are presented in Fig. 4,
distinguished only by jacket type (FRP or steel) without
separation by cross-sectional shape. Instead of illustrating
the distribution of measured deformation data directly,
parameter a (plastic drift capacity) was chosen as a directly
meaningful parameter to represent the nonlinear drift data
in accordance with ASCE/SEI 41-13. Parameter a was
determined for each jacketed column in the database as the
difference between the drift corresponding to a 20% reduc-
tion in lateral load and the drift at yield defined as described
previously. The data show differences in parameter a
between FRP-jacketed and steel-jacketed columns. FRP-jack-
eted columns have a range of parameter a from 0.0073 to
0.0882—a mean value of 0.0380 and a ±1 standard devia-
tion range of 0.0213 to 0.0547. In contrast, steel-jacketed
columns have parameter a-values that range from 0.0149 to
0.0812, a mean value of 0.0413, and a ±1 standard deviation
range of 0.0246 to 0.0580. The drift data and the general
behavior observed from the reported hysteresis curves
also indicated that cross-sectional shape was an important
factor that influenced the deformation behavior of jack-
eted columns. The average plastic deformation capacity of
circular sections for both FRP- and steel-jacketed columns
was slightly higher than for rectangular columns, primarily
due to the higher confining and lap-splice clamping effi-
ciency of circular jackets over rectangular ones.
Cross section shape appears to have a meaningful effect
on structural response; therefore, Fig. 4 shows the data histo-
ACI Structural Journal/January 2018
Fig. 4—Histograms of parameter a for FRP- and steel-jacketed columns.
grams separated by cross-sectional shape. Although sepa-
rating data into groups in this way results in modest sample
sizes for each group, it is needed because the mechanics of
confinement differ for circular and rectangular columns and,
furthermore, the number of samples is not inconsistent with
the amount of data available for other reinforced concrete
components for which limited laboratory tests have been
conducted. A comparison of parameter a data with three
typical statistical distributions (lognormal, Weibull, and
Rayleigh) is presented in Fig. 5. These three distributions
were chosen because they all satisfy the physical constraint
that parameter a must always have values exceeding zero.
Although both the Weibull and lognormal distributions
provide good fits to central values of the data, the lognormal
distribution has a much heavier upper tail than is indicated
by the data, which would lead to significant probability of
highly unconservative values of drift. Therefore, the Weibull
distribution was selected as the appropriate statistical model
for drift, and fitting parameters for the data shown in Fig. 5 are
listed in Table 3. Using this distribution, parameter a values
were calculated for a range of probabilities of exceedance
between 50 and 95%, as given in Table 4. These data are
provided to allow for selection of design values for param-
eter a based on exceedance probability in place of the mean
value. For both jacketing materials, there is a clear differ-
ence between parameter a depending on section shape. It is
clear that plastic drift capacity (parameter a) is greater in
circular columns than in rectangular columns, as would be
expected given the higher confining and clamping efficiency
of jackets on circular cross sections. The estimated plastic
drifts that columns can sustain are comparable between jack-
eted circular columns independent of jacket material. For
rectangular columns, steel-jacketed columns are estimated
to reach higher plastic drifts than FRP-jacketed columns,
as would be expected because of the higher bending stiff-
ness of steel jackets. The comparisons are in line with the
anticipated differences in behavior of circular and rectan-
gular columns jacketed with either steel or FRP materials.
It should be noted that although a similar statistical analysis
of parameter b would be desirable, insufficient data exist
in the open literature to perform such an analysis because
many of the published tests were stopped prior to the column
ACI Structural Journal/January 2018
Table 3—Fitting parameters for Weibull distribution
of parameter a
Weibull fitting parameters
Jacketing material Cross section shape μ k
Circular 0.0553 2.91
FRP
Rectangular 0.0380 2.64
Circular 0.0471 4.74
Steel
Rectangular 0.0458 2.36
reaching residual capacity. This points to a need for further
testing that extends the testing protocol to the point where
the force decays to residual capacity.
RECOMMENDED BACKBONE PARAMETERS FOR
JACKETED COLUMNS
The ACI 369R-11 contains modeling parameters for
components of existing frame buildings and is in the process
of being updated by using recent test data and statistical
studies of available component tests. As noted previously,
modeling recommendations for retrofitted components are
lacking, so there is a need to provide this information. The
approach being followed by ACI 369R is to use mean defor-
mation values from tests to define modeling parameters. The
suggested values in this section were developed with this
approach in mind to facilitate future adoption of the tabulated
values in updates of the ACI 369R and ASCE/SEI 41 docu-
ments. It is important to note that, due to differences in the
ability to model force and drift levels for retrofitted columns
and differences in the availability of data for different param-
eters in the backbone curve, a hybrid approach to specifying
backbone curve parameters is used herein. For the force
levels at yield and peak, a mechanistic sectional model was
proposed and validated. For drift parameter a, experimental
data were analyzed and a probabilistic model was fit to the
data. Although herein the mean value is recommended to be
consistent with current recommended practice, the proba-
bilistic model is described so that in the future, it will be
possible to specify design values based on desired exceed-
ance probability rather than simply using the mean value.
For drift parameter b, recommendations for unretrofitted
columns were adopted in the absence of sufficient data for
61
Fig. 5—Distribution of parameter a and fitted lognormal, Weibull, and Rayleigh statistical distributions.
Table 4—Values of parameter a for selected base, mean data were only computed for parameter a and
probabilities of exceedance for columns tested under low values of axial force (P/(Agfc′)
Parameter a ≤ 0.10). Recommended values of other parameters (b and
c, the residual capacity ratio) are based on the similarities
FRP-jacketed Steel-jacketed
Probability of observed in the behavior of jacketed columns in comparison
exceedance, % Circular Rectangular Circular Rectangular with columns containing reinforcing details representative
95 0.0199 0.0123 0.0251 0.0130 of new design (code conforming columns). The similarities
90 0.0255 0.0161 0.0293 0.0176 in backbone behavior of an originally deficient column that
has been retrofitted using two different jacketing configura-
85 0.0296 0.0190 0.0321 0.0212
tions with the behavior of a similar code-conforming column
80 0.0330 0.0214 0.0343 0.0242 is shown in Fig. 6. The behavior of the jacketed columns
75 0.0360 0.0236 0.0362 0.0270 is remarkably similar to that of columns with well-detailed
70 0.0388 0.0256 0.0379 0.0295 reinforcement, providing support for using modeling param-
eters of these columns for jacketed columns where there are
65 0.0414 0.0275 0.0394 0.0320
no available laboratory data.
60 0.0439 0.0294 0.0408 0.0344 Very few jacketed column tests have been conducted to
55 0.0463 0.0312 0.0422 0.0368 the large displacement demands needed to generate axial
50 0.0487 0.0330 0.0436 0.0392
load failures. One of the few test series of jacketed column
tests conducted to the point of residual lateral load levels is
jacketed columns. This assumption is conservative because the work reported by Ghosh and Sheikh (2007). The tests
jacketing would be expected to increase drift relative to an included reinforced concrete columns containing short
unjacketed column. lap splices retrofitted with FRP jackets. Some of the tests
Based on the statistical analysis of jacketed columns, reported in this research reached rupture of the jacket and
mean values of parameter a for the two cross sections studied lateral load degradation of 90% from the peak strength at
(circular and rectangular) and the two jacket materials included a plastic drift of 11%. These tests demonstrate that plastic
in the database (FRP and steel) were computed (Table 5). drift at strength degradation reaches values that exceed those
These four categories are consistent with the way the data being proposed based on results of well-detailed columns
were organized and statistically studied. Because of the and provides support for the conservative adoption of param-
limitations found in the jacketed column experimental data- eter b values based on well-detailed but unjacketed columns.

62 ACI Structural Journal/January 2018

Table 5—Proposed modeling parameters for FRP- and steel-jacketed columns
Section parameters
Jacketing material Section shape P/(Agfc′)
≤0.1
Circular
≥0.6
FRP
≤0.1
Rectangular
≥0.6
≤0.1
Circular
≥0.6
Steel
≤0.1
Rectangular
≥0.6
Fig. 6—Comparison of backbone force-deformation behavior
of code-conforming and two different jacketed columns.
Should further test data become available, it may be possible
to increase the recommended value of drift parameter b for
jacketed columns.
CONCLUSIONS
This paper presents a procedure that can be used in prac-
tice to construct the nonlinear backbone force-deformation
response of FRP- or steel-jacketed columns. Backbone
curves of these retrofitted components can be used to eval-
uate the response of frame structures with selective jack-
eting of columns. Based on the results of this research, the
following conclusions and modeling recommendations can
be made:
1. Section (fiber) models can be used to determine the
lateral load at yield (Vy) and peak (Vpeak) of jacketed columns,
as long as the column shear strength exceeds the force
required to reach the nominal flexural strength of the section.
2. The contribution of FRP jackets to the flexural strength
of retrofitted sections can be neglected if fibers are oriented
in the transverse direction. Steel jacket contribution to flex-
ural strength can also be neglected because of potential slip-
ping between jacket and column cross section.
3. Jackets that extend throughout the plastic hinge region
of columns can provide concrete confinement and clamping
of short lap splices within this region. The compression
ACI Structural Journal/January 2018
Modeling parameters
Plastic rotations angle, rad Residual strength ratio
a b c
0.049 0.060 0.2
0.010 0.010 0.0
0.034 0.060 0.2
0.010 0.010 0.0
0.043 0.060 0.2
0.010 0.010 0.0
0.040 0.060 0.2
0.010 0.010 0.0
strength of confined concrete can be calculated using models
for confined concrete found in the literature for both FRP-
and steel-confined concrete.
4. In this study, the lack of deformation data for jacketed
columns subjected to large displacements was overcome by
conducting a statistical study of available data to estimate
plastic drift capacity.
5. The mean plastic displacement corresponding to a
lateral load degradation of 20% from peak can be used to
define parameter a to allow construction of backbone curves
that are consistent with ASCE/SEI 41-13.
6. There is a sparsity of test data for jacketed columns
subjected to large displacements that generate a significant
shear force drop from peak (more than 20%). Therefore, in
this research, parameter b was based on values that corre-
spond to the response of columns with code-conforming rein-
forcement details based on the similar hysteretic response
observed between well-detailed and jacketed columns.
AUTHOR BIOS
ACI member José C. Alvarez is a Visiting Assistant Professor at Quin-
nipiac University, Hamden, CT. He received his PhD from the University
of Massachusetts Amherst, Amherst, MA, in 2017, and his BS in civil and
environmental engineering from the University of Puerto Rico, Mayagüez
Campus, PR. His research interests include seismic design and rehabilita-
tion of new and existing structures.
Sergio F. Breña, FACI, is a Professor at the University of Massa-
chusetts Amherst. He received his MS and PhD from the Univer-
sity of Texas at Austin, Austin, TX. He is a member of ACI Commit-
tees 318-C, Safety, Serviceability and Analysis Structural Concrete
Building Code; 369, Seismic Repair and Rehabilitation; 374, Perfor-
mance-Based Seismic Design of Concrete Buildings; 440, Fiber-
Reinforced Polymer Reinforcement; and Joint ACI-ASCE Committee 445,
Shear and Torsion. His research interests include the behavior of strength-
ened concrete elements using fiber composites and performance evaluation
of structures through field instrumentation.
Sanjay R. Arwade is a Professor of civil engineering at the University of
Massachusetts Amherst. Educated at Princeton and Cornell, his research
interests include probabilistic material and structural mechanics and reli-
ability analysis of structural systems.
ACKNOWLEDGMENTS
The database assembled as part of the research reported in this paper
was made possible by funding received from the ACI Concrete Research
Council. The first author expresses his gratitude to the Northeast Alli-
ance for Graduate Education and the Professoriate (NEAGEP) at UMass
Amherst for granting him a fellowship to partially support his PhD studies.
63
 REFERENCES
Aboutaha, R. S.; Engelhardt, M. D.; Jirsa, J. O.; and Kreger, M. E., 1996,
“Retrofit of Concrete Columns with Inadequate Lap Splices by the Use of
Rectangular Steel Jackets,” Earthquake Spectra, V. 12, No. 4, pp. 693-714.
doi: 10.1193/1.1585906
Aboutaha, R. S.; Engelhardt, M. D.; Jirsa, J. O.; and Kreger, M. E., 1999a,
“Rehabilitation of Shear Critical Concrete Columns by Use of Rectangular
Steel Jackets,” ACI Structural Journal, V. 96, No. 1, Jan.-Feb., pp. 68-78.
Aboutaha, R. S.; Engelhardt, M. D.; Jirsa, J. O.; and Kreger, M. E.,
1999b, “Experimental Investigation of Seismic Repair of Lap Splice Fail-
ures in Damaged Concrete Columns,” ACI Structural Journal, V. 96, No. 2,
Mar.-Apr., pp. 297-307.
ACI Committee 369, 2011, “Guide for Seismic Rehabilitation of Existing
Concrete Frame Buildings (ACI 369R-11),” American Concrete Institute,
Farmington Hills, MI, 35 pp.
Alcocer, S. M., and Durán-Hernández, R., 2002, “Seismic Perfor-
mance of a RC Building with Columns Rehabilitated with Steel Angles
and Straps. Innovations in Design with Emphasis on Seismic, Wind, and
Environmental Loading,” Quality Control and Innovations in Materials/
Hot Weather Concreting, pp. 531-552.
Alvarez, J. C., and Breña, S. F., 2014, “Non-Linear Modeling Param-
eters for Jacketed Columns Used in Seismic Rehabilitation of RC Build-
ings,” Seismic Assessment of Reinforced Concrete Buildings, SP-297, K. J.
Elwood, J. Dragovich, and I. Kim, eds., American Concrete Institute, Farm-
ington Hills, MI. (CD-ROM)
ASCE/SEI 41-13, 2013, “Seismic Rehabilitation of Existing Buildings,”
American Society of Civil Engineers, Reston, VA.
Breña, S. F., and Schlick, B. M., 2007, “Hysteretic Behavior of Bridge
Columns with FRP-Jacketed Lap Splices Designed for Moderate Ductility
Enhancement,” Journal of Composites for Construction, ASCE, V. 11,
No. 6, pp. 565-574. doi: 10.1061/(ASCE)1090-0268(2007)11:6(565)
Chai, Y. H.; Nigel Priestley, M. J. N.; and Seible, F., 1991, “Seismic
Retrofit of Circular Bridge Columns for Enhanced Flexural Performance,”
ACI Structural Journal, V. 88, No. 5, Sept.-Oct., pp. 572-584.
Cho, J.-Y., and Pincheira, J. A., 2006, “Inelastic Analysis of Reinforced
Concrete Columns with Short Lap Splices Subjected to Reversed Cyclic
Loads,” ACI Structural Journal, V. 103, No. 2, Mar.-Apr., pp. 280-290.
Galal, K.; Arafa, A.; and Ghobarah, A., 2005, “Retrofit of RC Square
Short Columns,” Engineering Structures, V. 27, No. 5, pp. 801-813. doi:
10.1016/j.engstruct.2005.01.003
Ghosh, K. K., and Sheikh, S. A., 2007, “Seismic Upgrade With Carbon
Fiber-Reinforced Polymer of Columns Containing Lap-Spliced Reinforcing
Bars,” ACI Structural Journal, V. 104, No. 2, Mar.-Apr., pp. 227-236.
Harajli, M. H., and Dagher, F., 2008, “Seismic Strengthening of
Bond-Critical Regions in Rectangular Reinforced Concrete Columns Using
Fiber-Reinforced Polymer Wraps,” ACI Structural Journal, V. 105, No. 1,
Jan.-Feb., pp. 68-77.
Harajli, M. H., and Rteil, A. A., 2004, “Effect of Confinement Using
Fiber-Reinforced Polymer or Fiber-Reinforced Concrete on Seismic
Performance of Gravity Load-Designed Columns,” ACI Structural Journal,
V. 101, No. 1, Jan.-Feb., pp. 47-56.
Harries, K. A.; Ricles, J. R.; Pessiki, S.; and Sause, R., 2006, “Seismic
Retrofit of Lap Splices in Nonductile Square Columns Using Carbon Fiber-
Reinforced Jackets,” ACI Structural Journal, V. 103, No. 6, Nov.-Dec.,
pp. 708-716.
Haroun, M. A., and Elsanadedy, H. M., 2005a, “Fiber-Reinforced Plastic
Jackets for Ductility Enhancement of Reinforced Concrete Bridge Columns
64
with Poor Lap-Splice Detailing,” Journal of Bridge Engineering, ASCE,
V. 10, No. 6, pp. 749-757. doi: 10.1061/(ASCE)1084-0702(2005)10:6(749)
Haroun, M. A., and Elsanadedy, H. M., 2005b, “Behavior of
Cyclically Loaded Squat Reinforced Concrete Bridge Columns
Upgraded with Advanced Composite-Material Jackets,” Journal of
Bridge Engineering, ASCE, V. 10, No. 6, pp. 741-748. doi: 10.1061/
(ASCE)1084-0702(2005)10:6(741)
Iacobucci, R. D.; Sheikh, S. A.; and Bayrak, O., 2003, “Retrofit of Square
Concrete Columns with Carbon Fiber-Reinforced Polymer for Seismic
Resistance,” ACI Structural Journal, V. 100, No. 6, Nov.-Dec., pp. 785-794.
Lam, L., and Teng, J., 2003a, “Design-Oriented Stress-Strain Model
for FRP-Confined Concrete,” Construction and Building Materials, V. 17,
No. 6-7, pp. 471-489. doi: 10.1016/S0950-0618(03)00045-X
Lam, L., and Teng, J., 2003b, “Design-Oriented Stress-Strain Model
for FRP-Confined Concrete in Rectangular Columns,” Journal of Rein-
forced Plastics and Composites, V. 22, No. 13, pp. 1149-1186. doi:
10.1177/0731684403035429
Li, Y.-F.; Hwang, J.-S.; Chen, S.-H.; and Hsieh, Y.-M., 2005, “A
Study of Reinforced Concrete Bridge Columns Retrofitted by Steel
Jackets,” Zhongguo Gongcheng Xuekan, V. 28, No. 2, pp. 319-328. doi:
10.1080/02533839.2005.9670997
Mander, J. B.; Priestley, M. J. N.; and Park, R., 1988, “Theoret-
ical Stress-Strain Model for Confined Concrete,” Journal of Struc-
tural Engineering, ASCE, V. 114, No. 8, pp. 1804-1826. doi: 10.1061/
(ASCE)0733-9445(1988)114:8(1804)
Memon, M. S., and Sheikh, S. A., 2005, “Seismic Resistance of Square
Concrete Columns Retrofitted with Glass Fiber-Reinforced Polymer,” ACI
Structural Journal, V. 102, No. 5, Sept.-Oct., pp. 774-783.
Ozcan, O.; Binici, B.; and Ozcebe, G., 2010, “Seismic Strengthening of
Rectangular Reinforced Concrete Columns Using Fiber Reinforced Poly-
mers,” Engineering Structures, V. 32, No. 4, pp. 964-973. doi: 10.1016/j.
engstruct.2009.12.021
Priestley, M. J. N.; Seible, F.; Xiao, Y.; and Verma, R., 1994a, “Steel
Jacket Retrofitting of Reinforced Concrete Bridge Columns for Enhanced
Shear Strength—Part 1: Theoretical Considerations and Test Design,” ACI
Structural Journal, V. 91, No. 4, July-Aug. pp. 394-405.
Priestley, M. J. N.; Seible, F.; Xiao, Y.; and Verma, R., 1994b, “Steel
Jacket Retrofitting of Reinforced Concrete Bridge Columns for Enhanced
Shear Strength Part 2: Test Results and Comparison with Theory,” ACI
Structural Journal, V. 91, No. 5, Sept.-Oct., pp. 537-551.
Seible, F.; Priestley, M. J. N.; Hegemier, G. A.; and Innamorato, D., 1997,
“Seismic Retrofit of RC Columns with Continuous Carbon Fiber Jackets,”
Journal of Composites for Construction, ASCE, V. 1, No. 2, pp. 52-62. doi:
10.1061/(ASCE)1090-0268(1997)1:2(52)
Wu, Y.-F.; Wang, L.; and Liu, T., 2008, “Experimental Investigation on
Seismic Retrofitting of Square RC Columns by Carbon FRP Sheet Confine-
ment Combined with Transverse Short Glass FRP Bars in Bored Holes,”
Journal of Composites for Construction, ASCE, V. 12, No. 1, pp. 53-60.
doi: 10.1061/(ASCE)1090-0268(2008)12:1(53)
Xiao, Y., and Ma, R., 1997, “Seismic Retrofit of RC Circular
Columns Using Prefabricated Composite Jacketing,” Journal of Struc-
tural Engineering, ASCE, V. 123, No. 10, pp. 1357-1364. doi: 10.1061/
(ASCE)0733-9445(1997)123:10(1357)
Xiao, Y., and Wu, H., 2003, “Retrofit of Reinforced Concrete
Columns Using Partially Stiffened Steel Jackets,” Journal of Struc-
tural Engineering, ASCE, V. 129, No. 6, pp. 725-732. doi: 10.1061/
(ASCE)0733-9445(2003)129:6(725)
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S05
Shear-Friction Strength of Low-Rise Walls with 550 MPa
(80 ksi) Reinforcing Bars under Cyclic Loading
by Jang-Woon Baek, Hong-Gun Park, Byung-Soo Lee, and Hyun-Mock Shin
In the thick reinforced concrete walls of nuclear power plants, the
use of high-strength reinforcing bars for shear design is necessary
to enhance constructibility and economy. However, low-rise walls
subjected to cyclic lateral loading are susceptible to sliding failure
at the construction joints. In the present study, low-rise walls were
tested to verify the applicability of 550 MPa (80 ksi) bars to the
shear-friction design. The major test parameters were the grade
of reinforcing bars, wall aspect ratio, reinforcement ratio, and
surface condition of the construction joint. The test results showed
that the specimens were susceptible to sliding failure and the stress
of 550 MPa (80 ksi) shear-friction bars was not reached to the
yield strength. Particularly, the shear-friction strengths under
cyclic loading were smaller than those under monotonic loading
reported in previous studies. The applicability of current design
methods was evaluated for the shear-friction design of walls with
550 MPa (80 ksi) bars.
Keywords: cyclic loading; high-strength reinforcing bars; low-rise shear
walls; nuclear power plant walls; shear-friction strength.
INTRODUCTION
In the construction of nuclear power plants
(NPPs), many large-diameter reinforcing bars are used for
thick reinforced concrete (RC) walls, which can cause bar
congestion. Such bar congestion significantly degrades the
constructability and economy of the structures. Recently,
to reduce the number of reinforcing bars, the use of high-
strength 550 MPa (80 ksi) bars for flexural reinforcement
and shear reinforcement has been studied.1-5 However,
under cyclic loading, squat walls with aspect ratio hw/lw ≤
0.5, which are commonly used for nuclear power plants,
are vulnerable to shear sliding at the construction joint.6,7
Thus, in the design of squat walls, frequently, the number of
vertical reinforcing bars is increased to prevent shear sliding.
Currently, the construction joint interfaces are designed
based on the shear-friction strength specified in Eq. (11-25) of
ACI 3498 (or Eq. (22.9.4.2) of ACI 3189). The shear-friction
strength was developed based on experimental and theo-
retical studies by Birkeland and Birkeland.10 However, the
majority of the specimens were tested using the push-off
test setup, which differs from the actual loading condition of
low-rise walls under earthquake loading in that: 1) flexural
moment as well as shear force is applied to the walls; and
2) repeated cyclic loading degrades the shear-friction resis-
tance of the walls. To address such effects, seismic design
code, Eurocode 8,11 specifies stricter requirements than those
of general design code, Eurocode 2.12 On the other hand, in
ACI 3498 (or ACI 3189), no special design requirements are
specified for cyclic loading.
ACI Structural Journal/January 2018
TECHNICAL PAPER
The effects of flexural moment and cyclic loading on
shear-friction strength have been investigated in previous
studies.13-15 Mattock et al.13 performed a series of push-off
tests for specimens with an interface subjected to flex-
ural moment or axial tension. In the Mattock’s tests, the
yield strength of the reinforcement was 345 to 365 MPa
(50.0 to 52.9 ksi). The test results showed that flexural
moment did not decrease the shear-friction strength. The
reason was reported in the commentaries of ACI 318-14
(R22.9.4.6)9: flexural compression stress increased the shear-
friction strength while flexural tension stress reduced the
shear-friction strength.
Wood14 proposed a lower-bound shear strength using
a shear-friction analogy based on the existing test results
of 143 low-rise wall specimens (yield strength of shear-
friction reinforcement across a shear plane fyv = 280 to
570 MPa [40.6 to 82.7 ksi]) subjected to monotonic, repeated,
and reversed cyclic loads. The proposed lower-bound shear
strength (0.25Av fyv, where Av is area of shear-friction rein-
forcement) is significantly smaller than the shear-friction
strength of ACI 318 and ACI 349 (μAv fyv; the coefficient of
shear-friction μ = 0.6 to 1.4).
Orakcal et al.15 tested wall specimens with an intention-
ally weakened joint at the wall-foundation interface, where
the concrete cross section was reduced and a part of the
vertical reinforcement was discontinued to cause shear-fric-
tion failure (fyv = 352 MPa [51.0 ksi]). They reported that the
nominal shear-friction strength of ACI 318 overestimated the
test strengths (the ratio of the maximum tested strength to
the shear-friction strength predicted by ACI 349 [Eq. (A1)]
Vtest/Vsf = 0.87 to 0.89), and the strength overestimation was
attributed to the effect of cyclic loading. On the other hand,
Bass et al.16 reported that the effect of cyclic loading on
shear-friction strength was negligible (fyv = 420 MPa [60 ksi],
concrete compressive strength fc′ = 18.6 to 34.7 MPa [2.69 to
5.03 ksi]). However, in the test specimens, the shear-friction
bar ratio ρv was relatively small (= 0.21 to 0.31%). Further, the
number of load cycles (only 10 load cycles) was smaller than
that of conventional loading protocol.
Harries et al.17 studied the effect of high-strength rein-
forcement. Push-off tests under monotonic loading
were performed for shear-friction specimens with normal
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2016-232.R1, doi: 10.14359/51700915, received February 20, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
65
(fyv = 424 to 464 MPa [61.5 to 67.3 ksi]) and high (fyv = 868
to 965 MPa [125.9 to 139.9 ksi]) strength bars. The inter-
face between old and new concrete was roughened with an
amplitude of at least 6.4 mm (0.25 in.) according to ACI
3498 or Eurocode 2.12 The test result showed that the high-
strength bars did not increase the shear-friction strength. The
ratio of the maximum tested strength to the shear-friction
strength predicted by the ACI 349 equations (Vtest/Vsf [Eq.
(A1)]) were 1.55 to 2.53 for the specimens with normal-
strength reinforcement, and 0.72 to 1.26 for the specimens
with high-strength reinforcement. However, in all speci-
mens, the stresses in the both normal-strength and high-
strength reinforcing bars did not reach to the yield strength.
The test results indicate that under monotonic loading,
concrete cohesion significantly contributes to the shear-fric-
tion strength of walls.
Kono et al.18 performed push-off tests under reversed
cyclic loading (overall six to eight cycles) to study the effect
of high-strength concrete (44.1 to 112 MPa [6.4 to 16.2 ksi]),
high-strength bars (386 to 999 MPa [56.0 to 144.9 ksi]), and
interface roughness (smooth, roughened, and monolithic
surface). Similar to the study by Harries et al.,17 even in
specimens with normal-strength shear-friction bars (386 to
432 MPa [56.0 to 62.6 ksi]), the stresses did not reach
the yield strength at the measured maximum strength of
the specimens. Vintzeleou and Tassios19 also reported the
strength and stiffness degradation of dowels under cyclic
deformation. Such degradation of maximum strength was
also found under repeated loading (not reversed loading):
Figueira et al.20 reported that the shear-friction strength
decreased under repeated loading.
As such, the existing studies reported that the use of
high-strength reinforcement (fyv = 868 to 999 MPa [125.9 to
144.9 ksi]) did not increase the shear-friction strength
of walls. However, the range of the yield strength is far
beyond the yield strength targeted in the present study,
and the number of shear-friction tests for walls with Grade
550 MPa (80 ksi) reinforcement are limited, particularly for
squat walls (hw/lw ≤ 0.5) with construction joints. Further-
more, the effects of cyclic loading and flexural moment on
shear-friction strength should be meticulously investigated
for the seismic design of squat walls.
In the present study, the effect of high-strength bars on the
shear-friction strength of low-rise walls subjected to cyclic
loading was studied. The test parameters were the bar grades
(Grade 420 and 550 MPa [60 and 80 ksi] bars), aspect ratio
hw/lw (0.5 and 0.33), reinforcement ratio, surface condition
of construction joint, presence of axial compression, and use
of additional shear-friction bars.
RESEARCH SIGNIFICANCE
In the thick RC walls with small aspect ratios of NPPs,
shear sliding failure along the construction joints is the
major failure mode under earthquake loading. Thus, for the
use of high-strength bars for NPP walls, it is crucial to verify
the effect of high-strength bars on the shear-friction strength.
The present study experimentally evaluated the validity
of high-strength bars (Grade 550 MPa [80 ksi]) for shear
66
friction design. Based on the test results, the relevant design
recommendations were provided.
TEST PLAN
Major test parameters and specimen details
In NPP walls, because of the high safety requirements,
the shear reinforcement ratio for the design is close to the
maximum shear reinforcement ratio specified by current
design codes. Thus, most of the specimens tested in the
present study were designed with the permissible maximum
shear reinforcement ratio (refer to Appendix B for the
maximum permissible ratio). Further, in the case of NPP
walls with the permissible maximum shear reinforcement
ratio, uniformly distributed vertical reinforcement, 42 MPa
(6 ksi) compressive strength concrete, a construction joint,
and without axial compression, the nominal shear-friction
strength predicted by the ACI equations becomes smaller
than the nominal flexural strength, approximately when
M/Vlw ≤ 0.75. Thus, to investigate the shear-sliding failure
strength before flexural yielding, aspect ratios of hw/lw ≤ 0.5
(M/Vlw ≤ 0.66) were used for the wall specimens.
Thirteen specimens were prepared for cyclic loading tests.
Figure 1 shows the dimensions and details of the specimens,
and Table 1 presents the design parameters. The dimensions
of walls with hw/lw = 0.5 were 1500 mm (length) x 750 mm
(height) x 200 mm (thickness) (59.10 x 29.55 x 7.88 in.). The
dimensions of walls with hw/lw = 0.33 were 1500 mm (length)
x 500 mm (height) x 200 mm (thickness) (59.10 x 19.70 x
7.88 in.) (Fig. 1 and Table 1). The test parameters were the
grade of reinforcing bars, wall aspect ratio, reinforcement
ratio, surface condition of the construction joint, and addi-
tional details (axial load or additional shear-friction bars).
The names of the specimens indicate the test parameters. The
first letters, H and N, refer to higher-strength deformed bars
(Grade 550 MPa [80 ksi]) and normal-strength deformed bars
(Grade 420 MPa [60 ksi]), respectively (refer to Table 2). The
numbers 0.5 and 0.33 refer to the aspect ratio of the spec-
imens. The second letter indicates the horizontal bar ratio:
M, H, and Q refer to the maximum, half, and quarter of the
permissible reinforcement ratio of horizontal bars, respec-
tively. (Refer to Appendix B for the maximum permissible
reinforcement ratio for horizontal bar). The third letter indi-
cates the surface condition of the construction joint: U, R, and
G refers to untreated surface, roughened surface, and grooved
surface, respectively. The last letter(s) following the dash (-)
indicate the additional details: C for axial compression load
(0.06 to 0.07Acfc′, where Ac is area of wall cross section),
AS for additional short shear-friction bars (bar length =
160 mm [6.30 in.]), and AL for additional long shear-friction
bars (300 mm [11.82 in.]). For example, H0.5MU-C indicates
a specimen with Grade 550 MPa (80 ksi) shear bars, an aspect
ratio of 0.5, maximum shear reinforcement ratio, untreated
surface in the construction joint, and axial compression
loading (0.06 to 0.07Ac fc′).
For Specimen H0.5MU, Grade 550 MPa (80 ksi) rein-
forcing bars and aspect ratio hw/lw = 0.5 were used. For
horizontal and vertical reinforcement, two layers of HD13
bars (actual yield strength fyh and fyv = 667 MPa [96.7 ksi])
were used (Fig. 1(b), Table 2). The horizontal bar ratio ρh
ACI Structural Journal/January 2018
Fig. 1—Dimensions and reinforcement details of test specimens. (Note: 1 MPa = 0.145 ksi; 1 mm = 0.0394 in.)
was 0.68% (Table 1), which corresponds to the maximum
shear bar ratio specified in ACI 349 (using Vs [Eq. (B3)] =
Vsmax [Eq. (B4)]). The vertical bar ratio ρv was 1.01%. The
nominal flexural strength Vf (that is, horizontal force corre-
sponding to the flexural moment capacity) was designed to
be equal to the nominal shear strength Vn (Eq. (B1) to (B3)).
A construction joint without surface treatment was located
at the interface between the wall and foundation slab. Thus,
the friction coefficient of the nominal shear-friction strength
μ was 0.6 (Eq. (A1)). All the vertical bars were assumed to
contribute to the nominal shear-friction strength Vsf, which
was close to the nominal flexural strength Vf (or the nominal
shear strength Vn) (Vsf/Vf = 0.97).
In Specimens H0.5HU and H0.5QU, to investigate the effect
of the bar ratio, smaller bar ratios were used (Fig. 1(b)): ρh =
0.38%, ρv = 0.60% for H0.5HU and ρh = 0.19%, ρv = 0.43%
for H0.5QU (Table 1). The nominal shear-friction strength
Vsf was smaller than the nominal flexural strength Vf (or the
nominal shear strength Vn) (H0.5HU: Vsf/Vf = 0.82, H0.5QU:
Vsf/Vf = 0.76).
In Specimen H0.5MU-C, the effect of compressive axial
force (0.07Ac fc′) was studied (Fig. 1(b)). As the nominal
ACI Structural Journal/January 2018
flexural strength was increased by the axial force, the
number of flexural bars was decreased when compared to
that of H0.5MU (ρv = 0.68%, ρh = 0.68%). However, the
nominal shear-friction strength Vsf was also increased by the
compression and was close to the nominal flexural strength
Vf (Vsf/Vf = 1.02).
In Specimen H0.5MG-C, to prevent sliding failure, the
wall-foundation interface was intentionally moved approx-
imately 50 mm (1.97 in.) to the inside of the foundation
(Fig. 1(a)). To achieve this purpose, pocket metals were
installed at the wall-foundation interface. After pouring
concrete for the foundation, a groove was formed by the
pocket metals. Thus, the wall-foundation interface in
H0.5MG-C was regarded as a monolithic joint: the friction
coefficient μ = 1.4. Thus, the nominal shear-friction strength
was significantly increased to 2921 kN (657.2 kip), but was
limited to 1919 kN (431.8 kip) (Eq. (A2a), ACI 318) or
1650 kN (371.3 kip) (Eq. (A2b), ACI 349) by the permissible
maximum value specified by ACI 318 or ACI 349 (Vsf/Vf =
1.49 or 1.28), respectively. The other details of H0.5MG-C
were the same as those of H0.5MU-C (Fig. 1(b)).
67
Table 1—Design parameters of test specimens
Additional shear-
Wall
friction reinforcing
aspect
Wall concrete Horizontal reinforcing bar Vertical reinforcing bar bar
ratio
hw/lw fc′, Number and ρhfyh, Number
Specimens (M/Vlw) MPa P, kN Age Vs/Vsmax type ρh, % MPa Number and type ρv, % and type ρsf, % Vsf /Vf
H0.5MU 42.1 0 21 1.04 8-HD13 0.68 4.51 24-HD13 1.01 — — 0.97
H0.5HU 42.1 0 20 0.55 8-HD10 0.38 2.38 12-HD13/4-HD10 0.60 — — 0.82
0.5
H0.5QU 42.1 0 20 0.27 4-HD13 0.19 1.19 8-HD13/4-HD10 0.43 — — 0.76
(0.66)
H0.5MU-C 42.1 882 17 1.04 8-HD13 0.68 4.51 16-HD13 0.68 — — 1.02
H0.5MG-C 38.7 735 51 1.09 8-HD13 0.68 4.51 16-HD13 0.68 — — 1.28
H0.33MU 42.1 0 24 1.03 10-HD10 0.71 4.46 18-HD13 0.76 — — 0.68
H0.33HU 42.1 0 22 0.62 6-HD10 0.43 2.67 12-HD13 0.51 — — 0.64
H0.33MU-AS 42.1 0 27 1.03 10-HD10 0.71 4.46 18-HD13 0.76 16-HD10 0.38 1.00
H0.33HU-AS 0.33 42.1 0 23 0.62 6-HD10 0.43 2.67 12-HD13 0.51 10-HD10 0.24 0.93
H0.33MU-AL (0.50) 45.5 0 21 0.99 10-HD10 0.71 4.46 18-HD13 0.76 16-HD10 0.38 1.00
N0.33MU 44.5 0 19 1.16 8-D13 1.01 5.17 16-D16 1.06 — — 0.67
N0.33MU-AL 44.5 0 19 1.16 8-D13 1.01 5.17 16-D16 1.06 14-D13 0.59 1.07
H0.33MR 44.0 0 20 1.01 10-HD10 0.71 4.46 18-HD13 0.76 — — 1.13

Notes: M/Vlw is shear span ratio; fc′ is compressive strength; P is axial compression; Vs is contribution of shear reinforcement (Eq. (B3)); Vsmax is maximum shear strength contrib-
uted by shear reinforcement in ACI 349 (Eq. (B4)); Vsf is shear-friction strength prediction (Eq. (A1) and Eq. (A2b)); Vf is flexural strength prediction; HD is Grade 550 MPa
(80 ksi) deformed bars; D is Grade 420 MPa (60 ksi) deformed bars (refer to Table 2); 1 kN = 0.225 kip; 1 MPa = 0.145 ksi.

Table 2—Properties of reinforcement U-type bars were respectively placed at the wall-foundation
interface (Fig. 1(c) and 1(d)). Consequently, the nominal
Type HD10 HD13 D13 D16
shear-friction strength was close to the nominal flex-
Grade 550 MPa (80 ksi) 420 MPa (60 ksi) ural strength (Vsf/Vf = 1.00 for H0.33MU-AS and 0.93 for
9.5 mm 12.7 mm 12.7 mm 15.9 mm H0.33HU-AS). For the U-type bars, HD10 (yield strength
Bar diameter
(No. 3) (No.4) (No.4) (No.5) of additional shear-friction reinforcement fysf = 625 MPa
Yield strength 625 667 510 470 [90.6 ksi]) was used (Table 2). The U-type shear-friction
fy, MPa (ksi) (90.6) (96.7) (74.0) (68.2) reinforcement was assumed to not contribute to the flexural
Notes: HD is Grade 550 MPa (80 ksi) deformed bars; D is Grade 420 MPa (60 ksi)
strength Vf. The length of the U-type bars from the wall-foun-
deformed bars; 1 mm = 0.0394 in. dation interface to the end of the hook was 160 mm (6.30 in.)
to satisfy the development length of the hook (Fig. 1(c)).
In the case of walls with hw/lw = 0.5, except H0.5HU and In Specimen H0.33MU-AL, the development length of the
H0.5QU, the nominal shear-friction strength Vsf was close U-type bars from the wall-foundation interface to the end
to the nominal flexural strength Vf. On the other hand, in the of the hook was increased to 300 mm (11.82 in.)—approxi-
case of walls with hw/lw = 0.33, the nominal flexural strength mately 30-bar diameter. For the development length of addi-
Vf for a given vertical reinforcing bar area was increased due tional shear-friction reinforcement (straight reinforcing bars
to the smaller wall height. Therefore, the nominal shear-friction without hook), Wasiewicz7 recommended the greater value
strength was smaller than the nominal flexural strength. of 15 bar diameter and wall thickness. The other details of
In Specimen H0.33MU, the ratio of the nominal H0.33MU-AL were the same as those of H0.33MU-AS.
shear-friction strength to the nominal flexural strength was To directly compare the effects of Grade 550 and 420 MPa
significantly small: Vsf/Vf = 0.68. Like H0.5MU, the permis- (80 and 60 ksi) bars, in Specimen N0.33MU, D13 and D16
sible maximum shear bar ratio was used for shear reinforce- bars [Grade 420 MPa (60 ksi)] were used for the horizontal
ment (Table 1, ρh = 0.71%, ρv = 0.76%). For horizontal and bars and vertical bars, respectively (Fig. 1(d), Tables 1 and 2).
vertical reinforcement, HD10 (fyh = 625 MPa [90.6 ksi]) and The horizontal bar ratio was increased to 1.01% as the yield
HD13 (fyv = 667 MPa [96.7 ksi]) were used, respectively. strength of bars was decreased. The effective bar strength
In H0.33HU, to study the effect of bar ratios, the bar ratios ρhfyh = 5.17 MPa (0.75 ksi) was slightly greater than that of
were decreased (ρh = 0.43%, ρv = 0.51%). H0.33MU (ρhfyh = 4.46 MPa [0.65 ksi]). The vertical bar ratio
In Specimens H0.33MU-AS and H0.33HU-AS, addi- was ρv = 1.06%. In N0.33MU-AL, U-type D13 bars (fysf =
tional shear-friction bars were placed at the wall-foundation 510 MPa [74.0 ksi], ρsf = 0.59%) were additionally placed at
interface so that the nominal shear-friction strength Vsf was the wall-foundation interface. Similar to H0.33MU-AL, the
close to the nominal flexural strength Vf. In H0.33MU-AS length of the U-type bars from the wall-foundation interface
and H0.33HU-AS, to increase only the shear-friction to the end of hook was 300 mm (11.82 in.). The other details
strength rather than the flexural strength, eight and five short of N0.33MU-AL were the same as those of N0.33MU.

68 ACI Structural Journal/January 2018

Table 3—Mixture design of concrete
Unit weight, kgf/m3
Maximum
Compressive strength fc′ w/c, % W C FS S FA CA SP Slump, mm aggregate size, mm
42 MPa 33.3 161 340 58 85 749 956 4.1 120 25

Notes: W is water; C is cement; FS is fly ash; S is blast-furnace slag; FA is fine aggregate; CA is coarse aggregate; SP is superplasticizer (high-range water-reducing admixture);
1 mm = 0.0394 in.; 1 m = 3.281 ft; 1 kgf = 2.204 lbf.

Fig. 3—Lateral loading protocol. (Note: 1 mm = 0.0394 in.)

Fig. 2—Test setup. (Note: 1 mm = 0.0394 in.). ment transducers (LVDTs) for the measurement of lateral
displacements, flexural deformations, shear deformations,
In Specimen H0.33MR, to satisfy the “roughened surface”
and sliding at the wall base. Figures 1(a), 1(c), and 1(e) show
specified by ACI 3498 and Eurocode 2,12 the interface
the strain gauges for the reinforcing bars.
was roughened with surface treatment of 6 mm (0.24 in.)
(Fig. 1(d) and (e)). For the roughened surface, the fric-
TEST RESULTS
tion coefficient μ was increased to 1.0. The other details of
Load-displacement relationships
H0.33MR were the same as those of H0.33MU.
Figure 4 shows the lateral load-displacement (lateral drift
In the test specimens, the concrete compressive strength
ratio) relationships of the test specimens. The lateral displace-
was 42.1 to 45.5 MPa (6.10 to 6.60 ksi), except for 38.7 MPa
ment indicates the net displacement (L1-L2 in Fig. 2) excluding
(5.61 ksi) of H0.5MG-C (Table 1). The ages of the concrete
sliding of the wall base (L2 in Fig. 2). Figure 4 also shows
at the time of testing are shown in Table 1. Table 3 presents
the nominal shear strength Vn, nominal flexural strength Vf,
the mixture design of concrete. In the small-scale test spec-
and nominal shear-friction strength Vsf predicted by ACI 349
imens, the wall depth was not sufficiently large to provide
(ACI 318). In the case of H0.33MU-AS, H0.33HU-AS,
good anchorage of the horizontal bars. Thus, 90-degree
H0.33MU-AL, and N0.33MU-AL, the nominal flexural
hooks were used at the ends of the horizontal bars. All the
strength Vf was calculated assuming that the additional
vertical bars and U-type shear-friction bars satisfied the
shear-friction bars do not contribute to the flexural strength.
anchorage length in the foundation slab. The concrete clear
Due to the relatively short length, the flexural contribution of
cover was 20 mm (0.79 in.).
the additional vertical reinforcing bars is limited only to the
wall bottom, and disappears in the area above the wall bottom.
Test procedure and instrumentation
In the calculation of the nominal shear-friction strength,
A cyclic lateral load was applied using the test setup
the limitation of bar yielding strength (420 MPa [60 ksi])
shown in Fig. 2. In the case of H0.5MU-C and H0.5MG-C,
specified in ACI 349 was not used. In all the specimens
an axial compressive load of approximately 0.06 to 0.07Ac fc′
except H0.5MG-C, H0.33MU-AL, and N0.33MU-AL, the
was applied to the top of the wall (882 kN [198.5 kip] in
maximum tested strength Vtest was not reached to the nominal
H0.5MU-C (42.1 MPa [6.10 ksi]), and 735 kN [165.4 kip]
flexural strength Vf. This result indicates that the maximum
in H0.5MG-C (38.1 MPa [5.52 ksi])). The level of the axial
tested strength Vtest was determined by shear sliding failure
compressive force was maintained during cyclic lateral
before flexural yielding (Fig. 4(a) to 4(d), 4(f) to 4(j), and
loading. Generally, the actual level of axial load ratio for
4(m)). On the other hand, in H0.5MG-C showing flexural
NPP walls ranged from 0.05 to 0.20Ac fc′. The axial load
yielding, the test strength was close to the flexural strength
level of 0.06 to 0.07Ac fc′ was determined: 1) to consider
(Fig. 4(e)). In H0.33MU-AL and N0.33MU-AL, with addi-
limitation of the axial load capacity of the testing actuator;
tional shear-friction reinforcing bars of 300 mm (11.82 in.), the
and 2) to compare the test results with the results of previous
test strength was reached to the nominal flexural strength
studies4,21 by maintaining the same axial load ratio.
(Fig. 4(k) and 4(l)). However, it should be noted that the
The lateral loading protocol in Fig. 3 followed the
nominal flexural strength was calculated neglecting the
“Acceptance Criteria for Special Precast Concrete Struc-
contributions of the additional vertical reinforcing bars.
tural Walls.”22 Figure 2 shows the linear variable displace-
Thus, the actual flexural strength could be greater than

ACI Structural Journal/January 2018 69

Fig. 4—Lateral load-displacement relationships of specimens. (Note: 1 mm = 0.0394 in.; 1 kN = 0.225 kip.)
the nominal flexural strength. The flexural strength Vf was
calculated by sectional analysis considering the effect of
axial load.
In Specimen H0.5MU (Fig. 4(a)) with HD13 bars (fyh and
fyv = 667 MPa [96.7 ksi], ρh = 0.68%, ρv = 1.01%), the peak
strength occurred at the drift ratio of +0.45 and –0.60%. The
maximum tested strength Vtest was +744 and –831 kN (+167.4
and –187.0 kip) in the positive and negative loading directions,
respectively, which was only 65% of the nominal shear-fric-
tion strength predicted by the ACI 349 (or ACI 318) equation;
the nominal shear-friction strength significantly overestimated
the tested strength. This is mainly because the yield strength
of Grade 550 MPa (80 ksi) reinforcement was assumed to be
valid for the nominal shear-friction strength. However, in fact,
in the current ACI 349 (or ACI 318) equation, the design yield
70
strength of shear friction is limited to 420 MPa [60 ksi] (refer
to the section, “Strains of reinforcing bars”).
In H0.5HU and H0.5QU (Fig. 4(b) and 4(c)) with smaller
bar ratios, the maximum tested strength Vtest decreased:
in the case of H0.5HU, +450 and –544 kN (+101.3 and
–122.4 kip) at the drift ratios of +0.35 and –0.60%, and in the
case of H0.5QU, +441 and –406 kN (+99.2 and –91.4 kip)
at the drift ratios of +0.28 and –0.60%. The maximum tested
strength was approximately proportional to the vertical rein-
forcing bar ratio. However, the ratio of the maximum tested
strength to the nominal shear-friction strength by ACI 349
(Vtest/Vsf) was increased to 0.70 and 0.83 in H0.5HU and
H0.5QU, respectively. This result indicates that the nominal
shear-friction strength predicted by ACI 349 and ACI 318
equations is overestimated in the case of higher bar ratios.
ACI Structural Journal/January 2018
Table 4—Summary of test results
Ratio of test strengths
Test results to predictions Estimated bar yield stress
Specimens Vtest,s, kN Vtest,g, kN Vtest, kN Drift ratio at Vtest, % Failure mode Vtest,s/Vsf Vtest,s/Vf fsf [= (Vtest,s – μP)/(μAvall)], MPa
H0.5MU 744 831 787 +0.45/–0.6 0.61 0.59 408
H0.5HU 450 544 497 +0.35/–0.6 0.63 0.52 370
SF
H0.5QU 406 442 423 +0.28/–0.6 0.79 0.60 445
H0.5MU-C 1087 1089 1087 +1.0/–1.0 0.81 0.83 458
H0.5MG-C 1368 1440 1404 +1.0/–1.0 CC+DT 0.83 1.06 —
H0.33MU 575 718 646 +0.35/–0.45 0.63 0.43 420
SF
H0.33HU 309 535 421 +0.45/–0.2 0.51 0.33 339
H0.33MU-AS 1031 1060 1045 +1.0/–0.6 0.77 0.77 502
SW
H0.33HU-AS 709 743 726 +1.0/–0.6 0.81 0.75 497
H0.33MU-AL 1344 1492 1418 +1.25/–1.25 1.00 1.00 655
SF
N0.33MU 969 1040 1004 +0.45/–0.6 1.08 0.72 508
N0.33MU-AL 1321 1531 1426 +1.0/–1.0 SW 0.92 0.99 444
H0.33MR 1132 1351 1242 +1.0/–1.0 SF 0.74 0.84 497

Notes: Vtest,s, Vtest,g, and Vtest are smaller, greater, and average values of measured maximum loads in positive and negative loading directions, respectively; Vsf is shear-friction
strength predictions by ACI 349 (Eq. (A1)); Vf is flexural strength predictions; SF is sliding failure at wall-foundation interface; CC is concrete crushing failure; DT is diagonal
tension failure; SW is sliding failure at midheight of wall; 1 kN = 0.225 kip; 1 MPa = 0.145 ksi.

In H0.5MU-C (Fig. 4(d)), subjected to axial compression,
the maximum tested strength Vtest and the corresponding
lateral drift significantly increased: +1088 and –1086 kN
(+244.8 and –244.4 kip) at the drift ratios of +1.0 and
–1.0%, respectively. Assuming that the vertical reinforcing
bar ratio of H0.5MU-C (ρv = 0.68%) was identical to that
of H0.5HU (ρv = 0.60%), the shear-friction coefficient μ
for axial compression can be estimated as approximately
0.67 [= (1087 – 497)kN/882 kN)], which coincides with the
coefficient for the smooth surface (not intentionally rough-
ened) specified in the ACI shear-friction equation (refer to
Appendix A). Nevertheless, the maximum tested strength
was not reached to the shear-friction strength: Vtest/Vsf = 0.81.
In H0.5MG-C (Fig. 4(e)), with a groove construction joint
and axial compression loading, the maximum tested strength
Vtest exceeded the nominal flexural strength Vf: +1368 and
–1440 kN (+307.8 and –324.0 kip) at the drift ratios of +1.0
and –1.0%, showing ductile behavior until the drift ratios of
+2.0 and –2.0%, respectively.
In H0.33MU (Fig. 4(f)) with aspect ratio hw/lw = 0.33 (fyh =
625 MPa [90.6 ksi], ρh = 0.71%, fyv = 667 MPa [96.7 ksi], ρv =
0.76%), the maximum tested strength was smaller than that of
H0.5MU: +574 and –718 kN (+129.2 and –161.6 kip) at the
drift ratios of +0.35 and –0.45%, respectively. In H0.33HU
(Fig. 4(g)), with half of the permissible maximum bar ratio,
the maximum tested strength Vtest decreased: +309 and
–534 kN (+69.5 and –120.2 kip) at the drift ratios of +0.45
and –0.20%, respectively. The ratios of the maximum tested
strength to the nominal shear-friction strength predicted by
ACI 349 equation were Vtest/Vsf = 0.71 and 0.69 in H0.33MU
and H0.33HU, respectively. In H0.33HU, the test strength
(309 kN [69.4 kip]) in the positive direction was excessively
less than the test strength (534 kN [120.1 kip]) in the negative
direction, though the other specimens also showed a similar
trend (refer to Table 4). The excessive difference in the test
strength may be caused by an unexpected loading condition
such as torsion, which occurs under the pull-push lateral
loading. Similar to the specimens with hw/lw = 0.5 (H0.5MU,
H0.5HU, and H0.5QU), the maximum tested strength was
proportional to the vertical reinforcing bar ratio, which indi-
cates that the shear-friction strengths were not affected by the
wall aspect ratios between hw/lw = 0.5 and 0.33.
In N0.33MU (Fig. 4(h)) with D13 and D16 bars (fyh =
510 MPa [74.0 ksi], ρh = 1.01%, fyv = 470 MPa [68.2 ksi],
ρv = 1.06%), unlike the walls with HD10 and HD13 bars,
the maximum tested strength was reached to the nominal
shear-friction strength Vsf: +1039 and –968 kN (+233.8 and
–217.8 kip) at the drift ratios of +0.45 and –0.60%, respec-
tively. After the peak load, as the lateral drift increased, the
load-carrying capacity gradually decreased.
In H0.33MU-AS and H0.33HU-AS (Fig. 4(i) and 4(j))
using HD10 additional shear-friction bars (fysf = 625 MPa
[90.6 ksi], ρsf = 0.38% in H0.33MU-AS, and ρsf = 0.24%
in H0.33HU-AS), the maximum tested strength Vtest and
the corresponding lateral drift increased: in H0.33MU-AS
(maximum shear bar ratio), +1060 and –1031 kN (+238.5
and –232.0 kip) at the drift ratios of +1.0 and –0.60%, and in
H0.33HU-AS (smaller shear bar ratio), +708 and –743 kN
(+159.3 and –167.2 kip) at the drift ratios of +1.0 and –0.60%,
respectively. However, the maximum tested strength did not
reach the nominal shear-friction strength predicted by the
ACI 349 equation: Vtest/Vsf = 0.78 and 0.83 in H0.33MU-AS
and H0.33HU-AS, respectively. On the other hand, in
H0.33MU-AL (fysf = 625 MPa [90.6 ksi], ρsf = 0.38%) and
N0.33MU-AL (fysf = 510 MPa [74.0 ksi], ρsf = 0.59%) with
the longer shear-friction bars (300 mm [11.82 in.]) (Fig. 4(k)
and 4(l)), the maximum tested strength was reached to the
nominal shear-friction strength: for H0.33MU-AL, Vtest/Vsf  =
1.06 (Vtest = +1491 and –1344 kN [+335.5 and –302.4 kip]
at +1.25 and 1.25%, respectively), and for N0.33MU-AL,

ACI Structural Journal/January 2018 71

Fig. 5—Damage modes of specimens at end of test.
Vtest/Vsf = 0.99 (Vtest = +1530 and –1321 kN [+344.3 and
–297.2 kip] at +1.0 and –1.0%, respectively). This result
indicates that the development length of the shear friction
reinforcing bars should be greater than the ordinary devel-
opment length specified by current design codes. This is
because at the bottom of walls subjected to cyclic loading,
wide and severe flexure-shear cracks degrade the bond
strength of the shear friction reinforcing bars. In the present
study, development length of a 30-bar diameter is recom-
mended for the U-type Grade 550 MPa [80 ksi] shear-fric-
tion reinforcement.
In H0.33MR (Fig. 4(m)), with a roughened surface, the
maximum tested strength was +1350 and –1132 kN (+303.8
and –254.7 kip), which was 1.92 times that of H0.33MU
with the smooth surface. Nevertheless, the maximum tested
strength did not reach the nominal shear-friction strength:
Vtest/Vsf = 0.82 (m = 1.0).
Failure modes
Figure 5 shows the damage modes of the specimens at the
end of the tests. Table 4 summarizes the failure modes. In
H0.5MU and H0.5MU-C, although the design shear-friction
strength was the same as the design flexural strength, sliding
failure occurred at the wall-foundation interface (Fig. 5(a), SF
mode). The failure mode was the same as that of H0.5HU,
H0.5QU, H0.33MU (Fig. 5(b)), H0.33HU, and N0.33MU
(Fig. 5(c)), which were intentionally designed for sliding
failure (Table 1). The test results showed that the failure mode
of the specimens was determined by shear sliding before flex-
ural yielding irrespective of the vertical reinforcing bar area,
because the increase of vertical reinforcing bars increased
the flexural capacity as well as shear-friction capacity. In
H0.33MR, with the roughened construction joint (μ = 1.0),
sliding failure occurred at the wall-foundation interface,
although the design shear-friction strength was greater than
the design flexural strength (Vsf /Vf = 1.13, Table 1).
In Specimens H0.33MU-AS, H0.33HU-AS, H0.33MU-AL,
and N0.33MU-AL, despite the use of additional shear-fric-
tion bars, sliding failure occurred. In H0.33MU-AS and
H0.33HU-AS, the sliding failure surface was formed along
the end of the shear-friction bars, which was located at a third
72
of the wall height (167 mm [6.58 in.] from the wall-foundation
interface) (Fig. 5(d), SW mode). Similarly, in N0.33MU-AL,
with additional D13 shear-friction bars of 300 mm (11.82 in.)
anchorage length, the failure surface was formed along three-
fifths of the wall height (Fig. 5(e)¸ SW mode). On the other
hand, in H0.33MU-AL with additional HD10 shear-friction
bars of 300 mm (11.82 in.) anchorage length, the failure
surface occurred at the wall-foundation interface (SF mode).
In H0.5MG-C, with a groove construction joint, sliding
failure did not occur, and a mixed failure mode of shear
and flexure occurred: crushing in the compression zone and
diagonal tension cracking in the tension zone (Fig. 5(f),
CC+DT mode).
Figure 6 shows the crack patterns of H0.33MU,
H0.33MU-AS, and H0.5MG-C, which represent the failure
modes of the wall-foundation interface (SF mode), wall
interface (SW mode), and combined flexure and shear
(CC+DT mode), respectively. In the SF mode specimens
(Fig. 6(a)), at the drift ratio of 0.2%, vertical splitting tensile
cracks occurred at the wall-foundation interface at the ends
of the wall section. Although diagonal cracking occurred in
the web, the number and width of the diagonal cracks did not
increase after 0.6% drift ratio. After the drift ratio of 1.0%,
the extensive vertical splitting tensile cracks led to spalling
of the concrete in the wall ends, and additional vertical split-
ting tensile cracks occurred in the web of the wall. Spalling
of the concrete at the wall-foundation interface became
severe, and ultimately, sliding failure occurred along the
wall-foundation interface (Fig. 5(b)). In H0.5MU, with an
aspect ratio of 0.5, vertical splitting tensile cracks occurred
earlier (0.2% drift ratio) in the web as well as at the ends of
the wall. In H0.5HU, H0.5QU, and H0.33HU, with smaller
bar ratios, sliding failure occurred earlier. The number of
cracks was fewer than that of H0.5MU, H0.33MU, and
N0.33MU, which had greater bar ratios. In H0.33MR, with
the roughened construction joint (μ = 1.0), the failure mode
was the same as that of H0.33MU, except that extensive
diagonal cracking also occurred.
In the SW mode specimens (Fig. 6(b)), diagonal cracking
occurred at the beginning of the test and spread across
the entire wall panel until the failure of the wall. Flexural
ACI Structural Journal/January 2018
Fig. 6—Crack propagations of specimens.
cracking was initiated at the height where the additional
shear-friction bars were terminated (160 mm [6.30 in.] from
the wall-foundation interface in the case of H0.33MU-AS).
As the load increased, the width of the flexural cracks
significantly increased. At the maximum load, a new hori-
zontal crack was formed across the midheight of the wall
and sliding failure occurred along the horizontal crack. In
the case of H0.33MU-AL, like other SW mode specimens,
flexural cracking occurred at the end of the shear-friction
bars. However, sliding occurred at the wall-foundation inter-
face (SF mode).
In the CC+DT mode specimen (Fig. 6(c)), similar to the
SW mode specimens, diagonal cracking gradually propa-
gated across the entire wall panel. As the drift ratio increased,
the width of the diagonal cracks increased, and after flexural
cracking, spalling of concrete and buckling of the vertical
bars occurred at the end of the walls. Ultimately, concrete
crushing and diagonal tensile cracking occurred in the
compression zone and the tension zone (Fig. 5(f)), respec-
tively. Unlike the SF mode specimens, vertical splitting did
not occur during the test.
Strains of reinforcing bars
Figure 7 shows the strain distributions of the vertical bars
along the wall length at 0.5Vtest, 0.75Vtest, and 1.0Vtest, where
Vtest indicates the maximum tested strength. The locations of
the bars and strain gauges are shown in Fig. 7. The strains
that were measured near the failure interface are presented.
In all specimens, the strains of the vertical bars and additional
shear-friction bars were approximately linearly distributed
at 0.5 to 0.75Vtest because of the flexural action (Fig. 7).
At 1.0Vtest in H0.5MU (Fig. 7(a) and 7(b)), the strains of
the vertical bars were still linear, and bar yielding did not
occur. This test result contradicts the assumption that all
ACI Structural Journal/January 2018
shear-friction bars yield as the peak load is attained, similar
to the current ACI 349 provisions for bars with yield strength
up to 420 MPa (60 ksi). In H0.5HU, H0.5QU, H0.5MU-C,
H0.33MU, and H0.33HU, with HD13 and/or HD10 bars
and without additional shear-friction bars, the strains of the
vertical bars were similar to those of H0.5MU.
On the other hand, in N0.33MU (Fig. 7(c) and 7(d)), with
D16 and D13 bars, bar yielding occurred in the web and
compression zone as well as in the tension zone at 1.0Vtest. In
N0.33MU-AL and H0.33MU-AL, with the longer (300 mm
[11.82 in.]) additional shear-friction bars (Fig. 7(e) to 7(h)),
the strains of vertical bars were similar to those of N0.33MU.
Thus, in the cases of N0.33MU-AL and H0.33MU-AL, with
additional shear-friction bars, the assumption of ACI 349
shear-friction method was valid, and the maximum tested
strength was reached to the design shear-friction strength:
Vtest/Vsf = 0.99 and 1.06 in N0.33MU-AL and H0.33MU-AL,
respectively (Table 4). However, in the case of H0.33MU-AS,
with the shorter additional shear-friction bars (160 mm
[6.30 in.]), the maximum tested strength did not reach the
design shear-friction strength despite the use of additional
shear-friction bars: Vtest/Vsf = 0.83. Similar to H0.5MU, the
strains of the vertical bars were linearly distributed, and bar
yielding did not occur in the vertical bars or the additional
bars (Fig. 7(i) and 7(j)).
Figure 8 shows the strain distributions of the horizontal
bars along the wall height (Fig. 1(a)) at 0.5Vtest, 0.75Vtest, and
1.0Vtest. In H0.5MG-C (Fig. 8(a)), without sliding failure, the
strains of the horizontal bars were greater than those of the
specimens with sliding failure (Fig. 8(b) and 8(c)). However,
yielding of the horizontal bars did not occur because of: 1)
the smaller contribution of horizontal bars in the low aspect
ratio (hw/lw = 0.5); and 2) the higher-yield strain (≈0.0033) of
the high-strength bars.
73
Fig. 7—Measured strains of vertical bars and shear-friction bars in specimens. (Note: 1 mm = 0.0394 in.; 1 kN = 0.225 kip.)
Fig. 8—Measured strains of vertical bars in specimens.
(Note: 1 mm = 0.0394 in.; 1 kN = 0.225 kip.)
In Specimens H0.5MU (Fig. 8(b)), H0.5HU, H0.5QU,
H0.5MU-C, H0.33MU (Fig. 8(c)), and H0.33HU, with
the shear-sliding failure mode, the strain of the horizontal
bars was smaller when the displacement at sliding failure
was greater. This result indicates that the shear force was
limited by the sliding failure at the wall-foundation inter-
face. In H0.33MU-AS, H0.33HU-AS, H0.33MU-AL
(Fig. 8(d)), and N0.33MU-AL, with additional shear-friction
bars, the strain of the horizontal bars was greater than that of
the specimens without additional shear-friction bars.
74
EVALUATION OF DESIGN STRENGTH OF
SHEAR-FRICTION BARS
Figures 9(a) to 9(d) show the ratios of the maximum tested
strengths to the predictions of ACI 349,8 Eurocode 2,12
Eurocode 8,11 and Harries et al.17 (Eq. (A1), (A3), (A4), and
(A5)). The test results included results of a previous study.18
In most existing test specimens,6,15,21,23,24 sliding failure
occurred after flexural yielding. Thus, only the test results by
Kono et al.,18 which showed sliding failure before flexural
yielding, were included. Specimen H0.5MG-C, showing
flexural yielding without sliding, was excluded. In the ACI
349 prediction, the maximum limit of yield strength, fyv ≤
420 MPa (60 ksi), was not used. Figures 9(a) to 9(d) show
that in the majority of the specimens, the peak strength did
not reach the design shear-friction strength due to the effects
of cyclic loading and high strength re-bars.
The bar stress fsf at sliding failure can be estimated by
applying the maximum tested strength Vtest,s to the nominal
shear-friction equation by ACI 349, Vsf: fsf = (Vtest,s – μP)/
(μAvall), where Avall is sum of bar area of vertical bars and
additional shear-friction bars (that is, Avall = Av + Asf), and
P is axial force (positive in compression). The maximum
tested strength is defined as the smaller of the two maximum
strengths in the positive and negative directions, Vtest,s
(Table 4). In the evaluation, Avall were assumed to contribute
to the shear-friction strength. The bar stresses fsf were esti-
mated as 339 to 655 MPa (49.2 to 95.0 ksi) (Table 4), except
H0.5MG-C with flexural yielding. In most of the specimens,
the estimated bar strengths did not reach the yield strength
of 550 MPa (80 ksi).
Figure 10 shows the relationships between the test
strengths Vtest,s and the area of overall vertical bars Avall (in
ACI Structural Journal/January 2018
Fig. 9—Comparison of test shear strengths and predictions of current design codes. (Note: 1 mm2 = 0.0016 in.2)
3. Regardless of the wall aspect ratios, the maximum
tested strength was proportional to the area of overall
vertical bars. This result indicates that the shear-friction
strength was not directly related to the magnitude of flex-
ural moment, as pointed out by Mattock et al.,13 because
flexural moment induces both compression and tension in a
cross-section, which have a positive and negative effect on
the shear-friction resistance, respectively. Thus, the effect of
flexural moment can be neglected.
From the trend line and the test results of Characteristics
1 to 3 listed previously, the shear-friction strength can be
defined as follows. For untreated interface (μ = 0.6)

Vsf2 = 250Avall = μ(420)Avall (μ = 0.6) (SI)

(1a)
Vsf2 = 36Avall = μ(60)Avall (μ = 0.6) (U.S. customary)

Fig. 10—Evaluation of design strength of shear-friction For yield strengths less than 420 MPa (60 ksi), clearly the
bars. (Note: 1 mm2 = 0.0016 in.2; 1 kN = 0.225 kip; 1 MPa bar stress cannot be greater than the yield strength. Further,
= 0.145 ksi.) the effect of compression force, if any, should be considered.
Thus, the shear-friction strength can be redefined as follows
the case of compression, the effective area = [Avallfyv + P]/
fyv = Avall + P/fyv). In Fig. 10, regardless of the steel grade of
Vsf2 = μ(Avallfyv + P) (fyv ≤ 420 MPa [60 ksi]) (1b)
the bars, the maximum tested strength was approximately
proportional to the bar area. In the case of the axial compres-
Equation (1b) is identical to the current ACI 349 equation.
sion (0.07Acfc′, H0.5MU-C) and surface treatment (μ = 1.0,
Particularly, the yield strength of vertical bars is limited to
H0.33MR), the maximum strength was greater. Exception-
420 MPa (60 ksi), as defined in the shear-friction design of
ally in the case of H0.33MU-AL with HD10 additional
ACI 349. A previous study4 reported that Grade 550 MPa
shear-friction bars of 300 mm (11.82 in.) anchorage length,
(80 ksi) bars are applicable to the flexural strength and shear
the maximum tested strength was significantly greater than
strength of walls. However, the test result in this study indi-
the proportional trend line.
cates that the design yield strength of bars for shear-friction
From the test results of the specimens with untreated
design should be limited to 420 MPa (60 ksi), although
interfaces (Fig. 10), the characteristics of the shear-friction
550 MPa (80 ksi) bars are used for flexural and shear design.
strength under cyclic loading can be summarized as follows:
Figure 9(e) shows the strength ratios of the test specimens
1. In the case of Grade 420 and 550 MPa (60 and 80 ksi)
predicted by Eq. (1), including the test results by Kono et
bars, the shear-friction strength was proportional to the area
al.18 In Fig. 9(e), the shear-friction strength by Eq. (1) safely
of overall vertical bars, showing a trend line with the slope of
predicted the test strength. The exception was H0.5HU and
250 MPa (36.3 ksi), regardless of the yield strength of bars.
H0.33HU, where the strength ratio Vtest,s/Vsf2 was 0.88 and
2. The trend line passes through the origin, which indi-
0.81, respectively. In the specimen, the maximum loads in
cates that the cohesion resistance was negligible due to the
the positive and negative directions significantly differed,
effect of cyclic loading.

ACI Structural Journal/January 2018 75

which indicates that an unexpected loading condition such
as torsion occurred under the pull-push lateral loading.
CONCLUSIONS
To investigate the shear-friction strength of walls with
Grade 550 MPa (80 ksi) bars, five wall specimens with hw/lw
= 0.5 and eight wall specimens with hw/lw = 0.33 were tested
under cyclic lateral loading. In most specimens, surface
treatment in the interface of the construction joint was not
used: μ = 0.6 (ACI 349). The major findings of the present
study are summarized as follows:
1. Sliding failure occurred in all specimens with a construc-
tion joint at the wall-foundation interface. In the speci-
mens with additional shear-friction bars (H0.33MU-AS,
H0.33HU-AS, and N0.33MU-AL), a new horizontal crack
was formed along the end of the additional bars and sliding
failure occurred along the horizontal crack. This result indi-
cates that the major failure mode of low-rise walls with
construction joints is shear sliding failure rather than flexural
yielding mode.
2. On the other hand, H0.5MG-C, which had a groove
construction joint at the interface, showed a typical flex-
ure-shear failure mode without shear sliding.
3. In most of the specimens, the maximum tested strength
was proportional to the vertical reinforcing bar ratios,
regardless of the reinforcement grade and wall aspect ratio
(hw/lw = 0.33 or 0.5). The effect of axial compression on
the shear-friction strength agrees with the prediction of the
ACI 349 shear-friction equation. By roughening the surface
at wall-foundation interface complying with the “rough-
ened surface” specified by ACI 349 and Eurocode 2, the
maximum strength was increased approximately as twice as
that of the specimen with a “smooth surface.”
4. In the specimens with Grade 550 MPa (80 ksi) bars
and sliding failure mode, the maximum tested strength
was smaller than the nominal shear-friction prediction by
ACI 349 (Vtest/Vsf = 0.65 to 0.83; yielding of Grade 550 MPa
[80 ksi] bars was assumed to be valid). On the other hand,
in the specimens with Grade 420 MPa (60 ksi) bars, the
maximum tested strength was close to or greater than the
nominal shear-friction prediction by ACI 349 (Vtest/Vsf = 0.99
to 1.06).
5. In the specimens with additional shear-friction bars of
a long anchorage length, the strength ratios were greater
than those using the short anchorage length. The test result
indicates that the development length of the shear friction
reinforcing bars should be greater than the ordinary devel-
opment length specified by current design codes, because
in walls subjected to cyclic loading, wide and severe flex-
ure-shear cracks degrade the bond strength of the shear
friction reinforcing bars. In the present study, development
length of 30-bar diameter is recommended for the U-type
Grade 550 MPa (80 ksi) shear-friction reinforcement.
6. The shear-friction strengths of the test specimens with
a construction joint were proportional to μAvallfs (μ = 0.6,
fs = 420 MPa [60 ksi]), regardless of the yield strength of
bars and wall aspect ratio. The result indicates that: 1) the
design yield stresses of shear-friction bars should be limited
76
to 420 MPa (60 ksi), as specified in ACI 349 (or ACI 318);
2) the effect of flexural moment can be neglected, as pointed
out by Mattock et al.13; and 3) due to the effect of cyclic
loading, the cohesion resistance is negligible.
7. In low-rise walls with construction joints, the safety
margin of shear-friction strength is smaller than that of
flexural strength. Further, the use of additional vertical bars
increases the flexural strength as well as the shear-fric-
tion strength. Thus, the failure mode of shear walls with
construction joints is expected to be determined by shear-
sliding before flexural yielding.
AUTHOR BIOS
Jang-Woon Baek is a Researcher at the Institute of Engineering Research
at Seoul National University, Seoul, South Korea. He received his BE, MS,
and PhD from the Department of Architecture & Architectural Engineering
at Seoul National University.
ACI member Hong-Gun Park is a Professor in the Department of Architec-
ture & Architectural Engineering at Seoul National University. He received
his BE and MS in architectural engineering from Seoul National University
and his PhD in civil engineering from the University of Texas at Austin,
Austin, TX. His research interests include inelastic analysis and the seismic
design of reinforced concrete structures.
Byung-Soo Lee is a Researcher in the Plant Construction & Engineering
Laboratory at Central Research Institute of Korea Hydro & Nuclear Power
Co., Ltd., Gueongju, South Korea. He received his BE in architectural engi-
neering from Kyung Hee University, Seoul, South Korea.
Hyun-Mock Shin is a Professor in the Department of Civil and Environ-
mental Engineering at Sungkyunkwan University, Seoul, South Korea. He
received his BE and MS in civil engineering from Seoul National Univer-
sity and his PhD in civil engineering from the University of Tokyo, Tokyo,
Japan. His research interests include nonlinear structural analysis and the
seismic design of reinforced concrete structures.
ACKNOWLEDGMENTS
This work was supported by Korea Hydro & Nuclear Power Co. Ltd
(No. L16S140000) and by the Nuclear Power Core Technology Develop-
ment Program of the Korea Institute of Energy Technology Evaluation and
Planning (KETEP), granted financial resource from the Ministry of Trade,
Industry & Energy, Republic of Korea (No. 2014151010169B). The Insti-
tute of Engineering Research at Seoul National University also provided
research facilities for this work.
REFERENCES
1. Cheng, M.-Y.; Hung, S.-C.; Lequesne, R. D.; and Lepage, A., “Earth-
quake-Resistant Squat Walls Reinforced with High-Strength Steel,” ACI
Structural Journal, V. 113, No. 5, Sept.-Oct. 2016, pp. 1065-1076. doi:
10.14359/51688825
2. Kabeyasawa, T., and Hiraishi, H., “Tests and Analyses of High-
Strength Reinforced Concrete Shear Walls in Japan,” High-Strength
Concrete in Seismic Regions, SP-176, C. French and M. Kreger, eds., Amer-
ican Concrete Institute, Farmington Hills, MI, 1998, pp. 281-310.
3. Maier, J., and Thürlimann, B., “Bruchversuche an Stahlbetonscheiben
(in German),” Institut für Baustatik und Konstruktion, Eidgenössische
Technische Hochschule (ETH) Zürich, Zürich, Switzerland, 1985, 130 pp.
4. Park, H. G.; Baek, J. W.; Lee, J. H.; and Shin, H. M., “Cyclic Loading
Tests for Shear Strength of Low-Rise Reinforced Concrete Walls with
Grade 550 MPa Bars,” ACI Structural Journal, V. 112, No. 3, May-June
2015, pp. 299-310. doi: 10.14359/51687406
5. Saito, H.; Kikuchi, R.; Kanechika, M.; and Okamoto, K., “Experi-
mental Study on the Effect of Concrete Strength on Shear Wall Behavior,”
Proceedings of the Tenth International Conference on Structural Mechanics
in Reactor Technology, V. H08/05, Anaheim, CA, 1989, pp. 227-232.
6. Paulay, T.; Priestley, M.; and Synge, A., “Ductility in Earthquake
Resisting Squat Shearwalls,” ACI Journal Proceedings, V. 79, No. 4, Apr.
1982, pp. 257-269.
7. Wasiewicz, Z. F., “Sliding Shear in Low Rise Shear Walls under
Lateral Load Reversals,” master’s thesis, University of Ottawa, Ottawa,
ON, Canada, 1988, 127 pp.
ACI Structural Journal/January 2018
 8. ACI Committee 349, “Code Requirements for Nuclear Safety-Related d) Concrete anchored to as-rolled structural steel by
Concrete Structures (ACI 349-13) and Commentary,” American Concrete
Institute, Farmington Hills, MI, 2013, 196 pp.
headed studs or by reinforcing bars: μ = 0.7.
9. ACI Committee 318, “Building Code Requirements for Structural In ACI 318, the permissible maximum shear-friction strength
Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American is specified as follows: for normalweight concrete either placed
Concrete Institute, Farmington Hills, MI, 2014, 520 pp.
10. Birkeland, P. W., and Birkeland, H. W., “Connections in Precast
monolithically or placed against hardened concrete with clean
Concrete Construction,” ACI Journal Proceedings, V. 63, No. 3, Mar. 1966, surface free of laitance and intentionally roughened to a full
pp. 345-368. amplitude of approximately 6 mm (0.25 in.)
11. British Standards Institution, “Eurocode 8: Design of Structures for
Earthquake Resistance,” European Committee for Standardization, Brus-
sels, Belgium, 2003, 229 pp. Vsf,max = min[0.2fc′Acv, (3.3 + 0.08fc′)Acv, 11Acv] (SI) (A2a)
12. British Standards Institution, “Eurocode 2: Design of Concrete
Structures – Part 1-1: General Rules and Rules for Building (EN-1992-1-
1:2004:E),” European Committee for Standardization, Brussels, Belgium,
For all other cases
2004, 225 pp.
13. Mattock, A. H.; Johal, L.; and Chow, H. C., “Shear Transfer in Rein- Vsf,max = min[0.2fc′Acv, 5.5Acv] (SI) (A2b)
forced Concrete with Moment or Tension Acting across the Shear Plane,”
PCI Journal, V. 20, No. 4, 1975, pp. 76-93.
14. Wood, S. L., “Shear Strength of Low-Rise Reinforced Concrete where fc′ is cylinder strength of concrete (MPa); and Acv is
Walls,” ACI Structural Journal, V. 87, No. 1, Jan.-Feb. 1990, pp. 99-107. total sectional area of the shear interface (mm2). The yield
15. Orakcal, K.; Massone, L. M.; and Wallace, J. W., “Shear Strength
of Lightly Reinforced Wall Piers and Spandrels,” ACI Structural Journal,
strength of shear-friction bars is limited to 420 MPa (60 ksi)
V. 106, No. 4, July-Aug. 2009, pp. 455-465. in both ACI 3189 and ACI 349.8
16. Bass, R. A.; Carrasquillo, R. L.; and Jirsa, J. O., “Shear Transfer In ACI 349, for conservative design of NPP walls, the
across New and Existing Concrete Interfaces,” ACI Structural Journal,
V. 86, No. 4, July-Aug. 1989, pp. 384-394.
permissible maximum shear-friction strength is specified by
17. Harries, K. A.; Zeno, G.; and Shahrooz, B., “Toward an Improved using Eq. (A2b) for all cases.
Understanding of Shear-Friction Behavior,” ACI Structural Journal,
V. 109, No. 6, Nov.-Dec. 2012, pp. 835-844.
18. Kono, S.; Tanaka, H.; and Watanabe, F., “Interface Shear Transfer for
Eurocode 2
High Strength Concrete and High Strength Shear Friction Reinforcement,” In Eurocode 2,12 the shear strength at an interface between
Proceedings of High Performance Materials in Bridges, 2003, pp. 319-328. concrete cast at different times is defined as the sum of
19. Vintzeleou, E., and Tassios, T., “Behavior of Dowels under Cyclic Defor-
mations,” ACI Structural Journal, V. 84, No. 1, Jan.-Feb. 1987, pp. 18-30.
concrete cohesion and friction, as follows
20. Figueira, D.; Sousa, C.; Calçada, R.; and Neves, A. S., “Push-off
Tests in the Study of Cyclic Behavior of Interfaces between Concretes Cast VEC2 = cAcv fct + μf(Avf fyv + P) ≤ 0.5vfc′ (SI) (A3)
at Different Times,” Journal of Structural Engineering, ASCE, V. 142,
No. 1, 2016, 04015101. doi: 10.1061/(ASCE)ST.1943-541X.0001364
21. Salonikios, T. N.; Kappos, A. J.; Tegos, I. A.; and Penelis, G. G., where fct is tensile strength of concrete (MPa); v is strength
“Cyclic Load Behavior of Low-slenderness Reinforced Concrete Walls: reduction factor; and c and μf are factors that depend on the
Design Basis and Test Results,” ACI Structural Journal, V. 96, No. 4,
July-Aug. 1999, pp. 649-661.
roughness of the interface as follows.
22. Hawkins, N. M., and Ghosh, S. K., “Acceptance Criteria for Special a) Indented: A surface with indentations complying with
Precast Concrete Structural Walls Based on Validation Testing,” PCI Fig. 6.9 in Eurocode 2: c = 0.5 and μf = 0.9
Journal, V. 49, No. 5, 2004, pp. 78-92. doi: 10.15554/pcij.09012004.78.92
23. Greifenhagen, C., and Lestuzzi, P., “Static Cyclic Tests on Lightly
b) Rough: A surface with at least 3 mm (0.12 in.)
Reinforced Concrete Shear Walls,” Engineering Structures, V. 27, No. 11, roughness at about 40 mm (1.58 in.) spacing, achieved by
2005, pp. 1703-1712. doi: 10.1016/j.engstruct.2005.06.008 raking, exposing of aggregate or other methods giving an
24. Athanasopoulou, A., and Parra-Montesinos, G., “Experimental Study
on the Seismic Behavior of High-Performance Fiber-Reinforced Concrete
equivalent behavior: c = 0.40 and μf = 0.7
Low-Rise Walls,” ACI Structural Journal, V. 110, No. 5, Sept.-Oct. 2013, c) Smooth: A slipformed or extruded surface, or a free
pp. 767-777. surface left without further treatment after vibration: c =
0.20 and μf = 0.6
APPENDIX A: SHEAR-FRICTION STRENGTHS OF d) Very smooth: A surface cast against steel, plastic, or
CURRENT DESIGN CODES specially prepared wooden molds: c = 0.025 to 0.10 and
ACI 349 and ACI 318 μf = 0.5.
In ACI 3498 (or ACI 3189), the shear-friction strength of a Generally, the shear-friction predictions by Eq. (A3) are
construction interface is defined as follows greater than those by Eq. (A1) because of the inclusion of the
concrete cohesion term.
Vsf = μ(Avf fyv + P) (A1)
Eurocode 8
where Avf is area of shear-friction reinforcement across the In Eurocode 8,11 the shear resistance against sliding is
shear plane (mm2); fyv is yield strength of the shear-friction defined as the sum of the dowel resistance of vertical bars
reinforcement (MPa); P is applied axial compression force (Vdd), shear resistance of inclined bars (Vid) (at an angle φ to
(N); and μ is friction coefficient defined as follows. the potential sliding plane—for example, construction joint),
a) Concrete placed monolithically: μ = 1.4 and friction resistance (Vfd)
b) Concrete placed against hardened concrete with surface
intentionally roughened: μ = 1.0 VEC8 = Vdd + Vid + Vfd (A4a)
c) Concrete placed against hardened concrete not
intentionally roughened: μ = 0.6 (
Vdd = min 1.3∑ Asj f c′ f yv , 0.25 f yv ∑ Asj ) (SI) (A4b)

ACI Structural Journal/January 2018 77

 Vid = ∑Asi fyvcosφ (A4c)
Vfd = min(μf[(∑Asj fyv + Ned)ξ + MEd/z], 0.5η fc′ξlwbw) (SI) of the wall (mm); d is distance from the extreme compression
(A4d)
where ∑Asj is the sum of the areas of the vertical bars of
the web and of additional bars arranged in the boundary
elements specifically for resistance against sliding (mm2);
∑Asj is the sum of the areas of all inclined bars in both
directions (mm2); φ is an angle to the potential sliding plane
(rad); μf is concrete-to-concrete friction coefficient under
cyclic actions, which may be assumed equal to 0.6 for
smooth interfaces and to 0.7 for rough interfaces, as defined
in Eurocode 2; Ned and Med are the design axial load (N)
and design flexural moment (N·mm), respectively, obtained
from the analysis for seismic design; ξ is normalized neutral
axis depth; z is length of the internal lever arm (mm); η =
0.6(1 – fc′/250); lw is length of wall (mm); and bw is depth of
wall (mm). In Eurocode 2 and Eurocode 8, the coefficients
of c and μ for concrete monolithically cast are not defined.
Harries’s equation
Harries et al.17 studied the effect of high-strength shear-
friction bars, and proposed a shear-friction strength as follows
VHarries = cAcv fc′ + 0.002AvfEs ≤ 0.20Acvfc′ (SI) (A5)
where α = 0.075, 0.040, and 0 for the interface with
monolithically cast uncracked, cold-jointed, and monolithically
cast precracked conditions, respectively; and Es is elastic
modulus of vertical bars (MPa). In Eq. (A5), the shear-friction
strength is determined by the elastic modulus Es of the shear-
friction bars, regardless of the yield strength.
APPENDIX B: MAXIMUM SHEAR
REINFORCEMENT RATIO
In NPP walls, because of the high safety requirements,
the shear reinforcement ratio for the design is close to the
maximum shear reinforcement ratio specified by current
design codes. Thus, most of the specimens tested in the
present study were designed with the permissible maximum
shear reinforcement ratio.
In ACI 3498 (or ACI 3189), the shear strength of walls is
defined as the sum of the contributions of concrete Vc and
shear reinforcement Vs
Vn = Vc + Vs (SI) (B1)
Vc = 0.17 f c′hd (SI) (B2a)
Vc = 0.27 f c′hd + N u d /4lw (SI) (B2b)
 l (0.1 f c′ + 0.2 N u /(lw h)) 
Vc = 0.05 f c′ + w  hd (SI) (B2c)
 ( M u /Vu − lw /2) 
Vs = As fyhd/s (B3)
78
where fc′ is cylinder compressive strength of concrete
(MPa); lw is overall length of the wall (mm); h is thickness
fiber to the centroid of longitudinal tension reinforcement
(mm) (= 0.8lw in ACI 349); Nu is axial force (positive sign in
compression) [N]; Mu is applied flexural moment (N·mm); Vu
is applied shear force (N); As is area of shear reinforcement
(mm2) within spacing s (mm); and fyh is specified yield strength
of web shear reinforcement (MPa). Vn in Eq. (B1) shall not
exceed the permissible maximum shear strength 0.83√fc′hd
(Vn,max). In ACI 349 (or ACI 318), the concrete shear strength
Vc of walls is determined as the smaller of the values of
Eq. (B2b) and Eq. (B2c) unless Eq. (B2a) is used. Eq. (B2a)
is generally regarded as the basic concrete shear strength of
flexural members including squat walls. Thus, in the present
study, for simplicity, Eq. (B2a) was used for the concrete shear
strength. In this case, the permissible maximum shear strength
of horizontal web reinforcing bars is defined as follows
Vsmax = 0.83 f c′hd − Vc = 0.66 f c′hd (SI) (B4)
On the other hand, in the seismic provision of ACI 349
(ACI 318), the shear strength of walls is specified as follows
Vseis = Acv (α c f c′ + ρh f yh ) (B5)
The first term on the right-hand side indicates Vc, and
the second term indicates Vs. In Eq. (B5), Acv is the total
sectional area, and the coefficient αc is 0.25 (3.0 in U.S.
customary units) for hw/lw ≤ 1.5.
In the seismic provision, the permissible maximum shear
strength is specified as follows
Vseis,max = 0.66 f c′Acv (SI) (B6a)
In the case of walls with a rectangular cross-section, Acv =
hlw and d = 0.8lw (ACI 349 or ACI 318). Thus, the permissible
maximum shear strength of walls with a rectangular cross
section can be refined as follows
Vseis,max = 0.66 f c′Acv = 0.66 f c′hlw = 0.825 f c′hd (SI)
(B6b)
From Eq. (B5) and (B6), the permissible maximum shear
strength of horizontal web reinforcing bars for hw/lw ≤ 1.5 is
defined as follows
Vsmax = (0.66 − 0.25) f c′Acv = 0.5125 f c′hd (SI) (B7)
Comparing Eq. (B4) and Eq. (B7), the greater value
of Vsmax in Eq. (B4) was used for the design of the test
specimens. Substituting Vsmax in Eq. (B4) into Vs in Eq. (B3),
the permissible maximum horizontal reinforcing bar ratio ρh
was calculated. The vertical web reinforcing bar ratio ρv was
determined to be the same as ρh according to the seismic
provisions of ACI 349 (or ACI 318): ρv ≥ ρh unless the aspect
ratio of walls is greater than 2.0.
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S06
Behavior of Straight and T-Headed ASTM A1035/A1035M
Bar Splices in Flexural Members
by Sergio F. Breña, Jeffrey Messier, and Sean W. Peterfreund
Interest has increased in recent years in the use of high-strength
reinforcement in structural concrete members. This paper discusses
the behavior of flexural elements that contain coupled or lapped
high-strength reinforcement conforming to ASTM A1035/A1035M.
Behavior of elements containing couplers or spliced straight and
T-headed bars are included, adding to the number of test results
available for broadening the application of high-strength reinforce-
ment in structural concrete. In addition to general observations on
the behavior of specimens tested in this experimental program,
information on measured strain distributions along the splices
are provided from which key force-transfer mechanisms between
bars are identified. The benefits of having transverse reinforcement
within the splice of T-headed are highlighted.
Keywords: high-strength reinforcement; lap splices; T-headed reinforcement.
INTRODUCTION
Interest has increased in recent years in the use of high-
strength reinforcement in structural concrete to facilitate
construction operations and decrease congestion in heavily
reinforced concrete members (ACI Innovation Task Group 6
2010). Limitations on the use of high-strength reinforcing
bars conforming to ASTM A1035/A1035M (Grades 100
and 120) are placed in the ACI 318-14 building code (ACI
Committee 318 2014), both in type of application and
usable strength. ASTM A1035/A1035M bars do not exhibit
a well-defined yield plateau so yield strength is commonly
determined using the 0.2% offset method. These bars typi-
cally have a 0.2% offset yield strength that exceeds 100 ksi
(690 MPa). However, ACI 318-14 places limits on the
stress that can be used when designing with high-strength
reinforcement because of limited availability of test data.
For example, ACI 318-14 sets an upper value to the yield
strength of spiral reinforcement for confinement to 100 ksi
(690 MPa), and lower stress limits are imposed when using
high-strength reinforcement in other applications.
The use of high-strength reinforcement in flexural appli-
cations has been limited to date primarily because of lack of
design guidance for its use. Given the higher yield stress of
high-strength reinforcement, cracks forming at service load
levels could be wider than those that form when normal-
strength steel is used, raising concerns about ingress of water
and corrosive agents into the concrete element. Furthermore,
the longer embedment required to develop higher tensile
stresses might cause reinforcement congestion and generate
designs that become impractical. Therefore, methods to that
allow high-strength reinforcement to develop its strength
without creating congestion are needed.
In response to the need to increase the use of high-strength
reinforcement in structural concrete design applications,
ACI Structural Journal/January 2018
TECHNICAL PAPER
several documents have been developed that identify key
issues that need to be considered and those that need further
study (ACI Innovation Task Group 6 2010; Applied Tech-
nology Council 2014). For flexural applications, the ACI
ITG-6R-10 design guide (ACI Innovation Task Group 6 2010)
provides key information that should be considered for design
of beams containing ASTM A1035/A1035M reinforcement.
Pertaining to straight development length of ASTM A1035/
A1035M bars in tension, this document recommends using
Eq. (4-11a) in ACI 408R-03 (Joint ACI-ASCE Committee
408 2003) (rewritten as follows with factors for bar location,
epoxy coating, and lightweight concrete set equal to 1.0) with
a strength reduction factor ϕ equal to 0.80 instead of 0.82
 fy 
ld  f ′1/ 4 − φ2400ω
= c
(1)
db  cω + K tr 
φ76.3  
 d b
The yield strength fy used in this equation should be
established following the 0.2% offset method (refer to the
Notation section for definition of variables). Development
lengths obtained using this equation, however, can be signif-
icant, particularly for large-diameter bars.
T-headed reinforcement may be used to decrease the length
required to develop high-strength reinforcement, thereby
addressing some of the concerns associated with the long
development lengths of straight bars. Current provisions
in ACI 318-14 Section 25.4.4.2 are limited to reinforce-
ment with a nominal yield strength not exceeding 60 ksi
(420 MPa) because the experiments upon which the provi-
sions are based were limited to this reinforcement grade.
Furthermore, a clear spacing between adjacent bars less
than 4db is currently not allowed in ACI 318-14 because
of concerns that a change in the force-transfer mechanisms
between spliced bars might affect splice strength. This
spacing limit prevents T-heads from contacting adjacent bars
and requires a wider space to accommodate splices. Exper-
imental data are needed to broaden current design provi-
sions for spliced T-headed reinforcement to applications that
involve high-strength reinforcement.
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2016-298.R1, doi: 10.14359/51700782, received February 20, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
79
Fig. 1—Geometry and reinforcement of Group L specimens.
RESEARCH SIGNIFICANCE
This research provides needed data on the behavior of
flexural elements that contain ASTM A1035/A1035M bars
as the main tensile reinforcement. Laboratory experiments
focused on investigating the splice behavior of bars with
couplers, straight lap splices, and splices between T-headed
high-strength bars. The findings from this research will
complement existing and future investigations that may be
used to develop design provisions to increase use of high-
strength bars in a wider variety of applications than those
currently permitted in building and bridge codes.
EXPERIMENTAL PROGRAM
Specimen description
Laboratory tests in this research project included 15
beams containing ASTM A1035/A1035M longitudinal rein-
forcement on the tension side of the beams spliced in the
region of maximum moment. The specimens were organized
in three groups, depending on splice type. Six specimens
had straight-bar lap splices, each with a different overlap-
ping length (Group L). Reinforcement in eight of the nine
remaining specimens was spliced using threaded couplers or
T-heads (Groups C and T, respectively). One specimen was
built with continuous tension bars as a control for specimens
using No. 5 bars in Groups C and T.
Specimen notation consists of three groups of alphanumeric
characters separated by dashes. The first group denotes the
specimen group (L, C, or T), the second group of characters
denotes the bar size (No. 4, 5, or 6), and the last character
group denotes the coupler manufacturer (M or E), or the length
of the splice (in.) and presence of transverse reinforcement
along the splice (R). For example, Specimen T-5-12 denotes
a specimen in Group T (T-headed splice) with two No. 5 bars
as main tension reinforcement and a 12 in. (305 mm) splice.
Geometry and reinforcement of the specimens is illus-
trated in Fig. 1. The non-contact spliced bars in Group L
specimens had a transverse clear spacing of one bar diameter
db and a minimum of 2db between the groups of longitudinal
bars being spliced. Splices in specimens with T-headed rein-
forcement were constructed so that the head of the tee was
in contact with the adjacent bar being spliced. A clear cover
80
Fig. 2—Geometry and reinforcement of Groups C and T
specimens.
of 1.5 in. (38 mm) around all reinforcing bars was provided
in all specimens.
Specimens in Group L were longer than specimens in
Groups C and T in order to fit the required lap splice length
within the maximum moment region of the beams. Splicing
bars with mechanical couplers or T-heads allowed short-
ening of the maximum moment region of the beams and,
therefore, the overall length of these specimens was shorter
(Groups C and T).
A four-point testing configuration was used to generate
a constant moment region in the center of the beams. The
beams were tested in an inverted position (tension on top)
to facilitate observation and mapping of the progression of
cracking throughout the tests. Beams were supported using
rollers placed at the end of the constant moment region, and
loading was applied in the downward direction near the end
of the beams using hydraulic jacks. Bar splices were located
within the center region of the beams (refer to Fig. 1 and 2).
Specimen geometry and reinforcement are shown in Fig. 1
for Group L specimens and Fig. 2 for Groups C and T speci-
mens. Specimen dimensions are summarized in Table 1. The
span length of Group L specimens was 168 in. (4200 mm),
which included a length of the constant moment region equal
to 68 in. (1700 mm). The span length in Group T and C spec-
imens was 114 in. (2900 mm) with a length of the constant
moment region equal to 30 in. (760 mm). Cross-sectional
dimensions of Group L specimens were 13 in. wide by
9 in. deep (330 x 229 mm); those of Group C specimens were
either 11 or 9 in. wide (279 or 229 mm) by 14 in. (356 mm)
deep; and those of Group T specimens were 9 in. wide by
14 in. deep (229 x 356 mm). The small width variation in
Group C specimens was to provide the same lateral spacing
between bars as a function of bar diameter for the two different
bar sizes tested (No. 6 [19 mm] and No. 5 [16 mm]).
Main tensile reinforcement in Group L specimens consisted
of two No. 4 (13 mm) or two No. 5 (16 mm) bars. The main
reinforcement was lap spliced within the constant moment
region using different lengths in each specimen. For each
bar size (No. 4 [13 mm] or No. 5 [16 mm]), the longest lap
splice length chosen was consistent with ACI 318-14 devel-
opment length equations in Section 25.4.2.3, including the
ACI Structural Journal/January 2018
Table 1—Properties of laboratory specimens
Specimen b, in. h, in. As, in.2 ls, in. ld,* in. fc′, psi fy, ksi fu, ksi εy εu
L-4-24 13 9 0.4 24 6190 0.007 0.065
L-4-19 13 9 0.4 19 18.3 6210 120 198 0.007 0.065
L-4-14 13 9 0.4 14 6085 0.007 0.065
L-5-37 13 9 0.62 37 6190 0.007 0.060
L-5-29 13 9 0.62 29 23.5 6155 120 178 0.007 0.060
L-5-21 13 9 0.62 21 6830 0.007 0.060
Control CT 9 14 0.62 — — 5460 105 155 0.0056 0.060
C-5-M 9 14 0.62 — — 5140 105 155 0.0056 0.060
C-6-M 11 14 0.88 — — 5400 121 170 0.0062 0.050
C-5-E 9 14 0.62 — — 4280
T-5-12 9 14 0.62 12 4320
T-5-8 9 14 0.62 8 4410
105 155 0.0056 0.060
T-5-4 9 14 0.62 4 15.8 4640
T-5-8R 9 14 0.62 8 4490
T-5-4R 9 14 0.62 4 4280
*
Computed using ACI 318-14 Eq. (25.4.2.3a) for L-series specimens and Section 25.4.4.2(a) for T-series specimens.
Notes: 1 in. = 25.4 mm; 145 psi = 1 MPa; 1 ksi = 1000 psi.

modification for a Class B lap splice (ACI 318-14, Section

25.5.2.1) to theoretically reach a bar stress approximately
equal to yield (120 ksi [827 MPa]); refer to “Material prop-
erties” section. To promote splice failure in other specimens,
splice lengths were reduced to approximately 80% and 60%
of the longest splice for each bar size.
Tensile reinforcement of Group T specimens consisted of
two No. 5 (16 mm) bars. The reinforcement in these speci-
mens was spliced using non-contact lapped bars that termi-
nated with a threaded T-head. The splice lengths in these
specimens were shorter than estimated using ACI 318-14
Section 25.4.4.2 to investigate the behavior of specimens up
until the onset of splitting failure. The longest splice length,
corresponding approximately to 75% of the code deter-
mined value, was equal to 12 in. (305 mm); this length was
decreased to 8 and 4 in. (204 and 102 mm) in other speci-
mens within this group. Previous researchers (Thompson et
al. 2006) have demonstrated that the presence of the T-head
allows bars to develop stress in a much shorter distance
compared with a straight bar splice. The T-headed ends of
adjacent spliced bars were staggered in the splice region, as
illustrated in Fig. 2.
Group C specimens had two No. 5 (16 mm) or two No. 6
(19 mm) bars as main tensile reinforcement. The bars were
spliced using threaded couplers machined from the same
material as the reinforcement. Longitudinal reinforcement
on the compression side in all specimens consisted of two
No. 3 (9 mm) bars. Transverse reinforcement (stirrups) Fig. 3—Splice details for Group T Specimens: (a) transverse
outside the constant moment region was fabricated using reinforcement (Specimens T-5-4R and 8R); and (b) strain
bars at spacing necessary to reach flexural strength without gauge placement.
prior occurrence of shear failure. stirrup to be placed at the center of the splice and the first
Only two specimens in Group T included transverse stirrup next to the T-head outside the splice. All other spec-
reinforcement within the lap splice (Fig. 3(a)). Specimens imens in Groups L and T had no transverse reinforcement
T-5-8R and T-5-4R had No. 3 stirrups at a spacing of 8 and within the lap splice or inside the constant moment region.
4 in. (200 and 100 mm), respectively, which allowed one

ACI Structural Journal/January 2018 81

Material properties
Beams were designed using a nominal 28-day compres-
sive strength of concrete, fc′, equal to 5000 psi (35 MPa) for
specimens in Group L and 4000 psi (28 MPa) for specimens
in Groups C and T. The tensile and compressive strengths of
concrete were determined the day specimens were loaded to
failure by testing companion cylinders and plain concrete
beams that were cast and cured under the same laboratory
conditions as the specimens. The concrete compressive
strength at the day of testing varied between an average
minimum of 4300 psi (30 MPa) and a maximum of 5460 psi
(38 MPa), depending on the time that elapsed between casting
and testing date. Average results from concrete cylinder and
beam tests for each specimen are listed in Table 1.
Longitudinal tension reinforcement conformed to ASTM
A1035/A1035M. Three different bar sizes were used in the
testing program, and each bar size had slightly different
material stress-strain properties. The stress-strain prop-
erties of two of the bar sizes (No. 5 [16 mm] and No. 6
[19 mm]) were determined in the laboratory by tensile testing
of bar segments cut from the same longitudinal reinforce-
ment used in the specimens. The stress-strain properties of the
No. 4 (13 mm) bars listed in Table 1 were reported by the
manufacturer of the reinforcing steel. Compression steel and
transverse reinforcement (No. 3 [9.5 mm] bars) conformed to
ASTM A615 Grade 60 (Grade 420); mechanical properties
were not determined in the laboratory for these bars. Measured
stress-strain curves of No. 5 (16 mm) and No. 6 (19 mm) bars
are shown in Fig. 4(a). The yield stress of the reinforcement
was determined using the 0.2% offset method as illustrated in
the figure. For use in moment-curvature analysis as described
later in this paper, relationships that approximately describe Fig. 4—Stress-strain properties of ASTM A1035/A1035M
the measured stress-strain behavior of the bars used in this bars: (a) measured; and (b) bilinear approximation. (Note:
testing program were obtained for the No. 5 (16 mm) bars 1 ksi = 6.895 MPa.)
constructed using Eq. (2a) and (2b) is presented in Fig. 4(b).
 29, 000ε s (ksi) ε s ≤ 0.00275 This comparison reveals that the bilinear approximation
 (2a)
fs =  0.4 would result in reasonably accurate stresses, particularly
163 − ε + 0.00194 (ksi) 0.00275 < ε s ≤ 0.060 within the post-yield region. The two stress-strain modeling
 s
approaches were used to calculate the maximum moment
that the sections could theoretically develop if the bars were
and for the No. 6 (19 mm) bars
continuous, as will be discussed later in this paper.
Commercially available couplers and T-heads are typi-
 29, 000ε s (ksi) ε s ≤ 0.00275
cally qualified to develop at least 1.25fy of reinforcement

fs =  0.4 (2b) with a well-defined yield plateau. Because of the high
178 − ε + 0.00132 (ksi) 0.00275 < ε s ≤ 0.060 strength characteristics of the ASTM A1035/A1035M bars
 s
used in this research, mechanical devices commercially
Equations (2a) and (2b) have the same form as those available for ASTM A615 bars could potentially fracture at
proposed by Mast et al. (2008) but with coefficients that low strains. Therefore, threaded mechanical couplers and
were adjusted to better represent the measured proper- T-heads were machined from round stock that had the same
ties of the bars used in this research. Stress-strain rela- material properties as the tension longitudinal reinforcement
tions proposed by Seliem et al. (2009) and Shahrooz et al. in this research. The geometry of couplers and T-heads fabri-
(2014) were also evaluated but were not selected because cated for this research project is presented in Fig. 5. Coupler
these material models gave a sharper transition from the E was only used in one of the specimens (Specimen C-5-E),
linear portion to the hardening region of the stress-strain as listed in Table 1. Ends of reinforcing bars were threaded at
diagrams. The stress-strain curves of the tension reinforcing the fabrication shop prior to delivery in accordance with the
bars were also modeled using a simple bilinear approxi- coupler or T-heads they were designed to use. The resulting
mation to simplify calculations. A comparison between the diameter of the couplers and T-heads was approximately
bilinear approximations for each bar size and the curves 50% larger than those used for ASTM A615 bars because
the higher material hardness generated difficulties during

82 ACI Structural Journal/January 2018

Fig. 5—Couplers and T-head geometries.
milling. Pictures illustrating coupled and T-head reinforced
bars during assembly of two selected specimens are shown
in Fig. 6.
Instrumentation
External instrumentation on the specimens consisted of
load cells and displacement transducers (linear potentiom-
eters). Individual load cells were used for each hydraulic
ram at the specimen ends to measure the total load applied.
Differences in measured load at each of the load cells were
less than 2% of the total applied load throughout testing,
so loading was considered symmetrically applied. Linear
potentiometers were positioned at beam centerline and at the
locations of load application near the beam ends. Given the
test configuration, the total beam deflection was determined
by adding the average displacement measured near the ends
of the specimen to the displacement measured at centerline.
Bonded strain gauges were used to instrument reinforcing
steel at various locations depending on each specimen and
to instrument the compression surface of all specimens at
midspan. Steel strain gauges were mounted on both rein-
forcing bars of Specimen CT at midspan. In Group C spec-
imens, steel strain gauges were bonded to both reinforcing
bars 1.75 in. (44.5 mm) on either side from the center of
the coupler located at midspan. In specimens containing
splices (Group L and T), the spacing and number of steel
strain gauges varied for each specimen because of differ-
ences in splice length. Both bars in one of the two splices in
Group L specimens were instrumented with equally spaced
strain gauges. The spacing of gauges ranged from 2.75 in.
(70 mm) for the specimen that had the shortest splice length
(L-4-14) to 6.2 in. (157 mm) for the specimen containing
the longest splice length (L-5-37). Although more closely
spaced gauges are likely to better capture sudden changes in
strain along the bars, it was decided to have the same number
of strain points along all spliced bars in this group.
In Group T specimens, three out of the four spliced
bars were instrumented using strain gauges at equal inter-
vals along the splice. The first strain gauge was placed at a
distance equal to 2db (1.25 in. [32 mm]) from the inside face
ACI Structural Journal/January 2018
Fig. 6—Layout of coupled and T-headed reinforcing bar
cages showing strain gauges.
of the T-head with the goal to separate the contribution of
the T-head from the stress developed along the spliced bars,
but this was not always possible. Other strain gauges were
distributed at a constant spacing (between 2.5 and 3.0 in. [64
and 76 mm], if space was available) within the remaining
length of the splice (Fig. 3).
TEST RESULTS
A list of the failure modes observed after testing of each
specimen is given in Table 2. Control Specimen CT and
Group C (coupled bar) specimens, with exception of Spec-
imen C-5-M, exhibited flexural failure characterized by
crushing of the concrete on the compression face of the
beams after bar yielding. The bars in Specimen C-5-M frac-
tured next to the threaded ends when the specimen reached
99% of the estimated flexural strength using the simpli-
fied approach described later in this paper. Concrete in the
compression zone had started to flake at this load level, indi-
cating impending crushing. Crack patterns at failure of all
specimens containing splices were indicative of splitting
of concrete along the spliced bars. In specimens containing
straight bar splices (Group L), flexural cracks formed first
within the constant moment region typically at or near
midspan, and at the end of the splice. These cracks were
followed by formation of longitudinal splitting cracks that
appeared on the sides and/or on the tensile face of the beams
and gradually lengthened at higher loads. The splitting cracks
83
Table 2—Summary of test results
Test results M-ϕ analysis Simplified
Ptest, Mtest, fs test, Mcalc, Mcalc, Pcalc, fs calc,†
Specimen kip kip-in. εs,test* ksi kip-in. Mtest/Mcalc kip-in. Mtest/Mcalc kip ksi fs test/fs calc Observed failure mode
L-4-24 14.2 370.7 0.0094 126 0.93 1.05 119 1.06 Bottom splitting
Bottom splitting + one
L-4-19 14.1 368.2 0.0107 132 0.92 1.04 106 1.25
399 353 14.1 side splitting
Bottom splitting + side
L-4-14 11.3 298.2 0.0073 118 0.75 0.84 91 1.30
splitting
Side and bottom splitting
L-5-37 19.6 505.7 0.0085 122 0.92 0.97 118 1.03
(corner)
548 523 20.9 Side and bottom splitting
L-5-29 18.9 488.2 0.0064 113 0.89 0.93 105 1.08
(corner)
L-5-21 14.5 378.2 0.0053 103 0.69 0.72 91 1.13 —
Flexural – concrete
Control CT 45.4 966.6 0.0122 135 928 1.04 941 1.03 44.8 141 0.96
crushing
C-5-M 44.9 957.6 0.0139 138 914 1.05 974 0.98 46.4 141 0.98 Reinforcing bar fracture
Flexural – concrete
C-6-M 58.1 1238.0 0.0127 135 1311 0.94 1355 0.91 64.5 147 0.92
crushing
Flexural –concrete
C-5-E 41.4 883.0 0.0168 143 1.01 0.96 140 1.02
crushing
T-5-12 20.2 439.1 0.0048 98 0.50 0.48 79 1.24 Side splitting along splice
T-5-8 14.6 320.5 0.0018 52 0.37 0.35 53 0.98 Side splitting along splice

871 916 43.6 Side and bottom splitting

T-5-4 13.7 301.5 0.0023 59 0.35 0.33 27 2.17
along splice
Side and bottom splitting
T-5-8R 23.2 501.4 0.0039 90 0.58 0.55 54 1.68
along splice
Bottom splitting
T-5-4R 21.1 457.8 0.0027 69 0.53 0.50 26 2.64
diagonally to T-head
*
εs,test is average of strains measured end of splice of all instrumented bars.
†
fs calc was determined using Eq. (3b) for Group L specimens, simplified flexural strength method for Group C specimens, and Eq. (4b) for Group T specimens.
Notes: 1 kip = 4.45 kN; 1 kip-in. = 113 kN-mm; 1 ksi = 6.895 MPa.

that formed during loading eventually led to the formation of
a mechanism and failure of the splice by splitting.
Specimens in Group T also exhibited cracks that formed
at the end of the splice region behind the T-headed end of
alternate bars in addition to cracking at midspan. No longi-
tudinal cracking along the reinforcing bars was apparent
throughout testing in these specimens. Failure of Group T
specimens was triggered by widening of the cracks at the
splice ends and sudden formation of longitudinal splitting
crack throughout the splice region. Inspection of the splice
region after failure of these specimens revealed the occur-
rence of a large displacement of the back side of the T-head
as concrete crushed in front of the T-head (refer to Fig. 7).
It is apparent that crushing of concrete released the T-head,
causing a sudden increase in bond stresses along the bars,
which led to the sudden splitting failure observed. The
formation of a truncated concrete cone between diagonal
cracks in front of the T-head was also visible in these spec-
imens (Fig. 7). The length of the cones measured in several
specimens ranged between 2.1 and 2.2 in. (53 and 56 mm),
resulting in an angle of inclination between the bar axis and
the diagonal cracks of approximately 21 degrees.
Load-deflection response
Measured peak loads and corresponding moments for all
the specimens are summarized in Table 2. The moments
listed in the table were adjusted to include the self-weight
moment computed at the end of the constant moment region.
Load-deflection responses of various groups of specimens
are compared in Fig. 8. Deflection data were lost during
testing of Specimen L-5-21, so the load-deflection curve
for this specimen is not included in Fig. 8(a) but the peak
measured load is listed in Table 2. It should be noted that
the figures are drawn at different scales so that detailed
features of the load-deflection plots of the different groups
of specimens can be easily observed. The loads calculated
under the assumption that specimens would reach their
corresponding flexural strength at the critical section under
one of the load points Pcalc are indicated in each plot with a
dashed horizontal line and are also listed in Table 2. These
calculated loads were estimated by iterating on the neutral
axis position using the Whitney stress block for concrete in
compression and a bilinear approximation for the steel in
tension in accordance with Fig. 3(b). The measured material
properties for different groups of specimens were used in
these calculations.

