Aci Structural Journal Jan.-Feb. 2015 v. 112 No. 01

V. 112, NO.
1
JANUARY-FEBRUARY 2015
ACI
STRUCTURAL
J O U R N A L
A JOURNAL OF THE AMERICAN CONCRETE INSTITUTE
CONTENTS
Board of Direction
ACI Structural Journal
President
William E. Rushing Jr.
January-February 2015, V. 112, No. 1
Vice Presidents
Sharon L. Wood
Michael J. Schneider
Directors
Roger J. Becker
Dean A. Browning
Jeffrey W. Coleman
Alejandro Durn-Herrera
Robert J. Frosch
Augusto H. Holmberg
Cary S. Kopczynski
Steven H. Kosmatka
Kevin A. MacDonald
Fred Meyer
Michael M. Sprinkel
David M. Suchorski
Past President Board Members
Anne M. Ellis
James K. Wight
Kenneth C. Hover
Executive Vice President
Ron Burg
Technical Activities Committee

Ronald Janowiak, Chair
Daniel W. Falconer, Staff Liaison
JoAnn P. Browning
Catherine E. French
Fred R. Goodwin
Trey Hamilton
Neven Krstulovic-Opara
Kimberly Kurtis
Kevin A. MacDonald
Jan Olek
Michael Stenko
Pericles C. Stivaros
Andrew W. Taylor
Eldon G. Tipping
Staff
Executive Vice President
Ron Burg
a journal of the american concrete institute

an international technical society
3
A Global Integrity Parameter with Acoustic Emission for Load
Testing of Prestressed Concrete Girders, by Francisco A. Barrios and
Paul H. Ziehl
13 Analysis of Rectangular Sections Using Transformed Square Cross
Sections of Unit-Length Side, by Girma Zerayohannes
23
Evaluation of Post-Earthquake Axial Load Capacity of Circular
Bridge Columns, by Vesna Terzic and Bozidar Stojadinovic
35 Shear Behavior of Reinforced Concrete Columns with High-Strength
Steel and Concrete, by Yu-Chen Ou and Dimas P. Kurniawan
47 Three-Parameter Kinematic Theory for Shear Behavior of Continuous
Deep Beams, by Boyan I. Mihaylov, Bradley Hunt, Evan C. Bentz, and
Michael P. Collins
59 Investigation of Bond Properties of Alternate Anchorage Schemes for
Glass Fiber-Reinforced Polymer Bars, by Lisa Vint and ShamimSheikh
69
Stress-Transfer Behavior of Reinforced Concrete Cracks and
Interfaces, by Ali Reza Moradi, Masoud Soltani, and Abbas Ali Tasnimi
81 Condition Assessment of Prestressed Concrete Beams Using Cyclic
and Monotonic Load Tests, by Mohamed K. ElBatanouny, Antonio
Nanni, Paul H. Ziehl, and Fabio Matta
91 Crack Distribution in Fibrous Reinforced Concrete Tensile Prismatic
Bar, by Yuri S. Karinski, Avraham N. Dancygier, and Amnon Katz
Glass Fiber-Reinforced Polymer-Reinforced Circular Columns
103
under Simulated Seismic Loads, by Arjang Tavassoli, James Liu, and
ShamimSheikh
115 Discussion
Engineering
Managing Director
Daniel W. Falconer
Bond Strength of Spliced Fiber-Reinforced Polymer Reinforcement. Paper by

Managing Editor
Khaled Nahlawi

Behavior of Epoxy-Injected Diagonally Cracked Full-Scale Reinforced
Ali Cihan Pay, Erdem Canbay, and Robert J. Frosch