84 ACI Structural Journal/January 2018

Fig. 7—Failure of Specimen T-5-8R.
Fig. 8—Selected comparisons of load-deflection responses within groups of specimens.
For lap-spliced specimens, only Specimens L-4-24 and
L-4-19 reached the estimated load Pcalc (Fig. 8(a)). These
specimens, however, failed by splitting along the lap splice
almost immediately after reaching the estimated flexural
strength without a sizeable increase in deflection beyond the
peak load. Specimens L-5-37 and L-5-29 failed at approx-
imately 94% and 90% of the loads corresponding to the
estimated flexural strength, respectively, while specimens
containing the shortest splice lengths (L-4-14 and L-5-21)
failed after only reaching approximately 80% and 70%
of the calculated flexural strength, respectively. The ratio
of splice length ls of each specimen divided by calculated
development length ld using ACI 318-14 Section 25.4.2.3
ACI Structural Journal/January 2018
with fy = 120,000 psi (830 MPa) and the measured concrete
compressive strength is presented in Table 1. Only Speci-
mens L-4-14 and L-5-21 had ls/ld ratios less than 1.0, but it
should be noted that the splice factor of 1.3 was not used in
the calculation for ld.
Specimens in Group C (coupled bar specimens) reached
or exceeded the load corresponding to flexural strength
calculated using measured material properties. Only Spec-
imen C-6-M fell short of the calculated flexural strength, by
approximately 10%.
Specimens in Group T were deliberately designed to fail
within the splice region prior to reaching the estimated flex-
ural strength of the cross section. Splice lengths were much
85
Fig. 9—Strains in maximum moment region of Specimen CT and Group C specimens.
shorter than the tension development length for headed bars
computed using ACI 318-14 Section 25.4.4.2, assuming a
yield stress of 105 ksi (730 MPa) and the average concrete
strength of all Group T specimens. As shown in Fig. 8(c)
and Table 2, all the specimens in Group T failed at a fraction
of the capacity Pcalc calculated assuming continuous bars.
Figure 8(d) shows that specimens that contained transverse
reinforcement within the splice (T-5-8R and T-5-4R) were
able to develop between 55 and 60% higher ultimate loads,
respectively, in comparison with corresponding specimens
that did not contain transverse reinforcement along the splice
(T-5-8 and T-5-4). Having transverse reinforcement within
the splice delayed failure and increased ductility, even with
the minimal amount of transverse reinforcement provided
and the extremely short splices used in these specimens.
Reinforcing bar and concrete strains
Strains were measured within the maximum moment
region of all specimens. The location of strain gauges for
each group of specimens was discussed in the “Instrumen-
tation” section.
Strains measured in Specimen Control CT and Group C
specimens are illustrated in Fig. 9. Failure associated
with concrete crushing after reinforcing bar yielding was
observed in all these specimens as discussed previously,
so measured strains were anticipated to be compatible with
this failure mode. Concrete strains are shown as negative
values and were measured in the extreme compression face
at midspan; steel strains are positive and were measured at
midspan for Specimen Control CT or next to the coupler
for Group C specimens. As observed in the plots in Fig. 9,
steel strains exceeded 0.012 in tension and compressive strains
measured in the concrete reached the value of 0.003 often asso-
ciated with concrete crushing. Initiation of concrete crushing was
86
observed in all tests including Specimen C-5-M, which exhibited
reinforcing bar fracture. Plots showing strains measured on the
two bars on both sides of the coupler of all Group C specimens
are presented in Fig. 9(b) through Fig. 9(d). In general, all curves
lie close to each other for individual specimens with the excep-
tion of strains in one bar on one side of the coupler in Speci-
mens C-5-M and C-6-M. The differences in specific values could
have been caused by the proximity of cracking to gauge loca-
tions. However, the general trends in all specimens are similar
and compatible in appearance with strains measured at centerline
of Control CT specimen. The similarity in observed strains indi-
cated that couplers were able to effectively splice high-strength
bars to reproduce the behavior of a beam with continuous bars.
The strains developed along the splice in the two spec-
imens that had the intermediate splice length (Specimens
L-4-19 and L-5-29) are shown in Fig. 10, and are repre-
sentative of other specimens in this group. Strains typically
increased from the end of the bar to the end of the splice
region. Bar strains increased nonlinearly along the splice
region, and local peaks were recorded at locations near
concrete cracks. The curves are labelled a through e, corre-
sponding to the instrument location starting at the beginning
of the splice as illustrated in the bottom right corner of each
plot. The small drops in load at different points along the
curves correspond to the load holds taken during testing to
mark cracks and document damage in the beams.
In Specimen L-4-19, the first change in slope occurred
in all the strain curves at approximately 3 kip (13.3 kN)
corresponds to first flexural cracking. Strains at location e
increased markedly at approximately 7 kip (31 kN), corre-
sponding to the occurrence of flexural cracking near the
instrument location. At loads nearing failure of the specimen
above 10 kip (44.5 kN), strains increased at a higher rate
for smaller increments of load. Because strain gauges were
ACI Structural Journal/January 2018
Fig. 10—Strains measured in selected Group L specimens.
Fig. 11—Strain distribution along splice in selected Group L specimens.
bonded at equal distances along the splice, the wider spacing
between curves at peak load observed for strains near the
end of the splice give evidence of the nonlinear development
of strain along the bar (refer also to Fig. 11(a)).
Measured strains in Specimen L-5-29 showed similar
trends as those discussed previously. Curves a, b, and c
shown in Fig. 10(b), corresponding to the first three strain
gauges starting from the end of the bar, exhibit a sudden
decrease at a load of approximately 17 kip (432 kN). The
decrease in strain near the end of the bar occurred at a load
corresponding to the formation of the first splitting crack
along the splice that apparently caused the bars near the
splice end to slip. At higher loads, strains measured in these
three gauges no longer increased at the same rate as prior
to formation of splitting cracks, further supporting the argu-
ment for bar slip.
Strain distributions along the splice for four different load
levels for the same two specimens discussed previously are
presented in Fig. 11. Strain variation along the splice for the
two lowest load levels presented (cracking [Pcr] and 50% of
yield [Py]) was not as pronounced as for the two highest load
levels (Py and Pmax). Strains increased with distance from
the end of the spliced bar as would be expected. A higher
strain gradient was observed in the region farthest from the
bar end, but the strain values varied significantly in the two
cases presented. This result is not surprising because the
value of measured strain is strongly affected by strain gauge
location with respect to the location of cracks.
In specimens of Group T, the development of bar strains
along spliced bars was strongly influenced by the presence of
ACI Structural Journal/January 2018
T-heads at the ends of bars. Strain development was depen-
dent on splice length and presence of transverse reinforce-
ment within the splice. Each graph in Fig. 12 presents six
curves, two for each of the three instrumented bars depicting
strains measured at two locations (end of splice and adjacent
to T-head).
Strain values developed in specimens containing trans-
verse reinforcement (T-5-8R and T-5-4R) within the splice
region were much higher than their companion specimens
(T-5-8 and T-5-4). The maximum strain developed in spec-
imens without transverse reinforcement was between 0.002
and 0.003 and, interestingly, sometimes occurred near the
T-head. The change in strain from the T-head (triangles) to
the end of the splice (circles) in these specimens was not
substantial (Fig. 12(a) and (b)). The presence of transverse
reinforcement in Specimens T-5-8-R and T-5-4R resulted in
larger strains developing in each bar, particularly at the end
of the splice region instead of near the T-head. Strain values
in these specimens reached between 0.004 and 0.0045. The
increase in strain along each bar between the T-head and the
splice end is also noted (Fig. 12(c) and (d)). The observa-
tions on strain behavior of specimens in Group T point to
differences in the force-transfer mechanism between adja-
cent lapped bars in specimens with and without transverse
reinforcement along the splice.
To better understand the differences in force-transfer mech-
anisms between specimens in Group T, strut-and-tie models
were constructed for Specimens T-5-4 and T-5-4R (Fig. 13).
Solid lines represent ties and dashed lines indicate struts; the
force in each element of the models is expressed as a fraction
87
Fig. 12—Comparison of measured bar strains in Group T specimens.
of the force in the exterior longitudinal bar Text. In Specimen
T-5-4, force transfer occurs primarily through the formation
of a diagonal strut between T-heads of two adjacent bars.
The tension force in longitudinal bars of this model is nearly
constant throughout the splice length consistent with the obser-
vation that measured strains next to the head and at the end of
the splice remained approximately constant in this specimen
(refer to Fig. 12(b)). The force in interior longitudinal bars of
the model is three times higher than the force in exterior bars.
This value is larger than the factor of 2.3 between interior and
exterior strains measured during the test, but the model serves
to illustrate the force-transfer mechanism that was observed.
The strut-and-tie model illustrating the force-transfer mecha-
nism of Specimen T-5-4R includes an intermediate tie corre-
sponding to the transverse reinforcement within the splice.
Presence of the middle tie increases the concrete strut inclina-
tion and allows the force in longitudinal ties to change within
the splice region, as was evidenced by changes in strains Fig. 13—Strut-and-tie models for Group T specimens
measured along bars during the experiments (compare strain containing 4 in. (102 mm) splice.
at head with strain at end of splice [Fig. 12(a)]). The force
ratio between interior longitudinal ties and exterior longitu- reinforcement. This assumption allowed determining the
dinal ties was 1.67 in this case, less than the ratio observed in fraction of the calculated flexural strength that each spec-
Specimen T-5-4. This force difference between interior and imen was able to reach depending on its splice length.
exterior bars was also noted in the strain measurements taken The moment-curvature response of each specimen was
during the test, as can be observed in Fig. 12(d). estimated using measured material properties in combina-
tion with the following assumptions. Concrete was consid-
Comparison of measured moments with estimated ered unconfined and modeled using the Mander et al. (1988)
flexural strength uniaxial stress-strain model with an unconfined strength
For comparison with experimentally determined moments, determined from the average measured compressive
flexural strength of all specimens was estimated using two strength of companion concrete cylinders. Tensile longitu-
procedures: a moment curvature analysis approach and a dinal reinforcement was modeled using Eq. (1a) for No. 4
simplified approach. Regardless of the procedure used, the (13 mm) and No. 5 (16 mm) bars, and Eq. (1b) for No. 6
longitudinal reinforcement was assumed to be continuous (19 mm) bars. The moment-curvature curve of each spec-
even in specimens containing lapped splices or T-headed imen was used to determine the flexural strength associated
with concrete crushing in compression after yielding of the
88 ACI Structural Journal/January 2018
tension reinforcement. Table 2 lists the calculated flexural
strength Mcalc and the ratio between the moment measured
during the test and the calculated flexural strength based on
moment-curvature analysis. The Mtest/Mcalc ratio for spec-
imens with continuous reinforcement (Group C) ranges
between 0.94 and 1.04. These ratios are less than 1.00 for
all specimens containing splices, as would be anticipated,
particularly for elements containing short splices.
An approximate approach was also followed to determine
flexural strength of the specimens as a way of computing a
reasonable estimate of strength without the need to conduct
a moment-curvature analysis. Being a sectional model, this
method also assumed that the reinforcement was continuous
in all specimens, so flexural strength was reached when
concrete crushed in compression after bar yielding. Concrete
stresses within the compression zone were represented using
the equivalent stress block described in ACI 318-14, Section
22.2.2.4. Tensile strength of concrete was neglected and
perfect bond between concrete and tension reinforcement
was assumed. The stress-strain response of longitudinal
reinforcement was represented using a bilinear approxi-
mation with properties as shown in Fig. 3(b). To compute
flexural strength, the neutral axis depth was determined by
iteration until internal force equilibrium between tensile and
compressive forces was achieved within a reasonable toler-
ance. Computed flexural strengths of all specimens using this
simplified approach are listed in Table 2. Flexural strengths
calculated using the simplified method described in this
paper were not significantly different from those obtained
using the more rigorous moment-curvature approach. The
maximum difference was approximately 10% for L-4 spec-
imens between the two approaches; differences in calcu-
lated strength of other specimen groups were below 5%. A
comparison between calculated and measured bar stresses
was then conducted using the simplified approach for spec-
imens with continuous reinforcement (Group C). As shown
in Table 2, computed stresses compared favorably with those
determined during the tests by entering the bar stress-strain
relation using average strains measured within the maximum
moment region.
EVALUATION OF LONGITUDINAL BAR STRESSES
Several existing equations to compute development length
of bars in concrete were used to evaluate their applicability
to ASTM A1035/A1035M bars. All these equations were
developed from statistical analyses of test data that included,
in most cases, beam or slab tests with most of the reinforce-
ment having a nominal yield stress of up to 60 ksi (420 MPa).
Only the development length equation proposed by Hosny
et al. (2012) included tests that had ASTM A1035/A1035M
bars. The applicability of the development length equations
evaluated by Hosny et al. is also limited to a maximum
concrete compressive strength of 10,000 psi (69 MPa),
because higher concrete strength was not used in the beams
reported by the researchers.
Many research studies have been devoted to develop-
ment length and lap splices of straight bars in tension.
Fewer studies have concentrated on studying behavior of
T-headed spliced reinforcement. Only specimens containing
ACI Structural Journal/January 2018
lapped splices (Group L) were evaluated using the develop-
ment length models proposed by Hosny et al. (2012). Bar
stresses of specimens in Group T were evaluated using ACI
318-14 Section 25.4.4.2, which is based on the research by
Thompson et al. (2006).
Hosny et al. (2012) proposed a development length equa-
tion for high-strength reinforcement conforming to ASTM
A1035/A1035M for unconfined conditions. The equation
proposed was based on the results of 66 beams containing
splices without transverse reinforcement along the splice.
This equation was subsequently used to determine stresses
for comparison with test results from a much larger number
of tests (a total of 213 beams and slabs) to determine its
validity for other types of reinforcement. The resulting
development length equation of bars in tension without
confining reinforcement was
ld (0.001 f s ) 2
= (3a)
db 2 cmin f c′
Solving for bar stress gives
1000 f c′1/ 4 ld cmin
fs = (3b)
db
where cmin is the minimum concrete cover; fs is the bar stress;
db is the spliced bar diameter; and fc′ (in psi) represents the
concrete compressive strength.
The tension development length for T-headed bars with
a nominal yield stress not exceeding 60,000 psi (420 MPa)
and concrete compressive strength not exceeding 6000 psi
(42 MPa), ACI 318-14, Section 25.4.4.2, requires that
ld 0.016 f y
= (4a)
db f c′
Replacing development length, ld, for splice length, ls, and
the yield stress, fy, for bar stress fs, an expression for bar
stress can be used to compare with stress values measured
during the tests
ls f c′ 1
fs = ⋅ (4b)
0.016 db
Bar stresses in Group L specimens were estimated using
Eq. (3b) while those in Group T specimens were estimated
using Eq. (4b), as listed in Table 2 under the column corre-
sponding to fs,calc. Bar stresses in Group C and the Specimen
CT were found as part of the simplified flexural strength
calculation described earlier in this paper, assuming a
bilinear approximation for the stress-strain relation of the
longitudinal reinforcement. Measured bar stresses in Groups
L and C exceeded calculated values in most cases except for
three specimens (CT, C-5-M, and C-6-M). This result indi-
cates that the procedures used to estimate bar stresses that
can then be used to estimate flexural strength were found
to be conservative. Measured (test) bar stresses in Group T
specimens were also higher than those calculated using
89
Eq. (4b), except for Specimen T-5-8, where the measured
stresses were 2% lower than estimated. In the majority of
Group T specimens, the ratio of test-to-calculated stress was
much higher than 1.0, indicating that Eq. (4b) was very conser-
vative, in some instances by over 100%. The largest ftest/fcalc
values correspond to specimens containing transverse rein-
forcement within the splice region of specimens in Group T.
This result differs from the findings of Thompson et al. (2006),
who showed that presence of transverse reinforcement had little
or no influence on splice strength of T-headed reinforcement.
ACI 318-14 provisions for T-headed reinforcement are based on
these researchers’ findings (Eq. (4a)). Larger-diameter T-heads
were used in the tests presented in this paper, causing them to
be in contact with adjacent bars. This may have constrained
lateral spread of compression stresses generated in front of
the head, promoting efficiency of transverse reinforcement in
increasing the strength of the splice.
CONCLUSIONS
Splice strength tests of straight and T-headed high-strength
reinforcement (ASTM A1035/A1035M) were conducted to
provide information on behavior, to understand the force-
transfer mechanisms within the splice region, and to deter-
mine the influence of splice length in stress developed in
longitudinal bars. The following conclusions may be drawn
as a result of these series of tests:
1. The specimens tested in this research containing straight
bar spliced high-strength reinforcement failed by splitting of
concrete. This failure mode could potentially be eliminated
if transverse reinforcement were present in the splice region.
2. Straight bar lap splices involving high-strength reinforce-
ment may become impractically long if the design goal is to
promote a ductile failure involving reinforcing bar yielding
followed by concrete crushing in compression, thereby
avoiding splitting failures. Again, the presence of transverse
reinforcement within the splice length would help promote
ductility by eliminating or delaying splitting failures.
3. Flexural strength of members containing high-strength
reinforcement without a well-defined yield plateau can be
estimated with sufficient accuracy by adopting simplified
models for concrete in compression (stress block) and repre-
senting the stress-strain relation of steel by a bilinear relation.
4. Couplers may be used to connect high-strength rein-
forcing bars and promote ductile flexural failure associ-
ated with bar yielding in tension and concrete crushing in
compression of elements containing this reinforcement.
5. The presence of transverse reinforcement within the
lap splice region of T-headed bars changes the force-transfer
mechanism between adjacent bars and leads to the develop-
ment of higher stresses along the splice and higher member
deformation capacity for equal lap lengths.
AUTHOR BIOS
Sergio F. Breña, FACI, is a Professor at the University of Massachusetts
Amherst, Amherst, MA. He received his MS and PhD from the University
of Texas at Austin, Austin, TX. He is a member of ACI Committees 369,
Seismic Repair and Rehabilitation; 374, Performance-Based Seismic
Design of Concrete Buildings; 440, Fiber-Reinforced Polymer Rein-
90
forcement; Joint ACI-ASCE Committee 445, Shear and Torsion; and ACI
Subcommittee 318-C, Safety, Serviceability, and Analysis (Structural
Concrete Building Code).
Jeffrey Messier is a Project Manager with Schaefer in Phoenix, AZ. He
received his BS from the University of Massachusetts Dartmouth, Dart-
mouth, MA, in 2002 and his MS in civil engineering (structures) in 2004.
Sean W. Peterfreund is an Associate with McMillen Jacobs Associates,
Seattle, WA. He received his BS from the University of Massachusetts
Amherst in 2003 and his MS from the University of Michigan, Ann Arbor,
MI, in 2004.
ACKNOWLEDGMENTS
The Commonwealth College at the University of Massachusetts provided
funds for the third author’s honors research. The authors would like to thank
MMFX Technologies Corporation for donating the ASTM A1035/A1035M
reinforcement subject of this research, Erico for fabricating and providing
mechanical couplers and terminators, and Barker Steel for providing trans-
verse and compression reinforcement used in the specimens.
NOTATION
c = spacing or cover dimension = cmin + db/2, in.
cmax = maximum concrete cover (side or bottom), in.
cmin = minimum concrete cover (side or bottom), in.
db = bar diameter, in.
fc′ = concrete compressive strength, psi
fs = reinforcing bar stress, psi
fy = yield stress of reinforcing bar, psi
Ktr = transverse reinforcement index = (0.52trtdAtr/sn)fc′
ld = development length of reinforcing bar, in.
ls = splice length of reinforcing bar, in.
Rr = relative rib area of the reinforcement ≈ 0.073
td = 0.78db + 0.22
tr = 9.6Rr + 0.28 ≤ 1.72
εs = strain in reinforcing bar
ϕ = strength reduction factor
ω = (0.1 cmax/cmin + 0.9) ≤ 1.25
REFERENCES
ACI Committee 318, 2014, “Building Code Requirements for Struc-
tural Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 519 pp.
ACI Innovation Task Group 6, 2010, “Design Guide for the Use of ASTM
A1035/A1035M Grade 100 (690) Steel Bars for Structural Concrete (ACI
ITG-6R-10),” American Concrete Institute, Farmington Hills, MI, 90 pp.
Applied Technology Council, 2014, “Roadmap for the Use of High-
Strength Reinforcement in Reinforced Concrete Design (ATC-115),”
Redwood City, CA, 194 pp.
Hosny, A.; Seliem, H. M.; Rizkalla, S.; and Zia, P., 2012, “Development
Length of Unconfined Conventional and High-Strength Steel Reinforcing
Bars,” ACI Structural Journal, V. 109, No. 5, Sept.-Oct., pp. 655-664.
Joint ACI-ASCE Committee 408, 2003, “Bond and Development of
Straight Reinforcing Bars in Tension (ACI 408R-03),” American Concrete
Institute, Farmington Hills, MI, 49 pp.
Mander, J. B.; Priestley, M. J. N.; and Park, R., 1988, “Theoret-
ical Stress-Strain Model for Confined Concrete,” Journal of Struc-
tural Engineering, ASCE, V. 114, No. 8, pp. 1804-1826. doi: 10.1061/
(ASCE)0733-9445(1988)114:8(1804)
Mast, R. F.; Dawood, M.; Rizkalla, S. H.; and Zia, P., 2008, “Flexural
Strength Design of Concrete Beams Reinforced with High-Strength Steel
Bars,” ACI Structural Journal, V. 105, No. 4, July-Aug., pp. 570-577.
Seliem, H. A.; Hosny, A.; Rizkalla, S.; Zia, P.; Briggs, M.; Miller, S.;
Darwin, D.; Browning, J.; Glass, G. M.; Hoyt, K.; Donnelly, K.; and Jirsa,
J. O., 2009, “Bond Characteristics of ASTM A1035 Steel Reinforcing
Bars,” ACI Structural Journal, V. 106, No. 4, July-Aug., pp. 530-539.
Shahrooz, B. M.; Reis, J. M.; Wells, E. L.; Miller, R. A.; Harries, K. A.;
and Russell, H. G., 2014, “Flexural Members with High-Strength Reinforce-
ment: Behavior and Code Implications,” Journal of Bridge Engineering,
ASCE, V. 19, No. 5, pp. 1-7. doi: 10.1061/(ASCE)BE.1943-5592.0000571
Thompson, M. K.; Ledesma, A.; Jirsa, J. O.; and Breen, J. E., 2006, “Lap
Splices Anchored by Headed Bars,” ACI Structural Journal, V. 103, No. 2,
Mar.-Apr., pp. 271-279.
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S07
End-Region Behavior of Pretensioned I-Girders Employing
0.7 in. (17.8 mm) Strands
by J. Salazar, H. Yousefpour, R. Alirezaei Abyaneh, H. Kim, A. Katz, T. Hrynyk, and O. Bayrak
This paper presents an experimental study on transfer length,
end-region cracking, and transverse end-region stresses in preten-
sioned concrete I-girders fabricated using 0.7 in. (17.8 mm) diam-
eter strands on a 2 x 2 in. (51 x 51 mm) grid. The full-scale specimens
consisted of two Tx46 and two Tx70 girders that were fabricated in
a controlled laboratory environment using different strand patterns
and concrete release strengths. The detailing for mild-steel rein-
forcement was done according to the current practice in Texas
for girders with smaller-diameter strands. The measured 24-hour
transfer lengths from the specimens exceeded estimates by both
AASHTO LRFD and ACI 318-14 provisions. The observed crack
widths in the specimens within 28 days after prestress transfer were
generally limited to 0.007 in. (0.18 mm), indicating satisfactory
performance for exposure to deicing chemicals according to ACI
224R guidelines. However, noticeably greater transverse forces
were observed in the end regions of the specimens compared to the
resistance required by AASHTO LRFD specifications.
Keywords: 0.7 in. strands; bursting stresses; end-region cracking; preten-
sioned; spalling stresses; splitting resistance; transfer length.
INTRODUCTION
Pretensioned concrete bridge girders are currently fabri-
cated using 0.5 or 0.6 in. (12.7 or 15.2 mm) diameter
prestressing strands. In recent years, however, it has become
of interest to use 0.7 in. (17.8 mm) diameter seven-wire
strands in pretensioned elements. These larger-diameter
strands have so far been used primarily for mining applica-
tions and cable bridges.1 In addition to reducing the number
of required strands, the increase in the strand diameter will
allow for a greater concentration of steel to exist closer to the
tensile face of the cross section. Therefore, nominal flexural
and shear capacities are expected to be greater in members
with larger-diameter strands. Such capacity increases may
offer improvements in attainable span capabilities, slen-
derness of bridge superstructure, and transverse spacing
between girders.2
In pretensioned concrete elements, the stress in the strand
increases from zero at the end face of the member to the full
prestress level over a distance referred to as “transfer length.”
A longer transfer length results in a more gradual develop-
ment of prestress with distance from the end face and, there-
fore, reduces the end-region damage at the time of prestress
transfer. However, reduced prestress levels developed within
the transfer length might negatively affect the strength of
pretensioned girders, especially under shear-critical loading
conditions.3 Therefore, the transfer length ltr is a critical
design parameter. In most design provisions, ltr is considered
primarily a function of strand diameter db. A value of 60db
is used in AASHTO LRFD 2014 Bridge Design Specifica-
ACI Structural Journal/January 2018
TECHNICAL PAPER
tions.4 In ACI 318-14 provisions,5 ltr is taken as (fse/3000)db,
in which fse is the effective stress in the strands after prestress
losses in psi, or (fse/21)db if fse is in MPa. Therefore, on the
basis of these provisions, an increase in the strand diameter
is expected to result in an increase in the transfer length.
However, studies supporting the aformentioned provisions
were done considering strands with diameters that were less
than or equal to 0.6 in. (15.2 mm),6 and the applicability of
such provisions to larger-diameter 0.7 in. (17.8 mm) strands
has yet to be verified.
To minimize the impact of transitioning to larger-diameter
0.7 in. (17.8 mm) strands on fabrication facilities, it is of
considerable interest to use these strands on the 2 x 2 in. (51 x
51 mm) grid that is typically used for 0.5 and 0.6 in. (12.7 and
15.2 mm) strands. However, incorporating 0.7 in. (17.8 mm)
strands in this configuration is believed to increase the stresses
that develop within the end region of pretensioned girders
and, as such, may lead to end-region damage that is not found
in girders constructed with smaller-diameter strands.
Transfer of prestress within girder end regions results in
transverse stresses that are generally categorized as bursting
and spalling stresses.7 Bursting stresses are mostly due to
Poisson’s effect at the time of release. When the strands
are released, they tend to return to their original diameter,
causing radial stresses in the surrounding concrete.8 These
stresses are typically located in the bottom flange of preten-
sioned girders and can result in longitudinal crack develop-
ment along the strands. Spalling stresses are due to compat-
ibility of the strains near the end face of the girders and are
dependent on the eccentricity of the strands. These stresses
might result in horizontal or inclined cracks that are close to
the end face of the member but are typically located at some
distance from the elevation of the strands.
To alleviate concerns regarding the serviceability and
durability of girders, appropriate reinforcement detailing is
essential for controlling the width of potential cracks due
to bursting and spalling stresses. Confining reinforcement
surrounding the strands and transverse reinforcement in the
web, near the ends of the member, should be used to resist
bursting and spalling stresses, respectively. The amount of
such reinforcement in current detailing practice has been
primarily based on experience or from the results of previous
experimental studies on members with 0.5 and 0.6 in. (12.7
and 15.2 mm) diameter strands. Therefore, the suitability of
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2016-309.R2, doi: 10.14359/51700783, received February 13, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
91
Table 1—Summary of previous studies on specimens with 0.7 in. (17.8 mm) strands
Reported transfer
lengths, in.
Release
Source Specimen type Nspec Strand layout fpi/fpu fci′, ksi method ltr,initial ltr,t t, days Notes
Tadros and
96 x 7 x 7 in. 4 One strand 0.75 6.0 G 26 to 28 26 to 28 28 Small-scale prism
Morcous1
One or two
Dang et al.9 216 x 6.5 x 12 in. 16 0.75 5.9 to 9.2 G 23 to 28 26 to 31 28 Small-scale prism
strands
Vadivelu10 AASHTO Type I 1 2.0 x 2.0 in. 0.75 10.0 F 21 to 22 — — —
T-girders 6 2.0 x 2.0 in. 0.75 9.0 G 19 to 25 21 to 26 28 —
Tadros and
NU900 1 2.0 x 2.0 in. 0.66 12.0 F 26 — — UHPC
Morcous1
BDT 1 2.0 x 2.0 in. 0.60 12.0 F 14 to 20 17 to 21 14 UHPC
Morcous et al. 11
NU900 1 2.2 x 2.3 in. 0.75 6.7 F 35 — — —
Morcous et al. 12
NU1350 2 *
2.0 x 2.0 in. 0.75 6.0 to 10.0 F 32 36 14 Bridge application
*
Morcous et al.12 reported transfer length measurements from two girders out of 20 fabricated girders.
Notes: Nspec is number of specimens tested; fpi is jacking stress; fpu is ultimate strength of prestressing steel; fci′ is compressive release strength; ltr,initial is transfer length immediately
after prestress transfer; ltr,t is transfer length at t; G is gradual release; F is flame cut; and BDT is bridge double tee; 1 in. = 25.4 mm; 1 ksi = 6.9 MPa.

such detailing for girders employing 0.7 in. (17.8 mm) diam-
eter strands needs to be investigated.
Relatively few experimental studies have addressed
prestress transfer or end-region behavior in girders
employing 0.7 in. (17.8 mm) diameter strands. A summary
of those studies is provided in Table 1. As can be seen in
this table, some of the studies were conducted on small-
scale specimens that were reinforced with only one strand.
Such studies do not provide a realistic representation of the
boundary conditions or of the interaction between adjacent
strands, and the applicability of results from such studies to
full-scale girders is questionable.
Full-scale specimens have been used in a few studies to
evaluate transfer length and constructability issues for girders
employing 0.7 in. (17.8 mm) diameter strands. The transfer
length, which was determined using mechanical measure-
ments of surface strains in all these studies, was reported to
be noticeably less than the estimates from ACI 318-14 and
AASHTO LRFD 2014. Very little information was reported
in any of these studies regarding the cracks developed within
the end regions of the specimens. Moreover, the magnitude of
bursting and spalling stresses was not measured in any of the
studies, providing little insight into any need to modify the
end-region detailing for use in girders with 0.7 in. (17.8 mm)
diameter strands. Furthermore, in most of these studies, the
prestress transfer was performed at a concrete release strength
considerably greater than what is commonly used in practice.
Therefore, the behavior of the specimens and the observed
transfer lengths may not be indicative of the performance of
actual pretensioned girders used in the field.
This paper presents an experimental research program on
full-scale pretensioned I-girders that were fabricated using
0.7 in. (17.8 mm) diameter strands in the controlled laboratory
environment. The compressive strength of concrete at the time
of release ranged from 5.2 to 8.3 ksi (35.8 to 57.2 MPa), and
the girders were extensively instrumented to monitor bursting
stresses, spalling stresses, and transfer length.
RESEARCH SIGNIFICANCE
The use of 0.7 in. (17.8 mm) diameter prestressing strands
might influence the serviceability and strength of preten-
sioned girders due to increased transfer length and greater
end-region stresses. The experimental study presented
in this paper is a major contribution to the knowledge
regarding the performance of full-scale girders employing
these larger-diameter strands at the time of prestress transfer,
in which different strand patterns, girder cross section sizes,
and concrete release strengths have been investigated. More-
over, this study presents the first set of data on bursting and
spalling stresses and progression of end-region cracking
within the first few weeks after the fabrication of girders
with 0.7 in. (17.8 mm) diameter strands.
EXPERIMENTAL PROGRAM
Four pretensioned Texas Bulb-Tee girders, commonly
referred to as Tx-girders, were fabricated using straight
0.7 in. (17.8 mm) diameter seven-wire strands that were
located on a 2 x 2 in. (51 x 51 mm) grid. These specimens
included two Tx46 and two Tx70 girders, each with a length
of 30 ft (9 m) and the cross-sectional geometry shown in
Fig. 1. The cross-sectional properties of the specimens are
listed in Table 2.
A summary of the design parameters for the specimens
is shown in Table 3. While the Tx46 specimens represent
the mid-sized bulb-tee cross sections that are widely used
in Texas, the Tx70 specimens represent the deepest girders
among this family of precast sections and have been shown
to likely experience the greatest benefit from using larger-
diameter strands.2 Different strand patterns were used among
the specimens, as shown in Fig. 2. The strands in Tx46-I
and the Tx70 specimens were placed conventionally—that
is, at the greatest possible eccentricity within the specimen
cross section—to generate the maximum spalling stresses
within the specimen end regions. In Tx46-II, the strands
were concentrated near the centroid of the cross section to
represent critical conditions for bursting stresses.

92 ACI Structural Journal/January 2018

Fig. 1—Standard detailing for mild-steel reinforcement in Tx-girders.13 (Note: Clear distance between first pair of S-bars and
end face: 1.5 in. [38 mm]; 1 in. = 25.4 mm.)

Table 2—Cross-sectional properties of Tx46 and Table 3—Specimen design parameters

Tx70 girders13 Tx46-I Tx46-II Tx70-I Tx70-II
Property Tx46 Tx70 Max. Max. Max.
Gross cross-sectional area, in.2 761 966 Design objective eccentricity Max. Pi eccentricity eccentricity

Distance from bottom fiber to center of gravity of Design fci′, ksi 5.5 5.2 5.5 7.8
20.1 31.9
girder, in. Top 4 4 4 4
No. of
Moment of inertia around x-axis, in.4 198,089 628,747 strands Bottom 24 30 28 42
Moment of inertia around y-axis, in.4 46,478 57,579 Top strands 44.0 44.0 68.0 68.0
Weight, lb/ft 819 1040 yp, in. Bottom
3.3 10.4 3.5 4.5
Notes: 1 in. = 25.4 mm; 1 in.2 = 645 mm2; 1 in.4 = 416,000 mm4; 1 lb/ft = 1.5 kg/m.
strands
Top strands 157.5 202.5 110.0 202.5
The specimens were designed based on AASHTO LRFD
fpi, ksi
Specifications.4 According to these specifications, the allow- Bottom
202.5 202.5 202.5 202.5
able stresses at the time of prestress transfer are 0.65fci′ in strands
compression and k√fci′ in tension, where fci′ is the compres- σpredicted, Top fiber 0.23 (T) 1.13 (C) 0.55 (T) 0.55 (T)
sive strength of concrete at the time of prestress transfer, ksi Bottom fiber 3.57 (C) 3.40 (C) 3.53 (C) 5.04 (C)
and k is 0.24 or 0.63, if fci′ is in ksi or MPa, respectively. In
Notes: Pi is initial prestressing force; yp is distance from bottom fiber of girder to
addition to the strands in the bottom flange, all specimens centroid of strands; fpi is jacking stress; σpredicted is predicted concrete stress after
included four 0.7 in. (17.8 mm) diameter strands within the prestress transfer; (C) is compression; (T) is tension;. 1 ksi = 6.9 MPa; 1 in. =
top flange to control the stresses within the cross section at 25.4 mm.

ACI Structural Journal/January 2018 93

Fig. 2—Strand patterns within bottom flange of specimens.
diameter, Grade 270 strands were procured from a commer-
cial prestressing steel manufacturer. Reusable strand chucks
were used for anchoring the 0.7 in. (17.8 mm) diameter
strands at each end, which had an external diameter of 2 in.
(51 mm). To remove the slack and ensure uniform stressing
of the strands, each strand was individually stressed to 2 kip
(9 kN), after which the main hydraulic rams were used to
“gang-stress” the strands—that is, stress all strands simul-
taneously. Gang-stressing of the strands was carried out
in a minimum of 10 increments. The prestress level was
controlled through measurements of hydraulic pressure in
the rams using pressure transducers. The elongation of the
strands was measured using a series of linear potentiome-
ters at each end of the prestressing facility and verified to be
within 5% of the calculated value after each increment. Once
Fig. 3—Prestressing facility at FSEL. the strands were stressed to the desired level, installation of
transverse reinforcement was carried out.
the time of prestress transfer. The stress level in these strands
Typically, each specimen contained 90 electrical resistance
was variable among the specimens to optimize the capacity
strain gauges (SGs) on the strands to measure the transfer
of the specimens and maximize the number of strands that
length. At each end of the specimen, four strands were
could be accommodated in the specimens. The end-region
instrumented, as shown in Fig. 4. Among the instrumented
and shear reinforcement detailing for the specimens followed
strands, the SGs were installed at two intervals—12.0 in.
the standard drawings developed by the Texas Department
(305 mm) and 6.0 in. (152 mm)—to examine any potential
of Transportation (TxDOT)13 for girders with 0.5 and 0.6 in.
changes in transfer length due to the presence of SGs. The
(12.7 and 15.2 mm) diameter strands, as shown in Fig. 1.
SGs were continued up to a distance of 60 in. (1.52 m) from
Specimens Tx46-I, Tx46-II, and Tx70-I were designed
the end face of the specimens. One SG was also installed
assuming concrete release strengths that are typically used
outside the specimen on each of the instrumented strands
for prestressed concrete superstructures in Texas. However,
to serve as a reference measurement representing the stress-
a greater concrete release strength was considered for the
free strain condition of the strand after prestress transfer. All
design of Tx70-II to investigate the behavior of girders fabri-
strand SGs were installed on the helical wires. An example
cated with a greater number of 0.7 in. (17.8 mm) diameter
of strain gauge application on a strand is shown in Fig. 5(a).
strands. As a result, Tx70-II could accommodate an addi-
Each specimen also contained a considerable number of
tional row of strands compared to Tx70-I.
SGs on the transverse reinforcement to determine bursting
The specimens were fabricated using the prestressing
and spalling stresses as well as strains at the time of future
facility at Ferguson Structural Engineering Labora-
structural tests. At a minimum, the first 15 stirrups (R-bars
tory (FSEL) (Fig. 3), which is designed to accommodate
in Fig. 1) from each end of the specimen were instrumented.
prestressing forces up to 2500 kip (11,120 kN). Steel bulk-
Three SGs were installed on each instrumented stirrup, as
heads measuring 12 in. (305 mm) thick are used at each
shown in Fig. 4. The stirrup leg that contained two SGs was
end of the prestressing facility to anchor the strands. The
alternated from one stirrup to the next. The initial installa-
bulkhead at one end is fixed, while the bulkhead at the other
tion of a SG on one of the stirrups, prior to applying protec-
end is supported by four 800 kip (3500 kN) hydraulic rams,
tive layers, is shown in Fig. 5(b).
which are extended for stressing the strands and retracted at
The specimens were also instrumented using three
the time of release. These two ends are herein referred to as
vibrating wire strain gauges (VWGs) that were embedded at
the dead and live ends, respectively. Two adjustable cross
the midspan of the specimen (Fig. 5(c)). Measurements from
beams are also used within the prestressing facility, making
VWGs were used to develop a strain profile at the midspan
it possible to apply prestressing forces of up to 300 kip
section and to estimate prestress losses due to elastic short-
(1330 kN) to top strands in different types of cross sections.
ening, shrinkage, and creep over the life of the specimen.
Fabrication of each specimen began by placing the 0.7 in.
To measure the hydration temperatures in the concrete, six
(17.8 mm) diameter strands in the prestressing facility and
thermocouples were placed in a section located within 2 ft
installing the anchorage devices. The 0.7 in. (17.8 mm)
(0.60 m) from the end face of the specimens. As shown in

94 ACI Structural Journal/January 2018

Fig. 4—SG locations for monitoring transfer length and transverse reinforcement stresses. (Note: 1 in. = 25.4 mm.)
Table 4—Summary of measured mechanical
properties
Test
Property method Tx46-I Tx46-II Tx70-I Tx70-II
ASTM
fci′, ksi 5.7 5.2 6.5 8.3
C3914
ASTM
Concrete Eci, ksi — 4940 4490 4900
C46915
ASTM
f ′c,28, ksi 7.3 7.0 10.7 12.7
C3914
fy, ksi 60.7 72.2
No. 4 bars
fu, ksi ASTM 99.3 111.6
fy, ksi A37016 74.0 67.7
No. 6 bars
fu, ksi 114.7 106.7
Ep, ksi 27,810
ASTM
Strands fpy, ksi 232.0
A106117
fpu, ksi 276.1

Notes: fci′ and Eci are compressive strength and modulus of elasticity of concrete at
prestress transfer, respectively; fc,28′ is 28-day compressive strength of concrete;
1 ksi = 6.9 MPa.

precast manufacturing plant and transported to FSEL. In all

Fig. 5—Instrumentation application.
Fig. 5(d), these thermocouples were distributed within the
cross section to capture the variability of concrete curing
temperatures within the end region of the girders.
The properties of concrete mixtures used within the test
program are provided in Appendix A.* Type III portland
cement and crushed limestone with a maximum aggregate
size of 3/8 in. (10 mm) were used in all mixtures. The clear
spacing between the prestressing strands was approximately
1.3 in. (33.0 mm), which exceeds 1.33 times the maximum
size of the aggregates, 1 in. (25.4 mm), and the strand
diameter. Therefore, the minimum spacing requirements of
ACI 318-14 and AASHTO LRFD 2014 were satisfied.
The mixtures used for the Tx46 specimens were batched
and mixed at FSEL. The concrete comprising the Tx70
specimens was batched and mixed by a nearby commercial
*
The Appendixes are available at www.concrete.org/publications in PDF format,
appended to the online version of the published paper. It is also available in hard copy
from ACI headquarters for a fee equal to the cost of reproduction plus handling at the
time of the request.
cases, a combination of internal and external vibration was
used to ensure satisfactory consolidation of concrete.
Forty-eight 4 x 8 in. (100 x 200 mm) match-curing concrete
cylinders were cast using the concrete comprising each spec-
imen. These cylinders were connected to a relay system that
maintained the same temperatures within the cylinders as
those measured from the embedded thermocouples. There-
fore, eight cylinders were match-cured based on each of
the six thermocouples shown in Fig. 5(d), which made it
possible to capture the variability of concrete strength within
the specimen cross section.
In the hours following the concrete placement, match-
cured specimens were periodically tested to identify the
appropriate timing for formwork removal and subsequent
prestress transfer. Prestress transfer commenced as soon as
the compressive strength of match-cured cylinders based
on all thermocouple measurements exceeded the desired
release strength. The prestressing strands were released
through gradual retraction of hydraulic rams in 20 steps,
which typically took 1 hour. As the strands were being
released, the compressive strength and modulus of elas-

ACI Structural Journal/January 2018 95

ticity of match-cured concrete specimens were measured to Table 5—Transfer length estimates from live and
obtain the mechanical properties at prestress transfer. The dead ends of specimens
average compressive strength and modulus of elasticity of Live end Dead end
the match-cured specimens at the time of prestress transfer
lplateau, ltr, lplateau, ltr,
are presented in Table 4.
in. in. n in. in. n
Along with the match-cured cylinders, additional cylin-
ders were prepared to measure mechanical properties of At Max 42 40 48 47
1 4
concrete at 28 days. Moreover, the mechanical properties of release Min. 42 40 42 40

Tx 46-I
the prestressing strands and the mild-steel reinforcing bars Max. 48 45 48 46
used in the construction of the girders were determined by 24 hour 1 3
Min. 48 45 42 41
performing ASTM-compliant tests at FSEL, which are also
presented in Table 4. Each mechanical property reported At Max. 42 35 48 41
4 4
in this table is an average value from a minimum of three release

Tx 46-II
Min. 36 29 36 29
samples tested by the research team. Max. 54 52 54 48
Data acquisition from the instrumentation commenced 24 hour 3 3
Min. 48 44 48 45
immediately prior to prestress transfer and was continued
for 24 hours after the end of the release operation. After the At Max. 42 39 42 41
4 3
release operation was completed, the specimens were care- release Min. 36 29 36 34

Tx 70-I
fully examined for end-region cracking. Max. 48 46 54 51
24 hour 3 2
Min. 36 31 36 35
RESULTS AND DISCUSSION
Transfer length At Max. 48 43 42 40
4 3
The transfer length was determined by comparing the data release
Tx 70-II Min. 36 34 36 34
obtained from the SGs before the release operation with those Max. 48 45 48 45
obtained immediately after release and 24 hours after release. 24 hour 2 3
Min. 48 45 48 41
To determine the transfer lengths, a modified version of the
95% average maximum strain (AMS) method introduced by At Max. 48 43 48 47
13 14
Summary

Russell and Burns6 was used. With the increase in distance release Min. 36 29 36 29
from the end face, the strains gradually increased from zero Max. 54 52 54 51
until reaching a plateau. For each end of each specimen, the 24 hour 9 11
Min. 36 31 36 35
strains in the plateaued region were averaged to determine
the average maximum strain. The transfer length is defined Notes: n is number of strands used for determining the transfer length; 1 in. = 25.4 mm.
as the distance at which the strain-versus-distance plot inter-
sects the 95% of AMS. The strains used in this procedure
were obtained from SGs that were installed on helical wires,
which are linearly correlated to the average axial strain in
the strand.18 Therefore, using these strains as opposed to the
average axial strain in the strand is not expected to affect the
transfer lengths.
Detailed information related to the evaluation of the transfer
lengths for the specimens is provided in Appendix B. Table 5
provides a summary of the distances corresponding to the start
of the plateau region, lplateau, and to the 95% AMS at dead and
live ends of each specimen. Because some of the strain gauges
did not function properly, determining the transfer length was
not possible for all instrumented strands. For each end region,
the number of strands from which a reliable transfer length Fig. 6—Comparison between measured transfer lengths and
could be determined is shown in the table. code predictions. (Note: 1 in. = 25.4 mm; 1 ksi = 6.9 MPa.)
In general, the transfer lengths obtained from live and which is consistent with the known effect of time-dependent
dead ends of each specimen were similar. This observation deformations of concrete on transfer length.19 At 24 hours
comes as no surprise because the gradual release of strands after release, the shortest and longest transfer lengths were
by hydraulic rams is expected to result in little difference recorded as 31 in. (790 mm) and 52 in. (1320 mm), obtained
between the live and dead ends of the specimens. Imme- from the live end of Tx70-I and the live end of Tx46-II,
diately after release, the shortest transfer length was 29 in. respectively.
(740 mm), which was obtained from both ends of Tx46-II. Figure 6 provides a comparison between the transfer
The longest transfer length at this time was 47 in. (1190 mm), lengths determined in this study with estimates from
which was found at the dead end of Tx46-I. AASHTO LRFD Specifications and ACI 318-14. The
Table 5 also shows a noticeable increase in the transfer vertical bars in this figure present the average transfer length
lengths measured within the first 24 hours after release, determined for each specimen at release and at 24 hours. The

96 ACI Structural Journal/January 2018

observed range between maximum and minimum transfer
lengths is also illustrated using the vertical black lines. The
transfer length calculation in ACI 318-14 is dependent on
effective stress in the strands after prestress losses and there-
fore varies among the specimens. The prestress loss due to
elastic shortening, ∆fES, was estimated using data obtained
from the VWGs and is shown in Table 6.
Figure 6 shows that the transfer lengths immediately
after release were generally shorter than those predicted by
ACI 318-14 and AASHTO LRFD provisions. However, after
24 hours, the transfer length in all specimens, except Tx70-I,
exceeded the estimates by both provisions. In Tx70-I, the
average 24-hour transfer length was very close to estimates
by both codes, but transfer lengths greater than code esti-
mates were also observed. A greater increase in the transfer
lengths is anticipated for all specimens over time. However,
monitoring the growth in transfer length after 24 hours was
not considered in this study.
The transfer lengths observed in this study were greater
than those reported in previous studies noted in Table 1. In
those studies, the overestimation of transfer length by the
design codes were deemed to be primarily due to the high
release strength of concrete, which is not explicitly consid-
ered in code equations. However, observations from Fig. 6
do not support that reasoning. As can be seen in the figure,
the transfer lengths immediately after and 24 hours after
release were not correlated with the release strength, indi-
cating that the improved bond due to greater compressive
strength is not the only factor affecting the transfer length.
The use of strain gauges in this study is believed to provide
a more precise picture of stress changes in the strands
compared to the mechanical measurements of surface
strains. However, the presence of strain gauges could also
potentially have a negative effect on the prestress transfer
within the instrumented strands. Thus, the transfer lengths
from strands instrumented at 6 and 12 in. (150 and 300 mm)
intervals were compared. The results showed that the
transfer lengths from strands with 6 in. (150 mm) spacing
were consistently greater than those with gauges placed
at 12 in. (300 mm) spacing. However, the difference was
generally limited to 6 in. (150 mm), which is the resolution
of the estimated transfer lengths. Therefore, the presence of
strain gauges does not seem to be the major contributor to
the greater transfer lengths measured in this study.
Cracking patterns
All specimens were carefully inspected for cracks imme-
diately after prestress transfer. Moreover, a final survey of
the specimens for cracks was conducted at least 28 days after
the specimen was cast. For Tx70-II, in addition to the initial
and final measurements, cracks were also measured and
documented at ages of 7 and 13 days. The age of the speci-
mens at the time of final crack measurements, t, is presented
in Table 6. This table also includes the prestress losses in the
bottom strands at the time of final measurements, ∆ftotal,t, as
determined from VWGs.
The measured cracking patterns of the specimens are
presented in Fig. 7. This figure shows both end faces but
only the side of each specimen containing the greatest crack
ACI Structural Journal/January 2018
Table 6—Short- and long-term prestress losses
Tx46-I Tx46-II Tx70-I Tx70-II
∆fES, ksi 21.5 18.9 24.1 30.7
t at final
131 40 39 28
measurement
∆ftotal,t, ksi 45.1 35.6 38.7 45.9
Note: 1 ksi = 6.9 MPa.
width observed. Similar figures reflecting the cracking
patterns on the other side of each specimen are included in
Appendix C. In Fig. 7, the small circles are used to present
the measured width of the cracks at each location at the
time of final crack measurements. In the regions where no
circles are shown, the crack width was less than or equal
to 0.004 in. (0.10 mm). The widths of the cracks in the first
three specimens were measured using a crack comparator
with a resolution of 0.002 in. (0.05 mm). For Tx70-II, a
7× magnifying loupe was used with reticles that provided
a crack measurement resolution of 0.0004 in. (0.01 mm).
However, for brevity and simplicity, crack widths shown for
Tx70-II are also categorized in a manner similar to that used
for the other specimens.
The three specimens constructed with a conventional
strand pattern revealed spalling cracks in their end regions.
As the prestressing force increased near the bottom fiber of
these specimens, the spalling cracks extended further into
the beam. Tx46-II, which was designed to accommodate
the greatest prestressing force, showed bursting cracks that
were primarily limited to the bottom flange. All specimens
showed near-continuous cracking at the interface between
the web and the bottom flange within the end region. In all
specimens except Tx46-II, such cracking was also observed
through the width of the web at the end faces. In Tx70-II,
cracks parallel to the outermost strands, which are indicative
of bond-related damage, were also observed.
The maximum crack width recorded over the course of
the entire test program was 0.008 in. (0.20 mm). This crack
width was observed immediately after prestress transfer in
the spalling cracks within the web of Tx46-I and Tx70-I.
Note that the crack widths in both specimens were recorded
with a resolution of 0.002 in. (0.05 mm). The maximum crack
widths in Tx46-II and Tx70-II were recorded as 0.004 in.
(0.10 mm) and 0.007 in. (0.18 mm), respectively. In Tx46-I
and Tx70-I, the reported maximum width was observed only
in an isolated length of few cracks. However, in Tx70-II,
crack widths between 0.006 and 0.007 in. (0.16 and
0.18 mm) were measured at several locations. The greatest
crack width was observed in the dead end of all specimens
except Tx70-I. However, the overall difference in patterns or
widths of cracks between the two ends was not significant.
All specimens demonstrated noticeable changes in their
cracking conditions over time. These changes included
growth in the length and width of cracks that were
detected immediately after prestress transfer, as well as the
development of new cracks, especially in the Tx70 speci-
mens. However, in most end regions, the widest crack did
not demonstrate noticeable growth between the time of
release and the time of final measurement. Periodic observa-
97
Fig. 7—Crack patterns and widths in specimens. (Note: 1 ft = 0.30 m; 1 mm = 0.039 in.)
tions from Tx70-II also showed that most of the new cracks
formed within the first week after the prestress transfer.
There is no globally accepted limit for the permissible
crack width within the end regions of pretensioned concrete
elements. To evaluate the performance of the specimens in
this test program, three references were used, as follows:
1. ACI 224R-01,20 a report by ACI Committee 224 on
concrete cracking, provides general guidelines on acceptable
crack widths in reinforced concrete flexural elements under
service loads.
According to these guidelines, the “reasonable” crack
width is 0.007 in. (0.18 mm) for elements exposed to deicing
chemicals and 0.012 in. (0.30 mm) for elements exposed to
humidity, moist air, and soil. Table 4.1 in these guidelines
notes that a portion of the cracks in the structure might have
widths that exceed these limits.
No cracks within the end regions of the girders exceeded
the limit recognized by ACI 224R as tolerable for humidity
and soil exposure. In Tx46-I and Tx70-I, isolated cracks
exceeding 0.007 in. (0.18 mm) were observed. However, as
noted previously, a few cracks with widths greater than the
listed limits are considered acceptable. Therefore, according
98
to ACI 224R, all girders comprising this test program met
the conditions for use in exposure to deicing chemicals.
2. NCHRP Report 654, a comprehensive study on the
acceptance criteria for the width of end-region cracks by
Tadros et al.21 recommends that no action be taken for any
end-region cracks that are 0.012 in. (0.30 mm) in width or
less. The observed crack widths in this study did not exceed
the recommended limit. Therefore, no repair is required
according to these guidelines.
3. TxDOT specifications for construction and maintenance
of highways, streets, and bridges22 require corrective action
if cracks exceeding 0.005 in. (0.13 mm) form within the end
regions of I-girders. While the specimens did not satisfy this
strict requirement, crack widths up to 0.010 in. (0.25 mm)
are occasionally observed in Tx-girders currently fabricated
using 0.6 in. (15.2 mm) diameter strands and containing
identical detailing for mild-steel reinforcement. Therefore,
the implementation of 0.7 in. (17.8 mm) diameter strands on
a 2 x 2 in. (51 x 51 mm) grid in Tx-girders does not appear
to cause new concerns regarding the serviceability or dura-
bility of pretensioned girders.
ACI Structural Journal/January 2018
Fig. 8—Stresses within end-region reinforcement of specimens. Gray shades in figure represent h/4 distance from end face.
(Note: 1 in. = 25.4 mm; 1 ksi = 6.9 MPa.)
Stresses in transverse reinforcement
Figure 8 provides a graphical representation of
stresses inferred from strain measurements in the transverse
reinforcement of all four specimens. Detailed information
regarding the stresses in the transverse reinforcement is
provided in Appendix D. As can be seen in Fig. 8, among
Tx46-I, Tx70-I, and Tx70-II, which had conventional strand
patterns, the increase in the transverse reinforcement stress
levels is correlated with the increase in the total prestress
force. The maximum stress level in the specimens was
26 ksi (179 MPa), which was observed in Tx70-II. The
greatest stresses in all specimens were observed at the inter-
face between the web and the bottom flange, mostly due to
the change in geometry and the flow of the stresses from
the bottom flange into the web. However, the stresses also
remained large at the centroids of these three specimens. In
ACI Structural Journal/January 2018
Tx46-II, which was designed to represent critical bursting
conditions, smaller stresses were detected compared to
Tx46-I, despite a greater prestressing force. Large stresses
were observed only at the interface between the web and the
bottom flange in this specimen.
In the specimens constructed with conventional strand
patterns, the stresses in transverse reinforcement were found
to diminish very quickly with distance from the end face of
the specimen. Stresses greater than 15 ksi (103 MPa) were
observed only in the first three stirrups, which were located
within 8.5 in. (220 mm) from the end face of the girder. Typi-
cally, diaphragms are not used at supports for Tx-girders.
However, in any bridge application, this distance is normally
in the overhang segment—that is, outside the main span.
Therefore, while excessive cracking in this region might
have negative effects on the durability of the girder, these
99
stresses are not expected to affect the load-carrying perfor-
mance under in-service conditions. Typical detailing used
for Tx-girders includes a 9 in. (230 mm) overhang segment,
and an 8 in. (200 mm) support width for Tx46 and a 9 in.
(230 mm) support width for Tx70 girders. Using this
detailing, the first four stirrups in the girders are not expected
to be mobilized under external loads. The stresses in other
stirrups were generally less than 10 ksi (69 MPa).
Tx46-II exhibited transverse stresses that extended over
3 ft (910 mm) into the beam. This distance was considerably
greater than that in the other three specimens. The stresses
were primarily concentrated along the web-bottom flange
interface, and only small stresses were detected at the centroid
depth and at the interface between the web and the top flange.
The maximum stresses were not observed in the first few
stirrups, but in the stirrups that were farther away from the
end face of the girder. Large stresses within this region are
known to increase the likelihood of horizontal shear distress.23
However, the magnitudes of these stresses were generally
limited to 10 ksi (69 MPa), which was small compared to
stresses observed in the other three specimens. These observa-
tions are consistent with the results obtained by O’Callaghan18
from Tx-girders that employed 0.6 in. (15.2 mm) diameter
strands on a layout resembling that of Tx46-II.
If the specimens with conventional strand patterns in
the current test program were to be fabricated using 0.5 or
0.6 in. (12.7 or 15.2 mm) diameter strands, more strands
would be needed to achieve the same flexural capacity.
Consequently, the distance between the centroid of strands
and the web-bottom flange interface would be decreased,
resulting in stress distribution similar to that in Tx46-II or
in girders reported by O’Callaghan.18 Such a change in the
configuration of the girder is expected to result in a greater
spread of transverse stresses and potentially large stirrup
stresses in the main span of the girder rather than in the
overhang region. Therefore, the use of 0.7 in. (17.8 mm)
diameter strands might be beneficial in containing the trans-
verse stresses due to prestress transfer within the overhang
region, therefore preventing these end-region stresses from
affecting the performance of the girder under shear-critical
loading conditions.
Section 5.10.10 in AASHTO LRFD4 describes require-
ments for resisting the tensile forces generated in the end
regions of pretensioned elements due to bursting and
spalling stresses. In these provisions, the resistance of the
end region against vertical tensile forces is referred to as
“splitting resistance”, and no distinction is made between
forces generated due to bursting and spalling stresses. Split-
ting resistance Pr is taken as fsAs,h/4, where fs is the stress in
steel, not to exceed 20 ksi (138 MPa); As,h/4 is the total area
of reinforcement within a distance of h/4 from the end of
the girder; and h is the height of the girder. AASHTO LRFD
provisions require this splitting resistance to be greater than
4% of the total prestressing force at transfer.
In this testing program, the h/4 distance is equal to 11.5
and 17.5 in. (292 and 445 mm) for the Tx46 and the Tx70
specimens, respectively. On the basis of the standard details
for Tx-girders, the steel required to provide the splitting
resistance is limited to the first four pairs of R- and S-bars in
100
Table 7—Splitting resistance in specimens
according to AASHTO LRFD4 compared to initial
prestressing force
Specimen As,h/4, in.2 Pr = fsAs,h/4, kip Pi, kip Pr/Pi, %
Tx46-I 5.12 102.4 1614 6.3
Tx46-II 5.12 102.4 2024 5.1
Tx70-I 7.68 153.6 1796 8.6
Tx70-II 7.68 153.6 2738 5.6
Notes: fs is taken as 20 ksi (138 MPa); 1 in.2 = 645 mm2; 1 kip = 4.4 kN.
Fig. 9—Comparison between transverse forces determined
in this study and those in database of bursting and spalling
stresses (adapted from Dunkman24 with additional data
points). (Note: 1 kip = 4.45 kN.)
the Tx46 specimens and the first six pairs of R- and S-bars in
the Tx70 specimens. Observations from Fig. 8 support this
selection. In specimens with conventional strand patterns,
noticeable stresses were observed in the reinforcement
located within the h/4 region, which is shown by the gray
shade in the figure.
Figure 8 shows that maximum detected stresses in the
Tx70 specimens in this test program exceeded 20 ksi
(138 MPa). However, because the stresses diminished
very quickly with distance from the end face, the average
stresses within the bars located in the first h/4 distance
from the end face was less than 20 ksi (138 MPa); there-
fore, all girders comprising this test program effectively met
the stress limit defined in AASHTO LRFD. It is important
to note that the area of mild-steel reinforcement provided
within the h/4 distance of Tx-girders considerably exceeds
the requirements in AASHTO LRFD. As shown in Table 7,
the use of closely spaced R- and S-bars within the end region
of Tx-girders results in a splitting resistance that ranges
between 5.1 and 8.6% of the initial prestressing force in
each specimen.
Figure 9 includes the transverse forces developed within
the distance of h/4 from the end face of the girders inves-
tigated in the current study and those from a database of
bursting and spalling stresses developed by Dunkman.24
This database includes the results from eight inverted-tee
specimens, four U-girders, and 53 I-girders fabricated with
0.5 or 0.6 in. (12.7 or 15.2 mm) diameter strands.
ACI Structural Journal/January 2018
 As can be seen in Fig. 9, the transverse force developed
over the distance of h/4 of all previous points in the data-
base was less than or equal to 0.04Pi, where Pi is the initial
prestress force in the bottom flange before elastic shortening
losses. However, in this study, noticeably greater bursting and
spalling forces were observed in all specimens except Tx46-II,
which had a nonconventional strand pattern. In both end
regions of Tx70-I, the magnitude of bursting and spalling
forces reached up to 5.7% of the initial prestressing force.
SUMMARY AND CONCLUSIONS
Four full-scale Tx-girder specimens employing 0.7 in.
(17.8 mm) diameter seven-wire strands on a 2 x 2 in. (51 x
51 mm) grid were designed and fabricated to investigate
their end-region behavior at the time of prestress transfer.
The specimens were extensively instrumented to provide a
detailed picture of transverse stresses within reinforcement,
transfer length, and prestress losses. Moreover, each spec-
imen was carefully examined for end-region cracking after
prestress transfer and at least 28 days after prestress transfer.
The primary conclusions regarding the end-region behavior
of the specimens were as follows:
Transfer length: The strain measurements from the
strands revealed that the 24-hour transfer length, which
is known to be less than the final transfer length, slightly
exceeded the transfer lengths predicted using AASHTO
LRFD and ACI 318-14 equations. While there is a possi-
bility that the presence of strain gauges contributed to an
increased transfer length, further research on full-scale spec-
imens is necessary to ensure accurate transfer lengths are
used for the design of girders that employ 0.7 in. (17.8 mm)
diameter strands.
Cracking patterns: The patterns of cracks developed
within the specimens were similar to those observed in
typical Tx-girders employing 0.5 or 0.6 in. (12.7 or 15.2 mm)
diameter strands. In all specimens, there was noticeable
cracking at the web-bottom flange interface. Most cracks in
the test program had a width that was less than 0.007 in.
(0.18 mm), with a width of 0.008 in. (0.20 mm) observed
in a few isolated locations in two specimens. Within the
first 28 days after prestress transfer, the length and width of
end-region cracks increased, and new cracks appeared in all
girders. However, the maximum width of the cracks within
each specimen did not show a noticeable change.
Stresses within transverse reinforcement: The greatest
stresses in all specimens were observed at the interface
between the web and the bottom flange. A maximum stress
level of 26 ksi (179 MPa) was observed in Tx70-II. In speci-
mens with conventional strand patterns, large stirrup stresses
were observed only in the overhang region of the girders.
For the nonconventional strand pattern used within Tx46-II,
the maximum stresses were not observed in the first few stir-
rups but in stirrups located further away from the end face
of the girder.
Necessity of modifications to detailing of end-region
reinforcing bars: Because the typical crack widths in the
specimens did not exceed those frequently observed in
girders with smaller-diameter strands, the use of 0.7 in.
(17.8 mm) diameter strands on the standard 2 x 2 in. (51 x
ACI Structural Journal/January 2018
51 mm) grid does not seem to trigger a need to modify the
end-region detailing in Tx-girders. Based on the guidelines
of ACI 224R-01, all specimens in this test program met the
criteria for use in exposure to deicing chemicals.
End-region reinforcement requirements: The trans-
verse force observed within the first h/4 length of specimens
with conventional strand patterns exceeded 4% of the initial
prestressing force. The relatively satisfactory performance
of the girders within this test program might be attributed
to the use of considerably greater amounts of transverse
reinforcement compared to that required by AASHTO
LRFD. The observations of this test program indicate a
potential need to reevaluate the requirements for end-region
reinforcement in AASHTO LRFD provisions when used for
girders employing 0.7 in. (17.8 mm) diameter strands.
A few alternatives for modifications to reinforcement
detailing in Tx-girders have been examined computation-
ally by Alirezaei Abyaneh25 to reduce the width of end-
region cracks, potentially below the strict 0.005 in. (0.13 mm)
limit required by TxDOT. Investigating the performance of
girders employing 0.7 in. (17.8 mm) diameter strands under
applied loads is beyond the scope of the current paper. The
shear behavior of such girders may be negatively affected by
diminished anchorage capacity due to greater transfer and
development lengths as well as the need to transfer a greater
horizontal shear force between the web and the bottom
flange. The development of near-continuous cracks at the
interface between the web and the bottom flange of the spec-
imens comprising this test program could also be a source
of concern regarding the transfer of horizontal shear. These
effects are evaluated in detail by Katz et al.3
AUTHOR BIOS
Jessica L. Salazar is a Bridge Designer with FIGG, Dallas, TX. She
received her BSCE from New Mexico State University, Las Cruces, NM, and
her MSE from the University of Texas at Austin, Austin, TX. Her research
interests include end-region behavior of prestressed concrete elements
and the effects of using larger-diameter strands on the performance and
economy of pretensioned concrete girders.
ACI member Hossein Yousefpour is an Assistant Professor at Babol
Noshirvani University of Technology, Babol, Iran. He received his
BSc and MSc from the University of Tehran, Tehran, Iran, and his PhD
from the University of Texas at Austin. His research interests include
prestressed concrete, structural health monitoring, and shear strength and
time-dependent behavior of concrete structures.
ACI member Roya Alirezaei Abyaneh is a Structural Engineer with
Building Diagnostics, Inc., Houston, TX. She received her BAppS from
the University of Waterloo, Waterloo, ON, Canada, and her MSE from the
University of Texas at Austin. Her research interests include prestressed
concrete and computational simulation of reinforced concrete structures.
Hyun su Kim is a PhD Student at the University of Texas at Austin. He
received his BSc and MSc from Seoul National University, Seoul, Korea, in
2008 and 2010, respectively. His research interests include shear strength
and end-region behavior of prestressed concrete girders.
Alex T. Katz is a Project Engineer with Walker Restoration Consultants,
Chicago, IL. He received his BS from the University of Virginia, Char-
lottesville, VA, and his MSE from the University of Texas at Austin. His
research interests include prestressed concrete and shear behavior of
concrete structures.
ACI member Trevor D. Hrynyk is an Assistant Professor in the Depart-
ment of Civil, Architectural and Environmental Engineering at the Univer-
sity of Texas at Austin. He is a member of Joint ACI-ASCE Committees 421,
Design of Reinforced Concrete Slabs, and 447, Finite Element Analysis of
101
Reinforced Concrete Structures; and ACI Subcommittee 445-C, Punching
Shear. His research interests include performance assessment of rein-
forced concrete structures, analysis and design of slabs and shells, and
nonlinear modeling.
Oguzhan Bayrak, FACI, is a University Distinguished Teaching Professor
in the Department of Civil, Architectural and Environmental Engineering
and holds the Charles Elmer Rowe Fellowship in Engineering at the
University of Texas at Austin. He is a member of ACI Committees 341,
Earthquake-Resistant Concrete Bridges, and E803, Faculty Network Coor-
dinating Committee; and Joint ACI-ASCE Committees 441, Reinforced
Concrete Columns, and 445, Shear and Torsion.
ACKNOWLEDGMENTS
The authors gratefully acknowledge TxDOT for financially supporting
this study, as well as Coreslab Structures in Cedar Park, TX, and Alamo
Cement for donating concrete and cement that were used for the fabrication
of the specimens. Thanks are also due to R. Boehm for her assistance with
the performance of the experimental program. The findings and opinions
presented in this paper are those of the authors and do not necessarily reflect
the views of TxDOT.
REFERENCES
1. Tadros, M. K., and Morcous, G., “Impact of 0.7 Inch Diameter Strands
on NU I-Girders,” Technical Report, Nebraska Department of Roads,
Lincoln, NE, 2011, 203 pp.
2. Salazar, J.; Yousefpour, H.; Katz, A.; Alirezaei Abyaneh, R.; Kim, H.;
Garber, D.; Hrynyk, T.; and Bayrak, O., “Benefits of Using 0.7 in. (18 mm)
Diameter Strands in Precast, Pretensioned Girders: A Parametric Investiga-
tion,” PCI Journal, V. 62, No. 6, Nov-Dec. 2017, pp. 59-75.
3. Katz, A.; Yousefpour, H.; Kim, H.; Alirezaei Abyaneh, R.; Salazar, J.;
Hrynyk, T.; and Bayrak, O., “Shear Performance of Pretensioned Concrete
I-Girders Employing 0.7 in. (17.8 mm) Strands,” ACI Structural Journal,
V. 114, No. 5, Sept.-Oct. 2017, pp. 1273-1284.
4. American Association of State Highway and Transportation Officials
(AASHTO), “AASHTO LRFD Bridge Design Specifications,” seventh
edition with 2015 and 2016 Interim Revisions, Washington, DC, 2016,
2150 pp.
5. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318R-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 2014, 520 pp.
6. Russell, B. W., and Burns, N. H., “Measured Transfer Lengths of 0.5
and 0.6 in. Strands in Pretensioned Concrete,” PCI Journal, V. 41, No. 5,
1996, pp. 44-65. doi: 10.15554/pcij.09011996.44.65
7. Comité Euro-International Du Beton (CEB), “Anchorage Zones of
Prestressed Concrete Members,” CEB Bulletin No. 181, 1987, 137 pp.
8. Hoyer, E., and Friedrich, E., “Beitrag zur Frage der Haftspannung in
Eisenbetonbauteilen [Contribution to the Question of Bond in Concrete
Elements],” Beton und Eisen, V. 30, No. 6, 1939, pp. 107-110.
9. Dang, C. N.; Floyd, R. W.; Hale, M.; and Marti-Vargas, J. R.,
“Measured Transfer Lengths of 0.7 in. (17.8 mm) Strands for Pretensioned
Beams,” ACI Structural Journal, V. 113, No. 3, May-June 2016, pp. 85-94.
102
10. Vadivelu, J., “Impact of Larger Diameter Strands on AASHTO/PCI
Bulb-Tees,” master’s thesis, University of Tennessee, Knoxville, TN, 2009,
114 pp.
11. Morcous, G.; Hanna, K.; and Tadros, M. K., “Use of 0.7-in. Diameter
Strands in Pretensioned Bridge Girders,” PCI Journal, V. 56, No. 4, 2011,
pp. 65-82. doi: 10.15554/pcij.09012011.65.82
12. Morcous, G.; Assad, S.; Hatami, A.; and Tadros, M. K., “Imple-
mentation of 0.7 in. Diameter Strands at 2.0 x 2.0 in. Spacing in Preten-
sioned Bridge Girders,” PCI Journal, V. 59, No. 3, 2014, pp. 145-158. doi:
10.15554/pcij.06012014.145.158
13. Texas Department of Transportation (TxDOT), “Concrete I-Girder
Details, Standard Drawings by the Bridge Division, Austin, Texas,” Austin,
TX, 2015, 2 pp.
14. ASTM C39/C39M-14, “Standard Test Method for Compressive
Strength of Cylindrical Concrete Specimens,” ASTM International, West
Conshohocken, PA, 2014, 7 pp.
15. ASTM C469/C469M-14, “Standard Test Method for Static Modulus
of Elasticity and Poisson’s Ratio of Concrete in Compression,” ASTM
International, West Conshohocken, PA, 2014, 5 pp.
16. ASTM A370-16, “Standard Test Methods and Definitions for
Mechanical Testing of Steel Products,” ASTM International, West Consho-
hocken, PA, 2016, 49 pp.
17. ASTM A1061/A1061M-16, “Standard Test Methods for Testing
Multi-Wire Steel Prestressing Strand,” ASTM International, West Consho-
hocken, PA, 2016, 4 pp.
18. O’Callaghan, M. R., “Tensile Stresses in the End Regions of Preten-
sioned I-Beams at Release,” master’s thesis, University of Texas at Austin,
Austin, TX, 2007, 265 pp.
19. Barnes, R.; Grove, J.; and Burns, N., “Experimental Assessment of
Factors Affecting Transfer Length,” ACI Structural Journal, V. 100, No. 6,
Nov.-Dec. 2003, pp. 740-748.
20. ACI Committee 224, “Control of Cracking in Concrete Structures
(ACI 224R-01),” American Concrete Institute, Farmington Hills, MI, 2001,
46 pp.
21. Tadros, M. K.; Badie, S. S.; and Tuan, C. Y., “Evaluation and Repair
Procedures for Precast/Prestressed Concrete Girders with Longitudinal
Cracking in the Web,” NCHRP Report 654, Transportation Research Board,
Washington, DC, 2010, 66 pp.
22. Texas Department of Transportation (TxDOT), “Standard Specifica-
tions for Construction and Maintenance of Highways, Streets, and Bridges,”
Austin, TX, 2014, 919 pp.
23. Hovell, C.; Avedano, A.; Moore, A.; Dunkman, D.; Bayrak, O.; and
Jirsa, J., “Structural Performance of Texas U-Beams at Prestressed Transfer
and under Shear-Critical Loads,” Technical Report, Center for Transporta-
tion Research, Austin, TX, 2011, 355 pp.
24. Dunkman, D., “Bursting and Spalling in Pretensioned U-Beams,”
master’s thesis, University of Texas at Austin, Austin, TX, 2009, 242 pp.
25. Alirezaei Abyaneh, R., “Computational Modeling of Prestress
Transfer, End-Region Cracks, and Shear Behavior in Prestressed Concrete
I-Girders Employing Large-Diameter Strands,” master’s thesis, University
of Texas at Austin, Austin, TX, 2016, 95 pp.
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S08
Stepwise Bond Model Including Unconfined and Partially
Confined Hooks
by Armin Erfanian and Alaa E. Elwi
A contact interface model including hook reinforcement under
monotonic, unconfined, in-plane compression and tension is
presented, which is in close agreement with experimental test
results. Unconfined in-plane compression accompanies unconfined
hooks under the bent portion of a hook and creates splitting in the
same plane as the hook. Although the effect of pressure on bond
capacity of unconfined concrete is investigated in the literature,
there is little information on reliable bond-slip behavior for this
case. fib 2010 does not provide bond-slip equations for unconfined
and partially confined concrete under compression, although it
considers the case of confined concrete under pressure. Laborious
effort was made in this research to formulate a procedure that
works in the three-dimensional finite element environment without
impractical limitations. In addition, a new method of computing
bond stress increments is proposed to be used in cases involving
slip-dependent normal pressure such as joints.
Keywords: bond; confinement; in-place pressure; partial confinement; slip;
splitting; stepwise.
INTRODUCTION
The present research aims at providing a research tool as
a substitution to lab experimentation. Therefore, this analyt-
ical tool should be able to simulate a range of configurations
regardless of whether they are currently common in prac-
tice. Likewise, this research tool should be able to discretize
the effects of influencing parameters on the behavior of the
structure. This paper is an effort to develop such a research
tool in the field of discrete bond modeling. In contrast, the
smeared crack and implicit modeling assumes full compat-
ibility of steel and concrete using tension stiffening to
account for bond and slip.1,2 This research uses the discrete
type of modeling.
One of the earliest finite element approaches with discrete
modeling of bond has been developed by Ngo et al.,3 who
modeled the bond between concrete and steel surfaces with
spring elements. Nilson4 proposed a curve for bond in terms
of slip and used springs to connect the nodes of steel and
concrete elements interacting by bond. Hofstetter5 reported
that Keuser et al.6 and Mehlhorn7 developed bond elements
that worked better than spring elements, while Miguel et al.8
used a combination of bond elements and spring elements
to model steel-concrete interaction. Lowes et al.9 proposed
a bond-slip model and an interface element for confined
concrete. fib 201010,11 presents discrete relationships for
bond and slip of unpressured unconfined and unpressured
partially confined concrete as well as pressure-free and
transversely pressured fully confined concrete. It is noted
that the term “confinement” is used for stirrup confinement.
ACI Structural Journal/January 2018
TECHNICAL PAPER
fib 201010,11 allows an increase of the design bond strength
for the case of transverse pressure perpendicular to the split-
ting plane. However, fib 201010,11 does not have a clear local
bond-slip relationship for splitting failure of unconfined
and partially confined concrete under transverse pressure,
whether perpendicular or parallel to the splitting plane. In
addition, fib 201010,11 does not provide an ultimate value
for the splitting failure slippage of unconfined and partially
confined concrete under compression beyond which bond is
zero. In other words, the bond-slip relationship for uncon-
fined and partially confined concrete in transverse pressure
is unknown in fib 201010,11 model code and models that use
fib 201010,11 as a basis such as the works of Zhang et al.,12 Xu
et al.,13 and Costa et al.14
The deficiency in fib-based models makes the modeling
of unconfined and partially confined hooks impossible.
Unconfined hooks are permitted by ACI 318-1415 if certain
requirements are met. Nonetheless, hooks in confined joints
meeting the requirements of ACI 318-1415 are not neces-
sarily fully confined according to fib 201010,11 local bond
model provisions. In some cases, confined joints complying
to ACI 318-1415 may meet the fib 201010,11 local bond model
requirements for full confinement. In the rest of the cases,
hooks in ACI-complying confined joints fall into the cate-
gory of partially confined concrete based on fib 201010,11
local bond model. Analysis of these partially confined
hooked joints cannot be accomplished using fib-based
models. This is because fib 201010,11 does not provide a clear
relationship for partially confined concrete under transverse
pressure, which is in this case produced under the bend.
In addition, finite element models similar to the one by
Zanuy et al.,16 which are able to account for unconfined
transversely pressured concrete behavior, use the cohe-
sive zone theory instead of the bond-slip theory, which is a
related but a different field of research.
Further, the reliability information so far published on fib
2010 model10,11 does not focus on discrete bond modeling17
of unconfined transversely pressured concrete. Accepting
the reliability of the fib-based models does not help create a
finite element model for unconfined concrete with transverse
pressure based on fib 2010.10,11
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2016-328.R2, doi: 10.14359/51700949, received March 19, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
103
 For the case of fully confined concrete, there is an abun-
dance of test results and complete bond-slip relationships
such as the experimental results by Eligehausen et al.18 and
Shima et al.,19 who provided confinement by stirrups and
massive covers, respectively. However, the case of uncon-
fined concrete is more complicated than the fully confined
case. For the bond capacity of unconfined concrete under
transverse pressure, however, because of large discrepan-
cies, some test results from previous researchers18,20-23 were
not used herein. In addition, there are reports on the slip for
unconfined concrete with and without transverse pressure at
or up to the peak load.20,22,24 These reports lack the complete
slip behavior up to failure for transversely pressured uncon-
fined and partially confined concrete. In addition, there exist
numerous experiments that result in bond-slip relationships
similar to the fib 2010 model10,11 for unconfined, trans-
versely unpressured, and fully confined concrete.25-29 Addi-
tionally, some tests were conducted to explain the confining
stresses produced by stirrups, distribution of cracks, and
crack-width-dependent bond-slip relationships for uncon-
fined concrete with zero transverse pressure.30,31
Gambarova32,33 has carried out numerous tests using
specimens with performed splitting cracks and a confining
frame that keeps the crack opening width constant. None-
theless, the confining frame in Gambarova’s32,33 experiment
eliminated the possibility of failure by splitting mode, as all
Gambarova’s32,33 specimens failed in pullout shearing mode.
Thus, Gambarova’s32,33 specimens cannot be considered
unconfined concrete, as the corresponding test results do
not represent the splitting mode failure of transversely pres-
sured unconfined concrete. Instead, Gambarova’s32,33 results
represent the case of fully confined concrete by means of a
confining frame and external transverse pressure.
The aforementioned discussion is important for the case
of unconfined transversely pressured concrete in general
and particularly for the case of unconfined concrete under
hook pressure. The reason is that not all values of transverse
pressure fully confine splitting. Specifically, the normal
pressure developed under a hook parallel to the splitting
plane does not contain the splitting failure, no matter how
large it becomes. Hence, because of these reasons, a bond-
slip relationship for the splitting failure of unconfined trans-
versely pressured concrete cannot be created according to
Gambarova’s32,33 test results.
Therefore, a finite element model for the analysis of
transversely pressured unconfined and partially confined
concrete cannot be constructed solely based on previous
experiments. This is especially true for the case of uncon-
fined hooks, where the pressure under the bend is parallel to
the splitting plane. In this case, either of the terms “in-plane
pressure” or “in-plane stress” is a more suitable term and is
suggested herein. Accordingly, this research presents a finite-
element-based bond model that accounts for bond capacity
and slip in transversely pressured unconfined and partially
confined concrete. In addition, the present model considers
the effect of in-plane tension, varying slip-dependent normal
pressure on the bars and hooks, and splitting of concrete for
both pressured and unpressured conditions in a three-dimen-
sional (3-D) finite element environment.
104
In this research, test results by Eligehausen et al.18 were
used for best-fitting purposes for the case of fully confined
concrete. Eligehausen et al.18 carried out a comprehen-
sive experimental program consisting of 125 specimens of
confined concrete with and without transverse pressure. In
contrast to using steel tubes or large concrete covers, the
confinement was provided by stirrups and vertical rein-
forcement simulating practical situations. This is the most
comprehensive set of tests for confined concrete under trans-
verse pressure available so far. These test results have been
referenced and used by numerous researchers in the past. A
recent implementation of Eligehausen et al.18 test results was
done by fib 2010,10,11 which largely adapted these test results
for bond-slip modeling.
Further, the present model proposes a new method of
computing the increments of bond other than the sole use
of the slope of the bond-slip curve for the cases involving
slip-dependent normal pressure. The procedure to imple-
ment the present model is explained along with a flowchart
for efficient application. Experience has shown using an
algorithm that updates bond stresses incrementally with
respect to the current contact interface configuration is more
efficient in terms of convergence to accurate results, espe-
cially for models including hooked bars. Thus, the proposed
approach is updated in each analytical step to improve its
precision. Additionally, in the present approach, bond-slip
modeling of the splitting for unconfined in-plane pressure
has priority because using a proper concrete material model
with 3-D elements eliminates the necessity to model the
shearing failure explicitly. An experimental test was carried
out in this research to help in development and optimiza-
tion and verification of the model. Further, the present model
was validated through comparisons with other researchers’
experimental and analytical results.
RESEARCH SIGNIFICANCE
An efficient finite element bond model is presented for
discrete modeling of bond at the contact interface of steel
and concrete for monotonic loading. The proposed model
produces accurate results in different geometrical and
confinement conditions. In contrast to other models, the
present model considers unconfined in-plane pressure and
tension for hooks in a stepwise analysis. Using the present
model along with 3-D finite element package enables
researchers to simulate controlled test conditions to study
the behavior of reinforcing bars without building the actual
specimens in a laboratory, which helps save a considerable
amount of time and resources.
EXPERIMENTAL TEST
To investigate the behavior of unconfined concrete under
in-plane hook pressure, the specimen of Fig. 1 was tested in
a universal servo-controlled machine. The loading method
was displacement control with an approximate rate of
loading of 0.001 mm/s (3.94 × 10–5 in./s). Because a joint
is a disturbed region, the strut-and-tie method was used to
design it. Weldable, low-alloy, Type 400W bars conforming
to CSA G30.18-M92 with the nominal yield stress, fy of
400 MPa (58 ksi) were used. The actual yield stress fya, based
ACI Structural Journal/January 2018
Fig. 1—Present experimental test specimen. (Note: Dimensions are in mm; 1 mm = 0.0394 in.)
on coupon tests, was 450 MPa (65.25 ksi). The compressive
strength of concrete, fc′, was 35 MPa (5.07 ksi). The diam-
eters of the tension longitudinal and shear reinforcement
were 35 and 15 mm (1.38 and 0.6 in.), respectively. There
were two layers of 20 mm (0.78 in.)-diameter longitudinal
compression bars. The tension longitudinal reinforcement is
bent by 90 degrees in the middle with a diameter of 254 mm
(10 in.) and is hooked by 180 degrees at both ends with the
same diameter as the middle bend.
The present specimen was designed for a capacity up to
yielding of bars at the face of the joint. However, a hook
development length of approximately 60% of the full length
Fig. 2—Failed experimental test specimen.
was provided to ensure bars cannot fully develop yield
strength and, therefore, they have to fail by the loss of bond. load-deflection and stress diagrams from this test are used
Thus, a study of the effect of unconfined in-plane pressure to develop bond-confinement and bond-slip relationships.
on bond capacity becomes possible because other modes These diagrams are also used to make comparisons of the
of failure have been eliminated. Unconfined in-plane pres- developed model with the present experiment.
sure is single-sided normal pressure produced in the bend
region, resulting in splitting in unconfined concrete in the INTERFACE BOND MODEL
same plane as the hook. Thus, at the time, one face of steel is Assuming the local bond stress can be presented as a func-
under normal pressure the other is not. tion of slip and an externally applied constant normal pres-
The failure, shown in Fig. 2, was by splitting of the concrete sure, the i-th increment of the bond stress would be equal
at the joint region at the peak load and, as a result, the side to the product of a parameter iΓs and the increment of slip,
covers spalled off and then a descending branch occurred where iΓs would be the slope of the bond-slip relationship.
on load versus loading end displacement diagram in Fig. 3. However, in a more practical case of a hook pullout test, the
The strains were measured by a set of strain gauges placed normal pressure is applied internally as a consequence of slip-
on the bars from the face of the joint to the beginning of the page and depends on the slip. In this case, using the product
bend. Figures 4 and 5 show the values of average strains at of iΓs and the slip increment to compute the bond stress incre-
the beginning of the bend and at the face of the joint. The ment is not a complete solution and renders erroneous results.

ACI Structural Journal/January 2018 105


Fig. 3—Comparison of present model and with present test
results.
Fig. 4—Present model versus present test for stresses at
beginning of bend.
Fig. 5—Present model versus present test for stresses at face
of joint.
Therefore, to have a mathematically correct solution, a new
additional parameter iΓn is introduced in this research, which
is associated to the changes in slip-dependent normal stress.
Equation (1) presents the preceding discussion
τ = iΓs · is + iΓn · in (1)
i
where iτ and is denote the incremental values of the bond
stress τ and slip s. The normal stress parameter in is equal
to the increment of the normal stress parameter n, which is
defined as the ratio of the of normal stress to the compressive
strength of concrete, fc′. The normal stress parameter depends
on the steel-concrete contact surfaces. When the local normal
stress σl is compressive or zero, the normal stress parameter
is computed as the stated ratio. In contrast, when the local
106
normal stress is in tension, the normal stress parameter would
be equal to zero, as stated in the following equation
 σl
 σl ≥ 0
n =  f c′ (2)
0 σl ≤ 0

Likewise, the bond stress is dependent on the state of the
two contacting surfaces. However, the bond stress depends
on the magnitude of the local normal stress as well. Tensile
normal stress exceeding the interface chemical adhesion
results in zero bond stress. This is because the contacting
surfaces are separated. Further, if the local normal tensile
stress does not exceed the chemical adhesion capacity, the
bond stress equals to the accumulation of all bond stress
increments up to the current state of analysis. The following
equation expresses this concept
Σ i τ σl ≥ −m
τ= (3)
0 σl ≤ −m
where μ is the normal chemical adhesion stress capacity.
Chemical adhesion stress parallel to the bars can be found in
the literature.34 However, in the present research, the chem-
ical adhesion stress is used in the direction perpendicular to
the steel-concrete interface. The present treatment of tension
helps maintain the stability of some cases of analysis with
in-plane tension in bent bars. In-plane tension can occur
in bent bar conditions when the contact between steel and
concrete is lost on the outside face of a hook or inside face
of a hook tail near the failure load, or when displacements
become large and the contacting surfaces start to separate.
To distinguish between full, partial, and no confinement, a
new measure Λ is introduced. In cases of full and no confine-
ment, Λ would be 1 and 0, respectively. Thus, any values
between 0 and 1 represent a partially confined case. There-
fore, as the volumetric ratio of confining reinforcement
ρ becomes larger, the parameter Λ gets closer to 1. Like-
wise, reducing the confinement volumetric ratio results in
Λ getting closer to 0. Considering f(ρ) as a function of ρ, an
expression such as (1 – e–f(ρ)) would be asymptotic to 1 and
0 for large and small values of f(ρ), respectively. Therefore,
the following equation applies
Λ = 1 – e–f(ρ) (4)
f(ρ) = 10(ρ ·ρs–1)1.5 (5)
Having the values of bond capacities for unconfined
concrete under normal pressure, τupc, and confined concrete
under normal pressure, τcpc, the bond capacity for partially
confined concrete under normal pressure is obtained as follows
τec = τcpcΛ + τupc(1 – Λ) (6)
ACI Structural Journal/January 2018
 As a comparison, in the fib 2010 model,10,11 computation
of the bond capacity for partially confined concrete under
normal pressure is not provided. Considering the concept
of Eq. (1) and (6), Fig. 6 illustrates the advantage of using
Eq. (1), proposed in this research, for the computation of the
bond stress incrementally. It is clear in Fig. 6 that the imple-
mentation of Eq. (1) is more precise than the conventional
application of the total values of slip and normal pressure
into a bond-slip relationship. For simplicity of comparison, in
Fig. 6, a hypothetical bilinear bond-slip relationship is shown.
It is expected that in every step of the analysis at different
locations, different normal pressures and, consequently,
different bond capacities, exist. The proposed bond capacity
for unconfined concrete under normal pressure is given by
Eq. (7), which is obtained by a large number of trials and
optimization of the parameters on the finite element model
of the present specimen bent bar region, which is stirrup-free
but develops unconfined in-plane normal pressure. Fig. 6—Incremental versus total computation of bond with
hypothetical bilinear bond-slip relation.
τupc = τuc(α1n3 + α2n2 + α3n + 1) (7)
η = 2.4τucΛ(3β1n2 + 2β2n + β3) + τuc(1 – Λ)(3α1n2 + 2α2n + α3)
where τuc is the unconfined bond capacity. Parameters α1, α2, (11)
and α3 are optimized as 0.4086, –2.4855, and 3.4769, respec-
tively. The unconfined bond capacity τuc, which is dependent where
on the cover to bar diameter ratio and the tensile strength
of concrete, can be obtained from the literature.36 Equations τ ec
sm = (12)
(1) and (7) make hook and bent bar analysis possible for κ
unconfined concrete where there is normal pressure parallel
to the splitting failure at the inside face of the bent bar. This sm2 = c1(sm – sc) + sc (13)
task cannot be accomplished with the fib 2010 model,10,11 as
it does not provide an equation to account for τupc, nor does su = [c2(sm – sc) + sc](1 – Λ)–1 (14)
it provide a relationship of bond with slip for unconfined
concrete under normal pressure. In the aforementioned equations, splitting of uncon-
To model the behavior of confined concrete under pres- fined and partially confined concrete under pressure are
sure, Eq. (8) is proposed by the authors best-fitting the modeled explicitly, and su is the amount of slip beyond
experimental test results by Eligehausen et al.18 which bond is zero. For fully confined concrete, su generates
a very large number before which the shearing failure will
τcpc = 2.4τuc(β1n3 + β2n2 + β3n + 1) (8) happen. However, there is no need to model the shearing
failure in the present approach because the shearing mode
where parameters β1, β2, and β3 are best-fitted to 2.98, –3.75, is not a bond failure, but a concrete failure. In other words,
and 1.61, respectively. It is obvious that n has a limit, but using an appropriate material model for concrete, when the
there is no need to enforce an explicit limit to Eq. (7) and concrete at the interface reaches its critical value of shearing
(8). Since using an acceptable material model, the concrete deformation, it fails even without explicit modeling of the
under the hook will crush under extremely large values of shearing mode.
normal pressure. As the result of parameter optimization, κ is a constant that
Here, parameters iΓs and iΓn are defined is calibrated as 250 N/mm3 (9.2 × 105 lbf/in.3) to produce an
ascending branch as close as possible to the load-deflection
κ s ≤ sm diagram of Fig. 3. This value results in acceptable analytical
0 responses for normal-strength concrete. The constants c1, c2,
 sm < s ≤ sm 2
and sc are obtained as 1.468, 21.11, and 0.018 mm (7.09 ×
i
Γ s =  τ ec (9) 10–4 in.), respectively, by calibration of the parameters over
 s − s (1 − Λ ) sm 2 < s ≤ su
a large number of analyses of the presented specimen.
 m2 u
0 s > su Figure 7 presents the flowchart for the proposed procedure
implemented to each analysis step. In this flowchart, the
area associated with each node is specified first. Then, the
0 s ≤ sm orientation of the local frame of reference is given using the
η sm < s ≤ sm 2 direction cosines of the tangent to the bar. The increments of
 normal pressure and slip are computed next.
i
Γn =  η
 η + s − s ( s − sm 2 )(1 − Λ ) sm 2 < s ≤ su (10)
 m2 u

0 s > su
ACI Structural Journal/January 2018 107
Fig. 7—Flowchart.
The normal pressure is considered proportional to the
amount of penetration by the steel surface into the concrete
surface. If the reference coordinate system for the inter-
face is different from the local coordinate system, which is
usually the case, a conversion needs to be made to obtain the
slip and its increment in the local frame of reference using
the direction cosines of the interface reference frame and the
local reference system assigned at the beginning of the flow-
chart. A proper contact algorithm should be used wherein
the interface coordinate system is updated with the contact
surface deformations.
The next step is to calculate the bond capacity based on
the values obtained in preceding steps. Next is to compute
the increment of bond stress depending on the increments of
slip and normal stress. The last stage is to store the current
values of bond in the local coordinate system for every loca-
tion, as they will be used in the next step.
VALIDATION OF MODEL
The presented model and algorithm was used to analyze
different specimens with straight and bent bars from present
and past research. Experimental specimens from the present
researchers Untrauer and Henry,22 Lormanometee,20 Elige-
hausen et al.18 and Minor and Jirsa37 were considered for
validation. In addition, analytical models from Luccioni et
al.,38 Lowes et al.,9 Desir et al.,39 Lettow,40 and Mendes and
Castro41 were used for comparison. Table 1 shows the cases
studied as well as the comparison of test and analysis results.
108
Furthermore, explicit dynamic analysis in quasi-static form
was carried out using the software ABAQUS.42 In all cases,
3-D solid elements were implemented to model concrete and
main tension steel while embedded truss elements were used
to model stirrups and compression bars. In addition, tension
steel and concrete are connected through a contact interface.
Further, the present interface bond model was programmed
into the software as a user subroutine. Nonetheless, the
damaged plasticity material model43,44 was implemented
for concrete. In contrast to the bond action, the formation
of cracks was not modeled discretely. In other words, cracks
were indirectly considered through the concrete material
model. To eliminate the analysis sensitivity to the mesh
size, the fracture energy was divided by the characteristic
length.43,45 Variable mass scaling42 was implemented to
limit the stress wave transit of a time step to less than the
smallest element dimension. The stable time steps used in
this procedure are shown in Table 2 along with other model
configurations. The increments of the prescribed displace-
ment are generally small, as shown in Table 2. This type of
analysis is not iterative. Additionally, symmetry of speci-
mens is accounted for by modeling half of the specimens
with respect to each plane of symmetry and applying the
appropriate boundary conditions on the planes of symmetry.
For example, Fig. 8 shows the finite element mesh used to
model the present specimen.
In this research, priority is given to the modeling of
unconfined concrete and, thus, most of the verification cases
ACI Structural Journal/January 2018
Table 1—Specimens modeled
Case Specimen Type db/dba fc′/fca′ fy/fya n0 Pmax-Test, kN (kip) Pmax-FE, kN (kip) Pmax-FE/Pmax-Test
1 Present specimen J 1.40 1.17 1.00 0.00 970.00 (211.50) 921.00 (207.00) 0.95
2 36625 22
SP 0.75 0.83 1.59 0.25 118.26 (26.61) 111.89 (25.18) 0.95
3 46600 22
SP 0.75 1.07 1.59 0.00 81.87 (18.42) 80.05 (18.01) 0.98
4 4762022 SP 0.75 1.10 1.59 0.20 148.28 (33.36) 127.36 (28.65) 0.86
5 6-6-52-0-A420 SP 0.75 1.19 1.00 0.00 80.05 (18.01) 87.33 (19.65) 1.09
6 6-6-52-48-A2 20
SP 0.75 1.19 1.00 0.48 172.84 (38.89) 179.21 (40.32) 1.04
7 6-6-52-24-A6 20
SP 0.75 1.19 1.00 0.24 171.93 (38.68) 154.65 (34.80) 0.90
8 6-6-39-0-D120 SP 0.75 0.90 1.00 0.00 73.68 (16.58) 71.86 (16.17) 0.98
9 9-6-41-0-G120 SP 1.30 0.95 1.00 0.00 85.19 (19.17) 84.23 (18.95) 0.99
10 9-6-48-43-J6 20
SP 1.30 1.10 1.00 0.44 255.58 (57.51) 212.98 (47.92) 0.83
11 1.4.2 9,18,38,39
SP 1.00 1.00 1.33 0.00 60.80 (13.68) 55.74 (12.54) 0.92
12 6.49,18,40,41
SP 1.00 1.00 1.33 0.45 172.28 (38.76) 172.28 (38.76) 1.00
13 7-8.5-90-1.5 A37 HP 0.88 1.27 1.00 0.00 155.25 (34.90) 150.59 (34.98) 0.97
14 7-8.5-90-1.5 B 37
HP 0.88 1.33 1.00 0.00 155.25 (34.90) 155.25 (34.90) 1.00
15 7-8.5-90-2.0 A 37
HP 0.88 0.87 1.00 0.00 168.45 (37.87) 178.56 (40.14) 1.06
16 7-8.5-90-2.0 B 37
HP 0.88 1.08 1.00 0.00 181.84 (40.88) 198.20 (44.55) 1.09
17 5-4.5-90-1.5 A37 HP 0.63 1.24 1.00 0.00 86.29 (19.40) 80.25 (18.04) 0.93
18 9-8.3-90-3.0 A 37
HP 1.13 0.76 1.00 0.00 236.24 (53.11) 229.16 (51.52) 0.97
19 9-8.3-90-4.5 A 37
HP 1.13 0.62 1.00 0.00 191.50 (43.05) 177.15 (39.82) 0.93
20 Present Model HP 1.18 1.00 1.00 0.00 282.74 (63.56 )
* *
279.91 (62.93) 0.99
*
Capacity from ACI 318-14.
Notes: db is bar diameter; Pmax-Test is maximum load from test; Pmax-FE is maximum load from analysis; db is bar diameter; dba = 25.4 mm (1 in.); fca′ = 30 MPa (4.35 ksi); fya = 400 MPa
(58 ksi); n0 is external pressure divided by fc′; J is joint; SP is straight bar pullout; HP is hook pullout.

involve unconfined concrete. Therefore, except for Elige-
hausen’s Specimen 6.418 and the straight arms of the present
specimen outside the joint, which are confined, the rest of
the specimens are unconfined. In the analysis, the nominal
yield stress of steel and compressive strength of concrete
given by the corresponding researcher is used.
Figures 3 to 5 and Fig. 9 to 15 illustrate the stated compar-
isons. Figure 14 compares the present model results with
experimental and analytical results from other researchers.
This figure presents the analysis for unconfined concrete
without transverse pressure. It is evident from Fig. 14 that
the present model results are in close agreement with the
experimental results. A few other presented cases of compar-
ison include unconfined concrete under transverse pressure
including unconfined hooks. These analysis cases were
compared to experimental results. Currently the authors do
not know of any other bond-slip model capable of modeling
unconfined concrete under transverse pressure, in-plane
and out-of-plane. In Fig. 15, the present model exhibits a
softer behavior than the experiment at large slip values.
This is because the shearing failure in the present model
is dependent on the material properties. In other words, an
improvement is expected if the exact material curves from
the test are applied to the analysis. However, some of the
material properties including the stress-fracture energy
curve have been approximated according to the literature.45
This is because of the unavailability of such data from the
test results. Therefore, the present approach is sensitive to
material properties at large slip values associated with the
pullout shearing failure. Overall, the comparisons in the
aforementioned tables and figures show that the present
model renders results with close agreement to the present
and past experiments. Figure 16 shows an unconfined hook
pullout simulation created with the present model compared
with the hook capacity from ACI 318-14.15 For the model in
Fig. 16, the required hook development length for yielding of
steel is provided according to ACI 318-14,15 with two hooks
spaced at 280 mm (11 in.) and top, side, and hook tail back
clear covers of 68, 45, and 100 mm (2.68, 1.77, and 3.94 in.),
respectively. Bar size and material properties for the model
of Fig. 16 are given in Table 1, as load and strength reduction
factors are set to unity and the boundary conditions were
applied in the same manner as Minor and Jirsa37 enforced
them. It is clear from Fig. 16 that the predicted capacity from
the present model and ACI 318-1415 are in close agreement.
SUMMARY AND CONCLUSIONS
A stepwise constitutive model for the bar-concrete inter-
face is proposed THAT accounts for all influential param-
eters, including bond-slip for unconfined and partially
confined in-plane pressure and tension. The model is
suggested for 3-D finite element step-by-step analysis to
enable researchers to simulate laboratory experiment results
without expending the associated time and resources. The
proposed equations are to be used in conjunction with a
contact algorithm. An efficient procedure on how to imple-

ACI Structural Journal/January 2018 109

Table 2—Configurations of specimens
Δδp × 104,
CN CET Nec Nnc Dnc SET Nes Nns Dns Δts × 104, s Δt × 104, s t, s Ns × 10–2 mm (in.) δp, mm (in.)
1 B/W 3062/888 8/6 3/3 B/T 82/172 8/2 3/3 5.0 3.0 10.0 200 4.20 (0.17) 14.00 (0.55)
2 W 372 6 3 B 12 8 3 5.0 3.0 10.0 200 0.38 (0.01) 1.25 (0.049)
3 W 372 6 3 B 12 8 3 5.0 3.0 10.0 200 0.18 (0.01) 0.60 (0.024)
4 W 372 6 3 B 12 8 3 5.0 3.0 10.0 200 0.27 (0.01) 0.91 (0.036)
5 W 372 6 3 B 12 8 3 50.0 5.0 10.0 200 0.48 (0.02) 0.95 (0.038)
6 W 372 6 3 B 12 8 3 50.0 5.0 10.0 200 1.00 (0.04) 2.00 (0.078)
7 W 372 6 3 B 12 8 3 50.0 5.0 10.0 200 1.00 (0.04) 2.00 (0.078)
8 W 372 6 3 B 12 8 3 50.0 5.0 10.0 200 0.35 (0.01) 0.69 (0.027)
9 W 372 6 3 B 12 8 3 50.0 5.0 10.0 200 0.30 (0.01) 0.60 (0.024)
10 W 372 6 3 B 12 8 3 50.0 5.0 10.0 200 0.48 (0.02) 0.96 (0.038)
11 W 6200 6 3 W 40 6 3 10.0 3.0 20.0 200 0.60 (0.02) 4.00 (0.16)
12 W 6200 6 3 W/T 40/76 6/2 3/3 10.0 3.0 20.0 200 1.38 (0.05) 9.16 (0.36)
13 W 5130 6 3 W 38 6 3 50.0 50.0 5.0 10 8.50 (0.33) 0.85 (0.33)
14 W 5149 6 3 W 38 6 3 50.0 50.0 5.0 10 10.00 (0.39) 1.00 (0.039)
15 W 5149 6 3 W 38 6 3 50.0 50.0 5.0 10 13.70 (0.54) 1.37 (0.054)
16 W 5149 6 3 W 38 6 3 50.0 50.0 5.0 10 12.20 (0.48) 1.22 (0.048)
17 W 5149 6 3 W 38 6 3 50.0 50.0 5.0 10 14.70 (0.58) 1.47 (0.058)
18 W 5149 6 3 W 38 6 3 50.0 50.0 5.0 10 13.70 (0.54) 1.37 (0.054)
19 W 5149 6 3 W 38 6 3 50.0 50.0 5.0 10 10.90 (0.43) 1.09 (0.043)
20 W 7332 6 3 W 60 6 3 20.0 4.0 5.0 25 2.80 (0.11) 3.50 (0.14)

Notes: CN is case number; CET is concrete element type; Nec is number of concrete elements; Nnc is number of nodes per concrete element; Dnc is degrees of freedom per concrete
node; SET is steel element type; Nes is number of steel elements; Nns is number of nodes per steel element; Dns is degrees of freedom per steel node; Δts is stable time step; Δt
is time increment; ts is total time period; Ns is number of steps; Δδp is increment of prescribed displacement; δp is total prescribed displacement; B is brick element; W is wedge
element; T is truss element.

Fig. 9—Present finite element approach versus

Fig. 8—Finite element mesh of present specimen.
ment the relationships is presented. Although the bond-slip
parameters were optimized for the present experiment, it is
shown that they work well for other cases without modifi-
cation. The present procedure leads to accurate results for
different types of reinforcement configurations. Compari-
sons have been made between this work and experiments
from the literature. These comparisons render close agree-
Lormanometee’s20 test results.
ment between the presented approach and test results. A
flowchart of the present procedure is also demonstrated.
A sequel to this research applies the present model to a
parametric study where the response of bars in specific
bent and straight configurations independent of stirrups is
studied along with the assessment of the corresponding ACI
318-1415 provisions and the recommendation of appropriate
modifications. It is imperative in a parametric study to assess
the influence of each parameter on the behavior of structure
independently of other parameters, and the present bond
model provides a means to that purpose. In other words, this

110 ACI Structural Journal/January 2018

Fig. 10—Present finite element approach versus
Lormanometee’s20 test results.
Fig. 14—Comparison with Eligehausen’s test18 for uncon-
fined concrete.

Fig. 11—Present finite element approach versus

Lormanometee’s20 test results.

Fig. 15—Comparison with Eligehausen’s test18 for fully

confined concrete.

Fig. 12—Present finite element approach versus Minor’s37

test results.
Fig. 16—Comparison of the present hook pullout model with
ACI 318-14.15
parametric study could not be completed without the imple-
mentation of the present model.

AUTHOR BIOS
Armin Erfanian is a Senior Structural and R&D Engineer at AETECH
Consulting, Calgary, AB, Canada. He received his BS from Amirkabir
University of Technology, Tehran, Iran; his MS from Ferdowsi University
of Mashhad, Mashhad, Iran; and his PhD from the University of Alberta,
Edmonton, AB, Canada. His research interests include anchorage of bars
in reinforced concrete and advanced finite element modeling and analysis.

Alaa E. Elwi is a Structural Engineering Professor Emeritus at the Univer-

sity of Alberta. He received his BS from the University of Cairo, Cairo,
Egypt; and his MS and PhD from the University of Alberta. His research
Fig. 13—Present finite element approach versus Minor’s37 interests include bond in reinforced concrete, composite structures,
test results. dynamics of structures, and advanced finite element method.

ACI Structural Journal/January 2018 111

 ACKNOWLEDGMENTS
The authors wish to express their gratitude and sincere appreciation to the
Natural Sciences and Engineering Research Council of Canada and Cement
Association of Canada for funding this research.
REFERENCES
1. Link, R. A.; Elwi, A. E.; and Scanlon, A., “Biaxial Tension Stiffening
Due to Generally Oriented Reinforcing Layers,” Journal of Engineering
Mechanics, ASCE, V. 115, No. 8, 1989, pp. 1647-1662. doi: 10.1061/
(ASCE)0733-9399(1989)115:8(1647)
2. He, X. G., and Kwan, A. K. H., “Modeling Dowel Action of Rein-
forcement Bars for Finite Element Analysis of Concrete Structures,”
Computers & Structures, V. 79, No. 6, 2001, pp. 595-604. doi: 10.1016/
S0045-7949(00)00158-9
3. Ngo, D., and Scordelis, A. C., “Finite Element Analysis of Reinforced
Concrete Beams,” ACI Journal Proceedings, V. 64, No. 3, Mar. 1967,
pp. 152-163.
4. Nilson, A. H., “Nonlinear Analysis of Reinforced Concrete by the
Finite Element Method,” ACI Journal Proceedings, V. 65, No. 9, Sept.
1968, pp. 757-766.
5. Hofstetter, G., and Mang, H. A., Computational Mechanics of Rein-
forced Concrete Structures, Vieweg & Sohn Verlagsgesellschaft mbH,
Wiesbaden, Germany, 1995, 366 pp.
6. Keuser, M., and Mehlhorn, G., “Finite Element Models for Bond
Problems,” Journal of Structural Engineering, ASCE, V. 113, No. 10, 1987,
pp. 2160-2173. doi: 10.1061/(ASCE)0733-9445(1987)113:10(2160)
7. Mehlohrn, G., “Some Developments for Finite Element Analysis of
Reinforced Concrete Structures,” Proceedings of the Second International
Conference on Computer Aided Analysis and Design of Concrete Struc-
tures, Pineridge Press, Swansea, UK, 1990, pp. 1319-1336.
8. Miguel, P. F.; Jawad, M. A.; and Fernandez, M. A., “A Discrete-
Crack Model for the Analysis of Concrete Structures,” Proceedings of the
Second International Conference on Computer Aided Analysis and Design
of Concrete Structures, Pineridge Press, Swansea, UK, 1990, pp. 847-908.
9. Lowes, L. N.; Moehle, J. P.; and Govindjee, S., “Concrete-Steel Bond
Model for Use in Finite Element Modeling of Reinforced Concrete Struc-
tures,” ACI Structural Journal, V. 101, No. 4, July-Aug. 2004, pp. 501-511.
10. “fib Model Code for Concrete Structures 2010,” Ernst & Sohn,
Berlin, Germany, 2013. 500 pp.
11. Cairns, J., “Bond and Anchorage of Embedded Steel Reinforcement
in the fib Model Code 2010,” Structural Concrete, V. 16, No. 1, 2015,
pp. 45-55. doi: 10.1002/suco.201400043
12. Zhang, X.; Wang, L.; Zhang, J.; and Liu, Y., “Model for Flexural
Strength Calculation of Corroded RC Beams Considering Bond-Slip
Behavior,” Journal of Engineering Mechanics, ASCE, V. 142, No. 7, 2016,
pp. 1-11. doi: 10.1061/(ASCE)EM.1943-7889.0001079
13. Xu, L.; Hai, T.; and King, L. C., “Bond Stress-Slip Prediction
under Pullout and Dowel Action in Reinforced Concrete Joints,” ACI
Structural Journal, V. 111, No. 4, July-Aug. 2014, pp. 977-988. doi:
10.14359/51686816
14. Costa, R.; Providência, P.; and Dias, A., “Anchorage Models for
Reinforced Concrete Beam-Column Joints under Quasi-Static Loading,”
ACI Structural Journal, V. 113, No. 3, May-June 2016, pp. 503-514. doi:
10.14359/51688759
15. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 2014, 520 pp.
16. Zanuy, C.; Curbach, E. H. M.; and Lindorf, A., “Finite Element
Study of Bond Strength between Concrete and Reinforcement under
Uneven Confinement Condition,” Structural Concrete, V. 14, No. 3, 2013,
pp. 260-270. doi: 10.1002/suco.201200019
17. Cervenka, V., “Reliability-Based Non-Linear Analysis According to
fib Model Code 2010,” Structural Concrete, V. 14, No. 1, 2013, pp. 19-28.
doi: 10.1002/suco.201200022
18. Eligehausen, R.; Popov, E. P.; and Bertero, V. V., “Local Bond Stress-
Slip Relationships of Deformed Bars under Generalized Excitations,”
Report No. UCB/EERC-83/23, Earthquake Engineering Research Center,
University of California, Berkeley, Berkeley, CA, 1983, 169 pp.
19. Shima, H.; Chou, L. L.; and Okamura, H., “Micro and Macro Models
for Bond in Reinforced Concrete,” Journal of the Faculty of Engineering,
University of Tokyo, Tokyo, Japan, V. 39, No. 2, 1987, pp. 133-194.
20. Lormanometee, S., “Bond Strength of Deformed Reinforcing Bar
under Lateral Pressure,” University of Texas at Austin, Austin, TX, Jan.
1974, 64 pp.
21. Malvar, L. J., “Bond of Reinforcement under Controlled Confine-
ment,” ACI Materials Journal, V. 89, No. 6, Nov.-Dec. 1992, pp. 593-601.
112
22. Untrauer, R., “E., and Henry, R. L., “Influence of Normal Pressure on
Bond Strength,” ACI Journal Proceedings, V. 62, No. 5, May 1965, pp. 577-586.
23. Robins, P. J., and Standish, I. G., “The Influence of Lateral Pressure
upon Anchorage Bond,” Magazine of Concrete Research, V. 36, No. 129,
1984, pp. 195-202. doi: 10.1680/macr.1984.36.129.195
24. Rucka, M., and Wilde, K., “Experimental Study on Ultrasonic
Monitoring of Splitting Failure in Reinforced Concrete,” Journal of
Nondestructive Evaluation, V. 32, No. 4, 2013, pp. 372-383. doi: 10.1007/
s10921-013-0191-y
25. Solomos, G., and Berra, M., “Rebar Pullout Testing under Dynamic
Hopkinson Bar Induced Impulsive Loading,” Materials and Structures,
V. 43, No. 1-2, 2010, pp. 247-260. doi: 10.1617/s11527-009-9485-z
26. Ghandehari, M.; Krishnaswamy, S.; and Shah, S., “Bond-Induced
Longitudinal Fracture in Reinforced Concrete,” Journal of Applied
Mechanics, V. 67, No. 4, 2000, pp. 740-748. doi: 10.1115/1.1313822
27. Murcia-Delso, J., and Shing, P., “Bond-Slip Model for Detailed
Finite-Element Analysis of Reinforced Concrete Structures,” Journal of
Structural Engineering, ASCE, V. 141, No. 4, 2015, p. 04014125. doi:
10.1061/(ASCE)ST.1943-541X.0001070
28. Desnerck, P.; Lees, J. M.; and Morley, C. T., “Bond Behaviour of
Reinforcing Bars in Cracked Concrete,” Construction and Building Mate-
rials, V. 94, 2015, pp. 126-136. doi: 10.1016/j.conbuildmat.2015.06.043
29. Lahnert, B. J.; Houde, J.; and Gerstle, K. H., “Measurement of Slip
Between Steel and Concrete Core,” ACI Journal Proceedings, V. 83, No. 6,
May-June 1986, pp. 974-982.
30. Maeda, M.; Otani, S.; and Aoyama, H., “Effect of Confinement
on Bond Splitting Behavior in Reinforced Concrete Beams,” Struc-
tural Engineering International, V. 5, No. 3, 1995, pp. 166-171. doi:
10.2749/101686695780601042
31. Lura, P.; Plizzari, G.; and Riva, P., “Splitting Crack Propagation
in Pull-out Tests,” Proceedings of FramCos IV, Fracture Mechanics of
Concrete Structures, R. de Borst, J. Mazars, and J. Pijaudier-Cabot, eds.,
Balkema, Paris, France, 2001, pp. 999-1006.
32. Gambarova, P.; Rosati, G. P.; and Zasso, B., “Steel-to-Concrete Bond
after Concrete Splitting: Test Results,” Materials and Structures, V. 22,
No. 1, 1989, pp. 35-47. doi: 10.1007/BF02472693
33. Gambarova, P.; Rosati, G. P.; and Zasso, B., “Steel-to-Concrete Bond
after Concrete Splitting: Constitutive Laws and Interface Deterioration,”
Materials and Structures, V. 22, No. 5, 1989, pp. 347-356. doi: 10.1007/
BF02472505
34. Feldman, L. R., and Bartlett, M., “Bond Strength Variability in
Pullout Specimens with Plain Reinforcement,” ACI Structural Journal,
V. 102, No. 6, Nov.-Dec. 2005, pp. 860-867.
35. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-11) and Commentary,” American Concrete Institute,
Farmington Hills, MI, 2011, 503 pp.
36. Tepfers, R., “Cracking of Concrete Cover along Anchored Deformed
Reinforcing Bars,” Magazine of Concrete Research, V. 31, No. 106, 1979,
pp. 3-12. doi: 10.1680/macr.1979.31.106.3
37. Minor, J., and Jirsa, J., “Behavior of Bent Bar Anchorages,” ACI
Structural Journal, V. 72, No. 4, Apr. 1975, pp. 141-149.
38. Luccioni, B. M.; López, D. E.; and Danesi, R. F., “Bond-
Slip in Reinforced Concrete Elements,” Journal of Structural Engi-
neering, ASCE, V. 131, No. 11, 2005, pp. 1690-1698. doi: 10.1061/
(ASCE)0733-9445(2005)131:11(1690)
39. Désir, J.-M.; Romdhane, M. R. B.; Ulm, F.-J.; and Fairbairn,
E. M. R., “Steel-Concrete Interface: Revisiting Constitutive and Numerical
Modeling,” Computers & Structures, V. 71, No. 5, 1999, pp. 489-503. doi:
10.1016/S0045-7949(98)00308-3
40. Lettow, S., “Ein Verbundelement für nichtlineare Finite Elemente
Analysen - Anwendung auf Übergreifungsstöße. Dissertation,” Institut für
Werkstoffe im Bauwesen, Universität Stuttgart, Stuttgart, Germany, 2006,
196 pp.
41. Mendes, L. A. M., and Castro, M. S. S., “A New RC Bond Model
Suitable for Three-Dimensional Cyclic Analyses,” Computers & Struc-
tures, V. 120, 2013, pp. 47-64. doi: 10.1016/j.compstruc.2013.01.007
42. ABAQUS, “Abaqus/CAE User’s Manual.” Dassault Systémes, 2012.
43. Oller, S.; Onate, E.; Oliver, J.; and Lubliner, J., “Finite Element
Nonlinear Analysis of Concrete Structures Using a Plastic-Damage Model,”
Engineering Fracture Mechanics, V. 35, No. 1-3, 1990, pp. 219-231. doi:
10.1016/0013-7944(90)90200-Z
44. Lubliner, J.; Oliver, J.; Oller, S.; and Onate, E., “A Plastic-Damage
Model for Concrete,” International Journal of Solids and Structures, V. 25,
No. 3, 1989, pp. 299-326. doi: 10.1016/0020-7683(89)90050-4
45. Hofstetter, G., and Meschke, G., Numerical Modeling of Concrete
Cracking, Springer Science & Business Media, Berlin, Germany, 2011,
327 pp.
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S09
Seismic Performance of Innovative Reinforced Concrete
Coupling Beam—Double-Beam Coupling Beam
by Youngjae Choi, Poorya Hajyalikhani, and Shih-Ho Chao
Diagonally reinforced coupling beams (DCBs) are commonly used
as seismic force-resisting members for medium- to high-rise build-
ings. The diagonal reinforcing bars in DCBs are most effective
when the beam has a span-depth ratio of less than 2. However,
modern construction typically requires span-depth ratios between
2.4 and 4, which leads to a very shallow angle of inclination for
the diagonal reinforcement. The lower angles of inclination, when
combined with the detailing requirements specified in ACI 318,
result in reinforcement congestion and construction difficulties.
These issues can be considerably minimized by using an innova-
tive and simplistic reinforcing scheme consisting of two separate
cages similar to those used for typical beams in reinforced concrete
special moment frames. The proposed coupling beam acts like a
conventional coupling beam under small displacements. Upon the
occurrence of large displacements, cracks begin developing at the
beam’s midspan and midheight area where the narrow unreinforced
concrete strip is located, gradually propagating toward the beam’s
ends. The cracks eventually separate the coupling beam into two
relatively slender beams where each has nearly twice the aspect
ratio of the original coupling beam. This split essentially trans-
forms the shear-dominated deep beam behavior into a flexure-dom-
inated slender beam behavior. Because damage initiates from the
center of the beam and then spreads toward the ends, the beam’s
ends maintain their integrity even under very large displacements,
thereby eliminating the sliding shear failure at the beam-to-wall
interface. Testing results on half-scale specimens with span-depth
ratios of 2.4 and 3.3 showed that the proposed coupling beam not
only has high ductility and shear strength, but can significantly
reduce construction issues in conventional DCBs. In addition, the
proposed coupling beam arrangement has great architectural flexi-
bility, allowing utility ducts to be placed inside the coupling beams
where the gap between the two steel cages is located.
INTRODUCTION
Reinforced concrete (RC) structural walls are used
commonly as the primary seismic-force-resisting system
in buildings. Based on the architectural requirements, these
walls often have numerous openings for features such as
elevators, windows, and doors, which divide a single wall into
more slender walls connected by substantial beams. These
beams are known as coupling beams. The use of the coupled
wall system leads to a more efficient and economical structure
system than single walls because properly designed coupled
wall systems possess significantly higher strength, stiffness,
and energy dissipation capacity. For the desired behavior of
the coupled wall system to be attained, the coupling beams are
required to sustain high shear forces while undergoing large
displacement. However, the coupling beams must also yield
before the wall piers, behave in a ductile manner, and exhibit
significant energy-dissipating characteristics.
ACI Structural Journal/January 2018
TECHNICAL PAPER
Two types of coupling beams are allowed by ACI 318-14
(ACI Committee 318 2014) for coupled wall system. They
are conventional and diagonally reinforced coupling beams,
and their application depends on the span-depth ratio (or
aspect ratio ln/h) and shear demands. Prior studies (Barney et
al. 1980; Naish at el. 2013; Lim et al. 2016) have shown that
conventionally reinforced coupling beams (reinforced with
vertical and longitudinal reinforcing bars) with span-depth
ratios between 2.5 and 4 and maximum shear stress between
3.4√fc′ and 7.9√fc′ exhibited fast strength degradation after
approximately 4% beam chord rotation. The major source of
the strength degradation was the sliding shear at the beam-
to-wall interface.
Another important factor when considering strength
degradation is rotational capacity. Prior nonlinear time-
history analyses (Harries and McNeice 2006) indicated that
coupling beams would need average rotational capacities of
approximately 3% and 6% for design basis earthquakes
(DBEs) and maximum considered earthquakes (MCEs),
respectively. These rotational capacities help maintain the
integrity of a coupled wall system. Note that DBE has a 10%
probability of exceedance in 50 years and MCE has a 2%
probability of exceedance in 50 years.
Based on the shear resistance and adverse failure mecha-
nisms of conventional coupling beams, Paulay and Binney
(1974) proposed diagonally reinforced coupling beams
(DCBs) with reinforcement detailing consisting of two
intersecting diagonal reinforcement groups combined with
closely spaced transverse reinforcement (Fig. 1(a)). In this
reinforcement detail, the diagonal bars need to be well
confined by transverse reinforcement and carefully anchored
in the walls. In a design using this type of coupling beam, the
entire shear-transfer mechanism is resisted by heavily rein-
forced diagonal cages. Experimental results have shown that
diagonal reinforcement detailing can significantly improve
deformation and energy dissipation capacity compared
with conventional detailing for coupling beams subjected
to reverse cyclic loading (Barney et al. 1980; Galano and
Vignoli 2000; Naish et al. 2009; Paulay and Binney 1974).
For coupling beams with a span-depth ratio of approxi-
mately 1.0, diagonal reinforcement placed with a large incli-
nation has proven to be the most efficient solution (Harries
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2016-368.R3, doi: 10.14359/51700951, received March 10, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
113

Fig. 1–Reinforcement detail: (a) DCB with aspect ratio
greater than 2.0, according to ACI 318-14 (confinement of
individual diagonals); (b) DCB with aspect ratio greater
than 2.0, according to ACI 318-14 (full confinement of diag-
onally reinforced concrete beam section); and (c) proposed
reinforcement detailing (DBCB) for RC coupling beam.
et al. 2005). On the other hand, modern architectural
specifications typically require span-depth ratios between
2.4 and 4. Prior research shows that for DCBs with span-
depth ratios between 2.4 and 3.3, the majority of damage
occurred at the beam-to-wall interface due to slip of the
diagonal reinforcing bars from adjacent walls (Naish et al.
2013). This type of bond failure is typically difficult to repair
(Engindeniz et al. 2005). In addition to the slip, damage at
the DCB ends revealed full-depth cracking, which leads to
separation of the beam and wall (Paulay 1977). In addition,
coupling beams with span-depth ratios between 2.4 and 4
lead to a very shallow angle of inclination for the diagonal
reinforcement (can be as low as approximately 10 degrees).
The lower angles of inclination, combined with the detailing
requirements specified in ACI 318-14, can cause several
major issues for both design and construction (Harries et al.
114
2005; Moehle et al. 2011; Naish et al. 2009; Moehle 2015).
Five prominent issues are listed as follows:
1. A small angle of inclination significantly decreases
the efficiency of diagonal reinforcement in resisting shear
forces, which in turn requires an even smaller angle to
accommodate the reinforcing bars. Thus, more reinforcing
bars are needed, which ultimately increases the difficulty
of construction. There is significant difficulty in placing
the diagonal reinforcing bars because they can be easily
obstructed by transverse reinforcement used to confine the
diagonal bars (Fig. 1(a)). Additional reinforcement detailing
is also required when the extension of the diagonal bars is
bent at the top of the wall or at openings in the wall.
2. Although ACI 318-14 Section 18.10.7.4(d) allows alter-
native full beam section confinement to reduce congestion
issues, there is still the obvious difficulty of passing the diag-
onal bars through the hoops and crossties (Fig. 1(b)).
3. It can be very challenging and time-consuming to thread
the diagonal reinforcement through the congested vertical
and horizontal bars in the wall’s boundary elements
(Fig. 1(a) and 1(b)).
4. The minimum width requirement for diagonal elements
causes interlock of the two diagonal elements. This in turn
demands an increased clear distance between reinforcing
bars in order for one diagonal element to pass through the
other. Often, the diagonal reinforcing bars need additional
space to prevent the conflict with the boundary element rein-
forcement of the wall pier. The minimum dimensions and
required reinforcement clearances can make the coupling
beam very wide, which controls the wall width.
5. Modern construction also calls for the additional archi-
tectural requirement of having utility ducts go through the
coupling beams (Hooper 2014). This is generally a diffi-
cult task when using diagonally reinforced coupling beams
(DCBs) due to reinforcement congestion or the concern of
weakening the local region when ducts are present.
Other alternative reinforcement schemes have been inves-
tigated (Tassios et al. 1996), such as the addition of dowels
at the ends of the coupling beams or a diagonal reinforce-
ment located only at the beam-wall interface. However, it
was experimentally demonstrated that coupling beams with
these alternative reinforcement details did not exhibit satis-
factory seismic behavior, and/or they posed construction
difficulties (Tassios et al. 1996).
The aforementioned construction and design issues with
DCBs can be considerably minimized by using an innova-
tive and simplistic reinforcing scheme investigated in this
study, as illustrated in Fig. 1(c). This new coupling beam—
double-beam coupling beam (DBCB)—consists of two
separate cages similar to those used for beams in reinforced
concrete (RC) special moment frames. The new coupling
beam acts like a conventional coupling beam under small
displacements. With larger displacements, cracks begin
developing at the beam’s midspan and midheight where
the narrow unreinforced concrete strip is located, gradually
propagating toward the ends. The cracks eventually separate
the beam into two relatively slender beams (ACI 318-14,
Section 18.6.2.1, defines a beam with clear span ln extending
at least four times the effective depth d as a relatively slender
ACI Structural Journal/January 2018
member) where each has nearly twice the aspect ratio of
the original coupling beam. This essentially transforms the
shear-dominated behavior into flexure-dominated behavior
common to conventional beams. Because damage initiates
from the center of the beam, then spreads toward the ends,
the beam ends maintain their integrity even under very
large displacements, thereby eliminating the sliding shear
failure at the beam-to-wall interface, as is commonly seen
in conventional beams (Aktan and Bertero 1981; Aristiza-
bal-Ocfaoa 1987). In addition, because the cracks always
initiate at midspan and midheight, the damage location can
be easily predicted, which makes repair work, if needed,
much easier after moderate earthquakes. Furthermore, the
gap between the two cages allows for utility ducts through
the beam without affecting the beam’s seismic behavior.
For performance-based design, a structural component’s
damage at various performance levels needs to be investi-
gated to evaluate its potential impact on life and economic
loss, down time, and occupancy. Post-earthquake damage
states at various performance levels for DCBs, as outlined by
ACI 374.2R (ACI Committee 374 2013) and ASCE/SEI 41
(ASCE 2014), were used to evaluate the performance of
DBCBs. The three levels of performance are:
• Immediate Occupancy Performance Level (defined
as the post-earthquake damage state in which a structure
remains safe to occupy and essentially retains its
pre-earthquake strength and stiffness): Coupling beams
experience diagonal cracking, and the maximum accept-
able beam rotation is 0.6% rad.
• Life Safety Performance Level (defined as the
post-earthquake damage state in which a structure has
damaged components but retains a margin against the
onset of partial or total collapse): Coupling beams have
extensive shear and flexural cracks with some crushing,
but concrete generally remains in place. The maximum
acceptable beam rotation is 1.8% rad.
• Collapse Prevention Performance Level (defined as
the post-earthquake damage state in which the building
is on the verge of partial or total collapse. Substantial
damage to the structure has occurred, potentially
including significant degradation in the stiffness and
strength of the lateral-force-resisting system): Coupling
beams are shattered and virtually disintegrated. The
maximum acceptable beam rotation is 3% rad.
RESEARCH SIGNIFICANCE
This research investigated the seismic performance of RC
double-beam coupling beams (DBCBs) used in RC coupled
core walls (CCWs). Innovative configurations were used
that can significantly improve and simplify constructibility,
design, and repair work. The new coupling beams also
exhibited excellent strength, stiffness, and ductility under
large displacement reversals. The reinforcement layout of
the new DBCBs is similar to that used for typical beams in
reinforced concrete special moment frames. Results from
this research show how the new detailing approach provides
simplified construction. DBCBs also have seismic behavior
equivalent or superior to that provided by the diagonally
ACI Structural Journal/January 2018
Fig. 2—Force-resisting mechanism in DBCBs.
reinforced coupling beams with span-depth ratios between
2.4 and 3.3.
EXPERIMENTAL PROGRAM
Force-resisting mechanism of DBCBs
When a DBCB is subjected to small shear forces, the
vertical shear stress distribution along the beam height
follows that predicted by elastic theory. Associated hori-
zontal shear stress also acts along the longitudinal axis of the
beam. Both stresses are the highest at the midheight of the
beam. Thus, maximum shear strain takes place at midheight.
Because no shear reinforcement is present along the beam’s
midheight, concrete at this location cracks first. The shear
deformation gradually breaks the concrete and a gap
develops from the midspan toward the ends, and eventually
separates the DBCB into two relatively slender beams. Note
that although the shear stress is theoretically the same at
beam ends (Section A in Fig. 2), the restraint imposed by the
adjacent walls prevents the DBCB from separation at the
ends. Therefore, the separation starts at midspan (Section B
in Fig. 2), then propagates toward both ends as displacement
increases. This damage process is opposite from that of
conventionally reinforced coupling beams in which the
damage starts from the beam-to-wall interface, thereby
drastically reducing the ability of transferring shear to the
walls and eventually leading to sliding shear failure or bond
slip failure.
Reinforcement details
Reinforcement details of the test DBCBs are illustrated in
Fig. 3. Each DBCB has two separate steel cages, and each
consists of three layers of vertically distributed longitudinal
reinforcing bars. Each steel cage is confined by closed trans-
verse reinforcement. The advantages of using distributed
longitudinal reinforcing bars are: 1) they delay the propa-
gation of flexure-shear cracks from becoming the domi-
nating cracks before the gap opens and separates the DBCB.
The diagonal cracks resulting from cyclic shear forces can
rapidly diminish the concrete’s shear resistance (Park and
115
Fig. 3—Reinforcement details of specimens. (Note: Dimensions are in inches; 1 in. = 25.4 mm.)
Table 1—Specimen information
Transverse reinforcement†
Unreinforced ρplastic hinging, % ρnon-plastic hinging, % ρl,‡ % (main bar,
Specimen ln/h ln/d concrete strip, in. Vnormalized* (bar size, spacing) (bar size, spacing) intermediate bar) fcm,§ ksi
R2.4-SC-1 2.4 5.76 1 8.34 4.9 (No. 4, 1.5 in.) 2.9 (No. 4, 2.5 in.) 2.4 (No. 6, No. 6) 5.7
R2.4-NC-1 2.4 5.76 1 8.34 4.9 (No. 4, 1.5 in.) 2.9 (No. 4, 2.5 in.) 2.4 (No. 6, No. 6) 6.3
R3.3-SC-1 3.3 7.98 1 7.47 4.9 (No. 4, 1.5 in.) 1.8 (No. 4, 4.0 in.) 3.2 (No. 7, No. 7) 5.7
R2.4-SC-0.25 2.4 5.49 0.25 10.1 4.9 (No. 4, 1.5 in.) 2.9 (No. 4, 2.5 in.) 3.0 (No. 7, No. 6) 6.0
R2.4-SC-2-P 2.4 6.33 2 9.43 3.7 (No. 3, 1.0 in.) 2.6 (No. 3, 1.4 in.) 3.5 (No. 7, No. 6) 4.0
*
Normalized shear stress: Vnormalized = Vdesign /( f c′bw h) , where fc′ = 5 ksi (design compressive strength).
†
Transverse reinforcement ratio ρ is calculated by Av/(sbw). Here, Av is total transverse reinforcement area within transverse reinforcement spacing s; and bw is width of beam. Note
that hoops for R2.4-C-1, R2.4-NC-1, and R2.4-SC-0.25 are overlapped by two C-shaped stirrups so that area of one leg is two times original area of No. 3 stirrup.
‡
ρl is reinforcement ratio of either top or bottom longitudinal bars of each individual beam; ρl = As/(bwd). Intermediate bars are not included in ratio.
§
fcm is measured concrete compressive strength.
Notes: 1 in. = 25.4 mm; 1 ksi = 6.895 MPa.

Paulay 1975); 2) vertically distributed longitudinal rein-
forcing bars form a pattern of multiple finer cracks, which
helps in the development of a more uniform distribution of
stresses along the transverse reinforcement (Paulay 1969);
3) vertically distributed longitudinal reinforcing bars can
prevent sliding shear failure near the beam-to-wall bound-
aries (Xiao et al. 1999); and 4) a beam section with vertically
distributed longitudinal reinforcing bars can add additional
flexural strength and ductility (Wong et al. 1990). Therefore,
the width of the coupling beam can be reduced and wider
spacing between reinforcing bars can be used in the top of
the beam during construction for placing and controlling
vibration of concrete.
Specimen design
Five half-scale DBCBs were fabricated and tested under
reversed cyclic loads. The information of each DBCB is
shown in Table 1. The name of each specimen was desig-
nated according to test variables. Primary test variables were
span-depth ratios, displacement (loading) protocol, and the
size of the unreinforced concrete strip. For instance, Spec-
imen R2.4-SC-1 represents a DBCB with a span-depth ratio
of 2.4 (R2.4) and a 1 in. (25.4 mm) unreinforced concrete
strip (1), tested under symmetrical cyclic (SC) displacement
protocol. Specimen R3.3-SC-1 has a span-depth ratio of
3.3. The ratios were selected to reflect modern architectural
specifications, which typically require coupling beams with
span-depth ratios of approximately 2.4 for residential build-
ings and 3.3 for office buildings. Ratios were also selected to
compare the test results of DBCB specimens with previous
research on DCBs with the same span-depth ratios (Naish
et al. 2009). However, instead of using a beam width of
12 in. (305 mm) as is the case with DCB specimens (Naish
et al. 2009), the width of DBCBs was reduced by 50% (6 in.
[152 mm]) because this width is sufficient to accommodate
longitudinal reinforcement. Also, this smaller width can
impose high shear stress on the DBCBs. In addition, the
development length of the longitudinal reinforcement is
only about 50% (for No. 7 bars) to 60% (for No. 6 bars)
of that required by ACI 318-14 Section 18.8.5.3(b). This is
because the beam-to-wall boundary did not suffer severe
damage (to be discussed later). This is opposite to DCBs,
where the major damage is at the beam-to-wall boundary
due to the slip and extension of the diagonal reinforcing bars
(Naish et al. 2013).
All specimens have a 1 in. (25.4 mm) unreinforced
concrete strip (clear distance between the ends of the trans-
verse reinforcement) at midheight, except for Specimens
R2.4-SC-0.25 and R2.4-SC-2-P (with PVC pipes going
through the gap), which had a 0.25 in. (6.35 mm) and 2 in.

116 ACI Structural Journal/January 2018

(51 mm) unreinforced concrete strip, respectively. The sizes
of the unreinforced concrete strip were selected based on
preliminary sizes taken from a nonlinear finite element
analysis conducted by VecTor2 (Wong and Vecchio 2002),
which is explained later in this paper. The analysis shows that
a wide unreinforced concrete strip could lead to a reduced
moment arm for each cage; consequently, its moment capacity
as well as the overall shear strength of the DBCB is reduced.
On the other hand, if the unreinforced concrete strip is too
narrow, the beam might not separate into two beams before the
major shear cracks dominate the behavior, as in conventionally
reinforced coupling beams (using only vertical and longitudinal
reinforcement) and can lead to early strength degradation.
When a DBCB separates into two relatively slender
beams, the individual beams will have a span-depth ratio
that meets ACI Code criteria for beams in special moment
frames—that is, the span of the beam shall be at least four
times greater than that of the effective depth of the beam
(ACI 318-14 Section 18.6.2.1). Therefore, each half of the
DBCB specimens was designed according to ACI Code
except when enforcing the requirement for longitudinal
reinforcement ratio, which is limited up to 2.5% for beams
of special moment frame. The top or bottom reinforcement
ratio of each separated beam was slightly higher than 2.5%
(Table 1). This was done to increase the shear stress of the
specimens. As shown in Fig. 3, reinforcement details of
Specimens R2.4-SC-1 and R2.4-NC-1 were identical, and
the only difference between these two specimens was the
applied loading protocols. For these two specimens, No. 6
bars (284 mm2) were used for both the main and interme-
diate longitudinal reinforcement. In Specimen R3.3-SC-1,
because this specimen had greater span length, No. 7 bars
(387 mm2) were used for longitudinal reinforcement to
increase the shear stress to equal that in Specimens R2.4-
SC-1 and R2.4-NC-1. A pilot analytical study using VecTor2
indicated that if the unreinforced concrete strip is too narrow,
the beam might not separate into two beams. To verify the
analysis result, Specimen R2.4-SC-0.25 was fabricated with
a 0.25 in. (6.35 mm) unreinforced concrete strip. This spec-
imen was also used to gauge the maximum shear stress and
rotation a DBCB can carry when it is not separated. No. 7
bars (387 mm2) and No. 6 bars (284 mm2) were used as the
beam’s main and intermediate longitudinal reinforcement,
respectively. Specimen R2.4-SC-2-P used the same bars as
used in R2.4-SC-0.25. However, the nominal shear strength
of Specimen R2.4-SC-2-P was lower than that of Specimen
R2.4-SC-0.25. This is because of the decreased moment arm
caused by using the 2 in. (50.8 mm) unreinforced concrete
strip. Specimen R2.4-SC-2-P aimed at reaching a shear stress
demand of approximately 9.5√fc′ (psi), which is slightly
lower than the ACI limit on the maximum nominal shear
strength of DCBs. The required transverse reinforcement
was calculated based on the maximum probable moment at
beam ends. An overstrength factor of 1.25 was multiplied to
the specified yield strength of steel reinforcing bars when
calculating the probable moment of all specimens. Note that
following ACI 318’s assumption with a maximum useable
concrete strain of 0.003 would lead to a small compres-
sive stress in the compression steel, which in turn leads to
ACI Structural Journal/January 2018
a smaller probable beam end moment and shear demand.
It was observed in this study’s preliminary testing that the
maximum compression reinforcement strain reached 0.008
without concrete crushing. This is because a large amount
of confinement was provided by closely placed transverse
and longitudinal reinforcement. Therefore, a maximum
usable concrete strain of 0.006 was used for the calculation
of the moment capacity of DBCBs, which leads to a higher
maximum probable moment. It should be also noted that the
restraint from the adjacent walls can provide non-negligible
compression, which in turn increases the bending strength of
the beam. Then, the expected peak shear stress (design shear
stress Vdesign) can be calculated from the sum of maximum
probable beam end moments divided by the beam’s span.
The normalized peak shear stress of each specimen is calcu-
lated using the expected peak shear stress divided by the
square root of design concrete compressive strength (5 ksi
[34.4 MPa]). These shear stress calculation results are listed
in Table 1. However, the required transverse reinforcement
of each specimen was controlled by ACI 318’s confinement
requirements rather than by shear demand. Except for Spec-
imen R2.4-SC-2-P, all specimens used No. 3 transverse
reinforcement at a spacing of 1.5 in. (38 mm) in the plastic
hinging region, which was assumed to have a length equal
to the height of the DBCB. The first two hoops from the
beam-to-wall interfaces were placed with a spacing of 1.0 in.
(25.4 mm). This was done to increase confinement to the
concrete where the maximum moments occurred. Beyond
the plastic hinging region, the spacing of transverse rein-
forcement was 2.5 in. (63.5 mm) for all specimens except
for Specimen R3.3-SC-1, which had 4 in. (101.6 mm) of
spacing. A wider spacing in Specimen R3.3-SC-1 was used
considering the smaller shear demand due to the slender
aspect ratio. Except for Specimen R2.4-SC-2-P, the closed
hoops were made of two overlapping U-shaped stirrups by
welding (Fig. 3). Therefore, clear spacing of the transverse
reinforcement was less than using one single No. 3 hoop. For
Specimen R2.4-SC-2-P, No. 3 hoops with a 135-degree hook
were spaced more closely together (1 in. [25.4 mm]) than the
other specimens. The plastic hinging length was assumed as
half of the beam height. Beyond the plastic hinging region,
1.4 in. (35.6 mm) spacing was used. For all specimens, 2.5 in.
(63.5 mm) spacing was used at the midspan to help the
initiation of the separation.
Specimen construction
Each specimen consisted of a coupling beam as well as
a pair of large and small reinforced concrete blocks repre-
senting adjacent structural walls. Laying on one side on the
strong floor, concrete was placed for all specimens. Form-
works were made of plywood reinforced with 2 x 4 in. (51 x
102 mm) lumber. DBCB formwork was inserted in between
the formwork for the concrete blocks. Top and bottom cages
of the DBCB were built separately and were easily slid into
the cage of the concrete blocks (Fig. 4(a)). This construction
process was much easier than that required for DCBs. Ready
mixed concrete was used to cast all specimens. The specified
concrete compressive and steel yield strength was 5000 and
60,000 psi (34.5 and 414 MPa), respectively. The measured
117
Fig. 4—(a) Construction photo; and (b) test setup.
Fig. 5—Loading protocols: (a) symmetrical loading protocol; and (b) near-collapse loading protocol.
average concrete compressive strengths fcm obtained on the
testing dates are listed in Table 1.
Test setup and instrumentation
The specimens were cast horizontally, then rotated and
placed in the test setup, with the large block fixed to the
strong floor (Fig. 4(b)). Displacement reversals were applied
via a vertical actuator, with the line of action of the actu-
ator forces passing through the test specimen’s midspan to
produce an anti-symmetrical moment pattern in the coupling
beam and zero moment at the beam’s midspan. The actu-
ator was connected to the small block through a wide flange
steel section. The load was transferred to the small block by
means of direct bearing and unbonded threaded bars passing
through the small block. Two steel links were used to provide
some moderate axial restraints for the beams because the
adjacent structural walls and surrounding slab would provide
non-negligible resistance to beam expansion upon cracking
(Teshigawara et al. 1998; Lequesne 2011; Barbachyn et al.
2012). Each steel link consisted of two channel sections and
was pin-connected to a T-stub. The T-stub was tightened to
each block by four high-strength threaded rods. To deter-
mine the shear force carried by the links, one of the coupling
beam specimens was cut in the middle after testing and the
loading block was displaced 4 in. (102 mm) upward. It was
found that the maximum shear force carried by the links
was 44 lb (195 N), which is negligible compared with the
peak strength of the coupling beams. The coupling beam
118
specimens were subjected to cyclic loading in a displace-
ment control mode, which produced predefined reversed
cyclic displacement patterns. Two loading protocols were
used, starting from a coupling beam chord rotation of
0.25% and reaching a maximum rotation of 12%. The first
loading protocol consisted of symmetric cyclic (SC) loading
using two to three cycles per deformation level (Fig. 5(a)).
However, this type of loading is not representative of a
near-collapse (NC) earthquake response, which would be
unsymmetrical and would contain fewer cycles of loading;
hence, the loading protocol should contain displacements
that are representative of the ratcheting effect that leads to
structural collapse. Such a protocol was developed based
on preliminary nonlinear analyses (Fig. 5(b)). Linear vari-
able differential transformers (LVDTs) were placed at the
boundary of the DBCB and the small block to measure beam
chord rotations and at the outer face of the large block to
gauge its rotation and movement. Strains of transverse and
longitudinal reinforcement were measured by strain gauges
attached to the reinforcement.
EXPERIMENTAL RESULTS
Cracking and damage pattern
Table 2 summarizes experimental test results for all spec-
imens. The progressive cracking patterns and the failure
mechanisms are shown in Fig. 6. The specimens with a
1 in. (25.4 mm) unreinforced concrete strip separated into
two beams that had nearly twice their original span-depth
ACI Structural Journal/January 2018
Table 2—Test results
Specimen θyield,* % Vmax, kip Vmax/Vdesign
R2.4-SC-1 2 68.5 1.29
R2.4-NC-1 2 89 1.68
R3.3-SC-1 2 68 1.43
R2.4-SC-0.25 4 68 0.97
R2.4-SC-2-P 2 63.5 1.06
*
θyield is beam chord rotation when longitudinal bar yields.
†
Normalized shear stress Vnormalized is Vmax /( f cm bw h) , where fcm is measured compres-
sive strength.
Notes: 1 in. = 25.4 mm; 1 kip = 4.45 kN.
ratio. The cracking and damage patterns of the specimens
were very similar to one another. During the first positive
cycle loading, diagonal tension cracks (shear cracks) were
observed at the midspan and midheight of the specimens
where the 1 in. (25.4 mm) unreinforced concrete strip
was located. As the load reversed, the same crack pattern
appeared at the same location, producing diagonal grids of
cracks crossing each other. No flexural cracks were observed
at that time. As the load increased, those cracks spread
toward both ends of the specimens (Fig 6(a)). As cyclic
loads continued increasing, concrete at the unreinforced
concrete strip was crushed due to the abrasion resulting
from the sliding of the top and bottom beams. This crushing
initiated from midspan, progressing toward the beam ends,
because the blocks provided the DBCBs with a confinement
effect near the boundaries. The relative displacement of
the top and bottom beams due to the horizontal sliding is
clearly seen by a marked grid on the specimen (Fig. 6(a)). At
1.5% chord rotation, diagonal cracks widened up to 0.16 in.
(4 mm). Canbolat et al. (2005) reported that diagonal cracks
were first observed during the cycles at 0.25% rotation for
their DCB specimen, and the diagonal cracks widened up
to 0.12 in. (3 mm) at 1.5% rotation. This indicated that the
occurrence of initial cracks in DBCBs and their widths are
similar to that of conventional DCBs. At 2% rotation, the end
of the separation is approximately 6 in. (152 mm) away from
the beam-to-wall boundaries. Diagonal cracks at the end of
the separation spread toward the boundaries in a fan-shaped
pattern. Concrete crushing at all corners of the specimens
was initiated at approximately 2% to 3% rotations. At this
time, the measured strain in longitudinal reinforcement
at the corners indicated that the reinforcement yielded in
compression. Beyond 4% rotation, the concrete crushing at
the corners caused all the cover concrete in the plastic region
to come off. Beyond 6% to 8% rotation, the core concrete
crumbled due to the combined normal and shear stresses,
leading to significant degradation in the shear resistance.
This was the primary reason for the strength drop of Speci-
mens R2.4-SC-1 and R3.3-SC-1. For Specimen R2.4-NC-1,
which was subjected to the near-collapse protocol (Fig. 6(b)),
the damage was not as severe as the damage observed in
Specimen R2.4-SC-1. Note that all the specimens main-
tained their integrity at the beam-to-wall interfaces up to
the end of the tests, which effectively eliminated the sliding
shear failure.
ACI Structural Journal/January 2018
The cracking pattern of Specimen R2.4-SC-0.25 (with a
θmax,% Vnormalized†
smaller unreinforced concrete strip height) was similar to the
other specimens in the early stage of loading. However,
8 10.1
the concrete crushing in the unreinforced concrete strip
11 12.5 progressed in a slower manner. Before the separation
8 10.0 occurred, a considerable number of flexural-shear cracks had
2 9.75
propagated across the entire depth of the beam, leading to a
drastically different shear stress distribution from that shown
6 11.16
in Fig. 2. In addition, horizontal cracks along the outermost
longitudinal reinforcing bars developed after considerable
shear cracks occurred. This is believed to be due to dowel
failure as many shear cracks developed, causing loss of the
aggregate interlock after the width of the cracks widened.
Only the middle half of the unreinforced concrete strip sepa-
rated, which made the portions above and below the strip to
behave like beams with very short span-depth ratios. This
aggravated the shear-dominated damage in the separated
region. Cover concrete came off in this region and the spec-
imen eventually failed by shear dominated behavior
(Fig. 6(a)). As shown in Fig. 6(e), Specimen R2.4-SC-2-P
has a more complete separation toward the beam ends,
which was believed to be attributed to the presence of the
utility pipes as well as the larger-size unreinforced concrete
strip. However, it had more shear cracks because Specimen
R2.4-SC-2-P had higher shear stress demand and the transverse
reinforcement ratio was less than the other specimens
(Tables 1 and 2). These results indicate that the separation of
the unreinforced concrete strip is affected by the span-depth
ratio and the size of the strip. Additionally, placing the utility
pipes at the ends of DBCBs seems to help the separation.
While a seismic-resistant coupling beam is expected to
experience severe damage at the Life Safety and Collapse
Prevention performance levels, it is expected to retain most
of its pre-earthquake design strength with light damage
(ACI 374.2R; ASCE/SEI 41) at the Immediate Occupancy
Performance Level. For the Immediate Occupancy Perfor-
mance Level, the maximum accepted beam chord rotation
for DCBs is from 0.6% according to ACI 374.2R. Figure 6
shows that at 0.75% rotation, DBCBs only had cosmetic
concrete cracking and the peak strengths had not been
reached. Furthermore, the beam-to-wall interfaces were
undamaged. Therefore, DBCBs can be easily repaired by
patching with new concrete to regain the elastic stiffness
after moderate earthquakes.
Shear-versus-rotation responses
Shear force (and stress) versus beam chord rotation
responses for all specimens are illustrated in Fig. 7. For
comparison purposes, ACI 318’s upper-bound shear stress
Vu = ϕ10√fc′Acw is also plotted as dashed lines in Fig. 7. The
beam chord rotation was obtained by using the net vertical
displacements, which were measured between the two
beam-to-wall interfaces, excluding the fixed-end rotation or
displacement, if any. This measurement was then divided
by the beam span. It is seen in Fig. 7(a) that Specimen
R2.4-SC-1 was able to maintain a very high shear stress
(10.1√fcm; fcm is the measured concrete compressive strength,
or 10.8√fc′; fc′ is the design concrete compressive strength
= 5 ksi [34.5 MPa]) without significant strength degrada-
119
Fig. 6—Cracking and damage pattern of DBCBs: (a) R2.4-SC-1; (b) R2.4-NC-1; (c) R3.3-SC-1; (d) R2.4-SC-0.25; and
(e) R2.4-SC-2-P. Note: These do not represent residual cracking (that is, at zero rotation).
tion up to a rotation of 6% (approximate demand for MCE previously. Specimen R2.4-SC-2-P retained a very high
level ground motions). Also, it could still resist 80% of the shear stress (11.16√fcm or 10√fc′) up to 6% rotation without
peak stress at 8% rotation. In addition, Specimen R2.4-NC-1 significant strength drop. However, after 6% rotation, its
showed no strength degradation up to 11% rotation, while shear strength dropped rapidly.
shear stress increased (12.5√fcm or 14√fc′) when subjected
to the near-collapse loading protocol, as shown in Fig. 7(b). Strains in steel reinforcement
Furthermore, the shear strength in Specimen R3.3-SC-1 did The yielding in the top and bottom layers of the longitu-
not drop until 8% rotation, maintaining a shear stress of dinal reinforcing bars of the two steel cages at the beam-to-
10√fcm (or 10.7√fc′) (Fig. 7(c)). The shear stress of Specimen wall interface indicates the complete separation of DBCBs
R2.4-SC-0.25 reached 9.75√fcm (or 10.7√fc′). However, it and a mechanical behavior shifting from shear to flexure. For
experienced significant shear strength degradation after 2% Specimens R2.4-RC-1, R2.4-NC-1, and R2.4-SC-2-P, the
rotation due to the shear dominated behavior, as discussed yielding of longitudinal reinforcing bars commenced at 2%

120 ACI Structural Journal/January 2018

Fig. 7—Hysteresis curves of DBCBs: (a) R2.4-SC-1; (b) R2.4-NC-1; (c) R3.3-SC-1; (d) R2.4-SC-0.25; and (e) R2.4-SC-2-P.
(Note: 1 in. = 25.4 mm; 1 kip = 4.45 kN; normalized measured shear force is Vtest /( f cm b w h) , where fcm is measured concrete
compressive strength.)

Fig. 8—Strain of longitudinal bar inside loading block (R2.4-SC-2-P). (Note: 1 in. = 25.4 mm.)
rotation, while it was 3% in Specimen R3.3-SC-1. For Spec-
imen R2.4-SC-0.25, no yield in the longitudinal reinforcing
bars was measured at the beam-to-wall interface. In addition,
strains in the top and bottom longitudinal reinforcing bars of
each individual cage at the beam-to-wall interface showed
opposite signs, indicating no residual tensile strains affecting
the behavior of DBCBs. This is because the reinforcing bars
in DBCBs do not experience large inelastic deformation
even at large displacement. On the other hand, due to the
residual tensile strains after the reinforcing bars experience
large inelastic deformation, both the top and bottom rein-
forcing bars in conventionally reinforced coupling beams
are usually subject to tension at large displacement. This
does not allow the interface crack to be entirely closed when
subject to compression, thereby causing degradation of
shear-transfer mechanism and sliding shear failure. Yielding
in the longitudinal reinforcement spread up to approximately
7.5 in. (191 mm) (which is 50% of the height of DBCBs)
away from the beam-to-wall interface for Specimens R2.4-
SC-1, R-2.4-NC-1, and R2.4-SC-2-P, and approximately 15 in.
(38 mm) for Specimen R3.3-SC-1. This indicates that the
plastic hinging region of DBCBs depends on the span-depth
ratio. No yielding in transverse reinforcement was recorded
for Specimen R2.4-SC (NC)-1. However, only one hoop in
the plastic hinge region of Specimens R3.3-SC-1 and R2.4-
SC-2-P experienced yielding at 4% and 6% rotation, respec-
tively. In addition, all intermediate longitudinal reinforce-
ment experienced yielding, thereby contributing toward the
strength of the DBCBs. Figure 8 illustrates the strain distri-
bution from four strain gauges on a longitudinal reinforcing
bar inside the concrete block (representing the wall pier) of
Specimen R2.4-SC-2-P. As shown in the figure, strains in

ACI Structural Journal/January 2018 121

Fig. 9—Normalized shear stress versus beam chord rotation: (a) span-depth ratio of 2.4; and (b) span-depth ratio of 3.3. (Note:
Normalized shear stress is shear stress normalized by maximum shear stress of each specimen.)

Fig. 10—Comparison of DBCBs with DCBs: (a) span-depth ratio of 2.4; and (b) span-depth ratio of 3.3. (Note: Normalized
measured shear force is Vtest /( f cm b w h) , where fcm is measured concrete compressive strength.)
the reinforcing bars at the interface and at 4 in. (102 mm)
inside the block increased by a similar amount up to a 2%
beam rotation. However, at larger rotations, only the strain
at the interface showed a dramatic increase. Although the
specimen was subjected to many large displacement rever-
sals, yielding did not penetrate beyond approximately 4 in.
(102 mm) inside the wall pier. The embedded length of the
longitudinal reinforcing bars in Specimen R2.4-SC-2-P is
18 in. (457 mm), which is only 50% of that required by
ACI 318-14 Section 18.8.5.3(b), which is 37 in. (940 mm).
The minor yielding or complete elastic strain beyond 4 in.
(102 mm) inside the wall pier indicates that DBCBs did
not suffer bond deterioration, as commonly seen in DCBs
(Naish et al. 2013). This is because, unlike DCBs, damage
of a DBCB starts from the center of the beam and then
gradually propagates toward the end. Figure 8 also shows
that, along the embedded length, the steel reinforcement
experienced completely elastic or minor inelastic cyclic
deformation up to approximately 3% beam rotation.
Stiffness
Figures 9(a) and (b) illustrate the normalized shear stress
versus beam chord rotation response for DBCBs with the
symmetrical cycling load and DCBs tested in prior research
(Naish et al. 2009). The normalized shear stress is the stress
divided by the maximum shear stress of each specimen.
These figures indicate that DBCBs have an equivalent elastic
stiffness and act like a conventional coupling beam before
inelastic behavior occurs. In other words, DBCB’s elastic
stiffness is not affected by its unique separation mechanism.
In any case, a beam’s elastic stiffness can be easily adjusted
by optimizing its dimensions.
Comparison between DBCB and DCB
The performance of DBCBs was compared with that of
DCBs tested by Naish et al. (2009). Both DCB and DBCB
specimens were approximately 1/2-scale replicas of the
coupling beams in typical residential and office buildings.
The only major difference was that DBCBs were half the
width of DCBs. The design of the DCBs followed current
ACI provisions (ACI 318-14). Their DCBs can be divided
into two groups: one group has transverse reinforcement
around diagonal bar groups (CB24D or CB33D), and the
other group has transverse reinforcement around the entire
cross section (CB24F or CB33F). Because the performance
of the latter group, in terms of strength and ductility, was
slightly better (Naish et al. 2009), the performance of DBCBs
is compared to CB24F and CB33F in Fig. 10(a) and (b). The
shear stress, normalized by the square root of the measured
concrete compressive strength, fcm, was used for comparison
because the specimens have different concrete compressive
strengths and cross-sectional areas. Two beam chord rota-
tions, 3% and 6%, are highlighted in the figures because they
represent approximately the upper bound rotational demands
of coupling beams for the DBE and MCE level ground
motions (Harries and McNeice 2006). The specimens with a
span-depth ratio of 2.4 for both the DCBs and DBCBs
exceeded a shear stress level of 10√fcm (psi), which is the
unfactored shear strength limited by the ACI provision.
Although the strength in DBCB R2.4-SC-1 slightly

122 ACI Structural Journal/January 2018

Fig. 11—(a) Comparison of experiment and VecTor2 analysis; and (b) size effect of unreinforced concrete strip. (Note: 1 in. =
25.4 mm; 1 kip = 4.45 kN.)
decreased after 3% beam rotation, the overall ductility and Table 3—Material models
strength of DBCB R2.4-SC-1 were similar to that of the
Concrete models Concrete models
DCB specimen (CB24F) up to a beam rotation of 6% (MCE).
For the specimens with a span-depth ratio of 3.3, CB3.3F Compression Hognestad Crack width Agg/2.5 max.
prepeak (Parabola) check crack width
has a shear strength of 6.6√fcm (psi), and its shear strength
degradation begins after 3% beam rotation. On the other Compression Modified Crack slip Walraven
post-peak Park-Kent calculation (monotonic)
hand, DBCB R3.3-SC-1 could sustain much higher shear
stress (10√fcm [psi]) with no strength degradation up to 8% Compression
Vecchio 1992-A
Creep and
Not available
beam rotation. This shows that DBCBs can provide higher softening relaxation
safety by maintaining a stable global response of the struc- Tension Modified Bentz Hysteretic Nonlinear with
ture without loss of coupling between walls due to their high stiffening 2003 response plastic offsets)
plastic rotational capacity (Harries and McNeice 2006). Tension
Linear Concrete bond Eligehausen
stiffening
ANALYTICAL STUDY FRC tension Not considered — —
Nonlinear finite element analysis
Confined strength Kupfer/Richart Reinforcement
Nonlinear analysis using finite element software VecTor2
(Wong and Vecchio 2002) was conducted to study the effects Variable – Hysteretic Bauschinger
Dilation
Kupfer response Effect (Seckin)
of the width of the unreinforced concrete strip. A finite
element model was built for Specimen R2.4-SC-1. The trian- Cracking criteria
Mohr – Coulomb
Dowel action
Tassios
gular elements in VecTor2 were used for modeling concrete. (stress) (crack slip)
To simulate steel reinforcement, truss elements were used Crack stress Basic (DSFM/
Buckling
Refined Dhakal-
rather than smeared reinforcement. All the reinforcing bars calculation MCFT) Maekawa
simulated by truss elements were placed at the same posi-
tions as that in the actual specimen. A perfect bond between of the specimen in Fig. 11(a) is overestimated by VecTor2,
reinforcing bars and concrete was assumed. A smoothed the peak strengths and the beam rotations associated with the
triangle mesh was used with a mesh size of 0.63 in. (16 mm) starting of strength degradation reasonably match the exper-
based on a convergence study. The loading block and the imental results. It should be noted that VecTors2 stopped
large block were assumed to be rigid elements. Because running when severe cracking occurred at the unreinforced
VecTor2 does not provide any type of hinge boundary, the concrete strip location, although the strength of the DBCB
steel links, which connect the loading block to the large model did not drop significantly.
block for restraining the rotation of the loading block and
beam elongation, were omitted. Instead, to prevent the block Effect of height of unreinforced concrete strip
from rotating, displacement was imposed evenly on the top As shown in the experimental results, whether the DBCB
of the blocks. The material models used in this study are separates or remains unseparated depends on the height
shown in Table 3. A detailed explanation of the models’ of the unreinforced concrete strip. To assess the impact of
properties and functions can be found in the VecTor2’s User this factor, five models were studied. All models were the
Manual (Wong and Vecchio 2002). same in dimension and reinforcement detailing as Specimen
Figure 11(a) compares the simulation result from VecTor2 R24-SC-1, except they each had a different height unrein-
and the test result. The damage pattern from the analysis forced concrete strip. The model with a 0.25 in. (6.35 mm)
is very similar to the result of the experiment (refer to unreinforced concrete strip failed earlier than the others,
Fig. 11(a)), showing the beam separation and plastic hinges exhibiting the highest strength but poor ductility. On the
at the two separated beams (Fig. 11(a)). Comparison for other hand, the model with a 2 in. (51 mm) unreinforced
hysteresis response up to approximately 4% beam chord concrete strip exhibited the highest ductility, but lowest
rotation is shown in Fig. 11(a). Even though unloading stiffness strength (due to the reduced moment arm of each beam;

ACI Structural Journal/January 2018 123

however, strength can be easily raised by using larger
reinforcing bars with wider DBCBs). It was observed in the
analytical study that elastic stiffness of the beam is nearly
the same when the size of the unreinforced concrete strip
changes. In addition, the strength of the models increased
as the height of the unreinforced concrete strip decreased
(Fig. 11(b)). An additional model with a 2 in. (51 mm)
unreinforced concrete strip was analyzed. This model used
No. 7 reinforcing bars to replace No. 6 longitudinal bars at
the top and bottom of each cage, thereby compensating for
the reduced strength with larger strip height. The analysis
(Fig. 11(b)) shows that, this design was able to effectively
increase the strength while maintaining a high ductility.
Based on the experimental and analytical results, for each
individual beam in a DBCB, the span-depth ratio should be
at least 5.0 to ensure that the unreinforced concrete strip is
large enough to allow the DBCB to separate. The depth of
each beam was calculated as 0.5 × (depth of entire DBCB
minus height of the unreinforced concrete strip).
SUMMARY AND CONCLUSIONS
This study investigated an innovative and simplistic rein-
forcing layout for RC coupling beams that can significantly
reduce design and construction difficulties when using
diagonally reinforced coupling beams. The new double-
beam coupling beam (DBCB) consists of two separate
cages much like those used for typical beams in reinforced
concrete special moment frames. The two cages are sepa-
rated by a small spacing from 1 to 2 in. (25.4 to 50.8 mm).
Only vertical and horizontal reinforcing bars are needed.
Upon large displacements, cracks begin developing at a
DBCB’s midspan and midheight, then gradually propagate
toward the beam’s ends. The cracks eventually separate the
coupling beam into two relatively slender beams where each
has nearly twice the aspect ratio of the original coupling
beam. This split essentially transforms the shear-dominated
single deep beam behavior into a flexure-dominated slender
beam behavior. Because damage initiates from the center of
the beam, and then spreads toward the ends, the beam ends
are able to maintain their integrity even under very large
displacements, thereby eliminating the sliding shear failure
at the beam-to-wall interface. Although DCBs have excel-
lent seismic performance, they are very difficult to construct
with a beam span-depth ratio higher than 3. On the other
hand, the proposed DBCBs possess simultaneously high
shear strength and ductility with a much simpler design and
construction process. The width of adjacent walls does not
have to be increased due to the required width of coupling
beams because the width of DBCBs can be largely reduced
compared to DCBs. In addition, the placement of longitu-
dinal reinforcing bars in DBCBs can be easily adjusted to
accommodate the locations of vertical longitudinal rein-
forcing bars in the wall pier’s boundary elements. Further-
more, diagonal bars in DCBs need to be bent at the top floor
in order not to protrude outside of the concrete wall. This
could make the construction even harder and affect the
force-transferring mechanism. On the other hand, DBCBs
can easily resolve these issues.
The following conclusions are drawn from the study:
124
1. The test results on half-scale coupling beam specimens
with a span-depth ratio of 2.4 and 3.3 showed that coupling
beams with the proposed reinforcement scheme were able
to simultaneously sustain high shear stresses (10 to 12√fcm
[psi]) and large rotations (6 to 11%) before strength degra-
dation occurred.
2. Experimental results showed that both DBCB and ACI
compliant diagonally reinforced coupling beams (DCBs)
with a 2.4 span-depth ratio were able to sustain a high shear
stress of 10√fcm (psi) up to 6% beam rotation (approximate
rotational demand for MCE ground motion and Collapse
Prevention Performance Level). However, for the speci-
mens with 3.3 span-depth ratio, DBCBs not only showed
better ductility (up to 8% beam rotation) than the DCBs (up
to 6% beam rotation), but also reached greater shear strength
(10√fcm versus 6.6√fcm [psi]). These improvements offered
higher safety by maintaining the structure’s stable global
response without loss of coupling between walls.
3. The beam-to-wall interface in DBCBs experienced
much less damage when compared to that of DCBs; there-
fore, a smaller development length is required for the longi-
tudinal reinforcing bars (approximately 60% of that required
by ACI 318-14 Section 18.8.5.3(b)).
4. Experimental and nonlinear FEA shows that the height
of the unreinforced concrete strip does not have a signifi-
cant effect on the elastic stiffness of the DBCBs. However,
a smaller height (0.25 in. [6.35 mm]) could not separate the
two beams before the major diagonal cracks developed,
thereby reducing the ductility of the DBCBs. On the other
hand, a large height will decrease the moment arm of each
beam, leading to a smaller capacity. Nevertheless, this can be
easily compensated for by using slightly larger reinforcing
bars or wider DBCBs with slightly more reinforcing bars.
5. Because the cracks at DBCBs always initiate at their
midspan and midheight, the damage location can be easily
accessed, which makes repair work easier after moderate
earthquakes. Test results showed that at 1% beam rotation
(rotational demand for Immediate Occupancy Performance
Level [ACI 374.2R]) DBCBs only had cosmetic concrete
cracking and the peak strengths had not yet been reached. The
beam-to-wall interfaces were essentially undamaged. This
makes repair work easier after moderate earthquakes. This is
opposite to DCBs, where the major damage is at the beam-
wall boundary as a result of the slip and extension of the diag-
onal bars (Naish et al. 2013), which is difficult to repair.
6. Based on the experimental and analytical results, for each
individual beam in a DBCB, the span-depth ratio should be at
least 5.0 to ensure that the unreinforced concrete strip is large
enough to allow the DBCB to separate. The depth of each beam
was calculated as 0.5 × (depth of entire DBCB minus height of
the unreinforced concrete strip). A detailed design and modeling
of DBCBs will be presented in future publications.
AUTHOR BIOS
ACI member Youngjae Choi is a PhD Student and Research Assistant in
the Department of Civil Engineering at the University of Texas at Arlington,
Arlington, TX. His research interests include design and behavior of rein-
forced concrete structures subjected to extreme loads.
Poorya Hajyalikhani is a Structural Engineer at Toll Brothers, Grapevine,
TX. He received his PhD in structural engineering from the University of
ACI Structural Journal/January 2018
Texas at Arlington. His research interests include the seismic behavior and
design of reinforced concrete coupling beams and squat walls with innova-
tive reinforcement configurations.
ACI member Shih-Ho Chao is a Professor at the University of Texas at
Arlington. He is a member of ACI Committee 239, Ultra-High-Performance
Concrete, and 544, Fiber-Reinforced Concrete. He received the ACI Chester
Paul Siess Award for Excellence in Structural Research in 2011. His research
interests include high-performance fiber-reinforced cementitious materials
and seismic behavior of reinforced concrete structures.
REFERENCES
ACI Committee 318, 2014, “Building Code Requirements for Struc-
tural Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 520 pp.
ACI Committee 374, 2013, “Guide for Testing Reinforced Concrete
Structural Elements under Slowly Applied Simulated Seismic Loads (ACI
374.2R-13),” American Concrete Institute, Farmington Hills, MI, 18 pp.
Aktan, A. E., and Bertero, V. V., 1981, “The Seismic Resistant Design of
R/C Coupled Structural Walls,” Report No. UCB/EERC-81/07, Earthquake
Engineering Research Center, University of California, Berkeley, Berkeley,
CA, 196 pp.
Aristizabal-Ocfaoa, J. D., 1987, “Seismic Behavior of Slender Coupled
Wall Systems,” Journal of the Structural Division, ASCE, V. 113, No. 10, Oct.,
pp. 2221-2234. doi: doi10.1061/(ASCE)0733-9445(1987)113:10(2221)
ASCE, 2014, “Seismic Evaluation and Retrofit of Existing Buildings
(ASCE/SEI 41-13),” American Society of Civil Engineers, Reston, VA,
518 pp.
Barbachyn, S. M.; Kurama, Y. C.; and Novak, L. C., 2012, “Analytical
Evaluation of Diagonally Reinforced Concrete Coupling Beams under
Lateral Loads,” ACI Structural Journal, V. 109, No. 4, July-Aug.,
pp. 497-507.
Barney, G. B.; Shiu, K. N.; Rabbat, B. G.; Fiorato, A. E.; Russell, H.
G.; and Corley, W. G., 1980, “Behavior of Coupling Beams under Load
Reversals (RD068.01B),” Portland Cement Association, Skokie, IL, 22 pp.
Canbolat, B. A.; Parra-Montesinos, G. J.; and Wight, J. K., 2005, “Exper-
imental Study on Seismic Behavior of High-Performance Fiber-Reinforced
Cement Composite Coupling Beams,” ACI Structural Journal, V. 102,
No. 1, Jan.-Feb., pp. 159-166.
Engindeniz, M.; Kahn, L. F.; and Zureick, A.-H., 2005, “Repair and
Strengthening of Reinforced Concrete Beam-Column Joints: State of the
Art,” ACI Structural Journal, V. 102, No. 2, Mar.-Apr., pp. 187-197.
Galano, L., and Vignoli, A., 2000, “Seismic Behavior of Short Coupling
Beams with Different Reinforcement Layouts,” ACI Structural Journal,
V. 97, No. 6, Nov.-Dec., pp. 876-885.
Harries, K. A.; Fortney, P. J.; Shahrooz, B. M.; and Brienen, P. J., 2005,
“Practical Design of Diagonally Reinforced Concrete Coupling Beams—
Critical Review of ACI 318 Requirements,” ACI Structural Journal, V. 102,
No. 6, Nov.-Dec., pp. 876-882.
Harries, K. A., and McNeice, D. S., 2006, “Performance-Based Design
of High-Rise Coupling Wall Systems,” Structural Design of Tall and
Special Buildings, V. 15, No. 3, pp. 289-306. doi: 10.1002/tal.296
ACI Structural Journal/January 2018
Hooper, J. D., 2014, “Couplings—Current Practice, Issues, and Oppor-
tunities,” ID ORAL20D, Tenth U.S. National Conference in Earthquake
Engineering, Earthquake Engineering Research Institute, Anchorage, AK,
July 21-25.
Lequesne, R. D., 2011, “Behavior and Design of High-Performance
Fiber-Reinforced Concrete Coupling Beams and Coupled-Wall Systems,”
PhD dissertation, Department of Civil and Environmental Engineering,
University of Michigan, Ann Arbor, MI, 277 pp.
Lim, E.; Hwang, S. J.; Cheng, C. H.; and Lin, P. Y., 2016, “Cyclic Tests
of Reinforced Concrete Coupling Beam with Intermediate Span-Depth
Ratio,” ACI Structural Journal, V. 113, No. 3, May-June, pp. 515-524. doi:
10.14359/51688473
Moehle, J., 2015, Seismic Design of Reinforced Concrete Buildings,
McGraw-Hill Education, New York, 760 pp.
Moehle, J. P.; Ghodsi, T.; Hooper, J. D.; Fields, D. C.; and Gedhada, R.,
2011, “Seismic Design of Cast-In-Place Concrete Special Structural Walls
and Coupling Beams: A Guide for Practicing Engineers,” NEHRP Seismic
Design Technical Brief No. 6, NIST GCR 11-917-11REV-1, produced by
the NEHRP Consultants Joint Venture, NIST, Gaithersburg, MD, 37 pp.
Naish, D.; Fry, J. A.; Klemencic, R.; and Wallace, J., 2013, “Reinforced
Concrete Coupling Beams—Part I: Testing,” ACI Structural Journal,
V. 110, No. 6, Nov.-Dec., pp. 1057-1066.
Naish, D.; Wallace, J.; Fry, J. A.; and Klemencic, R., 2009, “Reinforced
Concrete Link Beams: Alternative Details for Improved Constructability,”
Report to Charles Pankow Foundation, UCLA-SGEL, 103 pp.
Park, R., and Paulay, T., 1975, Reinforced Concrete Structures, John
Wiley & Sons, Inc., New York, 769 pp.
Paulay, T., 1969, “The Coupling of Shear Walls,” PhD dissertation,
Department of Civil Engineering, University of Canterbury, Christchurch,
New Zealand, 432 pp.
Paulay, T., 1977, “Ductility of Reinforced Concrete Shearwalls for
Seismic Areas,” Reinforced Concrete Structures in Seismic Zones, SP-53,
N. M. Hawkins and D. Mitchell, eds., American Concrete Institute,
Farmington Hills, MI, pp. 127-148.
Paulay, T., and Binney, J. R., 1974, “Diagonally Reinforced Coupling
Beams of Shear Walls,” Shear in Reinforced Concrete, SP-42, American
Concrete Institute, Farmington Hills, MI, pp. 579-598.
Tassios, T. P.; Moretti, M.; and Bezas, A., 1996, “On the Behavior and
Ductility of Reinforced Concrete Coupling Beams of Shear Walls,” ACI
Structural Journal, V. 93, No. 6, Nov.-Dec., pp. 711-720.
Teshigawara, M.; Kato, M.; Sugaya, K.; and Matsushima, Y., 1998,
“Energy Absorption Mechanism and the Fluctuation of Shear Force in the
Coupled Shear Walls,” Structural Engineering World Wide 1998—Proceed-
ings, Paper Number T-186-5, Elsevier Science Ltd., 8 pp.
Wong, P. K. C.; Priestley, M. J. N.; and Park, R., 1990, “Seismic Resis-
tance of Frames with Vertically Distributed Longitudinal Reinforcement in
Beams,” ACI Structural Journal, V. 87, No. 4, July-Aug., pp. 488-498.
Wong, P. S. and Vecchio, F. J., 2002, “VecTor2 & Formworks User’s
Manual,” Aug.
Xiao, Y.; Esmaeily-Ghasemabadi, A.; and Wu, H., 1999, “High-Strength
Concrete Short Beams Subjected to Cyclic Shear,” ACI Structural Journal,
V. 96, No. 3, May-June, pp. 392-400.
125
 in
Your
Classroom
Integrate
your classroom!
To support future leaders, ACI has launched several initiatives to engage
students in the Institute’s activities and programs – select programs that
may be of interest to Educators are:
• Free student membership – encourage
students to sign up
• Special student discounts on ACI 318
Building Code Requirements for Structural
Concrete, ACI 530 Building Code Require-
ments and Specification for Masonry
Structure, & Formwork for Concrete manual.
• Access to Concrete International – free to
all ACI student members
• Access to ACI Structural Journal and ACI
Materials Journal – free to all ACI student
members
• Free sustainability resources – free copies
of Sustainable Concrete Guides provided to
universities for use in the classroom
• Student competitions – participate in ACI’s
written and/or team-based competitions
into
• Scholarships and fellowships – students
who win awards are provided up to $10,000
and may be offered internships and paid
travel to attend ACI’s conventions
• ACI Award for University Student Activities –
receive local and international recognition
for your University’s participation in concrete
related activities
• Free access to ACI Manual of Concrete
Practice – in conjunction with ACI’s chapters,
students are provided free access to the
online ACI Manual of Concrete Practice
• ACI online recorded web sessions and
continuing education programs – online
learning tools ideal for use as quizzes or
in-class study material

https://www.concrete.org/educatorsandresearchers/aciinyourclassroom.aspx
 ACI STRUCTURAL JOURNAL TECHNICAL PAPER
Title No. 115-S10

A New Approach to Modeling Tension Stiffening in

Reinforced Concrete
by Angus Murray, Raymond Ian Gilbert, and Arnaud Castel

A novel approach to the modeling of tension stiffening is proposed

based on considerations of the highly non-uniform strain regions
that occur in the concrete surrounding embedded reinforcement
bars near the primary cracks. A finite element study is undertaken to
determine the effect of these disturbed regions on the overall defor-
mation of reinforced concrete tension members. Simple hand calcu-
lations are presented to determine an effective cross-sectional area
of concrete, which describes the overall contribution to axial rigidity
of the cracked concrete for the ideal case of perfect bond. A scalar
damage parameter is then introduced to account for the additional
reduction in the tension stiffening effect caused by deterioration of
bond at the reinforcement-concrete interface. Experimental tension
stiffening studies in the literature (including a recent study by the
authors) are used to calibrate the evolution of interface damage
according to factors such as applied loading and shrinkage.

Keywords: bond; disturbed regions; finite element analysis; reinforced

concrete; tension stiffening.

INTRODUCTION
Load sharing approach
Under service loads, the stiffness of a reinforced concrete
(RC) member is strongly affected by the sharing of tension
forces between the reinforcement and the concrete within the
cracked region of the member (that is, the tension stiffening
effect). Accurate models of tension stiffening are therefore
crucial for the successful design of serviceable RC structures.
Analytical models based on the principle of load sharing
are prominent in the literature because they are conceptu-
ally simple and allow a variety of serviceability effects to
be predicted, often with hand calculations alone. The typical
approach employed by this type of model is to consider the
variations of stress and strain in reinforcement and concrete,
which are governed by assumptions of the magnitude and
distribution of bond stresses τb at the interface between the
two materials (Fig. 1(a)). In this way, the tension stiffening
effect may be quantified as the difference between the strain
in the bare reinforcement, εs2, and the average reinforcement
strain within the cracked region of the member, εsm. The
tension stiffening effect is illustrated in Fig. 1(b) for an RC
tension member subjected to a service load Ns. Fig. 1—(a) Distribution of bond stress τb and reinforcement
Various approaches have been taken in the literature to strain εs for a cracked RC tension member; and (b) load-
characterize the interface behavior between reinforcement deformation responses of cracked reinforced concrete and
and concrete. For example, Marti et al.1 adopted a rigid- bare reinforcement.
plastic bond model leading to a linear variation of strain in
the reinforcement and concrete between adjacent cracks. ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2016-404.R4, doi: 10.14359/51700952, received March 28, 2017, and
Good agreement was achieved with experimental results, reviewed under Institute publication policies. Copyright © 2018, American Concrete
although the success of this model depends largely on the Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
specific choice of bond stress, for which little justification closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.

ACI Structural Journal/January 2018 127

is provided. A similar model was presented by Gilbert and
Ranzi,2 in which the magnitude of the bond stress was modi-
fied to account for the deterioration of tension stiffening that
occurs with increasing load. The effects of other parameters on
the magnitude of the bond stress were also modeled, including
bar diameter and the time-dependent deformation of concrete
due to creep and shrinkage. Alternatively, models based on
linear bond stress distributions (and therefore quadratic vari-
ations of strain) have also been used successfully to describe
the in-service behavior of RC structures.3-6
Assumptions of both uniform and linear bond distributions
broadly corroborate experimental measurements of the strain
variation along embedded reinforcement bars.7,8 However, the
shape of the bond stress distribution is distinctly nonlinear,
and at any point between cracks depends on the local state
of damage at the reinforcement-concrete interface. With
increasing load, regions of high local bond stress (particularly
near the primary cracks) exhibit a softening effect, as internal
cone-shaped cracks and longitudinal splitting cracks begin to
form around the embedded bars.9,10 The resulting bond stress
distribution is both nonlinear and load-dependent, and this is
reflected in some models in the literature.11,12
Another type of approach is to adopt a local bond stress-
slip (τb-s) relationship.13 For simple τb-s models, explicit
solutions may be found to characterize the tension stiff-
ening effect.14 However, most common τb-s models are
piecewise-defined, and neat analytical solutions do not exist
to quantify the tension stiffening effect for cases in which
multiple branches of the τb-δ curve must be considered simul-
taneously. To this end, numerical approaches are usually
favored,15 although these require significant computational
effort, particularly for the analysis of entire structures.
Notwithstanding, the tension stiffening effect is almost
always presented as though it were strictly a one-dimensional
problem, with variations of stress and strain being consid-
ered in the direction of the embedded reinforcement only.
While this is a convenient simplification for the basis of an
analytical model, it necessarily implies that the deformation
of concrete at a specific cross section may be represented
by a single value of stress and strain. However, even at the
service level, it is difficult to justify such an assumption.
Plane sections do not remain plane
In the plane of a primary crack, the entire tension force is
carried by the reinforcement and the concrete is free of stress
and strain. Through the action of bond, a portion of the total
tension force is gradually transferred from the reinforcement
to the surrounding concrete as the distance from the crack
increases. The difference between the bare reinforcement strain
and the average reinforcement strain between the primary
cracks (εs2 – εsm) represents the tension stiffening effect.
However, irrespective of the actual bond stress distri-
bution, the resultant bond force acting upon the concrete
occurs over a region that is many times smaller than the
dimensions of the concrete cross section. This highly local-
ized loading induces a warping effect in the concrete cross
sections, particularly near the primary cracks where bond
stresses are relatively large. Over a sufficient distance from
a crack, Saint-Venant’s principle indicates that the concrete
128
Fig. 2—Concrete axial stress and strain variation close to
and remote from a primary crack.
cross sections should exhibit a gradual transition from
highly non-uniform deformation to essentially plane defor-
mation (Fig. 2). For RC tension members like those consid-
ered in this paper, this transition occurs over a distance
that is roughly in the order of the width of cross section16
(≈2c + db), where c and db are the cover dimension and
reinforcement bar diameter, respectively. After a stabilized
crack pattern has formed, however, significant portions of a
member surrounding the cracks (if not the entire member)
are likely to exhibit this warping effect; the usual assumption
that plane sections should remain plane does not hold true
after the formation of primary cracks.
Some important aspects of the in-service behavior of RC
structures can only be explained by considering this behavior.
One example is the curved surfaces of primary cracks (Fig. 2),
which account for the difference in the measured crack width
at the surface of a member compared to that at the level of
reinforcement.17-19 This non-uniform deformation engenders
a dependence of crack width on the cover dimension, and
this fact is reflected in many crack models in the
literature.2,20,21
The formation of internal cone-shaped cracks near the
primary cracks9 is also largely a result of non-uniform
concrete axial stress variations transverse to the reinforce-
ment. High bond stresses in the area of the primary cracks
induce regions of high tensile stress in the concrete at the
reinforcement-concrete interface and cause cone-shaped
cracks to form at regular intervals around the embedded
reinforcement. Near the surface of the member, however, the
concrete stress is small (Fig. 2) and may even be compres-
sive.17,19 In these regions near the primary cracks, prop-
agation of the internal cracks is effectively arrested and
they typically remain confined within the cover concrete.
Conversely, regions remote from existing primary cracks are
characterized by essentially plane deformation and there-
fore allow for the formation of new primary cracks, which
(unlike internal cracks) occur almost instantaneously over
the entire cross section.
Finally, for RC tension members tested in the laboratory
(whose ends represent initial primary cracks), there is a
significant and measurable difference in stiffness compared
to the ideal uncracked case, which may be observed during
the very first application of load. During the earliest stages
of loading, bond is mainly assured through a combination of
ACI Structural Journal/January 2018
chemical adhesion and micromechanical interlock between
the reinforcement and the concrete,10 and a condition of
“perfect bond” occurs everywhere (that is, no slip occurs along
the reinforcement-concrete interface). However, a condition
of perfect bond is incompatible with a state of plane defor-
mation near a primary crack because it necessarily requires
the concrete in this region to sustain a uniform tensile stress
over its entire cross section. The concrete is essentially free of
stress and strain close to the face of a crack. This incompat-
ibility may only be overcome if the concrete should exhibit
non-plane deformation near a primary crack.
RESEARCH SIGNIFICANCE
The model proposed in this paper provides a more
informed picture of the influence of non-plane deforma-
tion occurring near primary cracks on the stiffness of RC
structures under service loads. The approach outlined in
this paper may be extended to characterize other aspects
of serviceability in RC structures, including cracking and
time-dependent effects.
CHARACTERIZATION OF DEFORMATION
Local axial rigidity
For an RC tension member subjected to an axial force
N, the reinforcement strain εs at a specific cross section is
governed by the local axial rigidity RA (that is, εs = N/RA). At
a point that is sufficiently remote from the abrupt disconti-
nuities in loading (bond stresses) and geometry presented by
a primary crack, a condition of essentially plane deforma-
tion occurs over the entire cross section and the local axial
rigidity is equal to RA1 = EsAs + EcAc, where Es and Ec and
As and Ac are the elastic moduli and cross-sectional areas of
the reinforcement and concrete, respectively. This leads to
the familiar expression for the reinforcement strain in the
uncracked concrete, εs1
N N
ε s1 = = (1)
RA1 Es As + Ec Ac
In the plane of a primary crack, the concrete carries no tension
and the local axial rigidity is simply that of the bare reinforce-
ment (RA2 = EsAs), leading to a reinforcement strain εs2 of
N N
εs2 = = (2)
RA 2 Es As
In the intermediate region between a primary crack and a
cross section characterized by plane deformation, a gradual
transition must occur in the concrete’s contribution to local
axial rigidity (that is, from 0 to EcAc). At any point between
adjacent cracks, this contribution determines the extent to
which the total axial force N is shared between the reinforce-
ment and the concrete, and, therefore, governs the variations
of stress and strain in both materials.
Effective axial rigidity
Whereas the local reinforcement strain εs is governed by
the local axial rigidity RA, it is the average reinforcement
ACI Structural Journal/January 2018
strain εsm that is of primary interest because it determines
the overall tension stiffening effect. Therefore, an effective
axial rigidity RA,ef is introduced that defines the relationship
between an axial force and the average reinforcement strain
(that is, εsm = N/RA,ef). In general, the relationship between
the effective and the local axial rigidity is given by
1 1 b 1
= ∫ dz (3)
RA, ef b − a a RA ( z )
where RA(z) is the local axial rigidity between some points
a and b along the reinforcement. Because RA(z) is bounded
by the limiting values of RA2 and RA1, it follows from Eq. (3)
that RA2 ≤ RA,ef ≤ RA1, as expected. The effective axial rigidity
RA,ef of a cracked RC tension member may be considered to
be governed by two primary effects, namely: 1) the spacing
of primary cracks; and 2) the deterioration of bond at the
reinforcement-concrete interface. In this paper, the influ-
ences of these two effects on RA,ef are considered separately.
First, the individual effect of primary crack spacing on
RA,ef may be highlighted by considering the special case of
perfect bond, for which no slip occurs at any point along
the reinforcement-concrete interface. Perfect bond is
assured during the earliest stages of loading, during which
the primary bond mechanism is a combination of chemical
adhesion and micromechanical interaction between the rein-
forcement and the concrete.10 This condition of zero slip is
significant because it enforces the greatest possible interac-
tion between the reinforcement and the concrete, and there-
fore yields the greatest tension stiffening effect for a given
spacing of primary cracks. For the case of perfect bond,
the effective axial rigidity may be written as RA,ef = EsAs +
EcAc,ef, where Ac,ef (0 ≤ Ac,ef ≤ Ac) is an effective cross-sec-
tional area of concrete.
Where the spacing of primary cracks is relatively wide,
most of a member’s cross sections are located in regions
characterized by plane or nearly plane deformation with local
axial rigidity RA ≈ RA1. The cross sections near the primary
cracks, having local axial rigidity RA ≈ RA2, represent only a
small portion of the member’s total length. As a result, the
member’s effective axial rigidity RA,ef is close to (but less
than) that of an uncracked member and the effective cross
sectional area of concrete, Ac,ef, is close to (but less than) the
actual cross-sectional area of concrete, Ac. Conversely, where
primary cracks are finely spaced, the concrete carries little of
the total tension force, and most cross sections exhibit highly
non-uniform deformation with local axial rigidity RA ≈ RA2.
In this case, RA,ef is close to (but greater than) that of the
bare reinforcement and the effective cross-sectional area of
concrete, Ac,ef, is close to (but greater than) 0.
However, only a modest increase in load is required to
break the chemical and micromechanical bonds that assure a
condition of perfect bond, and slip between the reinforce-
ment and the concrete facilitates a transition of the bond
mechanism to one characterized by bearing of the reinforce-
ment lugs on the cover concrete. With increasing load, internal
cone-shaped cracks and splitting cracks begin to form along
the interface, and this is accompanied by a reduction in
129
the effectiveness of the bond.10 Eventually, damage to the
reinforcement-concrete interface becomes so extensive that
the tension stiffening effect is exhausted, and RA,ef approaches
RA2, regardless of the spacing of primary cracks. The influence
of bond deterioration may be captured by a scalar damage
parameter λ (0 ≤ λ ≤ 1), which modifies the contribution to
axial rigidity of the effective area of concrete EcAc,ef. This
approach is similar to studies by Castel et al.,4 Xu et al.,5 and
Murray et al.6 The case of perfect bond is represented by λ = 0,
whereas complete bond deterioration is represented by λ = 1.
Thus, for the general case in which both primary crack spacing
and bond deterioration are considered, the effective axial
rigidity may be expressed as RA,ef = EsAs + (1 – λ)EcAc,ef and
the average reinforcement strain is given by
N N
ε sm = = (4)
RA, ef Es As + (1 − λ ) Ec Ac , ef
Together, Ac,ef and λ describe the two major mechanisms of
the tension stiffening effect in a cracked RC member. In the
following sections, a finite element (FE) approach is used to
determine Ac,ef for a wide variety of section geometries and
average primary crack spacings. The results of experimental
tension stiffening studies in the literature (including a recent
study by the authors) are then used to calibrate the evolution of
λ according to factors such as applied loading and shrinkage.
FINITE ELEMENT MODELING OF EMBEDDED
REINFORCEMENT
In the previous section, the effective cross-sectional area
of concrete, Ac,ef, was used to describe the influence of
primary cracks alone on the tension stiffening effect in RC
tension members—that is, in the absence of any bond deteri-
oration. Accordingly, Ac,ef is determined for the special case
of perfect bond (λ = 0). Rearranging the expression for RA,ef
and making the substitution λ = 0 yields
RA, ef − Es As
Ac , ef = (5)
Ec
where RA,ef may be determined by a suitable FE analysis. In
this section, the simple case of a cylindrical tension member
is considered to determine values of Ac,ef for a wide variety
of reinforcement ratios and crack spacings. For members
comprising square cross sections, such as those consid-
ered later in this paper, values of Ac,ef have previously been
shown22 to be practically identical to those of cylindrical
segments having the same reinforcement ratio ρ.
The software ABAQUS23 was used to create an axisym-
metric FE model of the shaded region in Fig. 3(a), repre-
senting a segment of a cylindrical RC tension member
situated between two adjacent primary cracks. To present the
results of the FE study in a dimensionless form, the geometry
of the segment is defined in terms of its reinforcement ratio
ρ = As/Ac and the ratio of the crack spacing to the reinforcement
bar diameter s/db. The dimensionless form of Eq. (5) is then
130
Fig. 3—(a) Segment of a cylindrical RC tension member; (b)
schematic diagram of axisymmetric FE model.
1
ρef = (6)
n(ε s 2 /ε sm − 1)
where ρef = As/Ac,ef is the effective reinforcement ratio; n =
Es/Ec is the modular ratio; and εs2/εsm is the ratio between
the bare reinforcement strain εs2 and the average reinforce-
ment strain of an embedded bar εsm, to be determined by
FE analysis.
A schematic diagram of the FE model is shown in
Fig. 3(b). For the case of λ = 0, the dominant bond mecha-
nism is a combination of chemical and physical adhesion,
although some portion of the total bond may be attributed to
bearing action between the reinforcement lugs and the cover
concrete.10 Despite this, a sufficiently accurate picture of the
overall deformation is assumed to be obtained by ignoring
the presence of the reinforcement lugs. This simplification
also avoids unnecessary complication of the numerical
model, which would otherwise require several additional
parameters to define the geometry of the reinforcement lugs.
The validity of this assumption will be tested against exper-
imental results in following sections. The reinforcement is
therefore idealized as a round bar and the element nodes are
tied along the boundary between the reinforcement and the
concrete to satisfy the condition of full strain compatibility.
Both reinforcement and concrete were modeled as linear-
elastic isotropic materials and a range of typical values for
the modular ratio (n = 5 to 8) was considered. A value of Es =
200 GPa (29,000 ksi) was adopted in the analyses, although
the magnitude of both Es and Ec is entirely arbitrary for any
given value of n. Poisson’s ratios for steel and concrete were
taken to be υs = 0.3 and υc = 0.2, respectively. Values of ρ
between 0.001 and 0.1 were considered, corresponding to a
cover dimension in the range of c ≈ 1db to 15db. Values of
s/db were taken within the range of 2 to 50. The combina-
tions of n, ρ, and s/db considered herein cover practically all
realistic cases for cracked RC members. Both reinforcement
ACI Structural Journal/January 2018
and concrete were modeled with a suitably fine mesh of
eight-node axisymmetric elements (CAX8), the dimension
db taken to be the width of 16 finite elements. Based on the
results of a convergence study, this degree of discretization
was found to represent the best compromise between accu-
racy and computation time. For each combination of n, ρ,
and s/db, a value of ρef was determined according to Eq. (6).
These represented approximately 750 separate FE analyses.
For a very fine spacing of primary cracks (s/db → 0), little
opportunity exists for axial force to be transferred from
the reinforcement to the concrete and the effective area of
concrete approaches zero (ρef → ∞). For very widely spaced
primary cracks (s/db → ∞), the end regions characterized by
non-plane deformation are small compared to the total length
of the segment, and the uncracked behavior is approached
(ρef → ρ). A function that satisfies these criteria is given by
ρ( s /db )1.2 + A
ρef = (7)
( s /db )1.2
where the exponent of 1.2 was found to provide the best
overall fit to the entire family of curves produced by the FE
analyses. The variable A has been determined by regression
analysis for each combination of n and ρ, and may be taken as
1.13ρ + 0.0458
A= − 0.048n (8)
ρ + 0.0255
Figure 4 shows the FE results for the case of n = 6 along-
side the curves generated by Eq. (7). The error associated
with Eq. (7) is in the order of 2% for the range of n, ρ, and
s/db considered herein.
CHARACTERIZATION OF SCALAR DAMAGE
PARAMETER
Whereas the influence of primary cracks on the overall
stiffness of an RC tension member may be readily deter-
mined by numerical methods (as in the previous section), the
effect of deteriorating bond on the tension stiffening effect
is governed by a combination of complex, nonlinear effects,
including10: 1) the gradual formation of a series of closely
spaced internal cracks around the embedded reinforcement;
2) the interaction between the resulting comblike concrete
and the reinforcement lugs; 3) localized crushing of the
concrete in front of the reinforcement lugs; and at higher
load levels; and 4) a loss of confinement accompanying the
formation of splitting cracks.
Various models are available in the literature to describe
the deterioration of bond (for example, bond stress-slip
models13-15 and numerical models focused at the meso-
scale24,25), but the computational demand of many of these
approaches is prohibitive and not conducive to hand calcu-
lations. The scalar damage parameter λ may instead be char-
acterized directly based on the results of tension stiffening
experiments in the literature. In this section, the results
of five tension stiffening studies on axially loaded pris-
matic members are used to describe the evolution of λ with
ACI Structural Journal/January 2018
Fig. 4—FE results for the case of n = 6.
increasing load and under the influence of shrinkage. These
are summarized in Table 1.
From each of the studies listed in Table 1, selected tension
stiffening tests were chosen based on the completeness of
the published data, particularly with respect to the load-
deformation curves, material properties, and the spacing of
primary cracks at various load levels. For one study, the
elastic modulus of concrete had to be estimated based on the
observed stiffness of the members before cracking. However,
in most cases, sufficiently detailed data were provided such
that the evolution of λ could be determined directly from the
relevant publications.
Both normal-strength and high-strength concrete members
were investigated with mean compressive strengths fcm
between 21.6 and 81.0 MPa (3130 and 11,750 psi). In most
cases, tension members comprised a square cross section
with a single centrically embedded ribbed steel reinforce-
ment bar. A range of reinforcement bar diameters (12 to
19.5 mm [0.47 to 0.77 in.]) and reinforcement ratios (0.0099
to 0.0309) was investigated. The tests by Fronteddu26 and
Fields and Bischoff27 comprised tension members with
multiple reinforcement bars, for which the tributary concrete
cross section surrounding each bar was square.
The loading regime is characterized by one of two types:
purely monotonic loading (M) or a combination of mono-
tonic loading and cycles of loading and unloading (M-C).
Whereas tests performed with M-C type loading provide a
direct measure of the instantaneous stiffness, tests performed
with M type loading must make an adjustment to the apparent
axial rigidity to account for any deformation associated with
early shrinkage. This is discussed in the following section.
Effect of early shrinkage on residual deformation
For an RC tension member subjected to early shrinkage,
an appropriate adjustment must be made to account for the
shrinkage-induced deformations present in the member
prior to first loading. Figure 5 shows the typical relation-
ship between load and average reinforcement strain for an
RC tension member subjected to an early shrinkage strain of
magnitude, εsh. The shrinkage induces an initial compressive
strain εsm,sh in the reinforcement and shifts the origin of the
monotonic loading curve from O to O′. For members that
131
Table 1—Summary of tension-stiffening experiments analyzed in this study
Cross section Early shrinkage
Study Loading Member fcm, MPa (psi) n db, mm (in.) dimensions, mm (in.) ρ (×10–6)
UC1 100 x 100* (3.94 x 3.94) 0.0309 –234
Fronteddu26 M-C UC2 33.5 (4860) 6.0 19.5 (0.77) 125 x 125* (4.92 x 4.92) 0.0196 –191
UC5 175 x 175 (6.89 x 6.89)
*
0.0099 –130
NS:15M 41.2 (5980) 6.9 –232
16.0 (0.63) 0.0130
Fields and HS:15M 81.0 (11,750) 4.9 –337
M 125 x 125* (4.92 x 4.92)
Bischoff27 NS:20M 54.9 (7960) 6.9 –261
19.5 (0.77) 0.0196
HS:20M 81.0 (11,750) 5.2 –337
N 35 24.8 (3600) 9.9 †
0
Lee and
M M 35 60.7 (8800) 6.6 †
19 (0.75) 150 x 155 (5.91 x 6.10) 0.0124 0
Kim28
H 35 80.4 (11,660) 6.6 †
0
STN12 12 (0.47) 0.0114 0
21.6 (3130) 8.9
Wu and STN16 16 (0.63) 0.0205 0
M 100 x 100 (3.94 x 3.94)
Gilbert29 STS12 12 (0.47) 0.0114 –249
24.7 (3580) 9.3
STS16 16 (0.63) 0.0205 –249
T1 –222
T2 0
T3 0
Present study M-C T4 35.7 (5180) 7.3 16 (0.63) 100 x 100 (3.94 x 3.94) 0.0205 0
T5 0
T6 0
T7 0
*
For tributary cross section.
†
Estimated.
Notes: M is monotonic; and M-C is monotonic with cycles of loading and unloading.

are initially uncracked, the age-adjusted effective modulus

method (AEMM)2,30 may be used to determined εsm,sh

1
ε sm , sh = ε sh (9)
1 + ne ρ

where n̄ e is the age-adjusted effective modular ratio and is

given by n̄ e = n(1 + χφ), where χ is an aging coefficient (and
may be taken as approximately 0.8 for problems involving
relaxation2); and φ is the tensile creep coefficient associ-
ated with the period of early shrinkage. For a short member
whose ends represent initial primary cracks, an appropriate
value of ρef may replace ρ in Eq. (9).
The curve OʹVX therefore represents the monotonic enve-
lope for a tension member subjected to early shrinkage, where
V corresponds to the formation of the first primary crack. For
a sufficiently long member (for which the influence of the end
regions may be ignored), the initial branch of the curve OʹV
is defined by the uncracked axial rigidity RA1 = EsAs + EcAc.
Fig. 5—Load-deformation response of RC tension member Alternatively, where the influence of the end regions is to be
subjected to early shrinkage. accounted for, an appropriate value of ρef may be determined
and RA,ef = EsAs + EcAc,ef applies. An assumption of perfect
bond (λ = 0) is reasonable for at least the initial portion of the
loading, although some localized interface damage is likely to
occur even under moderate loading for members whose ends

132 ACI Structural Journal/January 2018

represent initial primary cracks. This is discussed in greater
detail in the following section. The branch VX describes the
gradual formation of primary cracks and deterioration of bond
with increasing load.
As the tension stiffening effect diminishes under
increasing load, so too does the degree of restraint provided
by the embedded reinforcement to the surrounding concrete.
In turn, any residual deformation (that is, the residual
strain under zero load) associated with the period of early
shrinkage must also reduce. For the case of complete bond
deterioration (λ = 1), the reinforcement strain is unaffected
by any shrinkage in the concrete and the residual deforma-
tion must be close to zero.
Suppose that the tension member was subjected to an axial
force corresponding to the point W along the curve OʹVX,
followed by cycles of loading and unloading at this load level
(that is, M-C type loading). The load-deformation response
departs from the monotonic envelope and instead follows
a new path WY. Near point W, this branch of the curve is
linear and its slope represents the effective axial rigidity for
the current state of cracking and bond damage RA,ef = EsAs +
(1 – λ)EcAc,ef. As the tension force in the member approaches
zero, the rough surfaces of the primary cracks meet and are Fig. 6—(a) Testing of member T1; (b) schematic diagram of
unable to fully close. In experiments, this is observed as test setup showing external measurement frame (all dimen-
an apparent increase in axial rigidity and a deviation from sions in mm) and LVDTs (front and rear LVDTs not shown);
the linear part of the loading-unloading curve. The point and (c) end plate detail indicating positions of rods in plan
Yʹ therefore represents the theoretical residual strain due and locking bolts to secure frame to reinforcement bar.
to early shrinkage for a member with smooth crack faces, (Note: 1 mm = 0.0394 in.)
whose magnitude may be taken as2,30 (λ = 0) is assured at the reinforcement-concrete interface.
For the present experimental study, precise measurements
1 of the average reinforcement strain εsm allowed the initial
ε sm
′ , sh = ε sh (10)
ne ρef axial rigidity RA,exp to be accurately calculated. Unlike many
1+
1− λ tension stiffening studies in the literature (which measure
strain at the concrete surface), average reinforcement strains
where ρef = As/Ac,ef is the effective reinforcement ratio for the were directly measured using a stiff external frame attached
current state of cracking. Equation (10), therefore, accounts to the exposed ends of the reinforcement (Fig. 6(a)). Four
for the load-dependent deterioration of restraint provided LVDTs (one at each of the members’ faces) measured the
by the embedded reinforcement to the now cracked and total elongation of the reinforcement bars and an appro-
damaged concrete. priate adjustment was made to account for the additional
For purely monotonic loading (M), the apparent axial elongation of the bars’ exposed ends (Fig. 6(b) and 6(c)).
rigidity represented by the secant OW would overesti- The members in this study were tested with a 1 MN (225 kip)
mate RA,ef in cases where early shrinkage occurs and is capacity universal testing machine using a combination of
not measured. Rather, RA,ef is defined by the instantaneous monotonic loading and cycles of loading and unloading
response YʹW. For tests involving monotonic loading and (M-C type loading). Members T1 and T2 were initially
cycles of loading and unloading (M-C), this distinction subjected to a monotonic load of 35 kN (7.9 kip) followed
becomes apparent and RA,ef may be determined from the by five cycles of loading and unloading at each of several
slope of YʹW directly. Thus, the adjustment in Eq. (10) allows increasing load levels of 5 kN (1.1 kip) increments until
the effective axial rigidity of cracked RC tension members a maximum load of 100 kN (22.5 kip), close to the yield
to be directly compared, regardless of the type of loading point of the reinforcement. Members T3 to T7, which form
performed or of any early shrinkage that has occurred prior part of an ongoing study, were subjected to a monotonically
to first loading. In any case, after the formation of a stabi- increasing load of 50 kN (11.2 kip) followed by five cycles
lized crack pattern, much of the initial restraint is lost and the of loading and unloading at that load level only. For each of
effect of early shrinkage on residual deformations is small. the seven members tested, RA,exp was determined by fitting
a straight line of best fit to the initial portion of the load
Initial behavior (λ = 0) versus average reinforcement strain graph prior to cracking
The initial response of the RC tension members to mono- for loads between approximately 0 to 5 kN (0 to 1.1 kip).
tonic loading presents a special opportunity for which the The members tested in the present study had nominal
validity of Eq. (7) may be tested. This is because for a spec- dimensions of 100 x 100 x 900 mm (3.93 x 3.93 x 35.43 in.)
imen not previously loaded, a condition of perfect bond and were reinforced with a single 16 mm (0.63 in.) diameter

ACI Structural Journal/January 2018 133

 Fig. 8—Full load-deformation response of member T1.
Primary cracks indicated by numbers 1 to 5; formation of
first splitting cracks indicated by S.

Fig. 7—Initial load-deformation response of members T1 to Table 2—Loads (kN) and positions (mm) of
T7. Early shrinkage for member T1 has been ignored here. primary cracks for present experimental study
Member Crack 1 Crack 2 Crack 3 Crack 4 Crack 5
steel reinforcement bar (ρ = 0.0205). The initial crack
spacing may be taken as the length of the prism and, there- T1 23.8; 218 25.9; 656 32.2; 356 35.0; 519 39.0; 123
fore, s/db = 56.25. Values of Es and Ec were found to be T2 33.3; 505 39.6; 303 45.0; 664 49.6; 142 —
203 and 27.7 GPa (29,400 and 4040 ksi), respectively (n = 7.3).
From Eq. (7), the effective reinforcement ratio is calculated T3 41.5; 676 46.4; 365 — — —
as ρef = 0.0296. Assuming perfect bond (λ = 0), the theoret- T4 42.0; 230 46.0; 407 47.2; 674 — —
ical initial effective rigidity for these members is RA,ef = T5 41.8; 581 44.5; 322 — — —
229 MN (51,500 kip). Even for such a wide spacing of
primary “cracks”, the influence of the non-uniform deforma- T6 40.1; 404 47.2; 607 — — —
tion in the end regions is significant; for comparison, the T7 43.9; 628 46.0; 273 — — —
axial rigidity of the uncracked member is RA1 = 312 MN
Notes: Crack positions are measured from the top of each member; 1 kN = 0.225 kip;
(70,100 kip), some 36% greater. 1 mm = 0.0394 in.
Figure 7 shows part of the load-deformation curves for
each of the seven members tested in this study. The initial Figure 8 presents an overview of the total load-deformation
portions of the curves closely match the effective axial response of member T1 tested in the present experimental
rigidity of RA,ef = 229 MN (51,500 kip) predicted by Eq. (7), study. A summary of the formation of primary cracks for the
whereas the uncracked rigidity RA1 = 312 MN (70,100 kip) present study is given in Table 2. For clarity, only the
obviously overestimates the initial response. The average ascending portion of the fifth load cycle at each load level is
initial rigidity for members T1 to T7 is measured to be 233 plotted. In addition to the primary cracks, splitting cracks
MN (52,300 kip), just 2% greater than that suggested by were observed to form along each of the seven tension
Eq. (7). This result confirms the previous assumption that members tested in the present study (T1 to T7). These were
the effect of the reinforcement lugs on the overall deforma- first observed at loads of approximately 50 kN (11.2 kip),
tion is negligible for the case of perfect bond. and for members T1 and T2, the splitting cracks increased in
length and width with increasing loads up to 100 kN
Deterioration of bond (λ > 0) (22.5 kip). After the formation of a stabilized crack pattern,
In Fig. 7, for a tension force less than approximately 10 kN the stiffness of the load-deformation response is seen to
(2.2 kip), the initial slopes of the loading curves closely match gradually reduce with each load increment. At a load of
the theoretical rigidity for the case of λ = 0. Soon thereafter, a 100 kN (22.5 kip), which is close to the yield point of the
softening effect is observed and the curves gradually begin to reinforcement, the response of the cracked tension member
deviate from the case of perfect bond. Because the spacing of is almost indistinguishable from that of the bare reinforce-
the primary (end) “cracks” is constant before the formation of ment, indicating a near-complete deterioration of bond. The
the first primary crack, any deviation from the initial rigidity apparent increase in axial rigidity due to the closing cracks
must be a result of bond deterioration (that is, λ > 0). At higher (as highlighted in Fig. 5) is evident in the response of T1.
loads, this effect is observed to continue until the tension stiff-
ening effect is almost entirely exhausted.

134 ACI Structural Journal/January 2018

Effect of shrinkage on bond deterioration
As concrete shrinks, it is restrained by embedded
reinforcement and as a result develops tensile stress. This
restraint is provided by bond stresses that develop at the
reinforcement-concrete interface near the primary cracks. For
a member subjected to early shrinkage, therefore, a certain
degree of bond is demanded even before external loads are
applied. Thus, for a given load level, a member that has been
subjected to early shrinkage demands a greater degree of bond
compared to a member that is free of shrinkage. In turn, a
greater degree of bond deterioration would be expected for the
member subjected to early shrinkage.
One way to account for this effect is to consider the shrink-
age-induced tensile stress in the concrete to be the result
of a fictitious mechanical load applied to the ends of the
reinforcement. In this way, the additional deterioration of
bond may be determined by superimposing the fictitious
shrinkage load with existing mechanical loads in the deter-
mination of λ. This is the basic methodology of Murray et al.6
Further explanation of this approach may be found in that
study. The magnitude of this fictitious axial force is determined
using the AEMM,2,6,30 and may be expressed in terms of the
resulting strain increment in the bare reinforcement εs,sh
Fig. 9—Evolution of bond damage with increasing load level.
1 + nρef sm is the average crack spacing. In other words, for deter-
ε s , sh = − ε sh (11)
ne ρef mining values of λ from the load-deformation responses of
1+ the members, bond damage may be deemed to occur within
1− λ
a maximum distance of approximately 10db either side of
where the shrinkage strain εsh is negative. In Eq. (11), λ a primary crack. Elsewhere, a condition of perfect bond is
represents the degree of bond deterioration that exists prior assumed. Importantly, this adjustment is only necessary for
to the development of the shrinkage strain. Equation (11) the initial stages of loading; for each of the tension members
therefore correctly accounts for the load-dependent deterio- analyzed in this study, the primary crack spacing at the stabi-
ration of restraint provided by the embedded reinforcement. lized crack stage satisfied the requirement of sm/db < 20.
Where limited early shrinkage occurs before first loading, a
value of λ = 0 may be safely assumed. Evolution of λ
Each of the tension-stiffening tests summarized in Table 1
Adjustment for wide crack spacing was analyzed to determine values of λ over a wide range of
For a segment located between widely spaced primary load levels. For each test, Eq. (4) was used to yield approx-
cracks, it is difficult to characterize the state of interface imately 10 values of λ based on the observed load-deformation
damage of the entire segment with a single value of λ. Near response and the precise spacing of primary cracks, as
the primary cracks, where reinforcement slip is greatest, reported by the respective authors. For tests conducted with
extensive internal cracking and splitting of the cover concrete purely monotonic loading (M), an adjustment was made in
may occur, while a state of perfect bond may exist in regions accordance with Eq. (10) to account for the influence of any
remote from the cracks. This is particularly so during initial residual deformations associated with early shrinkage on the
loading, when the average primary crack spacing is rela- axial rigidity of the member. Where cycles of loading and
tively large yet interface damage demonstrably begins to unloading were also performed (M-C), the instantaneous
occur even for modest loading (Fig. 7). To account for this, response of the member was used directly. For members
the size of the region near a primary crack that is considered subjected to early shrinkage, a fictitious strain εs,sh (Eq. (11))
to exhibit bond deterioration may be limited. was added to the bare reinforcement strain εs2 to accompany
Because the extent of bond damage (as characterized by each value of λ. A plot of λ versus the adjusted reinforcement
λ) should ideally be independent of the spacing of primary strain (εs2 + εs,sh) is shown in Fig. 9.
cracks (as it is intended to be), it follows that values of λ Overall, there appears to be little discernible influence
determined immediately before and after the formation of on the evolution of λ of concrete strength grade, reinforce-
a single primary crack should be approximately equal. In ment ratio, or reinforcement bar diameter (albeit within the
other words, at a constant load level, the formation of a new limited ranges of these parameters considered in this paper).
primary crack should be able to account for the entire reduc- In many cases, variation of λ at a given load level is greater
tion in stiffness that is observed. For the tension members among nominally identical members than among members
analyzed in this paper, this was found to be the case, from different studies. As demonstrated previously for the
provided that sm/db is less than approximately 20, where present study (Fig. 7), initiation of bond damage is seen in

ACI Structural Journal/January 2018 135

Fig. 9 to occur soon after first loading. Not surprisingly, for
an adjusted reinforcement strain (εs2 + εs,sh) less than approx-
imately 300 × 10–6, values of λ were found to be approx-
imately zero. Thereafter, the degree of bond deterioration
increases rapidly and then gradually approaches a value of
λ = 1 near the yield point of the reinforcement (εs2 ≈ 2500 ×
10–6). A curve of best fit is given by Eq. (12). Approximately
90% of the values of λ calculated from the experimental
results fall within ±0.1 of this curve
λ = 0 for εs2 + εs,sh < 300 × 10–6 (12a)
λ = 1 – exp[–1600(εs2 + εs,sh – 300 × 10–6)]
for εs2 + εs,sh ≥ 300 × 10–6 (12b)
VALIDATION OF MODEL
The tension-stiffening model proposed in this paper may
be used in conjunction with any deterministic or stochastic
model of primary crack formation. Alternatively, the general
approach outlined in this paper may be used to develop a
new model of primary crack formation based on consider-
ations of highly non-uniform strain regions near existing
primary cracks. In this section, the proposed model is used to
simulate the load-deformation responses of three RC tension
members tested by Abrishami and Mitchell.31 The spacing of
primary cracks at the stabilized cracking stage (as reported
by the authors) is ignored in this analysis. Instead, a simple
code-based model of primary crack formation is adopted.32
The first primary crack is assumed to occur at a load of Ncr
when the average concrete stress in a region of plane defor-
mation reaches the effective tensile strength of the concrete
(that is, accounting for the influence of early shrinkage).
Prior to the formation of the first primary crack, the tension
member is treated as a single segment located between end
“cracks,” and the initial response is calculated according to
Eq. (7) assuming that s = L, where L is the total member
length. An assumption of uncracked behavior during the
early loading period would lead to a significant overestima-
tion of initial rigidity compared to the experimental response
(refer to Fig. 7).
At a load greater than or equal to 1.5Ncr, a stabilized crack
pattern is assumed to have formed, for which the average
primary crack spacing (in mm) may be taken as
db
sm = 50 + 0.25k1k2 (13)
ρ 3. For members whose end regions represent initial
where, for a tension member with deformed reinforcement
bars, k1 = 0.8 and k2 = 1.0.32 The number of cracks is assumed
to be proportional to the magnitude of load between Ncr and
1.5Ncr (that is, sm is taken to be inversely proportional to the
load within this range). Overall, good agreement is achieved
between the proposed model and the test results (Fig. 10).
Particularly, the current model improves markedly upon that
proposed in the study by Abrishami and Mitchell.31
136
Fig. 10—Simulation of three tension-stiffening tests by
Abrishami and Mitchell.31
FURTHER RESEARCH
Further experimental investigation would be valuable to
characterize the influence of a wider range of concrete
strength grades, reinforcement ratios, and bar diameters on
the evolution of the bond damage parameter λ. The effects of
long-term shrinkage and sustained loading on time-dependent
changes in tension stiffening form the basis of ongoing
experimental studies by the authors.
CONCLUSIONS
Based on the findings of this study, the following conclu-
sions can be made:
1. The separate influences of primary crack formation and
bond deterioration on the overall load-deformation response
of RC tension members were modeled based on the results
of a FE study and experimental tension stiffening studies in
the literature, respectively.
2. For the case of perfect bond (that is, where no slip
occurs along the reinforcement-concrete interface), the pres-
ence of primary cracks alone represents a major mechanism
of the tension stiffening effect. This is attributed to the exis-
tence of highly non-uniform strain regions in the concrete
surrounding the embedded reinforcement near the primary
cracks. FE modeling for the case of perfect bond is shown
to accurately predict the initial load-deformation response of
RC tension members tested in a recent study by the authors.
primary cracks, deterioration of bond is found to begin soon
after the first application of load. For other types of members,
whose end regions are not subjected to significant tension
forces, deterioration of bond begins immediately after the
formation of the first primary crack. This bond deterioration
is attributed to the formation of internal cone-shaped cracks
around the embedded reinforcement, and at higher loads to
the effects of crushing of concrete in front of the reinforce-
ment lugs and splitting of the cover concrete.
4. Bond deterioration is characterized by the introduction
of a scalar damage parameter λ that depends on the magni-
tude of applied loading and shrinkage. The evolution of λ
with increasing load appears to be insensitive to factors such
ACI Structural Journal/January 2018
as concrete strength grade, reinforcement ratio, and rein-
forcement bar diameter (within the ranges of these parame-
ters considered in this paper).
AUTHOR BIOS
Angus Murray is a PhD Candidate in the School of Civil and Environmental
Engineering at the University of New South Wales, Kensington, Australia.
His research interests include serviceability of concrete structures.
ACI member Raymond Ian Gilbert is Emeritus Professor and Deputy
Director of the Centre for Infrastructure Engineering and Safety in the
School of Civil and Environmental Engineering at the University of New
South Wales. His research interests include serviceability of reinforced
and prestressed concrete structures. He is a member of ACI Committees
209, Creep and Shrinkage in Concrete; and 318, Structural Concrete
Building Code.
Arnaud Castel is an Associate Professor at the School of Civil and Envi-
ronmental Engineering at the University of New South Wales. His research
interests include durability of construction materials, low-carbon concrete,
and corrosion effects in concrete structures.
ACKNOWLEDGMENTS
This research was supported by the Australian Research Council through
ARC Discovery Projects DP110103028 and DP14010052. The authors
wish to thank A. Macken for his assistance with the experiments.
REFERENCES
1. Marti, P.; Alvarez, M.; Kaufmann, W.; and Sigrist, V., “Tension Chord
Model for Structural Concrete,” Structural Engineering International, V. 8,
No. 4, 1998, pp. 287-298. doi: 10.2749/101686698780488875
2. Gilbert, R. I., and Ranzi, G., Time-Dependent Behaviour of Concrete
Structures, Spon Press, London, UK, 2011.
3. Kwak, H. G., and Song, J. Y., “Cracking Analysis of RC Members
Using Polynomial Strain Distribution Function,” Engineering Structures,
V. 24, No. 4, 2002, pp. 455-468. doi: 10.1016/S0141-0296(01)00112-2
4. Castel, A.; Gilbert, R. I.; and Ranzi, G., “Instantaneous Stiffness of
Cracked Reinforced Concrete Including Steel-Concrete Interface Damage
and Long-Term Effects,” Journal of Structural Engineering, ASCE, V. 140,
No. 6, 2014, p. 04014021 doi: 10.1061/(ASCE)ST.1943-541X.0000954
5. Xu, T.; Castel, A.; Gilbert, R. I.; and Murray, A., “Modeling the
Tensile Steel Reinforcement Strain in RC-Beams Subjected to Cycles of
Loading and Unloading,” Engineering Structures, V. 126, 2016, pp. 92-105.
doi: 10.1016/j.engstruct.2016.07.043
6. Murray, A.; Castel, A.; Gilbert, R. I.; and Chang, Z.-T., “Time-
Dependent Changes in the Instantaneous Stiffness of Reinforced Concrete
Beams,” Engineering Structures, V. 126, 2016, pp. 641-651. doi: 10.1016/j.
engstruct.2016.08.025
7. Scott, R. H., and Gill, P. A. T., “Short Term Distributions of Strain and
Bond Stress Along Tension Reinforcement,” The Structural Engineer,
V. 65B, No. 2, 1987, pp. 39-43.
8. Kenel, A.; Nellen, P.; Frank, A.; and Marti, P., “Reinforcing Steel
Strains Measured by Bragg Grating Sensors,” Journal of Materials in
Civil Engineering, ASCE, V. 17, No. 4, 2005, pp. 423-431. doi: 10.1061/
(ASCE)0899-1561(2005)17:4(423)
9. Goto, Y., “Cracks Formed in Concrete around Deformed Tension
Bars,” ACI Journal Proceedings, V. 68, No. 4, Apr. 1971, pp. 244-251.
10. Fédération Internationale du Béton, Bulletin 10: Bond of reinforce-
ment in concrete, fib, Lausanne, Switzerland, 2000.
11. Somayaji, S., and Shah, S. P., “Bond Stress versus Slip Relationship
and Cracking Response of Tension Members,” ACI Journal Proceedings,
V. 78, No. 3, May-June 1981, pp. 217-225.
12. Chan, H. C.; Cheung, Y. K.; and Huang, Y. P., “Crack Analysis
of Reinforced Concrete Tension Members,” Journal of Structural
ACI Structural Journal/January 2018
Engineering, ASCE, V. 118, No. 8, 1992, pp. 2118-2132. doi: 10.1061/
(ASCE)0733-9445(1992)118:8(2118)
13. Fédération Internationale du Béton, “Model Code for Concrete
Structures 2010,” Ernst & Sohn, Berlin, Germany, 2013.
14. Muhamad, R.; Ali, M. S. M.; Oehlers, D. J.; and Griffith, M., “The
Tension Stiffening Mechanism in Reinforced Concrete Prisms,” Advances
in Structural Engineering, V. 15, No. 12, 2012, pp. 2053-2069. doi:
10.1260/1369-4332.15.12.2053
15. Yankelevsky, D. Z.; Jabareen, M.; and Abutbul, A. D., “One-Dimen-
sional Analysis of Tension Stiffening in Reinforced Concrete with Discrete
Cracks,” Engineering Structures, V. 30, No. 1, 2008, pp. 206-217. doi:
10.1016/j.engstruct.2007.03.013
16. Saint-Venant, A. J. C. B., “Mémoire sur la torsion des prismes,”
Mémoires des Savants Étrangers, V. 14, 1856.
17. Watstein, D., and Mathey, R. G., “Width of Cracks in Concrete at the
Surface of Reinforcing Steel Evaluated by Means of Tensile Bond Speci-
mens,” ACI Journal Proceedings, V. 56, No. 7, July 1959, pp. 47-56.
18. Broms, B. B., “Crack Width and Crack Spacing in Reinforced
Concrete Members,” ACI Journal Proceedings, V. 62, No. 10, Oct. 1965,
pp. 1237-1255.
19. Yannopoulos, P. J., “Variation of Concrete Crack Widths through the
Concrete Cover to Reinforcement,” Magazine of Concrete Research, V. 41,
No. 147, 1989, pp. 63-68. doi: 10.1680/macr.1989.41.147.63
20. Gergely, P., and Lutz, L. A., “Maximum Crack Width in Rein-
forced Concrete Flexural Members,” Causes, Mechanism, and Control of
Cracking in Concrete, SP-20, American Concrete Institute, Farmington
Hills, MI, 1968, pp. 1-17.
21. Frosch, R. J., “Another Look at Cracking and Crack Control in
Reinforced Concrete,” ACI Structural Journal, V. 96, No. 3, May-June 1999,
pp. 437-442.
22. Murray, A.; Gilbert, R. I.; and Castel, A., “Study of Disturbed Strain
Regions for Serviceability Analysis of Cracked RC Flexural Members,”
24th Australasian Conference on the Mechanics of Structures and Mate-
rials, H. Hong, H. and C. Zhang, eds., 2017, pp. 97-102.
23. Dassault Systèmes Simulia Corp, ABAQUS/Standard User’s Manual
V. 6.12, Providence, RI, 2012.
24. Michou, A.; Hilaire, A.; Benboudjema, F.; Nahas, G.; Wyniecki, P.;
and Berthaud, Y., “Reinforcement-Concrete Bond Behavior: Experimenta-
tion in Drying Conditions and Meso-Scale Modeling,” Engineering Struc-
tures, V. 101, 2015, pp. 570-582. doi: 10.1016/j.engstruct.2015.07.028
25. Hayashi, D., and Nagai, K., “Meso-Scale Approach to Investigate
the Bond Performance of Anchorage in Reinforced Concrete,” 9th Interna-
tional Conference on Fracture Mechanics of Concrete and Concrete Struc-
tures, V. Saouma, J. Bolander, and E. Landis, eds., University of California,
Berkeley, Berkeley, CA, 2016. doi: 10.21012/FC9.123
26. Fronteddu, L. F., “Response of Reinforced Concrete to Reverse Cyclic
Loading,” master’s thesis, University of British Columbia, Vancouver, BC,
Canada, 1992.
27. Fields, K., and Bischoff, P. H., “Tension Stiffening and Cracking
of High-Strength Reinforced Concrete Tension Members,” ACI Structural
Journal, V. 101, No. 4, July-Aug. 2004, pp. 447-456.
28. Lee, G.-Y., and Kim, W., “Cracking and Tension Stiffening Behavior
of High-Strength Concrete Tension Members Subjected to Axial Load,”
Advances in Structural Engineering, V. 11, No. 5, 2008, pp. 127-137.
29. Wu, H. Q., and Gilbert, R. I., “An Experimental Study of Tension
Stiffening in Reinforced Concrete Tension Members under Short-Term and
Long-Term Service Loads,” UNICIV Report No. R-449, University of New
South Wales, Kensington, NSW, Australia, 2008, 31 pp.
30. Trost, H., “Auswirkungen des Superpositionsprinzips auf Kriech-
und Relaxations-Probleme bei Beton und Spannbeton,” Beton- und Stahl-
betonbau, V. 62, No. 10, 1967, pp. 230-238., 261-269.
31. Abrishami, H. H., and Mitchell, D., “Influence of Splitting Cracks on
Tension Stiffening,” ACI Structural Journal, V. 93, No. 6, Nov.-Dec. 1996,
pp. 703-710.
32. British Standards Institution, “Eurocode 2: Design of Concrete Struc-
tures – Part 1: General Rules and Rules for Buildings – DD ENV 1992-1-
1:1992,” European Committee for Standardization, Brussels, Belgium, 1992.
137
CALL FOR ACTION
ACI Invites You To...
Do you have EXPERTISE in any of these areas?
• BIM
• Chimneys
• Circular Concrete Structures Prestressed by Wrapping
with Wire and Strand
• Circular Concrete Structures Prestressed with
Circumferential Tendons
• Concrete Properties
• Demolition
• Deterioration of Concrete in Hydraulic Structures
• Electronic Data Exchange
• Insulating Concrete Forms, Design, and Construction
• Nuclear Reactors, Concrete Components
• Pedestal Water Towers
• Pipe, Cast-in-Place
• Strengthening of Concrete Members
• Sustainability

Then become a REVIEWER for the

ACI Structural Journal or the ACI Materials Journal.
How to become a Reviewer:
1. Go to: http://mc.manuscriptcentral.com/aci;
2. Click on “Create Account” in the upper right-hand corner; and
3. Enter your E-mail/Name, Address, User ID and Password, and
Area(s) of Expertise.

Did you know that the database for MANUSCRIPT

CENTRAL, our manuscript submission program,
is separate from the ACI membership database?
How to update your user account:
1. Go to http://mc.manuscriptcentral.com/aci;
2. Log in with your current User ID & Password; and
3. Update your E-mail/Name, Address, User ID and Password,
and Area(s) of Expertise.