Concrete Girders. Paper by Matthew T. Smith, Daniel A. Howell, Mary Ann T.
Triska, and Christopher Higgins
Staff Engineers
Matthew R. Senecal
Gregory M. Zeisler
Jerzy Z. Zemajtis
Publishing Services
Manager
Barry M. Bergin
Editors
Carl R. Bischof
Tiesha Elam
Kaitlyn Hinman
Kelli R. Slayden
Discussion is welcomed for all materials published in this issue and will appear ten months from
this journals date if the discussion is received within four months of the papers print publication.
Discussion of material received after specified dates will be considered individually for publication or
private response. ACI Standards published in ACI Journals for public comment have discussion due
dates printed with the Standard.
Annual index published online at http://concrete.org/Publications/ACIStructuralJournal.
Copyright 2015 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: $166 per year (U.S. and possessions), $175 (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-February 2015
Contributions to
MEETINGS
2015
JANUARY
6-9Building Innovation 2015
Conference & Expo, Washington, DC,
www.nibs.org/?page=conference2015
11-15TRB 94th Annual Meeting,
Washington, DC, www.trb.org/
annualmeeting2015/annualmeeting2015.
aspx
20-22National Association of Home
Builders International Builders Show,
Las Vegas, NV, www.buildersshow.com/
Home
FEBRUARY
1-62015 Mason Contractors Association
of America Convention, Las Vegas, NV,
www.masoncontractors.org/convention/
index.php
2-3The International Concrete
Polishing and Staining Conference, Las
Vegas, NV, www.icpsc365.com/icpsc2015
2-6World of Concrete, Las Vegas, NV,
www.worldofconcrete.com
14-17Interlocking Concrete
Pavement Institute 2015 Annual
Meeting, San Antonio, TX, www.icpi.
org/2015annualmeeting
15-18Geosynthetics 2015, Portland, OR,
http://geosyntheticsconference.com
20-22National Concrete Masonry

Association Annual Convention, San
Antonio, TX, http://ncma.org/events
MARCH
1-32015 National Ready Mixed
Concrete Association Annual Convention,
Orlando, FL, www.nrmca.org/conferences_
events/annualconvention
4-8Precast/Prestressed Concrete
Institute Winter Conference, Orlando,
FL, www.pci.org/pci_events/pci_winter_
conference
5-7The Precast Show 2015, Orlando, FL,
http://precast.org/theprecastshow
17-21International Foundations
Congress & Equipment Exposition 2015,
San Antonio, TX, www.ifcee2015.com
25-27International Concrete Repair
Institute 2015 Spring Convention, New
York City, NY, www.icri.org/ebents/
upcomingevents.asp
MARCH/APRIL
30-2Concrete Sawing & Drilling
Association Convention and Tech Fair,
St. Petersburg, FL, www.csda.org/events/
event_details.asp?id=444478&group
APRIL
13-15BEST Conference Building
Enclosure Science & Technology,
Kansas City, MO, www.nibs.org/?page=best
18-192015 ICON-Xchange, San Antonio,

TX, http://iconxchange.org
UPCOMING ACI CONVENTIONS

The following is a list of scheduled ACI conventions:
2015April 12-16, Marriott & Kansas City Convention Center, Kansas City, MO
2015November 8-12, Sheraton Denver, Denver, CO
2016April 17-21, Hyatt & Wisconsin Center, Milwaukee, WI
For additional information, contact:
Event Services, ACI
38800 Country Club Drive, Farmington Hills, MI 48331
Telephone: +1.248.848.3795
e-mail: conventions@concrete.org
ON COVER: 112-S03, p. 27, Fig. 6Axial test setup.
Permission is granted by the American Concrete Institute for libraries and other users registered with the Copyright
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
$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.
The Institute is not responsible for statements or opinions expressed in its publications. Institute publications are not able
to, nor intend to, supplant individual training, responsibility, or judgment of the user, or the supplier, of the information
presented.
Papers appearing in the ACI Structural Journal are reviewed according to the Institutes Publication Policy by individuals
expert in the subject area of the papers.
2
The ACI Structural Journal is an open

forum on concrete technology and papers
related to this field are always welcome.
All material submitted for possible publication must meet the requirements of
the American Concrete Institute Publication Policy and Author Guidelines
and Submission Procedures. Prospective
authors should request a copy of the Policy
and Guidelines from ACI or visit ACIs
website at www.concrete.org prior to
submitting contributions.
Papers reporting research must include
a statement indicating the significance of
the research.
The Institute reserves the right to return,
without review, contributions not meeting
the requirements of the Publication Policy.
All materials conforming to the Policy
requirements will be reviewed for editorial
quality and technical content, and every
effort will be made to put all acceptable
papers into the information channel.
However, potentially good papers may be
returned to authors when it is not possible
to publish them in a reasonable time.
Discussion
All technical material appearing in the
ACI Structural Journal may be discussed.
If the deadline indicated on the contents
page is observed, discussion can appear
in the designated issue. Discussion should
be complete and ready for publication,
including finished, reproducible illustrations. Discussion must be confined to the
scope of the paper and meet the ACI Publication Policy.
Follow the style of the current issue.
Be brief1800 words of double spaced,
typewritten copy, including illustrations
and tables, is maximum. Count illustrations
and tables as 300 words each and submit
them on individual sheets. As an approximation, 1 page of text is about 300 words.
Submit one original typescript on 8-1/2 x
11 plain white paper, use 1 in. margins,
and include two good quality copies of the
entire discussion. References should be
complete. Do not repeat references cited
in original paper; cite them by original
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.
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
requirement will be returned or referred to
the author for private reply.
Send manuscripts to:
http://mc.manuscriptcentral.com/aci
Send discussions to:
Journals.Manuscripts@concrete.org
ACI STRUCTURAL JOURNAL
TECHNICAL PAPER
Title No. 112-S01
A Global Integrity Parameter with Acoustic Emission for