QUESTIONS?
E-mail any questions to Journals.Manuscripts@concrete.org.
 ACI STRUCTURAL JOURNAL
Title No. 115-S11
Punching Shear Strength of Slabs and Influence of Low
Reinforcement Ratio
by Susanto Teng, Khatthanam Chanthabouala, Darren T. Y. Lim, and Rhahmadatul Hidayat
Presented in this paper are the punching shear tests of 12 high-
strength concrete slabs (fc′ > 100 MPa [14,500 psi]) with various
flexural reinforcement ratios from 0.28 to 1.43% and supported on
columns having various aspect ratios of 1 x 1, 1 x 3, and 1 x 5. The
resulting test data were also used to verify the suitability of the
ACI 318 and Eurocode 2 punching shear equations, especially for
slabs with low reinforcement ratios.
Other influencing factors such as concrete strength and size effect
were also addressed. The authors’ new experimental results will be
combined with 355 existing published data and together they will
be used to evaluate the accuracies and safety of the ACI 318 and
Eurocode 2 methods for punching shear, as well as some methods
proposed by other researchers. A new general method and a simpli-
fied method are also proposed; they are shown to be accurate
and reliable.
Keywords: building codes; column aspect ratio; concrete slabs; flexural
failure; high-strength concrete; punching shear; reinforcement ratio; shear
failure; size effect.
INTRODUCTION
Punching shear failure is one of the critical failure modes
that can happen in a flat-plate floor system. Some of the
important design parameters that can influence the punching
shear strength of concrete slabs include concrete strength,
amount of flexural reinforcement, column rectangularity
(aspect) ratio, and size effect.
One of the parameters that has not been sufficiently inves-
tigated in the past is the influence of the amount of flex-
ural reinforcement ratio ρ (=As/bd), especially the low rein-
forcement ratio. Floor slabs with low reinforcement ratios
are frequently encountered in the design of flat-plate floors
for lightly loaded buildings, such as apartment and condo-
minium buildings or office buildings. The typical values of
the required flexural reinforcement ratios ρ (=As/bd) in the
column strips of the flat-plate floors in those buildings can
vary from approximately 0.6 to 0.8%. It has been known that
when a floor slab is provided with a low flexural reinforce-
ment ratio (ρ of approximately 0.7% or less), the slab may
not be able to attain its punching shear capacity as calcu-
lated using the ACI 318-141 equations for punching shear.
That is because it may fail at a lower load due to widespread
yielding of the flexural reinforcement. Essentially, that is a
flexural failure under punching shear load. A further increase
in loading beyond the peak load will only increase slab rota-
tion or deflection before it reaches the final failure in what
looks like a punching failure—albeit a ductile punching
failure. So far, the ACI 318 or the Eurocode 22 (EC2) equa-
tions for punching shear have not been sufficiently verified
for cases of slabs with low reinforcement ratios where flex-
ACI Structural Journal/January 2018
TECHNICAL PAPER
ural failure under punching load may occur ahead of the
normal punching shear failure.
One typical way to treat this lightly reinforced slab case
in practice is to calculate the punching strength of the slab
using the ACI punching shear equations and compare it with
the failure load (ultimate shear force) that causes flexural
failure of the slab as calculated using the yield line theory.
The lower of these two failure loads will be the punching
shear capacity of the slab. However, it would be much
easier if Code method can treat cases of low reinforcement
ratio directly.
To address the issues mentioned previously, the authors
present an experimental program involving the testing of 12
high-strength concrete slabs having various flexural rein-
forcement ratios and various column aspect (rectangularity)
ratios subjected to concentric punching loads. The authors
have previously tested similar slabs made of normal-strength
concrete to investigate the influence of column aspect ratio
as well as opening.3 Those earlier tests will be useful in
current research. In addition, the authors’ new experimental
results will also be combined with 355 existing published
data and together they will be used to evaluate the accura-
cies of the ACI 318-14 and Eurocode 2 methods, as well as
the methods proposed by other researchers.4-6 A new general
method and a simplified method are also proposed; they are
shown to be accurate and reliable.
RESEARCH SIGNIFICANCE
A new way to treat punching shear strength of concrete
slabs is presented herein. The influence of low flexural
reinforcement ratio is considered rationally, resulting in
two proposed methods: a proposed general method and a
proposed simplified method. The new proposed methods
were verified for accuracy using the authors’ new experi-
ment, which is also presented in this paper, as well as 355
slab data from literature. The proposed methods were also
thoroughly compared with existing methods proposed by
other researchers as well as with the methods of ACI 318
and Eurocode 2.
EXPERIMENTAL PROGRAM
The slab specimens in this experimental program were
designed to cover a low range of reinforcement ratios as
well as some of the higher ranges of reinforcement ratios to
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2016-432, doi: 10.14359/51701089, was received December 8, 2016, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
139
Fig. 1—(a) General dimensions and loading positions; and (b) reinforcement details.
provide a more complete set of data. The dimensions of the
specimens represent an actual model of the column zone of
flat-plate slabs having column-to-column spans of approxi-
mately 5.0 to 6.0 m (16.4 to 19.7 ft).
Specimen details and material properties
The 12 slab specimens tested in this experiment were cate-
gorized into three main groups: S11 series, S13 series, and
S15 series to represent their corresponding column aspect
ratios of 1 x 1, 1 x 3, and 1 x 5, respectively. Figure 1(a)
shows the general dimensions of the slab specimens and
their loading positions while Fig. 1(b) shows the reinforce-
ment details of the 12 slab specimens. Table 1 summarizes
140
important properties of the specimens. The overall dimen-
sions (L1 x L2 x h) of the specimens were 2.2 x 2.2 x 0.15 m
(87 x 87 x 5.9 in.) for the S11 and S13 series, and 2.7 x 2.2 x
0.15 m (106 x 87 x 5.9 in.) for the S15 series. The cross-
sectional dimensions of the column stubs were 200 x
200 mm (7.9 x 7.9 in.) for the S11 series, 600 x 200 mm
(23.7 x 7.9 in.) for the S13 series, and 1000 x 200 mm (39.5 x
7.9 in.) for the S15 series. The height of the column stub was
200 mm (7.9 in.) for all specimens. In addition to the main
top reinforcement, all the specimens were provided with
bottom reinforcement in the form of 10 mm (3/8 in.) diam-
eter bars that were distributed at 260 mm (10.2 in.) spacing,
denoted as T10@260 mm (No. 3@10.2 in.) in Fig. 1(b).
ACI Structural Journal/January 2018
Table 1—Properties of slab specimens
fy, MPa Top reinforcement bar size @
No. Slab ID Dimensions, m (in.) Column size, m (in.) d, mm (in.) fc′, MPa (ksi) (ksi) ρ, % spacing, mm (in.)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
1 S11-028 120 (4.7) 459 (66.5) 0.28 T10@ 260 (No. 3 @ 10.2)

S11 2 S11-050 2.2 x 2.2 x 0.15 0.2 x 0.2 117 (4.6) 537 (77.9) 0.50 T13@ 235 (No. 4 @ 9.2)
112.0 (16.2)
Series 3 S11-090 (87 x 87 x 5.9) (7.9 x 7.9) 117 (4.6) 537 (77.9) 0.90 T13@ 118 (No. 4 @ 4.6)
4 S11-139 114 (4.5) 501 (72.6) 1.39 T16@ 118 (No. 5 @ 4.6)
5 S13-028 120 (4.7) 459 (66.5) 0.28 T10@ 260 (No. 3 @ 10.2)

S13 6 S13-050 2.2 x 2.2 x 0.15 0.2 x 0.6 117 (4.6) 537 (77.9) 0.50 T13@ 235 (No. 4 @ 9.2)
114.0 (16.5)
Series 7 S13-090 (87 x 87 x 5.9) (7.9 x 23.7) 117 (4.6) 537 (77.9) 0.90 T13@ 118 (No. 4 @ 4.6)
8 S13-143 114 (4.5) 501 (72.6) 1.43 T16@ 118 (No. 5 @ 4.6)
9 S15-028 120 (4.7) 459 (66.5) 0.28 T10@ 260 (No. 3 @ 10.2)

S15 10 S15-050 2.7 x 2.2 x 0.15 0.2 x 1.0 117 (4.6) 537 (77.9) 0.50 T13@ 235 (No. 4 @ 9.2)
97.0 (14)
Series 11 S15-090 (106.4 x 87 x 5.9) (7.9 x 39.5) 117 (4.6) 537 (77.9) 0.90 T13@ 118 (No. 4 @ 4.6)
12 S15-143 114 (4.5) 501 (72.6) 1.43 T16@ 118 (No. 5 @ 4.6)
Notes: Concrete cover = 20 mm (0.8 in.); maximum aggregate size = 20 mm (0.8 in.); same reinforcement is provided in both directions; T10 @ 260 mm = 10 mm bars at 260 mm
spacing or No. 3 bar (3/8 in.) at 10.2 in. spacing; d is average effective depth fc′ is cylinder compressive strength of concrete; fy is yield strength of flexural reinforcement; ρ is average
reinforcement ratio (ρx + ρy)/2; ρx = 100Asx/(Lx × dx).

Fig. 2—(a) Typical test setup; and (b) Specimen S13-028 during testing.
The specimen notation represents the main properties of
the slab specimens. For example, Specimen S13-143 indi-
cates a slab specimen with a column aspect ratio of 1 x 3
(β = 3) and flexural reinforcement ratio ρ of 1.43%.
The strengths of the concrete used in all the speci-
mens were approximately 100 MPa (14,500 psi) (refer
to Column 7 of Table 1). The maximum aggregate size is
20 mm (0.8 in.). Some of the advantages of using high-
strength concrete compared to normal-strength concrete
include an increase in cracking load of the slab and a reduc-
tion in deflection at service load level due to higher tensile
strength and higher elastic modulus of higher-strength
concrete. Higher concrete strength also leads to higher dura-
bility in adverse environments.
Instrumentation and loading procedure
A typical test setup and a photograph of a slab during
testing are shown in Fig. 2. Each specimen was placed on
a steel support block and then vertically loaded downward
through four hydraulic jacks that were secured onto the
laboratory strong floor. Each hydraulic jack would apply
the loading by pulling down the steel rod, which transferred
the pulldown force to the spreader beams and then onto
the loading plates (points) on the slab. The actual positions
of the spreader beams and loading points (Fig. 1(a)) were
determined using a finite element software such that the
distributions of stresses near the column zone were close to
those stress distributions in the same slab when loaded under
uniform loading.
Strain gauges were installed on some of the top rein-
forcing bars in both directions and linear variable differen-
tial transformers (LVDTs) were placed below the slab along
the column center lines to measure vertical deflections at
every load increment.
Each specimen was loaded at 20 kN (4.5 kip) load incre-
ment or approximately 5 kN (1.12 kip) increment for each
jack. At every load increment, readings of vertical displace-
ments from LVDTs and steel strains were recorded all the

ACI Structural Journal/January 2018 141

Table 2—Comparison of design methods with current test results
Vexp/Vcalc
Proposed Proposed
c2 x c1, mm d, mm Vexp*, kN Failure ρfs‡, ρ/ρfs§, ACI Eurocode Peiris- General Simplified
Slab ID (in.) (in.) ρ, % (kip) mode† % (4)/(7) 318-14 2 Ghali CSCT Method Method
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
S11-028 120 (4.7) 0.28 280 (62.9) F 0.96 0.29 (F) 0.65 0.96 1.92 1.12 1.19 1.13
S11-050 200 x 200 117 (4.6) 0.50 394 (88.6) F 0.98 0.51 (F) 0.95 1.16 1.58 1.21 1.29 1.22
S11-090 (7.9 x 7.9) 117 (4.6) 0.90 440 (98.9) P 0.78 1.16 (P) 1.06 1.06 1.06 0.99 1.06 1.06
S11-139 114 (4.5) 1.39 454 (102.1) P 0.88 1.58 (P) 1.14 0.99 1.14 0.89 0.98 0.98
S13-028 120 (4.7) 0.28 308 (69.2) F 0.98 0.27 (F) 0.53 0.79 2.18 0.99 1.07 1.01
S13-050 200 x 600 117 (4.6) 0.50 418 (94.0) F 0.80 0.63 (F) 0.74 0.91 1.44 0.97 1.06 1.04
S13-090 (7.9 x 23.7) 117 (4.6) 0.90 558 (125.4) P 0.80 1.13 (P) 0.99 1.00 1.08 0.96 1.08 1.08
S13-143 114 (4.5) 1.43 718 (161.4) P 0.91 1.57 (P) 1.32 1.14 1.32 1.05 1.23 1.23
S15-028 120 (4.7) 0.28 322 (74.4) F 1.07 0.25 (F) 0.49 0.68 2.42 0.97 1.04 0.97
S15-050 200 x 1000 117 (4.6) 0.50 458 (103.0) F 0.85 0.58 (F) 0.70 0.79 1.59 0.95 1.04 1.00
S15-090 (7.9 x 39.5) 117 (4.6) 0.90 658 (147.9) P 0.85 1.06 (P) 1.00 0.93 1.28 1.00 1.12 1.12
S15-143 114 (4.5) 1.43 776 (174.4) P 0.97 1.47 (P) 1.22 0.98 1.22 0.99 1.17 1.17
Minimum 0.49 0.68 1.06 0.89 0.98 0.97
Maximum 1.32 1.16 2.42 1.21 1.29 1.23
Average 0.90 0.95 1.52 1.01 1.11 1.08
Coefficient of variation 0.30 0.15 0.29 0.084 0.082 0.083
*
Vexp is total failure load, including weight of specimen outside distance d from column face and weight of test equipment on top of specimen.
†
Observed failure mode: F is flexural failure; P is punching failure.
‡
ρfs is proposed limiting reinforcement ratio, calculated using Eq. (12a), ρfs is taken to be 0.007 for proposed Simplified Method as shown by Eq. (12b).
§
For ρ/ρfs > 1.0, the predicted failure mode is punching; for ρ/ρfs ≤ 1.0, the predicted failure mode is flexure. ρfs in Column (7) is for proposed Standard Method and calculated using
Eq. (12a).

way to failure. Crack widths were measured by using a crack

detector until near failure.

TEST RESULTS AND DISCUSSIONS

Failure loads and crack patterns
The failure loads Vexp of the 12 specimens are summarized
in Table 2. Each failure load includes the self-weight of the
slab outside the perimeter measured at d away from a column
face and the weight of test equipment placed on top of the
slab. Those slabs that failed abruptly with a sudden drop in
their load-deflection curves are indicated to have failed in
punching mode (P). Those slabs that failed in a more ductile
manner or with widespread yielding of reinforcement prior
to punching failure are indicated to have failed in flexural
(F) mode. All the slabs finally failed in what appears to be
punching failure, even though the slabs might have failed
earlier in flexural mode.
Figure 3 shows the photographs of the crack patterns
at the failure of the S11-series. The crack patterns of the Fig. 3—Crack patterns at failure: (a) S11-028; (b) S11-050;
other eight slabs are shown in Fig. 4. Figure 3 shows fairly (c) S11-090; and (d) S11-139.
clearly that slabs with low reinforcement ratios would have
different crack patterns at failure compared to slabs with the final circumferential crack that comes to the surface from
high reinforcement ratios. Typical crack patterns of slabs the internal inclined shear cracks.
failing in normal punching can be seen in Fig. 3(c) and (d), If the reinforcement ratio is low, however, the circular-
which show Specimens S11-090 and S11-139, respectively. fan-type of crack pattern may not form. Figure 3(a) shows
The crack pattern in each of these cases comprises a set of the crack pattern of Slab S11-028; it can be seen that the
closely spaced radial cracks or circular-fan-type cracks with crack pattern forms straight-line cracks nearly parallel

142 ACI Structural Journal/January 2018

Fig. 4—Crack patterns of S13 and S15 series at failure.
to column lines in both directions with perhaps one diag-
onal crack from the column corner toward a corner of the
slab. The parallel line crack pattern shows that widespread
yielding of the reinforcement has occurred. In the end, the
final failure of the slab would still be a punching failure with
the final circumferential crack occurring very close to the
column. The failure load is at a load no greater than the load
that caused earlier yielding of the flexural reinforcement.
Similar behavior and crack patterns can also be seen in other
specimens with low reinforcement ratios such as S13-028
(ρ = 0.28%) and S15-028 (ρ = 0.28%), as shown in Fig. 4.

Deflections
The load-deflection curves of the 12 specimens are shown
in Fig. 5. Each curve shows the average of four deflection
points located 100 mm (3.9 in.) from the slab edges. Before
cracking, the relationship between the load and deflection is
linear. After the first circumferential crack formed, the slope
of the load-deflection curve would change slightly. Upon
further loading, the change of slope becomes increasingly
more significant as the flexural stiffness of the slab drops
further due to more cracking or widening of cracks. As
expected, the flexural stiffness of the slab with a higher rein-
forcement ratio will degrade less after cracking—that is, its
load-deflection curves have steeper slopes compared to slabs
with low reinforcement ratios.
Figure 5 also shows that slabs with lower reinforcement
ratios are more ductile than slabs with higher reinforcement
ratios. Upon reaching the maximum load, any further load
increment to Specimen S11-028, S13-028, and S15-028
(very low reinforcement ratios) produces no additional Fig. 5—Load-deflection curves of 12 slabs: (a) S11-series;
increase in resistance, but their deformations continue to (b) S13-series; (c) S15-series.
increase until the final failure at the end. The slabs with
the highest reinforcement ratio (S11-139, S13-143, and Strains in flexural reinforcements
S15-143) are the most brittle. Nevertheless, the slabs with Figure 6 shows the summary of steel strain distributions
higher reinforcement ratios also have higher failure loads. at the ultimate load stage in the S11-series, S13-series, and
However, the influence of reinforcement ratio on punching S15-series. From Fig. 6, it can be seen that in slabs with
shear strength is neglected in the ACI 318-14.1 higher reinforcement ratios (ρ ≥ 0.9% or Sxy-090 and

ACI Structural Journal/January 2018 143

Fig. 6—Steel strains in several reinforcing bars near columns at failure loads for: (a) S11 series slabs; (b) S13 series slabs;
and (c) S15 series slabs.
Sxy-143), most of the steel strains drop considerably at
locations beyond 1.5h away from the column face. In slabs
with lower reinforcement ratios (ρ ≤ 0.5% or Sxy-028 and
Sxy-050), the initial failure mode is flexure, and most of the
steel strains can remain high or greater than the yield strain
even at locations near the edge of the slabs. Thus, in slabs
that fail in pure punching, the steel strains reach the yield
strain only near the column.
The consequence of the aforementioned behavior is that
the reinforcement that should be considered effective for
resisting punching load are those within approximately 1.5h
from the column faces or within a width of c2 + 3h for an
interior column.
DESIGN CODES AND EXISTING DESIGN METHODS
In this section and the next section, the ACI 318, Euro-
code 2 methods, and some existing design methods proposed
by researchers are discussed briefly.
ACI 318-141
According to ACI 318-14,1 the punching shear strength Vc
of slabs without shear reinforcement can be determined from
the lowest of the following expressions
1
Vc = (2 + 4 / β) f c′bo d (SI units)
12 (1)
Vc = (2 + 4 /β) f c′bo d U.S. units)
(U
1
Vc = (α s d /bo + 2) f c′bo d (SI units)
12 (2)
Vc = (α s d /bo + 2) f c′bo d (U.S. units) Vc = 16.6kCR(100ρfc′)1/3bod (U.S. units)
1
Vc = f c′bo d (SI units)
3 (3)
Vc = 4 f c′bo d (U.S. units)
where β is the ratio of long to short sides of the column; αs
is taken to be 40, 30, and 20 for interior, edge, and corner
columns, respectively; and bo is the length of the critical
perimeter located at 0.5d away from the column face. For
a circular column, the critical section can also be defined
144
by assuming a square column of equivalent area. The influ-
ence of flexural reinforcement and size effect are neglected
in the ACI Code equations. ACI 318-141 does not limit the
maximum concrete strength fc′ but the value of √fc′ used to
calculate shear strength is limited to 100 psi (8.3 MPa).
Eurocode 22
According to the Eurocode 2,2 the punching shear resis-
tance VRd,c of a slab without shear reinforcement is given in
Eq. (4)
VRd,c = 0.18k(100ρfc′)1/3u1d (SI units)
(4)
VRd,c = 5k(100ρfc′)1/3u1d (U.S. units)
where k is the size effect coefficient, = 1 + 200 / d ≤ 2.0 (SI
units) or = 1 + 8 / d ≤ 2.0 (U.S. units). The ρ is the average
flexural reinforcement ratio ( ρ = ρx ρ y ≤ 0.02 ). The crit-
ical shear perimeter u1 is located at a distance 2d away
from the column faces and it has round corners. Eurocode
2 neglects the effect of column rectangularity in symmetri-
cally loaded slabs.
Existing design methods
Teng et al.3 investigated the effects of slab openings
and column rectangularity in normal-strength reinforced
concrete slabs. They proposed a design equation for the
punching shear strength Vc, as shown in Eq. (5)
Vc = 0.6kCR(100ρfc′)1/3bod (SI units)
(5)
where
kCR = (b2/b1)1/4 < 1.0 (6)
where b1 and b2 are the longer and shorter sides of the crit-
ical shear perimeter, respectively. The use of b1 and b2 in the
column rectangularity factor, kCR, rather than c1 and c2, is
to take into consideration the thickness of the slab in influ-
encing the distribution of stresses around the column. The
critical shear perimeter bo has square corners for both square
and circular columns and is located at a distance d/2 away
ACI Structural Journal/January 2018
from column faces. For a circular column, an equivalent
square column of the same area can be used for bo calcu-
lations. The equation was verified3 with slab data having a
broad range of relevant parameters. However, the applica-
bility of the equation for slabs with very low reinforcement
ratios and very thick slabs has not been verified thoroughly.
Muttoni6 introduced the Critical Shear Crack Theory
(CSCT) for calculating the punching shear strength of slabs
without shear reinforcement. The method assumes that the
punching shear strength VRd of a slab depends on the rotation
ψ of the slab at failure, as well as crack width and aggre-
gate interlock along the critical shear crack. The CSCT
also considers that the rotation ψ of the slab depends on the
applied punching shear force V and a few other parameters.
Failure is reached when the applied punching shear force V
is equal to the punching shear strength of the slab VRd.
The method requires an iterative solution to two equations:
1) the relationship between the punching shear strength VRd
and rotation ψ; and 2) the relationship between the rotation ψ
and applied shear force V. Equations (7) and (8) show these
two equations
0.75
VRd = f c′bo d (SI units)
ψd
1 + 15
d go + d g
(7)
9 way as in Peiris and Ghali’s approach. If the reinforcement
VRd = f c′bo d (U.S. units)
ψd
1 + 15
d go + d g
1.5
r fy  V 
ψ = 1.5 s (8)
d Es  V flex 
where rs is the radius of the slab (equal to half of a slab
length or 0.22L, where L is column-to-column span). Vflex is
the shear force that will cause flexural failure of the slab. bo
is calculated at a distance d/2 away from column edges with
round corners. dgo is set to be 16 mm and dg is the maximum
aggregate size. Muttoni6 states that the size effect in CSCT is
a function of rs or slab span rather than the effective depth d.
The CSCT method can become complex, as it requires
an iterative procedure to solve for VRd. In addition, every
slab geometry, reinforcement ratio, and loading condition
would lead to different yield line pattern, requiring yield
line analysis to obtain Vflex. Consequently, for general design
purposes, Vflex can be approximated4,6 and taken equal to 8m,
where m is the moment capacity per unit width.
Peiris and Ghali5 proposed a straightforward method
to improve the ACI Code method by introducing a reduc-
tion factor, ρ/ρfs, to make it safer for slabs with low flexural
reinforcement ratios. ρfs is the limiting reinforcement ratio.
If the provided reinforcement ρ is reduced to below ρfs, the
slab is assumed to fail in flexural mode instead of the usual
punching mode. For slabs with provided reinforcement ρ >
ρfs, the punching shear strength of the slab Vc is determined
by the usual ACI equations (Eq. (1) to (3)). If the provided ρ
in the slab is reduced below ρfs, the punching shear strength
Vc becomes equal to Vflex—that is, the shear force that causes
ACI Structural Journal/January 2018
the slab to fail in flexural mode. For slabs with ρ < ρfs, Peiris
and Ghali5 assume that Vflex to be a linear function of ρ. So,
the failure load for slabs with ρ < ρfs is then equal to (ρ/ρfs)Vc.
The key lies in the term ρfs. The Vflex can be calculated by
using the yield line theory. For some simple cases, Vflex can
be approximated5 by 8m, where m is moment capacity per
unit width (ACI 318). By setting Vflex = Vc(ACI), the ρ becomes
ρfs, and the ρfs can be obtained. So
8m = 4 f c′bo d , where m = 0.95 fyd2ρ (9)
Then, solving for ρ or ρfs
4 f c′bo d
ρ fs = (U.S. units) (10)
8 × 0.95 f y d 2
DERIVATION OF PROPOSED EQUATION
The authors’ proposed equation shown in Eq. (15) is an
extension of the punching shear strength formula that was
previously introduced by Teng et al.,3 shown earlier in
Eq. (5). New reduction factors to consider low reinforce-
ment ratio and size effect are added.
Factor for low reinforcement ratio, kRR
The effect of low reinforcement ratio is treated in the same
ratio is less than ρfs, then the failure mode changes from
punching to flexural mode and a reduction factor for low ρ
is activated. A suitable yield line pattern is chosen based on
the authors’ experimental results to determine the flexural
strength Vflex of slabs under punching load. Then, by setting
the Vflex to be equal to Vc of Eq. (5) and taking into consid-
eration the reduction factors for size effect and column rect-
angularity, the reduction factor kRR for the effect of low rein-
forcement ratio can be written as (the complete derivation of
the factor kRR is given in the Appendix A,* which is included
in the electronic version of the paper)
kRR = (ρ/ρfs)1/6 (11)
where ρ is the provided reinforcement ratio; and ρfs is given by
3/ 2
 0.6kCR k SZ f c′1/ 3bo d 
ρ fs = (0.01)  2
(100)  (SI units)
 α o (0.95 f y d ) 
3/ 2
(12a)
16.6kCR k SZ f c′1/ 3bo d 
ρ fs = (0.01)  (100)  (U.S. units)
2
 α o (0.95 f y d ) 
where kCR and kSZ are the reduction factors for column rect-
angularity ratio and size effect, respectively (they will be
discussed in the following). αo = [2(c1 + c2)/r + 2π] with r
being the distance from the column face to the point load and
it can be taken as 0.2 times span length.
*
The Appendix is available at www.concrete.org/publications in PDF format,
appended to the online version of the published paper. It is also available in hard copy
from ACI headquarters for a fee equal to the cost of reproduction plus handling at the
time of the request.
145

Fig. 7—Performance of proposed column rectangularity
factor kCR.
Note, however, that the general formula for ρfs given by
Eq. (12a) can also be simplified considerably. By assuming
typical values encountered in practice (refer to Appendix A),
the simplified value of ρfs can be taken as
ρfs = 0.007 or 0.7% (12b)
Thus, the authors now have two proposed methods: 1) the
general method, which uses the ρfs of Eq. (12a); and 2) the
simplified method, which uses ρfs of Eq. (12b). The simplified
method will give equally accurate predictions of punching
failure loads but less accurate predictions of failure modes.
Size effect and column rectangularity ratio
Various approaches have been employed to study the size
effect in concrete structures. A fracture mechanics-based
approach pioneered by Bažant et al.7 leads to a reduction
factor that is proportional to d–1/2. However, due to the
variability of test results, properties of concrete members,
and the influence of reinforcement, an empirical approach
is often used, and this can lead to a reduction factor that is
proportional to d–1/3 to d–1/4 or others.
By observing the available experimental data and prac-
tical design considerations, the size-effect reduction factor
kSZ in the following is adopted for the proposed equation
kSZ = (300/d)1/2 ≤ 1.0 (SI units)
(13)
kSZ = (12/d)1/2 ≤ 1.0 (U.S. units)
kSZ is chosen to be applicable for an effective depth d >
300 mm (12 in.) simply because the current basic equation
for Vc is safe for d of up to 300 mm (12 in.). For d ≤ 300 mm
146
(12 in.), kSZ = 1.0. So, for the majority of slabs in practical
structural design, the size effect factor can be disregarded.
For the column rectangularity ratio, β (= c1/c2), greater
than 1.0, the corresponding reduction factor kCR in Eq. (6),
which was empirically derived, is retained with the exponent
adjusted to 1/3 to suit more data
kCR = (b2/b1)1/3 (14)
Figure 7 shows the graph depicting the normalized shear
stress at failure Vexp/{bod(100ρfc′)1/3} against β (the ratio of
long to short sides of the column, or c1/c2) for the authors’
current slab specimens and other existing slab data from the
literature.3,8-11 Each of selected set of data comprises slab
specimens with different column aspect ratios. The authors’
slab specimens were made of high-strength concrete, while
those from the literature were made of normal-strength
concrete. It can be seen from the trend line for each set of
data that the normalized shear stress decreases as the column
aspect ratio increases. The trend lines are all similar to each
other, including the one for the authors’ current high-strength
concrete slabs.
Figure 7 essentially shows the performance of the column
rectangularity factor, kCR, in comparison with experimental
data. It can be seen that kCR represents the trend line of the
experimental data (including high-strength concrete slabs)
very well. The way ACI treats rectangular column is by the
use of the β coefficient as shown in Eq. (1). It cannot be
shown together in Fig. 7 because ACI uses √fc′ instead of
fc′1/3 in its equation. Nevertheless, the ACI method of treating
rectangular columns as shown in Eq. (1) also represents the
experimental data very closely.
Proposed equation
The general form of the proposed equation can be written as
VC = 0.6kRRkCRkSZ(100ρfc′)1/3bod (SI units)
(15)
VC = 16.6kRRkCRkSZ(100ρfc′)1/3bod (U.S. units)
where kRR, kCR, and kSZ are the reductions factors for rein-
forcement ratio, column rectangularity ratio, and size effect
(refer to Eq. (11), (13), and (14)). ρ is As/bd and should not be
taken greater than 2.5%. Note that in most cases, the reduc-
tion factors need not be calculated, as they will be equal to 1.
For commonly used slabs supported on square columns
(c1/c2 =1) with reinforcement ratios ρ greater than 0.7%, and
effective depths d of 300 mm (12 in.) or less, the proposed
equation in Eq. (15) becomes simple as follows
VC = 0.6(100ρfc′)1/3bod (SI units)
(16)
VC = 16.6(100ρfc′)1/3bod (U.S. units)
COMPARISON WITH EXPERIMENTAL RESULTS
Comparison with authors’ 12 new high-strength
concrete slab specimens
For the purpose of these comparisons, all the load factors,
materials safety factors, and strength reduction factors are
ACI Structural Journal/January 2018
all set equal to 1.0. Shown in Table 2 are the punching shear
strengths of the 12 high-strength concrete slabs presented
earlier in this paper and the predictions of each of the methods
discussed previously. It can be seen in Column (9) that the
predictions of the ACI Code method are unconservative
(Vexp/Vcalc < 1.0) for slabs with low reinforcement ratios of
0.28% or 0.5%, as expected. For slabs with higher ρ, such as
Sxy-090 and Sxy-143 series, the ACI predictions are conser-
vative. The corrections to ACI method that was introduced
by Peiris and Ghali5 (Column (11)) make the corrected ACI
method very conservative for low reinforcement ratio cases.
For the Eurocode 2 predictions shown in Column (10), the
predictions are reasonably conservative, but it cannot take
into account the reduction in punching shear strengths of
slabs supported on rectangular columns (S15 series espe-
cially) and of slabs with very low reinforcement ratio of
< 0.3%, such as S11-028, S13-028, and S15-028. Muttoni
et al.’s6 CSCT method (Column 12) performs very well for
these 12 slabs, with a coefficient of variation of only 8.4%
and an average of Vexp/Vcalc of 1.01.
The authors’ proposed general prediction method (Column
13) is the most accurate, with an average of Vexp/Vc of 1.09
and a coefficient of variation of only 8.2%. The authors’
proposed general prediction method can also predict the
failure modes of the slabs accurately. Column (8) shows the
values of ρ/ρfs from the standard ρfs equation (Eq. 12(a)). If
the ρ/ρfs is less than 1.0, the corresponding slab would fail
in flexural mode (F); otherwise, if ρ/ρfs is equal or greater
than 1.0, the slab is expected to fail in punching (P). Column
(6) shows the actual failure modes of the slabs. Comparing
Column (8) and Column (6), it can be seen that the authors’
general method predicts the actual failure modes very accu-
rately—that is, all the failure modes are predicted correctly.
In fact, both the general and the simplified methods predict
the failure modes of the 12 HSC slabs correctly.
Comparison with slab database of 367 specimens
Table 3 summarizes the punching shear-strength predic-
tions of the methods mentioned previously for the 367
slabs in the database. The material properties and complete
details of the 367 slab data and the detailed comparisons of
the failure load of each slab with the predictions of all the
methods are given in Table A1 in Appendix B.
The current database of 367 slab data was from the authors’
earlier database3 with the addition of slab data from the data-
base of the ACI Subcommittee 445-C12 and new slab data
from the literature shown in Table A1 in Appendix B. All
the slab data in the current database is selected to make sure
that they satisfy all the relevant ACI design requirements.
Slabs having unusual geometries, loading arrangements,
or boundary conditions were excluded, as their punching
strengths would be quite different from those of normal slabs
in typical construction and they deserve special treatment of
their own.
The data are grouped according to three parameters (rein-
forcement ratio, effective depth, and compressive strength)
and divided into three or four different divisions or ranges as
detailed in Fig. 8 and 9 and also in Table 3. A good and reli-
ACI Structural Journal/January 2018
able method is expected to perform well in all the different
ranges or divisions.
Influence of reinforcement ratio—Figures 8(a) and (b)
show the comparison between the experimental failure load
Vexp and ACI 318-141 and Eurocode 22 predictions (Vcalc),
respectively. The ratio of the experimental failure load to the
calculated punching shear strength (Vexp/Vcalc) is plotted in
the y-axis, with the x-axis being three design parameters:
ρ, d, and fc′. Separate statistical analysis is done for each
of the divisions A1, A, B, and C. The horizontal dashed
line in each division represents the average values of the
Vexp/Vcalc for that division. Figures 9(a), (b), and (c) show the
comparisons of the experimental results with the predictions
of CSCT,6 Peiris-Ghali,5 and the authors’ general methods,
respectively. The statistical analysis results for each division
are summarized in a tabulated format in Table 3.
The performance of the ACI equations with respect
to reinforcement ratio is shown in the leftmost chart in
Fig. 8(a). It can be seen that as the reinforcement ratio
increases, the average of Vexp/Vcalc increases as well. Simi-
larly, Table 3 shows for the ACI318-14 predictions that as
the reinforcement ratio increases from Group A1 (Column
(4) or ρ < 0.6%) to C (Column (7) or ρ > 2%), the average
of the Vexp/Vcalc increase from 0.86 to 1.61. This is caused by
the neglect of the reinforcement ratio in the Vc formulas in
the ACI 318-14. Note also that for the region A1(ρ <0.6%)
in both Fig. 8(a) and Table 3, the average is 0.86 and most
of the points are below the 1.0 line. This confirms the earlier
conclusion from the 12 HSC slabs that the ACI 318 method
for punching shear is unconservative for slabs with low rein-
forcement ratio, specifically for ρ less than approximately
0.6% or 0.7%.
The correction procedure introduced by Peiris-Ghali to take
care of possible flexural failure prior to punching failure is
effective for the intended A1 region. From Table 3, in Peiris-
Ghali rows, the average of Vexp/Vcalc in division A1 increases
from 0.86 (ACI average) to a healthy 1.15. However, the
maximum Vexp/Vcalc increases from 1.19 to 2.42. This shows
that the correction can also lead to overconservatism of the
method. Both the Eurocode 2 (EC2) method and Muttoni
et al.’s CSCT method give good predictions with the CSCT
having slightly lower coefficients of variations (average
COV, that is; average of the four divisions = 0.16 compared
to 0.18 for EC2). Figure 9(a) shows that CSCT predictions
in terms of reinforcement ratio has narrower band (better)
compared to Fig. 8(b) for EC2. Note, however, that both
methods have a minimum Vexp/Vcalc of 0.64 for EC2 and 0.57
for the CSCT method. These are quite low.
The authors’ proposed method can be considered the best
with respect to reinforcement ratio. The combined average
of Vexp/Vcalc of the four divisions (sum of four averages
divided by 4) is approximately 1.2, with a combined average
COV of approximately 0.15. The leftmost chart of Fig. 9(c)
shows the authors’ predictions using the general method.
The average of one division is close to the averages of the
other three divisions, and the data are in a reasonably narrow
band, indicating a low COV.
Influence of effective depth—The middle charts in Fig. 8
and 9 and Columns 8 to 10 of Table 3 show the statistical
147
Table 3—Comparison of experimental and calculated failure load by all methods
Vexp/Vcalc Reinforcement ratio ρ, % Effective depth d, mm Concrete strength fc′, MPa
[A1] [A] [B] [C] [A] [B] [C] [A] [B] [C]
ρ< 0<ρ≤ 1.0 < ρ ≤ ρ> d ≤ 100 100 < d < d ≥ 300 fc΄ ≤ 50 50 < fc΄ ≤ fc΄ > 90
Division, units All data 0.6% 1.0% 2.0% 2.0% mm 300 mm mm MPa 90 MPa MPa
No. of data: 367 57 160 175 41 114 247 6 279 63 25
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
Minimum 0.49 0.49 0.49 0.76 1.06 0.70 0.49 0.64 0.51 0.59 0.49
Maximum 2.32 1.19 1.78 1.92 2.32 2.32 1.98 1.10 2.32 1.79 1.98
Vexp/Vcalc
ACI 318-14 Average 1.26 0.86 1.08 1.37 1.61 1.38 1.22 0.86 1.28 1.22 1.17
Coefficient
0.249 0.233 0.258 0.171 0.167 0.223 0.247 0.193 0.240 0.236 0.366
of variation
Minimum 0.64 0.68 0.68 0.64 0.83 0.79 0.64 0.79 0.69 0.64 0.68
Maximum 2.04 1.32 1.77 2.01 2.04 2.04 1.48 0.99 2.04 1.48 1.23
Vexp/Vcalc
Eurocode 2 Average 1.14 1.02 1.11 1.17 1.21 1.26 1.09 0.90 1.16 1.10 1.02
Coefficient
0.174 0.169 0.175 0.183 0.179 0.158 0.156 0.085 0.171 0.181 0.138
of variation
Minimum 0.57 0.64 0.64 0.57 0.76 0.65 0.57 0.68 0.64 0.57 0.89
Maximum 1.82 1.37 1.37 2.19 1.38 1.82 1.48 1.21 1.82 1.31 1.23
Vexp/Vcalc
CSCT Average 1.04 1.02 1.03 1.06 1.02 1.07 1.02 0.86 1.04 1.02 1.04
Coefficient
0.152 0.17 0.145 0.198 0.129 0.156 0.145 0.22 0.151 0.174 0.094
of variation
Minimum 0.75 0.76 0.75 0.76 1.06 0.90 0.76 0.75 0.75 0.76 0.94
Maximum 2.42 2.42 2.42 2.31 2.32 2.32 2.42 1.10 2.32 1.79 2.42
Vexp/Vcalc
Peiris-Ghali Average 1.31 1.15 1.19 1.37 1.61 1.42 1.28 0.89 1.32 1.23 1.47
Coefficient
0.214 0.27 0.217 0.177 0.167 0.185 0.215 0.152 0.201 0.217 0.282
of variation
Minimum 0.73 0.73 0.73 0.76 0.83 0.86 0.73 1.16 0.73 0.83 0.97
Vexp/Vcalc Maximum 1.69 1.55 1.67 1.69 1.50 1.69 1.67 1.28 1.69 1.56 1.43
proposed
Simplified Average 1.21 1.18 1.22 1.22 1.16 1.28 1.18 1.23 1.21 1.24 1.16
Method Coefficient
0.151 0.151 0.151 0.151 0.153 0.127 0.158 0.031 0.150 0.160 0.131
of variation
Minimum 0.76 0.81 0.81 0.76 0.83 0.86 0.76 1.13 0.76 0.83 0.98

Vexp/Vcalc Maximum 1.69 1.49 1.67 1.69 1.50 1.69 1.67 1.24 1.69 1.56 1.43
proposed Average 1.20 1.13 1.20 1.22 1.16 1.27 1.17 1.19 1.20 1.23 1.17
General Method
Coefficient
0.152 0.143 0.151 0.153 0.153 0.129 0.158 0.035 0.151 0.164 0.125
of variation

Notes: Regions A1, A, B, C correspond to the regions in Fig. 8 and 9. 1 mm = 0.0394 in., 1 MPa = 145.0 psi.

analysis results as they are influenced by the effective
depth, d. It can be seen from the middle chart in Fig. 8(a)
that the average of Vexp/Vcalc of the ACI 318-141 becomes
lower as the effective depth d increases. The ACI method
can be unconservative for slabs having an effective depth of
greater than 300 mm (12 in.). Obviously, this is caused by
the neglect of size effect in the ACI equations. The Peiris-
Ghali method, which is a correction to ACI method, does not
address size effect.
The EC22 equation, which has a size effect term, also does
not perform well when d is greater than 300 mm (12 in.)
(refer to the middle chart in Fig. 8(b)). The CSCT method
also does not predict well the punching shear strengths of
slabs that are thicker than 300 mm (12 in.).
Both the authors’ simplified and general methods show
very good agreements with the experimental results. The
average lines from the authors’ general method across the
different divisions of effective depths as shown in Fig. 9(c)
are very close to each other, indicating a very good match
between the equations and actual failure behavior of the
slabs. The average of the three divisions is approximately
1.2 and the average COV is approximately 0.11 for both the
simplified and standard equations.
Influence of compressive strength—The third charts (right-
most) in Fig. 8(a) shows the performance of ACI method
in terms of compressive strength fc′. It can be seen that the
average of Vexp/Vcalc in one division or range (represented
by the horizontal dashed line) is nearly the same as that of

148 ACI Structural Journal/January 2018

Fig. 8—Failure load predictions of 367 test data by: (a) ACI 318-14; and (b) Eurocode 2 2004.

Fig. 9—Failure load predictions of 367 test data by: (a) CSCT; (b) Peiris and Ghali; and (c) Authors’ proposed General Method.
the other divisions. This is true for the other four methods
shown in Fig. 8(b) and 9(a), (b), and (c). However, the ACI
method has the widest spread of results with combined COV
across the three divisions of 0.28. The CSCT method provides
the best predictions with a combined average of Vexp/Vcalc of
1.03 and with a combined COV of 0.14. The authors’ simpli-
fied and general prediction equations both have a combined
average of 1.2 and a combined COV of 0.15. Both CSCT
and the authors’ methods can be considered equally accu-
rate. Table 3 (Column 13, for fc′ > 90  MPa [13,000 psi])
shows that the CSCT method predicts the punching shear
strength of HSC slabs (fc′ > 90 MPa [13,000  psi]) very
well with an average of 1.04 and a COV of 0.094. The next
best method for fc′ > 90 MPa [13,000 psi] is the authors’
general and simplified methods with the COVs of 0.125 and
0.131, respectively, followed by the EC2, Peiris-Ghali, and
ACI methods.
Other parameters, failure modes, and summary—It is
understood that influencing parameters can be interrelated
and they may also influence each other. These are consid-
ered as much as possible. Other parameters that might influ-
ence the punching shear strength include column size to slab
thickness effect (c/d or bo/d), maximum aggregate size da,
and also shear span-depth ratio (r/d or 0.2L/d). However,
those effects were insignificant.
From the aforementioned comparative study, it can be
seen that the authors’ proposed General Method and the
Simplified Method are accurate and reliable. The General

ACI Structural Journal/January 2018 149

Method predicts failure modes more accurately than the
Simplified Method (refer to Table A1 in Appendix B). Of the
27 slabs that failed in flexure, the general method correctly
predicted 20 of them (74% correct) while the simplified
method predicts 21 correctly (78%). Of the 340 slabs failing
in punching, the general method predicts mode of failure
correctly for 316 (93%) of them compared to 274 (81%) for
the simplified method. Nevertheless, both methods predict
the punching shear strength equally accurately and they
give the closest predictions to the experimental failure loads
compared to other methods. The simplified method is espe-
cially very simple to use.
CONCLUSIONS
Based on the authors’ experiments and the accompanying
discussion as well as the comparative study of six different
methods against the database of 367 slabs, the following
conclusions can be made; they are applicable to normal-
strength as well as high-strength concrete slabs:
1. Under a punching load, a slab provided with a low rein-
forcement ratio (approximately less than 0.6 to 0.7%) may
fail in flexure first before it fails in what looks like punching
failure in the end. The failure mode will be ductile and
the failure load will be the load that causes yielding of the
majority of the reinforcement. The failure load will be lower
than predicted by ACI 318-14 equations for punching shear.
2. Slabs failing in pure punching shear mode will tend to
form circular-fan-type crack patterns at failure, while those
failing in flexural mode will tend to form cracks parallel to
column lines. In this case, the final circumferential crack
tends to be very close to the column.
3. The higher the amount of reinforcement provided, the
higher the failure load. However, only those reinforcements
within the width of 1.5h from a column face (or total width
of c2 + 3h for an interior column) will be fully effective
in resisting punching shear stresses and, therefore, can be
considered in influencing the punching shear strength.
4. The ACI 318 method is unconservative for slabs with
low reinforcement ratios (especially for ρ < 0.7%). The
correction procedure proposed by Peiris and Ghali fixes
the problem but it may lead to over-conservatism. The
EC2 method can be unconservative for slabs with very low
reinforcement ratio and supported on rectangular columns.
5. The ACI 318, EC2, and CSCT methods may not be
safe for slabs with an effective depth of more than 300 mm
(12 in.), even though EC2 and CSCT do consider size effect.
ACI needs to consider the inclusion of size effect into the
equation.
6. The authors’ general method and simplified method
both have been shown to be most accurate and reliable with
uniform averages of Vexp/Vcalc and COV across different
ranges of design parameters. The simplified method is espe-
cially simple to use.
150
AUTHOR BIOS
ACI member Susanto Teng is an Associate Professor at the School of Civil
and Environmental Engineering, Nanyang Technological University (NTU),
Singapore. He is a member of ACI Committee 435, Deflection of Concrete
Building Structures, and Joint ACI-ASCE Committees 421, Design of Rein-
forced Concrete Slabs, and 445, Shear and Torsion. His research interests
include shear in beams, slab-column connections, finite element modeling,
and tall buildings.
Khatthanam Chanthabouala is a PhD Candidate in the School of Civil
and Environmental Engineering at NTU. His research interests include
punching shear strength of high-strength concrete and steel fiber-reinforced
concrete slabs.
Darren T. Y. Lim is a PhD Candidate in the School of Civil and Environ-
mental Engineering at NTU. His research interests include concrete tech-
nology and shear behavior of large concrete beams.
Rhahmadatul Hidayat is a Civil Engineer with Mott MacDonald Consul-
tants, Singapore. He received his master’s degree from NTU. His research
interests include tunnel construction for the underground Mass Rapid
Transit system in Singapore.
ACKNOWLEDGMENTS
This research was made possible by the funding provided by the National
Research Foundation (NRF) of Singapore and the support of Nanyang
Technological University, Singapore.
REFERENCES
1. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 2014, 519 pp.
2. CEN, “Eurocode 2—Design of Concrete Structures: Part 1-1—
General Rules and Rules for Buildings,” EN 1992-1-1, Brussels, Belgium,
2004, 225 pp.
3. Teng, S.; Cheong, H. K.; Kuang, K. L.; and Geng, J. Z., “Punching
Shear Strength of Slabs with Openings and Supported on Rectangular
Columns,” ACI Structural Journal, V. 101, No. 5, Sept.-Oct. 2004,
pp. 678-687.
4. Guandalini, S.; Burdet, O. L.; and Muttoni, A., “Punching of Slabs
with Low Reinforcement Ratios,” ACI Structural Journal, V. 106, No. 1,
Jan.-Feb. 2009, pp. 87-95.
5. Peiris, C., and Ghali, A., “Flexural Reinforcement Essential for
Punching Shear Resistance of Slabs,” Recent Development in Reinforced
Concrete Slab Analysis, Design, and Serviceability, SP-287, M. Mahamid
and F. Malhas, eds., American Concrete Institute, Farmington Hills, MI,
2012, pp. 1-16.
6. Muttoni, A., “Punching Shear Strength of Reinforced Concrete Slabs
without Transverse Reinforcement,” ACI Structural Journal, V. 105, No. 4,
July-Aug. 2008, pp. 440-450.
7. Bažant, Z. P.; Yu, Q.; Gerstle, W.; Hanson, J.; and Ju, J. W., “Justifi-
cation of ACI 446 Proposal for Updating ACI Code Provisions for Shear
Design of Reinforced Concrete Beams,” ACI Structural Journal, V. 104,
No. 5, Sept.-Oct. 2007, pp. 601-610.
8. Rosenthal, I., “Experimental Investigation of Flat Plate Floors,” ACI
Journal Proceedings, V. 56, No. 12, Dec. 1959, pp. 153-166.
9. Moe, J., “Shearing Strength of Reinforced Concrete Slabs and Foot-
ings under Concentrated Loads,” Bulletin D47, Portland Cement Associa-
tion, Research and Development Laboratories, Skokie, IL, 1961, 130 pp.
10. Hawkins, N. M.; Fallsen, H. B.; and Hinojosa, R. C., “Influence
of Column Rectangularity on the Behaviour of Flat Plate Structures,”
Cracking, Deflection, and Ultimate Load of Concrete Slab Systems, SP-30,
American Concrete Institute, Farmington Hills, MI, 1971, pp. 677-720.
11. Oliveira, D. R. C.; Regan, P. E.; and Melo, G. S. S. A., “Punching
Resistance of RC Slabs with Rectangular Columns,” Magazine of Concrete
Research, V. 56, No. 3, 2004, pp. 123-138. doi: 10.1680/macr.2004.56.3.123
12. Ospina, C. E.; Birkle, G.; Widianto; and Kuchma, D., “ACI
445C - Punching Shear - Collected Databank,” 2011, https://nees.org/
resources/3660/. (last accessed Dec. 18, 2017)
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S12
Biaxial Interaction and Load Contour Method of Reinforced
Concrete T-Shaped Structural Walls
by Tae-Sung Eom, Hye-Sung Nam, and Su-Min Kang
Reinforced concrete structural walls with T-shaped cross sections
(or T-shaped walls) have been used as an efficient lateral
force-resisting system for building structures. Such T-shaped walls
are subjected to axial compression and combined bending moments
about two orthogonal axes. In the present study, a straightforward
design method for biaxially loaded T-shaped walls is developed
by modifying the existing load contour method. First, a strain
compatibility section analysis method that can estimate the biaxial
bending resistances of arbitrary wall sections is developed and its
validity is verified through comparisons with test results. Then,
a parametric study is performed to investigate the interaction of
biaxial moments at a constant axial load in T-shaped walls. The
parametric results show that, due to the unsymmetrical geometry
of T-shaped sections, the biaxial interaction depends significantly
on the direction of moments and the magnitude of axial compres-
sion. Based on the results, the non-dimensional contour equations
of biaxial moments for T-shaped walls are proposed and a design
procedure of T-shaped walls is established.
Keywords: biaxial interaction; biaxially loaded walls; load contour
method; reinforced concrete; structural wall; T-shaped walls.
INTRODUCTION
In high-rise buildings, nonplanar walls with various
section geometries, such as T-shaped, C-shaped, and
L-shaped cross sections, have been used as efficient lateral
force-resisting structures against earthquake and wind loads.
Such nonplanar walls are subjected to axial compression
and combined lateral loads about two orthogonal directions.
Thus, for safety and economy in wall design, complicated
stress and strain distributions over the nonplanar wall cross
section, caused by a biaxial interaction of axial compression
and biaxial bending moments, should be taken into account
(refer to Fig. 1).
Various strain compatibility section analysis/design
methods have been developed based on advanced analysis
techniques and enhanced computational power (Lau et al.
1993; Ahmad and Weekoon 1995; Chen et al. 2001; de
Sousa and Caldas 2005; Rodriguez-Gutierrez and Aristiza-
bal-Ochoa 1999a,b; Fafitis 2001; Chiorean 2010; Ludovico
et al. 2010; Marmo et al. 2011; Papanikolaou 2012). The
existing methods have been focused on the development of
versatile numerical methods that can accommodate a variety
of section geometries (dimensions, cross-sectional shapes,
and reinforcement arrangements) and materials (such as
concrete, steel, and fiber-reinforced polymer). However,
such numerical methods are not appropriate for general use
in practice because they require a dedicated software incor-
porated with pre- and post-processes.
ACI Structural Journal/January 2018
TECHNICAL PAPER
Alternatively, the equivalent eccentricity method, which
was originally developed for biaxially loaded columns, can
be used for the biaxial design of nonplanar walls (Wight and
MacGregor 2012; Furlong et al. 2004; MacGregor 1973).
In this method, a fraction between 0.4 and 0.8 times the
weak-axis moment is added to the strong-axis moment, and
the wall is then designed for the axial load and increased
strong-axis moment treated as a case of uniaxial bending
and compression. More straightforward and approximate
methods treating the biaxial bending and compression
directly are the load contour method and reciprocal load
method (Bresler 1960; Parme et al. 1966; Portland Cement
Association 1966, 2013; Furlong et al. 2004; Cedolin et al.
2008; El Fattah et al. 2013). Particularly in the load contour
method, the three-dimensional (3-D) interaction surface of
biaxial moments and axial load is converted into a family of
two-dimensional (2-D) load contours representing the rela-
tionships between the moments about two axes at a constant
axial load. Thus, the load contour method is convenient to
understand the biaxial interaction more intuitively. In addi-
tion, through the process of normalization, the load contour
method, which was originally developed for biaxially loaded
columns with solid sections such as rectangular and circular
column sections, can be easily modified for biaxially loaded
walls with various nonplanar section geometries (Hsu 1985,
1987, 1989).
In fact, the behavior and design of biaxially loaded
columns and walls have been an issue in reinforced concrete
design, as summarized in detail in Furlong et al. (2004) and
El Fattah et al. (2013). Most previous studies, particularly
those including straightforward design approaches such as
the load contour method and reciprocal load method, have
been focused on biaxially loaded rectangular or circular
columns with symmetric solid sections. Relatively, research
on the biaxial behavior and design of nonplanar walls with
unsymmetrical thin-walled sections is limited. Hsu (1985,
1987, 1989) studied experimentally and analytically the
behavior of biaxially loaded L-shaped, T-shaped, and
C-shaped walls with various axial load levels and weak axis
to strong axis moment ratios, and reported that the same
load contour method as that used for rectangular columns
can be used for the design of the L-shaped, T-shaped, and
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2016-433.R1, doi: 10.14359/51700917, received January 27. 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
151
Fig. 1—Stresses and strains at ultimate limit state of nonplanar wall sections subjected to axial compression and combined
biaxial moments.
Fig. 2—Strain-based section analysis method: flowchart for computer coding and stress-strain relationships of concrete and
reinforcing steel.
C-shaped walls. However, recent studies using advanced
analysis methods have shown that the biaxial behavior and
load contour shape of nonplanar walls might fundamentally
differ from those of columns (Ludovico et al. 2010; Marmo
et al. 2011; Papanikolaou 2012). For example, as shown in
Fig. 1, because the C-shaped wall is unsymmetrical about the
y-axis, its biaxial behavior might be significantly affected by
the ratio of moments about the weak and strong axes.
In this study, a straightforward biaxial design method for
T-shaped walls is proposed by modifying the existing load
contour method. A section analysis program based on the
strain compatibility method is developed and its validity
is verified by comparisons with the existing tests. Then,
through a parametric study using the developed program,
the interaction of biaxial moment strengths at constant axial
loads in monosymmetric T-shaped wall sections is investi-
gated. Further, based on these investigations, the normal-
ization method and non-dimensional contour equations of
the biaxial moment strengths, which depend on the direction
of loading and the magnitude of axial loads, are proposed.
152
Finally, the procedure and an example to design a biaxially
loaded T-shaped wall are given.
RESEARCH SIGNIFICANCE
Nonplanar walls with flanged sections are mainly
subjected to biaxial moments due to their unsymmetrical
section geometry. The biaxial behaviors of these nonplanar
walls fundamentally differ from those of columns with rect-
angular and circular solid sections. Thus, traditional biaxial
design methods and design aids, which are mostly provided
for rectangular and circular columns, cannot be used for
nonplanar walls. In this study, characteristics of the biaxial
interaction occurring in T-shaped walls are investigated in
detail, and a modified load-contour method is proposed.
STRAIN COMPATIBILITY SECTION ANALYSIS
A strain compatibility section analysis method for
nonplanar walls with arbitrary section geometries is
developed. Figure 2(a) shows the flowchart of the section
analysis. First, input data such as geometric and mate-
ACI Structural Journal/January 2018
Fig. 3—Comparisons of calculated P-M interaction curves with test results. (Note: 1 in. = 25.4 mm; 1 ksi = 6.9 MPa; 1 kip =
4.448 kN; 1 kip-ft = 1.36 kN∙m.)
rial properties are given. Then, by assuming the angle of
the neutral axis (–180 degrees ≤ θ ≤ 180 degrees) and the
depth of the compression zone (c) (refer to Fig. 2(b)), the
stress and strain distributions over the wall cross section are
determined. By numerically integrating the internal tension/
compression forces and moments of concrete and reinforce-
ment elements (that is, δFi, δMxi, and δMyi of element i; refer
to Fig. 2(b)) over the entire wall section, the axial compres-
sion strength Pn and the bending strengths Mnx and Mny about
the x- and y-axes, respectively, are estimated. The x- and
y-axes denote the horizontal and vertical axes, respectively,
passing through the geometrical center of the wall section.
The calculation is repeated for increasing θ and c. Through
these iterative calculations, the complete three-dimensional
Pn-Mnx-Mny interaction surface of the wall section is obtained.
In addition, through post-processing, the Mnx-Mny contours
at a given axial compression strength Pn or the Pn-Mn inter-
action curve at a given biaxial eccentricity can be calculated.
In the section analysis, linear strain distribution is
assumed, as illustrated in Fig. 2(b). For concrete in compres-
sion, a parabolic-linear stress-strain relationship is used as
follows (Wight and MacGregor 2012; Rodriguez-Gutierrez
and Aristizabal-Ochoa 1999a,b; refer to Fig. 2(c))
  2ε  ε  2 
 f c′  c +  c   for − ε co ≤ ε c ≤ 0
  ε co  ε co  

fc = 
 and horizontal axes denote the axial compression load
 − f ′ 1 − (1 − η)  ε c + ε co   for − ε ≤ ε ≤ − ε
 c  − ε + ε   cu c co
  cu co 
ACI Structural Journal/January 2018
where fc and εc are the stress and strain of the concrete, respec-
tively; fc′ (>0) is concrete compressive strength; η is residual
strength ratio of the concrete considering the post-peak soft-
ening behavior; εcu (>0) is ultimate compressive strain of the
concrete; and εco (>0) is strain corresponding to fc′. The tensile
stress of concrete is neglected. The stress-strain relationship
of steel reinforcements is assumed to follow the elastic-per-
fectly plastic behavior with Young’s modulus Es (= 200 GPa
[29,000 ksi]) and yield strength fy (refer to Fig. 2(d)).
For verification, in Fig. 3 and Table 1, the P-M interaction
curves obtained from the proposed method are compared
with the existing wall tests. The wall specimens are biaxi-
ally-loaded C-shaped, T-shaped, and L-shaped wall speci-
mens tested by Hsu (1985, 1987, 1989). Section geometries,
dimensions, and reinforcement details are shown in Fig. 3. The
material properties such as concrete compressive strengths
fc′ and steel yield strengths fy are also shown in Table 1. In
the wall specimens, axial compression loads P are applied
with eccentricities ex and ey in the x- and y-directions,
respectively, from the center of geometry (refer to Fig. 3(a)).
Through those eccentricities, the biaxial moments Mx (= Pey)
and My (= Pex) increasing in linear proportion to the axial
loads P are applied to the walls. The direction angle of
biaxial moments is denoted as ψ, defined as tan–1(Mx/My) or
tan–1(ey/ex). The ex, ey, and ψ values of the wall specimens
are presented in Table 1 (Hsu 1985, 1987, 1989).
Figures 3(b) to 3(e) compare the P-M interaction curves
(1) obtained from the analyses with the test results. The vertical
and bending moment about the x- or y-axis, respectively.
In each figure, two types of theoretical P-M curves are
shown. The P-M curves calculated by the proposed method
153
Table 1—Test variables of nonplanar members considered for verification
Specimen Section Main bars fc′, MPa fy, MPa ex, mm ey, mm ψ, deg Pu, kN
1c 22 No. 3 25.2 357 45.7 76.2 65.6 482
2c 22 No. 3 25.2 357 45.7 69.9 63.1 532
C-shaped:
3c 22 No. 3 25.2 357 45.7 76.2 65.6 459
Type A
4c 22 No. 3 29.2 357 45.7 88.9 69.8 477
5c 22 No. 3 26.9 357 38.1 63.5 65.6 541
8y C-shaped: 18 No. 3 20.4 453 116 78.4 34.1 330
9y Type B 18 No. 3 20.4 453 126 85.6 34.1 294
T-2n 18 No. 3 29.3 448 189 38.1 11.4 216
T-shaped:
T-5n 18 No. 3 33.4 448 147 28.4 10.9 303
Type A1
T-6n 18 No. 3 33.4 532 76.2 14.2 10.6 573
T-1n 18 No. 3 29.3 532 60.2 63.5 46.5 482
T-3n T-shaped: 18 No. 3 29.3 532 53.1 55.4 46.2 619
T-4n Type A1 18 No. 3 33.4 532 58.7 60.2 45.7 528
T-7t 18 No. 3 29.3 532 60.3 63.5 46.5 487
2a L-shaped 14 No. 3 25.9 357 16.3 16.3 45.0 496
4b 14 No. 3 29.0 462 38.9 127 73.0 170
L-shaped
5b 14 No. 3 29.0 462 42.7 140 73.0 156

Notes: 1 ksi = 6.9 MPa; 1 in = 25.4 mm; 1 kip = 4.45 kN.

are denoted as black solid lines, while the theoretical P-M
curves reported by Hsu (1985, 1987, 1989) are denoted as
gray dotted lines. The test results are plotted as white circles
and triangles. It is noted that material strengths (fc′ and fy)
and moment directions (ψ) used for both theoretical P-M
curves slightly differed from those of the tests. The fc′, fy, and
ψ values used for the analyses are shown in Fig. 3(b) to 3(e),
while the actual fc′, fy, and ψ values of each wall specimen
are presented in Table 1. For example, as shown in Fig. 3(b)
and Table 1, the actual fc′ and ψ values of T-shaped wall
specimens 1c, 2c, 3c, 4c, and 5c varied slightly between
fc′ = 25.2 and 26.9 MPa (3.65 and 3.90 ksi) and between
ψ = 63.1 and 69.8 degrees; however, the theoretical Pn-Mnx
and Pn-Mny curves plotted in the figure are those corre-
sponding to fc′ = 24.1 MPa (3.50 ksi) and ψ = 59 degrees,
which are the same as the values used by Hsu (1989), so that
the analysis and test results of the five wall specimens can be
compared in the same figure (refer to Fig. 3(b)). In the esti-
mation of the theoretical Pn -Mnx and Pn -Mny curves by the
proposed method, the characteristic strains of the concrete
are taken as εco = 0.002 and εcu = 0.0038 (Rodriguez-Guti-
errez and Aristizabal-Ochoa 1999a,b). The residual
compressive strength of the concrete at εcu = 0.0038 is
assumed as 0.85fc′ (that is, η = 0.85; refer to Fig. 2(c) and
Eq. (1)).
As shown in Fig. 3(b) to 3(e), the agreements between the
theoretical Pn -Mnx and Pn -Mny curves of the present study
and Hsu’s previous studies are reasonable. In addition, in
most specimens, the test strengths represented as the white
circles and triangles are slightly greater than the theoretical
P-M curves by the proposed method, regardless of the types
of section geometries and the magnitude of axial compres-
sion loads. It is noted that slenderness effect is not included
in Fig. 3 because the specimens tested by Hsu were stocky.
CHARACTERISTICS OF BIAXIAL INTERACTION
OF T-SHAPED WALLS
Biaxial interaction surfaces of nonplanar walls with thin-
walled sections can completely differ to those of columns
with rectangular and circular solid sections. In this study,
T-shaped walls with monosymmetric sections were chosen
for a fundamental study to investigate the characteristics
of the biaxial interaction in nonplanar walls because their
section geometry is relatively simple.
The three-dimensional Pn-Mnx-Mny interaction surface of a
prototype T-shaped wall, obtained by the proposed section
analysis method, is shown in Fig. 4(a). Sectional and material
properties of the prototype wall are also shown in Fig. 4(a).
In fact, it is inconvenient to use such 3-D interaction
surface for wall design. Thus, in the load contour method, as
shown in Fig. 4(b), the interaction surface is converted into
two-dimensional Mnx-Mny contours, each of which represents
the nominal moment strengths about the x- and y-axes at a
constant axial strength Pn. The Mnx-Mny contours, repre-
sented with different line types and weights, correspond to
the axial compression ratios of n = 0.0, 0.1, 0.2, 0.3, 0.4,
and 0.5, where n = Pn/Agfc′. Each Mnx-Mny contour includes
results of section analyses for neutral axis angles varying in
the range of –180 degrees ≤ θ ≤ +180 degrees. The neutral
axis angle, defined along the counter-clockwise direction
about the positive x-axis, is denoted as θ (= –180 to +180
degrees; refer to Fig. 4(a)). The positive and negative angles
denote the direction of the compression zone. For example,
θ = +90 and –90 degrees indicate that the neutral axes are
parallel to the y-axis in both cases, but the compression
zones exist in the right and left sides of the neutral axis,
respectively. For distinction, the Mnx-Mny contours corre-
sponding to –180 degrees ≤ θ ≤ 0 degrees and 0 degrees ≤ θ
≤ +180 degrees are denoted as gray and black lines, respec-
tively. From Fig. 4(b), the characteristics of the biaxial inter-
action of mono-symmetric T-shaped walls can be summa-
rized as follows:

154 ACI Structural Journal/January 2018

Fig. 4—Conversion Pn-Mnx-Mny interaction surface into family of Mnx-Mny contours at constant axial loads. (Note: 1 in. = 25.4
mm; 1 kip-ft = 1.36 kN∙m.)
Fig. 5—Variation of center of compression depending on angles of neutral axis (n = 1 and 0.5).
1. Because the T-shaped wall section is unsymmetrical
about the y-axis, the Mnx-Mny contours also have unsym-
metrical geometries depending on the sign of Mny. This is
attributed to the unsymmetrical section geometry made up
of relatively thin flange and web, as follows. Figure 5(a)
shows three neutral axis angles of θ = –45, 0, and 45 degrees
in the prototype wall under the same axial compression
load of Pn = 0.1Agfc′ (that is, n = 0.1). In each case, the
compression zone is represented as the shaded area and the
center of internal compression forces are denoted as black
dot. In addition, the distances between the center of
compressive forces to the x- and y-axes are denoted as yo
and xo, respectively. If the neutral axis angle decreases
from θ = 0 to –45 degrees (refer to Fig. 5(a2)), the center of
compressive forces does not change significantly, and conse-
quently Mnx and M ny remain almost unchanged. Thus, as
shown in Fig. 4(c), the points corresponding to θ = 0 and
–45 degrees in the Mnx-Mny contour at n = 0.1 are almost in the
same location. In contrast, if the neutral axis angle increases
from θ = 0 to +45 degrees so that the neutral axis intersects
both the flange and web (refer to Fig. 5(a3)), the center of
compressive forces moves upward close to the x-axis and to
the right crossing the y-axis. In other words, Mnx decreases
and Mny changes from negative to positive. Thus, in the
Mnx-Mny contour at n = 0.1 of Fig. 4(c), the point corre-
sponding to θ = +45 degrees, denoted as a triangle, is
displaced to the first quadrant.
ACI Structural Journal/January 2018
2. The shapes of Mnx-Mny contours vary with the magni-
tude of axial compression loads. For example, as shown in
Fig. 4(c), the Mnx-Mny contour shape resembles a triangle at
n = 0.1, but changes close to a rectangle at n = 0.5. This can
be easily understood by comparing the location of the neutral
axis and the center of compressive forces at n = 0.1 with
those at n = 0.5 (refer to Fig. 5(a) and 5(b)). Figure 5(b) shows
three neutral axis angles of θ = –45, 0, and 45 degrees at the
same axial compression ratio of n = 0.5. The neutral axes
at n = 0.5 are above the web due to the increased axial
compression load. Thus, when the neutral axis angle
increases from θ = –45 to +45 degrees, Mnx remains almost
unchanged because yo does not change significantly. On the
other hand, Mny changes significantly from negative to posi-
tive as the center of compressive forces moves from the left
of the y-axis at θ = –45 degrees to the right at θ = +45 degrees.
3. In T-shaped wall sections, the geometrical center does
not coincide with the shear center. Thus, the neutral axis angle
θ differs from the direction of biaxial moments ψ, defined as
tan–1(Mnx/Mny). For example, as shown in Fig. 4(c), although
the neutral axis is parallel to the x axis (that is θ = 0 degrees),
the moment about the y-axis, Mny, cannot be zero (ψ ≠ 0
degrees). This indicates that the direction of bending differs
from the direction of lateral loading. Consequently, T-shaped
walls resist biaxial bending accompanied by torsion and
twisting deformation. In Fig. 6, although the lateral load acts
along the y-axis and the wall is thus subjected to uniaxial
155
Fig. 6—Twisting of cantilever T-shaped wall due to lateral load acting along y-axis.
Fig. 7—Mapping of Mnx-Mny contour into non-dimensional mx-my contour through line transformation.
bending about the x-axis, a twisting deformation occurs in
the wall.
CONTOUR EQUATIONS FOR T-SHAPED WALLS
Normalization of contours
The investigation results of the previous section show
that, to address the characteristics of biaxial interaction
in T-shaped walls, the effects of the direction of bending
moments and the magnitude of axial compression load on
the contour geometry should be taken into account.
Figure 7 shows the normalization process for unsym-
metrical Mnx-Mny contours of T-shaped walls. In the figure,
points on the Mnx-Mny contour corresponding to θ = 0, 90,
and –90 degrees are denoted as white and gray circles for
distinction. Mnx0 and Mny0 are the moment strengths about
the x- and y-axes at θ = 0 degrees, respectively, and Mny90
and M ′ny90 are the moment strengths about the y-axis at θ
= 90 and –90 degrees, respectively. Because T-shaped wall
sections are symmetrical about the x-axis, the contour shape
is the same for positive and negative Mnx. Thus, the contour
of only positive Mnx (that is, –90 degrees ≤ θ ≤ +90 degrees)
is considered for simplicity. For distinction, the contours for
–90 degrees ≤ θ ≤ 0 degrees and 0 degrees ≤ θ ≤ 90 degrees
are denoted as solid and dashed lines, respectively. As
shown in Fig. 7(a), the Mnx-Mny contours for –90 degrees ≤
θ ≤ 0 degrees and 0 degrees ≤ θ ≤ 90 degrees significantly
differ in size because the neutral axis angle θ differs from
the biaxial moment angle ψ (compare the white and shaded
areas). To relieve the asymmetry in the contour shape, the
156
portions at 0 degrees ≤ θ ≤ +90 degrees and –90 degrees ≤
θ ≤ 0 degrees of Mnx-Mny contour are mapped to the non-di-
mensional mx-my contours at the first and forth quadrants,
respectively, through a linear transformation defined as
follows (refer to Fig. 7(b))
M nx
mx = for − 90 deg ≤ θ ≤ 90 deg (2)
M nx 0
 M ny − M ny 0
 for 0 deg ≤ θ ≤ 90 deg [or M ny − M ny 0 ≥ 0]
M − M ny 0
my =  ny 90
 M ny − M ny 0
for − 90 deg ≤ θ ≤ 0 deg [or M ny − M ny 0 < 0]
 M ny′ 90 − M ny 0

(3)
As shown in Fig. 7, through the normalization by the
proposed linear transformation, the unsymmetrical char-
acteristics of the Mnx-Mny contours, depending on design
parameters, are significantly alleviated in the mx-my contours.
Elliptic contour equations
In the load contour method, the non-dimensional mx-my
contour is basically defined as the following elliptical equation.
(mx)α + (my)α = 1 (4)
where α is the factor affecting the shape of the contour varying
with design parameters such as material strength, section
ACI Structural Journal/January 2018
Fig. 8—Mnx-Mny and mx-my contours for prototype T-shaped wall sections varying with axial load ratios (n = 1 to 0.5). (Note:
1 in. = 25.4 mm; 1 ksi = 6.9 MPa; 1 kip-ft = 1.36 kN∙m.)
geometry, axial compression load, and biaxial moment
direction. According to Bresler (1960) and the PCA Notes on
ACI 318-11 (Portland Cement Association 2013), the values
of α in the rectangular and circular columns vary between
1.15 and 1.55. Hsu (1985, 1987, 1989) suggested α = 1.5
for L-shaped, T-shaped, and C-shaped walls. However, it is
inappropriate to use such a single value of α for T-shaped
walls, considering that the unsymmetrical contour shape is
significantly affected by the sign of moments (that is, posi-
tive and negative Mny) and the magnitude of axial loads (n).
Thus, α values appropriate for T-shaped walls are investi-
gated through a numerical parametric study as follows.
Figure 8(a) shows the cross sections of five T-shaped walls
used for the parametric study. In the figure, the symbols t, b,
and h denote the wall thickness (= 200 mm [8 in.]), web
length, and half the flange length, respectively. By varying
h and/or b between 500 and 4000 mm (19.7 and 157.5 in.),
the shape factor γ, which is defined as the ratio of the half
flange length to the web length (= b/h), is gradually increased
from 0.25 to 0.5, 1.0, 1.5, and 2.0. In the parametric study,
the concrete compressive strength and steel yield strength
are taken as fc′ = 30 MPa (4350 psi) and fy = 400 MPa
ACI Structural Journal/January 2018
(58,000 psi), respectively. In addition, it is assumed that
longitudinal reinforcements are uniformly distributed over
the entire wall section with a reinforcement ratio of ρ = 0.01.
The Mnx-Mny and mx-my contours of five T-shaped wall
sections are shown in Fig. 8(b) to 8(f). In each figure, the
contours are estimated as follows. By using the proposed
strain compatibility section analysis method, the three-
dimensional Pn-Mnx-Mny interaction surfaces of T-shaped
walls are obtained and the Mnx-Mny contours corresponding
to n = 0.1 to 0.5, shown in Fig. 8(b1) to 8(f1), are then plotted.
The Mnx-Mny contours are converted into non-dimensional
mx-my contours presented in Fig. 8(b2) to 8(f2) through the
linear transformation following Eq. (2) and (3). In Fig. 8, for
distinction, the contours for no or low axial compression
ratios of n = 0, 0.1, and 0.2 are denoted as thin solid lines
connecting solid diamonds, circles, and triangles, respec-
tively, while the contours for relatively high axial compres-
sion ratios of n = 0.3, 0.4, and 0.5 are denoted as thin dashed
lines connecting voided diamonds, circles, and triangles,
respectively. It should be noted that the maximum design
axial compression strength is ϕPn,max = ϕ · 0.8(0.85fc′[Ag –
As] + fy As) and, when substituting ϕ = 0.65, ϕPn,max is less
157
Fig. 9—Effects of other design variables such as reinforcement ratio, reinforcement arrangement, concrete strength, and steel
yield strength on normalized mx-my contour shapes. (Note: 1 in. = 25.4 mm; 1 ksi = 6.9 MPa.)
than 0.5Ag fc′. Thus, in this study, the axial compression ratio
is limited to 0 ≤ n ≤ 0.5.
As shown in Fig. 8(b2) through 8(f2), the mx-my contours of
T-shaped wall sections differ in their geometries depending
on the neutral axis angle θ. In the region of 0 deg ≤ θ ≤ +90
degrees (that is, in the first quadrant), the mx-my contours
expand as the axial compression ratio increases from n = 0
to 0.5. This indicates that a greater α can be used for high
axial loads. However, in the region of –90 degrees ≤ θ ≤
0 degrees (that is, in the fourth quadrant), the correlation
between the axial compression ratio n and the shape of the
mx-my contours (or α) is relatively weak. In both regions,
the effects of the shape factor γ (= b/h = 0.25 to 2.0) on the
mx-my contour shape are also limited. Thus, based on these
results, the α values for biaxially-loaded T-shaped walls are
proposed as follows
3n + 1.05 for 0 deg ≤ θ ≤ 90 deg
α= (5)
2.0 for − 90 deg ≤ θ ≤ 0 deg
In Fig. 8(b2) to 8(f2), the proposed mx-my contours, repre-
sented as thick solid and dashed lines, are compared with
the analysis results (that is, thin gray lines with solid/voided
diamonds, circles, or triangles). The proposed contours are
obtained by substituting the α values defined in Eq. (5) into
Eq. (4). For clarity, the proposed contours in the region of
0 degrees ≤ θ ≤ +90 degrees are plotted only for n = 0 (that
is, α = 1.05) and 0.5 (that is, α = 3.05). In most cases, the
proposed α values are conservative compared to the analysis
results. However, as presented in Fig. 8(e2) and 8(f2), the
proposed α = 2.0 in the region of –90 degrees ≤ θ ≤ 0 degrees
is slightly overestimated for n = 0. Such overestimation is
limited in the Mnx-Mny contour, when considering the fact
that the mx-my contours at low axial compression ratio are
158
significantly modified in the region of –90 degrees ≤ θ ≤ 0
degrees through the linear transformation (refer to Fig. 7).
Effects of other design parameters
According to the PCA Notes on ACI 318-11 (Portland
Cement Association 2013), the mx-my contours can also be
affected by other parameters such as reinforcement ratio (ρ),
reinforcement arrangement (distributed and end-concen-
trated), concrete strength (fc′), and steel yield strength (fy). The
effects of such design parameters are investigated through
additional analyses for the prototype wall with a shape factor
γ = 1.0. Figure 9(a) compares the mx-my contours for ρ = 0.005,
0.01, and 0.02. In the cases of ρ = 0.005 and 0.01, the differ-
ence in the contour shapes is not significant (refer to Fig. 9(a1)
and 9(a2)). However, as the reinforcement ratio increases to
ρ = 0.02, the mx-my contours expand in the region of 0 degrees ≤
θ ≤ +90 degrees (that is, in the first quadrant; refer to
Fig. 9(a3)). This indicates that the α values defined in Eq. (5)
are conservative for ρ = 0.005 to 0.02.
Figure 9(b) shows the mx-my contours of the prototype
T-shaped wall section with end-concentrated reinforce-
ment arrangement. As shown in Fig. 9(b), the reinforcement
ratio at the ends of the flange and web is increased to ρe =
0.023, but the overall reinforcement ratio is maintained as ρ
= 0.01. When compared with Fig. 9(a2), changes in the
contour shapes due to distributed and end-concentrated
reinforcement arrangements are almost negligible. Figures 9(c)
and 9(d) show the mx-my contours for an increased concrete
strength fc′ = 40 MPa (5800 psi) and for an increased steel
yield strength fy = 600 MPa (87 ksi), respectively. Compared
with the contours of the prototype wall shown in Fig. 9(a2),
changes in the contour shapes due to the increased concrete
strength and steel yield strength are also limited. In summary,
Fig. 9(b) to 9(d) show that the effects of the reinforcement
ACI Structural Journal/January 2018
Fig. 10—Proposed Mnx-Mny contour for biaxial design of T-shaped walls.
arrangement and material strengths on the contour shapes
are limited.
DESIGN APPLICATION
For biaxial design of T-shaped walls, the ϕMnx-ϕMny
contour at a constant axial compression strength ϕPn is
presented in Fig. 10 (ϕ = strength reduction factor). By
substituting Eq. (2) and (3) into Eq. (4), design elliptical
contour equations can be defined as follows.
For 0 degrees ≤ θ ≤ +90 degrees (or Muy – ϕMny0 ≥ 0):
 M ux 
α
 M uy − φM ny 0 
α
α α
m +m =
ux uy  +  ≤ 1.0 (6a)
 φM nx 0   φM ny 90 − φM ny 0 
For –90 degrees ≤ θ ≤ 0 degrees (or Muy – ϕMny0 < 0):
α α
 M ux   M uy − φM ny 0 
muxα + muyα =   +  ≤ 1.0 (6b)
 φM nx 0   φM ny′ 0 − φM ny 0 
In these equations, Mux and Muy are the design moments
about the x- and y-axes, respectively, acting on T-shaped
walls, and all nominal moment strengths are multiplied
by the ϕ factor for design. In addition, the α factor corre-
sponding to a design axial load Pu (that is, n = Pu/[Agfc′]) is
determined from Eq. (5). To draw a complete design contour
for T-shaped walls, characteristic moment strengths such
as Mnx0, Mny0, Mny90, and M n′ y90 at a given axial load Pu (=
ϕPn) should be estimated first, as denoted with white and
gray circles in Fig. 10. Such characteristic design moment
strengths can be determined without significant computa-
tional efforts because the neutral axes are parallel to the x-
or y-axis.
A design example of a biaxially loaded T-shaped wall
using the proposed method is presented as follows. Figure 11(a)
shows the section properties and material strengths of the
wall. The factored design axial load, moment about the x-axis,
and moment about the y-axis are Pu = 2832 kN (637 kip),
Mux = 5500 kN·m (4057 kip-ft), and Muy = –2000 kN∙m
ACI Structural Journal/January 2018
(–1475 kip-ft), respectively. The biaxial design of the
T-shaped wall is performed according to the following
procedures:
1. Assume a longitudinal reinforcement ratio ρ of the wall
section.
2. By performing section analyses for the neutral axis
angles θ = –90, 0, and +90 deg, two-dimensional ϕPn-ϕMnx
and ϕPn-ϕMny interaction curves are obtained. Then, the
characteristic moment strengths ϕMnx0, ϕMny0, ϕMny90, and
ϕM n′ y90 corresponding to the factored design axial load Pu are
estimated from the interaction curves (refer to Fig. 11(b)).
3. The α is determined from Eq. (5) using n = Pu/(Agfc′).
Then, the biaxial design contour of the wall section is drawn
in accordance with Eq. (6a) and (6b) and Fig. 10.
4. Check whether the biaxial moments Mux and Muy are
within the design contour. If not, try Steps 1 through 4 again,
changing the reinforcement ratio ρ.
Design results of the biaxially-loaded T-shaped wall are
shown in Fig. 11 and Table 2. As a first trial, the reinforce-
ment ratio is taken as ρ = 0.008. The ϕPn-ϕMnx and ϕPn-ϕMny
interaction curves for θ = –90, 0, and 90 degrees are shown
in Fig. 11(b). The strength reduction factor ϕ and maximum
axial compression strength Pn,max are determined in accor-
dance with ACI 318-14. The characteristic moment strengths
ϕMnx0 (= 7571 kN∙m [5584 kip-ft]), ϕMny0 (= –1682 kN∙m
[–1241 kip-ft]), ϕMny90 (= 5606 kN∙m [4135 kip-ft]), and
ϕM n′ y90 (= –2181 kN∙m [–1609 kip-ft]) corresponding to
the factored design axial load Pu (= 2832 kN [637 kip]) are
then estimated, as shown in Fig. 11(b) and Table 2. Note
that the moment about the y-axis, Muy (= –2000 kN∙m
[–1475 kip-ft]), is less than ϕMny0 (= –1682 kN∙m [–1241
kip-ft]) and, thus, the neutral axis angle exists in the region
of –90 degrees ≤ θ ≤ 0 degrees. In addition, the α values
corresponding to the axial compression ratio n = Pu/(Agfc′) =
0.1 are determined as 2.0 and 1.45 in the regions of –90
degrees ≤ θ ≤ 0 degrees and 0 degrees ≤ θ ≤ +90 deg, respec-
tively. Figure 11(c) shows the mx-my contour for the biaxial
design of the T-shaped wall, plotted in the non-dimensional
plane. As shown in Fig. 11(c), mux = 0.726 and muy = 0.637,
159
Table 2—Summary of biaxial design of T-shaped wall
Biaxial strengths at θ = 0, 90, and –90
Design loads deg
Mux, Muy, ϕMnx0, ϕMny0, ϕMny90,
Pu, kN kN·m kN·m kN·m kN·m kN·m
2832 5500 –2000 7571 –1682 5606
Notes: 1 kip-ft = 1.36 kN-m; 1 kip = 4.45 kN
Fig. 11—Flexure-compression design of biaxially loaded T-shaped wall using proposed load contour method. (Note: 1 in. =
25.4 mm; 1 ksi = 6.9 MPa; 1 kip = 4.45 kN; 1 kip-ft = 1.36 kN∙m.)
represented as a black dot in the fourth quadrant, satisfy
Eq. (6b): mux2 + muy2 = 0.7262 + 0.6372 = 0.933 < 1.0.
In this study, a modified load contour method for biaxial
design of T-shaped walls is proposed. Because the elliptical
contour equations and exponent α are proposed from the
theoretical investigation for limited ranges of design param-
eters, the application of the proposed method needs to be
limited accordingly. Thus, the proposed method is appli-
cable only to T-shaped walls with monosymmetric sections
where the shape factor representing the ratio of half flange
length-to-web length should be between γ = 0.25 and 2.0.
In addition, the results of the parametric study show that,
at least within ρ = 0.005 to 0.02, fc′ = 30 to 40 MPa (4350
to 5800 psi), and fy = 400 to 600 MPa (58 to 87 ksi), the
proposed α and elliptical contour equations are conservative.
It is noted that the wall length-to-thickness ratio of the flange
and web varies between 2.0 to 9.75, as shown in Fig. 8(a).
Thus, conservatively, it is recommended that the length-to-
thickness ratio of both the flange and web in T-shaped walls
should be 3 or greater.
160
Non-dimensional load contour
ϕM n′ y90, α α
kN·m n (–90 deg ≤ θ ≤ 0 deg) (0 deg ≤ θ ≤ +90 deg) mux muy
–2181 0.1 2.0 1.35 0.726 0.637
SUMMARY AND CONCLUSIONS
In this study, the load contour method for biaxially loaded
T-shaped walls is developed. Through the theoretical para-
metric study using the strain-compatibility section analysis,
the characteristics of the biaxial interaction of T-shaped
walls have been investigated. Based on the results, non-
dimensional contour equations are proposed. In addition, the
wall design procedure using the proposed contour equations
is given. Conclusions are as follows:
1. In T-shaped walls, the Mnx-Mny contours at constant axial
compression loads are unsymmetrical in shape and signifi-
cantly affected by neutral axis angles (that is, directions of
loading) and axial load levels. The unsymmetrical contour
shape is attributed to the mono-symmetrical section geom-
etry made up of thin flange and web. Because the geometric
centroid does not coincide with the shear center, T-shaped
walls resist biaxial bending accompanied by torsion. Overall
the Mnx-Mny contour shape is close to a triangle at low axial
compression load levels (n = 0.0 to 0.2), but turns close to
ACI Structural Journal/January 2018
a rectangle or square at high axial compression load levels
(n = 0.3 to 0.5).
2. To eliminate effects of torsion, the Mnx-Mny contour of
T-shaped walls is converted into the non-dimensional mx-my
contour through a linear transformation. The mx-my contours
change in shape depending on the axial compression ratio
n and neutral axis angle θ. As the axial compression ratio
n increases, the mx-my contours expand in the region of
0  degrees ≤ θ ≤ +90 degrees. However, in the region of
–90  degrees ≤ θ ≤ 0 degrees, the correlation between the
axial compression ratio n and the mx-my contour shape is
relatively weak. Other design parameters such as the shape
factor γ (= ratio of flange length-to-web length), reinforce-
ment ratio and arrangement, concrete compressive strength,
and steel yield strength do not have significant impacts on
the mx-my contour shape.
3. Elliptical load contour equations are proposed based on
the parametric study on prototype T-shaped walls. Because
the section geometry and contour shape are unsymmetrical
about the y-axis, different equations are used depending on
the neutral axis angle θ or (the sign and magnitude of the
y-axis moment Mny). The parametric study shows that the
proposed contour equations are valid for n = 0.0 to 0.5, γ =
0.25 to 2.0, ρ = 0.005 to 0.02, fc′ = 30 to 40 MPa (4350 to
5800 psi), and fy = 400 to 600 MPa (58 to 87 ksi).
4. The elliptical contour equations can be used for prelim-
inary design of biaxially loaded T-shaped walls in selecting
a trial section, because the load contour method does not
require significant computational efforts nor dedicated soft-
ware. The advantage of the proposed design method is that
the biaxial design of nonplanar walls can be performed by
using only uniaxial P-M interaction curve about each axis
(that is, ϕPn-ϕMnx and/or ϕPn-ϕMny interaction curves for θ =
–90, 0, and 90 degrees).
AUTHOR BIOS
Tae-Sung Eom is an Associate Professor in the Department of Architec-
tural Engineering at Dankook University, Yongin, South Korea. He received
his BE, MS, and PhD in architectural engineering from Seoul National
University, Seoul, South Korea. His research interests include nonlinear
analysis and seismic design of reinforced concrete structures.
Hye-Sung Nam is a Graduate Student in the Department of Architectural
Engineering at Dankook University, where she received her BE in architec-
tural engineering.
Su-Min Kang is an Assistant Professor in the Department of Architectural
Engineering at Chungbuk National University, Cheongju, South Korea.
He received his BE, MS, and PhD in architectural engineering from Seoul
National University. His research interests include earthquake design of
reinforced concrete structures.
ACKNOWLEDGMENTS
This research was conducted with research funds from the Mid-
Career Research Program of National Research Foundation of Korea
(NRF-2015R1A2A2A01003777) and from the Infrastructure and Transporta-
tion Technology Promotion Research Program (Code 17CTAP-C129746-01),
funded by the Ministry of Land, Infrastructure, and Transportation of Korea.
REFERENCES
ACI Committee 318, 2011, “Building Code Requirements for Structural
Concrete (ACI 318-11) and Commentary,” American Concrete Institute,
Farmington Hills, MI, 503 pp.
ACI Structural Journal/January 2018
Ahmad, S. H., and Weekoon, S. L., 1995, “Model for Behavior of
Slender Reinforced Concrete Columns Under Biaxial Bending,” ACI Struc-
tural Journal, V. 92, No. 2, Mar.-Apr., pp. 188-198.
Bresler, B., 1960, “Design Criteria for Reinforced Concrete Columns
under Axial Load and Biaxial Bending,” ACI Journal Proceedings, V. 57,
No. 5, May, pp. 481-490.
Cedolin, L.; Cusatis, G.; Eccheli, S.; and Roveda, M., 2008, “Capacity of
Rectangular Cross Sections under Biaxially Eccentric Loads,” ACI Struc-
tural Journal, V. 105, No. 2, Mar.-Apr., pp. 215-224.
Chen, S. F.; Teng, J. G.; and Chan, S. L., 2001, “Design of Biaxially
Loaded Short Composite Columns of Arbitrary Section,” Journal of
Structural Engineering, ASCE, V. 127, No. 6, pp. 678-685. doi: 10.1061/
(ASCE)0733-9445(2001)127:6(678)
Chiorean, C. G., 2010, “Computerised Interaction Diagrams and
Moment Capacity Contours for Composite Steel-Concrete Cross-Sections,”
Engineering Structures, V. 32, No. 11, pp. 3734-3757. doi: 10.1016/j.
engstruct.2010.08.019
de Sousa, J. B. Jr., and Caldas, R. B., 2005, “Numerical Analysis of
Composite Steel-Concrete Columns of Arbitrary Cross Section,” Journal of
Structural Engineering, ASCE, V. 131, No. 11, pp. 1721-1730. doi: 10.1061/
(ASCE)0733-9445(2005)131:11(1721)
El Fattah, A. M.; Rasheed, H.; and Esmaeily, A., 2013, “LRFD Soft-
ware for Design and Actual Ultimate Capacity of Confined Rectangular
Columns,” Report No. K-TRAN: KSU-11-3, Kansas Department of Trans-
portation, Topeka, KS, 220 pp.
Fafitis, A., 2001, “Interaction Surfaces of Reinforced-Concrete Sections
in Biaxial Bending,” Journal of Structural Engineering, ASCE, V. 127,
No. 7, pp. 840-846. doi: 10.1061/(ASCE)0733-9445(2001)127:7(840)
Furlong, R. W.; Hsu, C.-T. T.; and Mirza, S. A., 2004, “Analysis and
Design of Concrete Columns for Biaxial Bending—Overview,” ACI Struc-
tural Journal, V. 101, No. 3, May-June, pp. 413-423.
Hsu, C.-T. T., 1985, “Biaxially Loaded L-Shaped Reinforced Concrete
Columns,” Journal of Structural Engineering, ASCE, V. 111, No. 12,
pp. 2576-2595. doi: (errata, 1988, V. 114 No. 11, p. 2629)10.1061/
(ASCE)0733-9445(1985)111:12(2576)
Hsu, C.-T. T., 1987, “Channel-Shaped Reinforced Concrete Compres-
sion Members under Biaxial Bending,” ACI Structural Journal, V. 84, No. 3,
May-June, pp. 201-211.
Hsu, C.-T. T., 1989, “T-Shaped Reinforced Concrete Members under
Biaxial Bending and Axial Compression,” ACI Structural Journal, V. 86,
No. 4, July-Aug., pp. 460-468.
Lau, C. Y.; Chan, S. L.; and So, A. K. W., 1993, “Biaxial Bending Design
of Arbitrarily Shaped Reinforced Concrete Column,” ACI Structural
Journal, V. 90, No. 3, May-June, pp. 269-278.
Ludovico, M. D.; Lignola, G. P.; Prota, A.; and Cosenza, E., 2010,
“Nonlinear Analysis of Cross Sections under Axial Load and Biaxial
Bending,” ACI Structural Journal, V. 107, No. 4, July-Aug., pp. 390-399.
MacGregor, J. G., 1973, “Simple Design Procedures for Concrete
Columns,” Introductory Report, Symposium on Design and Safety of Rein-
forced Concrete Compression Members, Reports of the Working Commis-
sions, V. 15, International Association of Bridge and Structural Engineering,
Zurich, Switzerland, Apr., pp. 23-49.
Marmo, F.; Serpieri, R.; and Rosati, L., 2011, “Ultimate Strength
Analysis of Prestressed Reinforced Concrete Sections under Axial Force
and Biaxial Bending,” Computers & Structures, V. 89, No. 1-2, pp. 91-108.
doi: 10.1016/j.compstruc.2010.08.005
Papanikolaou, V. K., 2012, “Analysis of Arbitrary Composite Sections
in Biaxial Bending and Axial Load,” Computers & Structures, V. 98-99,
pp. 33-54. doi: 10.1016/j.compstruc.2012.02.004
Parme, A. L.; Nieves, J. M.; and Gouwens, A., 1966, “Capacity of Rein-
forced Rectangular Columns Subjected to Biaxial Bending,” ACI Journal
Proceedings, V. 63, No. 9, Sept., pp. 911-923.
Portland Cement Association, 1966, “Capacity of Reinforced Rectan-
gular Columns Subject to Biaxial Bending,” Publication EB011D, Portland
Cement Association, Skokie, IL.
Portland Cement Association, 2013, “Notes on ACI 318-11 Building
Code Requirements for Structural Concrete with Design Applications,”
Publication EB716, Portland Cement Association, Skokie, IL.
Rodriguez-Gutierrez, J. A., and Aristizabal-Ochoa, 1999a, “Biaxial
Interaction Diagrams for Short RC Columns of Any Cross Section,”
Journal of Structural Engineering, ASCE, V. 125, No. 6, pp. 672-683. doi:
10.1061/(ASCE)0733-9445(1999)125:6(672)
Rodriguez-Gutierrez, J. A., and Aristizabal-Ochoa, J. D., 1999b,
“Partially and Fully Prestressed Concrete Sections under Biaxial Bending
and Axial Load,” ACI Structural Journal, V. 97, No. 4, Mar.-Apr.,
pp. 553-563.
Wight, J. K., and MacGregor, J. G., 2012, Reinforced Concrete:
Mechanics and Design, sixth edition, Pearson Education, Inc., Upper
Saddle River, NJ.
161
APPLY FOR
ACI Foundation
Research Grants
The ACI Foundation annually funds research projects with grants
up to $50,000. The Foundation seeks to advance the concrete
industry through the funding of concrete research projects that
further the knowledge and sustainability of concrete materials,
construction, and structures.

• Topics are encouraged from all areas of concrete research;

• A letter of support of the research concept by an ACI Technical
Committee is required;
• Industry partnering and project cost sharing are encouraged;
• Principal investigators must follow the ACI Foundation’s
published Concrete Research Council Grant Proposal Guide.

The Foundation will begin accepting proposals at the end of

August through December 1. Applications are submitted online
at concreteresearchcouncil.org.
 ACI STRUCTURAL JOURNAL TECHNICAL PAPER
Title No. 115-S13

Eccentric Punching Shear of Waffle Slab

by Ahmed Faleh Al-Bayati, Lau Teck Leong, and Leslie A. Clark

This paper presents test results of six 1/10th-scale micro-concrete

waffle slabs subjected to internal punching shear in the presence
of moment transfer. Although the observed punching failure mech-
anism of a waffle slab was found to be very similar to that of a
flat slab, the shear capacity is relatively reduced because some of
the potential failure surface is lost when it extends into the waffle
section. Because the current codes of practice do not consider the
punching shear mechanism of waffle slabs, the comparisons with
the test results revealed that BS8110 and EC2 overestimated the
punching capacities of waffle slabs, whereas ACI’s prediction was
conservative. A prediction model based on the upper-bound plas-
ticity theory was proposed to predict the failure load, which gave
good agreement with the tests.

Keywords: design; internal punching shear; moment transfer; plasticity;

reinforced concrete; waffle slab.

INTRODUCTION
Waffle slabs have been widely used owing to their economic
benefits. It consists of a grid of ribs, distributed in orthogonal
directions, regularly spaced, and topped by a thin slab (refer
to Fig. 1). The overall slab is then supported by columns at
various positions, at which solid sections are introduced to
allow for load transfer from the slab to the columns.
Similar to a flat slab, a waffle slab can develop a local shear
failure known as “punching shear failure” (refer to Fig. 2).
At failure, a solid revolution of concrete (“I”) surrounded by
the inclined shear cracks separates normally from the slab, Fig. 1—Waffle slab.
leaving the rest of the slab (“II”) remaining uncracked.1
flat slabs. So far, there has been a very limited amount of
However, despite of being widely used, a small amount of
work conducted to investigate the punching mechanism
research2-7 has been conducted to investigate the punching
of waffle slabs. As a result, it is not clear how to apply the
behavior of waffle slabs. As a result, the shear design proce-
current design methods that are derived from tests on flat
dures of waffle slabs are not included in the current design
slab to design against the punching failure of waffle slabs.
codes.8-10 Therefore, it is not clear how to apply (if neces-
Therefore, there is a need to conduct experimental inves-
sary) the codes’ design clauses for flat slabs to waffle slabs
tigation to aid the understanding of the punching failure
because when the solid section are very wide or top slabs
mechanism of waffle slabs and hence to develop a relevant
are sufficiently thick, the punching failure surface could
prediction model.
form within the solid section (refer to Fig. 2(b) and (c)),
but when the solid section is narrower, the punching failure
EXPERIMENTAL INVESTIGATIONS
surface could pass through the reduced depth section (refer
Specimen’s details
to Fig. 2(a)).11 As a result, a smaller shear failure surface
Tests were performed on six (1/10th-scaled) micro-
could be mobilized, which would consequently lead to a
concrete waffle slab specimens to simulate the internal
lower punching shear capacity.
column-waffle slab connections in the presence of moment
A prediction model based on the upper-bound solution
transfer. Variables considered were the size of solid section
of plasticity theory was proposed to predict the eccentric
and the column’s eccentricity (refer to Table 1).
punching shear of the waffle slab. The proposed model
develops the concentric punching shear model11 to predict
the eccentric punching shear of waffle slabs. ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2016-434.R1, doi: 10.14359/51701090, was received January 19, 2017,
and reviewed under Institute publication policies. Copyright © 2018, American
RESEARCH SIGNIFICANCE Concrete Institute. All rights reserved, including the making of copies unless
Despite the popularity of waffle slabs, research to date permission is obtained from the copyright proprietors. Pertinent discussion including
author’s closure, if any, will be published ten months from this journal’s date if the
has focused on the punching shear failure mechanism of discussion is received within four months of the paper’s print publication.

ACI Structural Journal/January 2018 163

 The ratios of column eccentricity to column size, e/cx,
considered were 0.5, 1.0, and 1.5. All slab specimens had
an overall depth of 70 mm (2.8 in.) and a top slab of 20 mm
(0.8 in.) thick (refer to Fig. 3). These geometrical proper-
ties were chosen in accordance with Clause 5.3 in EC2.9 All
specimens were cast in an upright position to simulate the
casting position of the prototype structures. The moment
transferring column(s) were simulated with a steel column
stub with overhang bolted to the specimens (refer to Fig. 4).
Although such an arrangement is different from the actual
details of a column slab connection, it is believed that as long
as the punching shear failure occurs outside the column, the
punching resistance would not be affected.12

Concrete
All specimens were cast from micro-concrete having
a maximum aggregate size of 2 mm (0.08 in.), water-
cement ratio (w/c) of 0.56, and aggregate/cement ratio of
1.6. All the aggregates used were sieved and remixed to the
required grading13 (refer to Table 2). The concrete compres-
sion strengths were obtained from 50 mm (2 in.) cubes cast
simultaneously with slab specimens. However, for calculating
the shear capacities, the measured 50 mm (2 in.) cube concrete
compressive strengths have been modified to represent the
intended 15 mm (0.6 in.) cube compressive strength in accor-
dance to Endersbee’s equation14 to allow for size effects.

−0.106
 15 
f cu15 = f cu 50   (1)
 50 

where both fcu15 and fcu50 are in MPa.

Reinforcements
Due to the difficulties in obtaining small-diameter ribbed
steel bars, reinforcements were fabricated from plain steel
bars of 3.4 mm (0.13 in.) diameter, with an average yield
strength of 440 MPa (63.8 ksi) (refer to Fig. 5). All slab
specimens tested in this research have been reinforced with
a sufficient amount of steel bars with tension bars bent up
at both ends to avoid premature bond failure. Confidence
in the use of small-scaled specimens cast with plain scaled
steel bars further derived from the test results (Fig. 6),
indicating all slab specimens failed in punching with the
manner of their load-deflection curves similar to flat slabs
cast with high-tension reinforcement ratio reported in the
literatures.15 Tension reinforcements were placed at 11 mm
(0.4 in.) spacing across the top slab regions, while all ribs
were doubly reinforced. And 5 mm (0.2 in.) covers to rein-
forcements were maintained in all specimens.

Fig. 2—Punching shear mechanism of waffle slab.

Table 1—Details of waffle slab specimens

Size of solid section, c, mm d, mm d1, mm d2, mm fcu50, N/mm2 Column eccentricity e,
Slab No. mm (in.) (in.) (in.) (in.) (in.) ρ, % ρct, % ρcb, % (psi) mm (in.)
IWSM1 290 x 290 (11.4 x 11.4) 100 (4) 62 (2.44) 63 (2.48) 63 (2.48) 1.29 1.27 0.6 29.5 (4278) 100 (4)
IWSM2 470 x 470 (18.5 x 18.5) 100 (4) 62 (2.44) 63 (2.48) 63 (2.48) 1.29 1.27 0.6 34.7 (5033) 100 (4)
IWSM3 200 x 200 (7.9 x 7.9) 100 (4) 62 (2.44) 63 (2.48) 63 (2.48) 1.29 1.27 0.6 34.6 (5018) 100 (4)
IWSM4 250 x 250 (9.8 x 9.8) 100 (4) 62 (2.44) 63 (2.48) 63 (2.48) 1.29 1.27 0.6 35.6 (5163) 100 (4)
IWSM5 250 x 250 (9.8 x 9.8) 100 (4) 62 (2.44) 63 (2.48) 63 (2.48) 1.29 1.27 0.6 32.5 (4714) 50 (2)
IWSM6 250 x 250 (9.8 x 9.8) 100 (4) 62 (2.44) 63 (2.48) 63 (2.48) 1.29 1.27 0.6 31.7 (4598) 150 (6)
Notes: d is average effective depth of orthogonal tension reinforcement; d1 is effective depth of tension reinforcement in bending direction; d2 is effective depth of compression
reinforcement

164 ACI Structural Journal/January 2018

Fig. 3—Waffle slab specimens.

Fig. 4—Test setup.

Test procedures Table 2—Sieve analysis of aggregates10
All specimens were supported on rollers at four edges in Sieve size Mass retained, g % retained % passing
an inverse manner and loaded eccentrically by means of a
2.0 0.00 0.00 100.00
hand-operated hydraulic jack in increments of 4 kN (0.9 kip)
onto the steel column stub until failure (refer to Fig. 4). 1.18 209.50 42.19 57.81
Central deflections were recorded and cracks were marked, 0.6 57.00 11.48 46.33
incrementally, using two digital gauges positioned adjacent 0.3 21.50 4.33 42.00
to the column.
0.15 168.80 33.39 8.61
≤0.15 42.80 8.62 0.00

ACI Structural Journal/January 2018 165

Fig. 5—Flexural reinforcements.
Fig. 6—Typical load-deflection curves of Specimen IWS M2.
APPLICABILITY OF SMALL-SCALED SPECIMENS
In view of the fact that the slab specimens described are
1/10th-scaled specimens cast from micro-concrete having
a maximum aggregate size of 2 mm (0.08 in.), and rein-
forced with plain steel bars with small diameter 3.4 mm
(0.13 in.). A discussion with regards to the size effects on
concrete strength in shear and in compression, the behavior
of micro-concrete in punching shear, and the bond strength
of small steel bars has been presented in an earlier paper.11
TEST RESULTS
Failure modes
All slabs failed suddenly by punching shear in a mode
very similar to the mechanism observed in flat slabs.15-18 The
failure surface was characterized by shear cracks inclined at
approximately 22 degrees and intersected with the top slab at
a distance of approximately 2.5 times the overall depth of slab
from the column face(s) comparable to that observed in flat
slabs.19 For slabs loaded with shear span less than 2.5 times
the overall slab depth, the failure surface was observed to
propagate from the column faces to the supports as expected.
The shear cracks were limited to the front (heavily loaded)
region and two side regions of the column, while no cracks
were observed in the rear region (refer to Fig. 7). However,
unlike a flat slab, when the width of solid section is less than
five times the total depth of the slab, the failure surface was
an incomplete failure surface because some of the potential
failure surface was lost due to the presence of waffle sections
(refer to Figure 7(I)).
166
Flexural cracks were observed on the tension surface of
the specimens at approximately 43% of the failure loads for
ISW M6, 50% of the failure loads for ISW M1 to M4, and
56% of the failure loads for ISW M5, as the eccentricity
decreased from 150 mm (5.9 in.) to 100 mm (3.9 in.) to
50 mm (2 in.), respectively. This is to be expected because
tensile stress would be increased as the column eccentric-
ities increased. Similar to flat slabs,1,20 as the applied load
continued to increase, torsional cracks on the tension surface
were observed to develop from the column side faces at an
angle of approximately 45 degrees.
Failure loads
The punching failure loads from tests, Ptest, are presented
in Table 3. It is important to note that the self-weight of the
slab specimens has not been included due to the small size
of slab specimens (limited to the points of contraflexure).
The shear capacity of a waffle slab was observed to increase
as the size of the solid section increased. In comparison
with Specimen ISW M3, cast with a 200 mm (7.9 in.) solid
section, Specimens ISW M1 and M2 provided increments
in punching strength of 18% and 23% as the solid section
increased to 290 and 470 mm (11.4 and 18.5 in.), respec-
tively. This is to be expected because of the increase in the
revolution failure surface that has formed and mobilized
within the solid sections.
The shear capacity of the waffle slab(s) was observed to
decrease as the column eccentricity increased. In comparison
with Specimen IWS M5, loaded with 50 mm (2 in.) eccen-
tricity, Specimens IWS M4 and M6 exhibited reductions in
punching strength of 17% and 42% as the applied eccentric-
ities increased to 100 and 150 mm (3.9 and 5.9 in.), respec-
tively. It is believed that the observed reductions in shear
capacities were attributed to the increase in the moment trans-
ferred. In addition, when the applied moment exceeded the
connection moment resistance, it transformed into torsion
stresses on both sides of the column. As a result, nonuni-
formly distributed shear stresses occurred at the perimeter of
the column, thus leading to a lower punching shear capacity
to be mobilized.
Deflection
The defection of each specimen was measured using two
digital gauges placed at the front and the back faces of the
column after each load increment. A typical set of load-
ACI Structural Journal/January 2018
Fig. 7—Failure mechanism of waffle slab specimens.
Table 3—Comparisons between predicted and observed shear strengths
Slab Mtest*, kN.m (k.ft) Ptest†, kN (kip) Ppred, kN (kip) PACI, kN (kip) PEC2, kN (kip) PBS, kN (kip) PPred/Ptest PACI/Ptest PEC2/Ptest PBS/Ptest
IWSM1 4.21 (3.1) 42.1 (9.5) 33.5 (8) 27.5 (6.2) 46 (10.3) 58.0 (13) 0.89 0.65 1.09 1.38
IWSM2 4.42 (3.3) 44.2 (9.9) 47.0 (10.6) 30.5 (6.9) 48.6 (10.9) 61.3 (14.4) 1.06 0.69 1.10 1.39
IWSM3 3.58 (2.6) 35.8 (8.1) 36.8 (8.3) 34.7 (7.8) 51 (11.5) 63.9 (14.4) 1.03 0.97 1.42 1.79
IWSM4 4.21 (3.1) 42.1 (9.5) 39.5 (8.9) 32.3 (7.3) 49.7 (11.2) 62.6 (14.1) 0.94 0.77 1.18 1.49
IWSM5 2.53 (1.9) 50.5 (11.4) 44.7 (10) 38.3 (8.6) 53.1 (11.9) 66.2 (59.1) 0.92 0.76 1.05 1.31
IWSM6 4.42 (3.3) 29.47 (6.6) 37.2 (8.4) 28.2 (6.3) 46.8 (10.5) 59.1 (13.3) 1.26 0.96 1.59 2.00
Mean 1.02 0.80 1.24 1.47
Standard deviation 0.14 0.13 0.22 0.19
*
Mtest is applied moment.
†
Ptest is applied vertical load.

deflection curves recorded from the test of Specimen (IWS
M2) is presented in Fig. 6. Full deflection data are reported
in Reference 21. The load-deflection curve can be consid-
ered as two straight lines inclined at two different angles.22
The first slope corresponds to the stiffness of an uncracked
section, while the second slope corresponds to the stiffness
of a cracked section.22 Such a phenomenon is consistent
with flat slabs failed by eccentric punching shear.15,17,18 It is
believed that such a phenomenon is due to the fact that the
steel reinforcements become fully mobilized and effective
on nominal slip.
PROPOSED MODEL
General
The proposed model is based on the upper-bound plasticity
theory in predicting the punching shear strength of concrete
slabs. The failure surface of revolution is predicted to be a
catenary, possibly jointed by a straight line inclined at the
concrete angle of friction to the plane of slab.23 The punching
shear strength is obtained by equating the work done by the
applied load to the energy dissipated in the failure surface.
However, due to its complexity, it is not possible to obtain a
closed form solution. In view of this, Salim and Sebastian24
have developed a simplified approach based on a straight
line failure surface that represents the complex method
adequately but with a slightly greater scatter in the ratio of
test to predicted values. The simplified approach is adopted
in the current paper.
The proposed model (refer to Fig. 8) considers that any
change in slab thickness entered into the failure surface
incurs a reduction in the shear area and, consequently, a
reduction in the energy dissipation and hence the ultimate
punching capacity. Thus, the reduction in punching capacity
depends on the loss of shear area, which in turn depends on
the size of solid section. If the latter is wide enough to accom-
modate the punching revolution, then no reduction would be
required. The implication is that the model could predict the
punching shear resistance of a flat slab by assuming either

ACI Structural Journal/January 2018 167

Fig. 8—Proposed eccentric punching shear failure surface.
Fig. 9—Eccentric punching failure surface of waffle slab.
the size of solid section is sufficiently large or the top slab is
sufficiently thick. As shown in Fig. 8 and Appendix A,* the
punching shear strength of the waffle slab is computed based
on the following geometrical categories:
1. Failure surface extends into the waffle section causing a
reduction in the shear capacity;
2. Failure surface extends into the waffle section but not
resulting in loss in the shear capacity. This mechanism is
derived from the tests of Lau and Clark on ribbed slabs,25
and shown herein for illustration as a possible failure mech-
anism; and
3. The solid section is sufficiently wide, resulting in no
loss in the shear capacity.
By categorizing the shear failure surface, the potential
punching shear capacity, Ppot, can be computed from the
following equation
Pexδ = Pinδ = Ppot = αwi Ac (2)
where Pex is the external work done by the applied load in
kN; δ is the relative displacement field between two rigid
zones; α is a shear retention factor; wi is internal work dissi-
*
The Appendix is available at www.concrete.org/publications in PDF format,
appended to the online version of the published paper. It is also available in hard copy
from ACI headquarters for a fee equal to the cost of reproduction plus handling at the
time of the request.
168
pated on the failure surface23 in MPa; and Ac is the punching
failure surface in mm.
The details of calculation for shear failure surface areas and
summation of energy dissipated to determine the punching
capacities are given in Appendix A. The failure surface of
revolution was found to be dependent on the column eccen-
tricity e—that is, it was observed to be restricted to the front
(heavily loaded) region and the adjacent side regions of the
column when the eccentricity was 1.0 and 1.5 times the
column size cx (refer to Fig. 9 and 10). When the eccentricity
was 0.5 times the column size cx, the failure surface was
observed to extend to the rear (lightly loaded) region of the
column (refer to Fig. 9(a)). As such, it can be concluded that
the cracks formation is inversely proportional to the eccen-
tricities (refer to Fig. 11). These observed mechanisms are
replicated in the proposed punching failure surface Ac in the
form of an effectiveness factor ξ that allows for the reduction
on effective shear area in the rear (lightly loaded) regions.
The effective punching shear failure surface Ac is expressed
as follows
 π π 
Ac = h 1 + cot 2 θ  2cx + c y + h cot θ + ξ h cot θ (3)
 2 2 
where h is the total depth of slab; θ is the inclination angle
of the failure surface; cx is the column size parallel to the
ACI Structural Journal/January 2018
Fig. 10—Internal shear cracks of Specimens IWS M4, M5, and M6.
Fig. 11—Schematic sketches of observed eccentric punching shear mechanism.
direction of the column eccentricity; cy is the column size
perpendicular to the column eccentricity; ξ is the effective
shear area factor ξ = 1– e/cx, e/cx ≤ 1; e is the column’s
eccentricity, e = (Mtest/Ptest); Mtest is the transferred moment;
and Ptest is the vertical load.
Effectiveness factor
In applying plastic theory to concrete, it is necessary to
introduce an effectiveness factor vc to the concrete compres-
ACI Structural Journal/January 2018
sive strength to allow for the fact that concrete is not a
perfectly rigid plastic material.26,27 Although the effective-
ness factor has been proposed to be a function of concrete
compressive strength by Braestrup et al.,23 Salim and Sebas-
tian24 found that good correlation with the test results was
only achieved when Sigurdsson’s effectiveness factor28 was
used in their proposed models. As proposed earlier,11 the
effectiveness factor is to be a function of concrete strength,
flexural reinforcement ratio, and total depth of the slab
169
 of the 10 and 20 mm (0.4 and 0.8 in.) aggregates specimens
remained at their peak values, as indicated in Fig. 12. In view
of the fact that the tests described in the present paper were
carried out on micro-concrete slabs with a maximum aggre-
gate size of 2 mm (0.08 in.), a shear retention factor that
is dependent on the maximum aggregate size was adopted.
Values of 1.0 and 0.7 were adopted for normal concrete and
micro-concrete, respectively.

Moment transfer mechanism

Fig. 12—Shear stress-displacement curve.29,30
In addition to the observed mechanism on the formation
of the shear failure surface, the torsion factor is another
important factor to be considered in predicting the eccen-
tric punching capacity. This is because when the applied
moment Mtest exceeds the flexural capacity Mf, the excessive
moment, (Mtest – Mf), is assumed to be redistributed within
the column side strips as torsion, T (refer to Fig. 13); subse-
quently, an additional force ΔP is induced at the inner face of
the column. The total downwards applied force is therefore
equal to (PPot + ΔP), and consequently, the punching failure
is assumed to take place when the total applied force is equal
to the punching shear resistance of the slab. The implication
is that when the moment transfer exceeds the slab’s flexural
capacity, the punching failure would occur at a lower failure
load. This effect is accounted for in the proposed model
by the torsion factor β, which can be calculated using the
following equation

PPot
β= ; β ≤ 1 (6)
( PPot + ∆P )
where PPot is the potential punching shear strength to be
computed from Eq. (2), and ΔP is the additional downward
force induced by torsion.
The additional torsion shear stresses, τt, and the additional
induced downward force, ΔP, can be derived using the sand
heap analogy32 as

Fig. 13—Additional force ΔP induced by torsion stresses. 2T

τt = (7)
 h 
0.9  0.5  hmin 2  hmax − min 
vc =  0.7 +  (1 + 0.14ρ) (4)  3 
f cc h
Hence
where fcc is the cylindrical compressive strength of concrete
T
in MPa; h is the total depth of slab in meters; and ρ is the ∆P = if cx ≥ h (8)
average flexural reinforcement ratio. h
cx −
3
Shear retention factor α
In the current paper, a shear retention factor α was adopted  c 
T h − x 
from the shear tests of Boswell and Wong29 and Wong30 that  2
∆P = if cx < h (9)
were conducted on Mattock-type pushoff specimens31 cast  c 
cx  h − x 
with different maximum aggregate size of 2, 10, and 20 mm  3
(0.08, 0.4, and 0.8 in.). The results indicate that, although
the peak strength is essentially independent of the aggregate Finally, the eccentric punching shear capacity can be
size, the residual strength was dependent on the aggregate computed from the following equation.
size. With increasing shear displacements, the shear resis-
tance of 2 mm (0.08 in.) aggregate specimen reduces to PPred = βPpot (10)
approximately 70% of its peak value, while the resistance

170 ACI Structural Journal/January 2018

Flexural capacities Mf
In deriving the flexural resistance, Mf, reference is made
to the test observations of previous research,15,17,33,34 where
strains of the flexural reinforcements (parallel to the applied
moment) were measured to indicate the effective width of
the slab in resisting the applied moment. It was noted that
the tensile stress in the reinforcements decreased linearly up
to a distance equal to the effective depth, d, of the slab from
both sides of the column. From these observations, the effec-
tive breadth (cy + d) is therefore proposed.33,35,36 In view of
the fact that the slab specimens tested in the current research
consisted of moment transfer similar to those reported, and
hence, the proposed effective breadth (cy + d) is adopted.
The flexural resistance is based on the plasticity equa-
tions of pure bending derived for rectangular beams.27 The
total flexural resistance of the connection, Mf, is the sum of
the flexural resistances developed at front and back of the
column Mf1, Mf2
Mf = Mf1 + Mf2 (11)
Comparisons with test results
The proposing model develops the previous model11 to
predict the eccentric punching shear strength. The model
was first used to predict the failure load of 40 flat slabs cast
from normal concrete and subjected to eccentric punching
shear.15-18 The mean ratio of the estimated strengths to tests
was 0.94 with a standard deviation of 0.15 (refer to Table 4).
The confidence in adopting the shear retention factor of 0.7
was derived from the recent works of Lau and Clark.25 Hence
the value of 0.7 was adopted in the model when comparing
with the current data. Good agreements were achieved
between the estimated strengths Ppred and the test results
(refer to Table 3). The mean ratio of predicted strength to
test was 1.02, with a standard deviation of 0.14.
EVALUATION OF CURRENT CODES APPROACH
The current codes of practice such as ACI 318-14,8 EC2,9
and BS811010 have adopted the control surface approach for
design of flat slabs against punching shear mechanism. The
control failure surface is essentially a virtual vertical shear
surface at an assumed distance from the column faces. In the
presence of moment transfer, the shear stresses induced from
the vertical load combine with those from the transferring
moment. The ultimate punching resistance is therefore the
sum of all shear stresses on the control surface.
In Table 3, comparisons with test results were carried out
to evaluate the applicability of design codes8-10 on eccen-
tric punching mechanism of waffle slabs. These predictions
were calculated with the partial safety factor for the mate-
rial being set to unity and the shear retention factor α taken
as 0.7.
In comparisons with both EC29 and BS8110,10 ACI8
provided conservative predictions toward the observed
punching failure loads on waffle slabs despite the fact that its
critical section does not resemble the punching revolution,
and also does not consider the loss of failure surface on the
punching revolution. However, the conservativeness derived
from ACI8 is not limited to punching mechanism on waffle
ACI Structural Journal/January 2018
Table 4—Comparison between predicted eccentric
punching shear strength and test results of flat slabs
PPred/Ptest
Mean/standard deviation
Total number
Researcher of slabs Proposed model
Ghali et al.21
3 1.01/0.17
Regan et al.22 16 0.92/0.10
Hawkins et al.23 25 0.96/0.17
Ozden et al. 24
6 0.83/0.08
Total 50 0.94/0.15
slabs, as similar observations were made by others35-37 when
predictions derived from ACI8 were compared with flat slabs.
The current EC29 overestimated the punching capac-
ities of waffle slab specimens. This may be attributed to
changes in waffle slab thickness and changes in the column
eccentricity. Comparisons with specimens having a small
solid section (ISW M3) and larger solid section (ISW M2)
revealed that EC29 overestimated the punching capacity (up
to 42%) of that having smaller solid section and overesti-
mated the punching capacity (up to 10%) of those having
a larger solid section. In comparison with Specimen IWS
M5 (loaded with 50 mm column eccentricity), EC29 overes-
timated the punching shear strength of Specimens IWS M4
and M6 by 18% and 59% as the eccentricities increased to
100 and 150 mm (3.9 and 5.9 in.), respectively.
Similar overestimations were observed from the use of
the current BS811010 in predicting the punching capaci-
ties of waffle slabs. This may also be attributed to the two
aforementioned reasons. Comparisons between specimens
having a small solid section (ISW M3) and larger solid
section (ISW M2) revealed that BS811010 overestimated the
punching capacity (up to 79%) of those having a smaller
solid section and overestimated the punching capacity (up
to 39%) of those having a larger solid section. Comparisons
between specimens loaded with a small column eccentricity
(IWS M5) and specimen loaded with larger column eccen-
tricity (IWS M6) revealed that BS811010 overestimated the
punching capacities from 31% to 100% as the eccentricities
increased from 50 to 100 mm (2 to 3.9 in.), respectively.
CONCLUSIONS
1. The punching failure mechanism of a waffle slab was
found to be similar in nature to that of a flat slab. However,
for a slab with a width of solid section less than five times
the overall depth of slab, the observed shear failure surface
was incomplete due to the losses in the failure surface when
it extended into the waffle section.
2. In the presence of moment transfer, the punching failure
surface was observed to concentrate in the front heavily
loaded region and the side regions of the loading column.
In addition, the punching shear capacity of waffle slabs was
observed to decrease as the column eccentricity increased.
3. The proposed upper-bound plastic model based on Niel-
sen-Braestrup’s shear analysis, coupled with the simplified
failure mechanism, has shown good agreement with the
171
test results of flat slabs reported in the literature and the test
results of the waffle slabs presented in the current work.
4. Because the punching shear design procedure in the
current code of practice does not consider the punching shear
mechanism of waffle slabs, comparisons with the test results
revealed that both the BS8110 and EC2 overestimated the
punching failure loads, while ACI remained conservative.
AUTHOR BIOS
Ahmed Faleh Al-Bayati is a Lecturer at Al-Nahrain University, Baghdad,
Iraq. He received his PhD from the University of Nottingham Malaysia
Campus, Semenyih, Malaysia. His research interests include the design
of reinforced concrete structures and the punching shear strength of slab-
column connections.
Lau Teck Leong is an Associate Professor at the University of Nottingham
Malaysia Campus. He received his BEng and PhD from the University of
Birmingham, Birmingham, UK. His research interests include the design
of reinforced concrete structures and the punching shear strength of slab-
column connections.
Leslie A. Clark is an Emeritus Professor of structural engineering at the
University of Birmingham. His research interests include the design, anal-
ysis, assessment, and behavior of concrete structures.
ACKNOWLEDGMENTS
The authors would like to thank the University of Nottingham Malaysia
Campus for providing the financial support to this project.
NOTATION
Ac = assumed failure surface
As1&2 = top and bottom flexural reinforcement areas passing through
effective bending width
ax = ay shear spans in x- and y-axis
b = effective bending width, b = cy + d
brib = rib width
cx, cy = size of column parallel to x- and y-axis, respectively
Dx, Dy = sides of solid section parallel to x- and y-axis, respectively
d = effective depth of slab Code of Practice for Design and Construction, (BS 8110: Part 1: 1997),”
e = column’s eccentricity, e = (Mtest/Ptest)
fc = effective concrete compressive strength in plasticity analysis,
fc = vcfcc
fcb = effective compressive strength in bending, fcb = υbfcc
fcc = cylinder compressive strength of concrete
fcu15 = concrete compressive strength measured from 15 mm cube
fcu50 = concrete compressive strength measured from 50 mm cube
ft = effective concrete tension strength in plasticity analysis, ft/fc =
1/400
h1 = thickness of topping slab
h2 = total depth of slab
k = concrete parameter, where Φ = 37 degrees
Mf = flexural capacity of connection Mf = Mf1 + Mf2
Mf1 = flexural capacity at column front face,
 1   1 
2
M f 1 = 1 − ϕ As1 f y d1 = 1 − ϕ ϕb1d1 f cb
 2   2 
Mf2 = flexural capacity at column back
 1   1 
M f 2 = 1 − ϕ As 2 f y d 2 = 1 − ϕ ϕb2 d 2 2 f cb
 2   2 
Mtest = transferred moment
PACI = predicted failure load by ACI
PBS = predicted failure load by BS8110
PEC2 = predicted failure load d by EC2
PPot = potential punching shear strength
PPred = predicted failure load
Ptest = test failure load
T = torsion moment
vc = shear effectiveness factor
1 tion of Slab-Column Connections,” Canadian Journal of Civil Engineering,
wi = internal work dissipated wi = δf c ( λ − m cos θ )
α = shear retention factor 2
β = torsion factor
ΔP = additional downward force induced by torsion
δ = displacement field in plasticity analysis
172
Φ = internal friction angle of concrete = Φ ≅ 37 degrees
φ = mechanical degree of steel reinforcement, φ = (ρc/fcb)
λ = material properties, λ = 1 – (k – 1)ft/fc
μ = material properties, μ = 1 – (k + 1)ft/fc
θ = angle of failure surface
ρ = average flexural reinforcement of solid section
ρc = flexural reinforcement ratio passing through effective bending
width
τt = torsion stress
υb = effectiveness factor of concrete in bending, MPa
ξ = effective shear area factor, ξ = 1– e/cx, e/cx ≤ 1
REFERENCES
1. Regan, P. E., and Braestrup, M. W., “Punching Shear in Reinforced
Concrete: A State-of-the-Art Report,” Bulletin d’Information No. 168,
Comité Euro-International du Béton, Lausanne, Switzerland, 1985, 232 pp.
2. Magura, D. D., and Corley, W. G., “Tests to Destruction of a Multi
Panel Waffle Slab Structure 1964-1965 New York World’s Fair,” ACI
Journal Proceedings, V. 69, No. 9, Sept. 1971, pp. 699-703.
3. Xiang, X. Z., “Punching Shear Strength of Waffle Slabs at Internal
Columns,” PhD dissertation, University of Leeds, Leeds, UK, 1993.
4. Hussein, A. F., “Punching Shear Strength at Edge Columns in Ribbed
Flat Slabs,” PhD dissertation, University of Leeds, Leeds, UK, 1994.
5. Mosalam, K. M., and Naito, C. J., “Seismic Evaluation of Grav-
ity-Load-Designed Column-Grid System,” Journal of Structural
Engineering, ASCE, V. 128, No. 2, 2002, pp. 160-168. doi: 10.1061/
(ASCE)0733-9445(2002)128:2(160)
6. Benavent-Climent, A.; Cah’ıs, X.; and Catal’an, A., “Seismic
Behaviour of Interior Connections in Existing Waffle-Flat-Plate Struc-
tures,” Engineering Structures, V. 30, No. 9, 2008, pp. 2510-2516. doi:
10.1016/j.engstruct.2008.02.004
7. Souza, S. S. M., and Oliveira, D. R. C., “Reinforced Concrete Waffle
Flat Slabs under Shearing,” IBRACON Structures and Materials Journal,
V. 4, No. 4, 2011, pp. 610-625.
8. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-08) and Commentary,” American Concrete Institute,
Farmington Hills, MI, 2008, 473 pp.
9. Eurocode 2, “Design of Concrete Structures—Part 1-1: General Rules
and Rules for Buildings,” CEN, EN 1992-1-1, Brussels, Belgium, 2004,
225 pp.
10. British Standards Institution, “Structural Use of Concrete: Part 1,
London, UK, 1997, 120 pp.
11. Al-Bayati, A. F.; Lau, T. L.; and Clark, L. A., “Concentric Punching
Shear of Waffle Slab,” ACI Structural Journal, V. 112, No. 5, Sept.-Oct.
2015, pp. 533-542. doi: 10.14359/51687906
12. Binici, B., and Bayrak, O., “Upgrading of Slab-Column Connections
Using Fiber Reinforced Polymers,” Engineering Structures, V. 27, No. 1,
2008, pp. 97-107. doi: 10.1016/j.engstruct.2004.09.005
13. Johnson, R. P., “Strength Tests on Scaled Down Concrete Suitable
for Models with a Note on Mix Design,” Magazine of Concrete Research,
V. 14, No. 40, 1962, pp. 47-53. doi: 10.1680/macr.1962.14.40.47
14. Endersbee, L. A., “Discussion of ‘Size Effects in Small-
Scale Models,’” ACI Journal Proceedings, V. 63, No. 11, Nov. 1966,
pp. 1572-1573.
15. Hawkins, N. M.; Bao, A.; and Yamazaki, J., “Moment Transfer
from Concrete Slabs to Columns,” ACI Structural Journal, V. 86, No. 6,
Nov.-Dec. 1989, pp. 705-716.
16. Ghali, A.; Elmasri, M.; and Dilger, W., “Punching of Flat Plates
face, under Static and Dynamic Horizontal Forces,” ACI Journal Proceedings,
V. 73, No. 10, Oct. 1976, pp. 566-572.
17. Regan, P. E.; Walker, P. R.; and Zakaria, K. A. A., “Tests of Rein-
forced Concrete Flat Slabs,” CIRIA Project RP 220, School of the Environ-
ment Polytechnic of Central London, London, UK, 1979, 217 pp.
18. Ozden, S.; Ersoy, U.; and Ozturan, T., “Punching Shear Tests of
Normal-Strength and High-Strength Concrete Flat Plates,” Canadian
Journal of Civil Engineering, V. 33, No. 11, 2006, pp. 1389-1400. doi:
10.1139/l06-089
19. Regan, P. E., “Behaviour of Reinforced Concrete Flat Slabs,” CIRIA
Report 89, Construction Industry Research and Information Association,
London, UK, 1981, 89 pp.
20. Alexander, S. D. B., and Simmond, S. H., “Shear-Moment Interac-
V. 15, No. 5, 1988, pp. 828-833. doi: 10.1139/l88-108
21. Al-Bayati, A. F., “Punching Shear Mechanisms of Waffle Slabs,”
PhD dissertation, University of Nottingham Malaysia Campus, Malaysia,
2013.
ACI Structural Journal/January 2018
 22. Marzouk, H., and Hussein, A., “Experimental Investigation on the
Behavior of High-Strength Concrete Slabs,” ACI Structural Journal, V. 88,
No. 6, Nov.-Dec 1991, pp. 701-713.
23. Braestrup, M. W.; Nielsen, M. P.; Jensen, B. C.; and Bach, F.,
“Axisymmetric Punching and Reinforced Concrete,” Report No. R-75,
Structural Research Laboratory, Technical University of Denmark, Lyngby,
Denmark, 1976, 33 pp.
24. Salim, W., and Sebastian, W. M., “Plasticity Model for Predicting
Punching Shear Strengths of Reinforced Concrete Slabs,” ACI Structural
Journal, V. 99, No. 6, Nov.-Dec. 2002, pp. 827-835.
25. Lau, T. L., and Clark, L. A., “Shear Transfer between Ribbed Slab
and Internal Column,” Magazine of Concrete Research, V. 59, No. 7, 2007,
pp. 115-128. doi: 10.1680/macr.2007.59.7.507
26. Nielsen, M. P.; Brsestrup, M. W.; Jensen, B. C.; and Bach, F.,
“Concrete Plasticity-Beam Shear-Shear in Joints-Punching Shear,” Special
Publication, Danish Society for Structural Science and Engineering, Struc-
tural Research Laboratory, Technical University of Denmark, Lyngby,
Denmark, 1978, 129 pp.
27. Nielsen, M. P., and Hoang, L. C., Limit Analysis and Concrete Plas-
ticity, third edition, CRC Press, Boca Raton, FL, 2011, 799 pp.
28. Sigurdsson, T. G., “Punching Shear by Eccentric Loading,” MSc
dissertation, Technical University of Denmark, Lyngby, Denmark, 1991.
(in Danish)
29. Boswell, L. F., and Wong, S. S., “The Shear Behaviour of
Micro-Concrete and Its Inclusion in a Proposed Yield Criterion,” Proceed-
ACI Structural Journal/January 2018
ings of Cement and Concrete Association Research Seminar, Wexham
Springs, 1981.
30. Wong, S. S., “Collapse Behaviour of Micro-Concrete Box Girders
Bridges,” PhD thesis, City University, London, UK, 1998.
31. Mattock, A. H.; Ibrahim, I. O.; and Hofbeck, J. A., “Shear Transfer in
Reinforced Concrete,” ACI Journal Proceedings, V. 66, No. 2, Feb. 1969,
pp. 119-128.
32. Clark, L. A., Concrete Bridge Design to BS 5400, Construction Press,
London and New York, 1983, 186 pp.
33. Kanoh, Y., and Yoshizaki, S., “Strength of Slab-Column Connections
Transferring Shear and Moment,” ACI Journal Proceedings, V. 76, No. 3,
Mar. 1979, pp. 461-478.
34. Yamazaki, J., and Hawkins, N.M., “Behavior of Concrete Plates
Jointed to Columns,” Proceedings Japan Society of Civil Engineers,
No. 292, pp. 117-130.
35. Park, H., and Choi, K., “Improved Strength Model for Interior Flat
Plate-Column Connections Subject to Unbalanced Moment,” Journal
of Structural Engineering, ASCE, V. 132, No. 5, 2006, pp. 694-704. doi:
10.1061/(ASCE)0733-9445(2006)132:5(694)
36. Tian, Y.; Jirsa, J. O.; and Bayrak, O., “Strength Evaluation of Inte-
rior Slab-Column Connections,” ACI Structural Journal, V. 105, No. 6,
Nov.-Dec. 2008, pp. 792-700.
37. Broms, C. E., “Design Method for Imposed Rotations of Inte-
rior Slab-Column Connections,” ACI Structural Journal, V. 106, No. 5,
Sept.-Oct. 2009, pp. 636-645.
173
 ACI STRUCTURAL JOURNAL GENERAL INFORMATION

ACI Research and Academic Opportunities

This article details some of the opportunities for researchers and professionals upon becoming a part of the ACI community. This article
will outline the possibilities available to members, such as attending The ACI Concrete Convention and Exposition, viewing past technical
presentations, access to a vast abstract library, and ACI’s Call for Papers. Up-to-date information concerning these and additional
opportunites can be found at ACI’s website, www.concrete.org.

THE ACI CONCRETE CONVENTION ACI Convention Schedule

AND EXPOSITION City Location Dates
ACI Conventions give attendees the opportunity to partic-
Salt Lake City, UT, Grand America &
ipate in the development of industry codes and standards, March 25-29, 2018
USA Little America
learn about the latest in concrete technology, network
with leading concrete professionals, and fulfill potential Rio All-Suites Hotel &
Las Vegas, NV, USA October 14-18, 2018
Casino
continuing education requirements.
ACI technical and educational sessions, which are held Quebec City
Quebec City, QC,
Convention Centre & March 24-28, 2019
during ACI Conventions, provide attendees with the latest Canada
Hilton Hotel
research, case studies, best practices, and opportunities to
earn Professional Development Hours and Continuing Duke Energy
Convention Center &
Education Units. ACI committees, whose meetings take Cincinnati, OH, USA
Hyatt Regency
October 20-24, 2019
place during the ACI Convention, develop the standards, Cincinnati
reports, and other documents needed to keep those in the
Rosemont, IL, USA Hyatt Regency O’Hare March 29-April 2, 2020
industry up to date with the latest technology. Committee
meetings are open to all registered convention attendees. Raleigh Convention
Raleigh, NC, USA Center & Raleigh October 25-29, 2020
The ACI Convention takes place twice a year—once in Marriott
the fall and once in the spring. ACI reserves rooms at local
Hilton & Marriott
hotels and offers a discounted rate to members. Networking Baltimore, MD, USA March 28-April 1, 2021
Baltimore
and other nontechnical events are coordinated through ACI
and take place at each convention. Hilton Atlanta
Atlanta, GA, USA October 17-21, 2021
Downtown

TECHNICAL PRESENTATIONS AND DOCUMENTS

descriptions of submission requirements and policies can
Access to a vast abstract library, online educational
be found at www.concrete.org. ACI’s website also contains
presentations, webinars, and ACI education documents are
a detailed list of the date(s), sponsor(s), and location(s) of
often free for members or offered at a discounted rate. New
events calling for papers.
presentations and documents are always being added.
Guidelines for submitting technical papers for review
to either the ACI Structural Journal or the ACI Materials
CALL FOR PAPERS
Journal can be found at ACI’s website.
ACI is accepting the submission of papers for conven-
tions, committees, chapters, and subsidiaries. Detailed

1132 ACI Structural Journal/September-October 2016

174 ACI Structural Journal/January 2018

 ACI STRUCTURAL JOURNAL
Title No. 115-S14
Modeling Time-Dependent Deformations: Application
for Reinforced Concrete Beams with Recycled
Concrete Aggregates
by Adam M. Knaack and Yahya C. Kurama
This paper describes the development and validation of a time-
dependent fiber-based (that is, layered) numerical model for the
service-load deflection analysis of reinforced concrete structures.
Specifically, the model is applied to analyze the effect of recycled
concrete aggregates (RCA) on the deflections of beams. Previous
research has shown that increased deflections (rather than reduced
strength) may be a greater limitation for the use of RCA to replace
natural coarse aggregates (for example, gravel and crushed lime-
stone) for increased sustainability. Analysis and design tools are
needed to quantify this increase in deflections so that limits on the
use of recycled aggregates can be established. To aid in quanti-
fying these increased deflections, a new time-dependent concrete
fiber including creep and shrinkage strains was developed in the
open-source structural analysis program, OpenSees (Open System
for Earthquake Engineering Simulation). This paper describes the
validation of the model, including recent data from service-load
tests of slender cracked and uncracked RCA concrete beams.
The  model was able to predict the time-dependent deflections
of reinforced concrete structural members under various load
scenarios; however, the initial (instantaneous) deflections were
generally underestimated because of underestimations in the extent
of cracking. The model was also able to predict the total strains
and increased neutral axis depth over time as a result of creep and
shrinkage strains, except for the tension strains that were underes-
timated because of the inability of the model to accurately predict
the amount of cracking. It was found that the shrinkage strains
had a large effect on the time-dependent deflections of the beam
test specimens, which was not a finding available from the exper-
imental measurements. While the paper focuses on RCA concrete
applications, the numerical model is a general-purpose tool that
can be used to analyze the time-dependent axial-flexural deforma-
tions of conventional reinforced concrete structures as well.
Keywords: creep; deflection; fiber modeling; recycled concrete aggregate;
reinforced concrete; service load; shrinkage.
INTRODUCTION
Previous research on the numerical modeling of time-
dependent deformations of concrete structures is extensive
and has included both fiber element (that is, layered) as well
as continuum finite element approaches (for example, Lou
et al. [2016], Mercan et al. [2013], Bacinskas et al. [2012],
Sousa et al. [2012], Marí et al. [2010], Mazzotti and Savoia
[2009], Chong et al. [2008], Maekawa et al. [2006], and
Scanlon and Murray [1974]). Different from these previous
research efforts, the current paper describes a new time-de-
pendent fiber-based numerical model within the framework
of an open-source, general-purpose structural analysis soft-
ware (OpenSees 2006). Additionally, a specific application
ACI Structural Journal/January 2018
TECHNICAL PAPER
of the numerical model is presented to estimate the time-de-
pendent behavior of a series of 18 cracked and uncracked
reinforced concrete beams with recycled concrete aggre-
gates (RCA) as replacement for natural coarse aggregates
(for example, crushed stone and gravel).
RESEARCH SIGNIFICANCE
By recycling old concrete to replace natural aggregates,
it may be possible to substantially reduce the need for new
aggregates, thus helping with the preservation of forested
areas and riverbeds. The use of RCA in the United States has
been limited to non-structural applications such as roadway
subbase, although the quality of RCA is usually much
higher than is required in these applications and the material
is widely available in recycling plants across the country.
Increased service-load deflections are one of the greatest
limitations for the use of RCA in structural applications
because of greater effects of RCA on the concrete stiffness,
creep, and shrinkage than on the concrete strength (Knaack
and Kurama 2013a, 2015a). The analytical model described
in this paper is able to quantify this effect. The open-source
application can especially serve as a tool to design struc-
tures that satisfy allowable service-load deflection limits not
only for RCA concrete but also for conventional reinforced
concrete structures.
BACKGROUND
Deflections of reinforced concrete flexural members
can increase up to three to four times as a result of time-
dependent long-term load effects (Espion 1988). This
increase can be exacerbated when using RCA concrete. For
example, considerable increases have been observed in the
immediate and time-dependent deflections of RCA concrete
beam test specimens (Maruyama et al. 2004; Sato et al.
2007; Li 2009; Lapko and Grygo 2010; Malešev et al. 2010;
Ajdukiewicz and Kliszczewicz 2011; Choi and Yun 2013).
The results from these studies are generally limited to a few
tests using RCA and only one load level in each research
program. For example, for a service-load duration of approx-
imately 15 weeks, Lapko and Grygo (2010) showed 20%
greater deflections in RCA concrete beams with full (100%)
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2017-019.R1, doi: 10.14359/51701153, was received January 29, 2017,
and reviewed under Institute publication policies. Copyright © 2018, American
Concrete Institute. All rights reserved, including the making of copies unless
permission is obtained from the copyright proprietors. Pertinent discussion including
author’s closure, if any, will be published ten months from this journal’s date if the
discussion is received within four months of the paper’s print publication.
175
aggregate replacement as compared to natural aggregate
(NA) concrete beams. These tests were conducted on small-
scale beams with a cross section of 3 x 5 in. (76 x 127 mm).
Ajdukiewicz and Kliszczewicz (2011) found similar results
for beams with full aggregate replacement loaded for up to
1 year. Choi and Yun (2013) found the ratios of long-term
to immediate deflections for RCA concrete beams to be
smaller than the ratios for conventional natural aggregate
(NA) concrete beams. A more recent study involving 18
RCA concrete beams was conducted by Knaack and Kurama
(2013b; 2015b).
None of the aforementioned experimental studies included
numerical modeling of the time-dependent deflections of
RCA concrete structures. Various methods are available
for the prediction of the long-term service-load deflections
of reinforced concrete structures, including simple design
code-based methods, closed-form analytical solutions, and
detailed fiber element and continuum finite element models.
Code approaches such as in ACI 318-14 and CEN (2004)
are generally simple, but not accurate (Ghali 1993; Gilbert
1999). Many analytical and numerical approaches (Teng
and Branson 1993; Nie and Cai 2000; Kaklauskas 2004;
Rodriguez-Gutierrez and Aristizabal-Ochoa 2007; Balev-
ičius 2010) have been demonstrated to provide consid-
erable improvements on the deflection predictions. One
important limitation of the available numerical models is
that they have used commercial finite element programs (for
example, ABAQUS and DIANA) or specific in-house codes
developed by the researchers. Different from these
approaches, the model in this paper uses simplifying assump-
tions and approximations within a widely and freely available
structural analysis platform, OpenSees (2006). Furthermore,
to the best of the authors’ knowledge, this is the first detailed
numerical modeling study for RCA concrete beams.
TIME-DEPENDENT CONCRETE MATERIAL MODEL
To result in a widely available and widely applicable
analysis tool, a new concrete material model including
time-dependent creep and shrinkage strains under sustained
service loads was implemented by adapting the existing
Static Analysis and fiber element modeling framework in the
open-source structural analysis program, OpenSees (2006).
One of the advantages of fiber element models is the ability
to represent the axial-flexural behavior of concrete structures
using cross sections that are discretized into concrete and
steel fibers together with a uniaxial stress-strain relationship
for each fiber. More information on fiber element modeling
for structural analysis can be found in Taucer et al. (1991)
and Jiang and Kurama (2010).
All of the available material models in OpenSees are
time-invariant, and as such, there is no built-in capa-
bility for a time-dependent creep and shrinkage analysis.
However, the magnitudes of the applied loads during a Static
Analysis are defined using a “pseudo-time” concept (that
is, time associated with the analysis domain). For example,
consider a linear time series with a pseudo-time step of
0.1 days for an applied force, P reached at 1 day. After one
analysis step, the pseudo-time in the model is 0.1 days and the
applied load is 0.1P. After 10 analysis steps, the pseudo-time
176
is 1 day and the applied load is equal to the full load, P. The
existing pseudo-time concept in the OpenSees Static Analysis
was advantageous for the development of the new time-
dependent material model in the sense that time was already
defined in the program for use as the time domain in the new
material model.
To allow for the inclusion of concrete creep and shrinkage
deformations during a pseudo-time step, a new analysis type
called Creep Analysis was implemented based on the Static
Analysis function already defined within the OpenSees
framework, and a new uniaxial concrete material type called
TDConcrete was developed based on the existing Concrete02
material. Full information on the new capabilities imple-
mented in OpenSees (including source codes, sample input
and analysis files, and element recorders) can be found in the
final project report (Knaack and Kurama 2013b).
Overview
The new time-dependent TDConcrete material in
OpenSees uses the following assumptions:
1. Concrete is fully hardened and able to resist load 2 days
after casting.
2. Because the analysis is carried out under service loads,
concrete is assumed to remain in the linear elastic range in
compression.
3. Concrete fibers exceeding the cracking strain are expected
to have some residual stresses through tension stiffening.
4. As an important limitation, the strength and stiffness of
concrete are assumed to remain constant during the entire
analysis (that is, increases in the concrete strength and stiff-
ness with age are not included). This is a reasonable approx-
imation for structures loaded at or after 7 days of curing, as
investigated in this paper and as would be most common in
practice (due to presence of shoring). For structures loaded
prior to 7 days, the strength and stiffness gain of concrete can
be significant (Neville 2011), and thus, numerical models
that incorporate the effects of aging may be needed for more
accurate analysis.
5. There are no thermal strains (that is, no significant
temperature changes) during the analysis.
6. The stress in each fiber remains constant during each
pseudo-time step. As described later, this is a common
assumption that has been used in previous research (Zien-
kiewicz et al. 1968; Kabir 1977; Kawano and Warner 1996).
7. The concrete creep strain in tension has an equal
magnitude to the creep strain in compression for the same
magnitude of stress. While this is also a common assump-
tion that has been used in the past (Kabir 1977; Gilbert and
Ranzi 2011), according to Neville (2011), creep of concrete
in tension is generally greater than in compression, with
differences as high as 100% for concrete stored at a relative
humidity of 50% and loaded at early ages. As such, some of
the model discrepancies discussed later in this paper could
have developed from an underestimation of the creep strains
in tension. However, Neville (2011) also states that reli-
able conclusions about the relative magnitudes of creep in
tension and in compression cannot be made, citing contra-
dictory evidence that shows greater compression creep than
tension creep.
ACI Structural Journal/January 2018

Fig. 1—Mechanical stress-strain behavior of concrete
(compression negative). (Note: 1 MPa = 0.145 ksi.)
8. Once the concrete cracking strain has been exceeded,
creep and shrinkage still occur according to the same consti-
tutive relationships as an uncracked fiber, but under the
reduced post-cracking (that is, tension stiffening) stresses.
To analyze the time-dependent behavior of a concrete
structure, a series of Static and Creep Analyses are
conducted at discrete times throughout the loading duration.
For example, consider the following scenario for the loading
and construction of a reinforced concrete column:
1. t = 0 days: Start time of the analysis. Column is cast and
moist cured for 7 days.
2. t = 2 days: Column is assumed to be fully hardened and
stresses are induced due to self-weight. Creep strains begin
to accumulate.
3. t = 7 days: Moist curing of column ends and shrinkage
strains begin to accumulate.
4. t = 14 days: Additional loads applied from the slab above.
An OpenSees analysis of this column can be conducted
by modeling the concrete using the TDConcrete material.
To conduct a time-dependent analysis, the user must set the
pseudo-time of the model so that a combination of Static
Analyses and Creep Analyses can be conducted at the desired
times to determine the initial and long-term deflections of
the column. To set the pseudo-time, the OpenSees “setTime”
command is used. For the aforementioned column construc-
tion scenario, the analysis would progress as follows:
1. Static Analysis at pseudo-time, t = 2 days to apply
column self-weight. Because the column is cast at t = 0 days
and assumed to take load once fully hardened at t = 2 days,
the pseudo-time at the beginning of the analysis is set to
2 days.
2. A series of Creep Analyses starting at t = 2 days to accumu-
late creep deformations from t = 2 days to 7 days. Note that no
shrinkage strains would accumulate during moist curing.
3. A new series of Creep Analyses starting t = 7 days to
start accumulating shrinkage deformations in addition to
ACI Structural Journal/January 2018
creep deformations. The last Creep Analysis for this loading
duration should be conducted at t = 14 days.
4. Static Analysis at t = 14 days to apply additional loads
from the slab above.
5. A series of Creep Analyses starting at t = 14 days to
accumulate creep and shrinkage deformations from t = 14
days to the end of the loading duration.
It can be seen that a new series of Static Analysis and
subsequent Creep Analyses is conducted for each increment
of applied loads (for example, if additional dead loads or
live loads are applied to the column at later times). Typically,
the best practice for each series of Creep Analyses is to use
pseudo-time steps (by explicitly specifying the pseudo-time
for each analysis using the OpenSees “setTime” command)
that are uniformly spaced in logarithmic scale (Bažant and
Wu 1973) so that more analyses are conducted when the load
is first applied (that is, when the creep strains increase at a
faster rate). The analyses can then be run using the “analyze”
command.
The “analyze” command in OpenSees will run one Creep
Analysis at the pseudo-time set by the user. To run a series
of Creep Analyses in succession at additional time steps (to
continue accumulating creep and shrinkage deformations for
the remainder of the loading duration), the “setTime” and
“analyze” commands are put inside a loop in the OpenSees
input file so that the pseudo-time is set prior to each analysis
for the desired number of analyses. An example time-
dependent analysis loop using logarithmically spaced time
steps is provided in Appendix D of Knaack and Kurama
(2013b).
Note that when conducting a new series of Static
Analysis and subsequent Creep Analyses, the entire load
increment for that series must be applied instantaneously
during the Static Analysis and then held constant for the
duration of the subsequent Creep Analyses. This requires
the use of a constant time-series to define each applied
load increment in OpenSees. If additional load increments
are placed on the structure at a later time, a new series of
Static Analysis and subsequent Creep Analyses should be
conducted using a similar process.
The TDConcrete material defines the total concrete strain
(excluding thermal strains) at time, t as
total (t ) = �  m (t ) +  cr (t ) +  sh (t ) (1)
where ϵtotal is total strain; ϵm is mechanical strain; ϵcr is creep
strain; and ϵsh is drying shrinkage strain.
During a Creep Analysis step at time t, the strain compo-
nents for each TDConcrete fiber are determined by
carrying out iterations until a specified level of conver-
gence is achieved on the global unbalanced force vector
in OpenSees. The fiber shrinkage strain, ϵsh(t) is calcu-
lated using the difference between the time t for the current
analysis step and the time td at the start of concrete drying,
and remains constant during all iterations within the time
step. The creep strain, ϵcr(t) for the current analysis step is
calculated using the stress from the previous analysis step
and also assumed to remain constant for the time step. Then,
the mechanical strain is calculated by subtracting ϵsh(t) and
177
Fig. 2—Concrete strains: (a) strain components (excluding thermal strain) under constant stress;
and (b) superposition of
creep strains under time-varying stress.
ϵcr(t) from the total strain, ϵtotal(t), and the concrete mechanical
stress-strain constitutive relationship is used to determine the
stress, σ(t) in the fiber. Finally, the stresses of all fibers within
the cross section are integrated and the internal forces within
the element are passed on to the global analysis to check
convergence of the structure unbalanced force vector.
Provided that the Creep Analysis time steps are sufficiently
small so that there is only a small amount of stress change
within each time interval, the aforementioned method leads
to a reasonably accurate determination of the creep strains.
A possibly more accurate alternative to this approach would
be to assume a linearly varying concrete stress within each
time step; however, this approach was not used in this
research because it was deemed unjustified given the many
inherent uncertainties on the assumed material properties
(for example, concrete strength and stiffness).
Mechanical stress-strain behavior
Figure 1 shows the assumed concrete mechanical stress-
strain behavior. As stated previously, the model uses a
linear-elastic stress-strain relationship in compression. The
concrete stiffness, Ec and tension strength, ft′ are input by
the user and should typically be specified at the concrete age
when the superimposed service loads are first applied. The
tension-stiffening behavior after cracking uses a model from
Tamai et al. (1988). A value for the tension-stiffening expo-
nent, bts is recommended by Tamai et al. (1988) as bts = 0.4
for concrete with deformed reinforcing bars. This value is
based on a study with steel reinforcement in the post-yield
range, and users can specify other values of bts for the linear
steel material range if experimental data is available. Once
the concrete is cracked, any load reversal is assumed to
follow the secant stiffness from the maximum strain reached
178
on the tension-stiffening envelope towards the origin. Any
subsequent reloading in tension also occurs along this secant
stiffness until a new maximum tension strain is reached.
Creep strain constitutive relationship
Figure 2(a) shows the three components of the concrete
total strain, given initiation of drying at t = td and a constant
stress applied at t = t0. According to Gilbert and Ranzi
(2011), creep strain is approximately proportional to stress
when the sustained stress is within the linear elastic limit.
When the stress is assumed to remain constant with respect
to time (that is, dσ/dt = 0), the creep strain at time t can be
expressed as
 cr (t ) =  m (t ) ⋅ φ0 (t ) (2)
 m (t ) = σ (t ) / Ec (3)
where ϕ0(t) is the creep coefficient at time t.
Then the load-induced strain (that is, sum of mechanical
and creep strains) can be written as
 m (t ) +  cr (t ) =  m (t ) ⋅ 1 + φ0 (t )  (4)
As shown in Fig. 2(b) and explained in Gilbert and Ranzi
(2011), the creep strain at time t as a result of time-varying
stress can be determined using the principle of superposition as
 φ0 (t ) ⋅ ∆σ 0 / Ec , t0 ≤ t < t1

 φ0 (t ) ⋅ ∆σ 0 + φ1 (t ) ⋅ ∆σ1  / Ec , t1 ≤ t < t2
cr (t ) =   (5)
 
 φ0 (t ) ⋅ ∆σ 0 + …+ φ n (t ) ⋅ ∆σ n  / Ec , tn ≤ t

ACI Structural Journal/January 2018
Fig. 3—Validation under constant uniaxial stress: (a) model; and (b) comparisons for creep frame CS-11 from Knaack and
Kurama (2013b, 2015a). (Note: 1 mm = 0.0394 in.)
This relationship can also be written as
n
cr (t ) = ∑φi (t ) ⋅ ∆σ i / Ec (6)
i =0
Gilbert and Ranzi (2011) found that the principle of super-
position applied to creep strains agrees well with available
test data for increasing stress histories. However, for concrete
under decreasing stresses (that is, unloading), superposition
tends to overestimate the amount of creep recovery, thus
leading to reduced creep deformations. Regardless, the prin-
ciple of superposition provides a reasonable approximation
of the creep strains for most practical purposes. The process
is assumed to be valid for both uncracked and cracked
concrete. Note that in a real reinforced concrete structure,
the concrete immediately adjacent to a crack unloads and is
not capable of carrying any stress, which is not captured by
the tension stiffening modeling approach.
Note also that the summation in Eq. (6) requires the entire
stress history of each fiber to be stored during the anal-
ysis for each integration point along each element. While
not implemented in TDConcrete, the use of a Dirichlet
series approximation (Bažant and Wu 1973; de Borst and
van den Boogaard 1994; Kabir 1977; Taylor et al. 1970)
to the creep strain superposition could reduce computing
times by allowing the concrete state (that is, stress and total
strain) to be determined from the previous time step rather
than requiring integration at each time step over the entire
strain history.
The creep coefficient at time t is assumed to follow
ACI 209R-92 model as
φ i ( t ) = ( t − ti ) /  ψ cr , 2 + (t − ti )
ψ cr ,1
 
where ti is time at stress change; ψcr,1 and ψcr,2 are creep
time-function parameters; and ϕu is ultimate creep coeffi-
cient. The creep time-function parameters ψcr,1 and ψcr,2 and
the ultimate creep coefficient ϕu are specified by the user as
inputs to TDConcrete. Because the creep model in Eq. (7)
is for a standard condition corresponding to moist-cured
concrete loaded at an age of 7 days, all of the modifications
(multipliers) specified in ACI 209R for non-standard condi-
tions can be applied by adjusting ϕu.
ACI Structural Journal/January 2018
Shrinkage strain constitutive relationship
Similar to the creep constitutive relationship, TDConcrete
uses ACI 209R-92 drying shrinkage model as
 sh (t ) = (t − td ) / ( ψ sh + t − td ) ⋅  sh ,u (8)
where td is time at initiation of concrete drying; ψsh is
shrinkage time-function parameter; and ϵsh,u is ultimate
shrinkage strain. All ACI 209R-92 adjustment factors for
non-standard conditions can be applied by modifying the
ultimate shrinkage strain ϵsh,u provided as input.
VALIDATION OF TIME-DEPENDENT
CONCRETE MODEL
Constant stress under uniaxial load
The first validation study for the TDConcrete analysis
package implemented in OpenSees was based on the uniaxial
RCA concrete creep test data from Knaack and Kurama
(2013b, 2015a). In each creep test frame, three pairs of plain
concrete cylinders were stacked and a constant axial load
was applied at a specific age. The three pairs of cylinders
represented concrete mixtures with three different aggre-
gate volumetric replacement ratios (that is, R = 0%, 50%,
and 100%; where R = 0% means no RCA in the concrete
mixture and R = 100% means no natural coarse aggregate
in the mixture), and the results from each pair were aver-
aged to determine the total strain over the loading duration.
The analytical model for this experimental configuration is
shown in Fig. 3(a).
During each test, shrinkage strains were simultane-
ously measured using companion cylinder pairs under no
applied load, and the load-induced strain (that is, creep plus
 ⋅ φu (7)
ψ cr ,1
mechanical strain) of the creep cylinders was determined by
subtracting the shrinkage strain from the total strain. The
mechanical strain was taken as the first data point immedi-
ately after load application, and the creep strain was deter-
mined as the load-induced strain minus this mechanical
strain. The creep coefficient was calculated as the creep strain
divided by the mechanical strain and a nonlinear regression
analysis was conducted using Eq. (7) to provide the best fit
to the data. The resulting ψcr,1, ψcr,2, and ϕu parameters were
then provided as input to TDConcrete. The concrete stiff-
179
Fig. 4—Validation under time-varying stresses: (a) model; and (b) comparison with Kawano and Warner (1996). (Note: 1 mm =
0.0394 in.; 1 m = 39.4 in.; 1 MPa = 0.145 ksi.)
ness  Ec was calculated from the mechanical strain. In the
analysis, the load was applied at the same age and magnitude
as the test and was held constant over the entire duration.
As illustrative results, the analytical and experimental
load-induced strains for creep frame CS-11 (Knaack and
Kurama 2013b, 2015a) are shown in Fig. 3(b). The results
from the other creep frames exhibited similar comparisons.
It can be seen that the TDConcrete model provided an excel-
lent fit to the test data by producing the creep strain model
that was used as input. This validates that for a constant
stress, TDConcrete was able to correctly use the intended
creep formulation within the Creep Analysis implemented
in OpenSees.
Time-varying stress under uniaxial load
The validation of TDConcrete under variable uniaxial
stresses involved the analysis of a reinforced concrete
column previously analyzed by Kawano and Warner (1996).
In this analysis, the column was subjected to a constant
uniaxial load, but time-varying stresses were present as
redistribution of stresses between concrete and reinforcing
steel (reduction of concrete stresses and increase of steel
stresses) occurred due to concrete creep and shrinkage. As
shown in Fig. 4(a), the analytical model consisted of a single
element with one TDConcrete fiber and eight steel fibers
(using the Steel01 material in OpenSees) to represent the
column reinforcement. Figure 4(b) shows no differences
between the analysis predictions from Kawano and Warner
180
(1996) and TDConcrete for the varying concrete stresses or
the creep strains during the analysis.
Time-varying stress under flexural load
Validation of TDConcrete under flexural load was
conducted using the analytical example from Kabir (1977),
which was based on a beam specimen tested by Washa and
Fluck (1952). The beam was analyzed using the model in
Fig. 5(a). The concrete compression strength and modulus of
elasticity (at an age of 14 days, when superimposed loading
was applied), as well as the steel yield strength, were taken
from measured values. Cracking was observed under the
initial application of load; and thus, the concrete tension
strength was determined so that the initial deflection from
the analysis matched the measured deflection. A tension
stiffening exponent of bts = 0.4 was used as recommended
by Tamai et al. (1988). As shown in Fig. 5(b), the concrete
creep parameters, ψcr,1, ψcr,2, and ϕu were determined from
nonlinear regression analysis on the creep coefficient from
the concrete cylinders tested by Washa and Fluck (1952).
Similar to Kabir (1977), shrinkage strain was neglected in
the analysis.
The beam deflection results depicted in Fig. 5(c)
demonstrate an excellent fit with Kabir (1977)
over the entire time history. Both the TDConcrete and Kabir
analyses gave conservative (that is, greater) beam deflec-
tions in comparison with the test data from Washa and Fluck
(1952) for approximately the first 100 days of loading,
which could be because the effects of aging on the concrete
ACI Structural Journal/January 2018
Fig. 5—Beam under flexural load: (a) model; (b) regression fit to creep data; and (c) comparisons for midspan deflection.
(Note: 1 mm = 0.0394 in.; 1 m = 39.4 in.)
strength and stiffness were ignored. After this loading
period, both analyses provided reasonable estimates of the
measured data. These results support the use of TDConcrete
in the flexural analysis of reinforced concrete structures.
VALIDATION AND APPLICATION FOR BEAMS
WITH RCA CONCRETE
This section describes the application of TDConcrete to the
RCA concrete beams tested by Knaack and Kurama (2013b,
2015b). As shown in Fig. 6(a), each l = 4.0 m long (13 ft)
beam was subjected to four-point bending using a concrete
weight block to simulate superimposed service loads for a
period of at least 119 days. The beams were b = 150 mm
(6.0 in.) wide and h = 230 mm (9.0 in.) deep. In the specimen
nomenclature (UT, UC, and CC series beams), the first letter
“U” indicates beams that were subjected to a superimposed
load designed not to induce immediate cracking, whereas
the first letter “C” indicates beams that were subjected to
a greater superimposed load to induce immediate cracking.
For the second letter of the nomenclature, the letter “T” indi-
cates beams with tension (bottom) longitudinal bars only and
“C” indicates beams with compression (top) longitudinal
bars and transverse stirrups in addition to tension (bottom)
longitudinal bars. Subsequent numbers in the nomenclature
indicate specimens with different amounts of coarse aggre-
gate replacement (R = 100, 50, and 0% replacement) and
beams loaded at the age of t0 = 7 days versus 28 days (for
example, UT-100-28 for a UT series beam with 100% aggre-
gate replacement and loaded at 28 days).
The average measured (on the day of superimposed
loading) material properties fc′, ft′, and Ec, were used as input
to the TDConcrete analyses. Considering that the beams
ACI Structural Journal/January 2018
were subjected to service-level loads, the steel stresses
remained in the elastic range. The concrete in compression
also remained within the elastic range (that is, σ < 0.50fc′).
The tension-stiffening exponent was chosen as bts = 0.8 to
result in the best match to the initial and time-dependent
deflections. Note that the use of a greater bts value for these
specimens as compared with bts = 0.4 for the specimen in
Fig. 5 indicates reduced tension-stiffening stresses in the
concrete (refer to the tension-stiffening equation in Fig. 1).
This may be expected because the smaller total area and
greater diameter of the two M16 bars for the specimens
in Fig. 6 likely resulted in smaller tension stress transfer
(through bond) to the concrete than the four M13 bars for
the specimens in Fig. 5.
Companion concrete cylinder creep and shrinkage test data
for each beam, as described in Knaack and Kurama (2013b,
2015a), were used to determine the creep and shrinkage
input (using nonlinear regression) into the TDConcrete
model. Six beams (UT-0-28, UT-0-7, UT-50-28, CC-50-7,
CC-100-28, and UC-100-7) did not have any companion
creep and shrinkage measurements; therefore, the measured
creep and shrinkage data for other beams were used as input
for these specific beams. Although ACI 209-92 suggests the
use of a size adjustment factor for creep and shrinkage, Euro-
code 2 (CEN 2004) does not require a size adjustment factor
for beams of this size. Therefore, it was decided to neglect
any size adjustment factor in the creep and shrinkage models
for the beams. Also, because the creep and shrinkage cylin-
ders were kept in the same room conditions and cured the
same way as the corresponding beam specimens, no adjust-
ment factor for humidity or curing needed to be applied.
181
Fig. 6—Service-load deflection modeling of RCA concrete beams: (a) test elevation; (b) analytical model; and (c) loading
schedule. (Note: 1 mm = 0.0394 in.; 1 m = 39.4 in.)
As shown in Fig. 6(b), each beam was modeled in
OpenSees using a total of 20 displacement-based beam-
column elements with Gauss-Lobatto integration scheme
and three integration points along each element. The
Legendre integration scheme, also available in OpenSees,
provided the same results as the Gauss-Lobatto scheme. The
beam cross-sections were discretized into 20 concrete fibers
and up to two (top and bottom) steel fibers depending on
the specimen. The load was applied in two steps: 1) uniform
beam self-weight assumed to take effect at t = 2 days after
casting (in the laboratory, each beam was placed on wooden
end blocks immediately after removal from formwork at
2 days); and 2) two superimposed point loads applied at
±190 mm (±7.5 in.) from the midspan at a service loading
age of t0 = 7 days or 28 days, depending on the beam. It was
assumed that the drying shrinkage of each beam began at td =
2 days (age at which the beams were removed from the form-
work) and that the shrinkage strains were uniformly distrib-
uted throughout the entire beam volume (that is, all of the
concrete fibers were given the same shrinkage properties).
182
Figure 6(c) shows the loading sequence for the test beams.
From t = td = 2 days until t = t0 (that is, first application of
superimposed service load), 20 pseudo-time steps uniformly
spaced in logarithmic scale were used to run a Static
Analysis and subsequent series of Creep Analyses to deter-
mine the immediate and creep deflections under self-weight
as well as the shrinkage deflections during this period.
As discussed in Bažant and Wu (1973), the use of a loga-
rithmic scale allows more time steps when the load is first
applied and creep and shrinkage strains are increasing at a
faster rate. From t = t0 until the age at which the service
load was removed at t = tr, another Static Analysis and Creep
Analysis series with 50 pseudo-time steps uniformly spaced
in logarithmic scale were used to determine the additional
immediate, creep, and shrinkage deflections during this
stage. Finally, from t = tr until the end of measurements
at t = te, the number of time steps was varied based on the
duration. A few beams had an additional superimposed load
applied during the last 2 months of testing (dashed lines in
Fig. 6(c)). This stage of increased loading was modeled using
30 logarithmically spaced time steps. Preliminary analyses
ACI Structural Journal/January 2018
Fig. 7—Deflection comparisons for UT series beams: (a) UT-0-7; (b) UT-0-28;
(c) UT-50-7; (d) UT-50-28; (e) UT-100-7; and
(f) UT-100-28. (Note: 1 mm = 0.0394 in.)
were conducted to ensure that the selected number of beam-
column elements, fibers, and time steps were adequate for
the accuracy of the analyses.
The ACI 209R-92 loading age adjustment factor, with
additional modifications (Knaack and Kurama 2013b) to
better represent the specific loading schemes from the exper-
imental program, were used in the analyses as follows.
For loading ages later than 7 days for moist cured concrete,
ACI 209R specifies a loading age adjustment factor as
γ la , ACI 209 = 1.25 ⋅ (ti − tcast )
−0.118
(9)
where ti is time at stress change; and tcast is time at concrete
casting. The ACI 209R age adjustment factor takes a value
less than 1.0 for ti – tcast greater than 7 days, and is intended
to be applied as a multiplier to the standard creep relation-
ship in Eq. (7). For the purposes of the analyses herein, the
ACI 209R adjustment factor in Eq. (9) was adapted to a
more general relationship applicable to other loading ages as
γ la = γ la , ACI 209 / 1.25 ⋅ Tcr−0.118  (10)
where Tcr is the loading age of the cylinders used to develop
the creep relationship (that is, 7 or 28 days in this research).
In this more general relationship, γla = γla,ACI209 when Tcr =
7 days (that is, Eq. (10) is reduced to Eq. (9) when the creep
relationship is based on specimens loaded at an age of 7 days),
and γla = 1 when ti – tcast = Tcr (that is, Eq. (10) gives a creep
adjustment factor of 1.0 when the age of concrete at stress
change is the same as the loading age of the cylinders used
to develop the creep relationship). Because ACI 209R does
ACI Structural Journal/January 2018
not specify an adjustment factor for loading ages less than
7 days, Eq. (10) was also used for loading ages less than Tcr
(to determine the creep strains under self-weight), resulting
in adjustment factors greater than 1.0 for those conditions.
The results from the time-dependent analyses are shown in
Fig. 7, 8, and 9 for the UT, UC, and CC series beams, respec-
tively. Similarly, Table 1 compares the predictions from
TDConcrete with the measured deflections at t = t0 (imme-
diate deflection from the application of superimposed load
at t0) and at t = t0 +119 days (total accumulated deflection
from t0 until t0 +119 days), which were zeroed immediately
before the superimposed loading at time, t = t0. Note that
measured deflection data for beams CC-50-28 and UT-100-7
were not available due to failure of the displacement trans-
ducers; and thus, only the analytical predictions are shown
for these beams. Note also that the analytical data in Fig. 7 to
9 start at t = td = 2 days, whereas the measured data start at t
= t0. Because the laboratory measurements did not start until
the application of the superimposed load at t = t0, the beam
deflections from t = td = 2 days to t = t0 (due to shrinkage and
the immediate and creep deflections due to self-weight) were
not measured. However, to accurately simulate the evolution
of concrete and steel stresses analytically, it was necessary
to run the analyses over the entire loading history beginning
at t = td = 2 days. The resulting analytical deflections were
subsequently zeroed at the application of the superimposed
load at t = t0 to permit direct comparisons with the measured
data starting at t = t0.
The comparisons show that the analytical models were
able to predict the time-dependent service-load deflections
of the test beams reasonably well; however, the immediate
183
Fig. 8—Deflection comparisons for UC series beams: (a) UC-0-7; (b) UC-0-28;
(c) UC-50-7; (d) UC-50-28; (e) UC-100-7;
and (f) UC-100-28. (Note: 1 mm = 0.0394 in.)

Fig. 9—Deflection comparisons for CC series beams: (a) CC-0-7; (b) CC-0-28;
(c) CC-50-7; (d) CC-50-28; (e) CC-100-7; and
(f) CC-100-28. (Note: 1 mm = 0.0394 in.)
deflections due to the application of the superimposed load RCA generally resulted in increased beam deflections. Also
were generally underestimated. The analyses were consis- consistent with the experimental results, RCA had greater
tent with the experimental results that increased amounts of effect on the deflections of beams with less cracking. The

184 ACI Structural Journal/January 2018

Table 1—Comparison of analytical results with measured beam deflections

Immediate superimposed load deflection at t = t0, Δsi, mm Immediate plus time-dependent deflection at t = t0 + 119, Δt 0 + 119, mm
Beam ID *
Measured Analytical Analytical/Measured Measured Analytical Analytical/Measured
UT-0-28 0.86 0.74 0.85 5.00 4.27 0.85
UT-0-7 0.74 0.61 0.83 4.62 5.89 1.27
UC-0-28 0.66 0.56 0.85 3.51 2.97 0.85
UC-0-7 0.94 0.64 0.68 5.11 4.04 0.79
CC-0-28 3.15 2.90 0.92 10.19 8.66 0.85
CC-0-7 3.40 3.10 0.91 10.69 10.64 1.00
UT-50-28 0.91 0.74 0.81 5.38 5.08 0.94
UT-50-7 0.94 0.76 0.81 6.96 7.01 1.01
UC-50-28 0.86 0.64 0.74 4.70 3.38 0.72
UC-50-7 0.84 0.64 0.76 5.99 5.36 0.89
CC-50-28 4.93 3.81 0.77 — 10.24 —
CC-50-7 4.14 3.68 0.89 12.90 12.07 0.94
UT-100-28 1.24 0.97 0.78 7.39 8.41 1.14
UT-100-7 1.12 0.94 0.84 — 9.07 —
UC-100-28 0.97 0.84 0.87 5.94 5.92 1.00
UC-100-7 1.27 0.86 0.68 7.62 5.82 0.76
CC-100-28 5.11 6.02 1.18 12.27 13.11 1.07
CC-100-7 4.60 4.67 1.02 14.68 14.38 0.98
*
Beam ID descriptors (Knaack and Kurama 2013b, 2015b): Left descriptors—UT is for beams with bottom (tension) longitudinal reinforcement only and subjected to superim-
posed load designed not to induce immediate cracking; UC is for beams with bottom (tension), top (compression), and transverse reinforcement and subjected to superimposed
load designed not to induce immediate cracking; CC is for beams with bottom, top, and transverse reinforcement and subjected to greater superimposed load to induce immediate
cracking; middle descriptors—0, 50, 100 for R of 0%, 50%, 100%, respectively; right descriptors—28 and 7 for t0 of 28 and 7 days, respectively.
Note: 1 mm = 0.0394 in.

analytical models for Beams UT-0-7 and UC-100-7 were
the worst performing among the UT and UC series beams,
respectively, which may have occurred because of the
lack of accompanying creep and shrinkage test data. The
worst performing model from the CC series was for Beam
CC-0-28. Companion creep and shrinkage test data were
available for this beam; however, as a potential source of
discrepancy, the measured concrete tension strength from the
accompanying modulus of rupture tests was disproportion-
ately large for the measured concrete compression strength.
The test beam demonstrated a greater level of cracking
(and therefore greater displacements) than the analytical
prediction based on the measured (disproportionately large)
concrete tension strength.
Although the UC series beams were loaded with a
smaller superimposed load to avoid initial cracking, there
was still some time-dependent cracking (as induced by
creep and shrinkage strains) due to the stress redistribu-
tions between concrete and reinforcing steel. The analytical
simulations predicted this cracking in all of the UC series
beams, except for Beams UC-0-28 and UC-0-7. Increased
creep and shrinkage strains generally caused more cracking
in beams with higher amounts of RCA (that is, higher R).
Thus, the extent of creep/shrinkage-induced cracking in
beams with R = 0% (for example, Beams UC-0-7 and
UC-0-28) was small as compared to beams with R = 100%,
making the prediction of this small amount of cracking
particularly difficult.
To evaluate the predictions for local beam behavior,
comparisons between the measured total strain diagrams
for the CC series beams and the analytical model results
are shown in Fig. 10. The measured total strain trends
compared reasonably well with the plane sections assump-
tion used in the analytical models. Note that the analytical
strain diagrams were zeroed at time t = t0 so that they could
be directly compared with the measurements. The model
was generally capable of predicting the downward shift in
the neutral axis as a result of creep and shrinkage strains
over time, with more downward shift for beams loaded at a
younger age and for beams with increased amounts of RCA.
However, even for the initial placement of superimposed
load at t = t0, the tension strains were significantly underesti-
mated due to the inability of the model to accurately predict
the amount of cracking. This finding is consistent with the
beam deflection comparisons in Fig. 9, which show that the
initial deflections of the analytical models under superim-
posed load were generally smaller than the measured deflec-
tions. The discrepancies in the time-varying deflections
under sustained loading were also likely affected by the
limitations of the analytical model to accurately predict the
amount of cracking.
Figure 11 shows the change in the beam neutral axis
depth, c (normalized with respect to the total beam depth,

ACI Structural Journal/January 2018 185

Fig. 10—Total strain diagrams for CC series test beams: (a) CC-0-7; (b) CC-0-28; (c) CC-50-7;
(d) CC-50-28; (e) CC-100-7;
and (f) CC-100-28 (compression negative). (Note: 1 mm = 0.0394 in.)

Fig. 11—Neutral axis depths for CC series test beams: (a) CC-0-7; (b) CC-0-28;
(c) CC-50-7; (d) CC-50-28; (e) CC-100-7;
and (f) CC-100-28.
h) over time, based on the strain diagrams in Fig. 10. The measured strain diagrams, a linear regression analysis was
neutral axis depth was determined as the distance from the conducted on the group of strain measurements taken at a
top of the beam to the location of zero total strain. For the given time to approximately determine the zero intercept of

186 ACI Structural Journal/January 2018

Fig. 12—Midspan stress and strain results for Beam CC-50-7: (a) top fiber stress; (b) top fiber strain components; (c) strain
diagram after loading; and (d) strain diagram prior to unloading (compression negative).
the strains. The results further reinforce the trends discussed
previously that the neutral axis depth increased over time
as a result of the creep and shrinkage deformations of the
concrete. Because the TDConcrete models generally under-
estimated the amount of cracking in the beams, the neutral
axis depth predictions were consistently greater than the
measured results. The discrepancies were smaller for
beams with R = 100%, which may have been because of
the increased cracking in these beams, making the accurate
prediction of the extent of this cracking less important.
Although the companion creep and shrinkage cylinders
described in Knaack and Kurama (2013b, 2015a) showed
significantly greater creep strains than shrinkage strains,
the components of the total strain from the beam analyses
showed that the shrinkage strains had a very significant effect
on the beam deflections, which was a finding that could not
be obtained from the experimental data (it was not possible
to separate the measured beam deflections into creep and
shrinkage components). This is because the companion
creep cylinders were loaded to 40% of the concrete compres-
sion strength fc′, while the compression stresses in the beams
generally did not reach this level. For example, Fig. 12(a)
shows the extreme (that is, top) concrete compression fiber
stress at the midspan of beam CC-50-7 over the duration
of the analysis. A maximum concrete compression stress
of approximately 0.22fc′ was reached at the initial place-
ment of the superimposed load at t0 = 7 days, with a signif-
icant reduction in concrete stress (due to redistribution of
stresses into the reinforcing steel) at later times during the
ACI Structural Journal/January 2018
analysis. Looking at Fig. 12(b) for the extreme compression
fiber strain components at midspan, it can be seen that the
shrinkage strain was greater than the creep strain before the
application of the superimposed load and was similar to the
creep strain during the later stages of loading.
The corresponding strain distributions over the beam
depth at midspan are shown in Fig. 12(c) and 12(d) for
times immediately following the application of the super-
imposed load at t = t0 = 7 days and immediately prior to the
removal of the load at t = 126 days, respectively. The results
again show that the shrinkage strain component had a very
large effect on the strains over the beam cross section. With
this significant effect of shrinkage on the concrete defor-
mations, the assumption of uniform shrinkage throughout
the concrete volume and the uncertainty/variability in the
measured shrinkage data from the accompanying concrete
cylinders could have affected the deflection predictions
of the beam specimens. However, considering that a good
level of agreement was obtained between the analytical and
measured beam deflections in Fig. 7 to 9, it is concluded
that the models still provided reasonable estimates of the
shrinkage strains.
Monte Carlo analysis of beam deflections
As stated previously, the concrete property inputs in
the TDConcrete analyses of the beams were from average
companion material test data. Because there was significant
variability in this material test data, the average concrete
properties used to model the beams could have been signifi-
187
Fig. 13—Monte Carlo simulation results: (a) UT-0-7; (b) UT-100-28; (c) UC-50-7; 
(d) UC-100-7; (e) CC-0-28; and (f)
CC-100-28.
cantly different from the actual in-place beam properties. To
investigate this effect, a Monte Carlo study was conducted
by assuming the following concrete properties to be random
variables: tension strength ft′, secant modulus of elasticity
Ec, ultimate creep strain ϵcr,u, and ultimate shrinkage strain
ϵsh,u. All other concrete properties and all steel properties
were still based on the average measured values from each
set of companion material test data. The mean and standard
deviation for the selected random variables were also deter-
mined from the material test data for each beam (Knaack
and Kurama 2013b). A Kolmogorov-Smirnov (K-S) test
(Mendenhall and Sincich 2016) was conducted to ensure the
normality of each data set (ft′, Ec, ϵcr,u, and ϵsh,u).
By varying ft′, Ec, ϵcr,u, and ϵsh,u within their probability
distributions based on the calculated mean and standard
deviation values, each test beam was simulated 200 times
to estimate the range of possible deflections. The results
of these simulations for six selected beams are shown in
Fig.  13, along with the respective measured deflections.
These beams are among the ones with the greatest discrep-
ancies as compared to the original analytical models. The
results show that except for Beam UT-0-7, the measured
beam deflections fell within the bands created by the Monte
Carlo simulations, thus supporting the hypothesis that the
inherent variability in the concrete material properties could
have contributed significantly to the differences between the
analytical and measured results in Fig. 7 through 9. As stated
previously, there was no creep or shrinkage data available
for Beam UT-0-7 and the use of creep and shrinkage data
measured for other beams was the best option available, likely
resulting in the greater deflection discrepancies. Overall,
188
through these validations, it is concluded that the new time-
dependent concrete material model, TDConcrete, and the Creep
Analysis developed and implemented in OpenSees can be used
to predict the time-dependent behavior of reinforced concrete
structural members, including structures with recycled concrete
aggregates, in a variety of loading scenarios.
SUMMARY AND CONCLUSIONS
As a means of analyzing the behavior of reinforced
concrete structures subjected to sustained service-load condi-
tions, a new time-dependent concrete fiber material model
was implemented in the open-source OpenSees structural
analysis framework. The model was validated using data
from previous research as well as experimental data from
a series of RCA concrete beam specimens. The important
findings from the investigation are summarized as follows.
1. The new time-dependent concrete material model was
able to predict the long-term service-load deformations
of reinforced concrete structural members in a variety of
loading scenarios. The initial (instantaneous) deflections
were generally underestimated because of underestimations
in the extent of cracking.
2. The time-dependent concrete material model was also
able to predict the downward shift in the beam neutral axis
over time as a result of creep and shrinkage strains. However,
even for the initial load stage, the tension strains were under-
estimated due to the inability of the model to accurately
predict the amount of cracking.
3. The Monte Carlo investigations showed that the
inherent variability in the concrete material properties may
ACI Structural Journal/January 2018
have contributed significantly to the differences between the
analytical and measured beam deflections.
4. The analyses were consistent with the experimental
results that increased amounts of RCA resulted in increased
beam deflections. Also consistent with the experimental
results, RCA had a greater effect on the deflections of beams
with less cracking.
5. The time-dependent material model showed that the
shrinkage strains had a large effect on the long-term deflec-
tions of the beams that were tested, which was a finding that
could not be obtained from the experimental measurements.
6. While further research is certainly needed, a prelimi-
nary expectation from this paper is that RCA concrete may
be suitable for use in structural applications as long as tools
are available to quantify its effects, such as the increased
service-load deflections investigated in this paper.
7. While the paper focuses on applications with RCA
concrete, the new modeling capability in OpenSees is
a general purpose tool that can be used to analyze the
time-dependent axial-flexural deformations of conventional
reinforced concrete structures as well.
AUTHOR BIOS
ACI member Adam M. Knaack is a Project Engineer at Schaefer in Cincin-
nati, OH. He received his MS and PhD in civil engineering from the Univer-
sity of Notre Dame, Notre Dame, IN. He is a member of ACI Committee 435,
Deflection of Concrete Building Structures, and Joint ACI-ASCE Committee
421, Design of Reinforced Concrete Slabs. His research interests include
the design of concrete structures and use of recycled materials in concrete.
ACI member Yahya C. Kurama is a Professor of civil engineering at
the University of Notre Dame. He is a member of ACI Committees 374,
Performance-Based Seismic Design of Concrete Buildings, and 555,
Concrete with Recycled Materials, and Joint ACI-ASCE Committee 550,
Precast Concrete Structures. His research interests include the behavior
and design of reinforced concrete structures under extreme loading from
earthquake and fire, sustainable construction, and precast/prestressed
concrete structures.
ACKNOWLEDGMENTS
Partial funding for this work was provided by the National Science Foun-
dation (NSF) under Grant No. CMMI 1436758. The support of current NSF
Program Director Y. G. Hsuan and former Director K. I. Mehta is grate-
fully acknowledged. The authors thank D. Atkinson of Concrete Recycling
Center, South Bend, IN, and M. Zeltwanger of American Mobile Aggregate
Crushing, Nappanee, IN, for their help in acquiring the RCA. Additional
materials were provided by Buzzi Unicem, Sika Corporation, and Transit
Mix South Bend. Any opinions, findings, conclusions, and/or recommenda-
tions in the paper are those of the authors and do not necessarily represent
the views of the individuals or organizations acknowledged.
REFERENCES
ACI Committee 209, 1992, “Prediction of Creep, Shrinkage, and
Temperature Effects in Concrete Structures (ACI 209R-92),” American
Concrete Institute, Farmington Hills, MI, 47 pp.
ACI Committee 318, 2014, “Building Code Requirements for Struc-
tural Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 519 pp.
Ajdukiewicz, A., and Kliszczewicz, A., 2011, “Long-Term Behavior of
Reinforced-Concrete Beams and Columns Made of Recycled Aggregate
Concrete,” fib Symposium, Prague, Czech Republic, pp. 479-482.
Bacinskas, D.; Kaklauskas, G.; Gribniak, V.; Sung, W. P.; and Shih,
M. H., 2012, “Layer Model for Long-term Deflection Analysis of Cracked
Reinforced Concrete Bending Members,” Mechanics of Time-Dependent
Materials, V. 16, No. 2, pp. 117-127. doi: 10.1007/s11043-011-9138-9
Balevičius, R., 2010, “An Average Stress Strain Approach to Creep
Analysis of RC Uncracked Elements,” Mechanics of Time-Dependent
Materials, V. 14, No. 1, pp. 69-89. doi: 10.1007/s11043-009-9093-x
ACI Structural Journal/January 2018
Bažant, Z., and Wu, S., 1973, “Dirichlet Series Creep Function for Aging
Concrete,” Journal of the Engineering Mechanics Division, ASCE, V. 99,
No. 4, pp. 367-387.
CEN, 2004, “Eurocode 2: Design of Concrete Structures – Part 1-1:
General Rules and Rules for Buildings (EN 1992-1-1),” Comité Européen
de Normalisation, Brussels, Belgium, 225 pp.
Choi, W., and Yun, H., 2013, “Long-Term Deflection and Flexural
Behavior of Reinforced Concrete Beams with Recycled Aggregate,”
Materials & Design, V. 51, pp. 742-750. doi: 10.1016/j.matdes.2013.04.044
Chong, K. T.; Foster, S. J.; and Gilbert, R. I., 2008, “Time-Depen-
dent Modelling of RC Structures Using the Cracked Membrane Model
and Solidification Theory,” Computers & Structures, V. 86, No. 11-12,
pp. 1305-1317. doi: 10.1016/j.compstruc.2007.08.005
de Borst, R., and van den Boogaard, A., 1994, “Finite-Element Modeling
of Deformation and Cracking in Early-Age Concrete,” Journal of Engi-
neering Mechanics, ASCE, V. 120, No. 12, pp. 2519-2534. doi: 10.1061/
(ASCE)0733-9399(1994)120:12(2519)
Espion, B., 1988, “Long-Term Sustained Loading Tests on Reinforced
Concrete Beams: A Selected Data Base,” Bulletin du Service Génie, Civil
No. 88-1, Université Libre de Bruxelles.
Ghali, A., 1993, “Deflection of Reinforced Concrete Members: A Critical
Review,” ACI Structural Journal, V. 90, No. 4, July-Aug., pp. 364-373.
Gilbert, I., and Ranzi, G., 2011, Time-Dependent Behaviour of Concrete
Structures, Spon Press, New York.
Gilbert, R. I., 1999, “Deflection Calculation for Reinforced Concrete
Structures – Why We Sometimes Get it Wrong,” ACI Structural Journal,
V. 96, No. 6, Nov.-Dec., pp. 1027-1032.
Jiang, H., and Kurama, Y., 2010, “Analytical Modeling of Medium-Rise
Reinforced Concrete Shear Walls,” ACI Structural Journal, V. 107, No. 4,
July-Aug., pp. 400-410.
Kabir, A., 1977, “Nonlinear Analysis of Reinforced Concrete Panels,
Slabs, and Shells for Time-Dependent Effects,” PhD dissertation, Univer-
sity of California, Berkeley, Berkeley, CA.
Kaklauskas, G., 2004, “Flexural Layered Deformational Model of Rein-
forced Concrete Members,” Magazine of Concrete Research, V. 56, No. 10,
pp. 575-584. doi: 10.1680/macr.2004.56.10.575
Kawano, A., and Warner, R., 1996, “Model Formulations for
Numerical Creep Calculations for Concrete,” Journal of Struc-
tural Engineering, ASCE, V. 122, No. 3, pp. 284-290. doi: 10.1061/
(ASCE)0733-9445(1996)122:3(284)
Knaack, A., and Kurama, Y., 2013a, “Design of Concrete Mixtures with
Recycled Concrete Aggregates,” ACI Materials Journal, V. 110, No. 5,
Sept.-Oct., pp. 483-493.
Knaack, A., and Kurama, Y., 2013b, “Sustainable Concrete Structures
using Recycled Concrete Aggregate: Short-Term and Long-Term Behavior
Considering Material Variability,” Structural Engineering Research Report
#NDSE-2013-01, University of Notre Dame, Notre Dame, IN.
Knaack, A., and Kurama, Y., 2015a, “Creep and Shrinkage of Normal
Strength Concrete with Recycled Concrete Aggregates,” ACI Materials
Journal, V. 112, No. 3, May-June, pp. 451-462. doi: 10.14359/51687392
Knaack, A., and Kurama, Y., 2015b, “Sustained Service-Load Behavior
of Concrete Beams with Recycled Concrete Aggregates,” ACI Structural
Journal, V. 112, No. 5, Sept.-Oct., pp. 565-577.
Lapko, A., and Grygo, R., 2010, “Long Term Deflection of Recycled
Aggregate Concrete (RAC) Beams Made of Recycled Concrete,” 10th
International Conference: Modern Building Materials, Structures and
Techniques, Vilnius, Lithuania.
Li, X., 2009, “Recycling and Reuse of Waste Concrete in China Part II:
Structural Behavior of Recycled Aggregate Concrete and Engineering
Applications,” Resources, Conservation and Recycling, V. 53, No. 3,
pp. 107-112. doi: 10.1016/j.resconrec.2008.11.005
Lou, T. J.; Lopes, S. M. R.; and Lopes, A. V., 2016, “Time-Depen-
dent Behavior of Concrete Beams Prestressed with Bonded AFRP
Tendons,” Composites. Part B, Engineering, V. 97, pp. 1-8. doi: 10.1016/j.
compositesb.2016.04.070
Maekawa, K.; Soltani, M.; Ishida, T.; and Itoyama, Y., 2006, “Time-
Dependent Space-Averaged Constitutive Modeling of Cracked
Reinforced Concrete Subjected to Shrinkage and Sustained Loads,” Journal
of Advanced Concrete Technology, V. 4, No. 1, pp. 193-207. doi: 10.3151/
jact.4.193
Malešev, M.; Radonjanin, V.; and Marinkovic, S., 2010, “Recycled
Concrete as Aggregate for Structural Concrete Production,” Sustainability,
V. 2, No. 5, pp. 1204-1225. doi: 10.3390/su2051204
Marí, A. R.; Bairan, J. M.; and Duarte, N., 2010, “Long-Term Deflections in
Cracked Reinforced Concrete Flexural Members,” Engineering Structures,
V. 32, No. 3, pp. 829-842. doi: 10.1016/j.engstruct.2009.12.009
Maruyama, I.; Sogo, M.; Sogabe, T.; Sato, R.; and Kawai, K., 2004,
“Flexural Properties of Reinforced Recycled Concrete Beams,” Interna-
189
tional RILEM Conference on the Use of Recycled Materials in Building
Structures, Barcelona, Spain, pp. 526-535.
Mazzotti, C., and Savoia, M., 2009, “Long-Term Deflection of Rein-
forced Self-Consolidating Concrete Beams,” ACI Structural Journal,
V. 106, No. 6, Nov.-Dec., pp. 772-781.
Mendenhall, W., and Sincich, T., 2016, Statistics for Engineering and
the Sciences, sixth edition, CRC Press, Taylor & Francis Group, LLC, Boca
Raton, FL.
Mercan, B.; Schultz, A. E.; Stolarski, H. K.; and Magana, R. A.,
2013, “Long-Term Lateral Deflection of Precast, Prestressed Concrete
Spandrel Beams,” PCI Journal, V. 58, pp. 93-115. doi: 10.15554/
pcij.09012013.93.115
Neville, A., 2011, Properties of Concrete, fifth edition, Prentice Hall,
Upper Saddle River, NJ.
Nie, J., and Cai, C. S., 2000, “Deflection of Cracked RC Beams under
Sustained Loading,” Journal of Structural Engineering, ASCE, V. 126,
No. 6, pp. 708-716. doi: 10.1061/(ASCE)0733-9445(2000)126:6(708)
OpenSees, 2006, “The Open System for Earthquake Engineering Simu-
lation,” Pacific Earthquake Engineering Research Center, Berkeley, CA,
http://opensees.berkeley.edu. (last accessed Nov. 29, 2017)
Rodriguez-Gutierrez, J. A., and Aristizabal-Ochoa, J. D., 2007, “Short-
and Long-Term Deflections in Reinforced, Prestressed, and Composite
Concrete Beams,” Journal of Structural Engineering, ASCE, V. 133, No. 4,
pp. 495-506. doi: 10.1061/(ASCE)0733-9445(2007)133:4(495)
Sato, R., Maruyama, I., Sogabe, T., Sogo, M., 2007, “Flexural Behavior
of Reinforced Recycled Concrete Beams,” Journal of Advanced Concrete
Technology, V. 5, No.1, pp. 43-61.
190
Scanlon, A., Murray, D.W., 1974, “Time-Dependent Reinforced-
Concrete Slab Deflections,” Journal of the Structural Division, ASCE,
V. 100, No. 9, pp. 1911-1924.
Sousa, C.; Sousa, H.; Neves, A. S.; and Figueiras, J., 2012, “Numer-
ical Evaluation of the Long-Term Behavior of Precast Continuous Bridge
Decks,” Journal of Bridge Engineering, ASCE, V. 17, No. 1, pp. 89-96. doi:
10.1061/(ASCE)BE.1943-5592.0000233
Tamai, S.; Shima, H.; Izumo, J.; and Okamura, H., 1988, “Average
Stress-Strain Relationship in Post Yield Range of Steel Bar in Concrete,”
Concrete Library of JSCE, V. 11, pp. 117-129.
Taucer, F.; Spacone, E.; and Filippou, F., 1991, “A Fiber Beam-Column
Element for Seismic Response Analysis of Reinforced Concrete Struc-
tures,” Report No. UCB/EERC-91/17, Earthquake Engineering Research
Center, University of California, Berkeley, Berkeley, CA.
Taylor, R.; Pister, K.; and Goudreau, G., 1970, “Thermomechanical
Analysis of Viscoelastic Solids,” International Journal for Numer-
ical Methods in Engineering, V. 2, No. 1, pp. 45-59. doi: 10.1002/
nme.1620020106
Teng, S., and Branson, D., 1993, “Initial and Time-Dependent Defor-
mation of Progressively Cracking Nonprestressed and Partially Prestressed
Concrete Beams,” ACI Structural Journal, V. 90, No. 5, Sept.-Oct.,
pp. 480-488.
Washa, G., and Fluck, P., 1952, “Effect of Compressive Reinforcement
on the Plastic Flow of Reinforced Concrete Beams,” ACI Journal Proceed-
ings, V. 49, No. 10, Oct., pp. 89-108.
Zienkiewicz, O. C.; Watson, M.; and King, I. P., 1968, “A Numer-
ical Method of Visco-Elastic Stress Analysis,” International
Journal of Mechanical Sciences, V. 10, No. 10, pp. 807-827. doi:
10.1016/0020-7403(68)90022-2
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S15
Carbon Fiber-Reinforced Polymer-Strengthened Reinforced
Concrete Beams Subjected to Differential Settlement
by Yail J. Kim and Aiham Al-Kubaisi
This paper presents the behavior of reinforced concrete beams
strengthened with carbon fiber-reinforced polymer (CFRP)
composites subjected to differential settlement. Two types of
strengthening schemes—externally bonded (EB) and near-surface-
mounted (NSM) CFRP—are tested with various support condi-
tions, and their performance is compared against unstrengthened
counterparts. The properties of soils are identified by sieve analysis
and load-settlement relationships, including a grain distribution
and subgrade reaction moduli with drained and undrained condi-
tions. The post-yield response and energy dissipation characteris-
tics of the beams are influenced by the differential settlement. The
ductility of the strengthened beams is less susceptible to settlement
in comparison with the unstrengthened ones. Regardless of support
conditions, the NSM beams exhibit higher ductility than the EB
beams. The development of curvature along the loading span is
related to the extent of the differential settlement, which is associ-
ated with the occurrence of CFRP debonding and failure modes.
Functional requirements for the strengthened beams are evaluated
based on angular distortion limits. Moment redistribution mech-
anisms induced by the differential settlement are elaborated by
analytical modeling in conjunction with parametric investigations.
Keywords: carbon fiber-reinforced polymer (CFRP); differential settle-
ment; externally bonded; near-surface-mounted (NSM); strengthening.
INTRODUCTION
The condition of soil strata is important for structural
design because their deformation can result in unfavorable
distress—namely, deflections and stresses. Excessive settle-
ment degrades the functionality of constructed facilities from
serviceability and strength standpoints. Differential settle-
ment is critical to structural members compared with uniform
settlement. Differential settlement can also be engaged with
nonstructural concerns such as unbalanced building floors
and cracking architectural elements. The economic impact
of settlement-induced structural damage is more than double
that associated with natural hazards such as floods, earth-
quakes, and hurricanes.1 The movement of soils is typically
categorized into two stages: immediate settlement and long-
term consolidation. Because the behavior of soils is complex
and thus hard to predict (especially when groundwater is
involved), undesirable settlement often disrupts the opera-
tion of constructed structures. Consequently, the subject of
differential settlement influencing structural behavior has
been of interest over several decades.2-5
When the flexural capacity of a reinforced concrete
member is insufficient to carry structural loads, it may be
strengthened using carbon fiber-reinforced polymer (CFRP)
composites. Two methods are broadly employed: 1) CFRP
sheets are externally bonded along a concrete substrate (EB
strengthening); and 2) CFRP strips are inserted into precut
ACI Structural Journal/January 2018
TECHNICAL PAPER
grooves of a concrete soffit and adhesively bonded (near-sur-
face-mounted [NSM] strengthening). These are well accepted
in the rehabilitation community and have numerous advan-
tages: durable performance; high strength and modulus;
reduced maintenance costs; and prompt implementation with
minimal disruption to users.6,7 Extensive efforts have been
made to examine the performance of CFRP-strengthened
concrete members in various detrimental environments such
as fatigue,8 aggressive weather,9 earthquake,10 and fire.11
However, little is known about the implications of differential
settlement for CFRP-strengthened members. Accordingly,
existing design guidelines do not offer any technical discus-
sions or provisions related to this potential hazard. Given that
differential settlement can affect the behavior of the strength-
ened structures, the consequences of such settlement should
be elucidated to advance the state of the art of CFRP applica-
tions and, eventually, to develop design guidelines.
In this paper, reinforced concrete beams strengthened with
EB and NSM CFRP are tested under various boundary condi-
tions to examine the effects of differential settlement that can
happen after the beams are strengthened. The focus of the
present investigation is on the consequences of vertical settle-
ment of soils supporting the flexural members, rather than
lateral soil movement, including ductility, strain development,
failure characteristics, angular distortion, and moment redis-
tribution. Salient parameters are identified and further used to
analytically expand the experimental findings.
RESEARCH SIGNIFICANCE
The current practice of CFRP strengthening does not
consider the movement of foundations, which is crucial for
structures built in regions where unstable soil strata are in
existence. Although ACI 440.2R-0812 is a comprehensive
document on CFRP-based rehabilitation, it does not provide
any guidance on strengthened members distressed by differ-
ential settlement. This research aims to address the identi-
fied knowledge gap and to establish a foundation for design
recommendations by understanding the previously unknown
soil structure interaction mechanism and its implications.
EXPERIMENTAL PROGRAM
Materials
Concrete with a specified compressive strength of fc′
= 20 MPa (2900 psi) was mixed with a maximum aggre-
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2017-028, doi: 10.14359/51700986, received January 28, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
191
Fig. 1—Beam details: (a) cross section and strengthening schemes; (b) steel cage and preparation; and (c) test setup. (Note:
Units in mm; 1 mm = 0.0394 in.)
gate size of 19 mm (0.75 in.). In accordance with ASTM
C39/C39M,13 an average compressive strength of 21 MPa
(3045 psi) was obtained at 28 days. The longitudinal steel
reinforcement used was No. 3 deformed bars (As = 71 mm2
[0.11 in.2], where As is the cross-sectional area of each bar),
while the shear reinforcement was No. 2 plain bars (As =
32 mm2 [0.05 in.2]) placed at a spacing of 75 mm (3 in.) on
center (Fig. 1(a)). The nominal yield strengths of these bars
were 414 and 250 MPa (60 and 36 ksi), respectively. Two
types of composites were selected as strengthening mate-
rials: CFRP sheets and strips. The sheets required a wet layup
process with an epoxy adhesive, whereas the strips were
prefabricated. According to the manufacturer, the tensile
strength and modulus of the CFRP sheets are 3800 MPa
(550 ksi) and 227 GPa (32,900 ksi), respectively, in conjunc-
tion with an equivalent fiber thickness of 0.165 mm
(0.0065 in.). The epoxy adhesive comprises a resin and
a hardener mixed at a mass ratio of 3:1 to achieve a yield
strength of 54 MPa (7800 psi) with an elastic modulus of
3 GPa (435 ksi). The tensile strength and modulus of the
CFRP strip (16 mm [0.63 in.] wide by 2 mm [0.079 in.] thick)
are 2068 MPa (300 ksi) and 124 GPa (18,000 ksi), respec-
tively. As shown in Fig. 1(b), sand was sieved to determine a
grain distribution and was oven-dried to eliminate moisture
(details to be discussed).
Beam details
Concrete beams were cast with singly and doubly rein-
forced sections for the hinged and fixed supports, as shown
in Fig. 1(a). The top steel bars were sufficiently extended
over 725 mm (29 in.) to preclude the premature cracking
failure of the beam in negative bending (the inflection point
was 538 mm [21 in.] from the fixed end). The beams were
classified into three groups, depending upon strengthening
schemes: unstrengthened (UB); externally-bonded CFRP
sheets (EB); and near-surface-mounted CFRP strips (NSM).
For comparison of the strengthened beams’ behavior, the
axial rigidity of the CFRP sheet and strip (EfAf, where Ef
192
and Af are the elastic modulus and cross-sectional area of
the CFRP, respectively) was intended to be the same. The
substrate of the EB beams was prepared with an electric
grinder to improve bond between the CFRP and the concrete
(Fig. 1(b)), and a single layer of CFRP sheet was adhered
using the two-part epoxy. The groove of the NSM beams
(13 mm [0.5 in.] wide by 25 mm [1 in.] deep) was designed
as per ACI 440.2R-08,12 and was created by mounting
expanded polystyrene foam underneath the steel cage prior
to casting the concrete. After roughening the groove surfaces
of the hardened beams with the grinder, a CFRP strip was
inserted with the epoxy adhesive for permanent bonding.
The EB and NSM beams with a strengthened length of
900 mm (35 in.), as shown in Fig. 1(c), were cured at room
temperature for at least 7 days in compliance with the manu-
facturer’s recommendation. The identification code of these
test beams enumerated in Table 1 indicates strengthening
types (UB, EB, and NSM), beam numbers (1 to 12), and
support conditions (F is fixed, R is rigid, DS is dry sand, and
WS is wet sand). Because the behavior of soil particles is
often uncontrollable, three strengthened beams were repeat-
edly tested per category.
Soil test
Two kinds of soils were tested: dry sand and wet sand.
The aforementioned oven-dried sand represented the
dry sand category. For the wet sand, an experimental
parametric study was conducted to identify a critical
soil condition when water was involved, which led to
a maximum differential settlement effect. Three water
contents of 15%, 20%, and 25% of the soil mass were
taken into account in drained and undrained conditions
(Fig. 2(a)). A load cell and two linear potentiometers
were employed to establish a relationship between the
applied pressure and soil settlement, so that a modulus of
subgrade reaction (representing the stiffness of each soil
type) was attained.
ACI Structural Journal/January 2018
Table 1—Identification of test beams
Py, kN Pu, kN
Beam CFRP Soil Each Avgerage Each Avgerage Failure sequence
UB1-F-R None None 41.5 41.5 49.3 49.3 FC → SC → CC
EB2-F-R CFRP sheet None 45.2 53.8 FC → SC → SCT (H)
EB3-F-R CFRP sheet None 49.9 47.6 56.4 56.4 FC → SC → CD (F)
EB4-F-R CFRP sheet None 47.9 59.1 FC → SC → CD (F)
UB5-F-DS None Dry sand 44.6 44.6 48.5 48.5 FC → SC → CC
EB6-F-DS CFRP sheet Dry sand 53.4 59.7 FC → SC → CD (F)
EB7-F-DS CFRP sheet Dry sand 46.4 49.8 63.2 60.8 FC → SC → SCT (H)
EB8-F-DS CFRP sheet Dry sand 49.7 59.7 FC → SC → SCT (F)
UB9-F-WS None Wet sand 44.9 44.9 54.9 54.9 FC → SC → CC
EB10-F-WS CFRP sheet Wet sand 58.9 71.8 FC → SC → CC
EB11-F-WS CFRP sheet Wet sand 56.1 56.1 63.6 67.0 FC → SC → SCT (F)
EB12-F-WS CFRP sheet Wet sand 53.4 65.7 FC → SC → SCT (F)
NSM2-F-R CFRP strip None 51.9 62.5 FC → SC → CD (H)
NSM3-F-R CFRP strip None 53.7 53.9 67.9 67.9 FC → SC → CD (F)
NSM4-F-R CFRP strip None 56.2 73.4 FC → SC → CD (H)
NSM6-F-DS CFRP strip Dry sand 51.2 71.8 FC → SC → CD (F)
NSM7-F-DS CFRP strip Dry sand 51.4 52.7 61.3 66.3 FC → SC → SCT (F)
NSM8-F-DS CFRP strip Dry sand 55.6 65.9 FC → SC → CD (F)
NSM10-F-WS CFRP strip Wet sand 53.5 66.1 FC → SC → SCT (F)
NSM11-F-WS CFRP strip Wet sand 54.1 55.1 63.4 64.8 FC → SC → SCT (F)
NSM12-F-WS CFRP strip Wet sand 57.9 64.9 FC → SC → SCT (F)

Notes: FC is flexural crack; SC is shear crack; CC is concrete crushing; SCT is shear cracked initiated at CFRP termination; CD is CFRP debonding with concrete cover; (H) is CD
near hinge support; (F) is CD near fixed support; 1 kN = 0.225 kip.

Fig. 2—Test setup: (a) soil test; (b) flexural test (fixed rigid); (c) flexural test (fixed settlement); (d) linear potentiometer; (e) PI
gauges; and (f) saturated soil and linear potentiometer.
Test setup and instrumentation configuration was suitable to examine the effects of differen-
The default boundary condition of the beams was one-end- tial settlement. To avoid beam rotation at the fixed end, two
fixed with a hinge on the other side (Fig. 1(c)). This test hollow square steel sections (100 x 100 mm [4 x 4 in.]) were
positioned, which were connected to the rigid steel support

ACI Structural Journal/January 2018 193

Fig. 3—Soil pressure and settlement: (a) drained sand (D); and (b) undrained sand (UD). (Note: 1 kPa = 0.145 psi; 1 mm =
0.0394 in.)
using high-strength threaded rods (19 mm [0.75 in.] in diam-
eter). A steel plate (100 mm [4 in.] wide by 5 mm [0.2 in.] thick)
was placed on top of the hinge to mitigate concrete crushing
caused by stress concentration. Figure 2(b) exhibits the fixed
rigid (F-R) condition of an EB beam. For the soil-supported
hinges (F-DS and F-WS test groups), a circular steel container
with a diameter of 740 mm (29 in.) was used to accommodate
a soil depth of 450 mm (18 in.) without boundary effects, as
shown in Fig. 1(c) and 2(c). As will be discussed in a later
section, the undrained soil strata allowed more deflection than
the drained cases; hence, a membrane layer was placed inside
the container before pouring the oven-dried sand and water
for the F-WS beams. For each test, the soils were positioned
anew in the container.
A load cell, a linear potentiometer, and two displacement-type
strain transducers (so-called PI gauges) were placed at
midspan of each beam to measure the flexural behavior
(Fig. 2(d) and (e)). Two linear potentiometers at 180 degrees
apart were also positioned on top of the hinge support to
monitor soil settlement (Fig. 2(f)). In addition, six strain
gauges were bonded along the EB CFRP sheet at a typical
distance of 150 mm (6 in.), as illustrated in Fig. 1(c). It
is worth noting that strain gauges were not bonded to the
NSM CFRP strip because the placement of the gages might
influence the bond between the strip and adhesive inside the
narrow groove. All beams were monotonically loaded in
four-point bending to failure, and the behavior was recorded
by a computerized data acquisition system.
TEST RESULTS
Soil properties
The pressure-settlement relationship of the sand is plotted
in Fig. 3 (average of the two linear potentiometers is given),
including grain size variation acquired from the sieve analysis.
The modulus of subgrade reaction K of each test condition
was taken within a linear settlement range between 5 and
10 mm (0.2 and 0.4 in.). For the drained sand (Fig. 3(a)),
all soil responses were similar up to a pressure of approxi-
mately 100 kPa (15 psi), regardless of water content. As the
pressure increased, the soil specimen with a water content of
25% began to deviate from those with lower water contents.
The reason is that the soil stratum of the 25% water case
preserved a certain amount of residual water in the voids,
which allowed additional settlement in comparison with the
194
other cases that were fully drained. The K-value of the 25%
water sand was, accordingly, 34% and 40% lower than those
of the sand with 15% and 20% water contents, respectively.
In the case of the undrained sand (Fig. 3(b)), the specimen
with a 15% water content showed hardening after a pressure
of 30 kPa (4 psi), which was not observed in its counterparts
involving 20% and 25% water contents. The decreasing trend
of the reaction modulus was proportional to the amount of
water kept inside the soil voids. The undrained sand with a
water content of 25% (K = 8.4 MPa/m [30 psi/in.]) exhibited
more settlement than others; therefore, this soil condition
was selected for structural testing. The average modulus of
subgrade reaction of the dry sand was determined to be K =
13.7 MPa/m (49 psi/in.).
Load displacement
Unstrengthened beams—Figure 4(a) shows the load-
displacement behavior of the unstrengthened beams subjected
to various boundary conditions. The response of the beam
tested at fixed rigid supports (UB1-F-R) was that of a typical
under-reinforced concrete beam (yield load = 41.5 kN
[9.3 kip] and ultimate load = 49.3 kN [11.1 kip], as listed in
Table 1). The behavior of the beam with fixed dry soil supports
(UB5-F-DS) was similar to the behavior of UB1-F-R up to
a load of 15.8 kN (3.6 kip)—32% of the UB1-F-R beam’s
capacity, beyond which a bifurcation was observed owing
to the onset of soil softening. Although there was no notice-
able difference between the capacities of the UB1-F-R and
UB5-F-DS beams, the displacement of the latter at failure
was 33% more than that of the former. Additionally, in
comparison with the UB1-F-R beam, the UB5-F-DS beam
accompanied a shorter range of post-yield displacement
until failure. This is explained by the fact that, because of
the settlement of the dry soil stratum, the UB5-F-DS beam
substantially deflected before its yield load; consequently,
the beam’s available deflection was restricted. The ultimate
load and corresponding displacement of the beam with wet
soil (UB9-F-WS) were higher than those of the other beams.
The post-yield plateau of UB9-F-WS was not observed
(the wet sand dissipated the applied stresses in the beam),
which confirms the trend of the foregoing reduced post-yield
displacement range. Energy dissipated until the peak loads
of the beams was obtained by numerically integrating the
area under the load-displacement curves (Fig. 4(a), inset).
ACI Structural Journal/January 2018
Fig. 4—Load displacement of EB beams: (a) unstrengthened beams; (b) fixed rigid; (c) fixed dry sand; and (d) fixed wet sand.
(Note: 1 kN = 0.225 kip; 1 mm = 0.0394 in.)
The energy of UB5-F-DS and UB9-F-WS was 9.5% and
63.6% higher than the energy of UB1-F-R, respectively.
EB beams—The behavior of the EB-strengthened beams
is shown in Fig. 4(b) to (d), with a comparison against their
unstrengthened counterparts (for clarity, test results of one
representative beam per category are presented). The post-
yield stiffness of EB4-F-R was markedly higher than that
of UB1-F-R (Fig. 4(b)), and the failure modes of these
beams were distinguishable (that is, brittle for EB4-F-R
and ductile for UB1-F-R, which is typical in this kind of
comparison). A difference of 31.7% in average energy was
observed between these two categories (Fig. 4(b), inset). The
responses of EB7-F-DS and UB5-F-DS were analogous to
those of EB4-F-R and UB1-F-R, as shown in Fig. 4(c), except
for the divergence of the displacement at a lower load of
20.5 kN (4.6 kip). The regional peak loads of EB7-F-DS
after yielding were attributable to the progression of shear
cracks initiated at the CFRP termination point (detailed
failure modes are elaborated in a later section). Unlike the
rigid support case (F-R), the average energy of the EB beams
with dry sand (F-DS) was 49.8% higher than that of the
unstrengthened beam (Fig. 4(c), inset). The soil settlement
of the EB F-DS beams alongside the improved load-carrying
capacity owing to CFRP strengthening was responsible for
such energy dissipation. Figure 4(d) illustrates the load-dis-
placement of an EB beam supported by wet sand. The
strengthened EB10-F-WS beam failed at a 30.8% higher
load than the unstrengthened one in conjunction with more
displacement. The EB F-WS beams revealed almost double
energy compared with the UB F-WS beam: 1443 kN·mm
(12.8 kip-in.) versus 723 kN·mm (6.4 kip-in.), respectively,
on average. It is worth noting that, although the soils were
ACI Structural Journal/January 2018
placed with care, their unpredictable nature influenced the
energy dissipation of the test beams.
NSM beams—Figures 5(a) to 5(c) reveal the behavior of the
beams with NSM CFRP strips. The strengthened beam with
fixed rigid supports (NSM4-F-R) exhibited a precipitously
increasing slope up to the yielding of the steel reinforce-
ment (Fig. 5(a)), followed by an abrupt load drop at 73.4 kN
(16.5 kip). The average energy of the NSM F-R beams was
4.8% higher than that of the unstrengthened UB1-F-R beam,
which is attributable to the NSM beams’ increased flexural
capacity despite their brittle failure. The flexure of those with
fixed dry sand and fixed wet sand supports (NSM8-F-DS
and NSM11-F-WS in Figs. 5(b) and 5(c), respectively) was
similar to the case of the NSM4-F-R beam; however, because
of the soil settlement, the former’s stiffness was less than the
latter’s. The load-carrying capacity of the NSM beams tended
to decrease with an increase in settlement, as shown in
Table 1. For instance, the average failure loads of the NSM
F-R, F-DS, and F-WS were 67.9, 66.3, and 64.8 kN (15.3,
14.9, and 14.6 kip), respectively. This capacity-varying
propensity was opposite to the trend of the EB beams (that
is, the average capacities of 56.4, 60.8, and 67.0 kN [12.7,
13.7, and 15.1 kip] with the F-R, F-DS, and F-WS support
conditions, respectively). It is postulated that CFRP-bonded
area (EB versus NSM) influences a stress redistribution of the
strengthened beams, among other attributes such as failure
modes, which can alter the capacity of the beams.
Ductility
The ductility index of the tested beams (DI) was calcu-
lated using Eq. (1) and is summarized in Fig. 5(d)
195
Fig. 5—Load displacement of NSM beams: (a) fixed rigid; (b) fixed dry sand; (c) fixed wet sand; and (d) ductility comparison.
(Note: 1 kN = 0.225 kip; 1 mm = 0.0394 in.)
Fig. 6—Comparison of tensile strain at level of reinforcing bar: (a) EB beams; and (b) NSM beams.
Eu
DI = (1)
Ey
where Eu and Ey are the energy values dissipated up to the
ultimate and yield loads of the beam, respectively. The
unstrengthened beam with fixed rigid supports (UB-F-R)
revealed a ductility index of 4.8, which was higher than the
strengthened ones (EB-F-R and NSM-F-R showed DI = 2.2
and 2.4, respectively). This comparison verifies the well-
known fact that CFRP strengthening reduces the member’s
ductility.14 When differential settlement was involved,
ductility of the unstrengthened beams dwindled down
to DI = 1.5 and 1.9 for UB-F-DS and UB-F-WS, respec-
tively. The soil settlement affected the usable deflection of
the beams after yielding of the steel, as discussed earlier,
thereby decreasing the Eu component of Eq. (1). The effect
of differential settlement in the strengthened beams was not
196
as significant as in the unstrengthened beams (Fig. 5(d)).
The ductility indexes of the EB beams with dry sand (F-DS)
and wet sand (F-WS) were, respectively, 23% and 7% lower
than that of the rigid-support beam (F-R). Although similar
responses were observed, the NSM beams maintained
consistently higher ductility relative to the EB beams in all
circumstances, owing to the NSM CFRP’s enhanced bond
performance until failure.
Load strain
As shown in Fig. 6, where the tensile strain of selected
beams at midspan is summarized, the strain development
was controlled by boundary conditions. The general trend
found was that the slope of the load-strain curve became
stiffer with an increase in settlement. Namely, the degree of
bending in the vicinity of the maximum positive moment
zone was alleviated. The reason is that the soil-side portion
ACI Structural Journal/January 2018
Fig. 7—Soil pressure settlement: (a) EB beams; and (b) NSM beams. (Note: 1 kPa = 0.145 psi; 1 mm = 0.0394 in.)

Fig. 8—CFRP strain of EB beams: (a) fixed rigid (Beam EB2-F-R); (b) fixed dry sand (Beam EB7-F-DS); (c) fixed wet sand
(Beam EB12-F-WS); and (d) comparison at 50% of Beam EB2-F-R’s capacity. (Note: 1 kN = 0.225 kip.)
of the F-DS and F-WS beams shifted down like a rigid where P is the total load applied; a is the distance from the
body, unlike the other portion on the fixed side; therefore, fixed support to the adjacent loading point; b is the distance
the curvature of the beam was not uniform along the span between the two loading points; and L is the span length of
(further discussions are available in subsequent sections). the beam. As far as the EB beams are concerned (Fig. 7(a)),
The load-strain response of the beams appears to be influ- two distinct trends were noticed by soil conditions. The
enced by the strengthening methods; further testing is neces- beams with dry sand exhibited relatively stiffer responses
sary to generate conclusive observations. than those with wet sand, which aligns with the discussion
expounded in the soil property section. For the NSM beams,
Soil settlement a similar trend was recorded with less settlement in the wet
The relationship between the soil pressure and settlement sand condition (Fig. 7(b)). It should be noted that two beams
of the strengthened beams is given in Fig. 7. The soil pres- (EB5-F-DS and NSM10-F-WS) were not included in Fig. 7
sure was calculated using the reaction force at the hinged because the locally disturbed sand caused the unbalanced
support, RH (Eq. (2)), divided by the contact area to the soil displacement of the supporting block, leading to an uplift
under the steel support (Aca = 0.09 m2 [140 in.2]) measurement of the soil settlement.

P  3a (a + b)  CFRP strain
RH = 1 −  (2)
2 2L 2  Figure 8 plots strain development along the CFRP sheets
toward the hinged support (strains to the fixed support side

ACI Structural Journal/January 2018 197

Fig. 9—CFRP strain profile of EB beams: (a) fixed rigid (Beam EB2-F-R); (b) fixed dry sand (Beam EB7-F-DS); and (c) fixed
wet sand (Beam EB12-F-WS). (Note: 1 mm = 0.0394 in.)
Fig. 10—Failure mode of unstrengthened beams: (a) fixed rigid (UB1-F-R); and (b) fixed wet sand (UB5-F-WS).
are available in Fig. 9). For the fixed rigid beam (EB2-F-R in
Fig. 8(a)), the strains of Gauges H2 and H3 were similar up
to a load of 12.1 kN (2.7 kip), whereas the strain of Gauge
H1 was 67% lower than that of Gauge H2. This indicates
that the shear deformation of the adhesive layer (horizontal)
caused by the bending of the beam was yet local. When the
beam was further loaded, the slope of Gauge H3 was substan-
tially reduced, which means that CFRP debonding occurred
near midspan of the beam. Gauge H2 followed the same
trend at a load of 20.4 kN (4.6 kip), owing to the progression
of CFRP debonding. When diagonal tension cracks began
at 30 kN (6.8 kip) near the CFRP’s termination point and
propagated, the response of Gauge H1 became unstable. The
beam with fixed dry sand (EB7-F-DS) revealed a tendency
analogous to the fixed rigid beam (EB2-F-R), as shown in
Fig. 8(b); however, the strain development of Gauges H2
and H3 (EB7-F-DS) was more rapid. The initial response of
the beam with fixed wet sand (EB12-F-WS in Fig. 8(c)) was
stiffer than those of the other beams, and the strain develop-
ment was slow. In compliance with the rigid-body-like shift
explained previously, the shear span of the beam outside the
two-point loading region toward the soil did not experience
significant bending, corroborated by the low strain values of
Gauges H2 and H1 in Fig. 8(c). As summarized in Fig. 8(d),
there was an affinity between the CFRP strain near midspan
(Gauge H3) and the support condition at a typical service
state (for example, 50% of the EB2-F-R beam’s capacity).
The CFRP strains decreased when the support became softer:
the strains of the beams with dry sand (F-DS) and wet sand
(F-WS) were 13% and 71% lower than the strain of the rigid
198
beam (F-R), respectively, because of the reduced curvature
effect in tandem with the rigid-body-like shift. The strain
profiles of the CFRP are shown in Fig. 9. The profiles of the
beams with fixed rigid and fixed dry sand were generally
symmetric about midspan (Fig. 9(a) and (b), respectively).
By contrast, the profile of the beam with fixed wet sand
(Fig. 9(c)) was asymmetric, because the effect of the non-
uniform curvature was augmented by the wet sand stratum.
Failure mode
Figure 10 shows the failure modes of the unstrength-
ened beams (only selected beams are presented, while
detailed failure information of all other beams is available in
Table 1). The failure sequence of the rigid-support beam
was typical: flexural cracking, shear cracking, and concrete
crushing (Fig. 10(a)). Although the UB beams tested on soils
demonstrated the same failure sequence (Table 1), the extent
of shear damage was more pronounced (Fig. 10(b)), owing
to the settlement-induced non-uniform curvature effect (that
is, on account of the rigid-body-like shift, the beam portion
toward the fixed support experienced more damage relative
to the portion on the soil side). The failure of the strength-
ened beams is provided in Fig. 11. The EB beam with fixed
rigid supports (EB3-F-R; Fig. 11(a)) failed by end-peeling
of the CFRP accompanied by concrete-cover delamina-
tion. The failure of the EB beam with dry sand (EB8-F-DS,
Fig. 11(b)) was initiated by stress concentration at the CFRP
termination point, which was then connected to a large
diagonal tension crack, including another crack induced by
excessive negative bending. The NSM4-F-R beam failed
ACI Structural Journal/January 2018
Fig. 11—Failure mode of strengthened beams: (a) fixed rigid (EB3-F-R); (b) fixed dry sand (EB8-F-DS); (c) fixed rigid
(NSM4-F-R); and (d) fixed wet sand (NSM10-F-WS).
Fig. 12—Angular distortion of beams: (a) at angle of 0.0033; and (b) at angle of 0.0067.
by CFRP debonding near the rigid support (Fig. 11(c)),
the same as the EB3-F-R beam (Fig. 11(a)). Nonetheless,
because of the NSM CFRP’s position inside the beam, the
amount of concrete-cover delamination of NSM4-F-R was
more than that of EB3-F-R. Figure 11(d) reveals the failure
of the NSM beam with fixed wet sand (NSM10-F-WS),
including damage localization within its shear-span near
the fixed support. Unlike the fixed rigid case (NSM4-F-R),
cover delamination failure was not associated despite stress
concentration at the CFRP termination. These observations
denote that: 1) regardless of strengthening scheme, the beam
failure near the fixed support was dominant compared with
the failure near the soil side at occurrence rates of 24%:86%,
respectively, as listed in Table 1; and 2) CFRP debonding
with concrete-cover delamination was mitigated when the
strengthened beams were subjected to differential settlement
ACI Structural Journal/January 2018
on account of a stress redistribution within the beams, which
led to the shear-crack-dominated failure.
Angular distortion
The angular distortion (AD) of a beam is traditionally
defined as the ratio between differential settlement and span
length. This measurement is useful when evaluating the
performance of a flexural member with the following limits2:
AD = 0.0033 and 0.0067 for the cracking of non-load-bearing
architectural elements and structural damage in a building,
respectively. Based on the average differential settlement
of each test category, Fig. 12 shows normalized load ratios
(a load level of a beam divided by its ultimate capacity Pu)
corresponding to the angular distortion limits. At the archi-
tectural limit of AD = 0.0033, the load ratios of the UB, EB,
and NSM beams tested with dry sand (DS) were P/Pu = 0.37,
199
Fig. 13—Moment ratio with differential settlement in service condition (UB is unstrengthened beam; CFRP is strengthened
beam): (a) maximum negative moment; and (b) maximum positive moment.
0.38, and 0.24, respectively (Fig. 12(a)). When the support
P M 
condition changed to wet sand (WS), these ratios diminished M cs =  − Fs  a (9)
2 L 
by 73%, 71%, and 29%, respectively. Contrary to the UB
and EB categories in F-DS, the NSM F-DS case reached the
where MF is the fixed support negative moment of the beam;
AD limit earlier at P/Pu = 0.24, and the effect of wet sand
MF0 and MFs are the moments at the fixed support without and
altered this tendency. These observations can be explained
with differential settlement, respectively; Mc is the maximum
by the facts that: 1) the average failure load of the NSM
positive moment along the beam span; Mc0 and Mcs are the
F-DS beams was higher than those of the UB and EB beams
moments induced by the applied load P and the differential
(Table 1), resulting in the comparatively low load ratio; and
settlement ∆s, respectively, at which Mc is obtained; and EcI
2) the NSM beams’ load-carrying capacity decreased by
is the flexural rigidity of the transformed section consisting
3% with a change in soil type from dry sand to wet sand
of the concrete elastic modulus (Ec = 57,000√fc′ in psi [Ec =
(Table 1), which was different from the variation trend of
4700√fc′ in MPa]15) and the moment of inertia at an arbitrary
the other beams, as delineated previously. The load ratios of
loading stage within the elastic domain (gross and cracked).
the UB, EB, and NSM beams with dry sand at the structural
Figure 13 shows the effects of differential settlement on the
damage limit of AD = 0.0067 were P/Pu = 0.54, 0.61, and
service moment of the beams before and after cracking loads.
0.44, respectively, and their reduction rates concerned with
Because the total negative and positive moments (MF and Mc,
wet sand were reasonably preserved, as in the case of AD =
respectively) were normalized by the base components MF0
0.0033 (that is, 61%, 72%, and 20%, respectively). Further
and Mc0, as shown in the equations in Fig. 13, their ratios were
discussions on the relationship between the strengthening
maintained within the two load categories (that is, P ≤ Pcr and
scheme and differential settlement follow in a later section.
Pcr < P ≤ Py, where Pcr and Py are the cracking and yield loads
of the beam, respectively). It is worth again noting that the
Service moment due to differential settlement
transformed sectional properties of the EB and NSM beams
Fundamental structural analysis provides the following
were the same due to the identical axial rigidity of the CFRP
expressions in conjunction with the test configuration illus-
sheet and strip; hence, a unified expression “CFRP” was used
trated in Fig. 1
in Fig. 13. For the negative moment (Fig. 13(a)), the ratio of
the rigid support case (FR) was unity owing to the null contri-
MF = MF0 + MFs (3)
bution of the MFs and Mcs to the total moment. As the support
condition changed from dry sand (F-DS) to wet sand (F-WS),
Mc = Mc0 + Mcs (4)
the moment ratios before cracking increased to 2.4 and 3.3,
respectively. After cracking of the beams, the moment ratio
3Pa (a + b) was influenced by the presence of CFRP, because the contri-
MF0 = (5)
4L bution of the CFRP to the cracked beam section was not negli-
gible, unlike the case of the uncracked section for the gross
3Ec I moment of inertia. The post-cracked moment ratios of the
M Fs = ∆ s (6)
L2 strengthened beams were 8% and 9% higher than those of the
unstrengthened beams for the dry sand and wet sand condi-
RH tions, respectively. For the positive moment (Fig. 13(b)), the
∆s = (7) amounts of the increased moment ratios caused by the settle-
K s Aca
ment were not as high as those of the ratios in the negative
moment. Moreover, after cracking, the strengthened beams
Pa  3a (a + b)  with settlement revealed lower moment ratios than did the
M c0 = 1 −  (8)
2  2 L2  unstrengthened ones, opposite to the negative moment trend.
These facts signify that the positive moment caused by differ-

200 ACI Structural Journal/January 2018

Fig. 14—Effects of properties of cracked beams (UB is unstrengthened beam; CFRP is strengthened beam): (a) concrete
strength; (b) modulus of subgrade reaction; (c) elastic modulus of CFRP; and (d) cross-sectional area of CFRP. (Note:
1 MPa/m = 3.68 psi/in.; 1 MPa = 145 psi; 1 mm2 = 0.0016 in.2.)
ential settlement was redistributed to the negative moment,
when the strengthened beams cracked. It should be noted that
the moment redistribution resulting from differential settle-
ment is different from the conventional redistribution associ-
ated with the yielding of steel reinforcement.
PARAMETRIC STUDY
Formulation
Several parameters affect the negative and positive
moments of the one-end-fixed beam subjected to differential
settlement: the compressive strength of concrete, the elastic
modulus of CFRP, the number of CFRP layers, and the
modulus of subgrade reaction. Based on Eq. (3) to (9), the
effects of these parameters were studied within a practical
range, as shown in the abscissa of Fig. 14. The configura-
tion of the test beam (Fig. 1) was the default for the present
parametric investigation into moment responses with and
without CFRP strengthening. The applied load P in Fig. 1(c)
was set to represent a typical service condition (that is, 60%
of the yield load of the control beam, UB1-F-R). Because
there was no remarkable difference between the EB- and
NSM-strengthened beams in service as discussed previ-
ously, these strengthening methods were not differentiated.
Effects of parameters
The effect of compressive concrete strength fc′ is shown
in Fig. 14(a). This parameter is engaged with the elastic
modulus of the concrete Ec that influences the flexural
rigidity of the beam (Eq. (5) and (9)). With an increase in
concrete strength from 20 MPa to 50 MPa (2900 to 7250 psi)
at a differential settlement of 10.5 mm (0.41 in.) resulting
ACI Structural Journal/January 2018
from the beams’ default subgrade reaction modulus of
K = 8.4 MPa/m (31 psi/in.), the negative moment ratio of
the unstrengthened (UB) and strengthened (CFRP) beams
increased from 2.6 to 3.6 and from 2.9 to 3.9, respectively.
Although the CFRP beam revealed higher moment ratios
than did the UB beam, their response rates were identical.
These increased negative moment ratios imply that, when
differential settlement takes place, CFRP strengthening of
stiff beams (or high-strength concrete beams) exacerbates
the fixed end moment. In contrast, the positive moment
of the beams was lessened because of the aforementioned
moment redistribution mechanism. Figure 14(b) illustrates
the moment variation of the beams with variable subgrade
reaction moduli at a concrete strength of fc′ = 20 MPa
(2900 psi). As the reaction moduli rose, the negative
moment ratios were asymptotically reduced toward unity.
Given that the increased reaction modulus aligns with a
reduction in differential settlement, the extent of additional
negative moment caused by the settlement decreased and
should disappear if the reaction modulus is significantly high
like a rigid support. On the contrary, the positive moment
ratios increased with the increased reaction moduli, because
the moment redistribution mechanism was diminished. The
contribution of the CFRP’s moduli to the moment ratios is
provided in Fig. 14(c). When the CFRP became stiffer, both
the negative and positive moment ratios of the strengthened
beams deviated from the invariant ratios of the unstrength-
ened cases: the variation of the negative moment was more
pronounced than that of the positive moment. Similar obser-
vations were noted for the beams strengthened with variable
201
CFRP layers, as shown in Fig. 14(d), except that the CFRP
effect was augmented relative to the case of Fig. 14(c).
SUMMARY AND CONCLUSIONS
This experimental research has investigated the behavior
of CFRP-strengthened beams under differential settlement,
which was not well understood in the community. The effects
of strengthening schemes (EB and NSM) were comparatively
examined with various support conditions (rigid, dry sand, and
wet sand). Soil testing was conducted to study the influence of
water inside the soil strata with drained and undrained condi-
tions. Although the number of test beams was limited, consistent
observations were made in most cases. An analytical parametric
study was carried out to assess the contribution of typical engi-
neering parameters affecting the performance of the strengthened
beams. A follow-up study may be of interest with various sizes of
CFRP-strengthened structural members and connection details
subjected to differential settlement. The following is concluded:
• The amount of water inside the soil strata controlled the
subgrade reaction modulus, which influenced the extent
of settlement. The available post-yield displacement of
the unstrengthened beams until failure was reduced by
the occurrence of differential settlement when compared
with displacement of the beam tested at rigid supports.
Nonetheless, the settled beams showed higher energy
dissipation than the rigid beam. The support conditions
also affected energy dissipation characteristics of the
strengthened beams.
• The ductility of the tested beams decreased after CFRP
strengthening. The variation of ductility in the strength-
ened beams was not as susceptible to differential settle-
ment as the case of the unstrengthened beams. The
NSM beams demonstrated higher ductility than the EB
beams, irrespective of support condition.
• Differential settlement caused the rigid-body-like shift of the
beams, which disrupted the formation of uniform curvature
along the loading span. Accordingly, strain development in
the positive moment region of the beams was alleviated,
including the strains along the CFRP sheets. The onset of
CFRP debonding was mitigated by the settlement.
• Owing to the damage localization associated with diag-
onal tension cracks, most strengthened beams failed on the
fixed-end side, rather than on the soil-supported side. As far
as functional requirements were concerned, the strengthened
and unstrengthened beams reached the angular distortion
limits at lower loads because of the differential settlement.
• CFRP strengthening increased the negative moment of the
beams when subjected to differential settlement, because
the positive moment of the strengthened beams was
redistributed to the negative moment. With an increase in
concrete strength, the fixed end moment of the strength-
ened beams worsened due to the redistribution mechanism.
AUTHOR BIOS
Yail J. Kim, FACI, is a Professor in the Department of Civil Engi-
neering at the University of Colorado Denver, Denver, CO. He is Chair
of ACI Committee 345, Concrete Bridge Construction, Maintenance, and
Repair; and ACI Subcommittee 440-I, FRP-Prestressed Concrete. He is
202
a member of ACI Committees 342, Evaluation of Concrete Bridges and
Bridge Elements; 343, Concrete Bridge Design; and 440, Fiber-Reinforced
Polymer Reinforcement. His research interests include advanced composite
materials for structural rehabilitation; complex systems; uncertainty quan-
tification; and science-based structural engineering, including statistical,
interfacial, and quantum physics.
Aiham Al-Kubaisi is a former masters student in the Department of Civil
Engineering at the University of Colorado Denver. He received his BS in
civil engineering from Nahrain University, Baghdad, Iraq, in 2008. His
research interests include structural strengthening using composite mate-
rials and soil-structure interaction.
ACKNOWLEDGMENTS
The authors would like to acknowledge financial support from the Higher
Committee for Education Development in Iraq (HCED). Technical contents
presented in this paper are based on the opinion of the authors, and do not
necessarily represent that of others. Proprietary information such as product
names and manufacturers is not provided to avoid commercialism.
REFERENCES
1. Holtz, W. G., and Hart, S. S., “Home Construction on Shrinking and
Swelling Soils,” Special Publication 11, Colorado Geological Survey,
1978, 18 pp.
2. Skempton, A. W., and MacDonald, D. H., “The Allowable Settlements
of Buildings,” Proceedings — Institution of Civil Engineers, V. 5, No. 6,
1956, pp. 727-768. doi: 10.1680/ipeds.1956.12202
3. Moulton, L. K.; GangaRao, H. V.; and Halvorsen, G. T. “Tolerable
Movement Criteria for Highway Bridges,” Report No. FHWA/RD-85/107,
Federal Highway Administration, Washington, DC, 1985, 109 pp.
4. Hearn, G., and Nordheim, K., “Differential Settlements and Inelastic
Response in Steel Bridge Beams,” Transportation Research Record:
Journal of the Transportation Research Board, V. 1633, 1998, pp. 68-73.
doi: 10.3141/1633-09
5. Hebeler, G. L.; Ospina, R. I.; Howard, D. K.; and Leistikow, R.,
“Assessment and Remediation of Geotechnical and Structural Deficiencies
at a 5-Story Hotel Experiencing Differential Settlements,” Proceedings of
the 5th Congress on Forensic Engineering, 2009, pp. 278-287.
6. Bakis, C. E.; Bank, L. C.; Brown, V. L.; Cosenza, E.; Davalos, J. F.;
Lesko, J. J.; Machida, A.; Rizkalla, S. H.; and Triantafillou, T. C., “Fiber-
Reinforced Polymer Composites for Construction—State-of-the-Art
Review,” Journal of Composites for Construction, ASCE, V. 6, No. 2, 2002,
pp. 73-87. doi: 10.1061/(ASCE)1090-0268(2002)6:2(73)
7. Bank, L. C., Composites for Construction, John Wiley & Sons, Inc.,
Hoboken, NJ, 2006, 551 pp.
8. Zhang, W., and Kanakubo, T., “Local Bond Stress-Slip Relationship
between Carbon Fiber-Reinforced Polymer Plates and Concrete under
Fatigue Loading,” ACI Structural Journal, V. 111, No. 4, July-Aug. 2014,
pp. 955-966. doi: 10.14359/51686781
9. Deng, J.; Tanner, J. E.; Mukai, D.; Hamilton, H. R.; and Dolan, C. W.,
“Durability Performance of Carbon Fiber-Reinforced Polymer in Repair/
Strengthening of Concrete Beams,” ACI Materials Journal, V. 112, No. 2,
Mar.-Apr. 2015, pp. 247-257. doi: 10.14359/51687104
10. Parks, J. E.; Brown, D. N.; Ameli, M. J.; and Pantelides, C. P.,
“Seismic Repair of Severely Damaged Precast Reinforced Concrete Bridge
Columns Connected with Grouted Splice Sleeves,” ACI Structural Journal,
V. 113, No. 3, May-June 2016, pp. 615-626. doi: 10.14359/51688756
11. Williams, B.; Kodur, V.; Green, M. F.; and Bisby, L., “Fire Endurance
of Fiber-Reinforced Polymer Strengthened Concrete T-beams,” ACI Struc-
tural Journal, V. 105, No. 1, Jan.-Feb. 2008, pp. 60-67.
12. ACI Committee 440, “Guide for the Design and Construction of
Externally Bonded FRP Systems for Strengthening Concrete Structures
(ACI 440.2R-08),” American Concrete Institute, Farmington Hills, MI,
2008, 76 pp.
13. ASTM C39/C39M-16, “Standard Test Method for Compressive
Strength of Cylindrical Concrete Specimens,” ASTM International, West
Conshohocken, PA, 2016, 7 pp.
14. Spadea, G.; Swamy, R. N.; and Bencardino, F., “Strength and
Ductility of RC Beams Repaired with Bonded CFRP Laminates,” Journal
of Bridge Engineering, ASCE, V. 6, No. 5, 2001, pp. 349-355. doi: 10.1061/
(ASCE)1084-0702(2001)6:5(349)
15. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 2014, 520 pp.
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL TECHNICAL PAPER
Title No. 115-S16

Tests of Short Headed Bars with Anchor Reinforcement

Used in Beam-to-Column Joints
by Ján Bujn ák and Matúš Farbák

The proper design and execution of joints is one of the basic

preconditions for efficient functioning of precast concrete frames.
Headed reinforcement bars are an efficient technique to anchor
tensile forces in beam-to-column joints. In comparison to tradi-
tional anchorage techniques that consist in developing tensile bars
in the form of a bend or a hook, headed bars allow for optimiza-
tion of the amount of steel and prevent congestion of reinforcement
in the column. The present paper deals with tests of short headed
bars anchoring tensile forces in compact columns. Practical bene-
fits yielding from the research are shown on examples of industrial
building products.