Load Testing of Prestressed Concrete Girders
by Francisco A. Barrios and Paul H. Ziehl
Structural evaluation of existing infrastructure has become a critical
subject in civil engineering. In recent years, significant efforts have
been placed on developing nondestructive techniques such as acoustic
emission to monitor and effectively assess the integrity of a structure
without causing significant damage. However, acoustic emission
methods face challenges regarding the subjectivity of associated
performance evaluation criteria and a lack of measurable parameters directly related to the mechanical response of the system. It has
been previously suggested that an integrated approach of the cyclic
load testing method with acoustic emission techniques may overcome
these difficulties and constitute a more effective, nondestructive load
testing methodology. The current investigation analyzes experimental
data gathered from flexural testing of six full-scale prestressed girder
specimens (lightweight and normalweight) and presents a potential
approach for damage detection and assessment within the minor to
intermediate damage zones based on acoustic emission data.
Keywords: girder; prestress; structural load test.
INTRODUCTION
Acoustic emission (AE) evaluation used in combination
with cyclic load testing is a promising nondestructive, or minimally destructive, technique for the assessment of existing
reinforced or prestressed concrete structures (Colombo et al.
2005; Ridge and Ziehl 2006; Galati et al. 2008; Ziehl et al.
2008; Liu and Ziehl 2009; Barrios and Ziehl 2011, 2012; Xu
et al. 2013). Attention has been placed on reducing the subjectivity of the AE criteria and quantifying, in terms of structural
damage, the changes in measured AE activity. Both AE and
cyclic load testing methods have achieved promising results
(ACI Committee 437 2007; JSNDI 2000; Barrios and Ziehl
2012), and an integrated approach may offer advantages and
decrease the challenges that each method faces for practical
implementation (Ziehl et al. 2008). Due to limited research
data, however, an integrated standardized approach is lacking
and quantification of damage with AE remains somewhat
subjective. The approach outlined herein presents the integration of AE data with cyclic load testing for the specific case
of prestressed concrete girders within the minor to intermediate damage zones. Potential applications for this approach
include prestressed double tees such as those used in parking
garages, one- and two-way passively reinforced or post-tensioned building slab systems (Galati et al. 2008; Ziehl et al.
2008), and prestressed and post-tensioned bridge girders. In
the case of bridge girders, the loading profile would be simplified and the load magnitude reduced (Ziehl et al. 2009a).
RESEARCH SIGNIFICANCE
Objective structural integrity evaluation has proven to be
a complex subject. Challenges arise from the wide range of
potential structural responses due to differing construction
materials and techniques combined with a need for rapid and

reliable integrity assessment. Acoustic emission techniques
have demonstrated very high sensitivity for sensing damage,
but challenges remain for coherent integrity evaluation
during load testing. This investigation presents a methodology to relate acoustic emission data to the level of damage
present in prestressed concrete girders loaded in flexure,
with a particular focus on the transition from the minor to
intermediate damage zone.
BACKGROUND
Integrity assessment with cyclic load test method
With the cyclic load test (CLT) method performance of the
system is evaluated through the three criteria of deviationfrom-linearity index, permanency ratio, and residual deflection (ACI Committee 437 2012). The loading pattern is
grouped in loadsets and is executed in a stepped fashion.
Each loadset consists of two identical load cycles (CycleA
and Cycle B). For the current investigation, the deviationfrom-linearity index was used to define boundaries between
minor, intermediate, and heavy damage. This index is
defined as follows