Keywords: anchor reinforcement; beam-column joint; headed bars; tensile

tests.

INTRODUCTION
A number of design guidelines are currently available to
assess the performance of headed bars used in beam-column
joints. ACI 318-141 allows the assumption that headed bars
are equivalent to hooks or bends provided that ldt > d/1.5,
and the bars are developed to the side of the column opposite
of the joint. In all other cases (for example, short headed bars
with ldt < d/1.5), the resistance of the anchorage against cone
breakout needs to be verified. fib Bulletin 722 introduces an
approach similar to ACI 318-141 with a ratio of heff/d > 0.6
required to treat the headed bar in a manner equivalent to a
bend or a hook. Eurocode EN 1992-1-13 does not formu-
late methods for the assessment of headed bars. Within
current European legislation, the performance of products
and systems that use headed bars as anchorage elements
is usually assessed by product approvals that demonstrate
the conformity of the headed bars to technical specification
CEN/TS 1992-4-2,4 where the resistance of the headed bars
is typically limited by the concrete failure (for example,
concrete cone failure) even with relatively long headed bars.
Where the tensile capacity of the headed bar is governed
by the concrete cone breakout, the codes referred to herein
allow use of anchor reinforcement to increase the capacity
of the anchorage. Most typically, anchor reinforcement is
provided in the form of stirrups that are designed and detailed
to tie the potential breakaway cone to the rest of the concrete
body (Fig. 1). ACI 318-141 allows use of anchor reinforce-
ment provided that the anchorage length of the stirrup in the
concrete cone (equivalent to l1 from Fig. 1) is greater than
8ds or 152 mm (6 in.). The technical specification CEN/TS
1992-4-24 limits the capacity of the anchor reinforcement
with the following design value of resistance
Fig. 1—Headed bar with anchor reinforcement.
l1 ⋅ π ⋅ d s ⋅ f bd
N Rd , a = ∑ (1)
n α
Comprehensive information about the background of
this design model may be found in reference published by
Eligehausen et al.5 The logic and practical implications of
Eq. (1) for the design of headed anchors has been recently
researched by several authors. Henriques et al.6 tested indi-
vidual headed bars and groups of headed bars with embed-
ment depth heff = 150 mm (5.91 in.) coupled with anchor
reinforcement, observing failure loads significantly higher
in comparison to those predicted using Eq. (1). Similar
tests and conclusions have been published by Bujnak et al.7
for groups of headed bars with anchor reinforcement and
anchorage depth as short as heff = 65 mm (2.56 in.). From-
knecht et al.8 tested fastening plates with four headed bars,
heff = 150 mm (5.91 in.), and anchor reinforcement. The
plates were embedded in columns and loaded by eccentric
shear load. The tests allowed identification of various failure
modes with failure loads four to eight times higher compared
to the characteristic value of resistance determined in accor-
dance with CEN/TS 1992-4-2.4 Berger9 developed a new
design model for anchors with anchor reinforcement. The
model was verified using a comprehensive database of tests
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2017-049.R2, doi: 10.14359/51701121, received March 15, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.

ACI Structural Journal/January 2018 203

Fig. 2—Alternative methods to anchor tensile forces to a
column: (a) bend or hook; (b) short headed bar with anchor
reinforcement; and (c) long headed bar developed to oppo-
site side of column.
Fig. 3—Geometric properties of headed anchor.
available from literature as well as Berger’s own tests. The
RESEARCH SIGNIFICANCE
model formulates the anchorage failure of the anchor rein-
The experimental program has been performed to over-
forcement with a resistance value significantly higher than
come the lack of knowledge about the structural performance
the value provided by Eq. (1). Meanwhile, it also introduces
of deformed short headed bars (heff = 140 mm [3.94  in.])
a maximum resistance that is defined as a factored breakout
with a large diameter (25 mm [0.98 in.]) and short overlap
resistance of concrete
length with anchor reinforcement (l1 = 81 mm [3.18 in.]).
Test series A allowed optimization of the test procedure,
NRd,max = ψ ∙ NRd,c (2)
identifying the minimum amount of reinforcement needed
for a reliable functioning of the test setup. Test series B and
with the coefficient ψ having a maximum value of 2.6. The
C demonstrated the failure loads and modes of short headed
tests of headed bars with anchor reinforcement summarized
bars with anchor reinforcement. A significant difference
herein have been performed using headed anchors made
between the measured value and failure loads predicted by
of smooth, round bars with a maximum diameter 16 mm
available design methods was observed.
(0.63 in.). Except for a handful of tests with shorter embed-
ment depth, the embedment depth of anchor bars used in
EXPERIMENTAL INVESTIGATION
the tests is 150 mm (5.91 in.) or greater. All of the works
The experimental program consisted of three test series,
dealing with the assessment of the current design methods
where the tensile capacity of headed anchors cast in concrete
against results of experimental research conclude that design
columns was tested. Identical anchors with dimensions summa-
methods that use Eq. (1) are extremely conservative.
rized in Fig. 3 were used in all tested specimens. The headed
The use of industrially manufactured building products
bars are manufactured from European deformed bars B500B
might have a significant impact on the efficiency of precast
with characteristic yield strength 500 N/mm2 (72.52 ksi).
concrete construction. Within beam-column joints, such
The headed bars were welded to a distribution plate with
building products (corbels, reinforcing bar couplers, or
a round hole. The distribution plate was connected to a steel
others) carry efforts from the beam and anchor them in the
rod through a hinge. The load was applied to the steel rod
column. To secure the continuity of forces between the beam
by a hollow hydraulic jack (Fig. 4). The load was applied
and the column, the anchorage part of the building products
continuously up to failure and measured by a load cell
is typically manufactured from deformed bars with forms
placed on the top of the hydraulic jack. The displacements
illustrated in Fig. 2: (a) bend developed to the opposite side
were measured using gauges placed between the distribution
of the column; (b) short headed bar with anchor reinforce-
plates and the bottom of the concrete column.
ment; and (c) long headed bar developed to the opposite side
The concrete specimens were reinforced horizontal
of the column.
and transverse reinforcement. Both tests with and without
While options (a) and (c) are in line with current design
anchor reinforcement placed around the headed bar have
codes (for example, ACI 318-141) and their design is thus
been conducted.
quite straightforward, option (b) is preferable from indus-
trial point of view because it allows standardization of the
Test series A
manufacturing process of building products and thus reduces
The target of test series A was to evaluate the amount
their cost for the construction industry. At the same time, the
of horizontal and transverse reinforcement needed for the
available data from literature indicate that the use of CEN/TS
proper execution of test. Indeed, while the bending and shear
1992-4-24 for the design of such short headed anchors leads
reinforcement of the concrete member might positively
to a design that is inefficient from a cost point of view. This is
affect the local performance of the anchorage, it is neces-
what motivates the designers of short headed anchors to look
sary to avoid the global failure of the tested specimens by
for alternative design methods—that is, design by testing.
bending or shear. The geometry and material parameters of

204 ACI Structural Journal/January 2018

Fig. 4—Test setup.

Fig. 5—Geometric parameters of specimens, Series A.

the specimens tested in Series A are summarized in Fig. 5
and Table 1.
The load displacement behavior of Series A is illustrated
in Fig. 6 and the maximum values of load recorded during
the test are in Table 1. The only specimen where the failure
of the anchorage was observed was Specimen 3A (Fig. 7).
The failure of Specimens 2A and 4A was associated with the
development of diagonal cracks starting from the bottom of
the tested specimen (shear failure of the concrete member),
and the failure of Specimen 5A was associated with the
opening of vertical cracks in the top part of the specimen
(bending failure).
Fig. 6—Load-displacement curves of specimens, Series A.

ACI Structural Journal/January 2018 205

Table 1—Test specimens, series A
Concrete beam Beam reinforcement Anchor reinforcement
fc,cube, m, N/mm
2
Bars, mm Stirrups, mm Bars, mm Stirrups, Ntest, kN
Specimen l, mm (in.) b, mm (in.) h, mm (in.) (ksi) (in.) (in.) (in.) mm (in.) (kip)
1A Ø10 (0.39) — 134.5 (30.2)
2A Ø20 (0.79) — 246.4 (55.4)
1500 280 280 41.5 4 Ø10 2 Ø10
3A Ø20 (0.79) Ø8 (0.31) 302.6 (68.0)
(59.06) (11.02) (11.02) (6.02) (0.39) (0.39)
4A Ø12 (0.47) — 187.2 (42.1)
5A Ø12 (0.47) Ø8 (0.31) 213.6 (48.0)
Reinforcement, fy, N/mm2 fu, N/mm2
mm (in.) (ksi) (ksi)
Ø8 (0.31) 609 (88.3) 641 (93.0)
Ø10 (0.39) 532 (77.2) 610 (88.5)
Ø12 (0.47) 544 (78.9) 627 (90.9)
Ø25 (0.98) 555 (80.5) 651 (94.4)

Fig. 7—Specimens 2A, 3A, and 4A at failure load and Specimens 1A and 5A after failure.
Test series B and C
Based on observations yielding from test series A, the spec-
imens tested in series B and C were equipped with normal
reinforcement consisting of four reinforcing bars Ø25 mm
(0.98 in.) and a shear reinforcement intended to prevent the
shear failure of the specimen outside the anchorage area.
Test series B had a concrete with average cubic strength
(measured on cubes with 150 mm [5.91 in.] edges) fc,cube,m
= 40.07 N/mm2 (5.81 ksi); the average cubic strength in test
series C was fc,cube,m = 52.5 N/mm2 (7.61 ksi). The dimen-
sions and parameters of specimens tested in series B and C
are summarized in Fig. 8 and Table 2. The varied parame-
ters were the amount and diameter of anchor reinforcement
and transverse reinforcing bars placed in the corners of the
anchor reinforcement. Specimens 4B and 3C had no anchor
reinforcement; all other specimens were equipped with
anchor reinforcement.
The load displacement behavior of Specimens B and C is
illustrated in Fig. 9. The failure of Specimen 1C happened
in the welds between the headed bar and the distribution
plate. In all other specimens with anchor reinforcement, the
failure was associated with the development of cracks on
the surface of the concrete specimen around the headed bar
(Fig. 10). Headed bars without anchor reinforcement were
pulled out of concrete (Fig. 11).
EXPERIMENTAL RESULTS AND DISCUSSION
The direct comparison between tests with anchor rein-
forcement and reference tests without anchor reinforcement
indicates the positive effect of the anchor reinforcement
on the performance of the anchorage. Depending on the

206 ACI Structural Journal/January 2018

Fig. 8—Geometric parameters of specimens, Series B and C.

Table 2—Test specimens, Series B and C

Concrete beam Beam reinforcement Anchor reinforcement
fc,cube, m, Bars, mm Stirrups, mm Bars, mm Stirrups, mm Ntest, kN
Specimen l, mm (in.) b, mm (in.) h, mm (in.) N/mm2 (ksi) (in.) (in.) (in.) (in.) (kip)
1B 4 Ø10 (0.39) 2 Ø10 (0.39) 294.9 (66.3)
2B 280 280 4 Ø12 (0.47) 2 Ø10 (0.39) 303.7 (68.3)
3B (11.02) (11.02) 4 Ø10 (0.39) 4 Ø8 (0.31) 330.4 (74.3)
1500 40.07 Ø 25 Ø8
4B (59.06) (5.81) (0.98) (0.31) — — 138.4 (31.1)
5B 380 (14.96) 280 (11.02) 292.8 (65.8)
4 Ø10 (0.39) 2 Ø10 (0.39)
6B 380 (14.96) 380 (14.96) 293.7 (66.0)
Reinforcement, fy, N/mm 2
fu, N/mm 2

mm (in.) (ksi) (ksi)

Ø8 (0.31) 535 (77.6) 693 (100.5)
Ø10 (0.39) 524 (76.0) 617 (89.5)
Ø25 (0.98) 567 (82.2) 657 (95.3)
Beam
Concrete beam reinforcement Anchor reinforcement
fc,cube, m, Bars, mm Stirrups, mm Bars, mm Stirrups, mm Ntest, kN
Specimen l, mm (in.) b, mm (in.) h, mm (in.) N/mm2 (ksi) (in.) (in.) (in.) (in.) (kip)
1C 4 Ø10 (0.39) 2 Ø10 (0.39) 343.4 (77.2)
1500 280 280 52.5 Ø 25 Ø8
2C — 2 Ø10 (0.39) 305.6 (68.7)
(59.06) (11.02) (11.02) (7.61) (0.98) (0.31)
3C — — 167.7 (37.7)
Reinforcement, fy, N/mm2 fu, N/mm2
mm (in.) (ksi) (ksi)
Ø8 (0.31) 547 (79.3) 615 (89.2)
Ø10 (0.39) 549 (79.6) 640 (92.8)
Ø25 (0.98) 541 (78.5) 630 (91.4)

ACI Structural Journal/January 2018 207

amount of steel used (both anchor reinforcement and trans- where fbm = 2.25 ∙ η1 ∙ η2 ∙ fctm.
verse bars), the ratio is between 2.05 and 2.39 (Table 3). The The mean value of ktest = 3.66 with COV = 17.79%. The
direct comparison between tests 1C and 2C also illustrates value ktest is determined as
the beneficial effect of the transverse bars on the anchorage
capacity. This influence is currently neglected by available N test
design models. ktest = (4)
N R , calc
The comparison of the measured and calculated values indi-
cates that the available design models might not allow assess- where
ment of the performance of headed bars to be applied in a real-
istic manner such as the tested configuration. The ACI model Ac , N
is not applicable because the anchorage length of the stirrup in N R , calc = ⋅ (11.9 ⋅ f cm , cube ⋅ hef1.5 ) ⋅ ψ s , N ⋅ ψ re , N ⋅ ψ ec , N (5)
the concrete cone is smaller than the minimum value specified Ac0, N
by the code. The mean value of Ntest/NRm,a = 3.37 with COV =
11.40%, where the calculated value is determined as Ac,N = (3 ∙ heff) ∙ b (6)

l1 ⋅ π ⋅ d s ⋅ f bm A0c,N = (3 ∙ heff)2 (7)

N Rm , a = ∑ (3)
n α
where ψs,N, ψre,N, and ψec,N are factors according to CEN/TS
1992-4-2.4

PRACTICAL IMPLICATIONS
The tests presented herein have been used to optimize the
anchorage part of an innovative corbel system that is used as
a vertical support for reinforced concrete or steel-concrete
composite beams.10 It consists of an anchorage part that is
integrated in a precast concrete column and a steel bracket
that is bolted to the anchorage part after the removal of the
column’s formwork (Fig. 12).
The anchorage part consists of a system of vertical and
horizontal anchorage bars. Variants of the anchorage
part of the corbel used up to now in Europe are shown in
Fig. 13. While the anchorage in Fig. 13(a) complies with
EN 1992-1-1,3 the shape and length of the anchor bars
Fig. 9—Load-displacement curves, series C and B.

Fig. 10—Specimens 1B through 6B at failure load.

208 ACI Structural Journal/January 2018

Table 3—Summary of predictions and experimental results
Anchor r. Concrete
Specimen Ntest, kN (kip) Stirrups, mm (in.) l1, mm (in.) fcm,cube, N/mm (ksi)
2
fctm, N/mm2 (ksi) fbm, N/mm2 (ksi) NRm,a, kN (kip) Ntest/NRm,a
1B 294.9 (66.3) 2 Ø10 (0.39) 81 (3.19) 40.07 (5.81) 2.50 (0.36) 5.62 (0.82) 81.8 (18.4) 3.61
2B 303.7 (68.3) 2 Ø10 (0.39) 81 (3.19) 40.07 (5.81) 2.50 (0.36) 5.62 (0.82) 81.8 (18.4) 3.71
3B 330.4 (74.3) 4 Ø8 (0.31) 78 (3.07) 40.07 (5.81) 2.50 (0.36) 5.62 (0.82) 126.0 (28.3) 2.62
5B 292.8 (65.8) 2 Ø10 (0.39) 81 (3.19) 40.07 (5.81) 2.50 (0.36) 5.62 (0.82) 81.8 (18.4) 3.58
6B 293.7 (66.0) 2 Ø10 (0.39) 81 (3.19) 40.07 (5.81) 2.50 (0.36) 5.62 (0.82) 81.8 (18.4) 3.59
3A 302.6 (68.0) 2 Ø10 (0.39) 81 (3.19) 41.5 (6.02) 2.58 (0.37) 5.80 (0.84) 84.4 (19.0) 3.59
1C 343.4 (77.2) 2 Ø10 (0.39) 81 (3.19) 52.5 (7.61) 3.15 (0.46) 7.08 (1.03) 103.0 (23.2) 3.33
2C 305.6 (68.7) 2 Ø10 (0.39) 81 (3.19) 52.5 (7.61) 3.15 (0.46) 7.08 (1.03) 103.0 (23.2) 2.97
4B 138.4 (31.1) — — 40.07 (5.81) — — — —
3C 167.7 (167.7) — — 52.5 (7.61) — — — —

Specimen b, mm (in.) heff, mm (in.) ψs,N NR,calc, kN (kip) ktest — Ratio

1B 280 (11.02) 0.9 74.87 (16.83) 3.94 — Ntest,1B/Ntest,4B 2.13

2B 280 (11.02) 0.9 74.87 (16.83) 4.06 — Ntest,2B/Ntest,4B 2.19

3B 280 (11.02) 0.9 74.87 (16.83) 4.41 — Ntest,3B/Ntest,4B 2.39

5B 380 (14.96) 0.97 109.67 (24.66) 2.67 — Ntest,5B/Ntest,4B 2.12

6B 380 (14.96) 140 0.97 109.67 (24.66) 2.68 — Ntest,6B/Ntest,4B 2.12

3A 280 (11.02) (5.51) 0.9 76.19 (17.13) 3.97 — — —

1C 280 (11.02) 0.9 85.70 (19.27) 4.01 — Ntest,1C/Ntest,3C 2.05

2C 280 (11.02) 0.9 85.70 (19.27) 3.57 — — —

4B 280 (11.02) 0.9 74.87 (16.83) 1.85 — — —

3C 280 (11.02) 0.9 85.70 (19.27) 1.96 — — —

Fig. 11—Specimens 1C, 2C, and 3C after failure.
needs to be adapted to the dimensions of the column and
thus the product cannot be standardized. The anchorage in
Fig. 13(b) uses a short headed bar and complies with the
Fig. 12—Corbel.
requirements of CEN TS 1992-4-2.4 The use of short anchor
bars allows for optimization of the manufacturing process of
the product; on the other hand, the large amount of anchor
reinforcement yielding from Eq. (1) leaves the installa-
tion of the product relatively complicated. The anchorage
in Fig. 13(c) was designed based on tests presented in this
paper and approved by German Building Authority DIBt,11
which allows for standardized manufacturing of the product

ACI Structural Journal/January 2018 209


Fig. 13—Variants of corbel used in past and present:
(a) anchorage provided with bended deformed bar; (b)
anchorage provided by short headed bars using anchor
reinforcement determined in accordance with Eq. (1); and
(c) anchorage provided by short headed bars designed in
accordance with tests referenced in present paper.
and enables it to be installed into the column in a simple and
productive manner.
CONCLUSIONS
The presented tests demonstrated the performance of short
headed bars with anchor reinforcement embedded in concrete
columns. Such short headed bars are increasingly being used
in innovative building products because they allow for opti-
mization of the manufacturing process and the costs of such
products. The main conclusions of the research are:
• Short headed bars made of deformed bars with embed-
ment depth 140 mm (5.51 in.) may anchor tensile forces
in a concrete column.
• The anchorage capacity of such headed bars can be
significantly increased by combining the short headed
bars with anchor reinforcement.
• Anchor reinforcement can be activated even with short
overlap lengths (l1 = 81 mm [3.19 in.]).
• The ratio between the failure load of a headed bar with
and without anchor reinforcement is 2.05 and 2.39,
depending on the amount of anchor reinforcement. This
210
ratio is significantly higher in comparison to what can
be predicted by available design models.
NOTATION
d = effective depth of concrete member
ds = diameter of anchor reinforcement
fbd = design value of bond strength
fc,cube,m = mean value of concrete cube strength
fc,t,m = mean value of concrete tensile strength
heff = embedment depth of headed bar
ldt = anchorage length of headed bar
l1 = overlap length between headed bar and anchor reinforcement
NRd,a = anchorage resistance of anchor reinforcement
NRd,c = resistance against concrete cone failure
NRd,max = maximum resistance of anchor reinforcement
Ntest = maximum measured value of load
α = empirical factor (=1.0 for straight bars, =0.7 for bended bars)
AUTHOR BIOS
Ján Bujňák is a Research and Development Manager at Peikko Group
Corporation, Lahti, Finland. He received his PhD from Polytech Cler-
mont-Ferrand, Aubière, France, in 2007. His research interests include
bond, fastening technology, and punching shear in flat slabs.
Matúš Farbák is a Researcher at the University of Žilina, Žilina, Slovakia,
where he received his PhD in 2015. His research interests include the
behavior of headed bars in concrete and steel-concrete composite structures.
REFERENCES
1. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 2014, 519 pp.
2. Fédération internationale du béton, “fib Bulletin 72: Bond and
Anchorage of Embedded Reinforcement,” Lausanne, Switzerland, 2014,
161 pp.
3. EN 1992-1-1:2004, “Design of Concrete Structures—Part 1: General
Rules and Rules for Buildings,” European Committee for Standardization
(CEN), Brussels, Belgium, 2004, 225 pp.
4. CEN/TS 1992-4-2:2009, “Design of Fastenings for Use in Concrete—
Part 4-2: Headed Fasteners,” European Committee for Standardization
(CEN), Brussels, Belgium, 2009, 31 pp.
5. Eligehausen, R.; Malée, R.; and Silva, J. F., Anchorage in Concrete
Construction, Ernst & Sohn, Berlin, Germany, 2006, 391 pp.
6. Henriques, J.; Raposo, J. M.; Simões da Silva, L.; and Luís Costa
Neves, L., “Tensile Resistance of Steel-Reinforced Anchorages: Experi-
mental Evaluation,” ACI Structural Journal, V. 110, No. 2, Mar.-Apr. 2013,
pp. 239-249.
7. Bujnak, J.; Farbak, M.; Bahleda, F.; and Leinonen, T., “On the Resis-
tance of Fastening Plates with Supplementary Reinforcement,” fib Sympo-
sium, 2015, Copenhagen, Denmark.
8. Fromknecht, S.; Odenbreit, C.; and Dorka, U., “Versuche zur Trag-
fähigkeit von Ankerplatten mit einbetonierten Kopfbolzendübeln in
schmalen Stahlbetonstützen,” Beton- und Stahlbetonbau, V. 105, No. 6,
2010, pp. 362-370. doi: (in German)10.1002/best.201000016
9. Berger, W., “Load-Displacement Behavior and Design of Anchorages
with Headed Studs with and without Supplementary Reinforcement under
Tension Load,” dissertation, University of Stuttgart, Stuttgart, Germany,
2015, 267 pp. (in German).
10. Bujnak, J.; Roeser, W.; Matiasko, S.; and Böhm, M., “Experimental
Assessment of an Innovative Beam to Column Connection,” High Tech
Concrete: Where Technology and Engineering Meet, D. Hordijk and M.
Luković, eds., 2018, Springer, Cham, Switzerland.
11. Deutsches Institut fur Bautechnik (DIBt), “Allgemeine bauauf-
sichtliche Zulassung Z-21.8-2076: Peikko PCs Konsole,” 2016, Berlin,
Germany. (in German)
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S17
Evaluation of Flexural and Shear Stiffness of Concrete
Squat Walls Reinforced with Glass Fiber-Reinforced
Polymer Bars
by Ahmed Arafa, Ahmed Sabry Farghaly, and Brahim Benmokrane
Estimating the flexural and shear stiffness of concrete squat walls
reinforced with glass fiber-reinforced polymer (GFRP) bars is
important to evaluate lateral displacement. To address this issue,
five full-scale concrete squat walls, including four reinforced with
GFRP bars and one reinforced with steel bars, were tested to failure
under quasi-static reversed cyclic lateral loading. Decoupling flex-
ural and shear deformations of the tested specimens showed the
contribution of shear deformation to the lateral displacement. The
shear stiffness of the cracked wall can be estimated based on the
truss model with an acceptable level of conservatism. The shear-
crack angle and concrete shear strength were evaluated. The
flexural stiffness was estimated based on available expressions in
codes and guidelines related to the design of concrete members
reinforced with fiber-reinforced polymer bars, demonstrating their
adequacy with walls although they were established for beam and
slab elements. Based on regression analyses of the test results,
expressions that correlate flexural and shear stiffness to lateral
drift were proposed. Such expressions would be vital in the context
of displacement-based design.
Keywords: concrete squat walls; flexural and shear deformations; glass
fiber-reinforced polymer bars; seismic resistance; stiffness.
INTRODUCTION
The use of fiber-reinforced polymer (FRP) materials has
grown to overcome the usual problems induced by the corro-
sion of steel reinforcement in concrete structures. The inves-
tigation of FRP-reinforced concrete structures, however,
have focused mainly on the behavior under static-loading
conditions, focusing less frequently on seismic design.
The feasibility of using FRP as internal reinforcement for
lateral-resisting systems while preserving the stiffness and
deformation capacity has become prominent. Mohamed et
al. (2014a) tested mid-rise shear walls showing the stable
cyclic performance and high level of deformability achieved
by glass FRP (GFRP)-reinforced shear walls in comparison
to one reinforced with steel. Mohamed et al. (2014b) indi-
cated the potential of GFRP reinforcement in distributing
shear deformations along the wall height, owing to its elastic
nature, resulting in control shear distortion relatively to that
in the steel-reinforced wall in which shear distortion took
place simultaneously with occurrence of yielding of flex-
ural reinforcement and mobilized at the plastic hinge zone,
thereby deteriorating shear resistance.
The test results for the GFRP-reinforced mid-rise walls
paved the way for a new experimental series using GFRP
bars in squat walls (shear span-to-length ratio less than
2.0) in which the shear-deformation problem is frequently
ACI Structural Journal/January 2018
TECHNICAL PAPER
encountered (Paulay et al. 1982; Luna et al. 2015). Arafa et
al. (2016) reported experimental results on two squat walls:
one was reinforced with conventional steel bars, while
the second was reinforced with GFRP bars. The GFRP-re-
inforced squat wall attained satisfactory strength and stable
cyclic behavior as well as self-centering capacity that
contributed in preventing sliding shear, which occurred in
the steel-reinforced counterpart.
One of the most important aspects in squat-wall design
is estimating wall’s lateral displacement and limiting this
displacement to an acceptable level. This requires an appro-
priate estimation of wall lateral stiffness, which can signifi-
cantly affect the calculation of the natural period time and
the distribution of lateral forces among structure walls as
well. Accordingly, estimating both the flexural and shear
stiffness of GFRP-reinforced squat walls was the main focus
of this study.
RESEARCH SIGNIFICANCE
This paper focuses mainly on estimating the flexural and
shear stiffness of GFRP-reinforced concrete squat walls as a
lateral seismic element. Flexural and shear deformations were
decoupled, showing the significant effect of shear deforma-
tion on the total displacement. The flexural and shear stiffness
of the GFRP-reinforced concrete squat walls was evaluated.
Herein, these results are thoroughly discussed and compared
to the experimental results. In addition, to gain useful informa-
tion within the context of displacement-based seismic design,
expressions that directly correlate the squat-wall flexural and
shear stiffness with lateral-drift ratio were proposed.
SUMMARY OF EXPERIMENTAL PROGRAM
AND RESULTS
Five full-scale reinforced concrete squat walls were
constructed and tested to failure under quasi-static reversed
cyclic lateral loading. Four specimens were entirely rein-
forced with GFRP bars (G4-250, G4-160, G4-80, and G6-80)
and one was reinforced with steel bars (S4-80). Figures 1(a)
and 1(b) show the concrete dimensions and reinforcement
configuration of the test specimens. The boundary elements’
longitudinal and transverse reinforcement ratios and vertical
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2017-052.R1, doi: 10.14359/51700987, received March 5, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
211
Fig. 1—Concrete dimensions, reinforcement details, test setup, load history, and instrumentation. (Note: 1 mm = 0.0394 in.)
Fig. 2—Failure modes of test specimens.
web reinforcement were 1.43%, 0.89%, and 0.59%, respec-
tively, in all specimens. Four horizontal web reinforcement
ratios equal to 0.51%, 0.79%, 1.58%, and 3.58% were used
in G4-250, G4-160, G4-80, and G6-80 with using two layers
of No. 4 GFRP bars spaced at 250, 160, and 80 mm (9.8,
6.3, and 3.15 in.) or No. 6 GFRP bars spaced at 80 mm
(3.15 in.), respectively. Specimen S4-80 served as a refer-
ence for G4-80, so both specimens had identical reinforce-
ment configurations and ratios. One layer of bidiagonal No.
3 GFRP bars with spacing of 100 mm (3.94 in.) was added
to prevent sliding shear. Figures 1(c) and 1(d) show the test
setup and loading history, respectively. A series of linear
variable differential transducers (LVDTs) and bar strain
gauges was mounted on the specimens (Fig. 1(e)). Table 1
provides the mechanical properties of the reinforcement.
Premature sliding failure dominated the behavior of S4-80
due to flexural reinforcement yielding, which produced a
major horizontal crack (Fig. 2(a)). Such behavior, however,
was prevented in G4-80, which exhibited flexural compres-
212
sion failure (Fig. 2(b)). This was attributed to the elastic
nature of GFRP bars, which helped the cracks to realign and
lock up in the compression zone as well as distributing shear
deformations along the wall height. Figure 3 illustrates that
both specimens exhibited similar initial stiffness. Due to
the relatively low elastic modulus of GFRP bars, however,
G4-80 experienced softer behavior than S4-80 after initia-
tion of the first flexural crack. The two envelopes intersected
at 1.35% lateral drift. Then, the strength of S4-80 deterio-
rated due to localized sliding-shear deformations, while
G4-80’s strength kept increasing to achieve an ultimate load
and drift of almost 71% and 50% higher than those of S4-80,
respectively.
The failure of G6-80 was identified as flexural compres-
sion failure (Fig. 2(c)). The failure of G4-250 occurred by
sliding along a major diagonal shear crack (Fig. 2(d)) due
to the inadequacy of horizontal web reinforcement, while
G4-160 experienced sudden flexural rupture in the longitu-
dinal bars at the boundary element under tension (Fig. 2(e)).
ACI Structural Journal/January 2018
Table 1—Tensile properties of reinforcement
Tensile modulus of
Bar Designated bar diameter, mm Nominal area*, mm2 elasticity, GPa Tensile strength†‡, MPa Average strain at ultimate, %
Straight bars
No. 3 GFRP 9.5 71 65 1372 2.1
No. 3 steel 9.5 71 200 fy = 420 εy = 0.2
No. 4 steel 12.7 129 200 fy = 420 εy = 0.2
Bent No. 3 GFRP—rectilinear spiral
Straight 50 1065 2.1
9.5 71
Bent — 460 —
Bent No. 4 GFRP—horizontal bar
Straight 50 1020 2.0
12.7 129
Bent — 459 —
Bent No. 6 GFRP—horizontal bar
Straight 50 1028 2.0
19.1 285
Bent — 463 —
*
According to CSA S807 (CSA 2010).
†
Tensile properties were calculated using nominal cross-sectional areas.
‡
Guaranteed tensile strength: average value is three times standard deviation (ACI 440.1R-15).
Notes: 1 mm = 0.0394 in.; 1 MPa = 145 psi; fy is steel yielding strength; εy is steel yielding strain.

in the length of the two diagonals using Eq. (1a). The flex-
ural deformations, on the other hand, can be calculated based
on the two vertical LVDTs mounted at both boundaries with
height h using Eq. (1b) (Fig. 4(a))

(d1′ − d )d − (d 2′ − d )d
U s original = γ original h = (1a)
2hL

(VL − VR )
Uf original = θh = h (1b)
L

where γoriginal is the shear distortion over a height h; d1′ and

Fig. 3—Load-top displacement envelope curves. (Note: 1 mm
= 0.0394 in.; 1 kN = 0.225 kip.)
Figure 3 shows that the horizontal web reinforcement ratio
had a significant effect on increasing the ultimate strength
and drift ratio. This effect, however, appears to have been
insignificant when the wall was provided with more hori-
zontal web reinforcement than required for flexural resis-
tance (G6-80 compared to G4-80). Overall, the observed
behavior reveals the acceptable behavior of GFRP-rein-
forced walls as a lateral-resisting system in low to moderate
earthquake regions. Table 2 summarizes the failure progres-
sion of the test specimens.
DECOUPLING OF FLEXURAL AND SHEAR
DEFORMATIONS
Oesterle et al. (1979) suggested that the shear deformation
of a shear panel can be directly estimated from the changes
d2′ are the lengths of the diagonal LVDTs after deformation;
d is the original length of the diagonal before deformation;
L and h are the length and height of the panel, respectively;
and θ is the rotation over the height h.
The deformation calculated by this method, however, is
overestimated because of the variation in bending moment
along the wall height. Hiraishi (1984) demonstrated that this
method is only valid if the center of rotation is located at the
element center (Fig. 4(b)), which is not the case of structural
walls. Therefore, a portion of the change in diagonal lengths
has to be attributed to flexural deformations (Fig. 4(c)). More-
over, Eq. (1b) is only valid if the curvature is concentrated at
the wall base, which is not the case in many walls. Therefore,
Hiraishi (1984) suggested the following corrected equations
Us corrected = Us original – (α – 0.5)θh (2a)
Uf corrected = αθh = αUf original (2b)
∫0 θ ( y ) dy
h
α= (2c)
θh

ACI Structural Journal/January 2018 213

Table 2—Summary of test results
Initial flexural Initial shear Concrete cover Excessive cover
crack crack Steel yielding splitting spalling Peak capacity Failure
Wall fc′,
designation MPa P, kN d, % P, kN d, % P, kN d, % P, kN d, % P, kN d, % P, kN d, % P, kN d, %
G4-250 35 147 0.025 228 0.39 — — 510 1.6 615 2.0 678 2.65 678 2.65
G4-160 35 145 0.025 227 0.36 — — 528 1.5 639 2.0 708 2.8 708 2.8
G4-80 40 164 0.026 234 0.34 — — 567 1.5 710 2.0 912 2.75 810 3.0
G6-80 41 168 0.026 237 0.29 — — 582 1.5 739 2.0 935 2.9 815 3.1
S4-80 35 160 0.025 230 0.1 425 0.4 525 1.0 534 1.25 534 1.25 440 2.00

Notes: fc′ is concrete compressive strength; P is applied lateral load; d is drift ratio; 1 MPa = 145 psi; 1 kN = 0.225 kip.

Fig. 4—Decoupling of flexural and shear deformations.

Fig. 5—Method for estimating α (rotation profile over wall
height).
The center of rotation (α) accounts for the variation in curva-
ture along the panel height and can be estimated as shown
in Fig. 5. Mohamed et al. (2014b) stated that Eq. (1) over-
estimated shear deformations, while correcting the results
based on Eq. (2) produced consistent results. This confirms
the necessity of involving α in the calculation to obtain
corrected values for decoupled deformations. In this study,
the flexural and shear deformations were calculated over a
height h equal to the wall length lw based on the arranged
instrumentation (Fig. 1(e)). To assist in calculating the
center of rotation, average curvature and rotations over the
wall height were calculated (Fig. 6). α was found to be 0.62
for S4-80, 0.59 for G6-80, and 0.58 for the other specimens.
Original and corrected flexural and shear
deformations
As shown in Fig. 7, the total displacement was substan-
tially overestimated with Eq. (1), while the results with
Eq. (2) were in good agreement with the measured lateral
displacement. Figure 8 shows plots of the corrected flex-
ural and shear deformations. The behavior of all specimens
was dominated initially by flexural response. With the initi-
ation of the first shear crack, shear deformations began to
contribute in the total displacement. Apparently, the contri-
bution of shear deformation varied as a function of reinforce-
ment type (steel or GFRP) and horizontal web reinforcement
ratio. In S4-80, the percentage of shear deformation to the
total deformation corresponding to yielding of the longitu-
dinal reinforcement was 22% (0.4% drift) and increased to
46% after the localization of sliding shear corresponding
to concrete cover spalling (1.25% drift) and reached 64%
at failure (2% drift). In the GFRP-reinforced specimens,
the percentages were approximately 49%, 38%, 28%, and
15% for G4-250, G4-160, G4-80, and G6-80, respectively,
up to 1% drift. At cover spalling (2% drift), the percentage
increased slightly to 56%, 42%, 36%, and 20%, respectively,
which remained almost constant up to failure. The results
reveal that shear deformation should not be neglected, even
if the shear strength was twice the applied load, as is the case
with G6-80.
All specimens exhibited negligible initial flexural deforma-
tion because the gross flexural stiffness resisted the deforma-
tion, as shown in Fig. 9(a). By the onset of the first flexural

214 ACI Structural Journal/January 2018

Fig. 6—Calculated curvature and rotation profiles for test specimens. (Note: 1 mm = 0.0394 in.)

Fig. 7—Measured and calculated displacement at height equal to wall length. (Note: 1 mm = 0.0394 in.)

Fig. 8—Displacement components at height equal to wall length. (Note: 1 mm = 0.0394 in.)
crack, the specimens experienced a reduction in lateral stiff-
ness, resulting in a significant increase in flexural deforma-
tion. The flexural deformation in all GFRP-reinforced speci-
mens followed a similar trend, showing no significant effect of
changing the horizontal web reinforcement ratio. Figure 9(b)
shows the normalized secant flexural stiffness to the initial
elastic flexural stiffness. Due to the relatively low modulus
of elasticity of GFRP bars, the stiffness loss in the GFRP-re-
inforced squat walls was relatively pronounced compared to
the steel-reinforced one. It is worth mentioning, however, that
the softer behavior of GFRP-reinforced walls will increase the
displacement demand, which could be considered an advan-
tage because a softer structure attracts lower seismic forces.
Figures 9(c) and 9(d) show that, similar to the flexural-de-
formation response, initial high shear stiffness was followed
by a significant reduction in shear stiffness manifested with
the appearance of the first shear crack. The figures under-
line the link between steel yielding and shear deformations
in S4-80, which exhibited significant degradation in shear
stiffness after a few cycles of steel yielding associated with
substantial increasing in shear deformation. In contrast, the
shear deformation in GFRP-reinforced squat walls increased
almost linearly with loading. The figures also reveal the
effectiveness of the horizontal web reinforcement ratio in
reducing shear deformation.

ACI Structural Journal/January 2018 215


Fig. 9—Comparison between flexural and shear deforma-
tions in test specimens.
SHEAR STIFFNESS
Although predicting the shear stiffness of test specimens
beyond shear cracking is necessary, the FRP design codes and
guidelines (CSA S806 [2012], ACI Committee 440 [2015])
contain no seismic provisions. Therefore, the methods used
in estimating the cracked shear stiffness of steel-reinforced
squat walls were reviewed and the most appropriate method
to apply to GFRP-reinforced squat walls was selected.
The shear stiffness in S4-80 significantly decreased after
the first shear crack and reduced to 58% at the onset of flex-
ural-reinforcement yielding, followed by severe deteriora-
tion due to the development of sliding-shear deformations
(Fig.  9(d)), confirming past research outcomes (Salonikios
et al. 1999, Massone et al. 2009). According to Eurocode
8 (2004) and ASCE/SEI 43-05 (2005), the cracked shear
stiffness is taken as 50% of the elastic shear stiffness, while
ACI 318 (2014) and CSA A23.3 (2014) account only for
the reduction in flexural stiffness, allowing no reduction
in the cracked shear stiffness. Given this disagreement and
uncertainty, Li and Xiang (2011) proposed and experimen-
tally verified an analytical model based on truss analogy
to predict the cracked shear stiffness corresponding to the
yielding point (Fig. 10). Tang and Su (2014) undertook
further verification of the model.
The Li and Xiang (2011) model assumes that shear defor-
mation is the summation of tie elongation (horizontal web
reinforcement) and shortening of concrete struts, which
could be formulated as follows
sin 4 θ + nρh Vs ⋅ h
∆s = 2 2
(3)
ρh sin θ ⋅ cos θ b ⋅ d ⋅ Es
f s ⋅ dcotθ ⋅ Ah
Vs = (4)
s
216
Fig. 10—Truss model for shear-deformation estimation.
where Δs is the shear deformation at height h; θ is the angle
between the shear crack and the longitudinal axis of the
wall; n is the modular ratio = Es/Ec; Es and Ec are the elastic
moduli of the horizontal web reinforcement and concrete,
respectively; ρh is the horizontal web reinforcement ratio; Vs
is the shear force carried by horizontal web reinforcement;
b is the wall thickness; d is the effective shear depth, which
could be taken as 0.8 of the wall length; fs is the stress in
horizontal web reinforcement; and Ah is the area of hori-
zontal web reinforcement within a spacing s.
The shear stiffness is calculated as the ratio between the
applied load (V) and Δs, as follows
V ρ sin 2 θ ⋅ cos 2 θ b ⋅ d ⋅ Es ⋅ V
Ks = = h 4 (5)
∆s sin θ + nρh Vs ⋅ h
The model calculates the shear stiffness of steel-rein-
forced walls based on shear-reinforcement yielding and a
shear crack angle (θ) conservatively assumed to be equal
to 45 degrees. The model cannot be directly applied to
GFRP-reinforced squat walls due to the absence of a yielding
point and the significant deviation of θ from 45 degrees in
many test series. The current concept for calculating shear
resistance in most guidelines and codes is to find a rational
estimation for the shear carried by the concrete Vc; then, the
shear-reinforcement (Vs) portion can be calculated. ACI 318
(2014) assumes that Vc is equal to the first shear-cracking
load (Vcr). On the other hand, ACI 440.1R-15 (2015), CSA
S6-14 (2014), and CSA S806-12 (2012) provide formulas
for calculating Vc and θ.
Design codes and guides for shear-strength
predictions
ACI guide provision (ACI 440.1R-15)—The estimation of
Vc in ACI 440.1R-15 for FRP-reinforced concrete elements
is based on the assumption that θ is 45 degrees and that Vc
can be calculated as follows
Vc = 2/5√fc′bwc; c = kd, k = 2 ρ f n f + (ρ f n f ) 2 − ρ f n f ,
and n = Ef/Ec (6)
ACI Structural Journal/January 2018
Fig. 11—Effect of shear-crack angle.
where bw is the wall thickness; c is the neutral-axis depth It should be noted that θ in Eq. (12) is limited to 50 degrees
of the cracked transformed section; and d is the effective based on the limitation of εx in Eq. (11).
flexural depth.
CSA code provisions (S806-12 and S6-14)—According to Evaluation of shear-crack angle (θ)
CSA S806 (2012), the Vc of flexural members can be evalu- ACI 440.1R-15 assumes θ to be 45 degrees, while the
ated as follows calculation of θ based on CSA S6-14 and CSA S806-12
resulted in 50 and 60 degrees, respectively. A unified method
Vc = 0.05 km kr ka k s 3 f c′ bw d v (7) for calculating Vs was used (ACI Committee 318 2014) in
which the concrete shear strength (Vc) was set to be equal to
where the experimental cracking load (Vcr) (given in Table 2). The
shear deformation prior to cracking was calculated using the
0.11 f c′bw d v ≤ Vc ≤ 0.22 f c′bw d v ; km = V f d /M f ≤ 1.0; elastic shear stiffness defined by Park and Paulay (1975) for
prismatic elements as follows
kr = 1 + 3 E f ρ f ; ka = 2.5/(Mf/Vfd) ≥ 1.0;
Gc Ash
ks = 750/(450 + d) ≤ 1.0 (8) K se = (13)
f ⋅ hw
where dv is the effective shear depth, taken as the greater
of 0.9d or 0.72lw and not less than 0.8lw according to CSA where Gc is the concrete shear modulus (= 0.4Ec); Ash is the
A23.3-14; and Mf and Vf are the moment and shear applied effective shear area of an uncracked element (=(5 + 5ν)/(6 +
at the critical section of shear, respectively. 5ν)Ag), where Ag is the gross-section area and ν is Poisson’s
In contrast to ACI 440.1R-15, CSA S806-12 provides ratio; and f accounts for nonuniform distribution of the shear
Eq. (9) to estimate θ stresses and is equal to 1.2 for rectangular cross sections.
Figure 11 shows that adopting a shear-crack angle of 45
30 degrees ≤ θ = 30 + 7000εl ≤ 60 degrees; and 50 degrees as proposed in ACI 440.1R-15 and CSA
M f /d v + V f S6-14 underestimated the shear deformations by 53% and
εl = (9) 35%, respectively, because the experimentally measured
2 E f Af
angles ranged from 55 to 59 degrees. Thus, θ < 50 degrees
could lead to unsafe predictions. Increasing this limitation to
According to CSA S6-14, the Vc of flexural members using 60 degrees, as proposed in CSA S806-12, however, yielded
FRP as the main reinforcement can be evaluated as follows higher shear displacement than the experimental results with
a mean value of 36%. This revealed that 56 degrees was the
 0.4 1300  angle that produced the best fit with the experimental results
Vc =  × φc f c′ bw d v (10)
 1 + 1500 ε x 1000 + sxe  (Fig. 11). However, this angle might lead to unsafe predic-
tions in other test series. Therefore, the shear-crack angle
based on CSA S806-12 (θ = 60 degrees) is conservatively
M f /d v + V f 35 d v
where ε x = ≤ 0.003, sxe = ≥ 0.85 d v , recommended in this study.
2 E f Af 15 + ag
and ag is the aggregate size (11) Evaluation of concrete shear strength (Vc)
Vc was predicted based on ACI 440.1R-15, CSA S6-14,
CSA S6-14 provides an equation similar to that in CSA and CSA S806-12 as well as Vc = Vcr while setting θ to 60
S806-12 to calculate θ as follows degrees. Figure 12 shows that the predictions according to
ACI 440.1R-15 and CSA S6-14 overestimated shear defor-
θ = (29 + 7000εx)(0.88 + Sze/2500) (12) mations with mean values of 74% and 64%, respectively.
While the predicted Vc based on CSA S806-12 resulted in
closer agreement with the mean value of 13%, the prediction

ACI Structural Journal/January 2018 217

Fig. 12—Effect of concrete shear strength (at θ = 60 degrees). (Note: 1 mm = 0.0394 in.; 1 kN = 0.225 kip.)
Fig. 13—Concrete shear strength versus top displacement.
(Note: 1 mm = 0.0394 in.; 1 kN = 0.225 kip.)
may lead to unsafe estimation at elastic stage prior to shear
cracking load, which is significantly lower than the calcu-
lated shear strength and resulted in underestimated shear
deformation (Fig. 12).
The experimental concrete shear strength Vc was calcu-
lated as Vc = V − VFRP, where V is the applied load and VFRP
is the contribution of the FRP horizontal web reinforce-
ment calculated using the truss analogy (Eq. (4)) with the
average measured shear-strain values. Although using a
constant value for Vc is not representative in estimating shear
deformation as Vc varies with loading (Fig. 13), it could be
conservatively assumed that Vc = Vcr is equal to the first shear
cracking load. Therefore, it would be appropriate to deter-
mine a formula to estimate Vcr. In this regard, the ACI 318
(2014) procedure for concrete shear strength was followed
in which the cracking load is normalized to bwlwfc′0.5. Based
on the experimental shear-cracking-load values (Table 2),
the concrete shear strength (Vc) could be estimated as an
average value equal to 0.12bwlwfc′0.5.
Figure 14 confirms the underestimation using the equation
in ACI 440.1R-15 and CSA S6-14, while the CSA S806-12
equation initially overestimated the shear stiffness degrada-
tion and got noticeably better predictions at high load levels
exceeding Vcr. Furthermore, the calculation based on Vc =
0.12bwlwfc′0.5 produced better estimations with an acceptable
level of conservatism.
218
Proposed shear-stiffness model
Seismic design practice worldwide is moving toward
displacement-based design methods. Therefore, the seismic
demands of a structural system should be evaluated using an
effective natural period of a structure based on the effective
secant stiffness connected to a targeted displacement point
on the load-displacement envelope curve. Hence, it would
be preferable, within the context of a displacement-based
design method, to develop a simple model that directly
correlates the shear-stiffness degradation of a wall to its
top drift. Accordingly, the authors predicted the normalized
shear-stiffness degradation based on regression analysis.
The results in Fig. 15(a) generally reflect strong correla-
tion between normalized shear stiffness and drift ratio
(R2 = 0.87). Interpretation of the results also revealed that
normalized shear-stiffness degradation is a function of hori-
zontal web reinforcement ratio. As shown in Fig. 15(b), the
degradation decreased almost linearly with increasing rein-
forcement ratio at the same drift level. The relation appears,
however, to have significantly deviated with increasing drift.
This is clear in Fig. 15(c) and (d), which show the nonlinear
variation of relation constants A and B with the drift ratio,
indicating the interrelation between the drift ratio and hori-
zontal web reinforcement ratio.
Based on this discussion, the normalized shear-stiffness
degradation was set as a function of top drift ratio and hori-
zontal web reinforcement ratio as follows
Ks
= a ⋅ δ x ⋅ ρ + b ⋅ δ y (14)
K se
where Ks is the secant shear stiffness at a lateral top drift
equal to δ; Kse is the elastic shear stiffness and can be calcu-
lated with Eq. (13); ρ is the horizontal web reinforcement
ratio; and a, b, x, and y are constants. For the experimental
data, the coefficients a, b, x, and y, which correlate between
experimental and analytical results, were found to be equal to
0.04, 0.03, –1, and –1.6, respectively. Therefore, the normal-
ized shear stiffness (Ks/Kse) can be rewritten as follows
K s 0.04ρ 0.03
= + 1.6 (15)
K se δ δ
Figure 16 shows good agreement between the exper-
imental results and the proposed formula. More testing is
needed, however, to clarify the effect of other parameters
ACI Structural Journal/January 2018
 Fig. 15—Validation of proposed model for shear-stiffness
Fig. 14—Shear-stiffness degradation normalized with gross degradation.
shear stiffness versus drift ratio.
such as aspect ratio, axial loading, and the presence of
boundary elements on the normalized shear stiffness.

FLEXURAL STIFFNESS
In practice, code methods employ an effective flexural
stiffness to consider the reduction in stiffness caused by
cracking for reinforced concrete elements. In designing
steel-reinforced walls, a single reduction factor (α) is
commonly applied to the gross stiffness to reduce it to effec-
tive stiffness. For instance, the effective flexural stiffness for
a wall under zero axial load is proposed to be 25%, 35%, and
50% of the gross flexural stiffness by NZS 3101 (1995), ACI
318-14, and Eurocode 8 (2004), respectively. CSA A23.3-14
links α to the ductility- and overstrength-related force modi-
fication factors (Rd and Ro) as follows

α = 1 – 0.35(RdRo/γw – 1), 0.5 ≤ α ≤ 1 (16)

where γw is the wall overstrength factor and can be taken

equal to Ro. According to CSA A23.3-14 and NBCC (2010),
Rd and Ro for moderately ductile steel-reinforced squat walls Fig. 16—Comparison of predicted normalized shear stiff-
are 2 and 1.4, respectively. Substituting these values in ness (Eq. (5)) with experimental data.
Eq. (16) results in α = 0.65.
the same reduction factors to estimate the effective flex-
As mentioned earlier, the measured secant flexural stiff-
ural stiffness of the GFRP-reinforced squat walls. This was
ness for specimen S4-80 was almost 30% of the gross flex-
expected and can be attributed to the different mechanical
ural stiffness corresponding to steel yielding. Comparing the
and bond characteristics between steel and GFRP bars.
value at yielding with the aforementioned suggestions for
Instead, the methods for estimation of flexural stiffness for
stiffness-reduction values reveals that Eurocode 8 (2005)
FRP-reinforced elements specified in ACI 440.1R-15 and
and CSA A23.3 (2014) overestimated α by 67% and 116%,
CSA S806-12 were employed.
respectively, while ACI 318 (2014) and NZS 3101 (1995)
ACI 440.1R-15 provides an expression for estimating the
yielded better estimations, with differences of 17% higher
effective moment of inertia (Ie) to calculate the deflection of
and lower than the measured stiffness, respectively. Based
FRP-reinforced concrete elements; the formula is given as
on Fig. 9(b), however, it would appear inappropriate to use
follows

ACI Structural Journal/January 2018 219

Fig. 17—Predicted flexural displacement and normalized flexural stiffness versus experimental data.

I cr
Ie = ≤ I g (17)
1 − γ ( M cr /M a ) 2 [1 − I cr /I g ]

where Icr is the cracked moment of inertia; γ is a function of

the cracking moment (Mcr) relative to the applied moment
(Ma) and equal to [3 – 2(Mcr/Ma)] for a cantilever with
end-point load; and Ig is the gross moment of inertia.
The method proposed in CSA S806-12 is based on inte-
grating the moment-curvature relation, which is assumed to
stay linear under increased loading with a flexural rigidity
of EcIg in the uncracked stage and EcIcr in the cracked stage,
whereas there is no tension stiffening in the cracked zone
of a beam element. The effective moment of inertia can be
calculated as follows
Fig. 18—Normalized flexural stiffness versus drift ratio.

I cr together in Fig. 17 along with the results obtained experimen-

Ie = ≤ I g (18)
1 − η( Lg /L)3
where η equals (1 – Icr/Ig); and Lg is the distance from the
free end to the point at which M = Mcr.
Knowing the effective moment of inertia (Eq. (17) and
(18)), the flexural deformations at any load level can be
calculated using the elastic deflection equation as follows
P ⋅ x2
δ= (3l − x) (19)
6 EI e
where P is the applied load; δ is the lateral displacement; E
is the elastic modulus for concrete, which can be calculated
as 4700√fc′ MPa (57,000√fc′ psi); x is the distance from the
support to the point at which the displacement is calculated
and equal to 1500 mm (59 in.); and l is the shear span equal
to 2550 mm (100.4 in.).
The predicted flexural deformations using the effective
moment of inertia specified by ACI 440.1R-15 and CSA
S806-12 were calculated at different load levels and plotted
tally. Using the calculated flexural deformations, the secant
flexural stiffness at different drift levels was also calculated
and normalized to the gross flexural stiffness and plotted
in Fig. 17. It should be noted that calculations were made
using the experimentally obtained cracking moment corre-
sponding to the first flexural crack, as given in Table 2. It can
be inferred from the figure that using either ACI 440.1R-15
or CSA S806-12 in calculating flexural deformation and the
consequent normalized flexural stiffness yielded good agree-
ment with the experimental results, although CSA S806-12
appears to be more conservative.
Similar to shear stiffness and based on regression analysis,
the normalized flexural stiffness could be formulated as in
Eq. (20). Figure 18 verifies the proposed equation
Kf/Kfe = 0.1δ–0.6 (20)
CONCLUSIONS
The main purpose of this research was to evaluate the
flexural and shear stiffness of GFRP-reinforced concrete

220 ACI Structural Journal/January 2018

squat walls. The evaluation process and its outcomes can be
concluded as follows:
1. Decoupling flexural and shear deformation without
considering curvature variation can lead to overestimated
deformation, while correcting the deformations based on the
estimated center of rotation produced consistent results.
2. Shear deformation contributed significantly to the total
deformation and cannot be omitted, even if the shear strength
is greater than the applied load.
3. The cracked shear stiffness of GFRP-reinforced
concrete squat walls can be estimated based on the truss
model with an acceptable level of conservatism. To ensure
this conservatism, the shear-crack angle should be estimated
based on CSA S806-12 and the concrete shear strength is
recommended to be equal to 0.12bwlwfc′0.5.
4. In calculating flexural deformation and the consequent
normalized flexural stiffness, both ACI 440.1R-15 and CSA
S806-12 yielded results in good agreement with the exper-
imental results, although CSA S806-12 appears to be more
conservative.
5. Within the context of displacement-based design method,
a simple model that correlates the flexural and shear stiffness
degradation of the test walls to their drift was proposed.
AUTHOR BIOS
Ahmed Arafa is a Doctoral Candidate in the Department of Civil Engi-
neering at the University of Sherbrooke, Sherbrooke, QC, Canada. He
received his BSc and MSc degrees from Assiut University, Assiut, Egypt.
His research interests include testing of concrete structures reinforced with
fiber-reinforced polymers.
Ahmed Sabry Farghaly is a Postdoctoral Fellow in the Department of
Civil Engineering at the University of Sherbrooke. His research interests
include nonlinear analysis of reinforced concrete structures and behavior of
structural concrete reinforced with fiber-reinforced polymers.
Brahim Benmokrane, FACI, is Professor of civil engineering and NSERC
Research Chair in FRP Reinforcement for Concrete Infrastructure and
Tier-1 Canada Research Chair in Advanced Composite Materials for Civil
Structures in the Department of Civil Engineering at the University of Sher-
brooke. He is a member of ACI Committee 440, Fiber-Reinforced Polymer
Reinforcement.
ACKNOWLEDGMENTS
The experimental study was conducted with funding from the Tier-1
Canada Research Chair in Advanced Composite Materials for Civil Struc-
tures, and Natural Sciences and Engineering Research Council of Canada
(NSERC-Industry Research Chair program), and the Fonds de recherche du
Québec – Nature et Technologie (FRQ-NT). The assistance of the technical
staff of the new Canadian Foundation for Innovation (CFI) structural lab
at the University of Sherbrooke’s Department of Civil Engineering is also
acknowledged.
ACI Structural Journal/January 2018
REFERENCES
ACI Committee 318, 2014, “Building Code Requirements for Struc-
tural Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 519 pp.
ACI Committee 440, 2015, “Guide for the Design and Construction of
Concrete Reinforced with FRP Bars (ACI 440.1R-15),” American Concrete
Institute, Farmington Hills, MI, 44 pp.
Arafa, A.; Farghaly, A. S.; and Benmokrane, B., 2016, “Experimental
Investigation of GFRP-Reinforced Squat Wall Subjected to Lateral Load,”
7th International Conference on Advanced Composite Materials in Bridges
and Structures (ACMBS-VII), Vancouver, BC, Canada, 10 pp.
ASCE/SEI 43, 2005, “Seismic Design Criteria for Structures, Systems,
and Components in Nuclear Facilities (ASCE/SEI 43-05),” American
Society of Civil Engineers, Reston, VA.
CSA A23.3-14, 2014, “Design of Concrete Structures Standard,” Cana-
dian Standards Association, Mississauga, ON, Canada, 240 pp.
CSA S6-14, 2014, “Canadian Highway Bridge Design Code,” Canadian
Standards Association, Mississauga, ON, Canada, 732 pp.
CSA S806-12, 2012, “Design and Construction of Building Components
with Fiber-Reinforced Polymers,” Canadian Standards Association, Missis-
sauga, ON, Canada, 208 pp.
Eurocode 8, 2004, “Design of Structures for Earthquake Resistance, Part
1: General Rules, Seismic Actions and Rules for Buildings, ENV 1998-
1:2003,” Comité Européen de Normalisation, Brussels, Belgium.
Hiraishi, H., 1984, “Evaluation of Shear and Flexural Deformations of
Flexural Type Shear Walls,” Bulletin of the New Zealand National Society
for Earthquake Engineering, V. 17, No. 2, pp. 135-144.
Li, B., and Xiang, W., 2011, “Effective Stiffness of Squat Structural
Walls,” Journal of Structural Engineering, ASCE, V. 137, No. 12, pp. 1470-
1479. doi: 10.1061/(ASCE)ST.1943-541X.0000386
Luna, B. N.; Rivera, J. P.; and Whittaker, A. S., 2015, “Seismic Behavior
of Low-Aspect-Ratio Reinforced Concrete Shear Walls,” ACI Structural
Journal, V. 112, No. 5, Sept.-Oct., pp. 593-604. doi: 10.14359/51687709
Massone, L. M.; Orakcal, K.; and Wallace, J. W., 2009, “Modeling
of Squat Structural Walls Controlled by Shear,” ACI Structural Journal,
V. 106, No. 5, Sept.-Oct., pp. 646-655.
Mohamed, N.; Farghaly, A. S.; Benmokrane, B.; and Neale, K. W.,
2014a, “Experimental Investigation of Concrete Shear Walls Reinforced
with Glass-Fiber-Reinforced Bars under Lateral Cyclic Loading,” Journal
of Composites for Construction, ASCE, V. 18, No. 3.
Mohamed, N.; Farghaly, A. S.; Benmokrane, B.; and Neale, K. W.,
2014b, “Flexure and Shear Deformation of GFRP-Reinforced Shear Walls,”
Journal of Composites for Construction, ASCE, V. 18, No. 2.
NBCC, 2010, “National Building Code of Canada,” Canadian Commis-
sion on Building and Fire Codes, National Research Council of Canada,
Montreal, QC, Canada
NZS 3101, 1995, “Code of Practice for the Design of Concrete Struc-
tures, Part 1,” Standards New Zealand, Wellington, New Zealand.
Oesterle, R. G.; Aristizabal-Ochoa, J. D.; Fiorato, A. E.; Russell, H. G.;
and Corley, W. G., 1979, “Earthquake Resistant Structural Walls—Test of
Isolated Walls—Phase II.” Report ENV77-15333, National Science Foun-
dation, Arlington, VA.
Park, R., and Paulay, T., 1975, Reinforced Concrete Structures, John
Wiley & Sons, Inc., New York.
Paulay, T.; Priestley, M. J. N.; and Synge, A. J., 1982, “Ductility in
Earthquake Resisting Squat Shear Walls,” ACI Journal Proceedings, V. 79,
No. 4, Apr., pp. 257-269.
Salonikios, T. N.; Kappos, A. J.; Tegos, I. A.; and Penelis, G. G., 1999,
“Cyclic Load Behavior of Low-Slenderness Reinforced Concrete Walls:
Design Basis and Test Results,” ACI Structural Journal, V. 96, No. 4,
July-Aug., pp. 649-660.
Tang, T. O., and Su, R. K. L., 2014, “Shear and Flexural Stiffnesses of
Reinforced Concrete Shear Walls Subjected to Cyclic Loading,” The Open
Construction and Building Technology Journal, V. 8, pp. 104-121. doi:
10.2174/1874836801408010104
221
ARE YOU A RESEARCHER?
SIGN UP FOR TODAY!
ORCID provides a persistent digital identifer that distinguishes you
from every other researcher and, through integration in key research
workflows such as manuscript and grant submission, supports
automated linkages between you and your professional activities,
ensuring that your work is recognized.
Individuals may use ORCID services freely and it’s as easy as 1-2-3:

1 REGISTER

2 ADD YOUR INFO

3 USE YOUR ORCID ID

For more information and to register, visit:

WWW.ORCID.ORG
 ACI STRUCTURAL JOURNAL
Title No. 115-S18
Long-Term Multipliers and Deformability of Fiber-
Reinforced Polymer Prestressed Concrete
by Yail J. Kim and Raymon W. Nickle
This paper presents the flexural characteristics of highway bridge
girders prestressed with fiber-reinforced polymer (FRP) tendons.
Of interest are the technical challenges identified by ACI Subcom-
mittee 440-I (FRP-Prestressed Concrete): long-term multipliers,
deformability, and minimum reinforcement. Aramid and carbon
FRP (AFRP and CFRP, respectively) composites are used to
prestress concrete girders. Based on analytical models, new design
expressions are proposed for the aforementioned items, followed by
an assessment using laboratory test data and full-scale benchmark
bridges. The long-term multipliers calibrated per reliability theory
are mostly different from the empirical multipliers adopted in
ACI 440.4R-04. The girders prestressed with AFRP/CFRP suffi-
ciently deform in flexure, even though their moment-curvature
responses are not comparable with those of steel-prestressed
girders. The new deformability index specifies design require-
ments for AFRP/CFRP-prestressed members with either compres-
sion-controlled or tensioned-controlled sections. The importance
of a potential change in FRP modulus during the service life of
prestressed concrete girders is examined. A factored ultimate-to-
cracking moment ratio of ϕMn/Mcr = 1.2 is suggested for girders
prestressed with AFRP/CFRP, which aligns with the articles of
existing design manuals and specifications dedicated to prestressed
concrete.
Keywords: bridge; deformability; fiber-reinforced polymer (FRP); long-
term multiplier; minimum reinforcement; prestress.
INTRODUCTION
To increase the durability of prestressed concrete bridges,
it has been proposed to use advanced composites instead
of conventional steel strands.1,2 Fiber-reinforced polymers
(FRP), comprising synthetic fibers and a resin matrix, are
promising materials with numerous advantages—namely,
low density at high strength, noncorrosiveness, and noncon-
ductance.3 As prestressing tendons, aramid FRP (AFRP)
and carbon FRP (CFRP) are used in preference. Glass
FRP (GFRP) is inadequate for prestressing applications
because of unfavorable creep resistance.4 Nanni and Tani-
gaki5 showed one of the early endeavors in FRP-prestressed
concrete. After testing 21 beams with several variables
(tendon size, prestress level, and shear reinforcement), it was
concluded that FRP was usable for prestressing concrete;
however, further development was recommended. Grace and
Abdel-Sayed6 proposed a prestressing concept for highway
bridges with a combination of bonded and unbonded CFRP
tendons. A reduced-scale prototype bridge superstructure
was constructed and tested in static, dynamic, and cyclic
loadings. The performance of CFRP was satisfactory; for
example, no visual damage was noticed after 7 million
fatigue cycles. Dolan and Swanson7 examined the failure
ACI Structural Journal/January 2018
TECHNICAL PAPER
characteristics of vertically distributed CFRP tendons, when
the prestressed concrete beams were subjected to bending.
Unlike steel strands, the stress development of CFRP was
substantially influenced by the degree of curvature that led
to the progressive rupture of the distributed tendons. Saiedi
et al.8 evaluated the performance of CFRP-prestressed
concrete beams in an aggressive environment. Fatigue
loading at a low temperature of –28°C (–18°F) degraded the
beams’ stiffness and camber, which accompanied the slip
of the CFRP. Zhang et al.9 reported the axial response of
multiple CFRP tendons for prestressing concrete members.
Parameters affecting the CFRP behavior were tendon stiff-
ness, anchoring methods, and geometric properties. Further
details about FRP-prestressed members, such as bond and
transfer length, anchorage, durability, and field applications
are available in review papers.10,11
Concerning long-term deflections for FRP-prestressed
concrete, empirical multipliers are exploited in current
practice.3 The origin of these multipliers can be found in an
analysis using steel-prestressed concrete, reported over 40
years ago.12 Although the developer clarified the inherent
unpredictability of long-term deflections, the multipliers are
adopted in most prestressed concrete applications, including
those with FRP tendons, owing to the absence of relevant
research.
The deformability of FRP-prestressed members has
been studied by a few researchers. It is important to note
that the concept of ductility in traditional concrete struc-
tures is not valid for FRP-prestressed members due to
FRPs’ nonyielding nature; accordingly, an alternative term
called “deformability” is employed. Zou13 tested concrete
beams prestressed with AFRP and CFRP tendons, and
compared their behavior in terms of curvature development,
sectional rotation, and displacement. A deformability factor
was proposed by combining displacement and moment at
cracking and ultimate stages of the prestressed beams. The
applicability of this factor, however, appears to be limited
because rigorous computation is necessary to determine
the beams’ deflections. Au and Du14 developed a numerical
model to predict the flexural behavior of FRP-prestressed
beams. A comparative study was performed to assess the
variation trend of existing deformability expressions with
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2017-054, doi: 10.14359/51700988, received February 20, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
223
reinforcement ratio. Because the expressions were inde-
pendently developed, meaningful conclusions were not
accomplished. ACI 440.4R-043 specifies a deformability
equation, whereas its applicability is practically restricted
due to the facts that: 1) there are no acceptable/unacceptable
limits in deformability; and 2) only compression-controlled
failure is considered, although ACI 440.4R-043 states
another failure mode (tension-controlled). Overall, a unified
expression for deformability with specific application limits
is necessary to stipulate the allowable amount of flexural
deformation in designing FRP-prestressed concrete girders.
This paper aims to address the aforementioned chal-
lenges in FRP-prestressed concrete technologies: long-term
multipliers for deflection prediction and deformability. Also
involved is a study on minimum reinforcement ratio for
FRP-prestressed bridge girders that have been undervalued
in the infrastructure community.
RESEARCH SIGNIFICANCE
ACI Subcommittee 440-I (FRP-Prestressed Concrete)
has recommended several technical items to be updated in
a revised version of ACI 440.4R-04.3 Three crucial issues
related to the flexural design of FRP-prestressed members
were identified (long-term multipliers, deformability, and
minimum reinforcement), because of their largely empir-
ical, ambiguous, and discrepant backgrounds. For instance,
the present long-term multipliers were taken from steel-
prestressed concrete despite the fundamental difference
between FRP and steel materials. Research is conducted to
suggest new design information, which can replace/comple-
ment the existing design articles of ACI 440.4R-04.3
LONG-TERM MULTIPLIERS
A stochastic simulation is performed to propose long-
term multipliers for AFRP- and CFRP-prestressed concrete
bridge girders, which can supersede those of ACI 440.4R-043
determined by arbitrary empirical assumptions. Delineated
in the following are girder configurations and numerical
approaches that lead to the development of new multipliers.
Benchmark bridges
Pursuant to the American Association of State Highway
Transportation Officials (AASHTO) Load and Resistance
Factor Design (LRFD) Bridge Design Specifications,15 Kim
and Nickle16 designed 50 prestressed concrete bridges using
AFRP and CFRP tendons (25 bridges each). A summary
of the bridge details is provided as follows. Four super-
structure types were employed with variable span lengths
and deck widths, as listed in Table 1. These girder config-
urations were adopted from highway bridges constructed
in Colorado, Florida, and Texas. The compressive strength
of concrete (fc′) for the girders and decks was from 38 to
72 MPa (5.5 to 10.5 ksi) and from 25 to 31 MPa (3.6 to
4.5 ksi), respectively. The AFRP and CFRP tendons
embedded in the concrete, respectively, had the following
properties: tensile strength = 1250 and 2250 MPa (181 and
326 ksi); and elastic modulus = 65 and 140 GPa (9400 and
20,600 ksi). The AFRP and CFRP tendons were tensioned to
224
625 and 1460 MPa (90 and 212 ksi), respectively, in compli-
ance with ACI 440.4R-04.3
Simulation approach
Description of multiplier components—Martin12 proposed
a simple method to determine multipliers for predicting
long-term deflections of prestressed concrete structures.
That method is outlined below with some modifications to
adjust the original method when necessary. Initial calcula-
tions for the prestressed concrete deflection multipliers are
based on an equation from ACI 318-71,17 describing the
long-term deflections of reinforced concrete beams, which
is no longer in use.18 As such, the relationship between long-
term deflections and immediate deflections is adopted from
an updated version. Equation (1) provides the basis for the
deflection multiplier (μb) in accordance with ACI 318-1418
ξ
mb = (1)
1 + 50ρ′
where ξ is a time-dependent constant varying from 1.0 to
2.0; and ρ′ is the compression reinforcement ratio at midspan
of the member, regardless of boundary conditions. For the
present study, three reinforcement ratios were used to repre-
sent typical bridge design (that is, ρ′ = 0.001, 0.005, and
0.01) in tandem with ξ = 2.0 (5 years or more). For long-term
deflections from dead load (μdf), the base factor is modified
by the strength gain in the concrete, based on the elastic
moduli at release (Eci) and 28 days (Ec), respectively
Eci
m df = m b (2)
Ec
Upward displacements for the final camber or deflec-
tions are defined as a function of the prestressing force (μpf).
Initial displacements are calculated at release and are modi-
fied with the final effective prestressing force
Pfe
m pf = m df (3)
Po
where Po is the prestressing force immediately following
release; and Pfe is the effective prestressing force. The
losses of prestressing force may be determined as detailed in
ACI 440.4R-04,3 whose methods are analogous to those of
the AASHTO LRFD Bridge Design Specifications.15 Erec-
tion cambers and deflections may be defined as a portion
of the aforementioned final factor multipliers. Martin12
arbitrarily scaled down the μdf factor by 50% to define the
amount of deflection at erection (μde). In the current calibra-
tion, however, a variable is taken into consideration for the
μde factor, which leads to the deflection from dead load at the
time of erection to be
μde = κμdf (4)
where κ is the erection-adjustment factor to account for creep
and shrinkage at an early stage of the prestressed member.
ACI Structural Journal/January 2018
Table 1—Benchmark bridges
ID Type Span, m (ft) Span-depth ratio Girder number Deck width, m (ft) Skew angle, deg
1 B 17 (56) 27 20 25 (82) 0
2 B 17 (56) 27 20 25 (82) 0
3 U 27 (87) 13 4 17 (55) 0
4 B 33 (107) 30 19 27 (90) 40
5 B 33 (108) 30 19 27 (90) 40
6 U 33 (109) 19 4 22 (71) 5
7 U 33 (109) 19 4 22 (71) 5
8 BT 34 (111) 27 7 13 (43) 27
9 BT 36 (119) 23 8 16 (52) 26
10 U 37 (120) 18 4 17 (55) 0
11 U 37 (121) 23 3 9 (29) 8
12 I 38 (124) 21 16 48 (157) 0
13 I 38 (124) 21 16 48 (157) 0
14 B 38 (125) 28 12 20 (67) 0
15 U 38 (125) 19 7 39 (127) 25
16 U 38 (126) 19 8 37 (120) 0
17 B 39 (127) 28 12 20 (67) 0
18 U 39 (127) 19 7 39 (127) 25
19 U 39 (129) 21 8 37 (120) 0
20 B 40 (132) 29 10 13 (43) 56
21 B 41 (135) 30 24 46 (152) 20
22 B 41 (135) 30 24 46 (152) 20
23 U 44 (143) 22 4 17 (55) 0
24 U 44 (144) 22 8 37 (120) 0
25 BT 50 (164) 21 4 9 (28) 30

Notes: ID is identification of bridges; Type (B is box; U is open box; BT is bulb tee; I is I-shape).

Previous research reports that the adjustment factor κ ranges composite multipliers for the load (μdfc) and prestressing
between 0.4 and 0.619; accordingly, random sampling was effects (μpfc), respectively
carried out within this range (details to follow). The camber
component during construction is a multiplier that uses I 
the erection dead load factor as a basis, combined with the m dfc = m de + (m df − m de )  o  (7)
 Ic 
average prestressing force assuming a linear relationship
between pretensioning release and the final effective force.
The upward component from prestressing at erection (μpe) is I 
m pfc = m pe + (m df − m pe )  o  (8)
then defined as  Ic 

Po + Pf e
m pe = m de (5) where Io and Ic are the moments of inertia for noncom-
2 Po posite and composite sections, respectively. For the long-
term deflection of the member induced by placing topping
For the application of long-term superimposed dead loads, concrete, the long-term factor (μt) is expressed as
the base factor may be used to describe the long-term effects
(μsd), as stated in Martin12 I 
m t = m sd  o  (9)
 Ic 
μsd = μb (6)
It is important to note that these factors are intended to be
The use of composite construction is common for precast
additive to the initial deflections calculated; hence, they are
girders. Because composite construction alters the stiffness
increased by 1.0 to be used as direct multipliers, as summa-
of the structural system, the factors shown in Eq. (2) through
rized in Table 2. The factors discussed previously assume
(6) need to be modified. Equations (7) and (8) provide

ACI Structural Journal/January 2018 225

that the composite topping is constructed with a thickness • Statistical distributions: Probability distribution types
of 50 to 75 mm (2 to 3 in.); consequently, the reinforcement are assigned to the variables associated with the sampled
ratio of the system does not change significantly (denoted properties, including statistical characteristics such as
as “with unreinforced composite topping”). By contrast, the coefficients of variation (COVs) shown in Table 3.
most bridge girders are in service with a concrete deck with • Property simulation: Random variables are generated
a noticeable amount of reinforcement, which results in the as per the statistical properties determined in previous
need to consider composite compression reinforcement steps.
ratios (the ρ′ ratios explained previously). For this reason, an • Multiplier sampling based on Monte-Carlo simula-
additional set of multipliers referred to as “with reinforced tion: Using the methodology discussed earlier (Eq. (1)
structural slab” is proposed and determined in which the to (9)) and the variable properties attained, the deflec-
increased reinforcement ratio is accounted. tion multipliers of the noncomposite/composite girder
Random variables and simulation—The multipliers for systems are iteratively simulated until convergence is
long-term deflections enumerated earlier involve several accomplished (that is, the model prediction with random
variables that can be simulated stochastically. A numer- samples approaches a specific value).
ical method called Monte-Carlo simulation that randomly
samples statistical data from a certain probability distri-
bution was used to predict the variability of the bridge
systems. In other words, the simulation technique allows
for the modeling of numerous possible situations in the
performance of individual bridge girders. As an example,
Fig. 1 shows sampled values for concrete strength (fc′) and
prestressing force ratio (Pfe/Po). Following is a summary of
the simulation procedures:
• Property sampling: Technical information is acquired
from the 25 bridges listed in Table 1, including geometric
and material properties that will form the basis for the
statistical means.

Table 2—Deflection multipliers variable summary

with and without composite topping
Stage Component Variable
Deflection (S) 1 + μde
Erection
Camber (P) 1 + μpe
Deflection (S) 1 + μdf
Camber (P) 1 + μpf
Final
Deflection (SD) 1 + μsd
Deflection (CT) 1 + μt*
*
Valid only with composite topping. Fig. 1—Simulated properties for Bridge No. 5: (a) concrete
Notes: Deflection is downward; camber is upward. S is girder self-weight; P is strength fc′; and (b) prestressing force ratio Pfe/Po for AFRP
prestress; SD is superimposed dead load; CT is composite topping. tendons.

Table 3—Random variables

Variable Distribution Mean Bias COV Reference
fci′ Normal Based on material 1.14 to 1.40 0.1 Kwon et al.26
fc′ Normal Based on material 1.14 to 1.40 0.1 Kwon et al.26
Eci Normal 1820√fci′ (in ksi) 1 0.1 AASHTO15
Ec Normal 1820√fci′ (in ksi) 1 0.1 AASHTO15
Efrp Lognormal Based on material — 0.2 Atadero et al.27
Ep Lognormal 28,500 ksi 1.04 0.024 Kwon et al.26
Afrp Normal Based on material — 0.05 Sanchez-Heres et al.28
Ap Normal As designed — 0.0125 Kwon et al.26
d, de, and b Normal As designed — 0.03 Okeil et al.29

Note: 1 ksi = 6.9 MPa.

226 ACI Structural Journal/January 2018

Table 4—Long-term multipliers
AFRP CFRP
Without composite With unreinforced With reinforced Without composite With unreinforced With reinforced
Stage Component topping composite topping structural slab topping composite topping structural slab
Deflection (S) 1.85 1.85 1.85 1.85 1.85 1.85
Erection
Camber (P) 1.70 1.70 1.70 1.80 1.80 1.80
Deflection (S) 2.70 2.35 2.35 2.70 2.35 2.30
Camber (P) 2.00 1.90 1.85 2.55 2.25 2.20
Final
Deflection (SD) 2.90 2.90 2.85 2.90 2.90 2.85
Deflection (CT) — 2.05 2.05 — 2.05 2.05

Notes: Deflection is downward; camber is upward; S is girder self-weight; P is prestress; SD is superimposed dead load; CT is composite topping.

Fig. 3—Comparison of proposed long-term multipliers

Fig. 2—Calculated multipliers without composite topping:
(a) deflection at erection of AFRP-prestressed girders; and
(b) deflection at final of CFRP-prestressed girders.
• Deflection multipliers: Upon taking the average of
the simulated multiplier samples, representative deflec-
tion multipliers are determined for AFRP- and CFRP-
prestressed concrete girders with the three topping
conditions: without composite topping, with unrein-
forced composite topping, and with reinforced struc-
tural slab.
Recommended multipliers—After solving multipliers
for all 50 bridge cases with AFRP and CFRP tendons, the
results were sorted, as shown in Fig. 2. For design purposes,
conservative multipliers were selected and proposed (Table
4). Figure 3 compares the proposed multipliers with the
existing ones in ACI 440.4R-043 for both AFRP- and CFRP-
prestressed bridge girders. It should be noted that the compar-
ison focuses only on the cases without composite topping,
against ACI 440.4R-04.
because ACI 440.4R-043 does not specify multipliers with
composite topping. The deflection multipliers at Erection
and Final did not show any difference between the new and
existing multipliers. Substantial dissimilarity was, however,
observed in all other cases, particularly for the multipliers
of camber and deflection (superimposed dead load) at Final.
This discrepancy is attributable to the different material
characteristics between the prestressing steel and the AFRP/
CFRP composite tendons used in the stochastic simulation.
The multipliers suggested in Table 4 verify the qualitative
statement of ACI 440.4R-04,3 which signifies that the multi-
pliers without topping are conservative compared with those
with composite topping.
DEFORMABILITY
The degree of flexural deformation in FRP-prestressed
concrete is quantified and assessed to propose a design
expression. The formulation elaborated as follows is valid
for bonded applications that satisfy the strain compatibility
of a section. The application of unbonded FRP is not taken
into account, owing to the wide variability of a strain reduc-
tion factor, depending upon the geometry of a prestressed
member.
Derivation
Equation (10) defines a newly proposed deformability
index (DI) for an FRP-prestressed member
ε ultf
DI = (10)
ε crf

ACI Structural Journal/January 2018 227

where εfult and εfcr are the FRP strains at ultimate and Table 5—Lower limits of deformability index (DI)
cracking of the member, respectively. In accordance with Condition AFRP CFRP
strain compatibility for a bonded section, these prestressed
Tension-controlled 2.8 1.8
FRP strains consist of three components
Compression-controlled 1.5 1.5
εfult = εfe + εce + εfail (11)
ε fe + (d f /c − 1)ε cu
εfcr = εfe + εce + εfcr (12) DI = for compression-controlled section
(d f − c) f r
ε fe +
εfe = ffe/Ef (13) h − c Ef
(17b)
Pfe ( I g + e 2f Ag )
ε ce = (14) Equations (17a) and (17b) are further manipulated for
Ag I g Ec
practical implementation. Because the modulus of rupture
of concrete is much smaller than the tensile modulus of FRP,
where εfe is the effective FRP strain after prestress losses; the fr/Ef term is ignored. Substituting (df/c –1) for k gives
εce is the additional FRP strain when the cambered member
is decompressed (concrete strain at the FRP level becomes ε fu
zero under flexural loading, which results in the neutral posi- DI = for tension-controlled section (18a)
tion of the cross section); εfail and εfcr are the FRP strains ε fe
measured from the neutral position to failure and to cracking,
respectively; ffe and Ef are the effective stress and modulus of k ε cu
the FRP, respectively; Ig and Ag are the moment of inertia and DI = for compression-controlled section (18b)
ε fe
cross-sectional area of the uncracked member, respectively;
and ef is the distance from the neutral axis of the section to
the center of gravity of the FRP. The εfail component is deter- Application
mined as per the failure mode of the prestressed member, Conventional prestressed concrete beams can sufficiently
including the conventional similar triangles of strain devel- deform under flexural loading, if the following limit is
opment in compression and tension achieved20

εfail = εfu for tension-controlled section (15a) c

≤ 120ε cu (19)
h
(d f − c)
ε fail = ε cu for compression-controlled section (15b) Given that prestressed concrete members should exhibit
c acceptable flexural deformation, regardless of prestressing
material, Eq. (19) can be used for FRP-prestressed concrete.
where εfu is the ultimate FRP strain; c and df are the neutral Substituting Eq. (19) into Eq. (18b) yields
axis depth of the section and the effective depth of the FRP,
respectively, measured from the extreme compression fiber d f / (120h) − ε cu + ε fe
of the member; and εcu is the maximum useable concrete DI ≥ (20)
ε fe
strain (εcu = 0.003 [from Reference 18]). The εfcr component
is also attained from the similar triangles of the uncracked For a typical prestressed concrete section, the ratio
section’s stress profile between the effective depth and height of the section is
0.85.20 ACI 440.4R-043 specifies that AFRP and CFRP
(d f − c) f r tendons are tensioned up to 0.4ffu and 0.6ffu, respectively,
ε crf = (16)
h − c Ef including transfer losses. Because the long-term prestress
losses of FRP-prestressed concrete members may be esti-
where h is the height of the member; and fr is the modulus of mated by the methods for steel-prestressed concrete,3 the
rupture of concrete (fr = 7.5√fc′ in psi and 0.63√fc′ in MPa18). lump-sum prestress losses provided in the AASHTO LRFD
Combining Eq. (10) to (16) with the fact that the εce term is Bridge Design Specifications15 are adopted, which are less
negligible in most cases, the following expressions are obtained than 10% of the ultimate tendon strength. As a result, the
long-term prestress losses of FRP-prestressed concrete are
ε fu assumed to be 0.05ffu. It is worth noting that this long-term
DI = for tension-controlled section (17a) loss assumption is reasonable because FRP-prestressed
(d f − c) f r
ε fe + concrete experiences less prestress losses than its steel coun-
h − c Ef
terpart, due to low FRP modulus.3 Considering all these
components, the effective strains of AFRP and CFRP may

228 ACI Structural Journal/January 2018


Fig. 4—Prestressed concrete girder details: (a) dimension in
mm (midspan); and (b) prestressing properties. (Note: 1 mm =
0.0394 in.; 1 kN = 0.225 kip.)
be 35% and 55% of their ultimate strains, respectively.
Following ACI 440.4R-04,3 the lower bound of ultimate
strains in typical AFRP and CFRP tendons may be taken as
0.015. The limits of the deformability indices provided in
Eq. (18) are then calculated, and summarized in Table 5 with
rounded values for design convenience.
Assessment
Benchmark girders—Two common prestressed concrete
bridge girders were selected from the PCI Bridge Design
Manual21 to assess the deformability index proposed earlier:
AASHTO Box Beam BIII-48 and AASHTO/PCI Bulb Tee
BT-72. Figure 4(a) and Table 6 show the girders’ geometric
and material properties. For numerical parametric investiga-
tions, the BIII-48 and BT-72 girders were used as the default
sections with representative AFRP and CFRP tendons
whose properties were taken from ACI 440.4R-04,3 as listed
in Table 6. The amounts of AFRP and CFRP tendons were
determined to provide the flexural capacities of the sections
as close as possible to the capacities of the steel-prestressed
sections, with a condition that tendon stresses did not exceed
their ultimate strengths when the girders failed by concrete
crushing (compression-controlled). It should be noted that
ACI Structural Journal/January 2018
the deformability of tension-controlled sections was not
evaluated with these girders, because the index can readily
be attained using Eq. 18(a) without rigorous computation.
A wide range of prestressing levels (immediately after
transfer) were considered, varying from 0.1ffu up to 0.4ffu
and 0.6ffu for the AFRP and CFRP tendons, respectively, in
compliance with the prestressing limits of ACI 440.4R-04,3
as shown in Fig. 4(b).
Flexural behavior—Conventional sectional analysis
was conducted to predict the moment-curvature behavior
of the benchmark girders at midspan. Typical trilinear and
bilinear responses are provided in Fig. 5(a) for the steel- and
AFRP-prestressed BT-72 girders, respectively. Because of
low prestressing levels, the cracking moments of the AFRP
girders were lower than that of the steel girder (for example,
the cracking moments of BT-72 prestressed with AFRP at
0.4ffu and 0.1ffu were 26% and 72% lower, respectively).
With a reduction in prestressing level of the AFRP tendon,
the girders’ energy dissipatiion up to failure increased
(Fig. 5(a), inset), which was obtained by integrating the
area under the individual moment-curvature curve. Figure
5(b) exhibits the variation of a ratio between the neutral axis
depth and height of the girders at failure (c and h, respec-
tively). While the c/h ratios went up with a prestressing
level, irrespective of FRP and structural types, the girders
with CFRP revealed higher ratios than those with AFRP.
The reason is that the CFRP’s stress development was not as
susceptible to girder bending as was the AFRP’s; hence, the
neutral axes of the CFRP girders were consistently deeper.
The c/h ratios of all girders were within the limit of 120εcu
(Eq. (19)). This indicates that the girders had sufficiently
deformed until failure occurred, even though the curva-
tures of the AFRP-prestressed girders were less developed
(lesser curvature) in comparison with the curvature of the
steel-prestressed girder (Fig. 5(a)). The ratios between the
ultimate and cracking moments of the girders with AFRP/
CFRP (Mn and Mcr, respectively) are plotted in Fig. 5(c),
including those of BT-72 and BIII-48 with steel for compar-
ison (the constant Mn/Mcr ratios of steel are attributable to
the behavior of the steel-prestressed case; Fig. 5(a)). As a
prestressing level in the AFRP/CFRP tendons increased, the
Mn/Mcr ratios asymptotically decayed down to the ratios of
the steel-prestressed cases. The BT-72 girders maintained
higher Mn/Mcr ratios compared with the BIII-48 girders in
all tendon types due to the former’s longer moment arm
(BIII-48’s height is 53% of BT-72’s; Table 6), which caused
higher moment-carrying capacity Mn. Figure 5(d) shows
deformability indexes for the AFRP- and CFRP-prestressed
girders. The decreasing trend of deformability was similar to
that of the Mn/Mcr ratios, whereas there was no remarkable
difference between the BT-72 and BIII-48 girders. It is there-
fore stated that the proposed deformability index (Eq. (18a)
and (18b)) can be universally applicable to bridge girders
prestressed with either AFRP or CFRP tendons.
Parametric study—To evaluate the effects of a potential
change in FRP modulus on the deformability of prestressed
concrete girders, a parametric study was conducted using
BT-72 and BIII-48 girders with the preceding AFRP and
CFRP tendons. This investigation is meaningful from appli-
229
Table 6—Geometric and material properties of prestressed concrete members
Property Box Beam (BIII-48) Bulb Tee (BT-72)
Height (h) 990 mm (39 in.) 1829 mm (72 in.)
Effective depth (d) 874 mm (34.42 in.) 1653 mm (65.08 in.)
Cross-sectional area of section (Ac) 524,515 mm (813 in. )
2 2
494,838 mm2 (767 in.2)
Moment of inertia (I) 70.2 × 109 mm4 (168,367 in.4) 227 × 109 mm4 (545,894 in.4)
Tendon eccentricity at midspan (e) 374 mm (14.71 in.) 754 mm (29.68 in)
Distance from centroid to top of section (yt) 500 mm (19.71 in.) 899 mm (35.4 in.)
Distance from centroid to bottom of section (yb) 490 mm (19.29 in.) 930 mm (36.6 in.)
Radius of gyration (r) 366 mm (14.4 in.) 678 mm (26.7 in.)
Concrete strength at initial prestress (fci′) 28 MPa (4000 psi) 40 MPa (5800 psi)
Specified concrete strength at 28 days (fc′) 34 MPa (5000 psi) 45 MPa (6500 psi)
Concrete elastic modulus at transfer (Eci) 25 GPa (3600 ksi) 30 GPa (4340 ksi)
Concrete elastic modulus (Ec) 28 GPa (4030 ksi) 32 GPa (4595 ksi)
Area of prestressing steel (Ap) 2863 mm2 (4.437 in.2) 4738 mm2 (7.344 in.2)
Area of AFRP tendon (AAFRP) 4516 mm2 (7 in.2) 8387 mm2 (13 in.2)
Area of CFRP tendon (ACFRP) 2581 mm2 (4 in.2) 4194 mm2 (6.5 in.2)
Elastic modulus of prestressing steel (Ep) 197 GPa (28,500 ksi) 197 GPa (28,500 ksi)
Elastic modulus of AFRP tendon (EAFRP) 62 GPa (8992 ksi) 62 GPa (8992 ksi)
Elastic modulus of CFRP tendon (ECFRP) 150 GPa (21,756 ksi) 150 GPa (21,756 ksi)
Ultimate strength of prestressing steel (fpu) 1860 MPa (270 ksi) 1860 MPa (270 ksi)
Yield strength of prestressing steel (fpy) 1675 MPa (243 ksi) 1675 MPa (243 ksi)
Ultimate strength of AFRP tendon (fAFRPu) 1200 MPa (174 ksi) 1200 MPa (174 ksi)
Ultimate strength of CFRP tendon (fCFRPu) 2550 MPa (370 ksi) 2550 MPa (370 ksi)

Fig. 5—Flexural behavior of prestressed concrete members: (a) moment curvature of BT-72 with AFRP; (b) ratio between
neutral axis depth and girder height at failure of girders; (c) ratio between ultimate and cracking moments; and (d) assessment
of deformability. (Note: 1 kN·m = 738 lb-ft.)

230 ACI Structural Journal/January 2018

Fig. 6—Effect of FRP modulus on deformability: (a) moment curvature of bulb tee (BT-72) with CFRP tendons; (b) variation
of FRP strain; (c) deformability index of bulb tee (BT-72); and (d) deformability index of box beam (BIII-48).
cation standpoints because FRP-prestressed concrete girders
are designed without considering any possibility of altered
FRP properties3 that may take place within a service period
of 75 years. Although FRP composites are known as durable
materials, recent research shows that their elastic moduli
vary with time when subjected to environmental/mechanical
distress.22,23 According to experimental observations,22,23
the variation range of FRP modulus was determined to be
±10% of the nominal value. Additionally, it was assumed
that the rupture strain of FRP is invariant with time as per the
time-dependent strain failure criteria of polymeric composite
materials.24
Figure 6(a) exhibits the flexural responses of CFRP- Fig. 7—Assessment of deformability using benchmark bridges.
prestressed BT-72 in conjunction with variable moduli from that the deformability of FRP-prestressed girders would
90%Efo to 110%Efo, where Efo is the nominal modulus of not cause structural problems, as long as: 1) the girders
the CFRP (Table 6). The precracking behavior of the girder are properly designed per the proposed design recommen-
was not influenced by modulus variations; however, the dations; and 2) the tendons’ actual moduli are greater than
cracking moment was affected. Because the contribution those used in design (this is generally not a concern, because
of the CFRP modulus to the concrete girder’s stiffness was manufacturers’ recommended properties are conservative).
negligible, the post-cracking stiffness of the girder remained It is, however, important to note that care should be exer-
unchanged. An interesting finding is that the reduced CFRP cised when FRP’s modulus is expected to degrade in certain
modulus caused a transition in the girder’s failure mode from service environments that can alter the failure mode of the
compression-controlled to tension-controlled, as shown in prestressed girders, as mentioned earlier.
Fig. 6(b). This is attributed to the fact that the decreased
modulus brought about an increase in CFRP strain under Applicability
the same level of tensile stress. Figures 6(c) and 6(d) reveal The deformability indexes of the full-scale prestressed
the effects of CFRP modulus on deformability of the BT-72 concrete bridges with AFRP and CFRP tendons (Table 1)
and BIII-48 girders with AFRP and CFRP tendons, respec- were calculated to examine the applicability of the proposed
tively. For both girders, the deformability indexes gradu- limits (Table 5), as shown in Fig. 7. Because the low-mod-
ally dwindled with the increased modulus of the tendons. ulus AFRP was more deformable in principle, the indexes of
Nonetheless, the indexes of all compression-controlled cases the bridges with AFRP were higher than those of the bridges
were higher than the limit of 1.5. These observations denote with CFRP. Unlike the bridges with compression-controlled

ACI Structural Journal/January 2018 231

Table 7—Ratio between ultimate and cracking capacities
Experimental* Experimental*
Reference Cracking Ultimate Failure †
Reference Cracking Ultimate Failure†
G&S 643.9 kN 2,443 kN T N&T 11,454 Nm 27,199 Nm T
Stoll 178 kN 554 kN T N&T 12,850 Nm 26,905 Nm T
T&S 85 kN 195 kN T N&T 12,256 Nm 26,932 Nm C
T&S 85 kN 185 kN T N&T 11,825 Nm 27,687 Nm C
T&S 86 kN 155 kN T N&T 12,531 Nm 25,450 Nm C
T&S 84.7 kN 143 kN T N&T 8821 Nm 15,398 Nm T
N&T 5852 Nm 12,311 Nm T N&T 7729 Nm 14,635 Nm T
N&T 6239 Nm 12,243 Nm T N&T 8552 Nm 20,020 Nm C
N&T 5152 Nm 11,955 Nm T N&T 8402 Nm 20,625 Nm C
N&T 5818 Nm 14,607 Nm T N&T 6054 Nm 19,114 Nm T
N&T 5805 Nm 15,381 Nm T N&T 13,265 Nm 18,710 Nm T
N&T 6720 Nm 17,197 Nm C N&T 12,902 Nm 21,069 Nm T
N&T 6080 Nm 18,681 Nm C N&T 12,006 Nm 20,541 Nm T
N&T 5555 Nm 13,442 Nm T M&E 17.2 kNm 34.8 kNm T
N&T 5086 Nm 12,424 Nm T M&E 17.2 kNm 33.2 kNm T
N&T 6112 Nm 12,516 Nm T M&E 18.1 kNm 40 kNm T
N&T 6112 Nm 11,948 Nm T Niitani 55 kN 131 kN C
N&T 5490 Nm 14,561 Nm C Niitani 54 kN 129 kN C
N&T 5792 Nm 14,386 Nm T Niitani 54 kN 103 kN C
N&T 5871 Nm 14,557 Nm T T&M 7 kN 19 kN C
N&T 5331 Nm 15,257 Nm T T&M 9 kN 17.9 kN C
N&T 6759 Nm 16,295 Nm T T&M 9 kN 20.9 kN T
N&T 6613 Nm 16,911 Nm T T&M 8 kN 15 kN C
N&T 8821 Nm 23,388 Nm T T&M 8 kN 22.5 kN C
N&T 9267 Nm 24,306 Nm T T&M 8 kN 28 kN C
N&T 12,087 Nm 27,317 Nm T T&M 9 kN 18 kN C
*
Flexural moment or load.
†
T is tension-controlled; C is compression-controlled.
Notes: G&S is Grace and Singh30; Stoll is Stoll et al.31; T&S is Toutanji and Saafi32; N&T is Nanni and Tanigaki5; M&E is McKay and Erki33; Niitani is Niitani et al.34; T&M is
Taerwe and Mattys35; 1 kN = 0.225 kip; 1 Nm = 0.74 ft-lb; 1 kNm = 740 ft-lb.

sections satisfying the deformability limit, several cases with
tension-controlled sections demonstrated the indexes lower
than corresponding limits (Table 5), including maximum
margins of 4% and 5% for the AFRP- and CFRP-prestressed
bridges, respectively. This issue is attributable to the fact
that the bridges were initially designed without considering
deformability16; nonetheless, these insignificant margins
can still support that the proposed limits are applicable in
practice (the margins for the tension-controlled sections can
readily be addressed by slightly adjusting the amounts of the
AFRP and CFRP tendons).
MINIMUM REINFORCEMENT
Table 7 summarizes experimental data collated from
published literature on FRP-prestressed concrete beams.
To account for the significance of failure modes, either
compression- or tension-controlled, the experimental data
were sorted and presented in terms of ultimate-to-cracking
ratios (Fig. 8(a)). The average ratios of the beams failed by
concrete crushing and FRP rupture were 2.3 and 2.4, respec-
tively. Figure 8(b) shows the factored ratios of the experi-
mental responses. The strength reduction factors were taken
from Kim and Nickle,16 which were calibrated using reli-
ability theory, rather than the empirically determined reduc-
tion factors of ACI 440.4R-04.3 In all cases except one beam
failed by bond slip, the factored ratios were greater than
1.2, including average ratios of 1.7 and 1.9 for the tension-
controlled and compression-controlled beams, respectively.
While these ratios were obtained from laboratory-scale
beams, their variation range was in reasonable agree-
ment with the range of the full-scale girders (Fig. 5(c) and
Fig. 8(c) and 8(d)). The ratio between a factored flexural
resistance (ϕMn) and its cracking counterpart (Mcr) is there-
fore proposed to be greater than 1.2, which is identical to
the ratio specified in ACI 318-14,18 the PCI Design Hand-
book,25 and AASHTO LRFD Bridge Design Specifica-

232 ACI Structural Journal/January 2018

Fig. 8—Ratio between ultimate and cracking responses: (a) unfactored (experimental); (b) factored (experimental): (c) unfac-
tored (benchmark full-scale bridges); and (d) factored (benchmark full-scale bridges).
tions.15 It should be noted that ACI 440.4R-043 states a
ϕMn/Mcr ratio of 1.5 that was arbitrarily elevated from the
ratio of 1.2 used in steel-prestressed concrete sections to
avoid a concern associated with a tension-controlled FRP
section, which does not appear to be an issue considering the
high strength of FRP composites (in other words, adequately
designed FRP-prestressed concrete girders will not fail
abruptly immediately after cracking, regardless of failure
mode that is related to the ultimate limit state rather than at
a cracking stage). Further, because the amount of minimum
reinforcement is not necessarily engaged with FRP types,
the proposed ϕMn/Mcr ratio of 1.2 for AFRP- and CFRP-
prestressed members can be integrated in design.
SUMMARY AND CONCLUSIONS
This paper has addressed technical challenges regarding
FRP-prestressed concrete bridge girders, particularly for
those of ACI 440.4R-043: long-term multipliers, deform-
ability index, and minimum reinforcement. New design
information was proposed based on analytical derivations
along with an experimental assessment. Fifty previously-
designed highway bridges prestressed with AFRP and CFRP
tendons were also employed to evaluate the proposed design
expressions. The long-term multipliers were refined into
three categories for AFRP and CFRP applications at Erec-
tion and Final without composite topping, with unreinforced
composite topping, and with reinforced structural slab. These
are different from the single category (without composite
topping) specified in ACI 440.4R-043 that was taken from
the empirically-determined values with steel strands. The
new deformability index with specific limits was appraised
in two possible failure conditions: tension-controlled and
ACI Structural Journal/January 2018
compression-controlled sections. The minimum reinforce-
ment requirement on the cracking stage of prestressed
concrete girders was revisited for both improvement and
consistency with other design specifications. The following
conclusions are drawn:
1. The proposed long-term multipliers for AFRP/
CFRP-prestressed concrete girders were the same as those of
ACI 440.4R-043 for deflections at erection and final stages,
whereas the multipliers were different in all other cases
because of the AFRP/CFRP’s unique material characteristics
associated with the stochastic calibration.
2. Although the moment-curvature behavior of the AFRP/
CFRP-prestressed girders was not comparable with that
of steel-prestressed counterparts, the girders sufficiently
deformed prior to failure in either concrete-crushing or
FRP-rupture. With an increase in the prestressing level, the
deformability indexes of the girders with AFRP and CFRP
were asymptotically reduced toward the suggested limits.
3. The alteration of the AFRP/CFRP moduli influenced the
girders’ cracking moments; however, corresponding effects
on precracking behavior and post-cracking stiffness were
negligible. The degraded moduli, however, changed the
failure mode of the girders from compression-controlled to
tension-controlled, which is an important consideration in
aggressive service environments.
4. The average ultimate-to-cracking ratios of the 52 exper-
imental AFRP/CFRP-prestressed concrete beams were 2.3
and 2.4 for the tension-controlled and compression-con-
trolled sections, respectively, which were in agreement with
the ratios of the full-scale bridge girders. The factored ulti-
mate-to-cracking moment ratio of 1.2 proposed was suffi-
cient to provide the low bound for AFRP/CFRP-prestressed
233
concrete, and is in conformance with the design expres-
sion of ACI 318-14,18 the PCI Design Handbook,25 and the
AASHTO LRFD Bridge Design Specifications.15
AUTHOR BIOS
Yail J. Kim, FACI, is a Professor in the Department of Civil Engi-
neering at the University of Colorado Denver, Denver, CO. He is Chair
of ACI Committee 345, Concrete Bridge Construction, Maintenance, and
Repair; and ACI Subcommittee 440-I, FRP-Prestressed Concrete. He is
a member of ACI Committees 342, Evaluation of Concrete Bridges and
Bridge Elements; 343, Concrete Bridge Design; and 440, Fiber-Reinforced
Polymer Reinforcement. His research interests include advanced composite
materials for structural rehabilitation, complex systems, uncertainty quan-
tification, and science-based structural engineering, including statistical,
interfacial, and quantum physics.
Raymon W. Nickle is a Structural Engineer at AECOM. He received his BS
and MS in civil engineering from Colorado State University, Fort Collins,
CO, in 2009 and the University of Colorado Denver in 2015, respectively.
His research interests include the design and analysis of transportation
structures.
ACKNOWLEDGMENTS
The authors gratefully acknowledge the members of ACI 440-I Subcom-
mittee, FRP-Prestressed Concrete, chaired by the first author of this paper,
for their rigorous review of technical contents and valuable suggestions.
Proprietary information such as product names and manufacturers is not
included herein.
REFERENCES
1. Abdelrahman, A. A.; Tadros, G.; and Rizkalla, S. H., “Test Model for
the First Canadian Smart Highway Bridge,” ACI Structural Journal, V. 92,
No. 4, July-Aug. 1995, pp. 451-458.
2. Fam, A. Z.; Rizkalla, S. H.; and Tadros, G., “Behavior of CFRP for
Prestressing and Shear Reinforcements of Concrete Highway Bridges,”
ACI Structural Journal, V. 94, No. 1, Jan.-Feb. 1997, pp. 77-86.
3. ACI Committee 440, “Prestressing Concrete Structures with FRP
Tendons (ACI 440.4R-04),” American Concrete Institute, Farmington
Hills, MI, 2004, 35 pp.
4. Youakim, S. A., and Karbhari, V. M., “An Approach to Determine
Long-Term Behavior of Concrete Members Prestressed with FRP Tendons,”
Construction and Building Materials, V. 21, No. 5, 2007, pp. 1052-1060.
doi: 10.1016/j.conbuildmat.2006.02.006
5. Nanni, A., and Tanigaki, M., “Pretensionded Prestressed Concrete
Members with Bonded Fiber Reinforced Plastic Tendons: Development
and Flexural Bond Lengths (Static),” ACI Structural Journal, V. 89, No. 4,
July-Aug. 1992, pp. 433-441.
6. Grace, N. F., and Abdel-Sayed, G., “Behavior of Carbon Fiber-
Reinforced Prestressed Concrete Skew Bridges,” ACI Structural Journal,
V. 97, No. 1, Jan.-Feb. 2000, pp. 26-35.
7. Dolan, C. W., and Swanson, D., “Development of Flexural Capacity
of an FRP Prestressed Beam with Vertically Distributed Tendons,”
Composites. Part B, Engineering, V. 33, No. 1, 2002, pp. 1-6. doi: 10.1016/
S1359-8368(01)00053-1
8. Saiedi, R.; Fam, A.; and Green, M. F., “Behavior of CFRP-Prestressed
Concrete Beams under High-Cycle Fatigue at Low Temperature,” Journal
of Composites for Construction, ASCE, V. 15, No. 4, 2011, pp. 482-489.
doi: 10.1061/(ASCE)CC.1943-5614.0000190
9. Zhang, K.; Fang, Z.; and Nanni, A., “Behavior of Tendons with
Multiple CFRP Rods,” Journal of Structural Engineering, ASCE, V. 142,
No. 10, 2016, p. 04016065. doi: 10.1061/(ASCE)ST.1943-541X.0001535
10. Soudki, K. A., “FRP Reinforcement for Prestressed Concrete Struc-
tures,” Progress in Structural Engineering and Materials, V. 1, No. 2, 1998,
pp. 135-142. doi: 10.1002/pse.2260010205
11. Lees, J. M., “Fibre-Reinforced Polymers in Reinforced and Prestressed
Concrete Applications: Moving Forward,” Progress in Structural Engi-
neering and Materials, V. 3, No. 2, 2001, pp. 122-131. doi: 10.1002/pse.60
12. Martin, L. D., “A Rational Method for Estimating Camber and
Deflection of Precast Prestressed Members,” PCI Journal, V. 22, No. 1,
1977, pp. 100-108.
13. Zou, P. X. W., “Flexural Behavior and Deformability of Fiber Rein-
forced Polymer Prestressed Concrete Beams,” Journal of Composites
for Construction, ASCE, V. 7, No. 4, 2003, pp. 275-284. doi: 10.1061/
(ASCE)1090-0268(2003)7:4(275)
234
14. Au, F. T. K., and Du, J. S., “Deformability of Concrete Beams with
Unbonded FRP Tendons,” Engineering Structures, V. 30, No. 12, 2008,
pp. 3764-3770. doi: 10.1016/j.engstruct.2008.07.003
15. AASHTO, “AASHTO LRFD Bridge Design Specifications,” seventh
edition with interim revisions, American Association of State Highway and
Transportation Officials, Washington, DC, 2016.
16. Kim, Y. J., and Nickle, R. W., “Strength Reduction Factors for Fiber-
Reinforced Polymer Prestressed Concrete Bridges in Flexure,” ACI
Structural Journal, V. 113, No. 5, Sept.-Oct. 2016, pp. 1043-1052. doi:
10.14359/51689028
17. ACI Committee 318, “Building Code Requirements for Reinforced
Concrete (ACI 318-71),” American Concrete Institute, Farmington Hills,
MI, 1971, 78 pp.
18. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 2014, 520 pp.
19. ACI Committee 209, “Prediction of Creep, Shrinkage, and Tempera-
ture Effects in Concrete Structures,” Designing for Effects of Creep,
Shrinkage and Temperature in Concrete Structures, SP-27, American
Concrete Institute, Farmington Hills, MI, 1971, pp. 51-93.
20. Skogman, B. C.; Tadros, M. K.; and Grasmick, R., “Ductility of
Reinforced and Prestressed Concrete Flexural Members,” PCI Journal,
V. 33, No. 6, 1988, pp. 94-107. doi: 10.15554/pcij.11011988.94.107
21. PCI, PCI Bridge Design Manual, Precast/Prestressed Concrete Insti-
tute, Chicago, IL, 2003.
22. Karbhari, V. M., “Durability Data for FRP Rehabilitation Systems,”
Report No. CA09-0248, California Department of Transportation, Sacra-
mento, CA, 2009.
23. Lees, J. M.; Toumpanaki, E.; Barbezat, M.; and Terrasi, G. P.,
“Mechanical and Durability Screening Test Methods for Cylindrical CFRP
Prestressing Tendons,” Journal of Composites for Construction, ASCE, V.
21, No. 2, 2017, p. 04016080 doi: 10.1061/(ASCE)CC.1943-5614.0000727
24. Guedes, R. M., “Time-Dependent Failure Criteria for Polymer Matrix
Composites: A Review,” Journal of Reinforced Plastics and Composites,
V. 29, No. 20, 2010, pp. 3041-3047. doi: 10.1177/0731684410370067
25. PCI, PCI Design Handbook, seventh edition, Precast/Prestressed
Concrete Institute, Chicago, IL, 2010.
26. Kwon, O.-S.; Kim, E.; Orton, S.; Salim, H.; and Hazlett, T., “Cali-
bration of the Live Load Factor in LRFD Design Guidelines,” Final Report
No. OR11-003, Missouri Department of Transportation, Jefferson City, MO,
2011.
27. Atadero, R. A., and Karbhari, V. M., “Calibration of Resistance
Factors for Reliability-Based Design of Externally-Bonded FRP Compos-
ites,” Composites. Part B, Engineering, V. 39, No. 4, 2008, pp. 665-679.
doi: 10.1016/j.compositesb.2007.06.004
28. Sanchez-Heres, L. F.; Ringsberg, J. W.; and Johnson, E., “Influence
of Mechanical and Probabilistic Models on the Reliability Estimates of
Fibre-Reinforced Cross-Ply Laminates,” Structural Safety, V. 51, 2014,
pp. 35-46. doi: 10.1016/j.strusafe.2014.06.001
29. Okeil, A.; El-Tawil, S.; and Shahawy, M., “Flexural Reliability of
Reinforced Concrete Bridge Girders Strengthened with Carbon Fiber-
Reinforced Polymer Laminates,” Journal of Bridge Engineering, ASCE,
V. 7, No. 5, 2002, pp. 290-299. doi: 10.1061/(ASCE)1084-0702(2002)7:5(290)
30. Grace, N. F., and Singh, S. B., “Design Approach for Carbon
Fiber-Reinforced Polymer Prestressed Concrete Bridge Beams,” ACI Struc-
tural Journal, V. 100, No. 3, May-June 2003, pp. 365-376.
31. Stoll, F.; Saliba, J. E.; and Casper, L. E., “Experimental Study of
CFRP-Prestressed High-Strength Concrete Bridge Beams,” Composite
Structures, V. 49, 2000, pp. 191-200. doi: 10.1016/S0263-8223(99)00134-8
32. Toutanji, H., and Saafi, M., “Performance of Concrete Beams
Prestressed with Aramid Fiber-Reinforced Polymer Tendons,”
Composite Structures, V. 44, No. 1, 1999, pp. 63-70. doi: 10.1016/
S0263-8223(98)00126-3
33. McKay, K. S., and Erki, M. A., “Flexural Behaviour of Concrete
Beams Pretensioned with Aramid Fibre Reinforced Plastic Tendons,”
Canadian Journal of Civil Engineering, V. 20, No. 4, 1993, pp. 688-695.
doi: 10.1139/l93-085
34. Niitani, K.; Tezuka, M.; and Tamura, T., “Flexural Behavior of
Prestressed Concrete Beams Using AFRP Pre-Tensioning Tendons,”
Proceedings of the 3rd International Symposium on Non-Metallic (FRP)
Reinforcement for Concrete Structures, 1997, pp. 663-670.
35. Taerwe, L., and Matthys, S., “Structural Behavior of Concrete Slabs
Prestressed with AFRP Bars,” Proceedings of the 2nd International Sympo-
sium on Non-Metallic (FRP) Reinforcement for Concrete Structures, 1995,
pp. 421-429.
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL TECHNICAL PAPER
Title No. 115-S19

Seismic Performance of Reinforced Concrete Columns

with Lap Splices in Plastic Hinge Region
by Chul-Goo Kim, Hong-Gun Park, and Tae-Sung Eom
For convenient bar placement, longitudinal bars in columns are
often lap-spliced at the bottom of the columns where plastic hinge
forms. In this study, the seismic performance of lap-spliced columns
was investigated through cyclic loading tests. The bar splice length
and shear span-to-height ratio were considered as the primary
test variables. The test results showed that after flexural yielding,
bond-splitting cracks propagated along the lap-spliced bars during
repeated load cycles. The strength, ductility, and failure mode of
the columns were significantly affected by the bar splice length
and shear span-to-height ratio (or moment gradient). Based on the
results, effects of moment gradient on the bond demand of spliced
bars were studied, and a reduced splice length addressing the
moment gradient was proposed. Furthermore, recommendations
for the design and detailing of lap splices in the column plastic
hinge region such as splice length, bar offset, and confining trans-
verse reinforcement were provided.