I DL = 1
tan (a i )
tan a ref
(1)
where tan(i) is the secant stiffness of any point i on the

increasing loading portion of the load-deflection envelope;
and tan(ref) is the slope of the reference secant line for the
load-deflection envelope (Fig. 1). A summary of the CLT
evaluation results for all six specimens is addressed in a
previous publication (Barrios and Ziehl 2012).
Integrity assessment with acoustic emission
Calm ratio versus load ratio (NDIS-2421)The Japanese
nondestructive inspection standard, NDIS-2421 (JSNDI
2000) proposes a graphical representation of the calm ratio
(CR) and load ratio (LR) values to classify the level of damage
in reinforced concrete flexural members. This method was
first studied and proposed for passively reinforced concrete
(Ohtsu et al. 2002); however, it has also been studied for
prestressed (Xu 2008; Xu et al. 2013) and post-tensioned
(Ziehl et al. 2008) concrete members. The terminology used
ACI Structural Journal, V. 112, No. 1, January-February 2015.
MS No. S-2011-056.R3, doi: 10.14359/51687294, was received February 6, 2014,
and reviewed under Institute publication policies. Copyright 2015, American
Concrete Institute. All rights reserved, including the making of copies unless
permission is obtained from the copyright proprietors. Pertinent discussion including
authors closure, if any, will be published ten months from this journals date if the
discussion is received within four months of the papers print publication.
Fig. 1Deviation from linearity (after ACI Committee 437

[2012]).
in NDIS-2421 is described in the following, and additional
terms related to AE are included at the end of this paper.
Load ratio (LR)Also known as Felicity ratio or concrete
beam integrity (CBI) ratio, the load ratio (LR) is a critical
parameter for AE evaluation and monitoring. This ratio
generally decreases with the accumulation of damage. It
is inversely related to the reduction in AE activity during
loading until the material is stressed beyond its previous
stress level, known as the Kaiser effect.
Calm ratio (CR)The calm ratio (CR) is related to the AE
activity during the unloading portion of the loading cycles
(Ohtsu et al. 2002). For the purposes of this paper, the CR
is calculated as the ratio of the total cumulative AE activity
throughout the unloading phase to the total AE activity during
the entire cycle. The CR can be calculated on either the loading
cycle or the reloading cycle of a particular loadset. Recent
experimental evidence suggests that the former computation
may provide a more consistent correlation with the accumulation of damage (Barrios and Ziehl 2011).
A crucial aspect for the computation of both the LR and
CR is the AE parameter selected for evaluation. Some authors
have used the number of hits during loading and unloading
for evaluation of the AE data (Ohtsu et al. 2002; Colombo etal.
2005). While this approach has shown promise, inspection of
the experimental data from the current investigation indicates
that signal energy or signal strength may be more appropriate for evaluation of the AE data than hits (Barrios and
Ziehl 2011). This can be explained by the observation that
the extent of damage is more closely related to the amount of
energy associated with each hit. Based on inspection of trends
in the data, the AE parameter of signal energy was found to be
promising and was therefore used in the current investigation.
The methodology for damage classification described in
NDIS-2421 is a graphical representation of CR versus LR for
each loadset (or load cycle) divided into four damage zones.
This method of damage classification is generally consistent
with damage trends. However, some challenges with this
approach may exist. First, the manner in which the graph is
partitioned allows for two distant damage levels (minor and
heavy) in very close proximity to one another, which does not
generally correspond to the nature of damage progression.
Second, previous investigations have generally relied on the
4
assistance of additional data, such as crack mouth opening

displacement (CMOD), to aid in locating the damage zones
(Ohtsu et al. 2002). This requires the installation of gauges
across existing cracks. To achieve this, existing cracks
must be present and visible, which may not be the case for
prestressed elements. Furthermore, the CMOD calibration
does not address other considerations such as crack density,
yielding, slippage of reinforcement, and concrete crushing.
These can be important for assessing integrity, particularly
for nonflexural modes such as shear and bond failure. Based
on CMOD data, threshold values for the CR and LR have
been derived for passively reinforced concrete flexural
members involved in a previous study (Ohtsu et al. 2002).
Other authors have located the boundaries intuitively so that
beams with similar damage plotted inside the corresponding
damage zone (Liu and Ziehl 2009; Colombo et al. 2005).
Experimental data from the current investigation indicate
that for the specific case of full-scale prestressed girders,
CR values do not increase proportionately with the accumulation of damage. Rather, as the plastic deformation of
the member increases, cracks remain open, reducing the AE
activity during unloading, hence decreasing the CR value.
This behavior complicates the classification of damage for
prestressed flexural members.
An assessment methodology that combines the CLT
criteria and the AE criteria, referred to as a global performance index IG, has been proposed (Ziehl et al. 2008). The
approach combines these criteria in a weighted manner and
includes a multiplier to account for knowledge of the structure (load history, previous load tests, and reinforcement
configuration), and the number of members being tested
compared to the total number of similar members in the
system. The AE portion of this index relies on the CR-versusLR index ICRLR. This approach has merit but does not eliminate the need for an external criterion to assess the level of
damage. In the absence of load-versus-displacement data, it
is not possible to quantify the magnitude of damage from the
ICRLR assessment.
The global integrity parameter (GIP) is a proposed loadtesting evaluation method (Barrios and Ziehl 2012) that aims
to quantify the amount of damage based only on deviation
from linearity IDL. The GIP identifies damage levels as fixed
percentages of the IDL at the nominal capacity of the member.
The GIP approach builds on the CLT evaluation criteria, but
differs in that the GIP approach uses specific member properties such as the fully cracked moment of inertia and elastic
stiffness of the member.
The preceding discussion provides an overview of recent
studies involving the evaluation of concrete members with
AE during in-place load testing, along with the GIP evaluation approach, which is based on load-versus-displacement
behavior. The discussion that follows is based on the specific
experimental results obtained during this investigation and
is focused on minor to intermediate damage in prestressed
flexural members. For this study, the damage level zones are
defined as: minor damage for IDL < 15%; intermediate damage
for 15% < IDL < 35%; and heavy damage for IDL>35%.
For IDL below 15%, cracks were barely visible and slight
nonlinearity was observed in the measured load-versusACI Structural Journal/January-February 2015
displacement response; at 35%, IDL cracks started to remain