Keywords: cyclic test; lap splice; offset bar; reinforced concrete column;
seismic performance.
INTRODUCTION
The seismic performance of reinforced concrete (RC)
columns such as ductility and energy dissipation capacity
can be degraded when lap splices of longitudinal reinforcing
bars are located in the plastic hinge zone. For this reason,
in ACI 318-14,1 lap splices are not permitted in potential
plastic hinge zones of special moment frames. On the other
hand, such a requirement is not imposed for ordinary and
intermediate moment frames. Thus, in low and moderate
seismic regions, for convenience of bar placement, lap
splices of longitudinal bars are usually located at the bottom
of columns, which is a plastic hinge zone (Fig. 1).
When lap splices of longitudinal bars are used at the plastic
hinge zone, the strength and ductility of columns are affected
by the lap-splice length. Figure 2 shows the required tension
splice length ls specified in ACI 318-141 and Eurocode 2.2 The
horizontal and vertical axes denote the concrete compressive
strength (fc′ = 18 ~ 30 MPa [2.61 ~ 4.35 ksi]) and required
splice length-to-bar diameter ratio (ls/db), respectively. In this
comparison, the lap-splice length specified in ACI 318-14 is
defined as ls = 1.3ld = 1.3(0.9dbfy/√fc′)/([cb + Ktr]/db) for Splice
Class B. In Eurocode 2, the lap-splice length is defined as
ls = α6lbd = α2α3α6(dbfyd)/(4fbd) by assuming α1 = α4 = α5 = 1.0,
where fyd is the design yield stress of the spliced bar, fbd is the
concrete bond strength (= 2.25fctd = 0.47fck2/3), and α2 (= 0.7 ~
1.0) and α3 (= 0.7 ~ 1.0) are the coefficients for concrete cover
and confining transverse reinforcement, respectively. If the
lap splices of column longitudinal bars are placed at the same
location (α6 = 1.5), the splice lengths are ls/db = 38.2 ~ 63.7
(ACI 318-14) and 28.0 ~ 57.2 (Eurocode 2) for fy = 400 MPa
Fig. 1—Location of lap splices in reinforced concrete
columns.
[58 ksi] and ls/db = 47.8 ~ 79.6 (ACI 318-14) and 35.1 ~ 71.5
(Eurocode 2) for fy = 500 MPa (72.5 ksi).
In previous studies on lap splices of column longitu-
dinal bars, Lynn et al.3 investigated the cyclic behavior of
columns with lap-spliced bars (ls = 20db). The test results
showed that the column behavior was significantly affected
by the spacing s of transverse hoops in the lap-splice region:
the column behavior was ductile for s = 0.67hmin (305 mm
[12 in.]), but brittle for s = hmin (457 mm [18 in.]). Aboutaha
et al.4 reported that in columns with lap-spliced bars of ls =
24db, bond-splitting failure occurred along the length of
the lap splice before flexural yielding. Melek and Wallace5
performed cyclic loading tests of columns with ls = 20db and
s = 0.67hmin (305 mm [12 in.]). Brittle failure occurred in all
columns at relatively small lateral drift ratios of 1.0% ~ 1.5%
because of the short lap-splice length. Similarly, Haroun and
Elsanadedy,6 Harries et al.,7 and Harajli8 reported that brittle
bond failure occurred at lateral drift ratios of 1.0% ~ 1.5% in
columns with ls = 20db, 22db, and 30db.
Furthermore, the strength and ductility of columns are
affected by lap-splice details. ACI 3159 specifies two
offset-bar details, bottom and top offset-bar splices (refer
to Fig. 3). In the bottom offset-bar splice, the bottom bars
from the lower story are offset inside. In the top offset-bar
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2017-071, doi: 10.14359/51701109, was received February 27, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.

ACI Structural Journal/January 2018 235

Fig. 2—Comparison of lap-splice lengths by ACI 318-14 and Eurocode 2.
Fig. 3—Types of lap splice using offset bars (ACI 315).
splice, on the other hand, the top bars are offset inside. As
the bottom bars are located inside the core concrete, the
moment-carrying capacity of the column with the bottom
offset-bar splice is less. However, columns with the bottom
offset-bar splice show better displacement ductility owing
to the thicker concrete cover of the bottom offset-bar. For
this reason, in current seismic design codes, the bottom
offset-bar splice is recommended (ACI 318-14, Section
10.7.5). On the contrary, if the top offset-bar splice is used,
there is no reduction in the moment capacity due to the bar
offset but the deformation capacity can be decreased due to
premature bond-splitting cracks occurring along the splice
length (Boyes et al.10). Even in the case that the top and
bottom bars are lap-spliced next to each other without offset,
the column performance such as the moment capacity and
deformation capacity is similar to that of columns with the
top offset-bar splice.8
The performance of lap splices is also affected by the
distributions of flexural moment and shear force along the
length of columns.11,12 The lap-splice length of ACI 318-14
was developed considering uniform bond stresses based on
the test results for beams with uniform moment (Fig. 4(a)).
On the other hand, in columns subjected to lateral loading
(Fig. 4(b)), the demand of bond stress varies along the
splice length according to the moment gradient. However,
236
bond-splitting cracks in the lap-splice region become severe
owing to the presence of shear force, particularly after diag-
onal shear cracking. Repeated cyclic loading may also cause
bond deterioration in the plastic hinge zone. Ferguson and
Krishnaswamy12 evaluated the performance of lap splices
under moment gradient for bearing walls. They reported that
the performance of lap splices was affected by the average
stress of two spliced bars, which was less than the yield stress
because of the moment gradient of the lap-splice region.
They proposed a reduction factor of (1 + k)/2, ranging from
0.75 to 1.0, where k is the ratio of the less of two spliced
bar stresses to the yield stress. As no data for k < 0.5 were
available, the reduction factor for splice length was limited
to not less than 0.75.
In the previous studies, existing columns designed and
constructed before the 1970s were of concern because such
columns had relatively small lap-splice lengths of 20db and
30db, which are the minimum compression lap-splice lengths
specified in ACI 318-5613 and ACI 318-77,14 respectively. The
previous tests demonstrate that, even with such small lap-splice
lengths, the nominal flexural strength of the columns were
attained, while the ductility and energy dissipation capacity
were degraded owing to premature bond-splitting failure.
In the present study, cyclic loading tests of columns with
lap-spliced longitudinal bars in the plastic hinge region were
performed. The lap-splice length of the column bars, and the
shear-span length (that is, moment gradient and shear force)
were taken into account as the primary test variables. With
the results of the cyclic loading tests, the strength, ductility,
and failure mode of the columns were investigated and the
effects of the test variables on the column performance were
discussed. Based on the results, the lap-splice length of
column bars was evaluated.
RESEARCH SIGNIFICANCE
Columns with lap splices in the plastic hinge region
are permitted to use in ordinary and intermediate moment
frames. However, in most cases, effects of lap splices on the
column behavior are not considered in seismic design and
evaluation. In this study, the effects of moment gradient and
shear force on the bond demand and splice performance were
investigated. Based on the test results, recommendations
were made for the design and evaluation of these lap splices.
The test results and design considerations could be useful in
ACI Structural Journal/January 2018
Fig. 4—Bending moments and shear forces at lap-splice region.
the evaluation of existing buildings because old buildings or Table 1—Mechanical properties of steel
buildings in low and moderate seismic zones usually have reinforcement
short lap splices in the plastic hinge zones of columns. Test No.* Bar type db, mm fy, MPa fu, MPa
SD400 D13 12.7 500 645
TEST PLAN 1
Material strength SD500 D25 25.4 571 700
Eight column specimens were prepared for the cyclic load SD400 D13 12.7 481 607
tests. Table 1 presents the yield and tensile strengths (fy and fu, 2
SD500 D25 25.4 550 685
respectively) of the reinforcing steel bars used for the spec-
SD400 D13 12.7 528 656
imens. The yield strength and tensile strength are fy = 475 ~ 3
590 MPa (68.9 ~ 85.6 ksi) and fu = 607 ~ 707 MPa (88.0 ~ SD500 D25 25.4 588 702
102.5 ksi), respectively. In the table, SD400 and SD500 SD400 D13 12.7 475 659
denote the nominal yield strength grades of deformed bars, 4 SD500 D25 25.4 590 707
400 and 500 MPa (58.0 and 72.5 ksi), respectively, and D10,
SD500 D29 28.6 521 671
D13, D22, D25, and D29 denote the diameters of deformed
bars, db = 9.53, 12.7, 22.2, 25.4, and 28.6 mm ([0.375, 0.5, Eight columns were tested over four experiments. Thus, yield and tensile strength
*

differed even for reinforcing bars of same grade and size.

0.875, 1.0, and 1.125 in.), respectively.
Table 2 presents the mixture proportion of the concrete.
The maximum aggregate size was 25.4 mm (1.0 in.).
Compression tests of cylinders (100 x 200 mm [3.94 x
7.87 in.]) were performed on the first day of testing. The
compressive strength of the concrete was fc′ = 25 ~ 37 MPa
(3.63 ~ 5.37 ksi).
Test specimens
The dimensions and reinforcement details of the eight
square column specimens are shown in Fig. 5. The test vari-
ables such as the shear span-to-height ratio (a/h), lap-splice
length (ls), and transverse hoop spacing (s) are presented in
Table 3. In the specimen names, the first letter “S” denotes
square cross section; the following letters “L00”, “L30”,
“L40”, and “L50” denote the length of lap splice, ls = 00db
(that is, without lap splice), 30db, 40db, and 50db, respec-
tively; “S1” and “S2” denote the shear spans, a = 1200 and
2400 mm (47.2 and 94.5 in.), respectively; and the last letter
“B” denotes the offset details of column bars (that is, bottom
offset bar splice, refer to Fig. 3(a)).
The cross-sectional dimensions of the columns were 400 x
400 mm (15.8 × 15.8 in.) (h = 400 mm [15.8 in.]). The shear
span lengths were a = 1200 mm (47.2 in.) (a/h = 3.0) or
a = 2400 mm (94.5 in.) (a/h = 6.0). SD500 D25 and D29
bars were used for the longitudinal reinforcement. In all
specimens, the straight top bars were lap-spliced at the
Notes: 1 mm = 0.039 in.; 1 MPa = 0.145 ksi.
plastic hinge region with the bottom offset bars embedded
in the base. Except for SL40S2B-1, the same diameter bars
were used for both the top and bottom bars. In SL40S2B-1,
the diameter of the bottom offset bars was increased to D29
to make up for the moment strength reduction owing to the
bottom bar offset. The lap-splice lengths of the longitu-
dinal bars varied from 30db to 50db, which were equivalent
to 57% ~ 108% of the required splice lengths of ACI 318-14
calculated using the actual material strength (fc′ and fy) and
bottom offset-bar diameters (db).
The lap-spliced longitudinal bars were confined by trans-
verse hoops with 90-degree end hooks for anchorage. SD400
D13 hoops were used at a spacing of s = 165 mm (6.5 in.)
(= 0.5d; d is the effective depth of the column), except for
SL50S2B-1 with s = 330 mm (13.0 in.) (=1.0d). The first hoop
was placed at a height of 80 mm (3.15 in.) from the base. As
shown in Fig. 5, the location of the 90-degree hook anchorage
of the hoops was alternated in accordance with ACI 318-14.
Loading method
Figure 6 shows the test setup for lateral and axial loading.
The cyclic lateral loading was controlled by the lateral
displacement of the actuator with a maximum stroke of
±250 mm (9.84 in.). The actuator was placed at 1200 or
2400 mm (47.2 or 94.5 in.) from the column base. During

ACI Structural Journal/January 2018 237

Table 2—Mixture design of concrete
Unit weight, kgf/m3
Test No. Compressive strength fc′ w/c W C FS S FA CA SP Slump
1 32 MPa 55.5% 155 153 56 70 923 950 1.95 120 mm
2 25 MPa 46.2% 173 243 56 75 807 887 2.24 150 mm
3 27 MPa 53.5% 159 149 74 74 907 893 1.78 120 mm
4 37 MPa 44.8% 168 319 56 - 832 915 2.25 150 mm
Notes: W is water; C is cement; FS is fly ash; S is blast-furnace slag; FA is fine aggregate; CA is coarse aggregate; SP is superplasticizer; 1 mm = 0.039 in.

Table 3—Summary of test variables

Specimen Test No. Shear span length a Splice length ls Hoop spacing s Number of hoops along lap splice Ratio of bottom splice offset bars
SL00S1B 1 N/A 0.5d — —
1200 mm
SL30S1B 1 30db 0.5d 4
(a/h = 3.0)
SL40S1B 1 40db 0.5d 6
SL30S2B 2 30db 0.5d 4 2.53% (8-D25)
SL40S2B 2 40db 0.5d 6
2400 mm
SL50S2B 4 50db 0.5d 8
(a/h = 6.0)
SL40S2B-1 4 40db 0.5d 7 3.21% (8-D29)
SL50S2B-1 3 50db 1.0d 4 2.53% (8-D25)
Note: 1 mm = 0.039 in.

Fig. 5—Dimensions and reinforcement details of specimens. (Note: 1 mm = 0.039 in.)

cyclic loading tests, the lateral drift ratio of the column was
increased to 0.25%, 0.35%, 0.5%, 0.75%, 1.0%, 1.5%, 2.0%,
2.5%, 3.5%, 5.0%, and 7.0%. The loading sequence was
planned in accordance with ACI 374.1.15 Thus, the target
lateral drift ratio of a loading step was determined as 1.25 ~
1.5 times the previous loading step, and load cycles were
repeated three times at each loading step. The loading rate
was increased from 20 mm/min (0.787 in/min) for δ = 0.25%
to 60 mm/min (2.36 in./min) for δ = 1.0%, and then main-
tained a constant value of 60 mm/min (2.36 in./min) after
δ = 1.0%.
An axial load N (= 0.15 to 0.185Agfc′) was applied to the
top of the specimen by post-tensioning two high-strength
bars (diameter = 47 mm [1.85 in.]). For post-tensioning,
two hydraulic jacks with a capacity of 1000 kN (225 kip)
were placed on the top of the column. During cyclic lateral
loading, the level of axial load was kept uniform by main-
taining the hydraulic pressure of the jacks. As shown in
Fig. 6, because pin joints are provided at the bottom of the
bars, the bars are free to move laterally. Because the axial
load of the bars always acts along the centerline of the
column, the second-order effect is negligible in the lateral
load-displacement relationship.

238 ACI Structural Journal/January 2018

Fig. 6—Test setup for axial and lateral loading. (Note: 1 mm = 0.039 in.)
TEST RESULTS
Cyclic behavior and failure mode
Figure 7 shows the base moment and lateral drift ratio
(M-δ) relationships of the columns. For a fair strength
comparison between the columns with different shear spans,
the base moment M (= Pa, where P is the lateral load of
the actuator and a is the shear span of the column) is used,
instead of the lateral load P. The lateral drift ratio δ is defined
by dividing the lateral displacement at the loading point by
the shear span. The maximum test moments Mu are denoted
as white circles, while the calculated moment strength Mn
without strength reduction factor is denoted as horizontal
dashed lines. The Mn of each specimen was calculated
from section analysis, considering the applied axial load N,
measured material strength, and location of the bottom offset
bars in the cross section.
In SL00S1B, without lap splices, the maximum loads
(Mu  = +407 and –400 kN∙m [+300 and –295 kip∙ft])
were greater than the calculated moment strength (Mn =
±374 kN∙m [±276 kip∙ft]). Strength degradation occurred at
δ = 2.0% after spalling of the concrete cover at the bottom
of the column. Ultimately, at δ = 3.5%, the load-carrying
capacity was significantly decreased because of hoop failure
at the 90-degree hook anchorages (Fig. 8(a)). Despite the
presence of the axial load of N = 870 kN (196 kip), severe
web shear cracking occurred. Concrete strut failure occurred
in the plastic hinge region.
In SL30S1B, with lap splice ls = 30db (63% of the minimum
splice length required by ACI 318-14), the maximum loads
(Mu= +350 and –354 kN∙m [+258 and –261 kip∙ft]) were
13% less than those of SL00S1B without lap splice. Never-
theless, Mu was slightly greater than the moment strength
(Mn = ±340 kN∙m [251 kip∙ft]). Because the bottom bars,
which resist flexure moment, were offset inside (Fig. 3(a)),
the test strength and nominal flexural strength of SL30S1B
were less than those of SL00S1B. Ultimately, failure
occurred at δ = 3.5% because of concrete crushing at the
ACI Structural Journal/January 2018
bottom of the column (Fig. 8(b)). In SL40S1B, with the
increased splice length ls = 40db, the maximum loads
(Mu = +393 and –417  kN∙m [+290 and –308 kip∙ft]) were
significantly increased when compared to SL30S1B. Despite
the lower nominal strength, the test strength was equivalent
to that of SL00S1B without lap splice. The enhanced strength
of SL40S1B is attributed to the increased splice length.
Spalling of the cover concrete began to occur at δ = 2.0%.
The load-carrying capacity was significantly degraded at the
first load cycle of δ = –5.0% owing to excessive concrete
crushing and web shear cracking (Fig. 8(c)).
Figures 7(d) and (e) show the cyclic behaviors of
SL30S2B and SL40S2B with the increased shear span a =
2400 mm (94.5 in.). The maximum test loads were approx-
imately 95% of the nominal strength (Mn = 294 kN∙m
[217 kip∙ft]), which indicates that the performance of the lap
splices was not successful. Despite the insufficient strength,
their ductile behaviors were maintained until δ = 7.0%. In
SL30S2B and SL40S2B, the column shear was significantly
reduced because of the increased shear-span length, when
compared to those of SL30S1B and SL40S1B with a shorter
shear span. Thus, as shown in Fig. 8(d) and (e), diagonal
shear cracking in the web was not significant even at large
inelastic deformations of δ = 5.0% or greater, and conse-
quently the displacement capacity was increased. However,
bond-splitting cracks occurred owing to the insufficient
splice lengths (that is, 57% and 76% of the required splice
length) and propagated along the spliced longitudinal bars
as the inelastic deformation increased. When the test results
of SL30S2B and SL40S2B are compared with those of
SL30S1B and SL40S1B, obviously, the shear span length a
affected the failure mode and ductility of the columns with lap
splice: when the shear-span ratio was decreased (a/h = 3.0),
even with the short splice length, the test strength exceeded
the nominal flexural strength. On the other hand, when the
shear span ratio was increased (a/h = 6.0), the test strength
was less than the nominal flexural strength. However, despite
239
Fig. 7—Base moment (M)-lateral drift ratio (δ) relationships of columns. (Note: 1 in. = 25.4 mm; 1 kip = 4.448 kN.)

Fig. 8—Cracking modes at δ = 2.0% and failure modes at end of tests.

240 ACI Structural Journal/January 2018

Table 4—Summary of test results
Lap-splice length (mm or ratio)
Specimen fc′, MPa ls ls/db ls,ACI* ls/ls,ACI lsm† ls/lsm Mu‡, kN∙m Mn, kN∙m Mu/Mn Δy§, mm Δu||, mm Failure mode#
SL00S1B 32 — — — — — — 404 374 1.08 9.3 40 FS
SL30S1B 32 750 30 1200 0.63 936 0.80 352 340 1.04 7.1 44 FS
SL40S1B 32 1000 40 1200 0.83 936 1.07 405 340 1.19 10.1 69 FS
SL30S2B 25 750 30 1308 0.57 1240 0.60 278 294 0.94 18.5 132 BS
SL40S2B 25 1000 40 1308 0.76 1240 0.81 281 294 0.95 17.9 168 —
SL50S2B 37 1250 50 1153 1.08 1266 0.99 371 367 1.01 23.8 168 —
SL40S2B-1 37 1160 40 1146 1.01 1128 1.11 363 393 0.92 22.8 168 —
SL50S2B-1 27 1250 50 1345 0.93 1196 0.97 342 312 1.10 22.2 120 BS
*
Length of Class B lap splice specified in ACI 318-14 on basis of bottom offset bars.
†
Modified lap-splice length in Eq. (3).
‡
Average of positive and negative maximum loads.
§
Δy = Pu/Ky, where Pu is maximum load and Ky is secant stiffness connecting origin and pre-peak point of 0.6Pu.
||
Δu is defined as average value of positive and negative maximum displacement corresponding to 80% of maximum load or as displacement at end of tests.
#
FS and BS, respectively, denote web shear failure after flexural yielding, bond-splitting failure along lap-splice length. In SL40S2B, SL50S2B, and SL40S2B-1, tests were termi-
nated without failure because of limited actuator stroke.
Notes: 1 in. = 25.4 mm; 1 kip = 4.448 kN; 1 ksi = 6.895 MPa.

the lower strength, the ductile behaviors of SL30S2B and
SL40S2B were maintained until large inelastic deformations.
In SL50S2B with ls = 50db, the maximum load (Mu =
+364 and –377 kN∙m [268 and 278 kip∙ft]) was signifi-
cantly increased, reaching to the nominal strength (Mn =
±367  kN∙m [±271 kip∙ft]). Because the splice length was
greater than the requirement of ACI 318-14, the strength
degradation during repeated load cycles was limited and the
energy dissipation capacity, represented as the enclosed area
by a full load cycle in the load-displacement curves shown in
Fig. 7, was significantly enhanced when compared to other
columns with ls = 30db and 40db. At δ = 7.0%, the test was
terminated because of lack of actuator stroke. As shown in
Fig. 8(f), bond-splitting cracking was not significant, which
indicates that the force transfer through the bond between
the concrete and spliced bars was successful.
In SL40S2B-1 with bottom offset bars of a greater bar diam-
eter D29 (ls = 40db and db = 28.6 mm [1.125 in.]), the maximum
loads (Mu = +342 and –383 kN∙m [252 and 282 kip∙ft]) were
less than the calculated moment strength (Mn = ±393 kN∙m
[±290 kip∙ft]) using the measured material strength (Fig. 7(g)).
This indicates that the lap-splice length ls = 40db was not suffi-
cient to develop the full yield strength of the D29 offset bars,
although ls/ls,ACI was greater than 1.0 (Table 4). However,
as the spliced bars were confined by transverse hoops with
a spacing s = 0.5d, the strength degradation was limited and
the hysteretic energy dissipation was significant. On the
other hand, in SL50S2B-1 with the increased hoop spacing
s = 1.0d, the maximum loads (Mu = +347 and –337  kN∙m
[256 and 249 kip∙ft]) were greater than the calculated moment
strength (Mn = ±312 kN∙m [±230 kip∙ft]), which indicates that
the lap-splice length ls = 50db satisfactorily developed the full
yield strength of the D25 offset bars. However, as the hoop
spacing s = 1.0d was not sufficient to ensure a post-yield
ductile behavior of the column, bond-splitting cracks occurred
along the lap splice at δ = 2.0% (Fig. 8(h)). Consequently, the
ductility and energy dissipation capacity of SL50S2B-1 were
significantly degraded.
Strains of lap-spliced bars
To investigate the force transfer between the spliced
longitudinal bars, the strains of the lap-spliced bars were
measured using strain gauges. Figure 9 shows the spliced
bar strains measured from SL30S2B, SL40S2B, SL50S2B,
SL40S2B-1, and SL50S2B-1. In each specimen, the strains
of the straight bar and offset bar are plotted in the right and
left figures, respectively. The horizontal and vertical axes,
respectively, denote the bar strains and height from the base
where the bar strains were measured. The bar strains in
Fig. 9 were the maximum tensile strains that each spliced bar
experienced during the load cycles repeated at each lateral
drift ratio. The bar strains corresponding to δ = 0.25%, 0.5%,
1.0%, 1.5%, 2.0%, and 2.5% are presented.
In Fig. 9, at each lateral drift ratio, the strains of the
bottom offset bars decreased as the height from the base
increased, while the strains of the straight top bars did not
significantly vary along the lap-splice length. Further, the
strains of the bottom offset bars were much greater than
those of the straight top bars. This result indicates that owing
to the moment gradient along the lap-splice length, the bond
demand of the bottom offset bars was much greater than that
of the straight top bars.
The magnitude of the bar strains was significantly
affected by the lap-splice length. The bar strains of SL30S2B
(ls/ls,ACI = 0.57) and SL40S2B (ls/ls,ACI = 0.76) in Fig. 9(a) and
(b) were smaller than the yield strain (εy = 0.00275). Because
of the short splice lengths, the stress of the bottom offset bars
was not fully transferred to the straight top bars. The bar
strains of SL40S2B with greater splice length were greater
than those of SL30S2B with a less splice length.
In SL50S2B (ls/ls,ACI = 1.08) and SL50S2B-1 (ls/ls,ACI =
0.93), on the other hand, the bar strains were significantly
increased by increasing the lap-splice lengths. The strains

ACI Structural Journal/January 2018 241

Fig. 9—Strains of spliced bars at lateral drift ratios 0.25% ~ 2.5%. (Note: 1 mm = 0.039 in.)
tensile stress fsO of the bottom offset bars and the tensile
stress fsS of the straight top bars. As shown in Fig. 10, in
the case of SL30S2B and SL40S2B (a/h = 6.0), the moment
Ms at the top end of the lap-splice region is greater, thereby
increasing stress fsS of the straight top bars and the average
bond demand along the lap-splice length (=[fsO + fsS]/2).
On the other hand, in SL30S1B and SL40S1B (a/h = 3.0),
the average bond demand is smaller. For this reason, the
nominal flexural strength was developed despite the short
splice length (ls/ls,ACI = 0.63 ~ 0.83).

Transverse reinforcement
In SL30S2B and SL40S2B with a hoop spacing of s =
Fig. 10—Moment gradient and spliced bar stresses. (Note: 0.5d, despite the short splice length, ductile behavior was
1 mm = 0.039 in.) maintained, although the test strength was less than the
nominal flexural strength. On the other hand, in SL50S2B-1
of the bottom offset bars exceeded the yield strains (εy =
with the increased hoop spacing of s = 1.0d, web shear
0.00263 or 0.00295). The strain distribution of SL40S2B-1
cracking became severe and vertical bond-splitting cracks
(ls/ls,ACI = 1.01) was similar, although the size of the bottom
propagated along the spliced bars, thereby decreasing the
offset bars was increased to D29 (db = 28.6 mm [1.125 in.]).
deformation capacity. This result indicates that closely
spaced hoops should be used to maintain ductility, avoiding
FACTORS AFFECTING PERFORMANCE OF
premature bond failure under cyclic loading. As illustrated
LAP SPLICE
in Fig. 10, the confining hoops can provide a clamping force
Moment gradient and lap-splice length
across the splitting crack.
As shown in Fig. 7 and 8, the performance of the columns
with lap splices at the plastic hinge was significantly affected
LAP-SPLICE LENGTH AFFECTED BY
by the shear span length. In SL30S1B and SL40S1B with a
MOMENT GRADIENT
smaller shear span length, a = 1200 mm (47.2 in.) (a/h = 3)
In columns, because the bond demand of the spliced bars
(refer to Fig. 7(b) and (c)), the maximum loads Mu were
varies with the moment gradient, the lap-splice length can
greater than the moment strength Mn, although the lap-splice
be redefined considering the moment gradient. The bond
lengths ls were only 63% or 83% of minimum splice lengths
demand for a lap-splice length can be defined as the average
ls,ACI specified in ACI 318-14. In contrast, in SL30S2B and
of the tensile stresses of the bottom offset bars fsO and
SL40S2B with a greater shear span length, a = 2400 mm
straight top bars fsS. fsO can be taken as the yield strength
(94.5 in.) (a/h = 6) (refer to Fig. 7(d) and (e)), the maximum
fyO of the bottom offset bars. On the other hand, fsS of the
loads Mu were less than the moment strength Mn. The lower
straight column bars, which is less than the yield strength
strength of SL30S2B and SL40S2B is attributed to the
fyS, can be estimated considering the moment gradient as
increased bond demand of the spliced bars, which involved
follows (refer to Fig. 10)
the moment gradient in the lap-splice region.
Figure 10 shows the effect of the moment gradient in the
lap slice region. The tensile stress gradients of the bottom  M   l 
f sS ≈ 1 − s  f yO = 1 − s  f yO ≤ f yS (1)
offset bars and straight top bars along the lap-splice length  Mn   a
are also illustrated in the same figure. The solid and dotted
where Mn is the nominal flexure strength at the bottom of
lines denote the moment gradients and spliced bar stresses
the column; Ms is the moment demand at the top of the lap
corresponding to a = 1200 and 2400 mm (47.2 and 94.5 in.)
splice (Mnls/a); a is the shear span length of the column; and
(a/h = 3.0 and 6.0), respectively. The bond demand over
ls is the length of the lap splice. It is noted that, in Eq. (1), fsS
the entire lap-splice length can be defined as the sum of the
of the straight column bars does not depend on fyS, but on fyO

242 ACI Structural Journal/January 2018

Fig. 11—Lap-splice and offset details of existing column specimens.
of the bottom offset bars considering yielding at the bottom
of the column.
By replacing fy in ACI 318-14’s Class B lap-splice length
equation with the average bond demand (fyO + fsS)/2 = fyO
(1 – lsm/[2a]), the modified lap-splice length lsm in columns
can be obtained as follows
 l 
lsm = 1.3 1 − sm  ldO (2)
 2a 
 2  than approximately 1.2. Δu and Δy are the ultimate and yield
or lsm =  ldO (3)
 1.54 + ldO a 
where ldO is the basic development length of the bottom
offset bars specified in ACI 318-14. In Eq. (3), the range of
the modified lap-splice length is defined as 0.87ldO ≤ lsm ≤
1.3ldO depending on the moment gradient. When shear span
a is equal to 1.3ldO, Eq. (3) yields the minimum value of the
lap-splice length (lsm = 0.87ldO). In the case of shear spans
a far greater than ldO, Eq. (3) yields the maximum value
of lsm = 1.3ldO. For conservative design and evaluation, the
shear span length a may be taken as the clear height between
the top and bottom floors. The proposed method is similar to
the result of Ferguson and Krishnaswamy.12
For verification, the modified splice length was compared
with the test results of the present and existing studies.
Figure 11 shows the lap-splice and offset details used in
the existing column specimens considered for the verifi-
cation. The test results of the present and existing studies
are summarized in Table 4 and Table 5, respectively. Twen-
ty-six columns were collected from 10 studies.3-8,10,16-18
The ranges of test parameters are 19.7 MPa (2.86 ksi) ≤ fc′
≤ 41.4 MPa (6.0 ksi), 315 MPa (45.7 ksi) ≤ fy ≤ 617 MPa
(89.5 ksi), 12.5 mm (0.49 in.) ≤ db ≤ 31.8 mm (1.25 in.), 105
mm (4.13 in., 0.21Hmin) ≤ s ≤ 457.2 mm (18.0 in., 1.0Hmin),
and 280 mm (11.0 in., 20db) ≤ ls ≤ 880 mm (34.6 in., 40db).
Figure 12(a) shows the relationships between the
maximum test load-to-nominal strength ratio (Mu/Mn) and
the provided-to-predicted lap-splice length ratio (lst/lsm),
where lst is the lap-splice length used in the tests, and lsm
ACI Structural Journal/January 2018
is the modified lap-splice length calculated by Eq. (3). In
general, the Mu/Mn ratio tends to increase as the lst/lsm ratio
increases. In all columns with lst/lsm ratios not less than 1.0,
except SL40S2B-1, which has different bar diameters in the
spliced bars, the Mu/Mn ratios were equal to or greater than
1.0. This result indicates that Eq. (3) safely estimated the
lap-splice lengths required to develop the nominal flexural
strength in columns with moment gradient. Furthermore, as
shown in Fig. 12(b), the ductility of the test specimens tends
to increase when the lst/lsm ratio increases. The displacement
ductility (Δu/Δy) exceeds 5.0 when the lst/lsm ratio is greater
displacements presented in previous studies, respectively.
SUMMARY AND CONCLUSIONS
Cyclic loading tests of columns with lap splices in the
plastic hinge region were performed. Based on the test results,
the effects of lap-splice length, moment gradient, and trans-
verse reinforcement on the failure mode and deformation
capacity were investigated. Further, a modified lap-splice
length for columns with moment gradient was proposed. The
major findings of this study are summarized as follows.
1. In the specimens with lap-splice length (ls = 30db or
40db) less than the requirement of ACI 318-14, when the
shear span ratio was a/h = 3.0 (SL30S1B and SL40S1B),
the test strength reached the nominal flexural strength,
despite the short splice length. On the other hand, when the
shear-span ratio was a/h = 6.0 (SL30S2B and SL40S2B),
the test strength was less than the nominal strength (Mu =
0.95Mn); however, ductile behavior was maintained until
large inelastic deformation.
2. In Specimen SL50S2B (ls = 50db), with the lap-splice
length satisfying the requirement of ACI 318, despite the
lap splice being located in the plastic hinge zone, the test
strength reached the nominal flexural strength and the ductile
behavior occurred. However, in Specimen SL50S2B-1,
with a greater spacing of hoops (s = 1.0d), the deformation
capacity was decreased.
Based on the present and previous test results, the
following considerations for the design and evaluation of
243
Table 5—Summary of existing tests of columns with lap splices
Specimen properties Lap-splice length Moment strength
Specimen h, mm fc′, MPa fy, MPa N/Agfc′ a, mm s/h db, mm ls, mm lsm, mm ls/lsm Mu, kN·m Mn, kN·m Mu/Mn
3SLH18 25.6 331 0.09 2946 1.0 31.8 635 1141 0.56 415 449 0.92
Lynn et al. 3
2SLH18 457 33.1 331 0.07 2946 1.0 25.4 508 772 0.66 454 333 1.36
3SMD12 25.5 331 0.28 2946 0.67 31.8 635 1091 0.58 590 505 1.17

Aboutaha FC4 19.7 414 0 2743 0.89 25.4 610 1141 0.53 488 553 0.88
457
et al.4 FC5 19.7 414 0 2743 0.89 25.4 610 1141 0.53 488 553 0.88
2S10M 36 510 0.1 1829 0.67 25.4 508 966 0.53 371 432 0.86
2S20M 36 510 0.2 1829 0.67 25.4 508 966 0.53 427 492 0.87

Melek and 2S30M 36 510 0.3 1829 0.67 25.4 508 966 0.53 522 514 1.02
457
Wallace5 2S20H 35 510 0.2 1676 0.67 25.4 508 953 0.53 452 492 0.92
2S20HN 35 510 0.2 1676 0.67 25.4 508 953 0.53 448 492 0.91
2S30X 35 510 0.3 1524 0.67 25.4 508 926 0.55 519 514 1.01
Haroun and
RF-A1 610 41.4 443 0.054 3429 0.21 19.0 381 895 0.43 0.832 1089 0.76
Elsanadedy6
Harries et al.7 L0 458 27.6 414 0.26 2440 0.78 22.2 490 873 0.56 426 389 1.09
C14 39 550 0 1400 0.5 1.0 420 591 0.71 111 119 0.93
Harajili8 C16 400 40 528 0 1400 0.5 16.0 480 629 0.76 135 132 1.02
C20 32 617 0 1400 0.5 20.0 600 1040 0.58 122 183 0.67
24L-300-2D 33.6 315 0.3 1624 0.67 25.4 600 667 0.90 406 406 1.00
Boyes et al.10 450
30L-300-2D 33.9 315 0.3 1624 0.67 25.4 750 664 1.13 426 406 1.05

Bournas L20d_C 27.8 523 0.28 1600 0.8 14.0 280 668 0.42 62 67 0.93
250
et al.16 L40d_C 25.8 523 0.28 1600 0.8 14.0 560 688 0.81 72 67 1.08

ElGawady AB-1 31 331 0.07 1803 0.33 12.5 445 401 1.11 135 112 1.21
381
et al.17 AB-2 31 331 0.07 1803 0.33 12.5 445 401 1.11 130 112 1.16
RL30S1T 32 566 0.17 1200 0.42 22.2 660 859 0.77 227 208 1.09
RL40S1T 32 566 0.10 1200 0.42 22.2 880 859 1.02 206 202 1.02
Kim et al.18 400
RL30S2T 25 565 0.175 2400 0.42 22.2 660 1149 0.57 191 183 1.04
RL40S2T 25 565 0 2400 0.42 22.2 880 1149 0.77 162 164 0.99
Notes: 1 mm = 0.039 in.; 1 MPa = 0.145 ksi.

Fig. 12—Moment strength ratio and ductility according to lap-splice length ratio.
the splice length are recommended for ordinary and interme-
diate moment frames.
1. When the lap-splice length satisfies the requirements of
ACI 318-14, a lap splice of longitudinal bars can be used in
the plastic hinge zone of columns.
2. In the columns with offset longitudinal bars for lap
splice, because the flexural strength can be decreased owing
to the location of the bottom offset bars, the nominal flexural
strength Mn should be calculated considering the locations of
the offset bars in the cross section.

244 ACI Structural Journal/January 2018

 3. To avoid premature bond failure, lap-spliced longitu-
dinal bars should be confined by closely spaced transverse
reinforcement. In this study, s = 0.5d (or 165 mm [6.50 in.])
is recommended for Grade 500 MPa (72.5 ksi) D25 longi-
tudinal reinforcing bars (bar diameter = 25.4 mm [1.0 in.]);
Otherwise, the deformation capacity and energy dissipation
are significantly decreased, even with the splice length spec-
ified in ACI 318-14.
AUTHOR BIOS
Chul-Goo Kim is a Researcher in the Institute of Engineering Research at
Seoul National University, Seoul, South Korea, where he received his BE,
MS, and PhD in architectural engineering. His research interests include
shear behavior and earthquake design of reinforced concrete structures.
ACI member Hong-Gun Park is a Professor in the Department of Architec-
ture & Architectural Engineering at Seoul National University. He received
his BE and MS in architectural engineering from Seoul National University
and his PhD in civil engineering from the University of Texas at Austin,
Austin, TX. His research interests include numerical analysis and seismic
design of reinforced concrete and composite structures.
Tae-Sung Eom is an Associate Professor in the Department of Architec-
tural Engineering at Dankook University, Gyeonggi-do, South Korea. He
received his BE, MS, and PhD in architectural engineering from Seoul
National University. His research interests include experiment and numer-
ical analysis of reinforced concrete and composite structures.
ACKNOWLEDGMENTS
This research was supported by a grant (Code 15AUDP-B066083-03)
from the R&D Policy Infra program funded by the Ministry of Land, Infra-
structure, and Transportation (MOLIT) of the Korean government, and the
research fund of National Research Foundation of Korea (Code R-2015-
00441, Mid-career Research Program). The Institute of Engineering
Research at Seoul National University also provided research facilities for
this work. The authors are grateful for their support.
NOTATION
cb = lesser of: (a) distance from center of bar to nearest concrete
surface; and (b) one-half center-to-center spacing of bars being
developed
d = distance from extreme compression fiber to centroid of longitu-
dinal tension reinforcement
db = nominal diameter of bar
fbd = concrete bond strength
fc′ = specified compressive strength of concrete
fck = characteristic compressive cylinder strength of concrete at 28 days
fsO = tensile stress of bottom offset bar
fsS = tensile stress of straight top bar
fu = ultimate strength for nonprestressed reinforcement
fy = specified yield strength for nonprestressed reinforcement
fyd = design value of yield strength of steel
fyO = yield stress of bottom offset bar
fyS = yield stress of straight top bar
hmin = minimum column dimension
Ktr = transverse reinforcement index
ldO = basic development length of bottom offset bars
ls = lap splice length
lsm = modified lap splice length
Mn = nominal flexural strength at section
Ms = moment demand at top of lap splice
Mu = maximum test moment
s = center-to-center spacing of transverse reinforcement
α1 = influencing factor for shape of bars
α2 = influencing factor for concrete minimum cover
α3 = influencing factor for confinement by transverse reinforcement
not welded to main reinforcement
ACI Structural Journal/January 2018
α4 = influencing factor for confinement by welded transverse
reinforcement
α5 = influencing factor for confinement by transverse pressure
α6 = influencing factor for lapped bars relative to total cross section
area
δ = drift ratio (=Δ/a)
εy = yield strain of steel bar
μ = ductility (=Δu/Δy)
REFERENCES
1. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 2014, 519 pp.
2. Eurocode 2, “Design of Concrete Structures: Part 1-1: General Rules
and Rules for Buildings,” British Standards Institution, London, UK, 2004.
3. Lynn, A. C.; Moehle, J. P.; Mahin, S. A.; and Holmes, W. T., “Seismic
Evaluation of Existing Reinforced Concrete Building Columns,” Earth-
quake Spectra, V. 12, No. 4, 1996, pp. 715-739. doi: 10.1193/1.1585907
4. Aboutaha, R. S.; Engelhardt, M. D.; Jirsa, J. O.; and Kreger, M. E.,
“Experimental Investigation of Seismic Repair of Lap Splice Failures
in Damaged Concrete Columns,” ACI Structural Journal, V. 96, No. 2,
Mar.-Apr. 1999, pp. 297-306.
5. Melek, M., and Wallace, J. W., “Cyclic Behavior of Columns with
Short Lap Splices,” ACI Structural Journal, V. 101, No. 6, Nov.-Dec. 2004,
pp. 802-811.
6. Haroun, M. A., and Elsanadedy, H. M., “Fiber-Reinforced Plastic Jackets
for Ductility Enhancement of Reinforced Concrete Bridge Columns with
Poor Lap-Splice Detailing,” Journal of Bridge Engineering, ASCE, V. 10,
No. 6, 2005, pp. 749-757. doi: 10.1061/(ASCE)1084-0702(2005)10:6(749)
7. Harries, K. A.; Ricles, J. M.; Pessiki, S.; and Sause, R., “Seismic
Retrofit of Lap Splices in Nonductile Square Columns Using Carbon
Fiber-Reinforced Jackets,” ACI Structural Journal, V. 103, No. 6, Nov.-Dec.
2006, pp. 874-884.
8. Harajli, M., “Seismic Behavior of RC Columns with Bond-Critical
Regions: Criteria for Bond Strengthening Using External FRP Jackets,”
Journal of Composites for Construction, ASCE, V. 12, No. 1, 2008,
pp. 69-79. doi: 10.1061/(ASCE)1090-0268(2008)12:1(69)
9. ACI Committee 315, “Details and Detailing of Concrete Reinforce-
ment (ACI 315-99),” American Concrete Institute, Farmington Hills, MI,
1999, pp. 1-44.
10. Boyes, A.; Bull, D. K.; and Pampanin, S., “Seismic Performance of
Concrete Columns with Inadequate Transverse Reinforcement,” University
of Canterbury, Christchurch, New Zealand, 2008, 12 pp.
11. Paulay, T., “Lapped Splices in Earthquake-Resisting Columns,” ACI
Journal Proceedings, V. 79, No. 6, June 1982, pp. 458-469.
12. Ferguson, P. M., and Krishnaswamy, C., “Tensile Lap Splices-Part
2: Design Recommendations for Retaining Wall Splices and Large Bar
Splices,” Research Report No. 113-3, Center for Highway Research,
University of Texas at Austin, Austin, TX, July 1971, 67 pp.
13. ACI Committee 318, “Building Code Requirements for Reinforced
Concrete (ACI 318-56),” American Concrete Institute, Farmington Hills,
MI, 1956, pp. 1-73.
14. ACI Committee 318, “Building Code Requirements for Reinforced
Concrete (ACI 318-77),” American Concrete Institute, Farmington Hills,
MI, 1977, pp. 1-103.
15. ACI Committee 374, “Acceptance Criteria for Moment Frames
Based on Structural Testing and Commentary (ACI 374.1-05),” American
Concrete Institute, Farmington Hills, MI, 2005, pp. 1-9.
16. Bournas, D. A.; Triantafillou, T. C.; Zygouris, K.; and Stavropoulos, F.,
“Textile-Reinforced Mortar Versus Frp Jacketing in Seismic Retrofitting of
RC Columns with Continuous or Lap-Spliced Deformed Bars,” Journal of
Composites for Construction, ASCE, V. 13, No. 5, 2009, pp. 360-371. doi:
10.1061/(ASCE)CC.1943-5614.0000028
17. ElGawady, M.; Endeshaw, M.; McLean, D.; and Sack, R., “Retro-
fitting of Rectangular Columns with Deficient Lap Splices,” Journal of
Composites for Construction, ASCE, V. 14, No. 1, 2010, pp. 22-35. doi:
10.1061/(ASCE)CC.1943-5614.0000047
18. Kim, C. G., Park, H. G., Eom, T. S., and Kim, T. W., “Performance of
Lap Splices in Rectangular RC Columns,” Proceedings of Korea Institute
for Structural Maintenance and Inspection, Oct. 2015, pp. 251-254.
245
NEW
Symposium
Publications
SP-304, Sustainable Performance
of Concrete Bridges and Elements
Subjected to Aggressive
Environments: Monitoring,
Evaluation, and Rehabilitation
This collection contains 10 papers selected from
the three special sessions held at The Concrete
Convention and Exposition in Washington, DC,
October 2014. Topics include: Performance
Reliability of Reinforced Concrete Beams
Strengthened with Fiber-Reinforced Polymer
(FRP) Sheets, Reducing Deck Cracking in
Composite Bridges by Controlling Long-Term
Properties, and many more.
Available in CD/DVD or PDF format: $76.50
(ACI members: $46.00) ($30.50 savings)
+1.248.848.3700 • www.concrete.org
from ACI
SP-305, Durability and Sustainability
of Concrete Structures
In October 2015, the Italy Chapter – ACI (ACI IC) and
the Department of Civil, Chemical, Environmental, and
Material Engineering (DICAM) of the University of
Bologna sponsored the First International Workshop
on “Durability & Sustainability of Concrete Structures”
in Bologna, Italy. The proceedings of the workshop
consist of 48 papers concerning reduction in
greenhouse gases in cement and concrete industry,
recycled materials, innovative binders and
geopolymers, life-cycle cost assessment in concrete
construction, reuse and functional resilience of
reinforced concrete structures, repair and main-
tenance, testing, inspection, and monitoring.
Available in PDF format: $101.50
(ACI members: $61.00) ($40.50 savings)
   
 ACI STRUCTURAL JOURNAL
Title No. 115-S20
Conventional and High-Strength Steel Hooked Bars:
Detailing Effects
by J. Sperry, D. Darwin, M. O’Reilly, A. Lepage, R. D. Lequesne, A. Matamoros, L. R. Feldman,
S. Yasso, N. Searle, M. DeRubeis, and A. Ajaam
Findings from a study on the effect of hook bend angle, concrete
clear cover, and orientation of confining reinforcement on hook
anchorage strength are presented. The range of test parameters
was much broader than in previous studies. Bar stress at anchorage
failure ranged from 33,000 to 137,400 psi (228 to 947 MPa) and
concrete compressive strengths ranged from 4300 to 16,500 psi
(30 to 114 MPa). Anchorage strength of hooked bars was insensi-
tive to bend angle (90 or 180 degrees) and side cover (between 2.5
and 3.5 in. [65 and 90 mm]). Confining reinforcement was found to
increase anchorage strength for 180-degree hooked bars regardless
of orientation (parallel or perpendicular to the embedment length).
For 90-degree hooked bars, reinforcement oriented parallel to the
embedment length had a greater effect on anchorage strength than
reinforcement oriented perpendicular to the embedment length.
Keywords: anchorage; beam-column joints; bond; development length;
high-strength concrete; high-strength steel; hook bend angle; reinforced
concrete; side cover.
INTRODUCTION
Current design provisions for anchorage of hooked bars
in reinforced concrete (ACI Committee 318 2014; ACI
Committee 349 2006; AASHTO 2012) are based on several
assumptions about their behavior: among others, hooked bars
with 90- and 180-degree bend angles are assumed to have
similar strengths, hooked bars with side covers of 2.5 in.
(65  mm) or greater have similar strengths, and confining
reinforcement oriented parallel or perpendicular to the
straight portion of a 90-degree hooked bar (“straight portion
of the hooked bar” refers to the straight portion of the bar in
the direction of the embedment or development length) is
assumed to be equally effective in providing confinement,
but only confining reinforcement oriented perpendicular to
the straight portion of a 180-degree hooked bar is assumed
to be effective.
The design provisions are based on 38 tests by Marques
and Jirsa (1975) and Pinc et al. (1977) of beam-column
joint specimens containing Grade 60 (420) No. 7, No. 9, or
No. 11 (No. 22, No. 29, or No. 36) bars with standard hooks
(ACI Committee 318 2014), and concrete with compressive
strengths ranging from 3600 to 5400 psi (25 to 37 MPa).
Marques and Jirsa (1975) observed that the thickness of the
concrete cover had a significant effect on the slip and stress
at failure but indicated no advantage for covers greater than
2.5 in. (65 mm). None of the test specimens in these earlier
studies contained confining reinforcement perpendicular to
the straight portion of the hooked bars.
To validate the applicability of the earlier findings, tests
were performed in this study to evaluate the effects of
hook bend angle, concrete clear cover, and orientation of
ACI Structural Journal/January 2018
TECHNICAL PAPER
confining reinforcement on hook anchorage strength for a
broader range of steel and concrete strengths than used in the
earlier studies. Additional results and analyses are presented
by Sperry et al. (2015a,b; 2017a,b).
RESEARCH SIGNIFICANCE
The use of high-strength steel and concrete as a means
of reducing reinforcement congestion, member dimensions,
and material use has increased. The analysis reported herein
is the first to evaluate whether previous findings showing
that bars with 90- and 180-degree hooks have equivalent
anchorage strengths are valid for high-strength materials
and the first to establish the effect of increasing concrete
clear cover from 2.5 to 3.5 in. (65 to 90 mm). The effective-
ness of confining reinforcement oriented either parallel or
perpendicular to the straight portion of hooked bars is also
evaluated to confirm whether current ACI 318 provisions
are appropriate.
EXPERIMENTAL INVESTIGATION
Tests of 166 simulated beam-column joint specimens
containing two hooked bars, included as part of a larger
research program (Sperry et al. 2015a,b; 2017a,b), were used
to investigate the effect of bend angle, side cover, and orien-
tation of confining reinforcement on anchorage strength.
No. 5, 8, and 11 (No. 16, 25, and 36) hooked bars were
tested in normalweight concrete with compressive strengths
ranging from 4300 to 16,500 psi (30 to 114 MPa). Nominal
clear cover from the hooked bars to the outside of the column
(side cover) ranged from 2.5 to 3.5 in. (65 to 90 mm).
Bar stresses at anchorage failure ranged from 33,000 to
137,400 psi (228 to 947 MPa). The results of these tests are
reported and used in conjunction with the results of previous
studies (Marques and Jirsa 1975; Pinc et al. 1977; Hamad
et al. 1993; Ramirez and Russell 2008; Lee and Park 2010)
to determine the effect on hooked bar anchorage strength
of bend angle, concrete side cover, and confining reinforce-
ment orientation. The details of the experimental investiga-
tion are provided by Sperry et al. (2015a,b).
Test specimens
Figure 1 shows side and plan views of a typical specimen.
Specimens were designed to represent exterior beam-column
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2017-075, doi: 10.14359/51700920, was received March 3, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
247

Fig. 1—Simulated beam-column specimens: (a) side view;
(b) cross section without confining reinforcement; and
(c) cross section of specimen with confining reinforce-
ment parallel to straight portion of hooked bar (Sperry et
al. 2017a).
joints and were cast without the beam. They were similar
to the specimens used in the studies by Marques and Jirsa
(1975) and Pinc et al. (1977). The specimens analyzed
in this paper contained two hooked bars cast inside the
column longitudinal reinforcement. The out-to-out spacing
of the hooked bars was the same for each bar size—8, 12,
and 16.5 in. (203, 305, and 419 mm) for specimens with
No. 5, No. 8, and No. 11 (No. 16, 25, and 36) hooked bars,
respectively. Total column widths are given in Appendix A.*
In this paper, embedment length ℓeh refers to the distance
from the front of the column face to the back of the tail of
the hook. This is in contrast to the development length ℓdh,
which refers to the minimum length of anchorage required
in Section 25.4.3 of ACI 318-14 (ACI Committee 318 2014)
to ensure that a bar can develop its yield strength. Column
reinforcement was proportioned to resist the maximum
expected shear and moment assuming that both hooked bars
reached their expected strength simultaneously.
The specimens contained one of three quantities of
confining reinforcement, in all but six cases oriented parallel
to the straight portion of the hooked bar: 1) no confining
reinforcement; 2) two No. 3 (No. 10) hoops spaced along
the length of the tail of the hook; or 3) No. 3 (No. 10) hoops
spaced at three bar diameters (3db) along the tail and the
bend of the hook, where db is the diameter of the hooked bar.
No. 3 (No. 10) hoops spaced at 3db represents the amount
of confining reinforcement required in Section  25.4.3
of ACI 318-14 to allow the use of the confining reinforce-
ment modification factor ψr = 0.8 in the calculation of the
development length of hooked bars. At this spacing, five
*
The Appendix is available at www.concrete.org/publications in PDF format,
appended to the online version of the published paper. It is also available in hard copy
from ACI headquarters for a fee equal to the cost of reproduction plus handling at the
time of the request.
248
Fig. 2—Specimen with confining reinforcement perpendic-
ular to straight portion of hooked bar: (a) side view; and
(b) cross section (Sperry et al. 2017a).
No. 3 (No. 10) hoops were used to confine No. 5 (No. 16)
and No. 8 (No. 25) standard hooks and six No. 3 (No. 10)
hoops were used to confine No. 11 (No. 36) standard hooks.
For case (3), the first hoop was centered 2db from the top
of the hooked bar (1.5db from the longitudinal axis of the
straight portion of the hooked bar). For case (2), in speci-
mens with No. 5 or No. 8 hooked bars, the first hoop was
spaced 3 in. (76.2 mm) from the center of the straight portion
of the hooked bars; the second hoop was spaced 3 and 8 in.
(76.2 and 203.2 mm), respectively, from the first hoop. For
case (2) in specimens with No. 11 hooked bars, the first and
second hoops were spaced at 8 in. (203.2 mm) intervals from
the center of the straight portion of the hooked bars.
To evaluate the effect of reinforcement orientation,
six specimens were tested with confining reinforcement
oriented perpendicular to the straight portion of the hooked
bar, as shown in Fig. 2. Of the six, two contained two No. 3
(No. 10) hoops, two contained four No. 3 (No. 10) hoops,
and two contained five No. 3 (No. 10) hoops. The latter two
cases meet the requirement to allow the use of ψr = 0.8.
Specimen heights were chosen so that the support reac-
tions from the test frame had minimal effect on the hook
region during testing, as shown in Fig. 3. The column height
was 52-3/4 in. (1340 mm) for the specimens with No. 5
(No. 16) or No. 8 (No. 25) hooked bars and 96 in. (2438 mm)
for the specimens with No. 11 (No. 36) hooked bars. The
distances from the longitudinal axis of the straight portion of
the hooked bar to the upper compression member hcu were
18.5, 18.5, and 48.5 in. (470, 470, and 1232 mm) for No. 5,
No. 8, and No. 11 (No. 16, No. 25, and No. 36) hooked
bars, respectively, and the distances from the center of the
hooked bar to the bearing member, simulating the compres-
sion region of the concrete beam framing into the column
hcl were 5.25, 10, and 19.2 in. (133, 254, and 488 mm) for
No. 5, No. 8, and No. 11 (No. 16, No. 25, and No. 36) hooked
bars, respectively.
ACI Structural Journal/January 2018

Fig. 3—Testing frame and forces applied to specimens
during testing (Sperry et al. 2017b).
Material properties
Specimens were cast using non-air-entrained ready mixed
concrete with nominal compressive strengths of 5000, 8000,
12,000, and 15,000 psi (34, 55, 83, and 103 MPa). Measured
strengths obtained using 6 x 12 in. (150 x 300 mm) cylin-
ders cured in the same manner as the test specimen and
tested on the same day as the anchorage tests ranged from
4300 to 16,500 psi (30 to 114 MPa). The concrete contained
Type I/II portland cement, crushed limestone, or granite
with a maximum size of 0.75 in. (19 mm), Kansas river
sand, and a high-range water-reducing admixture. Pea
gravel was incorporated as a portion of the aggregate in the
12,000 psi (83 MPa) concrete to improve the workability
of the mixture. Silica fume and Class C fly ash were used
as supplementary cementitious materials for the 15,000 psi
(103 MPa) concrete. Polycarboxylate-based high-range
water-reducing admixtures were used in the mixtures.
Mixture proportions are presented in Appendix A and by
Sperry et al. (2015a,b, 2017a).
Hooked bars were fabricated using ASTM A615 Grade 80
(550 MPa) and A1035 Grade 120 (830 MPa) reinforce-
ment. For most specimens, the ancillary steel for column
and confining reinforcement consisted of ASTM A615
Grade 60 (420 MPa) reinforcing bars. ASTM A1035 Grade
120 (830 MPa) bars were used as the column longitudinal
steel for specimens that had a greater flexural demand (based
on the expected strength of the hooked bars) than could be
satisfied using ASTM A615 Grade 60 (420 MPa) reinforcing
bars. Yield strength, nominal diameter, deformation spacing
and height, gap width, and relative rib area for the deformed
steel bars used as hooked bars are presented in Appendix A
and in Sperry et al. (2017b). Appendix A also includes repre-
sentative stress-strain curves for the hooked bars.
ACI Structural Journal/January 2018
Test procedure
Specimens were tested using a self-reacting system config-
ured to simulate the boundary conditions of a beam-column
joint (Fig. 3). The test frame was a modified version of the
apparatus used by Marques and Jirsa (1975). The flange width
of the upper compression member and the bearing member
were 6-5/8 and 8-3/8 in. (168 and 213 mm), respectively.
For specimens with No. 5 and No. 8 (No. 16 and No. 25)
hooked bars, a constant column axial load of 30,000 lb
(133 kN) was applied to most of the specimens, corre-
sponding to a range in axial stress of 90 to 460 psi (0.6
to 3.2 MPa). For early tests, a constant force of 80,000 lb
(356 kN) was used, corresponding to an axial stress range
of 505 to 1930 psi (3.5 to 13.3 MPa) based on column size.
Specimens with No. 11 hooked bars had a constant axial
stress of 280 psi (1.9 MPa) applied. These axial stresses
were chosen based on the capacity of the axial load appli-
cation system. Marques and Jirsa (1975) found that changes
in axial stress up to 3000 psi (21 MPa) resulted in negligible
changes in the anchorage strength of the hooked bars; the
effect of varying axial stress was therefore not examined.
The load was applied monotonically to the hooked bars
using hydraulic jacks to simulate tensile forces in the beam
reinforcement at the face of a beam-column joint. Specimens
were loaded to approximately 80% of the expected failure
load in, for most cases, four increments. Between incre-
ments, cracks were marked and photographs were taken.
Above 80% of the expected failure load, the specimens were
loaded continually to failure. Tests lasted 15 to 25 minutes.
Detailed descriptions of the test frame and testing procedure
are provided by Peckover and Darwin (2013).
RESULTS AND ANALYSIS
Results from 166 tests of beam-column joint specimens
with No. 5, No. 8, and No. 11 (No. 16, No. 25, and No. 36)
hooked bars were selected from the data presented by
Sperry et al. (2015a,b, 2017b) to evaluate the effect of hook
bend angle, concrete side cover, and confining reinforce-
ment orientation on anchorage strength. These results were
combined with selected test results from Marques and Jirsa
(1975), Pinc et al. (1977), Hamad et al. (1993), Ramirez and
Russell (2008), and Lee and Park (2010). The forces applied
to the hooked bars at failure for the specimens included in
the analysis are presented in Appendix A. As described by
Sperry et al. (2017b), the average bar force at the peak load,
equal to the maximum total force applied to a specimen
divided by the number of hooked bars under load, is treated
as the failure load per hooked bar and used to calculate the
average bar stress at failure.
To limit the effect of concrete compressive strength
on anchorage strength and simplify the comparisons, the
average bar forces at failure were normalized with respect
to a concrete compressive strength of 5000 psi (34.5 MPa)
by multiplying the average bar forces at failure T by
(5000 fcm ) p 1 ( [34.5 fcm ] p 1). The result is reported as the
normalized average failure load TN. A value of p1 = 0.25 was
selected based on the observation by Sperry et al. (2015a,b;
2017a,b) that the power of 0.5, currently used in the provi-
sions of ACI 318-14 (ACI Committee 318 2014), greatly
249
Fig. 4—Bar force at failure normalized to concrete compres-
sive strength of 5000 psi (34.5 MPa) versus embedment
length for No. 5, 7, 8, and 11 (No. 16, 22, 25, and 36)
hooked bars without confining reinforcement. (Note: 1 kip =
4.448 kN; 1 in. = 25.4 mm.)
overestimates the effect of concrete compressive strength
on anchorage strength, and that values of p1 in the range of
0.25 to 0.29 provide a realistic representation of the effect
of concrete compressive strength on the anchorage strength
of hooked bars. This observation is further supported by test
results for straight bar lap splices (Darwin et al. 1995, 1996;
Zuo and Darwin 1998, 2000; Joint ACI-ASCE Committee
408 2003) and headed bars (Shao et al. 2016), which indicate
that a value of p1 = 0.25 characterizes the effect of concrete
compressive strength on both straight bar bond and headed
bar anchorage strength.
In the comparisons that follow, a regression analysis tech-
nique based on dummy variables (Draper and Smith 1981)
is used to identify trends in the data. Dummy variables anal-
ysis is a least-squares regression analysis method that allows
differences in populations to be taken into account when
formulating relationships between principal variables. For
example, the effect of embedment length ℓeh on bar force
at failure T can be found for different bar sizes based on the
assumption that the effect of changes in ℓeh on changes in T
is the same for the bar sizes considered, but that the absolute
value of T for a given ℓeh will differ for each bar size.
Effect of bend angle
Figure 4 shows the normalized average failure loads TN
as a function of embedment length for a subset of 58 beam-
column specimens, of which 39 are from the current study.
These specimens contained No. 5, No. 7, No. 8, and No. 11
(No. 16, No. 22, No. 25, and No. 36) hooked bars without
confining reinforcement in the joint region, with bend angles
of 90 or 180 degrees. The results for the nine No. 7 (No. 22)
hooked bar tests and 11 of the No. 11 (No. 36) hooked bar
tests were taken from studies by Marques and Jirsa (1975),
Pinc et al. (1977), Hamad et al. (1993), Ramirez and Russell
(2008), and Lee and Park (2010). The solid lines and data
points represent the results for 90-degree hooked bars, while
the broken lines and open data points represent the results
for 180-degree hooked bars. Both trend lines and data points
250
Fig. 5—Bar force at failure normalized to concrete compres-
sive strength of 5000 psi (34.5 MPa) versus embedment length
for No. 5 and 8 (No. 16 and 25) hooked bars confined by two
No. 3 (No. 10) hoops oriented parallel to straight portion of
hooked bar. (Note: 1 kip = 4.448 kN; 1 in. = 25.4 mm.)
are identified based on bar size. In this figure and those that
follow, the order of the results in the legend coincides with
the order of the lines in the figure. For each bar size, the range
of embedment lengths was similar for 90- and 180-degree
hooked bars. The embedment lengths ℓeh ranged from 6.3 to
21.1 in. (160 to 536 mm), the bar stresses at failure ranged
from 42,700 to 136,100 psi (294 to 938 MPa), and the
normalized average bar forces at failure ranged from 19,300
to 114,400 lb (86 to 509 kN). The concrete compressive
strengths ranged from 2570 to 16,500 psi (17.7 to 114 MPa).
As shown in Fig. 4, an increase in embedment length is
associated with an increase in the normalized average bar
force at failure. The results in Fig. 4 show no clear correla-
tion between anchorage strength and bend angle. For No. 5,
7, and 11 (No. 16, 22, and 36) hooked bars, the trend line
corresponding to a 90-degree bend angle is higher than the
trend line corresponding to a 180-degree bend angle. The
opposite trend is observed for No. 8 (No. 25) hooked bars.
The difference between intercepts of the trend lines corre-
sponding to 90- and 180-degree bend angles is greater for
the No. 11 (No. 36) bars than for the smaller bars. The results
were evaluated using Student’s t-test by comparing the inter-
cepts obtained by extending a line with the same slope as the
trend lines from each data point to the TN axis. That evalu-
ation shows that the differences in anchorage strength for
bars with bend angles of 90 and 180 degrees are not statis-
tically significant at the 95% confidence level (5% or lower
risk of concluding that a difference exists when there is no
actual difference) regardless of bar size. In the cases shown
in Fig. 4, the variable p ranges between 0.13 and 0.80,
where p less than 0.05 would indicate that the differences
in anchorage strength based on bend angle are significant
(Wonnacott and Wonnacott 1985).
The relationship between normalized failure load and
embedment length for 26 beam-column specimens (all from
the current study) containing No. 5 or No. 8 (No. 16 and
25) hooked bars with 90- or 180-degree hooks and with two
No. 3 (No. 10) hoops in the joint region is shown in Fig. 5.
ACI Structural Journal/January 2018
Fig. 6—Bar force at failure normalized to concrete compres-
sive strength of 5000 psi (34.5 MPa) versus embedment
length for No. 8 and 11 (No. 25 and 36) hooked bars with
confining reinforcement conforming to Section 25.4.3.2
of ACI 318-14 and oriented parallel to straight portion of
hooked bar. (Note: 1 kip = 4.448 kN; 1 in. = 25.4 mm.)
The two hoops were oriented parallel to the straight portion
of the hooked bars. Two hoops are insufficient to satisfy ACI
Code (ACI 318-14) requirements for the use of a develop-
ment length modification factor ψr = 0.8 for hooked bars, and
hoops oriented parallel to the straight portion of the hooked
bar, regardless of number or spacing, are not considered by
the Code to increase the anchorage strength of bars with
180-degree hooks. Contrary to the Code, however, Sperry
et al. (2015a,b; 2017a,b) have shown that hoops placed
parallel to the straight portion of hooked bars provide similar
increases in anchorage strength for both 90- and 180-degree
hooked bars. The embedment lengths ℓeh ranged from 5.6 to
17.3 in. (142 to 439 mm), the normalized average bar forces
at failure TN ranged from 20,000 to 87,500 lb (89 to 389 kN),
the bar stresses at failure ranged from 68,000 to 137,400 psi
(469 to 947 MPa), and the concrete compressive strengths
ranged from 4300 to 15,800 psi (30 to 109 MPa).
The trend lines for anchorage strength nearly coincide for
the 90- and 180-degree No. 5 (No. 16) hooked bars, while the
180-degree No. 8 (No. 25) hooked bars had a lower strength
than the 90-degree No. 8 (No. 25) hooked bars. The results
of Student’s t-test show that the differences in anchorage
strength for No. 5 and 8 (No. 16 and 25) bars with 90- or
180-degree hooks are not statistically significant at the 95%
confidence level, with p = 0.81 and 0.12, respectively.
Figure 6 compares the anchorage strengths of No. 8
and No. 11 (No. 25 and 36) bars with 90- and 180-degree
hooks confined by No. 3 (No. 10) hoops oriented parallel
to the straight portion of the hooked bar and spaced at 3db,
which satisfies the requirements for use of the development
length modification factor ψr = 0.8 for 90-degree hooks in
ACI 318-14, Section 25.4.3.2. The data in Fig. 6 represent
18 specimens, all tested as part of this study, containing
No. 8 or No. 11 (No. 25 or No. 36) hooked bars with 90- and
180-degree bend angles. The embedment lengths ℓeh ranged
from 9.4 to 20.4 in. (239 to 518 mm), the normalized average
bar forces at failure TN ranged from 51,700 to 133,600 lb (230
ACI Structural Journal/January 2018
Fig. 7—Bar force at failure normalized to concrete compres-
sive strength of 5000 psi (34.5 MPa) versus embedment
length for No. 5 (No. 16) hooked bars with different amounts
of No. 3 (No. 10) bar confining reinforcement oriented
parallel to straight portion of hooked bar and side cover
(sc). (Note: 1 kip = 4.448 kN; 1 in. = 25.4 mm.)
to 594 kN), the bar stresses at failure ranged from 69,400 to
113,900 psi (478 to 785 MPa), and the concrete compressive
strengths ranged from 5420 to 15,800 psi (37.4 to 109 MPa).
For both the No. 8 and No. 11 (No. 25 and No. 36) hooked
bars, the anchorage strength of the hooked bars with a bend
angle of 180 degrees was slightly lower than the strength of
the hooked bars with a 90-degree bend angle, although the
results of Student’s t-test show that the differences are not
statistically significant (p = 0.54 and 0.50, respectively).
Because differences between the anchorage strengths
of hooked bars with 90- and 180-degree bend angles were
found to be small and not statistically significant, hooked
bars with either bend angle, and with all other parameters
the same, should be treated as having the same anchorage
strength, as reflected in the design provisions of ACI 318-14.
Further, confining reinforcement parallel to the straight
portion of hooks is shown to provide the same contribution
to the anchorage strength for both 90- and 180-degree stan-
dard hooks.
Effect of side cover
Based on the observations that bend angle has no measur-
able effect on anchorage strength, the comparisons in this
section include specimens with both 90- and 180-degree
bend angles. The relationship between normalized anchorage
strength TN and embedment length ℓeh for 39 beam-column
joint specimens containing No. 5 (No. 16) hooked bars is
shown in Fig. 7. The specimens in Fig. 7 were tested as part
of this study and had nominal side covers of 2.5 in. (65 mm)
(solid lines) or 3.5 in. (90 mm) (broken lines). Three different
quantities of confining reinforcement (parallel to the straight
portion of the hooked bar) were investigated: no confining
reinforcement; two No. 3 (No. 10) hoops within the joint
region; and No. 3 (No. 10) hoops spaced at 3db (satisfying
the requirements for ψr = 0.8). The embedment lengths ℓeh
ranged from 3.75 in. to 10.5 in. (95 to 267 mm). The normal-
ized failure load TN ranged from 14,000 to 41,500 lb (62 to
251
Fig. 8—Bar force at failure normalized to concrete compres-
sive strength of 5000 psi (34.5 MPa) versus embedment
length for No. 8 (No. 25) hooked bars with different amounts
of No. 3 (No. 10) bar confining reinforcement oriented
parallel to straight portion of hooked bar and side cover
(sc). (Note: 1 kip = 4.448 kN; 1 in. = 25.4 mm.)
185 kN), the bar stresses at failure ranged from 60,300 to
136,100 psi (416 to 938 MPa), and the concrete compressive
strengths ranged from 5190 to 15,800 psi (36 to 109 MPa). The
trend lines in Fig. 7 show that anchorage strength increased
with increasing embedment length and amount of confining
reinforcement. Regardless of the amount of confining rein-
forcement, the results shown in Fig. 7 indicate that there was
a decrease in strength as the side cover increased from 2.5
to 3.5 in. (65 to 90 mm), although the results from Student’s
t-test show that this decrease was not statistically significant
at the 95% confidence level, with p = 0.72, 0.08, and 0.30
for specimens without confining reinforcement, specimens
with two No. 3 (No. 10) hoops, and specimens with No. 3
(No. 10) hoops spaced at 3db, respectively.
The relationship between normalized failure load TN and
embedment length ℓeh for beam-column joint specimens with
No. 8 (No. 25) hooked bars with nominal side covers of 2.5
or 3.5 in. (65 or 90 mm) is shown in Fig. 8. The data represent
78 specimens tested as part of this study. The embedment
lengths ℓeh ranged from 6.1 to 18.7 in. (155 to 475 mm), the
concrete compressive strengths ranged from 4300 to 16,500
psi (30 to 114 MPa), the bar stresses at failure ranged from
44,430 to 120,700 psi (306 to 832 MPa), and the normal-
ized failure load TN ranged from 28,200 to 93,600 lb (125 to
416 kN). The trend lines in Fig. 8 show that the normalized
failure load increased with increasing embedment length and
amount of confining reinforcement. For specimens without
confining reinforcement, increasing the side cover from
2.5 to 3.5 in. (65 to 90 mm) led to increases in anchorage
strength. Specimens with confining reinforcement included
configurations with two No. 3 (No. 10) hoops and with five
No. 3 (No. 10) hoops spaced at 3db in the joint region. For
both of these configurations, specimens with 3.5  in. (90
mm) side cover had normalized failure loads slightly lower
than those of specimens with 2.5 in. (65 mm) side cover.
The results of Student’s t-test show that the differences in
anchorage strength associated with changes in cover for
252
Fig. 9—Bar force at failure normalized to concrete compres-
sive strength of 5000 psi (34.5 MPa) versus embedment length
for No. 11 (No. 36) hooked bars with different amounts of
No. 3 (No. 10) bar confining reinforcement oriented parallel
to straight portion of hooked bar and side cover (sc). (Note:
1 kip = 4.448 kN; 1 in. = 25.4 mm.)
specimens with two No. 3 (No. 10) hoops and No. 3 (No.
10) hoops spaced at 3db are not statistically significant at the
95% confidence level, with p equal to 0.32 and 0.47, respec-
tively. For specimens without confining reinforcement, the
difference in anchorage strength between hooked bars with
2.5 or 3.5 in. (65 or 90 mm) side cover is statistically signif-
icant (p = 0.03).
The relationship between normalized bar force at failure
TN and embedment length ℓeh is shown in Fig. 9 for 49 beam-
column joint specimens with No. 11 (No. 36) hooked bars
with nominal side covers of 2.5 or 3.5 in. (65 or 90 mm)
tested as part of this study. The embedment lengths ℓeh
ranged from 9.5 to 26.0 in. (241 to 660 mm), the concrete
compressive strengths ranged from 4910 to 16,180 psi (34 to
112 MPa), the bar stresses at failure ranged from 33,000 to
136,700 psi (228 to 943 MPa), and the normalized average
bar forces at failure TN ranged from 39,800 to 174,400 lb
(177 to 776 kN). Similar to the test results for the smaller
hooked bars, anchorage strength increased with embed-
ment length and the amount of confining reinforcement. For
specimens without confining reinforcement and specimens
with two No. 3 (No. 10) hoops, the change in the normal-
ized failure load was negligible as side cover increased from
2.5 to 3.5 in. (65 to 90 mm). For specimens with six No. 3
(No. 10) hoops spaced at 3db, the trend lines show that there
was a small decrease in normalized failure load as side cover
increased from 2.5 to 3.5 in. (65 to 90 mm). Because there
was only one specimen with 3.5 in. (90 mm) side cover, the
most meaningful conclusion that can be drawn from the data
is that the normalized failure load was comparable for the
two configurations. The results of Student’s t-test indicate
that the differences in anchorage strength associated with
changes in side cover for specimens without confining rein-
forcement and specimens with two No. 3 (No. 10) hoops
are not statistically significant at the 95% confidence level
(p = 0.56 and 0.82, respectively). Student’s t-test cannot
be performed for the specimens with No. 3 (No. 10) hoops
ACI Structural Journal/January 2018
Fig. 10—Plan view of hooked bars with confining reinforcement oriented perpendicular to bar being developed satisfying
maximum spacing requirement in ACI 318-14, Section 25.4.3.2: (a) four No. 3 (No. 10) hoops; and (b) five No. 3 (No. 10)
hoops. (Note: 1 in. = 25.4 mm.)
spaced at 3db because there was only one specimen with
3.5 in. (90 mm) side cover.
Of all the Student’s t-tests performed, in only one instance,
No. 8 (No. 25) hooked bars without confining reinforcement
was the value of p less than 0.05, indicative of a statistically
significant difference between the anchorage strength of
hooked bars with 2.5 in. (65 mm) side cover and hooked bars
with 3.5 in. (90 mm) side cover. Overall, the results indicate
that anchorage strength was not affected by differences in
side cover in the range of 2.5 to 3.5 in. (65 to 90 mm).
Effect of orientation of confining reinforcement
To take advantage of the modification factor ψr = 0.8 for
development length of hooked bars with a 90-degree bend
angle, ACI 318-14, Section 25.4.3.2, requires confining
reinforcement spaced at ≤3db and placed perpendicular or
parallel to the straight portion of the bar being developed.
For hooked bars with a 180-degree bend angle, ψr = 0.8 can
only be applied for confining reinforcement oriented perpen-
dicular to the straight portion of the bar. The equivalence of
the contribution of confining reinforcement oriented parallel
to the straight portion of a hooked bar to the anchorage
strength of bars with 90- and 180-degree hooks has already
been demonstrated in this paper. Prior to this study, no tests
had been performed to evaluate the contribution to anchorage
strength of confining reinforcement oriented perpendicular
to the straight portion of a hooked bar.
To address this question, a test series was planned with
12 similar beam-column joint specimens cast with the same
batch of concrete. This set of specimens contained No. 8
(No. 25) hooked bars with bend angles of 90 or 180 degrees.
The specimens had embedment lengths ℓeh with nominal
values of 10, 11, and 12.5 in. (254, 279, and 318 mm). The
corresponding column cross-sectional dimensions were
17 x 12 in., 17 x 13 in., and 17 x 14.5 in. (432 x 305 mm,
432 x 330 mm, and 432 x 368 mm). Compressive strengths
on the day of the tests were similar, ranging from 11,800 to
12,000 psi (81 to 83 MPa). Average measured embedment
lengths ranged from 9.4 to 12.8 in. (234 to 325 mm), bar
stresses at failure ranged from 66,600 to 95,200 psi (459 to
656 MPa), and average failure loads ranged from 60,200
to 75,200 lb (268 to 335 kN). Of the 12 specimens, six
contained hooked bars with a 90-degree bend angle and six
ACI Structural Journal/January 2018
contained hooked bars with a 180-degree bend angle. For
both sets of six, one specimen was cast without confining
reinforcement, one contained two hoops oriented parallel to
the bars being developed (parallel hoops), one contained two
hoops oriented perpendicular to the bars being developed
(perpendicular hoops), one contained parallel hoops spaced
at 3db, and two contained perpendicular hoops spaced at less
than 3db.
In the specimens with parallel hoops placed along the tail
of the hook, a minimum of five hoops were needed to meet
the 3db spacing requirement. In the specimens with perpen-
dicular hoops, the specimen dimensions and the depth of the
joint were such that only four hoops were needed to meet
the 3db spacing requirement specified in the code provisions
(ACI 318-14, Section 25.4.3.2). To evaluate the effect of
hoop orientation on anchorage strength without favoring the
parallel orientation, two different reinforcement configura-
tions were used in specimens with perpendicular hoops satis-
fying the 3db maximum spacing requirement—one with the
minimum of four hoops needed to achieve the 3db spacing
and one with five hoops to match the area of confining rein-
forcement used in the specimens with parallel hoops spaced
at 3db. The two configurations are shown in Fig. 10. For
specimens with a 180-degree bend angle, parallel hoops
were placed throughout the region defined by the bend and
tail of a 90-degree hooked bar, as shown in Fig. 1.
Because all of the specimens in this batch had different
embedment lengths ℓeh and slightly different concrete
compressive strengths fcm, the average bar forces at failure
T were normalized with respect to an embedment length
of 10 in. (254 mm) and a concrete compressive strength of
12,000 psi (83 MPa) by multiplying the average bar forces at
failure T by (10  eh )(12,000 f cm ) ( 254  eh )(83 f cm ) 
0.25 0.25
to obtain the normalized average failure load TN.
The values of TN for the specimens with two No. 3
(No. 10) hoops and the specimens with No. 3 (No. 10) hoops
spaced at ≤3db are shown in the bar graph in Fig. 11. The
first set of four bars in the graph shows TN for the 90- and
180-degree hooked bars confined by two hoops oriented
parallel or perpendicular to the straight portion of the bar.
As shown for these four specimens, the value of TN for the
specimen containing 90-degree hooked bars confined by
parallel hoops was 1.27 times the value of TN for the spec-
253
Fig. 11—Bar force at failure normalized to embedment
length of 10 in. (254 mm) and concrete compressive strength
of 12,000 psi (83 MPa) for specimens containing No. 8
(No.  25) hooked bars with No. 3 (No. 10) confining rein-
forcement oriented parallel or perpendicular to straight
portion of hooked bar and 90- or 180-degree bend angles.
(Note: 1 kip = 4.448 kN.)
imen containing 90-degree hooked bars confined by perpen-
dicular hoops. For the specimens containing 180-degree
hooked bars, TN for the specimens with the two orientations
of confining reinforcement had similar strengths—TN for the
specimen with perpendicular hoops was 1.01 times TN for
the specimen with parallel hoops.
The second and third sets of four bars in Fig. 11 show the
results for specimens with hoops spaced ≤3db. In each of
these sets, two specimens were cast with hoops oriented in
the parallel direction, one with 90-degree hooked bars and
the other with 180-degree hooked bars, both having five
hoops spaced ≤3db. Because there were no specimens with
four hoops with a parallel orientation, specimens with five
hoops are shown in the first and third columns of the second
set for reference. The trends for the specimens with hoops
spaced ≤3db are similar to those observed for specimens with
two hoops. The value of TN for the specimen with 90-degree
hooked bars confined by parallel hoops is higher than TN for
the specimens containing 90-degree hooked bars confined
by perpendicular hoops, although the difference is smaller
than observed for the specimens with two hoops. TN for the
90-degree hooked bar specimen with five parallel hoops
was, respectively, 1.21 and 1.16 times TN for the 90-degree
hooked bar specimens with four and five perpendicular
hoops. For the 180-degree hooked bar specimens, the spec-
imen with four perpendicular hoops had the highest strength,
while the specimen with five perpendicular hoops had the
lowest. TN for the 180-degree hooked bar specimen with five
parallel hoops was, respectively, 0.97 and 1.05 times the
failure loads of the companion specimens with four and five
perpendicular hoops. TN for the 180-degree hooked bar spec-
imen with five parallel hoops was equal to 0.95 of TN for the
90-degree hooked bar specimen with parallel hoops and 1.15
and 1.11 times TN for the 90-degree hooked bar specimens
254
Fig. 12—Ratio of anchorage strength of No. 8 (No. 25)
hooked bars with confining reinforcement oriented parallel
to straight portion of bar to anchorage strength of No. 8
(No. 25) hooked bars with confining reinforcement oriented
perpendicular to straight portion of bar. Strengths normal-
ized to embedment length of 10 in. (254 mm) and concrete
compressive strength of 12,000 psi (83 MPa).
with, respectively, four and five perpendicular hoops. In this
limited test series, increasing from four to five perpendicular
hoops had little effect on the strength of hooked bars with
bend angles of either 90 or 180 degrees.
Figure 12 shows the ratio of the anchorage strength of the
hooked bars confined by parallel hoops to the anchorage
strength of hooked bars confined by perpendicular hoops.
The figure indicates that for 90-degree hooked bars, all of
which had ratios greater than unity, parallel hoops provided
a greater increase in anchorage strength than perpendicular
hoops. In contrast, for 180-degree hooked bars, all had ratios
below unity, indicating that perpendicular hoops provided
the greater increase in strength. The greater contribution
of parallel hoops for 90-degree hooked bars may result
because the hoops serve as anchor reinforcement that resists
a breakout failure. In this configuration, hoop orientation is
optimal for carrying a direct tensile force, which helps keep
the joint region adjacent to the 90-degree hooked bar intact.
In contrast, perpendicular hoops do not develop a direct
tensile force to counteract the tensile force in the hooked
bars, causing the hoops to be less efficient than hoops with
a parallel orientation in preventing a breakout failure of the
joint region. Confining reinforcement oriented perpendic-
ular to the straight portion of the hooked bar, however, may
be more efficient in limiting splitting of the concrete caused
by slip of the hooked bars—splitting that may be greater for
180-degree hooked bars than for 90-degree hooked bars.
Greater slip was observed for 180-degree hooked bars by
Marques and Jirsa (1975) and Hamad et al. (1993). Splitting
due to slip is also key in developing straight bar reinforce-
ment, where the resistance to the wedging action of the bar
due to slip is a function of the amount of confining rein-
forcement oriented perpendicular to the bar and the concrete
compressive strength. This suggests that the confinement
ACI Structural Journal/January 2018
provided by perpendicular hoops may be similar to that of
the confinement provided by reinforcement perpendicular
to straight bars (Darwin et al. 1995, 1996; Zuo and Darwin
1998, 2000, Joint ACI-ASCE Committee 408 2003).
Figure 13 shows the ratio of the anchorage strength of
90-degree hooked bars to that of 180-degree hooked bars for
specimens with both hoop orientations. The ratio for speci-
mens with parallel hoops ranged from 1.01 to 1.06, while the
ratio for specimens with perpendicular hoops ranged from
0.80 to 0.89. For specimens with parallel hoops, anchorage
strength ratios were very close to 1.0, which indicates that
regardless of the number of hoops in the specimens, placing
the confining reinforcement parallel to the straight portion
of the hooked bar resulted in a similar increase in strength
for hooked bars with 90- and 180-degree bend angles. For
Fig. 13—Ratio of anchorage strength of No. 8 (No. 25) hooked
bars with 90-degree bend angle to anchorage strength of
No. 8 (No. 25) hooked bars with 180-degree bend angle for
hooks confined by No. 3 (No. 10) hoops oriented parallel or
perpendicular to straight portion of bar. Strengths normal-
ized to embedment length of 10 in. (254 mm) and concrete
compressive strength of 12,000 psi (83 MPa).
Fig. 14—Failure modes of specimens with 90-degree hooked bars with different orientations of confining reinforcement: (a)
hoops parallel to straight portion of hooked bars pinning triangular-shaped block to compression zone; and (b) hoops perpen-
dicular to straight portion of hooked bar with concrete block being pulled toward front of column.
ACI Structural Journal/January 2018
specimens with perpendicular hoops, the average anchorage
strength ratio was approximately 0.85, which shows that
placing confining reinforcement perpendicular to the
straight portion of the hooked bar resulted in lower increases
in anchorage strength for 90-degree hooked bars than for
180-degree hooked bars.
The observed failure modes (Fig. 14(a) and (b)) support
the theory that confining reinforcement oriented parallel
to the straight portion of a hooked bar is more effective in
keeping the concrete in the joint region together by acting as
anchor reinforcement, limiting the potential for a concrete
breakout failure at the front of the column. As observed by
Searle et al. (2014) and Sperry et al. (2015a,b; 2017a), the
tensile force in the hooked bars pulls out a triangular-shaped
block of concrete (with its base at the front of the column),
as shown in Fig. 14(a), with the tensile force in the hoops
directly opposing that breakout force and pinning the trian-
gular-shaped block to the compression zone at the back of
the column. When reinforcement oriented perpendicular to
the straight portion of the bar is used to confine hooked bars
with 90-degree bend angles, it does help to keep the concrete
in the joint together, although it is less effective because it
no longer acts as anchor reinforcement. Failure occurs as a
block of confined concrete is pulled toward the front of the
column with the hoops, as shown in Fig. 14(b).
Based on these results, it is concluded that the anchorage
strength of 180-degree hooked bars with either orientation
of confining reinforcement is similar to that of 90-degree
hooked bars with confining reinforcement oriented parallel to
the straight portion of the hooked bar. Confining reinforce-
ment oriented perpendicular to the straight portion of the bar
is less effective for 90-degree hooked bars. Considering the
limited amount of test data in this study to address the effect
on anchorage strength of confining reinforcement perpen-
dicular to the straight portion of a hooked bar, more research
on the effect of confining reinforcement with this orientation
is needed.
255
 SUMMARY AND CONCLUSIONS
Test results for 166 simulated exterior beam-column joints
with two hooked bars were used to investigate the effects
of bend angle, concrete side cover, and confining reinforce-
ment orientation on the anchorage strength of hooked bars.
The simulated beam-column joint specimens contained
No. 5, No. 8, and No. 11 (No. 16, No. 25, and No. 36)
hooked bars with 90- or 180-degree bend angles. The clear
concrete side cover ranged from 2.5 to 3.5 in. (65 to 90 mm).
The specimens were cast using normalweight concrete with
compressive strengths ranging from 4300 to 16,500 psi (30
to 114 MPa). Bar stresses at failure ranged from 33,000
to 137,400 psi (228 to 947 MPa). A set of specimens was
cast with confining reinforcement oriented either parallel or
perpendicular to the straight portion of the hooked bar in
the joint region, with all other parameters held constant to
study the effect of confining reinforcement orientation on
anchorage strength. Data from prior studies were included in
the analysis where applicable.
The following conclusions are based on the data and
analysis presented herein:
1. Hooked bars with 90- and 180-degree bend angles have
similar anchorage strengths and can be used interchange-
ably. This includes hooked bars with a 180-degree bend
angle confined by parallel reinforcement following Section
25.4.3.2 of ACI 318-14 to allow use of the 0.8 modification
factor for calculating the development length of hooked bars.
2. For hooked bars with a 90-degree bend angle, confining
reinforcement placed perpendicular to the straight portion of
the bars results in a lower anchorage strength than confining
reinforcement with a similar spacing placed parallel to the
straight portion of the bars.
3. Increasing concrete side cover from 2.5 to 3.5 in. (65
to 90 mm) does not increase the anchorage strength of
hooked bars.
AUTHOR BIOS
ACI member Jayne Sperry is a Graduate Engineer with Walter P Moore
in Orlando, FL. She received her BS, MS, and PhD from the University of
Kansas, Lawrence, KS.
ACI Honorary Member David Darwin is the Deane E. Ackers Distin-
guished Professor and Chair of the Department of Civil, Environmental,
and Architectural Engineering at the University of Kansas.
ACI member Matthew O’Reilly is an Assistant Professor of civil, environ-
mental, and architectural engineering at the University of Kansas.
Andrés Lepage, FACI, is an Associate Professor and Director of Labora-
tories in the Department of Civil, Environmental, and Architectural Engi-
neering at the University of Kansas.
ACI member Rémy D. Lequesne is an Assistant Professor of civil, environ-
mental, and architectural engineering at the University of Kansas.
Adolfo Matamoros, FACI, is the Peter T. Flawn Distinguished Professor
of Civil and Environmental Engineering at the University of Texas at San
Antonio, San Antonio, TX.
Lisa R. Feldman, FACI, is an Associate Professor of civil and geological
engineering at the University of Saskatchewan, Saskatoon, SK, Canada,
and Director of the Saskatchewan Centre for Masonry Design.
ACI member Samir Yasso is a Faculty Member at the University of Mosul,
Mosul, Iraq. He received his BSc and MSc from the University of Mosul and
his PhD from the University of Kansas.
256
ACI member Nathaniel Searle is a Structural Design Engineer with S. A.
Miro, Inc. in Denver, CO. He received his BS and MS degrees from the
University of Kansas.
ACI member Michael DeRubeis is a Structural Design Engineer with S. A.
Miro, Inc., in Denver, CO. He received his BS from the University of Mich-
igan, Ann Arbor, MI, and his MS from the University of Kansas.
ACI member Ali Ajaam is a Faculty Member at the University of Babylon,
Babylon, Iraq. He received his BSc and MSc from the University of Babylon
and his PhD from the University of Kansas.
ACKNOWLEDGMENTS
Support for the study was provided by the Electric Power Research Insti-
tute, Concrete Reinforcing Steel Institute Education and Research Foun-
dation, University of Kansas Transportation Research Institute, Charles
Pankow Foundation, Commercial Metals Company, Gerdau Corporation,
Nucor Corporation, and MMFX Technologies Corporation. Additional
materials were supplied by Dayton Superior, Midwest Concrete Materials,
and W. R. Grace Construction. Thanks are due to K. Barry and M. Ruis,
who provided project oversight for the Advanced Nuclear Technology
Program of the Electric Power Research Institute, and to N. Anderson,
C. Kopczynski, M. Mota, J. Munshi, and C. Paulson, who served as
industry advisors.
REFERENCES
AASHTO, 2012, AASHTO LRFD Bridge Design Specifications, sixth
edition, American Association of State Highway and Transportation Offi-
cials, Washington, DC, 1672 pp.
ACI Committee 318, 2014, “Building Code Requirements for Struc-
tural Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 519 pp.
ACI Committee 349, 2006, “Code Requirements for Nuclear Safety
Related Concrete Structures (ACI 349-06),” American Concrete Institute,
Farmington Hills, MI, 157 pp.
ASTM A1035/A1035M, 2014, “Standard Specification for Deformed
and Plain Low-Carbon, Chromium, Steel Bars for Concrete Reinforce-
ment,” ASTM International, West Conshohocken, PA, 7 pp.
ASTM A615/A615M, 2015, “Standard Specification for Deformed and
Plain Carbon-Steel Bars for Concrete Reinforcement,” ASTM Interna-
tional, West Conshohocken, PA, 8 pp.
Darwin, D.; Tholen, M. L.; Idun, E. K.; and Zuo, J., 1995, “Splice
Strength of High Relative Rib Area Reinforcing Bars,” SL Report 95-3,
University of Kansas Center for Research, Lawrence, KS, May, 58 pp.
Darwin, D.; Zuo, J.; Tholen, M. L.; and Idun, E. K., 1996, “Development
Length Criteria for Conventional and High Relative Rib Area Reinforcing
Bars,” ACI Structural Journal, V. 93, No. 3, May-June, pp. 347-359.
Draper, N. R., and Smith, H., 1981, Applied Regression Analysis, second
edition, Wiley, New York, 709 pp.
Hamad, B. S.; Jirsa, J. O.; and D’Abreu de Paulo, N. I., 1993, “Effect of
Epoxy Coating on Bond Anchorage of Reinforcing in Concrete Structures,”
ACI Structural Journal, V. 90, No. 1, Jan.-Feb., pp. 77-88.
Joint ACI-ASCE Committee 408, 2003, “Bond and Development of
Straight Reinforcing Bars in Tension (ACI 408R-03),” American Concrete
Institute, Farmington Hills, MI, 8 pp.
Lee, J., and Park, H., 2010, “Bending—Applicability Study of Ultra-Bar
(SD 600) and Ultra-Bar for Rebar Stirrups and Ties (SD 500 and 600) for
Compression Rebar,” Korea Concrete Institute, Aug., 504 pp. (translated
from Korean)
Marques, J. L., and Jirsa, J. O., 1975, “A Study of Hooked Bar Anchor-
ages in Beam-Column Joints,” ACI Journal Proceedings, V. 72, No. 5, May,
pp. 198-209.
Peckover, J., and Darwin, D., 2013, “Anchorage of High-Strength Rein-
forcing Bars with Standard Hooks: Initial Tests” SL Report 13-1, University
of Kansas Center for Research, Lawrence, KS, 47 pp.
Pinc, R.; Watkins, M.; and Jirsa, J. O., 1977, “The Strength of the Hooked
Bar Anchorages in Beam-Column Joints,” CESRL Report No. 77-3, Depart-
ment of Civil Engineering-Structures Research Laboratory, University of
Texas at Austin, Austin, TX, 67 pp.
Ramirez, J. A., and Russell, B. W., 2008, Transfer, Development, and
Splice Length for Strand/reinforcement in High-strength Concrete, Trans-
portation Research Board, National Research Council, Washington, DC,
pp. 99-120.
Searle, N.; DeRubeis, M.; Darwin, D.; Matamoros, A.; O’Reilly, M.; and
Feldman, L., 2014, “Anchorage of High-Strength Reinforcing Bars with
Standard Hooks - Initial Tests,” SM Report No. 108, University of Kansas
Center for Research, Inc., Lawrence, KS, Feb., 120 pp.
ACI Structural Journal/January 2018
 Shao, Y.; Darwin, D.; O’Reilly, M.; Lequesne, R. D.; Ghimire, K.; and
Hano, M., 2016, “Anchorage of Conventional and High-Strength Headed
Reinforcing Bars,” SM Report No. 117, University of Kansas Center for
Research, Lawrence, KS, Aug., 234 pp.
Sperry, J.; Al-Yasso, S.; Searle, N.; DeRubeis, M.; Darwin, D.; O’Reilly, M.;
Matamoros, A.; Feldman, L.; Lepage, A.; Lequesne, R.; and Ajaam, A., 2015a,
“Anchorage of High-Strength Reinforcing Bars with Standard Hooks,” SM
Report No. 111, University of Kansas Center for Research, Inc., Lawrence,
KS, June 2015, 266 pp.
Sperry, J.; Darwin, D.; O’Reilly, M.; and Lequesne, R., 2015b,
“Anchorage Strength of Conventional and High-Strength Hooked Bars in
Concrete,” SM Report No. 115, University of Kansas Center for Research,
Lawrence, KS, Dec., 266 pp.
Sperry, J.; Darwin, D.; O’Reilly, M.; Matamoros, A.; Feldman, L.;
Lepage, A.; Lequesne, R.; and Yasso, S., 2017a, “Conventional and High-
ACI Structural Journal/January 2018
Strength Hooked Bars—Part 2: Data Analysis,” ACI Structural Journal,
V. 114, No. 1, Jan.-Feb., pp. 267-276. doi: 10.14359/51689457
Sperry, J.; Yasso, S.; Searle, N.; DeRubeis, M.; Darwin, D.; O’Reilly,
M.; Matamoros, A.; Feldman, L.; Lepage, A.; and Lequesne, R., 2017b,
“Conventional and High-Strength Hooked Bars—Part 1: Anchorage
Tests,” ACI Structural Journal, V. 114, No. 1, Jan.-Feb., pp. 255-266. doi:
10.14359/51689457
Wonnacott, R., and Wonnacott, T., 1985, Introductory Statistics, Wiley,
New York, 649 pp.
Zuo, J., and Darwin, D., 1998, “Bond Strength of High Relative Rib
Area Reinforcing Bars,” SM Report No. 46, University of Kansas Center
for Research, Lawrence, KS, Jan., 350 pp.
Zuo, J., and Darwin, D., 2000, “Splice Strength of Conventional and
High Relative Rib Area Bars in Normal and High-Strength Concrete,” ACI
Structural Journal, V. 97, No. 4, July-Aug., pp. 630-641.
257
 Regular Price: $69.50
Member Price: $39.00

Only available in PDF version

44
• Design guidelines and specifications;
• Material properties for design;
• Behavior and design of beams and
columns;
• Behavior and design of slabs and
other structures;
• Behavior and design of foundations and
papers are organized underground components; and finally,

6
• Applications in structure and under-
into themes: ground construction projects.

Held at Polytechnique Montreal, Canada, on July 24 and 25, 2014, the workshop
demonstrated the state-of-the-art progress attained in terms of specifications
and applications of fiber-reinforced concrete.

| +1.248.848.3700 | www.concrete.org |
 ACI STRUCTURAL JOURNAL
Title No. 115-S21
Thermomechanical Relaxation of Reinforced Concrete
Beams Strengthened with Carbon Fiber-Reinforced
Polymer
by Yail J. Kim and Abdulaziz Alqurashi
This paper presents the behavior of reinforced concrete beams
strengthened with carbon fiber reinforced polymer (CFRP) sheets
subjected to relaxation induced by thermal and mechanical load-
ings that are applied simultaneously (thermomechanical loading).
The range of elevated temperatures varies from 75 to 150°C (167 to
302°F) and the mechanical load represents a typical service condi-
tion (60% of the strengthened beam’s capacity at 25°C [77°F]),
both of which are sustained for 30- to 75-minute periods. The heat
conduction of the beams is measured by an infrared imaging tech-
nique to understand their thermal responses. Upon completion
of the thermomechanical relaxation test, all beams are loaded to
failure for residual capacity investigations. The progression of
thermal loading across the CFRP-concrete interface exhibits a
so-called latent heat effect that takes place with thermal conduction
along the interface. During residual testing, the degraded interface
reveals partial debonding that accompanies the retraction of CFRP
strains. Thermomechanical relaxation influences the residual
capacity, load displacement, and interfacial cracking patterns of
the strengthened beams. According to analytical modeling, the
elastic modulus of CFRP and the magnitude of a mechanical load
affect the degree of thermomechanical relaxation. A design factor is
proposed based on multiple regression to quantify the implications
of thermomechanical relaxation for CFRP-strengthened beams.
Keywords: carbon fiber-reinforced polymer (CFRP); relaxation; retrofit;
strengthening; thermal.
INTRODUCTION
For the last two decades, carbon fiber-reinforced polymer
(CFRP) sheets have been employed to strengthen existing
structural members. Although susceptible to elevated
temperatures, these polymer-based materials are usable
for buildings that are potentially vulnerable to fire hazards
because insulation layers can protect the strengthened
members and retard heat propagation into the CFRP system
for a certain period of time (fire rating). Williams et al.1
conducted a fire test with reinforced concrete T-beams
strengthened with CFRP sheets in a furnace. Insulation
materials (thickness = 25 to 38 mm [1 to 1.5 in.]) were
sprayed to protect the beams prior to testing. The maximum
furnace temperature was approximately 900°C (1652°F)
at 2 hours and the corresponding temperature of the
CFRP-concrete interface inside the insulation was less than
200°C (392°F). Kodur and Yu2 proposed a design approach
for CFRP-strengthened members subjected to a fire. A rela-
tionship between the members’ capacity and fire exposure
time was elaborated. Recommendations included that the
approach used in current practice is conservative and equiv-
ACI Structural Journal/January 2018
TECHNICAL PAPER
alent sections can be used to represent the effect of insulation
layers. The CFRP-concrete interface is degraded by thermal
loading primarily due to a change in the tangled cross-
linking structure of polymeric adhesives.3 The interfacial
behavior is thus concurrently controlled by temperature and
exposure time. Gamage et al.4 examined the characteristics
of CFRP sheets bonded to a concrete substrate at elevated
temperatures. The interface was insulated and subjected
to temperatures from 600 to 1200°C (1112 to 2192°F) for
up to 3 hours in an oven. The interfacial capacity was not
controlled by CFRP-bond length, but was affected by the
thermal properties of insulation. Leone et al.5 evaluated
the bond of CFRP-concrete interface at elevated tempera-
tures varying from 20 to 80°C (68 to 176°F). The interfacial
capacity was remarkably reduced when the applied tempera-
ture exceeded the adhesive’s glass transition temperature of
55°C (131°F), which was associated with a state change from
solid to amorphous. The failure mode of the interface was
also influenced by temperatures: concrete cohesion failure at
50°C (92°F) and adhesive adhesion failure at 80°C (176°F).
It was recommended that thermal loading be considered in
CFRP-strengthening design.
Structural materials experience stress relaxation in
prescribed boundary conditions. Upon the onset of relax-
ation, the materials’ performance may be different from
what they were intended. When a building is subjected to
a fire, reinforced concrete beams are in a stress relaxation
state caused by the out-of-plane displacement of adjacent
columns in tandem with temperature-induced loading (that
is, mechanical and thermal loads are applied simultane-
ously). Thermomechanical relaxation is, therefore, of interest
from structural integrity perspectives, including the case of
CFRP-strengthened beams. Unlike conventional relaxation,
thermomechanical relaxation is not necessarily engaged
with stabilized stresses because the functionality of a relaxed
material in thermal distress, such as with fire, is terminated
within a relatively short period of time. As reviewed earlier,
the temperature-dependent behavior of CFRP-strength-
ened concrete members alongside CFRP-concrete interface
has been studied; however, their thermomechanical relax-
ation responses have not been reported. Consequently, it is
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2017-088, doi: 10.14359/51701091, was received March 11, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
259
unknown how CFRP systems perform when a strengthened
member is in thermomechanical relaxation and how the
degraded systems alter the efficacy of structural strength-
ening. This paper experimentally investigates the mecha-
nism of thermomechanical relaxation for CFRP-strength-
ened beams and analytically characterizes their behavior in
various conditions.

RESEARCH SIGNIFICANCE
Stress relaxation is considered important for structural
members consisting of multiple built-up components such
as externally bonded CFRP sheets that require appropriate
stress transfer to increase the load-carrying capacity of a
strengthened beam. The performance of a viscoelastic poly-
meric adhesive linking the concrete substrate and CFRP is
susceptibly influenced by thermal distress, and synergetic
degradation is expected when mechanical load is sustained.
This crucial load-carrying mechanism for CFRP-strength-
ened beams has been undervalued and is not considered
in design because no guidelines have been available. The
present research aims to understand the time- and tempera-
ture-dependent relaxation of reinforced concrete beams
strengthened with CFRP sheets, which results in analytical
expressions that can be used in practice.

EXPERIMENTAL INVESTIGATION
A test program is conducted to examine the thermome-
chanical relaxation and residual behavior of CFRP-strength-
ened beams in flexure. Summarized in the following sections
are the materials, specimen preparation, loading schemes,
and instrumentation applied.
Materials
Concrete was mixed for a compressive strength of fc′ =
20 MPa (2900 psi), close to the average 28-day cylinder
strength of 19.4 MPa (2813 psi). As primary reinforcement,
Grade 60 No. 3 steel bars were employed (yield strength σy
= 414 MPa [60 ksi] and cross-sectional area As = 71 mm2
[0.11 in.2]). Shear stirrups were A36 plain steel rods (6 mm
[0.24 in.] in diameter with a yield strength of 250 MPa
[36 ksi]). The unidirectional CFRP sheets used require
a wet-lay-up process (that is, dry fibers are impregnated
in a resin matrix), and have a nominal tensile strength of
σfu = 3800 MPa (551 ksi) and a modulus of Ef = 227 GPa
(32,900 ksi) based on an equivalent thickness of tf =
0.165 mm (0.0065 in.). The resin and hardener components
of a two-part structural epoxy were blended at a mass ratio of
3:1 until a homogeneous mixture was attained. The specified
yield strength and modulus of the epoxy are σey = 54 MPa
(7830 psi) and Ee = 3 GPa (440 ksi), respectively, including
a glass transition temperature of Tg = 71°C (160°F) with a
thermal conductivity of ke = 0.21W/(m·K) (1.45 Btu·in./
(h·ft2·°F)).
Beam details
Singly reinforced concrete beams were cast with dimen-
sions of 100 mm (4 in.) wide by 165 mm deep (6.5 in.) by
1200 mm (48 in.) long, as illustrated in Fig. 1. After 28 days
of curing in an environmental chamber (relative humidity =
Fig. 1—Beam details: (a) dimension; and (b) cross section.
(Note: Unit in mm; 1 mm = 0.0394 in.)
95% at 20°C [68°F]), the beams were taken out for drying
and surface preparation using a wire brush to improve the
bond between the CFRP and substrate. The blended epoxy
was applied along the tensile soffit of the beam, followed
by the impregnation of single-layer dry carbon fiber
fabric (100 mm [4 in.] wide by 900 mm [35 in.] long). In
compliance with the manufacturer’s recommendation, the
CFRP-strengthening system was cured at room temperature
for at least 7 days before testing. To preclude the premature
end-peeling failure of the longitudinal sheet, CFRP U-wraps
(200 mm [8 in.] by 400 mm [16 in.]) were additionally
bonded (Fig. 1).
Test procedure
The strengthened beams were simply supported with
a span length of 1000 mm (3.3 ft). To generate thermal
loading, a silicon rubber heat pad (150 mm [6 in.] long
by 100 mm [4 in.] wide) was attached to the tensile soffit
of the beam at midspan (Fig. 1) by tightening steel wires,
which is the most vulnerable location in forming a plastic
hinge when mechanically loaded. The pad is composed of
fiberglass reinforcement and perfluoroalkoxy lead wires for
heat generation. Five steady-state temperature categories
were tested from 75 to 150°C (167 to 302°F) an interval
of 25°C (45°F). Because the thermal behavior of the adhe-
sive at temperatures below its glass transition temperature
(Tg = 71°C [160°F]) is invariant, temperatures lower than
75°C (167°F) were not taken into consideration except for a
control temperature of 25°C (77°F). To identify the flexural
capacity of the strengthened section (Puc) under three-point

260 ACI Structural Journal/January 2018

Fig. 2—Test setup: (a) instrumentation; and (b) thermocouple reading. (Note: °F = °C(9/5) + 32.)

Table 1—Test specimens relaxation periods, all beams were loaded to failure at a rate
of 2 mm/min (0.08 in./min) while maintaining the sustained
Load, kN (kip)
Thermomechanical heat for residual strength investigations.
Beam ID relaxation time, min. Yielding Ultimate
B25-0 0 42.7 (9.6) 52.4 (11.8) Instrumentation
B75-30 30 38.9 (8.7) 50.5 (11.3) To measure the applied load and displacement of the
respective beams, a load cell and a linear potentiometer were
B100-30 30 38.9 (8.7) 51.8 (11.6)
placed at midspan, as shown in Fig. 2(a). Displacement-type
B125-30 30 37.8 (8.4) 49.7 (11.1) strain transducers (PI gauges) were also mounted onto one
B150-30 30 37.3 (8.3) 47.6 (10.7) side of each beam to monitor tension and compression
B150-45 45 36.6 (8.2) 47.5 (10.6) strains. The progression of heat toward the beam concrete
was quantified by an infrared camera (Fig. 2(a)). The camera,
B150-60 60 35.8 (8.0) 47.3 (10.6)
built upon multi-spectral dynamic imaging (MSX) technol-
B150-75 75 35.2 (7.9) 43.9 (9.8) ogies, has a 19,200 pixel resolution with a thermal sensi-
tivity of less than 0.06°C (0.11°F) and a frame rate of 9 Hz.
bending (Fig. 2(a)), one beam was monotonically loaded Thermocouple wires (accuracy = ±2.2°C [±3.9°F]) were
to failure at 25°C (77°F). Other beams were preheated at positioned in between the CFRP and the heat pad to measure
midspan (Fig. 2(b)) to achieve the predefined temperatures, the applied temperatures (Fig. 2(b)). The thermomechanical
and were subjected to a load of 60%Puc for a default period response of the CFRP-concrete interface was studied by six
of 30 minutes during which the stroke of the universal strain gauges bonded outside the heating region (Fig. 1).
testing machine was fixed for a relaxation test. To examine
the effects of thermomechanical relaxation time, additional TEST RESULTS
loading periods up to 75 minutes were adopted to the beams The results of a two-phase experimental investigation
exposed to 150°C (302°F). These test schemes were summa- (thermomechanical relaxation and residual behavior) are
rized in Table 1 with the beams’ identification. For example, discussed with a focus on heat propagation, interfacial char-
B125-30 indicates a beam loaded at a temperature of 125°C acteristics, flexural capacity, and failure modes.
(257°F) for 30 minutes. The preceding thermomechanical
relaxation periods spanned over 40% of a 3-hour fire rating, Thermomechanical loading
which are sufficient from practice standpoints because Figure 3(a) exhibits the time-history plot of the thermo-
the restrained boundary condition of a flexural member is mechanical and residual tests for the beams subjected to
typically sustained for approximately 30 to 60 minutes in 150°C (302°F). The load was precipitously increased up to
a building under the occurrence of a fire.6,7 It should be approximately 30 kN (6.7 kip), at which the machine stroke
noted that the temperatures delineated herein are intended to was held for examining the beams’ relaxation behavior. The
represent those inside insulation layers enclosing structural applied loads gradually decayed until a residual capacity
beams when a fire takes place (that is, the relevant tempera- test was conducted. The thermal loading associated with
ture is the one at the interface level, instead of temperatures this mechanical testing scheme was reasonably maintained,
outside the structural member). Because the objective of as shown in the thermocouple measurement (Fig. 2(b)).
the present experimental program was to evaluate the influ- During the relaxation period, the mechanical loading rate
ence of thermomechanical relaxation on the behavior of did not markedly vary (Fig. 3(b)), including an alteration
CFRP-strengthened beams, rather than to study an increase range of less than 0.2 kN/s (0.05 kip/s). This indicates that
in load-carrying capacity after strengthening, unstrength- the sustained load relaxed in a consistent manner, which
ened beams were not tested. Upon completion of the planned confirms the adequacy of the present test setup.

ACI Structural Journal/January 2018 261

Fig. 3—Load-time history of beams: (a) response at 150°C (302°F); and (b) thermomechanical relaxation loading at 150°C
(302°F) for 45 minutes. (Note: 1 kN = 0.225 kip.)
Fig. 4—Heat propagation based on infrared thermal imaging: (a) Location 1; (b) Location 2; and (c) Location 3. (Note: °F
= °C(9/5) + 32.)
Heat progression
The interface-level heat that propagated up the concrete is
provided in Fig. 4, where the measured temperatures (contours
in the infrared images) were compared under thermal loading
for 30 minutes. At Location 1 (Fig. 4(a)), the heat progres-
sion rate of the beam subjected to 150°C (302°F) was more
rapid than that of the other beams. The progression rates of
the beams at Location 2 were virtually identical up to approx-
imately 20 minutes, regardless of interfacial temperature, after
which bifurcations were noticed. This observation implies
that a so-called latent heat effect exists in the concrete above
the CFRP bond-line. In other words, the applied heat energy
at the CFRP-concrete interface level does not immediately
change the temperature of nearby concrete until the energy
is sufficiently accumulated, which is a function of distance
from the interface. The far-field temperature response at Loca-
tion 3 was marginal; nevertheless, the interfacial heat kept
conducting with time, as shown in Fig. 4(c).
Interfacial response during relaxation
Figure 5 shows the time-dependent strain of the CFRP
sheets. When the relaxation loading began, CFRP strains
close to the heat pad were generally higher than the strains
of other locations. This is attributable to a curvature gradient
along the loading span (that is, the curvature at midspan of
the beam developed more rapidly than the curvature in the
shear spans). With an increase in relaxation time, the strains
consistently dwindled. The degree of strain drop appeared
to be related to elevated temperatures, particularly the ones
262
near the heat pad. For instance, reductions of 9.7%, 23.4%,
40.3%, and 32.9% were recorded at the location of Gauge 4
during the 30-minute relaxation period for the beams
exposed to 75, 100, 125, and 150°C (167, 212, 257, and
302°F), respectively. Furthermore, heat progression along
the horizontal bond line was responsible for weakening the
CFRP-concrete interface, which will be elaborated in the
subsequent section.
Residual behavior
The full loading history of the thermomechanically loaded
beams at midspan is provided in Fig. 6, including residual
load-displacement responses (the B25-0 beam’s behavior was
similar to others; hence, it was not plotted for clarity). The
U-wrap anchors prevented end peeling of the longitudinal
CFRP, which led to the ductile behavior of the strengthened
beams. It should be noted that the strains provided by the PI
gauges were not useable on most occasions because the adhe-
sive employed for bonding the gauges to the concrete surface
was affected by the heat propagation. For the cases subjected
to the relaxation period of 30 minutes (Fig. 6(a)), there was no
marked difference in flexural response up to approximately
30 kN (6.7 kip). Thereafter, some divergence was noticed
owing to the adjustment of displacement along with relaxation
testing (further discussions will follow). Despite the likeness
of load-displacement curves, the beams’ yield and ultimate
loads were influenced by elevated temperatures, as enumer-
ated in Table 1. With an increase in temperature from 75 to
150°C (167 to 302°F), the yield and ultimate loads decreased
ACI Structural Journal/January 2018
Fig. 5—CFRP strain during thermomechanical relaxation: (a) B75-30; (b) B100-30; (c) B125-30; and (d) 150-30. (Note: °F
= °C(9/5) + 32.)
Fig. 6—Load-displacement at midspan: (a) elevated temperatures; and (b) variable thermomechanical relaxation periods.
(Note: 1 kN = 0.225 kip; 1 mm = 0.0394 in.)
by 4.1% and 5.7%, respectively. These were attributable to the
deteriorated CFRP-concrete interface, which lowered the effi-
cacy of strengthening (that is to say, the extent of composite
action between the CFRP and concrete beam was dimin-
ished). Figure 6(b) exhibits the flexural behavior of the beams
exposed to 150°C (302°F) for variable relaxation periods.
As before, the load-displacement responses were generally
similar; however, their load-carrying capacities were a func-
tion of thermal exposure time: decreases of 5.6% and 7.8%
of the yield and ultimate loads of the beams were recorded,
respectively, when the thermomechanical relaxation period
changed from 30 to 75 minutes (Table 1). A summary of the
temperature effects on altering the beams’ residual capacity is
shown in Fig. 7. Except for the minor discrepancy in B75-30
and B100-30 that failed at 50.5 kN (11.3 kip) and 51.8 kN
(11.6 kip), respectively, the decrease ratio of the ultimate load
almost linearly varied with elevated temperatures. In terms of
ACI Structural Journal/January 2018
relaxation time at 150°C (302°F), the decrease ratio did not
change considerably from 30 to 60 minutes, whereas a rapid
reduction in the ultimate load was observed during the expo-
sure period of 75 minutes. It is thus stated that thermomechan-
ical relaxation is crucial for CFRP-strengthened beams, and
should be taken into consideration in design, wherever neces-
sary, with an emphasis on the degradation of the CFRP-con-
crete interface that controls the beams’ performance.
Figure 8(a) demonstrates CFRP-strain profiles for the
B25-0 beam at an interval of 25% of the ultimate load (Pu).
Following the increasing flexural load, the profiles gradu-
ally developed. At failure of the beam by concrete crushing
(100%Pu), erratic strain gradients were observed, which
denote that the longitudinal CFRP partially debonded. The
U-wrap anchors preserved the interfacial integrity; conse-
quently, the strains at Gauges 1 and 6 in the vicinity of the
anchors (G1 and G6, respectively) were lower than others.
263
The propensity for strain development, however, changed
when the beams were exposed to thermomechanical loading,
as shown in Fig. 8(b). The B150-75 beam’s strain profiles
were similar to those of B25-0 until a load level of 50%Pu
was accomplished (the flexural load promptly increased in
the nascent loading stage [Fig. 3(a)]; therefore, the effect
of elevated temperatures was not yet apparent). Passing
through the relaxation period, the strain profile became
asymmetric (75%Pu) and, eventually, the beam failed with a
rapid increase in strain on the G1 gauge side (100%Pu). This
fact signifies that the CFRP-concrete interface weakened
during thermomechanical relaxation, and was disintegrated
by the residual loading with interfacial slippage between the
CFRP and concrete. It is worthwhile to note that debonding
strains recorded in these beams were lower than the effec-
tive strain determined by ACI 440.2R-088: εfd = 0.0095.
Fig. 7—Variation of residual capacity. (Note: °F = °C(9/5)
+ 32.)
Fig. 8—Load-CFRP strain up to failure: (a) profile of B25-0; (b) profile of B150-75; (c) development of B75-30; and (d) relax-
ation effect. (Note: 1 kN = 0.225 kip; 1 mm = 0.0394 in.)
264
Possible reasons for that are 1) the empirical ACI equation
was calibrated without due consideration about the use of
U-wrap anchors; and 2) the present CFRP properties were
outside the calibration range. In either case, an improvement
in the ACI expression appears to be necessary. The CFRP-
strain variation caused by the thermomechanical relaxation
is given in Fig. 8(c) and (d). The B75-30 beam revealed a
strain retraction from a load of 33.3 kN (7.5 kip) down to
26.8 kN (6.0 kip), accompanying a 24.7% strain drop in
strain Gauge 6 (G6 in Fig. 8(c)). The retraction of the CFRP
strains was pronounced with the increased temperature and
relaxation time, as shown in Fig. 8(d), which impacted the
integrity of the CFRP-concrete interface and, thereby, the
effectiveness of CFRP-strengthening.
Failure mode
The representative failure modes of the test beams are
provided in Fig. 9. The interfacial integrity of B25-0 loaded
without thermomechanical relaxation was preserved until the
concrete crushed (Fig. 9(a)). In conjunction with the insignif-
icant development of diagonal tension cracks, intermediate
crack-induced CFRP-debonding (known as IC-debonding)
was not observed along the loading span. For the beam
subjected to thermomechanical loading (Fig. 9(b)), the
CFRP-concrete interface was deteriorated during the relax-
ation period; subsequently, minor cracks formed upon the
residual loading. The cracks initiated at the junction of the
longitudinal sheet and the U-wrap anchor were connected to
the primary diagonal tension crack, which progressed toward
the loading point where concrete crushing eventually took
place. The tendency of the interface deterioration was obvi-
ously augmented with the increased temperature and relaxation
ACI Structural Journal/January 2018
Fig. 9—Failure mode: (a) B25-0; (b) B100-30; and (c) B150-75.
period, as shown in Fig. 9(c). In accordance with these failure
modes, it is confirmed that the thermal loading progressed
along the interface (although no physical CFRP-debonding
was associated during the relaxation period) and degraded the
interfacial characteristics that brought about strain localization
adjacent to the U-wrap anchors. The presence of the anchors
precluded end-peeling of the CFRP; accordingly, the load-car-
rying capacity of the beams was maintained with a maximum
variation margin of 7.8%, in spite of the detrimental thermo-
mechanical effects, as discussed earlier.
ANALYTICAL MODELING
An analytical model is developed based on conventional
relaxation theory9 to predict the thermomechanical relaxation
of CFRP-strengthened beams. Upon calibrating the model
using test data, parametric investigations are conducted and
design recommendations are proposed.
Development
The time-dependent strain of the CFRP at temperature T
[εf(t,T)] is expressed as
εf(t,T) = εfe(t,T) + εfc(t,T) (1)
where εfe(t,T) and εfc(t,T) are, respectively, the elastic and
creep components of the CFRP strain (hereafter, t and T indi-
cate time- and temperature-dependent properties, respec-
tively). According to previous research,10 the tempera-
ture-dependent modular ratio [nf(T)] between typical
CFRP and concrete (Ef(T) and Ec(T), respectively) within a
temperature range from 25 to 200°C (77 to 392°F) may be
obtained by (T in Celsius)
E f (T ) E fo exp( −0.0009T )
n f (T ) = = (2)
Ec (T ) Eco ( −0.001282T + 1.025641)
where Efo and Eco are the initial CFRP modulus and concrete
modulus (Eco = 57,000√fc′ in psi [4730√fc′in MPa11]), respec-
tively. In a thermomechanical relaxation condition of the
CFRP-strengthened beam (that is, deformation of the CFRP
remains unchanged), the strain rate becomes zero
d ε f e (t , T ) d ε f c (t , T )
0= + (3)
dt dt
ACI Structural Journal/January 2018
Taking an exponential form for the creep-strain rate of
Eq. (3) (dεfc(t,T)/dt = σf(t,T)m) along with the CFRP’s consti-
tutive relationship [εfe(t,T) = σf(t,T)/ Ef(T)] yields
d σ f (t , T )
= − dt (4)
σ f (t , T ) m
where m is a relaxation constant. Integrating Eq. (4)
(1− m)
σ f (t , T )
= −t + C (5)
1− m
where C is the integration constant to be solved by the CFRP
stress immediately before thermomechanical relaxation
loading (σf(0,T) at t = 0 minutes). Rearranging Eq. (5) gives
σf(t,T) = [(m – 1)t + σf(0,T)(1 –m)]1/(1 –m) (6)
Calibration and assessment
To calibrate the relaxation constant m, a transformed
section of the strengthened beam at midspan may be used
n f (T ) M (t , T ) y f (T )
σ f (t , T ) = (7)
I t (T )
where nf(T) is the CFRP modular ratio for the transformed
section at temperature T; M(t,T) is the time-dependent
moment applied; yf(T) is the distance from the neutral axis
to the CFRP level; and It(T) is the moment of inertia of the
transformed section. It is assumed that, when determining
the moment of inertia It(T), the modulus of steel reinforce-
ment is virtually temperature-independent below 150°C
(302°F).12 Substituting Eq. (6) into Eq. (7) with the boundary
condition employed in the test program (Fig. 1)
P(t,T) = kΔ for 0 ≤ t ≤ tR (8a)
4σ f (t , T ) I t (t , T )
P (t , T ) = for tR < t ≤ tl (8b)
n f (t , T ) y f (t , T ) L
where P(t,T) is the load associated with thermomechanical
relaxation at time t in min. (Fig. 3(a)); k is the stiffness of
the strengthened beam; Δ is the displacement of the beam
265
before the relaxation begins at time tR; and tl is the time when
relaxation ends. The experimental load variations of the
beams subjected to a thermomechanical relaxation period
of 30 minutes (Fig. 3(a)) were substituted into Eq. (6) and
(8b) in tandem with the temperature-dependent variables.
The unknown temperature-dependent constant m was then
determined by a generic algorithm. This computational
technique is useful to solve a complex problem based on a
mutation-and-selection approach with randomly generated
data samples. Further details on the solution algorithm and
its theoretical background are available elsewhere.13Figure
10 summarizes the average values of the constant from the
initiation of the relaxation (tR) to its end (tl) at the respective
temperatures. As the temperature rose, the m values gradually
increased, which resulted in the following best-fit equation (T
in Celsius)
m = 0.1406ln(T) – 0.4695 (9)
With the calibrated constant, Eq. (8a) and (8b) were
appraised with the experimental data, as shown in Fig. 11.
It should be noted that the load at the initiation of relaxation
(Eq. (8a)) was taken from the test data because the beams’
response before and after the relaxation (initial loading and
residual loading, respectively) was outside the interest of the
present study. Figure 11(a) assesses the predicted thermome-
chanical relaxation behavior of the beam at a temperature of
75°C (167°F) against its experimental counterpart. The insig-
Fig. 10—Determination of thermomechanical relaxation
constant. (Note: °F = °C(9/5) + 32.)
Fig. 11—Assessment of predictive model: (a) comparison at 75°C (167°F); and (b) comparison at all temperatures during
thermomechanical relaxation periods. (Note: °F = °C(9/5) + 32; 1 kN = 0.225 kip.)
266
nificant discrepancy between the tested and predicted load
variations (for instance, a margin of 6% at 30 minutes) was
due to the use of the regression-based constant (Eq. (9)). All
tested and predicted thermomechanical relaxation responses
during the 30-minute period (tR < t ≤ tl) at an interval of 1
second were compared in Fig. 11(b), including an average
absolute difference of 4.6%. Overall, the model prediction
was in agreement with the experimentally-attained relax-
ation data, and can be useable for parametric investigations
detailed in the ensuing section.
Parametric study
During a thermomechanical relaxation period (tR < t ≤ tl),
the amount of relaxation in CFRP is determined by
(σ f (0, T ) − σ f (t , T ))
λR = (10)
σ fu
where λR is the thermomechanical relaxation factor. Figure 12
exhibits the variation of the relaxation factor with several param-
eters that typically influence the behavior of CFRP-strengthened
beams. For comparison, unless otherwise stated, the present test
configuration depicted in Fig. 1 alongside the aforementioned
material properties was the default within a temperature range
from 75 to 150°C (167 to 302°F) for a 30-minute relaxation
period. As illustrated in Fig. 12(a), elevated temperatures were
crucial in augmenting the degree of thermomechanical relax-
ation. The relaxation factor associated with 150°C (302°F) was
174% higher than the factor with 75°C (167°F) at 30 minutes
(Fig. 12(a), inset). This is explained by the increased mobility
of polymeric chains in the epoxy resin with temperature.14
Because the relaxation was most noticeable at 150°C (302°F),
the predicted results of the parametric study at this tempera-
ture are only discussed hereafter. The effects of CFRP modulus
are provided in Fig. 12(b). The higher the CFRP modulus, the
more the relaxation amount was resulted. The reason is that
the CFRP stress σf(0,T) of the transformed section increased
with the increased CFRP modulus (Eq. (7)). It is important
to note that this response trend in the strengthened beam does
not necessarily mean CFRP itself is susceptible to relaxation.
The thermomechanical relaxation of the strengthened beam
was marginally influenced by concrete strength (Fig. 12(c)),
whose contribution to the transformed properties of the beam
was negligible. When a CFRP stress level at time tR increased,
ACI Structural Journal/January 2018
Fig. 12—Effects of parameters on thermomechanical relaxation: (a) temperature; (b) CFRP modulus; (c) concrete strength;
and (d) initial stress level (fraction of CFRP strength). (Note: °F = °C(9/5) + 32; 1 kN = 0.225 kip; 1 MPa = 145 psi.)

Table 2—Regression constants

Thermomechanical relaxation time
Constant 10 min 20 min 30 min Application range
c1 0.0007 0.0015 0.0023
c2 –0.0001 –0.0005 –0.0010 75°C ≤ T ≤ 150°C
100 GPa ≤ Ef ≤ 300 GPa
c3 0.0004 0.0007 0.0010
20 MPa ≤ fc′ ≤ 40 MPa
c4 0.0060 0.012 0.0170 0.1ffu ≤ σf(0,T) ≤ 0.5ffu
c5 –0.1500 –0.2900 –0.4320

Notes: °F = °C(9/5) + 32; 1 MPa = 145 psi.

the strengthened beam experienced more relaxation, as shown

in Fig. 12(d). At a stress level of 0.5σfu, the relaxation factor
increased up to 4.6%, which is approximately 50% greater than
the factor at a stress level of 0.1σfu. This fact signifies that the
magnitude of mechanical load sustained prior to relaxation is
influential on the relaxation of CFRP-strengthened beams.
The aforementioned parametric study results were inte-
grated by the multiple regression method so that a unified
expression was acquired at a specific relaxation period (for
convenience, the following units are used: Ef in GPa, fc′ in
MPa, σf(0,T) in MPa, and T in Celsius)
λR(%) = c1Ef + c2fc′ + c3σf(0,T) + c4T + c5 (11)
where ci is a regression constant, available in Table 2. Figure 13
evaluates the accuracy of the regression equation using the test
data, both of which agreed with an average absolute difference
of 2.1%.
Fig. 13—Assessment of regression equation.
SUMMARY AND CONCLUSIONS
This paper has discussed the thermomechanical relaxation
and residual behavior of reinforced concrete beams strength-
ened with CFRP sheets. Test parameters encompassed
temperatures and relaxation periods. An infrared imaging
technique was employed to examine the heat conduction
of the strengthened beams across the CFRP-concrete inter-
face. The development of CFRP strains in conjunction
with various failure modes of the beams was emphasized.
Analytical modeling based on the temperature-dependent
characteristics of the CFRP system was conducted to predict
the degree of thermomechanical relaxation, including a
calibrated expression for relaxation constant, and was used
for parametric investigations. A thermomechanical relax-
ation factor was proposed based on the multiple regression

ACI Structural Journal/January 2018 267

method, which can assist practitioners involved in CFRP-
based rehabilitation for building structures.
• During thermomechanical relaxation, a latent heat effect
was noticed above the CFRP-concrete interface, which
retarded temperature progression into the concrete
beam. Thermal conduction along the CFRP weakened
the strength of the interface.
• Thermomechanical relaxation caused partial CFRP-
debonding when mechanically loaded in residual testing
and the measured debonding strains were lower than the
effective strain calculated by ACI 440.2R-08. The extent
of CFRP-strain retraction representing the integrity of the
CFRP-concrete interface (composite action) was affected
by elevated temperatures and relaxation time.
• The residual capacity of the strengthened beams was
reduced by thermomechanical relaxation, which also
influenced the load-displacement behavior and interfa-
cial cracking patterns of the beams. The trend of capacity
reduction was proportional to the degree of temperatures.
• Unlike concrete strength, the elastic modulus of CFRP
and the magnitude of mechanical load augmented the
thermomechanical relaxation of the strengthened beams.
The proposed relaxation factor is recommended to
address the potential degradation of CFRP systems when
subjected to a thermomechanical loading condition.
AUTHOR BIOS
Yail J. Kim, FACI, is a Professor in the Department of Civil Engineering
at the University of Colorado Denver, Denver, CO. He is the Chair of ACI
Committee 345, Concrete Bridge Construction, Maintenance, and Repair,
and ACI Subcommittee 440-I, FRP-Prestressed Concrete. He is a member of
ACI Committees 342, Evaluation of Concrete Bridges and Bridge Elements,
and 440, Fiber-Reinforced Polymer Reinforcement; and a member of Joint
ACI-ASCE Committee 343, Concrete Bridge Design. His research inter-
ests include advanced composite materials for structural rehabilitation,
complex systems, uncertainty quantification, and science-based structural
engineering including statistical, interfacial, and quantum physics.
Abdulaziz Alqurashi was a Graduate Student in the Department of Civil
Engineering at the University of Colorado Denver, where he received his
MS in civil engineering in 2015. His research interests include the thermal
behavior of structural concrete with advanced materials.
268
ACKNOWLEDGMENTS
The writers gratefully acknowledge financial support from Saudi Arabian
Cultural Mission and the U.S. Department of Transportation through the
Mountain-Plains Consortium. Proprietary information such as manufac-
turers and product names was not shown to avoid commercialism. Tech-
nical contents presented in this paper are based on the opinion of the writers
and do not necessarily represent that of others.
REFERENCES
1. Williams, B.; Kodur, V.; Green, M. F.; and Bisby, L., “Fire Endurance
of Fiber-Reinforced Polymer Strengthened Concrete T-Beams,” ACI Struc-
tural Journal, V. 105, No. 1, Jan.-Feb. 2008, pp. 60-67.
2. Kodur, V. K. R., and Yu, B., “Rational Approach for Evaluating Fire
Resistance of FRP-Strengthened Concrete Beams,” Journal of Compos-
ites for Construction, ASCE, V. 20, No. 6, 2016, 04016041. doi: 10.1061/
(ASCE)CC.1943-5614.0000697
3. Hu, W., Polymer Physics: A Molecular Approach, Springer, Berlin,
Germany, 2013.
4. Gamage, J. C. P. H.; Al-Mahaidi, R.; and Wong, M. B., “Bond Charac-
teristics of CFRP Plated Concrete Members under Elevated Temperatures,”
Composite Structures, V. 75, No. 1-4, 2006, pp. 199-205. doi: 10.1016/j.
compstruct.2006.04.068
5. Leone, M.; Matthys, S.; and Aiello, M. A., “Effect of Elevated Service
Temperature on Bond between FRP EBR Systems and Concrete,” Compos-
ites. Part B, Engineering, V. 40, No. 1, 2009, pp. 85-93. doi: 10.1016/j.
compositesb.2008.06.004
6. Tan, K.-H.; Toh, W.-S.; Huang, Z.-F.; and Phng, G. H., “Structural
Responses of Restrained Steel Columns at Elevated Temperatures,” Engi-
neering Structures, V. 29, No. 8, 2007, pp. 1641-1652. doi: 10.1016/j.
engstruct.2006.12.005
7. Sadaoui, A., and Khennane, A., “Effect of Transient Creep on the
Behaviour of Reinforced Concrete Columns in Fire,” Engineering Structures,
V. 31, No. 9, 2009, pp. 2203-2208. doi: 10.1016/j.engstruct.2009.04.005
8. ACI Committee 440, “Guide for the Design and Construction of Exter-
nally Bonded FRP Systems for Strengthening Concrete Structures (ACI
440.2R-08),” American Concrete Institute, Farmington Hills, MI, 2008, 76 pp.
9. Kraus, H., Creep Analysis, John Wiley & Sons, New York, 1980.
10. Kim, Y. J.; Hmidan, A.; and Choi, K.-K., “Residual Behavior of
Shear-Repaired Concrete Beams Using CFRP Sheets Subjected to Elevated
High Temperatures,” Journal of Composites for Construction, ASCE, V. 16,
No. 3, 2012, pp. 253-264. doi: 10.1061/(ASCE)CC.1943-5614.0000244
11. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-14) and Commentary,” American Concrete Institute,
Farmington Hills, MI, 2014, 520 pp.
12. ACI Committee 216, “Guide for Determining the Fire Endurance
of Concrete Elements (ACI 216R-89) (Reapproved 2001),” American
Concrete Institute, Farmington Hills, MI, 1989, 48 pp.
13. Mitchell, M., An Introduction to Genetic Algorithms, MIT Press,
Cambridge, MA. 1998.
14. Casalini, R.; Fioretto, D.; Livi, A.; Lucchesi, M.; and Rolla, P. A.,
“Influence of the Glass Transition on the Secondary Relaxation of an Epoxy
Resin,” Physical Review B: Condensed Matter and Materials Physics,
V. 56, No. 6, 1997, pp. 3016-3021. doi: 10.1103/PhysRevB.56.3016
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S22
Practicability of Large-Scale Reinforced Concrete Beams
Using Grade 80 Stirrups
by Jung-Yoon Lee, Jae-Hoon Lee, Do Hyung Lee, Seong-Jun Hong, and Ho-Young Kim
ACI 318-14 specifies that the yield strength of shear reinforcement
is limited to 420 MPa (60 ksi) for shear-strength calculation of rein-
forced concrete beams to control failure mode and diagonal crack
width for such members. However, in the case where high-strength
shear reinforcement is employed in such members, the amount of the
shear reinforcement can be reduced and wider spacing is feasible.
This may lead to an easy placement of concrete and, thus, improved
durability. Thus, the use of high-strength shear reinforcement can
be particularly effective for structures having large-scale members
such as skyscrapers, long-span bridges, and nuclear power plants.
To investigate the practicability of the high-strength shear rein-
forcement, shear tests are carried out for large-scale reinforced
concrete beams using Grade 80 stirrups. For this purpose, a total
of 16 reinforced concrete beams are tested and categorized into
four groups according to the parameters: two levels of stirrup yield
strength and four levels of cross section depth. Current test results
reveal that all large-scale beams fail in shear tension, experiencing
yielding of stirrups prior to concrete web crushing. Measured
shear strength is greater than theoretical shear strength calculated
by ACI 318-14 in all members. However, the margin between the
two strengths is reduced as the cross section depth increases. The
diagonal crack width is influenced by the stirrup yield strength and
spacing, and the cross section depth. The crack width increases as
the cross section depth and stirrup spacing increase.
Keywords: diagonal crack width; Grade 80 stirrups; large-scale reinforced
concrete beams; shear strength; size effect.
INTRODUCTION
Many studies on high-strength and high-performance
concrete have been extensively carried out over the last
30 years. Subsequently, high-strength concrete of over
80 MPa (11.6 ksi) has been used in design practice. However,
contrary to high-strength concrete, a few studies have been
conducted with respect to the employment of high-strength
steel in reinforced concrete structures. The use of high-
strength steel in reinforced concrete structures can have
beneficial effects on cost and constructability. Specifically,
special structures such as nuclear power plants, skyscrapers,
long-span bridges, and liquid nitrogen gas (LNG) storage
tanks require very large amounts of steel. For example,
the total amount of steel for the construction of a nuclear
power plant containment vessel is approximately 70,000 to
100,000 tons. In addition, large-diameter steel reinforce-
ments of over D35 (No. 11) are employed in reinforced
concrete skyscrapers over 150 m (492 ft) in height. The
use of high-strength steel in such structures could lead to a
reduction of the total steel amount and, hence, a reduction
in cost. Moreover, the use of high-strength steel may result
in a wider arrangement of reinforcement spacing, leading to
ACI Structural Journal/January 2018
TECHNICAL PAPER
easier concrete placement and, thus, an improved concrete
quality.
Application of the high-strength steel in reinforced concrete
structures requires various investigations regarding material
characteristics, member level performance (that is, flexure,
shear, bond, anchorage, axial compression, and seismic
performance). In particular, because primary national codes
for reinforced concrete structures (ACI 318-14,1 EU-04,2 and
CSA-043) are mostly based on empirical design formulas,
experimental evidence is required for the performance evalu-
ation of such structures employing high-strength steel.
For these reasons, much research has been dedicated to the
development of high-strength steel over the last decade in
many countries. In the United States, a great deal of exper-
imental research on shear, bond, and development length
has been conducted for the application of Grade 100 steel to
reinforced concrete structures.4,5 In Japan, similar research
regarding the development of high-strength steel has been
carried out, including an investigation of the characteristics
of axial compression, shear, and bond, as well as performing
finite element analysis.6 In Taiwan, high-strength steel has
been investigated through a national project: the Taiwan
New RC Project.7,8 In Korea, several research works have
been conducted on high-strength steel since 2006.9-12
In addition, works on the application of high-strength steel
to shear reinforcement have been conducted by Munikrishma
et al.,13 Selim et al.,14 and Lee et al.10-12 These works revealed
that, even for members in which Grade 80 high-strength stir-
rups are employed, the stirrups reached yield strain before
maximum strength was attained, resulting in shear tension
failure. Also, the average diagonal crack width was less than
0.41 mm (0.016 in.), which is the recommended crack width
according to ACI 224R-01.15 These results imply that the
use of Grade 80 high-strength steel as stirrups is promising.
However, in the above works, the members are somewhat
limited in cross section size because most of the reinforced
concrete beams are not larger than 600 mm (23.6 in.) in
section depth. Consequently, the shear behavior on large-
scale beams in which high-strength stirrups are employed
needs to be evaluated, considering that such members are, in
general, used in high-rise buildings, nuclear structures, and
other special structures such as long-span bridges. Nonethe-
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2017-089.R1, doi: 10.14359/51701092, was received March 26, 2017,
and reviewed under Institute publication policies. Copyright © 2018, American
Concrete Institute. All rights reserved, including the making of copies unless
permission is obtained from the copyright proprietors. Pertinent discussion including
author’s closure, if any, will be published ten months from this journal’s date if the
discussion is received within four months of the paper’s print publication.
269
Table 1—Details of specimens and material properties
Longitudinal tensile reinforcement Stirrup
Specimen fc′, MPa d, mm h, mm b, mm ρl, % fyl, MPa s, mm ρv, % fyt, MPa
BS1-1 — — —

Group BS1-2 130 0.219 459

43.6 501.5 600 500 1.61 695
BS1 BS1-3 180 0.158 674
BS1-4 130 0.219 674
BS2-1 — — —

Group BS2-2 40.0 130 0.219 459

751.5 850 500 1.61 695
BS2 BS2-3 180 0.158 674
BS2-4 46.2 130 0.219 674
BS3-1 — — —

Group BS3-2 130 0.219 459

48.5 1001.5 1100 500 1.61 695
BS3 BS3-3 180 0.158 674
BS3-4 130 0.219 674
BS4-1 — — —

Group BS4-2 130 0.219 459

54.8 1251.5 1350 500 1.61 695
BS4 BS4-3 180 0.158 674
BS4-4 130 0.219 674

Notes: 1 mm = 0.0394 in.; 1 MPa = 145 psi.

less, very few studies have been carried out for the applica-
tion of Grade 80 high-strength stirrups to those members.
Among such studies, works by Kani,16 Zararis and Papa-
dakis,17 Bentz,18 and Bažant19 demonstrated that the shear
behavior of large-scale reinforced concrete beams was influ-
enced by the size effect, and the size effect was closely asso-
ciated with the effective depth of a section and shear span-
depth ratio. Qiang and Bažant20 investigated the size effect
of reinforced concrete beams with stirrups, and concluded
that, while stirrups were able to mitigate the size effect,
they were not able to suppress it completely. Much research
has been dedicated to diagonal crack width for reinforced
concrete beams in the last few decades. Subsequently, the
influencing parameters and occurrence mechanism on the
crack have been investigated.21-25 Previous research has indi-
cated that the spacing of the crack is reduced by an increase
of shear reinforcement ratio, and the diagonal crack width is
increased by wider stirrup spacing. Lee et al.11,12 performed
a test on the shear behavior of reinforced concrete beams
with high-strength stirrups. They demonstrated that the
diagonal crack width is not proportional to the yield strength
of stirrups but is influenced by a number of stirrups (ρv fyt).
From a test on large-scale beams, Shoya et al.26 showed that
shear crack width was increased as the cross section of a
member increased. However, the aforementioned studies
are not sufficient enough to evaluate the size effect of rein-
forced concrete beams. This is because most of the afore-
mentioned studies are limited to either large-scale members
with normal-strength stirrups or relatively small beams with
section depths of less than or equal to 600 mm (23.6 in.).
Therefore, research on the shear behavior of large-scale
beams of not less than 1000 mm (39.4 in.) in section depth
with Grade 80 high-strength stirrups is urgently needed.
In the present study, experimental work was conducted on
large-scale reinforced concrete beams with Grade 80 high-
strength stirrups. The yielding of high-strength stirrups,
diagonal crack width, and size effect are observed and, thus,
the applicability of high-strength stirrups to large-scale rein-
forced concrete beams is evaluated.
RESEARCH SIGNIFICANCE
Most studies on the applicability of high-strength shear
reinforcement focus on structural members with relatively
small section dimensions. However, high-strength shear
reinforcement is generally employed in skyscrapers or large-
scale structures rather than in mid- or low-rise structures.
Therefore, investigation on the applicability of high-strength
shear reinforcement to large-scale members is important
and necessary. When high-strength shear reinforcement
is employed in a large-scale member, the total amount of
reinforcement can be reduced and, thus, a wider spacing
between the reinforcements is possible. This can result in
an improved quality of concrete due to easy placement of
concrete. However, a detrimental effect may also occur,
whereby diagonal crack width can be increased in large-
scale members with high-strength shear reinforcement, and
the shear strength provided by the high-strength shear rein-
forcement can vary due to the size effect. To investigate this
phenomenon, a total of 16 large-scale reinforced concrete
beams employing Grade 80 high-strength stirrups are tested
in the present study. The test results are expected to provide
insight into the shear behavior of such members, as well
as useful information on the feasibility of employing these
high-strength stirrups in such members.

270 ACI Structural Journal/January 2018

 TEST SPECIMEN AND PROCEDURE
A total of 16 large-scale reinforced concrete beams were
constructed to evaluate the applicability of high-strength
stirrups to the specimens. Details of the beam specimens and
material properties are summarized in Table 1. The primary
parameter of the specimens is the cross section size. Four
different section sizes are employed in this study: 500 x
600 mm (19.7 x 23.6 in.), 500 x 850 mm (19.7 x 33.5 in.),
500 x 1100 mm (19.7 x 43.3 in.), and 500 x 1350 mm (19.7
x 53.1 in.), as shown in Fig. 1 and 2. For the stirrups, two
different types of steel, Grade 60 and Grade 80, are used in
the fabrication of the specimens. Tensile tests on both types
of steel indicate that the actual yield strengths of Grade 60
and Grade 80 steel are 459 and 674 MPa (66.6 and 97.8 ksi),
respectively, as shown in Table 1 and Fig. 3. The actual yield
strength of Grade 80 steel used for longitudinal reinforce-
ment of No. 11 bar is 695 MPa (100.8 ksi). The shear span-
Fig. 1—Test setup and measurement.
Fig. 2—Overall dimensions of specimens and locations of linear variable differential transformers (LVDTs). (Note dimensions
in mm; 1 mm = 0.0394 in.)
ACI Structural Journal/January 2018
depth ratio is designed to be 2.5 in all specimens, being shear
critical. Also, the tensile reinforcement ratio is designed to
be 1.61%, which is identical in all specimens. The inten-
tion for the identical ratio is to ensure the same influence
of the tensile reinforcement on size effect. The ratio is set
as relatively high to prevent flexure failure prior to shear
failure. Meanwhile, it is worth noting that longitudinal skin
reinforcements are not arranged in the specimens of 500
x 1100 mm (19.7 x 43.3 in.) and 500 x 1350 mm (19.7 x
53.1 in.), although ACI 318-141 specifies that the longitu-
dinal skin reinforcement needs to be uniformly distributed
along both side faces of a beam in case where section depth
of the beam exceeds 914.4 mm (36 in.). This is to evaluate
directly the influence of high-strength stirrups on crack
width for specimens with different section depths.
Test specimens are categorized into four different groups
(BS1, BS2, BS3, and BS4) depending on the cross section
sizes, and each group is composed of four specimens
in which the primary parameters are spacing and yield
strength of stirrup. The specimens numbered as 1- in each
group have no stirrups, while those numbered with 2- have
Grade 60 stirrups with 130 mm (5.12 in.) spacing, and those
numbered with 3- have Grade 80 stirrups with 180 mm
(7.09 in.) spacing. This arrangement leads to an identical
stirrup amount (ρv fyt) of 1.0 MPa (145 psi) in both speci-
mens. For the specimens numbered with 4-, Grade 80 stirrups
with 130 mm (5.12 in.) spacing are employed. The purpose
of these overall arrangements is to compare and evaluate
the effects of high strength stirrups on shear strength and
diagonal crack width in large-scale beams. Meanwhile, the
compressive strength of concrete is designed to be 45 MPa
(6635 psi), and the strength given in Table 1 represents the
actual strength at the date of the test.
All specimens are designed in accordance with the shear
equations given in ACI 318-141 (Eq. 22.5.1.1, 22.5.6.1(a),
and 22.5.10.5.3) and in accordance with provisions regarding
271
maximum and minimum stirrup ratio, and maximum stirrup
spacing in ACI 318-14.1 Particular emphasis is placed on the
stirrup ratio, which is as low as possible to investigate the
size effect of large-scale beams. However, closely spaced
high-strength stirrups are arranged in both ends of the spec-
imens to prevent support failure. In addition, to avoid local
concrete crushing or bearing failure at both loading points
and supports, rectangular steel plates of 100 x 100 mm (3.94
x 3.94 in.) are placed at such a location.
Loading was applied using a universal testing machine
(UTM) of 5000 kN (1,124,045 lb) capacity with displace-
ment control, and the applied loading was recorded in a data
acquisition instrument (brand name; remove). Strain gauges
were installed in both longitudinal tensile reinforcements
and stirrups, as illustrated in Fig. 2 to investigate the yielding
of both reinforcements, and, thus, the failure modes of the
specimens. The average deformation of the specimens was
estimated using 12 linear variable differential transformers
(LVDTs): three horizontal, two vertical, and one diagonal
Fig. 3—Tensile stress-tensile strain relationship of steel
bars. (Note: 1 MPa = 145 psi.)
Fig. 4—Representative crack patterns after test for specimens of Group BS4.
272
in both shear-critical areas. For the diagonal crack width of
concrete struts in the shear critical areas, two schemes were
employed to measure the width precisely. Demac gauges and
crack magnifiers having a minimum scale division of 0.001
and 0.01 mm (0.000039 and 0.00039 in.), respectively, were
used in the test. All the tests were terminated when the load
reached 90% of the maximum strength after the maximum
strength was achieved.
STRUCTURAL BEHAVIOR
Crack patterns
Shear failures occurred in all specimens in which maximum
strength was attained before the strain gauges attached to
the longitudinal tensile reinforcements reached yield strain.
Representative specimens of the BS4 group are shown in
Fig. 4, where typical shear cracks are clearly shown. Flex-
ural cracks were initially formed in the portion of pure flex-
ural moment between loading points. More flexural cracks
developed near loading points where the maximum flexural
moment and shear force occurred simultaneously. These flex-
ural cracks then changed to flexure-shear cracks as the load
was increased. When the load was near maximum strength,
shear cracks occurred in a web between the support and the
loading point. For the specimens without stirrups, diagonal
cracks having a large width occurred after the formation of
flexural cracks. The load was then slightly increased after
the formation of the diagonal cracks. However, the load
was then decreased immediately after additional diagonal
cracks developed. Many diagonal cracks in specimens with
stirrups were relatively greater than those in specimens
without stirrups. Diagonal crack width in specimens with
stirrups increased up to 2 mm (0.079 in.) when the maximum
strength was almost reached. The estimated diagonal crack
ACI Structural Journal/January 2018
Fig. 5—Shear force-deflection relationship of each Group. (Note: Rectangular section dimensions are in mm; 1 mm = 0.039 in.;
1 kN = 224.8 lb.)
Table 2—Shear strength of specimens
width was in general proportional to the cross section size.
Vn-test, Vn-ACI, Vn-EC2,
Nonetheless, many cracks were almost constant. Specimen kN kN kN V /V V /V n-test n-ACI n-test n-EC2

BS1-1 524.9 293.1 303.8 1.79 1.73

Shear force-deflection relationship
The shear force-deflection relationship for all speci- Group BS1-2 833.5 544.6 567.1 1.53 1.47
mens of the four groups is shown in Fig. 5. In general, the BS1 BS1-3 727.7 559.8 600.8 1.30 1.21
maximum shear strength is proportional to the cross-section
BS1-4 807.1 662.3 832.8 1.22 0.97
depth. The shear strength developed in specimens without
stirrups started to decrease immediately after two to three BS2-1 424.9 422.8 411.0 1.01 1.03
diagonal cracks had formed. However, the deflection in Group BS2-2 1239.1 799.6 849.8 1.55 1.46
such specimens was not significant. Similar behavior was BS2 BS2-3 915.5 822.4 900.3 1.11 1.02
observed in specimens with stirrups where the shear strength
BS2-4 1130.4 1004.5 1247.9 1.13 0.91
was reduced abruptly once the maximum shear strength was
attained. Figure 5 also depicts the stiffness for each group. BS3-1 442.9 614.6 557.4 0.72 0.80
Whereas the maximum shear strength is achieved in speci- Group BS3-2 1421.8 1116.7 1132.6 1.27 1.26
mens with Grade 60 stirrups for Groups 1 and 2, stiffness in BS3 BS3-3 1178.5 1147.1 1199.8 1.03 0.98
such members is less than that in specimens with Grade 80
BS3-4 1490.6 1351.9 1663.1 1.10 0.90
stirrups. However, this is not the case for Groups 3 and 4. As
observed, the maximum shear strength is attained in speci- BS4-1 886.0 811.8 701.9 1.09 1.27
mens with Grade 80 stirrups, but stiffness in such members is Group BS4-2 1684.1 1439.2 1415.3 1.17 1.19
less than that in specimens with Grade 80 stirrups. Maximum BS4 BS4-3 1683.3 1477.2 1499.3 1.14 1.12
shear strength developed in the test is summarized in Table 2.
BS4-4 1795.8 1733.1 2078.2 1.04 0.86
A comparison of the maximum shear strength divided by the
cross section area is discussed in detail later. Notes: Vn-ACI is ACI 318-14 ; Vn-EC2 is EC2-04 ; 1 kN = 224.8 lb.
1 2

possibility of concrete web crushing prior to yielding of the

Shear failure mode
stirrup. Previous test data10-12 for beams less than 600 mm
The basic principle in primary design codes based on the
(23.6 in.) in cross section depth reported that beams with
truss model is that the yielding of a stirrup occurs prior to the
high-strength stirrups of 650 MPa (94.3 ksi) greater than
web crushing of concrete, leading to shear tension failure.
Grade 60 (420 MPa [60.0 ksi]) specified by ACI 318-141
For beams with high-strength stirrups, yield strain is linearly
failed in shear tension. The previous test results also eval-
increased as yield strength increases. This implies a strong
uated whether shear tension failure occurred in large-scale

ACI Structural Journal/January 2018 273

Fig. 6—Strain distribution of stirrups. (Note: 1 mm = 0.039 in.)
beams with high strength stirrups, having a large concrete
contribution to shear strength. To investigate the failure mode
of the current test specimens, stirrup strains were evaluated
and their distribution at five different loading levels is shown
in Fig. 6. The x- and y-axes in Fig. 6 represent the loca-
tion of stirrups between two supports and the stirrup strains,
respectively. Vmax in the legend refers to the maximum shear
strength obtained from the test. As observed, the stirrups
underwent yield strain in all specimens prior to reaching
the maximum shear strength. It is therefore assumed that all
specimens fail in shear tension required by ACI 318-14.1
For a further assessment of failure mode in relation
to cross section size, a comparison was made in terms
of using stirrup strains in both the current and previous Fig. 7—Comparison of normalized stirrup strains for
various levels of ρvfyt√fc′.

274 ACI Structural Journal/January 2018


Fig. 8—Normalized shear strength for various section
depths. (Note: 1 mm = 0.0394 in.)
test results.10-12 Figure 7 shows the stirrup strains for the
current test with large-scale cross section and the previous
test10-12 with a section depth less than or equal to 600 mm
(23.6 in.). The y-axis in Fig. 7 represents the stirrup strain
εpeak measured at the maximum shear strength divided by
the yield strain of stirrup εy—that is, εpeak/εp ≥ 1 indicates
yielding of stirrup. Shear failure mode is closely associated
with the yield strength of stirrup fyt, stirrup ratio ρv, and
compressive strength of concrete fc′. In the case where the
value of ρv fyt/√fc′ for a member is large, there is a strong
possibility that the member fails in shear compression—
that is, concrete web crushing occurs prior to yielding of
stirrup. However, for the case of a small value of ρv fyt/√fc′,
the member is highly expected to fail in shear tension—that
is, concrete web crushing occurs after yielding of stirrup.
In Fig. 7, the stirrup yield strength of fyt varies between 459
and 750 MPa (66.6 and 100.1 ksi), and the maximum stirrup
ratio satisfies the ACI 318-141 provision of ρv fyt/√fc′ ≤ 2/3.
In Fig. 7, it is observed that εpeak/εy decreases as ρv fyt/√fc′ is
increased. It is worth noting that, as shown in Fig. 7, shear
compression failure occurs in two specimens with both high-
strength stirrup and high value of ρv fyt/√fc′. It is also note-
worthy that, as shown in Fig. 7, εpeak/εy is more influenced
by ρv fyt/√fc′ than by the cross section size. The variance of
εpeak/εy measured in large-scale specimens of the current test
exhibits a similar trend to that in the small-scale specimens
ACI Structural Journal/January 2018
Fig. 9—Normalized shear strength for various levels of
ρvfyt√fc′. (Note: 1 mm = 0.0394 in.; 1 MPa = 145 psi.)
of the previous test. It is thus observed from Fig. 7 that no
significant change occurs in shear failure mode, even for
large-scale members with high-strength stirrups.
DISCUSSION
Shear strength
Figure 8(a) shows the shear strength variation according
to cross section depth. For comparison, the normalized shear
strength ratio is used and summarized in Table 2. The normal-
ized shear strength ratio in the y-axis is the measured shear
strength Vn-test divided by the theoretical shear strength Vn-ACI
calculated using the ACI 318-141 equations: Eq. (22.5.1.1),
(22.5.6.1(a)), and (22.5.10.5.3). The strength reduction factor
of 1.0 and the actual measured yield strength of stirrups are
employed in the calculation of the theoretical shear strength.
For specimens without stirrups (BS1-1, BS2-1, BS3-1, and
BS4-1) shown in Fig. 8(a), while the normalized ratios in
Specimens BS1-1 and BS4-1 are greater than 1.0, those in
Specimens BS2-1 and BS3-1 are not greater than 1.0. In
particular, although the normalized ratio of BS4-1 is greater
than 1.0, the margin is as low as 1.1 approximately. This
implies that Eq. (22.5.1.1) of ACI 318-141 is likely to give an
unconservative shear strength prediction for specimens with
a cross section depth of not less than 850 mm (33.5 in.). In
general, the normalized ratio for specimens without stirrups
tends to decrease as the cross section size increases.
The normalized ratios for all specimens with stirrups are
greater than 1.0. However, they asymptotically decrease as
the cross section size increases. The normalized ratio was
a particularly low value of 1.03 and 1.04 for Specimens
BS3-3 and BS4-4, respectively. The decreasing tendency in
the normalized ratio reveals the unfavorable aspect whereby
theoretical strength by ACI 318-141 is likely to exceed the
measured strength for specimens with a cross section depth
greater than 1350 mm (53.1 in.). This suggests that the shear
equation by ACI 318-141 may lead to an erroneous overesti-
mation in the theoretical shear strength as the cross section
size increases. This is due to the fact that the calculation
of shear strength provided by concrete Vc in the equation
does not take into account the influence of size effect. The
normalized ratio for specimens with Grade 60 stirrups is
generally greater than that for specimens with Grade 80 stir-
275
Fig. 10—Shear-critical areas and schemes for diagonal crack measurement. (Note: Dimensions are in mm; 1 mm = 0.0394 in.)
rups. The margin of both ratios decreases as the cross section
size increases.
Further comparison is made in terms of using the predic-
tions by EC2-04.2 Theoretical shear strength Vn-EC2 calcu-
lated by EC2-042 shear design equations and the normal-
ized shear strength ratio Vn-test/Vn-EC2 are also given in
Table 2. In general, similar trends as in the case of compar-
ison with ACI 318-141 are observed. However, the theoret-
ical predictions by EC2-042 are generally greater than those
by ACI 318-14.1 This is attributed to the fact that shear
design equations of EC2-042 account for shear resistance by
stirrups only and the angle between the concrete compres-
sion strut and the beam axis perpendicular to the shear force.
The limiting values of the angle are varied and can be found
in EC2-04.2 This leads to an increase of the shear predictions
as the cross section depth increases and, thus, a decrease in
the normalized shear strength ratio is prominent for beams
with a large cross section depth. In particular, the normalized
shear strength ratio is less than 1 for specimens with a cross
section depth with 1350 mm (53.1 in.). This indicates that
shear equation of EC2-042 is likely to overestimate the shear
strength with a cross section depth greater than 1350 mm
(53.1 in.). Consequently, further assessment is required when
Grade 80 stirrups are employed in a large-scale member with
a size greater than that in the present study.
Figure 8(b) shows the measured shear strength provided by
stirrups divided by the calculated shear strength provided by
stirrups using Eq. (22.5.10.5.3) of ACI 318-141 for cross section
depths. The measured shear strength provided by stirrups is
obtained from the shear strength of a specimen with stirrups
subtracted by that of such specimen without stirrups in each
group. Figure 8(b) shows that the normalized ratio of shear
strength provided by stirrups for members with Grade 80 stir-
rups is lower than that for such members with Grade 60 stirrups.
Figure 9 shows a comparison of the relationship between
normalized shear strength ratio Vn-test/Vn-ACI and amount of
stirrups divided by the compressive strength of concrete
ρv fyt/fc′. Depicted in Fig. 9 are the previous test results8,9,25
for beams with high-strength stirrups and a depth not greater
than 600 mm (23 in.) and the current test results. A total of
276
78 specimens are employed in the comparison. Parameter
ranges vary widely: between 334 and 750 MPa (48.4 and
108.8 ksi) for yield strength of stirrup; between 25.0 and
81.4 MPa (3.6 and 11.8 ksi) for compressive strength of
concrete; and between 350 and 1350 mm (13.8 and 53.1 in.)
for cross section depth. Details of the specimens used in the
comparison are summarized in the Appendix.*
As observed in Fig. 9, the normalized shear strength ratio for
specimens with a depth not greater than 600 mm (23.6 in.) is
generally decreased as the value of ρv fyt/fc′ increases. More
specifically, while the ratio is 1.3 at 0.05 of ρv fyt/fc′, the ratio
decreases to 1.05 at 0.1 of ρv fyt/fc′. In addition, the ratio for
specimens with high-strength stirrups is lower than that
for specimens with normal-strength stirrups at an identical
value of ρv fyt/fc′. Meanwhile, it is worth noting that the value
of ρv fyt/fc′ is a relatively low level of 0.025 for the current test
specimens with Grade 80 stirrups and a cross section depth
of not less than 850 mm (33.5 in.). Nonetheless, the normal-
ized shear strength ratio of such specimens is approximately
1.1, which is less than that of specimens with a cross section
depth not greater than 600 mm (23.6 in.).
Diagonal cracks
The primary national design codes for reinforced concrete
structures including ACI 318-141 limit the yield strength of
stirrups. One of the main reasons is to control crack width.
In particular, when high-strength stirrups are employed in a
reinforced concrete structure, there is a high possibility that
diagonal crack width is increased. Shoya et al.26 revealed
that the width can be increased as the section depth is
increased. In the present study, measurements for the width,
angle of inclination, and the number of diagonal cracks are
taken in shear-critical areas. The shear-critical areas are
assumed to be the portion of the concrete web between a line
equal to half the section depth from the loading point and
a location extending the section depth from the line along
*
The Appendix is available at www.concrete.org/publications in PDF format,
appended to the online version of the published paper. It is also available in hard copy
from ACI headquarters for a fee equal to the cost of reproduction plus handling at the
time of the request.
ACI Structural Journal/January 2018
Fig. 11—Measured diagonal crack width. (Note: 1 mm = 0.0394 in.)
the member axis. These areas are illustrated as the shaded
areas in Fig. 10. When measuring the crack width, a straight
line perpendicular to a diagonal crack is drawn as shown in
Fig. 10. Then, Demec gauges and crack magnifiers are used
to measure the diagonal crack width at each loading level.
Cracks formed in the top and bottom cross sections are not
included in the measurement because they are mostly flex-
ural cracks. In addition, cracks formed close to the supports
or loading points are also excluded. Thus, only diagonal
cracks occurring in the middle web of half the section depth
are evaluated. Cracks satisfying the following conditions in
the shear-critical areas are regarded as diagonal cracks:
1. In the case where a line between the starting and ending
points of cracks occurred in the upper and lower parts of the
ACI Structural Journal/January 2018
shear critical areas has an angle of inclination between 20
and 60 degrees.
2. In the case where the vertical length of a crack in the
shear critical areas is greater than or equal to d/6.
Figure 11 shows the change of diagonal crack width
measured in the shear-critical areas at various load levels.
Diagonal crack barely occurred up to 40% of the load level.
Once the load level exceeded 40%, a diagonal crack formed
in the concrete web, and the crack width increased as the
load level increased. The crack width varied according to
the occurrence of the crack. In general, the earlier the occur-
rence, the larger the crack width at the maximum load level.
The difference between the maximum and minimum crack
widths is approximately 1 mm (0.039 in.) in identical spec-
277
Fig. 12—Average diagonal crack width at service load level.
(Note: 1 mm = 0.0394 in.)
imens. Figure 12 shows the average width of the diagonal
cracks measured at 60% of shear strength calculated by ACI
318-14.1 It is generally observed that the crack width and
spacing between the cracks increase as the stirrup spacing
increases from 130 to 180 mm (5.12 to 7.09 in.). In partic-
ular, the average diagonal crack width in specimens with
Grade 80 stirrups is larger than that in specimens with
Grade 60 stirrups. The width increases as the cross section
depth increases. For specimens with a cross section depth
not greater than 600 mm (23.6 in.) in the previous test,10-13
the width of such specimens with Grade 80 stirrups is not
greater than 0.41 mm (0.017 in.), which is the upper bound
recommended by ACI 224R-0115 at the service load level.
However, for a cross section depth not less than 1200 mm
(47.2 in.) in the current test, the width exceeds 0.41 mm
(0.017 in.). This implies that large-scale specimens with
high-strength stirrups may exceed 0.41 mm (0.017 in.) of
the crack width. Fig. 13—Average angle of inclination and average number
Figures 13(a) and 13(b) show the average angle of inclina- of diagonal cracks at service load level. (Note: 1 mm =
tion and the number of diagonal cracks measured in the shear 0.0394 in.)
areas, respectively. As observed, a number of diagonal cracks
increased as the cross section depth was increased. However, ment, additional test results are required respect to a number
the angle of inclination of diagonal cracks slightly reduced of the diagonal cracks and the angle of the cracks in relation
as the cross section depth was increased. Although the shear to cross section depth.
span-depth ratio is identical in all specimens, the area of
concrete web is relatively increased as cross section depth SUMMARY AND CONCLUSIONS
increases. Thus, numerous diagonal cracks passing through The applicability of Grade 80 high-strength stirrups to
the relatively large area of concrete web increased when the reinforced concrete members has been actively studied over
cracks formed due to force redistribution. The angle of incli- the last decade. However, most previous works focused
nation of diagonal cracks in group BS1 with the smallest on members with relatively small-scale cross sections,
cross section depth increased because the cracks initiated at and very few studies have been carried out for large-scale
a certain distance away from support and propagated toward members with Grade 80 stirrups. In this paper, evaluation
a loading point. However, the angle of cracks in Group BS4 was performed with respect to the shear behavior of large-
with the largest cross section depth decreased because the scale reinforced concrete beams with Grade 80 stirrups. For
cracks formed initially near of support and extended toward this purpose, a total of 16 beam specimens were tested in the
a loading point. In addition, the angles of cracks are closely current study. The conclusions derived from the current test
associated with a stirrup spacing together with the cross results are presented as follows:
section depth. The angle of cracks decreased as the stirrup 1. All of the large-scale reinforced concrete beams with
spacing increased. The above observation is derived from Grade 80 stirrups failed in shear tension—that is, yielding of
the present test results. Subsequently, for a further assess- stirrups occurs prior to concrete web crushing. The measured
shear strength is greater than the theoretical shear strength

278 ACI Structural Journal/January 2018

calculated by ACI 318- 141 in all specimens, even for speci-
mens of 1350 mm (53.1 in.) section depth in which Grade 80
stirrups were employed. It is thus promising that Grade 80
high-strength stirrups can be employed in large-scale rein-
forced concrete beams. It is, however, worth noting that the
normalized shear strength ratio Vn-test/Vn-ACI decreases as the
cross section depth increases.
2. Previous and current test results on 78 reinforced
concrete beams with high-strength stirrups indicate that the
normalized shear strength ratio of large-scale beams is lower
than that of relatively small-scale beams. In the present test
however, a low level of 0.025 for ρv fyt/fc′ is employed in the
large-scale beams. Hence, additional investigation is needed
to determine the applicability of high-strength stirrups to
large-scale beams with a relatively high level of ρv fyt/fc′.
3. The average diagonal crack width reduces as the
number of stirrups is increased. The width in beams with
Grade 80 stirrups is larger than that in beams with Grade 60
stirrups and increased as the stirrup spacing was increased.
In addition, the width increased as the cross section depth is
increased. In particular, for beams with a cross section depth
of 1350 mm (53.1 in.), the width is greater than 0.41 mm
(0.017 in.), which is the crack width recommended by ACI
224R-0115 at the service load level.
4. The angle of inclination of diagonal cracks is reduced
as the cross section depth increases and the stirrup ratio is
reduced. The number of diagonal cracks increased as the
cross section depth is increased. Although it is rather prema-
ture to generalize a tendency, the aforementioned phenomena
can be used as useful information regarding the evaluation
of shear behavior on large-scale beams, in the lack of avail-
able test data on the angle of inclination and the number of
diagonal cracks for such members.
Conclusions discussed previously are derived from a total
of 16 specimens. Subsequently, additional tests are required
further to generalize a practicability of the Grade 80 stirrups
and propose some suggestion in design formulas. In partic-
ular, influence of concrete compressive strength, stirrup
spacing, and ratio on shear behavior of large-scale beams
needs to be evaluated in the further tests.
AUTHOR BIOS
ACI member Jung-Yoon Lee is a Professor in the School of Civil, Architec-
tural Engineering, and Landscape Architecture at Sungkyunkwan Univer-
sity, Seoul, South Korea. His research interests include shear behavior and
seismic design of reinforced and prestressed concrete structures.
ACI member Jae-Hoon Lee is a Professor in the Department of Civil Engi-
neering at Yeungnam University, Gyongsan, South Korea. He received
his BE and MS from Seoul National University, Seoul, South Korea, and
his PhD from the University of Wisconsin–Madison, Madison, WI. His
research interests include behavior, analysis, and design of reinforced and
prestressed concrete structures.
Do Hyung Lee is a Professor in the Department of Civil, Environmental
and Railroad Engineering at Pai Chai University, Daejeon, South Korea.
He received his BSC and MSc from Hanyang University, South Korea, and
his MSc and PhD from Imperial College London, London, UK. His research
interests include seismic design and performance of concrete structures:
analytical and experimental studies.
Seong-Jun Hong is an Engineer at Samsung C&T, Engineering &
Construction Group, South Korea. He received his MS from Sungkyunkwan
University. His research interests include the structural analysis and design
of concrete structures.
ACI Structural Journal/January 2018
Ho-Young Kim is a PhD Candidate in the Department of Civil Engineering
at Yeungnam University, Gyongsan, South Korea, where he received his BE
and MS in civil engineering.
ACKNOWLEDGMENTS
This research was supported by the Korea Hydro & Nuclear Power Co.,
LTD (KHNP, contract number: L16S140000) and the Basic Science Research
Program through the National Research Foundation of Korea (NRF) funded
by the Ministry of Education (grant number: 2015R1D1A1A01056612).
NOTATION
b = width of beam cross section
d = effective depth of beam cross section
fc′ = compressive strength of concrete
fyl = yield strength of longitudinal tensile reinforcement
fyt = yield strength of stirrup
h = depth of beam cross section
s = stirrup spacing
V = shear strength
Vc = shear strength provided by concrete
Vn-ACI = shear strength calculated by ACI 318-14
Vn-test = measured shear strength by test
Vs = shear strength provided by stirrups
ρv = stirrup ratio
ρv fyt = amount of stirrups
REFERENCES
1. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 2014, 519 pp.
2. Comite European de Normalisation (CEN), “Eurocode 2: Design of
Concrete Structures, Part 1-1 General Rules and Rules for Buildings (BS
EN 1992-1-1),” Lausanne, Switzerland, 2004, 211 pp.
3. CSA Committee A23.3-04, “Design of Concrete Structures for Build-
ings (CAV3-A23.3-04),” Canadian Standards Association, Toronto, ON,
Canada, 2004, 232 pp.
4. Hosny, A.; Seliem, H. M.; Rizkalla, S.; and Zia, P., “Development
Length of Unconfined Conventional and High-Strength Steel Reinforcing
Bars,” ACI Structural Journal, V. 109, No. 5, Sept.-Oct. 2012, pp. 655-664.
5. Briggs, M.; Miller, S.; Darwin, D.; and Browning, J., “Bond Behavior
of Grade 100 AST A1035 Reinforcing Steel in Beam-Splice Specimens,”
SL Report 07-1, The University of Kansas Center for Research, Lawrence,
KS, 2007, 83 pp.
6. Aoyama, H., Design of Modern Highrise Reinforced Concrete Struc-
tures, Imperial College Press, London, UK, 2001, 442 pp.
7. Li, Y.-A., and Hwang, S.-J., “Prediction of Shear Strength for Rein-
forced Concrete Short Columns Using High Strength Material,” The
15th Korea-Japan-Taiwan Joint Seminar on Earthquake Engineering for
Building Structures, Taipei, Taiwan, 2013, pp. 41-50.
8. Ou, Y. C., and Kurniawan, D. P., “Effect of Axial Compression on
Shear Strength of Reinforced Concrete Columns with High-Strength Steel
and Concrete,” The 15th Korea-Japan-Taiwan Joint Seminar on Earthquake
Engineering for Building Structures, Taipei, Taiwan, 2013, pp. 105-114.
9. Choi, W.-S.; Park, H.-G.; Chung, L.; and Kim, J. K., “Experimental
Study for Class B Lap Splice of 600 MPa (87 ksi) Reinforcing Bars,” ACI
Structural Journal, V. 111, No. 1, Jan.-Feb. 2014, pp. 49-58.
10. Lee, J.-Y., and Choi, I.-J., “Shear Behavior of Reinforced Concrete
Beams with High-Strength Stirrups,” ACI Structural Journal, V. 108,
No. 5, Sept.-Oct. 2011, pp. 620-629.
11. Lee, J.-Y.; Lee, D. H.; Lee, J.-E.; and Choi, S.-H., “Shear Behavior
and Diagonal Crack Width for Reinforced Concrete Beams with High
Strength Shear Reinforcement,” ACI Structural Journal, V. 112, No. 3,
May-June 2015, pp. 323-333. doi: 10.14359/51687422
12. Lee, J.-Y.; Choi, S.-H.; and Lee, D. H., “Structural Behavior of Rein-
forced Concrete Beams with High Yield Strength of Stirrups,” Magazine
of Concrete Research, V. 68, No. 23, 2016, pp. 1187-1199. doi: 10.1680/
jmacr.15.00344
13. Munikrishma, A.; Hosny, A.; Rizkalla, S.; and Zia, P., “Behavior of
Concrete Beams Reinforced with ASTM A1035 Grade 100 Stirrups under
Shear,” ACI Structural Journal, V. 108, No. 4, July-Aug. 2011, pp. 34-41.
14. Seliem, H.; Hosny, A.; Rizkalla, S.; Zia, P.; Briggs, M.; Miller, S.;
Darwin, D.; Browning, J.; Glass, G.; Hoyt, K.; Donnelly, K.; and Jirsa, J.,
“Bond Characteristics of ASTM A1035 Steel Reinforcing Bars,” ACI Struc-
tural Journal, V. 106, No. 4, July-Aug. 2009, pp. 530-539.
15. ACI Committee 224, “Control of Cracking in Concrete Structures
(ACI 224R-01),” American Concrete Institute, Farmington Hills, MI, 2001,
46 pp.
279
 16. Kani, G. N. J., “How Safe Are our Large Reinforced Concrete
Beams?” ACI Journal Proceedings, V. 64, No. 3, Mar. 1967, pp. 128-141.
17. Zararis, P., and Papadakis, G. C., “Diagonal Shear Failure and Size
Effect in RC Beams without Web Reinforcement,” Journal of Structural
Engineering, ASCE, V. 127, No. 7, 2001, pp. 733-742. doi: 10.1061/
(ASCE)0733-9445(2001)127:7(733)
18. Bentz, E. C., “Shear Strength of Members without Transverse Rein-
forcement,” Structures Congress, 2009, pp. 1604-1611.
19. Bažant, Z. P., and Yu, Q., “Designing Against Size Effect on Shear
Strength of Reinforced Concrete Beams without Stirrups: I. Formulation,”
Journal of Structural Engineering, ASCE, V. 131, No. 12, 2005, pp. 1877-
1885. doi: 10.1061/(ASCE)0733-9445(2005)131:12(1877)
20. Yu, Q., and Bažant, Z. P., “Can Stirrups Suppress Size Effect on Shear
Strength of RC Beams?” Journal of Structural Engineering, ASCE, V. 137,
No. 5, 2011, pp. 607-617. doi: 10.1061/(ASCE)ST.1943-541X.0000295
21. Regan, P. L., and Placas, A., “Shear Failure of Reinforced Concrete
Beams,” ACI Journal Proceedings, V. 68, No. 10, Oct. 1971, pp. 763-773.
280
22. Adebar, P., “Diagonal Cracking and Diagonal Crack Control in Struc-
tural Concrete” Design and Construction Practices to Mitigate Cracking,
SP-204, E. G. Nawy, F. Barth, and R. Frosch, eds., American Concrete Insti-
tute, Farmington Hills, MI, 2001, pp. 85-116.
23. Brooms, B., “Crack Width and Crack Spacing in Reinforced
Concrete Members,” ACI Journal Proceedings, V. 62, No. 10, Oct. 1965,
pp. 1237-1256.
24. Hassan, H. M.; Farghaly, S. A.; and Ueda, T., “Displacements at
Shear Crack in Beams with Shear Reinforcement under Static and Fatigue
Loadings,” Proceedings of the JSCE, V. 19, Aug. 1991, pp. 215-222.
25. Zakaria, M.; Ueda, T.; Wu, Z.; and Meng, L., “Experimental Inves-
tigation on Shear Cracking Behavior in Reinforced Concrete Beams with
Shear Reinforcement,” Journal of Advanced Concrete Technology, V. 7,
No. 1, 2009, pp. 79-96. doi: 10.3151/jact.7.79
26. Shoya, T.; Iguro, M.; Nojiri, Y.; Akiyama, H.; and Okada, T., “Shear
Strength of Large Reinforced Concrete Beams,” Fracture Mechanics:
Application to Concrete, SP-118, V. C. Li and Z. P. Bažant, eds., American
Concrete Institute, Farmington Hills, MI, pp. 259-280.
ACI Structural Journal/January 2018
 ACI STRUCTURAL JOURNAL
Title No. 115-S23
Use of Anchored Carbon Fiber-Reinforced Polymer Strips
for Shear Strengthening of Large Girders
by William A. Shekarchi, Wassim M. Ghannoum, and James O. Jirsa
The objective of this research was to investigate the feasibility of
increasing the shear capacity of large reinforced concrete girders
using anchored carbon fiber-reinforced (CFRP) strips. In an exper-
imental study consisting of nine tests, the effects of unidirectional
and bidirectional CFRP layouts, anchored and fully wrapped
systems, retrofitting uncracked and precracked sections, and
loading conditions on the shear performance of large girders with
relatively wide webs were investigated.
Test results indicated that CFRP anchors were able to develop
the fracture strength of the CFRP strips despite some anchors being
placed in flexural tension zones. The unidirectional CFRP layouts
were able to improve the shear capacity by as much as 56% while
the bidirectional layouts improved the inclined cracking shear load
by as much as 22%. Moreover, the measured lower-bound average
CFRP fracture strain was found to be significantly larger (0.007)
than the maximum permitted effective strain (0.004) currently
recommended for design.
Keywords: carbon-fiber reinforced polymer (CFRP); CFRP anchors; rein-
forced concrete (RC); shear behavior; shear strengthening.
INTRODUCTION
Many structures in the United States are reaching their
expected service life due to increasing load demands, occu-
pancy requirements, exposure to corrosive environments,
and sometimes-inadequate structural designs. The costs
associated with replacing a large element in a bridge or a
building are often too high to be a feasible option. Effec-
tive and efficient rehabilitation techniques offer an attractive
solution for rehabilitating aging or inadequate elements if
the cost of the installation and disruption to the use of the
structure can be minimized.
Carbon fiber-reinforced polymer (CFRP) is an ideal mate-
rial for rehabilitating existing members due to its high tensile
strength-to-weight ratio, adaptability to a variety of member
shapes, and rapid installation that minimizes disruption to the
use of the structure. However, without proper anchorage, the
interfacial bond between the CFRP and the surface concrete
is lost at CFRP stresses that are less than half of the CFRP
fracture stress.1-4 Consequently, debonding failures limit the
efficiency and shear contribution of the CFRP material.
Rather than rely solely on the interfacial bond strength,
anchorage for CFRP strips has been studied for cases where
the CFRP strips cannot be fully wrapped due to obstruc-
tions (that is, a bent cap girder supporting I-beams or beams
that are cast monolithically with floor slabs). Much of the
research conducted on anchorage systems was based on
the use of mechanical anchors that clamp the CFRP strips
in place.5-7 Clamping anchorage systems are susceptible
ACI Structural Journal/January 2018
TECHNICAL PAPER
Fig. 1—Shear crack in bent cap girder. (Courtesy of the
Texas Department of Transportation.)
to buckling when a continuous anchorage plate is placed
within a member’s compression region and slippage of
the CFRP strips due to inadequate clamping forces. More
recently, research on the use of CFRP spike anchors, hence-
forth referred to as CFRP anchors, has begun to grow due
to the ability of properly detailed anchors to develop the
full tensile capacity of the CFRP strips and to their ease
of installation.8-14
Many of the experimental studies related to CFRP shear
strengthening have been conducted on small-scale, thin-
webbed beams that do not meet minimum transverse steel
reinforcement requirements.2,3,11 Hence, the observed shear
strength gains of these members likely do not represent the
shear strength gains of large in-place members. Therefore,
the feasibility of improving the shear capacity of large-scale
relatively wide-webbed reinforced concrete (RC) bent cap
girders using CFRP strips and CFRP anchors was studied
in this project. Shear distress in a bridge bent cap girder is
shown in Fig. 1. The effects of the following variables on the
shear performance of the square 32 in. (813 mm) deep and
wide bent cap girders were studied: 1) unidirectional and
bidirectional CFRP layouts; 2) anchored and fully wrapped
systems; 3) retrofitting uncracked and precracked sections;
and 4) loading conditions. The effects of additional variables
have been studied in other parts of the experimental program
and are discussed in Kim et al.10 and Jirsa et al.14
ACI Structural Journal, V. 115, No. 1, January 2018.
MS No. S-2017-093, doi: 10.14359/51701092, was received March 14, 2017, and
reviewed under Institute publication policies. Copyright © 2018, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including author’s
closure, if any, will be published ten months from this journal’s date if the discussion
is received within four months of the paper’s print publication.
281
Fig. 2—Cross sections of bent cap girder. (Note: 1 in. = 25.4 mm.)
Fig. 3—Description of tests and nomenclature.
RESEARCH SIGNIFICANCE
Reinforced concrete members may experience shear defi-
ciencies due to increasing load demands, inadequate struc-
tural designs, or inadvertent damage. In most cases, it is
more economical to strengthen rather than replace members.
Externally bonded CFRP strips present an ideal approach
for strengthening existing members due to their high tensile
strength-to-weight ratio and expedited installation times.
However, the efficiency of the CFRP layout depends on the
anchorage system. In this study, CFRP anchors were used to
develop the strength of the U-wrap (or discontinuous) CFRP
strips. Substantial shear strength gains were observed for the
large-scale bent cap girders using anchored CFRP strips.
TEST CONFIGURATION
Specimen design
To meet the constraints of the test facilities, the test spec-
imens had a 32 in. (813 mm) deep by 32 in. (813 mm) wide
cross section and were 27 ft 8 in. (8433 mm) long.
The loading and support conditions of in-place bent
cap girders generally result in double, or reverse, curva-
ture. Hence, the members will have positive and negative
moments along the shear span. Seven of the test specimens
were subjected to loading that produced double curvature,
and two tests were conducted using loading that resulted in
single curvature. Thus, the experimental results from the
double curvature tests could eventually be compared with
results from other experimental studies.10,14
The proportion of the longitudinal reinforcement varied
between the single and double curvature specimens due
to differences in the magnitude of the maximum applied
moments. Specifically, the maximum moment in the double
curvature specimen was approximately half of the maximum
single curvature moment for the same applied shear force
in the test span. Consequently, the single curvature spec-
imen had a longitudinal reinforcement ratio (ρ) of 3.6% for
an effective depth (d) of 27 in. (686 mm) while the double
282
curvature specimen had a longitudinal reinforcement ratio
(ρ) of 1.9% for an effective depth (d) of 28 in. (711 mm).
As a result, the single and double curvature specimens had
similar strains in the longitudinal reinforcement at the same
applied shear load. The large longitudinal reinforcement
ratios were necessary to ensure that the shear capacity of the
strengthened girders did not exceed the applied shear load
that would cause flexural yielding.
Regardless of the test configuration, the spacing of the
transverse steel reinforcement in the test span was deter-
mined such that the specimens met the minimum trans-
verse reinforcement requirements in accordance with
ACI and AASHTO specifications.15,16 All the specimens
were designed to have span-depth ratios (av/d) equal to
3.0, where the shear span (av) was measured as the clear
distance between the loading plate and reaction bearing
pad. However, it should be noted that the span-depth ratio
for a concrete member loaded in double curvature is gener-
ally implied to be reduced due to the existence of a point of
inflection between the positive and negative moments.15
Typical cross sections are shown in Fig. 2. Note that
the outer bundled bars in the double curvature specimen
(Fig. 2(b)) were moved inward for the last three tests in
an attempt to mitigate bond-shear cracks caused by a high
strain gradient in the compression reinforcement, as will be
discussed later.
Nomenclature
A nomenclature was developed to help define the param-
eters in each test. A graphical representation of the nomen-
clature can be seen in Fig. 3. The first term indicates how the
girder was loaded. The second term indicates if the girder
was previously tested and, thus, precracked. The third and
fourth group of terms indicate whether vertical or horizontal
strips were used and if so, how the CFRP strips were devel-
oped. The nomenclature and a short description of each test
is provided in Table 1.
ACI Structural Journal/January 2018
Fig. 4—CFRP layouts. (Note: 1 in. = 25.4 mm.)

Table 1—Specimen description CFRP anchors

A significant amount of research complementary to this
Nomenclature Description
project has been conducted on the optimization of CFRP
S-U-VN-HN Control anchors.8-10,14,17,18 In particular, Jirsa et al.14 proposed
S-C-VA-HN Anchored unidirectional, cracked comprehensive CFRP anchor design recommendations that
D-U-VN-HN Control determine the required amount of CFRP anchor material,
anchor hole diameters, chamfer bend radii, anchor fan bond
D-C-VF-HN Fully wrapped unidirectional, cracked
lengths, and anchor embedment depths for user fabricated
D-U-VF-HN Fully wrapped unidirectional, uncracked and premanufactured CFRP anchors.
Fully wrapped vertically and anchored horizontally, For all tests, 10 in. (254 mm) long premanufactured CFRP
D-U-VF-HA
uncracked anchors were used. The CFRP anchors on the boundary of
D-U-VN-HN*† Redundant control the vertical and horizontal CFRP strips provided an anchor-
D-C-VA-HN †
Anchored unidirectional, cracked
to-strip cross-sectional area ratio of 2.9 while the interme-
diate horizontal strip anchors provided a ratio of 2.4 after
D-U-VA-HA† Anchored bidirectional, uncracked
the intermediate anchors had been divided into two fans
*
Redundant double curvature control test. that were splayed in opposing directions over the strips,
†
Modified cross section and test setup. as shown in Fig. 4(c) and 4(d). Generally, providing an
anchor-to-strip cross-sectional area ratio of 2.0 was found
CFRP layouts to develop the tensile strength of the CFRP strips.10,14 Note
Four CFRP layouts were used to strengthen the bent cap that the intermediate horizontal anchors were spaced at a
girders in shear, as shown in Fig. 4. The CFRP layouts were distance approximately equal to the effective depth of the
designed such that each layout had the same cross-sectional section to minimize the elongation of the horizontal CFRP
area of CFRP crossing an assumed 45-degree shear crack, strips, which in turn was likely to reduce shear crack widths.
that is, half of the vertical CFRP strips in the unidirectional To simplify the anchor installation procedure used by Kim et
layouts were removed and placed in the horizontal direction al.,12 the vertical and horizontal CFRP strips were discontinued
for the bidirectional layouts. just short of the 9/16 in. (14 mm) diameter and 4 in. (102 mm)
All the CFRP strips in this experimental study had a deep boundary anchor holes, as shown in Fig. 5(a). Conversely,
nominal 3 in. (76 mm) width. Each CFRP strip location was the horizontal strips were placed over the 3/4 in. (19 mm) diam-
composed of two individual CFRP strips spliced together at eter and 4 in. (102 mm) deep intermediate anchor holes similar
the top and/or bottom of the girder. For instance, the anchored to Kim et al.,12 making insertion of the anchors through the
U-wrap (or discontinuous) CFRP strips were spliced only on strip troublesome. After installing the CFRP anchors, two
the bottom of the girders whereas the fully wrapped CFRP patches were laid over the holes and anchors in a perpendic-
strips were spliced on the top and bottom of the girders. As ular and parallel fiber direction to mitigate stresses within
a result, it was easier to ensure that air entrapped under the the anchors. The boundary and intermediate CFRP anchor
strip was removed. All the CFRP strip splices had at least an dimensions are shown in Fig. 5.
8 in. (203 mm) overlap length, which provided a bond stress By using the double curvature loading conditions, repre-
less than the CFRP manufacturer’s maximum allowable sentative of many in-place bent cap girder conditions, the
bond stress. Moreover, all sharp edges (corners and anchor performance of the CFRP anchors in a tension region could
holes) were rounded to a bend radius of 0.5 in. (13 mm) to be investigated because approximately half of the CFRP
mitigate stress concentrations in the CFRP. anchors were placed within a flexural tension zone. Many
of the CFRP anchors were installed through shear or bond-
shear cracks in the precracked specimens.

ACI Structural Journal/January 2018 283


Fig. 5—CFRP anchor detail. (Note: 1 in. = 25.4 mm.)
Material properties
A ready mixed concrete supplier provided a concrete
mixture with a test day strength of approximately 4500 psi
(31 MPa). The longitudinal reinforcement was ASTM A615
Grade 75 No. 11 reinforcing bars whereas the transverse
reinforcement was ASTM A615 Grade 60 No. 5 reinforcing
bars. The average measured yield strength of the transverse
reinforcement was 61 ksi (421 MPa). The yield strength of
the longitudinal reinforcement was not measured because
the reinforcement did not yield. The CFRP laminate used
in this study had unidirectional fibers and exhibited a linear-
elastic behavior until fracture. The CFRP laminate proper-
ties (parallel to the fiber direction), as reported by the CFRP
manufacturer, are shown in Table 2.
Transverse reinforcement comparison
Traditionally, the total amount of transverse reinforce-
ment is reported as a reinforcement ratio. However, a trans-
verse reinforcement ratio is meaningless because the steel
and CFRP reinforcement have different material properties
and stress-strain relationships. Instead, the vertical shear
stress contribution can be used as a surrogate for the trans-
verse reinforcement ratio because the material stress is
taken into account.
Because CFRP anchors were provided, the U-wrap CFRP
layouts were assumed to be equivalent to fully wrapped
CFRP layouts. Consequently, the effective stress of both
CFRP layouts was assumed, at the specimen design stage,
to be 40% of the CFRP fracture stress as proposed by ACI
440.2R-081 for fully wrapped layouts.
Using this approach, the control specimens had a vertical
shear stress contribution of 66 psi (0.46 MPa) or 1.0√fc′ (in psi
units), while the unidirectional specimens had a contribution
from both steel and CFRP of 100 psi (0.69 MPa) or 1.5√fc′
284
Table 2—CFRP laminate properties reported by
manufacturer
Typical test value Design value
Tensile elastic modulus 13,900 ksi 11,900 ksi
Tensile ultimate strength 143 ksi 121 ksi
Ultimate elongation at fracture 1% 0.85%
Thickness 0.02 in. 0.02 in.
Notes: 1 ksi = 6.9 MPa; 1 in. = 25.4 mm.
(in psi units) and the bidirectional specimens had a vertical
contribution of 83 psi (0.57 MPa) or 1.2√fc′ (in psi units). As a
point of reference, ACI 318-1415 limits the shear contribution
of the transverse reinforcement to 8√fc′ (in psi units).
Test setup
In the single curvature test setup shown in Fig. 6(a), the
applied shear in the test span was measured directly at the
south support using two load cells. In the double curvature
test setup shown in Fig. 6(b), an additional loading point
produced positive and negative moments in the test span.
Approximately equal maximum moments at both ends of the
test span were achieved by using a load maintainer, which
produced a load at the north ram approximately twice that at
the south ram. The shear in the test span was calculated as
the south reaction minus the load applied by the south ram.
Larger loading plates and reaction bearing pads were
used in the double curvature test setup compared to the
single curvature test setup to prevent concrete crushing of
the direct struts and nodes in the transfer spans outside of
the test span, which was between the loading and reaction
points. The large loading plates and reaction bearing pads
are representative of typical dimensions of prestressed
I-beams that rest on top of bent cap girders and columns that
support the girders.
The number of load cells used in the south reaction of the
double curvature test setup was changed from four to two so
that the rotation of the member could be better accommodated.
At the same time, the cross-sectional detail of the double
curvature specimens was altered as shown in Fig. 2(b). The
changes to the cross section of the specimens and the south
reaction did not affect the shear capacity of the specimens.
Load was applied to the control and strengthened speci-
mens in stages until a loss of shear strength (that is, shear
failure) was experienced or the load-deflection curve indi-
cated an increase in deflection with no change in the load.
Upon reaching the peak shear load, the control specimens
were immediately unloaded. The control specimens were
then strengthened. The behavior and performance of the bent
cap girders were monitored using strain gauges on the longi-
tudinal and transverse reinforcement, linear potentiometers
at the locations of the rams and reactions, and an optical
measurement system to measure three-dimensional (3-D)
displacements and element strains on the surface of the
concrete and CFRP strips. The determination of the surface
displacements and element surface strains is discussed
in-depth by Sokoli et al.19 and Shekarchi.20
ACI Structural Journal/January 2018
Fig. 6—Test configuration. (Note: 1 ft = 304.8 mm; 1 in. = 25.4 mm.)

Table 3—Summary of test results

Average concrete Cracking shear, Cracking shear Peak shear
Test name strength, ksi kip Peak shear, kip increase, % increase, % Crack angle, deg Failure mode*
S-U-VN-HN 4.71 165 295 — — 25 DT
S-C-VA-HN 4.71 — 365 — 24% 25 SF
D-U-VN-HN 4.40 148 293 — — 34 DT
D-C-VF-HN 4.40 — 424 — 45% 34 SF
D-U-VF-HN 4.41 161 456 9% 56% 31 SF
D-U-VF-HA 5.11 180 446 22% 52% 40 SF
D-U-VN-HN †
4.56 149 313 — — 31 DT
D-C-VA-HN 4.56 — 427 — 37% 31 SF
D-U-VA-HA 4.58 177 367 19% 17% 26 SF
*
Failure mode: DT is diagonal tension, and SF is CFRP strip fracture.
†
Redundant double curvature control test.
Notes: 1 ksi = 6.9 MPa; 1 kip = 4.45 kN, 1 deg = 0.017 rad.

EXPERIMENTAL RESULTS AND DISCUSSION
As predicted, all the specimens failed in shear. In all the
strengthened specimens, the CFRP strips that crossed the
critical shear crack fractured when the peak shear capacity
was reached. Such failures represent the optimal failure
mode because the full strength of the CFRP strips was used.
Fracture of the CFRP strips demonstrated the viability of
using CFRP anchors. Shear strength gains up to 56% were
observed for the strengthened specimens relative to compa-
rable control specimens, as shown in Table 3.
In the subsequent discussion, an overview of the shear
behavior of a typical uncracked member strengthened using
an anchored CFRP layout is presented. The behavior of an
uncracked and cracked member is discussed. Finally, the
effects of unidirectional and bidirectional CFRP layouts,
anchored and fully wrapped strips, strengthening of
uncracked and cracked sections, and loading conditions on
the shear performance of the bent cap girders are discussed.

ACI Structural Journal/January 2018 285

Fig. 7—Single curvature applied shear-displacement
response. (Note: 1 kip = 4.45 kN; 1 in. = 25.4 mm.)
Overview of shear behavior of strengthened
specimens
The following overview is based on observations herein
and available literature.10,20-24 Prior to the formation of a
shear crack, the concrete resists the applied shear force.
Immediately after a shear crack forms, a fraction of the
applied shear force is distributed to the transverse steel and
CFRP reinforcement. The force in the CFRP reinforcement
causes the strips to debond locally across the shear crack. As
the member is loaded further, the shear crack grows and the
CFRP strains increase causing the debonding of the strips
to spread outward from the shear crack toward the points of
anchorage. As the shear crack widens, the beneficial effects
of aggregate interlock are reduced. Once the transverse steel
reinforcement crossing the shear crack yields, any additional
applied shear force is resisted by the concrete and CFRP,
causing the rate of CFRP debonding to accelerate until the
Fig. 8—Double curvature applied shear-displacement
strips have fully debonded between the points of anchorage.
response. (Note: 1 kip = 4.45 kN; 1 in. = 25.4 mm.)
Hence, the anchors need to resist and transfer the strip forces
into the concrete. Adequately designed anchorage systems control specimens had lower cracking shear loads than the
enable the CFRP strips to fracture at the peak shear capacity single curvature control specimen even though the speci-
of the member, which is immediately followed by the loss of mens were designed similarly (Table 3). The cracking load
aggregate interlock. of the bidirectionally strengthened double curvature speci-
The only difference in behavior between an uncracked and mens was improved by as much as 22%.
cracked member is that the CFRP strains, and thus debonding, The addition of the transverse CFRP reinforcement
begin to increase immediately as the shear force is applied. improved the shear stiffness of the bent cap girder speci-
The degradation of bond between the concrete and CFRP mens after the formation of the shear crack. For instance,
strips in the uncracked and cracked members implies that the strengthened precracked specimen (S-C-VA-HN) had a
a strain compatibility between the two materials does not larger initial stiffness compared to the post-cracking shear
exist. However, displacement compatibility may be main- stiffness of the control specimen (S-U-VN-HN). A similar
tained after the anchored CFRP strips have fully debonded. trend was expected in the double curvature specimens but
bond-shear cracks formed along the compression flex-
Applied shear-displacement response ural reinforcement and reduced the stiffness of the control
The applied shear-displacement response of the bent cap specimens near the peak shear capacity. Note in Fig. 8 that
girder specimens offers insight into cracking loads, relative the strengthened double curvature specimens exhibited a
shear stiffness, and ultimate capacities. The relationship substantial change in stiffness when the transverse steel rein-
between the applied shear and displacement under the point forcement yielded.
load in the test span for the single and double curvature The strengthened precracked single curvature specimen
specimens are shown in Fig. 7 and 8, and the test results are (S-C-VA-HN) exhibited a 24% increase in the peak shear
summarized in Table 3. capacity, whereas the strengthened double curvature spec-
In this study, the inclined cracking shear load was defined imens exhibited up to a 56% increase in the peak shear
as the shear load that caused the shear-carrying mechanism capacity. The disparity in strength gains between the single
to change (that is, the steel transverse reinforcement began and double curvature specimens was likely the result of
to contribute to the shear resistance). The double curvature some damage due to loading the control single curvature

286 ACI Structural Journal/January 2018

Fig. 9—Double curvature midheight shear crack width
comparison. (Note: 1 kip = 4.45 kN; 1 in. = 25.4 mm.)
specimen (S-U-VN-HN) beyond the peak shear capacity (as
shown in Fig. 7), which was then strengthened and tested as
S-C-VA-HN.
Unidirectional and bidirectional CFRP layouts
The unidirectional layout provided a marginally larger
strength gain compared to the bidirectional layout even
though both layouts used a similar amount of CFRP on
the surface of the web. The better material use efficiency
of the unidirectional layout compared to the bidirectional
layout was likely due to the shallow shear crack angles
at the ultimate shear capacity, which were consistently
less than 45 degrees (Table 3). Consequently, the vertical
CFRP strips were able to develop fracture strength before
the horizontal strips. However, it should be mentioned that
the shear capacity of the anchored bidirectional specimen
(D-U-VA-HA) was lower than expected due to observed
undesirable strain concentrations between overlapping and
adjacent CFRP anchors, as shown in Fig. 4(d), which frac-
tured the horizontal strips prior to the vertical strips.
Using the optical measurement system, the width of the
principal shear crack (wc) was estimated to determine if the
crack width influenced the performance of the unidirec-
tional and bidirectional layouts. The width of the principal
shear crack at midheight of the section was used as a means
of consistently comparing crack widths between tests.
However, the midheight and maximum shear crack widths
had similar trends.
The width of the midheight shear crack was estimated by
subtracting half of the element cracking strain (εcr) from the
element principal tensile strain (ε1), which was then multi-
plied by the surface elements’ effective length perpendicular
to the crack (Leffective), as shown in Eq. (1). An in-depth discus-
sion on crack width estimates can be found in Shekarchi20
wc = (ε1 – εcr/2) × Leffective (1)
The midheight shear crack widths for the uncracked double
curvature tests are plotted in Fig. 9. The estimated crack
widths were compared to hand-measured crack widths at
various applied shear loads and locations to validate the
calculations. It was found that the bidirectional layout used
on D-U-VF-HA controlled the width of the shear crack
better at the ultimate shear capacity than the unidirectional
layout used on D-U-VF-HN. The shear crack was restrained
ACI Structural Journal/January 2018
Fig. 10—Uniform strain distribution. (Note: 1 in. = 25.4 mm.)
in the vertical and horizontal directions with the bidirec-
tional layout. The increased shear crack restraint may have
improved the concrete shear contribution.
Anchored and fully wrapped systems
Specimen D-C-VA-HN (with anchored vertical strips)
and D-C-VF-HN (with fully wrapped vertical strips) exhib-
ited similar shear strength capacities even though some of
the CFRP anchors were placed through preexisting cracks.
The relative efficiency of the anchored and fully wrapped
systems can be determined by investigating the strain distri-
bution within the CFRP strips and the average strains in the
CFRP strips across the critical shear crack at the ultimate
shear capacity.
One advantage of using CFRP anchors is that they bond well
with the CFRP strips, which helps minimize strain concentra-
tions at the interface of the anchor fan and strip. The limited
strain concentrations enabled the CFRP strips to exhibit a
uniform strain distribution along the height and across the
width of the 3 in. (76 mm) wide CFRP strips as shown in
Fig. 10 for Strip 8 of S-C-VA-HN. However, research14 has
shown that wider CFRP strips with one anchor exhibited a less
uniform strain distribution across the width of the strip. As a
result, the average strains in the CFRP strip across the critical
crack tend to be lower at the ultimate shear capacity.14,18
For design, a lower-bound average vertical CFRP strain
(that is, effective CFRP strain) is suggested for estimating
the CFRP shear contribution. The average CFRP strain is
287
Table 4—Average and minimum vertical CFRP
strains
Vertical CFRP strain
Test name Average Minimum
S-C-VA-HN 0.0093 0.0072
D-C-VF-HN 0.0079 0.0060
D-U-VF-HN 0.0107 0.0086
D-U-VF-HA 0.0106 0.0096
D-C-VA-HN 0.0076 0.0038
D-U-VA-HA 0.0077 0.0036

determined by averaging the vertical strains in all the CFRP

strips that cross the critical shear crack at the ultimate shear
capacity of the member. The measured average CFRP strip
strains, as well as the minimum CFRP strip strains, at first
strip fracture for all tests are shown in Table 4. The data indi-
cate that the CFRP anchors enabled the CFRP strips to reach
a nearly identical average CFRP strain as the fully wrapped
layout. Specifically, average CFRP strains for D-C-VA-HN
and D-C-VF-HN at first strip fracture reached 0.0076 and
0.0079, respectively.

Retrofitting uncracked and precracked sections

With a preexisting shear crack, the steel reinforcement and
the CFRP strips contributed to the shear capacity at the onset
of loading, as shown in Fig. 11(a) for D-C-VF-HN. In the
uncracked specimen (D-U-VF-HN), the steel contribution
did not increase until the specimen cracked and the CFRP
did not contribute significantly until the transverse steel
reinforcement yielded (Fig. 11(b)). Specimens D-C-VF-HN
(cracked) and D-U-VF-HN (uncracked), which were nomi-
nally identical tests except for precracking, had comparable
ultimate shear capacities.
However, the precracked member did not reach the same
average CFRP strain at failure as its uncracked companion.
In particular, D-C-VF-HN reached an average CFRP strain
of 0.0079 while D-U-VF-HN reached an average strain of
0.0107 at the ultimate shear capacity, which suggests that
the average CFRP strain across the critical crack may be
adversely affected by the width of the shear crack prior to the
CFRP installation. The precracked specimens had maximum
crack widths of approximately 0.03 in. (0.75 mm) when the
CFRP was installed.
For design, the differences in the average CFRP strains
can be accounted for by using a lower-bound effective CFRP
strain. Based on Table 4, a reasonable lower-bound effective
strain in the CFRP strips in cracked or uncracked members
at the ultimate shear capacity can be taken as 70% of the
CFRP laminate’s fracture strain (that is, 0.007 for the CFRP
used in this experimental study). The suggested average
strain is considerably larger than the ACI 440.2R-081
maximum recommended effective CFRP strain of 0.004 for
fully wrapped layouts. It is noted that the suggested lower-
bound effective CFRP strain was determined for specimens
that had a maximum coarse limestone aggregate size of 1 in.
(25 mm), which may influence aggregate interlock prior to
fracturing the CFRP strips.1,24
Fig. 11—Typical CFRP strain response. (Note: 1 kip =
4.45 kN.)
Loading conditions
The single and double curvature specimens had a span-
depth ratio (av/d) of 3.0, although the double curvature speci-
mens had an effective span-depth ratio (M/Vd) of 1.5 because
there was a point of inflection at the middle of the test span.
The effective span-depth ratio implies that the double curva-
ture specimens should have exhibited deep beam behavior.
However, the ultimate shear capacities of the single and double
curvature control specimens were nearly the same. Moreover,
the strengthened double curvature specimens experienced up
to a 56% increase in shear capacity using CFRP strips and
CFRP anchors despite past studies conducted by Kim et al.10
and Jirsa et al.14 showing that anchored CFRP strips cannot
effectively confine concrete struts or nodal zones. Had the
double curvature specimens exhibited deep beam behavior,
the anchored CFRP layouts would not have increased the
shear capacity of the members significantly.
Based on the similar control capacities, the significant
strength gains with CFRP strengthening, and the fact that
the double curvature tests did not form a direct compression
strut from the applied load to the reaction, it can be concluded
that the double curvature specimens exhibited sectional shear
behavior. Consequently, the material shear contributions from
the single and double curvature specimens can be directly
compared to identify the effects the loading conditions had on
the shear behavior of the bent cap girders.
The shear contributions of the concrete, steel stirrups, and
CFRP strips were determined from the measured transverse

288 ACI Structural Journal/January 2018

Fig. 12—Normalized shear contributions, or shear stress coefficients, at cracking and ultimate shear capacity.
steel and CFRP reinforcement strains, constitutive relations,
and vertical equilibrium across the critical shear crack. Using
this approach, the ability to estimate the response of the
material shear contributions depends on the strain measure-
ments close to the shear crack. For instance, the shear contri-
bution of the steel reinforcement may be underestimated
prior to the stirrups yielding if the strain gauges used to esti-
mate the internal steel forces were not at or near the critical
shear crack. However, given the elastic-plastic constitutive
relation of the steel, once the strain gauges recorded strains
beyond yield, their force became constant and equal to the
yield strength. Unlike the shear contribution of the steel, the
CFRP strains were more accurately measured at the crit-
ical crack location using the optical measurement system.
However, the vertical CFRP strip strains, and thus the shear
contribution of the vertical CFRP strips, may have been
marginally underestimated before debonding of the CFRP
strips spread from the crack over the selected gauge length
of the target elements. Nevertheless, once the CFRP strips
debonded beyond the tracked elements, the gauge length of
the elements did not affect the measured strains. In fact, the
CFRP strips debonded rapidly and fully after the transverse
steel reinforcement yielded (that is, at strains around 0.002),
which is a relatively low CFRP strain. The shear contribu-
tion of the transverse CFRP reinforcement was calculated
assuming a uniform strain distribution across the width of
the CFRP strips and using an elastic constitutive relation-
Fig. 13—Comparison of single and double curvature control
ship. The vertical component of the concrete shear contri-
tests. (Note: 1 kip = 4.45 kN.)
bution was estimated from vertical equilibrium, which indi-
cates that the concrete carried the remainder of the applied The shear stress coefficients shown in Fig. 12 suggest that
shear force after the steel and CFRP contributions were the double curvature control tests had higher concrete contri-
subtracted from the applied shear. A summary of the normal- butions but lower steel contributions when compared to the
ized material shear contributions at inclined cracking and the single curvature control test (S-U-VN-HN), even though all
measured ultimate shear capacity (VEXP) is shown in Fig. 12. the tests had similar ultimate shear capacities. This apparent
The effects of the loading conditions were investigated by load shift may be attributed to the fact that the single curva-
studying the control specimens (no CFRP strips) to minimize ture control test had a much shallower shear crack angle
the number of unknown variables. The variation in the shear (25 degrees from horizontal) compared to the double curva-
contribution responses, shown in Fig. 13, can be attributed ture control tests (34 and 31 degrees), which caused the
to differences in the flexural cross-sectional detailing and the critical shear crack in the single curvature tests to cross one
loading conditions. additional stirrup.

ACI Structural Journal/January 2018 289

 Similar trends were observed for the strengthened single
and double curvature specimens, as shown in Fig. 12. On
average, the strengthened double curvature specimens had a
63% higher concrete contribution compared to the strength-
ened single curvature specimen.
Dowel action caused by the cross-sectional detailing of
the flexural reinforcement can artificially inflate the concrete
contribution because the effects are lumped into a single
contribution. However, the single curvature specimen had
50% more overall flexural reinforcement compared to the
double curvature specimen and yet the single curvature
concrete contribution was lower. Because the transverse
reinforcement detailing was consistent between the single
and double curvature specimens and the specimens with
the least amount of flexural reinforcement experienced the
largest concrete contributions, it is likely that the lower
effective span-depth ratio of the double curvature specimens
improved the concrete contribution.
CONCLUSIONS
The objective of the bent cap girder experimental program
was to investigate the feasibility of strengthening relatively
wide-webbed reinforced concrete members in shear using
CFRP strips and CFRP anchors. In particular, the experi-
mental program was intended to provide insight into the effi-
ciency of anchored CFRP layouts relative to fully wrapped
layouts in uncracked and cracked reinforced concrete
members subjected to typical loading conditions.
The test results showed that relatively wide-webbed rein-
forced concrete bent cap girders can be strengthened in shear
effectively using anchored and fully wrapped CFRP layouts.
The following conclusions were based on findings from the
experimental study.
1. The anchored CFRP strips were able to develop the full
tensile strength of the CFRP strips. In fact, anchored CFRP
layouts were found to be as efficient as fully wrapped layouts
even when the anchors were placed in regions of flexural
tension. However, undesirable strain concentrations may
develop when anchors are overlapped or close to each other.
2. To engage the CFRP anchors fully, the CFRP strip must
fully debond. Hence, the strain compatibility between the
concrete and the CFRP strips cannot be maintained.
3. The post-cracking stiffness of the members prior to
strengthening increased after strengthening the members
with CFRP layouts.
4. Unidirectional layouts were found to provide larger
shear strength gains compared to bidirectional layouts
that had the same amount of CFRP surface area on the
web surface. However, the bidirectional layouts exhibited
increased cracking shear loads and reduced shear crack
widths compared to the unidirectional layouts. Specifically,
the unidirectional CFRP layouts were able to increase the
control test shear capacity by as much as 56% while the bidi-
rectional layouts increased the control test cracking shear
load by as much as 22%.
5. Strengthening precracked members in shear using CFRP
layouts was nearly as effective as strengthening uncracked
members. The presence of shear cracks prior to the CFRP
290
installation resulted in lower average CFRP strains across
the shear crack.
6. At failure, the measured lower-bound average (that is,
effective) CFRP strain was approximately 0.007 consid-
ering both cracked and uncracked specimens; a strain that
is significantly larger than the maximum permitted effective
CFRP strain in the ACI 440.2R-081 design guidelines (that
is, 0.004). Additionally, the measured effective CFRP strain
is considerably larger than the effective CFRP strain used in
ACI 440.2R-08 for bond-critical CFRP layouts, which illus-
trates that the CFRP material is more effectively used when
the strips are properly anchored. It should be noted that the
suggested effective strain was determined from data on beams
with a maximum coarse limestone aggregate size of 1 in.
(25 mm) and may be different for other aggregate types
or sizes.
7. The single and double curvature girders had similar ulti-
mate shear capacities. However, after investigating the mate-
rial shear contributions, it was observed that the internal force
distributions of the single and double curvature specimens
were different. The observed variances in the load distribution
were the result of differences in the shear crack angles and
the effective span-depth ratios. Moreover, the double curva-
ture loading was found to result in lower cracking shear loads
compared to the single curvature loading.
AUTHOR BIOS
ACI member William A. Shekarchi received his PhD from the University
of Texas at Austin, Austin, TX. His research interests include shear behavior
of reinforced concrete elements, structural rehabilitation, and the behavior
of fiber-reinforced polymers.
ACI member Wassim M. Ghannoum is an Associate Professor at the
University of Texas at San Antonio, San Antonio, TX. He is the Chair of ACI
Committee 369, Seismic Repair and Rehabilitation, and a member of ACI
Subcommittees 318-R, High-Strength Reinforcement (Structural Concrete
Building Code), and 440-F, FRP-Repair-Strengthening (Fiber-Reinforced
Polymer Reinforcement). His research interests include extending the life
span and increasing the resilience to damage of concrete structures through
the application of novel structural materials and retrofit techniques.
ACI Honorary Member James O. Jirsa holds the Janet S. Cockrell Centen-
nial Chair Emeritus at the University of Texas at Austin. He is an ACI
Past President and a member of ACI Committee 318, Structural Concrete
Building Code.
ACKNOWLEDGMENTS
The support of the Texas Department of Transportation for Project 0-6783
is appreciatively acknowledged. The contents of this paper reflect the views
of the authors, who are responsible for the facts and the accuracy of the data
presented herein. The contents do not necessarily reflect the official view
or policies of the Texas Department of Transportation. This paper does not
constitute a standard, specification, or regulation. The authors express their
thanks to the students, faculty, and staff at Ferguson Structural Engineering
Laboratory for their assistance with the project.
NOTATION
av = length of constant shear span in test region
d = section’s effective depth taken from extreme compression fiber
to centroid of tension steel
fc′ = specified 28-day concrete strength
Leffective = elements’ effective length perpendicular to the crack
M = moment in shear span being considered
V = shear force in shear span being considered
wc = width of principal shear crack
ε1 = element principal tensile strain
εcr = element cracking strain
ACI Structural Journal/January 2018
ρ = ratio of tensile longitudinal reinforcement area to effective area
of concrete
REFERENCES
1. ACI Committee 440, “Guide for the Design and Construction of Exter-
nally Bonded FRP Systems for Strengthening Concrete Structures (ACI
440.2R-08),” American Concrete Institute, Farmington Hills, MI, 2008, 80 pp.
2. Khalifa, A., and Nanni, A., “Rehabilitation of Rectangular Simply
Supported RC Beams with Shear Deficiencies Using CFRP Composites,”
Construction and Building Materials, V. 16, No. 3, 2002, pp. 135-146. doi:
10.1016/S0950-0618(02)00002-8
3. Zhang, Z., and Hsu, C. T. T., “Shear Strengthening of Reinforced
Concrete Beams using Carbon-Fiber-Reinforced Polymer Laminates,”
Journal of Composites for Construction, ASCE, V. 9, No. 2, 2005,
pp. 158-169. doi: 10.1061/(ASCE)1090-0268(2005)9:2(158)
4. Pellegrino, C., and Modena, C., “Fiber-Reinforced Polymer Shear
Strengthening of Reinforced Concrete Beams: Experimental Study and
Analytical Modeling,” ACI Structural Journal, V. 103, No. 5, Sept.-Oct.
2006, p. 720-728.
5. Aridome, Y.; Kanakubo, T.; Furuta, T.; and Matsui, M., “Ductility of
T-Shape RC Beams Strengthened by CFRP Sheet U-anchor,” Transactions
of the Japan Concrete Institute, V. 20, 1998, pp. 117-124.
6. Ortega, C. A.; Belarbi, A.; and Bae, S. W., “End Anchorage of Exter-
nally Bonded FRP Sheets for the Case of Shear Strengthening of Concrete
Girders,” Proceedings of the 9th International Symposium on Fiber Rein-
forced Polymer Reinforcement for Concrete Structures (FRPRCS-9),
Thomas Telford, London, UK, 2009.
7. Deifalla, A., and Ghobarah, A., “Strengthening RC T-Beams Subjected
to Combined Torsion and Shear Using FRP Fabrics: Experimental Study,”
Journal of Composites for Construction, ASCE, V. 14, No. 3, 2010,
pp. 301-311. doi: 10.1061/(ASCE)CC.1943-5614.0000091
8. Orton, S. L.; Jirsa, J. O.; and Bayrak, O., “Design Consid-
erations of Carbon Fiber Anchors,” Journal of Composites for
Construction, ASCE, V. 12, No. 6, 2008, pp. 608-616. doi: 10.1061/
(ASCE)1090-0268(2008)12:6(608)
9. Huaco, G. D.; Jirsa, J. O.; and Bayrak, O., “Quality Control Test for
Carbon Fiber Reinforced Polymer (CFRP) Anchors for Rehabilitation,”
Tenth International Symposium on Fiber-Reinforced Polymer Reinforce-
ment for Concrete Structures, SP-275, American Concrete Institute, Farm-
ington Hills, MI, 2011, pp. 1-18.
10. Kim, Y.; Quinn, K. T.; Satrom, C. N.; Garcia, J.; Sun, W.; Ghan-
noum, W. M.; and Jirsa, J. O., “Shear Strengthening of Reinforced and
Prestressed Concrete Beams Using Carbon Fiber Reinforced Polymer
(CFRP) Sheets and Anchors,” FHWA/TX-12/0-6306-1, Center for Trans-
portation Research, Austin, TX, 2012, 325 pp.
11. Koutas, L., and Triantafillou, T. C., “Use of Anchors in Shear
Strengthening of Reinforced Concrete T-Beams with FRP,” Journal of
Composites for Construction, ASCE, V. 17, No. 1, 2013, pp. 101-107. doi:
10.1061/(ASCE)CC.1943-5614.0000316
ACI Structural Journal/January 2018
12. Kim, Y.; Quinn, K.; Ghannoum, W. M.; and Jirsa, J. O., “Strength-
ening of Reinforced Concrete T-Beams Using Anchored CFRP Mate-
rials,” ACI Structural Journal, V. 111, No. 5, Sept.-Oct. 2014, p. 1027 doi:
10.14359/51686805
13. Kim, Y.; Ghannoum, W. M.; and Jirsa, J. O., “Shear Behavior of Full-
Scale Reinforced Concrete T-Beams Strengthened using CFRP Strips and
Anchors,” Construction and Building Materials, V. 94, June 2015, pp. 1-9.
14. Jirsa, J. O.; Ghannoum, W. M.; Kim, C.; Sun, W.; Shekarchi,
W.; Alotaibi, N.; Pudleiner, D. K.; Zhu, J.; Liu, S.; and Wang, H., “Use
of Carbon Fiber Reinforced Polymer (CFRP) with CFRP Anchors for
Shear-Strengthening and Design Recommendations/Quality Control Proce-
dures for CFRP Anchors,” FHWA/TX-16/0-6783-R1D, Center for Trans-
portation Research, Austin, TX, 2016, 232 pp.
15. ACI Committee 318, “Building Code Requirements for Structural
Concrete (ACI 318-14) and Commentary (ACI 318R-14),” American
Concrete Institute, Farmington Hills, MI, 2014, 520 pp.
16. American Association of State Highway and Transportation Offi-
cials, “AASHTO LRFD Bridge Design Specifications,” seventh edition,
Washington, DC, 2014, 2150 pp.
17. Sun, W., and Ghannoum, W. M., “Modeling of Anchored CFRP
Strips Bonded to Concrete,” Construction and Building Materials, V. 85,
March 2015, pp. 144-156. doi: 10.1016/j.conbuildmat.2015.03.096
18. Sun, W.; Jirsa, J. O.; and Ghannoum, W. M., “Behavior of Anchored
Carbon Fiber-Reinforced Polymer Strips Used for Strengthening Concrete
Structures,” ACI Materials Journal, V. 113, No. 2, Mar.-Apr. 2016,
pp. 163-172. doi: 10.14359/51688637
19. Sokoli, D.; Shekarchi, W. A.; Buenrostro, E.; and Ghannoum, W. M.,
“Advancing Behavioral Understanding and Damage Evaluation of Concrete
Members Using High-Resolution Digital Image Correlation Data,” Earth-
quakes and Structures, V. 7, No. 5, 2014, pp. 609-626. doi: 10.12989/
eas.2014.7.5.609
20. Shekarchi, W. A., “Shear Behavior of Reinforced Concrete Bridge
Pile Cap Girders Strengthened with Carbon Fiber Reinforced Polymer
(CFRP) Strips and CFRP Anchors,” PhD dissertation, The University of
Texas at Austin, Austin, TX, 2016, 317 pp.
21. Triantafillou, T. C., and Antonopoulos, C. P., “Design of Concrete
Flexural Members Strengthened in Shear with FRP,” Journal of Compos-
ites for Construction, ASCE, V. 4, No. 4, 2000, pp. 198-205. doi: 10.1061/
(ASCE)1090-0268(2000)4:4(198)
22. Chen, J. F., and Teng, T. G., “Shear Capacity of FRP Strengthened
RC Beams: FRP Debonding,” Construction and Building Materials, V. 17,
No. 1, 2003, pp. 27-41. doi: 10.1016/S0950-0618(02)00091-0
23. Bae, S. W., and Belarbi, A., “Behavior of Various Anchorage Systems
Used for Shear Strengthening of Concrete Structures with Externally
Bonded FRP Sheets,” Journal of Bridge Engineering, ASCE, V. 18, No. 9,
2013, pp. 837-847. doi: 10.1061/(ASCE)BE.1943-5592.0000420
24. Vecchio, F. J., and Collins, M. P., “The Modified Compression-Field
Theory for Reinforced Concrete Elements Subjected to Shear,” ACI Journal
Proceedings, V. 83, No. 2, Mar.-Apr. 1986, pp. 219-231.
291
 NOTES:

292 ACI Structural Journal/January 2018

 NOTES:

ACI Structural Journal/January 2018 293

 NOTES:

294 ACI Structural Journal/January 2018

 NOTES:

ACI Structural Journal/January 2018 295

 NOTES:

296 ACI Structural Journal/January 2018

ACI
STRUCTURAL
J O U R N A L
J O U R N

The American Concrete Institute (ACI) is a leading authority and

resource worldwide for the development and distribution of
consensus-based standards and technical resources, educational
programs, and certifications for individuals and organizations involved
in concrete design, construction, and materials, who share
a commitment to pursuing the best use of concrete.

Individuals interested in the activities of ACI are encouraged to

explore the ACI website for membership opportunities, committee
activities, and a wide variety of concrete resources. As a volunteer
member-driven organization, ACI invites partnerships and welcomes
all concrete professionals who wish to be part of a respected,
connected, social group that provides an opportunity for professional
growth, networking, and enjoyment.