open after unloading, and nonlinear behavior was clearly
present (Barrios and Ziehl 2012).
EXPERIMENTAL PROCEDURE
Experimental data were gathered during four-point
bending load testing of six full-scale prestressed girder specimens. The first set of specimens consisted of two girders
constructed using self-consolidating concrete (SCC-1 and
SCC-2) and one girder constructed with high-early-strength
concrete (HESC). The second set of specimens consisted
of two girders with self-consolidating lightweight concrete
(SCLC-1 and SCLC-2) and one girder with high-earlystrength lightweight concrete (HESLC). Specimen nomenclature is: concrete type (SCC, HESC, SCLC, or HESLC)specimen number associated with concrete type.
All girders were AASHTO Type III with a span of 58 ft
2 in. (17.7 m) and a depth of 45 in. (1140 mm). A concrete
mixture with a design compressive strength of 8000 psi
(55.2MPa) was used to cast the girders, and a 4000 psi
Fig. 2: Photograph and cross-section details (after Barrios

and Ziehl [2012]).
(27.6 MPa) concrete deck was cast in the laboratory prior to

testing. The concrete deck was 96 in. (2440 mm) wide and
8in. (203 mm) deep for the normalweight girders, and 30 in.
(762 mm) wide and 19 in. (483 mm) deep for the lightweight
girders (Fig. 2). All girders were prestressed using 18 bottom
and four top (all straight profile) low-relaxation strands with
a 0.5 in. (12.7 mm) nominal diameter and a tensile strength
of 270 ksi (1860 MPa). A 31,000 lb (138 kN) prestressing
force was applied to each strand prior to release. A summary
of the girder specimens is provided in Table 1 and further
information can be found in related South Carolina Department of Transportation reports (Ziehl et al. 2009b, 2010).
Internal and external instrumentation
Five vibrating-wire strain gauges were installed within
the girders to monitor the long-term strain for prestress loss
evaluation and other variables. Instrumentation was outfitted
externally on both the girder and deck to measure and record
displacement and concrete strain data. Strain gauges were
adhered to the top of the deck and at the bottom of the girder.
Cracks were marked on the girders with permanent markers
as they developed so that cracking patterns at different load
levels could be identified. Two draw-wire transducers placed
at midspan were used to measure deflection of the girders.
Acoustic emission activity was monitored continuously
during load testing. Six AE sensors (resonant in the vicinity
of 60 kHz with integral pre-amplification of 40 dB) were
mounted on one side of each girder specimen. The sensor
layout was symmetric about an axis located at the midpoint
between the loading points and a 48 dB test threshold was
used for all specimens (Fig. 3).
Loading protocol
The CLT load profiles for the normalweight girders are
shown in Fig. 4(a) and those for the lightweight girders are
shown in Fig. 4(b). Damage levels as determined through
the deviation-from-linearity index, as described previously,
combined with visual observations of cracking are shown
in the same figures. The resulting load-versus-displacement
behavior is shown in Fig. 5 and resulting cracking patterns
are shown in Fig. 6 (at 67% and 57% of nominal capacity Pn
for normalweight and lightweight specimens, respectively,
and also at the conclusion of the CLT loading protocols).
The loading profile was developed based on the calculated
nominal capacity of the normalweight girder specimens. A
total test load (maximum load applied) of 160 kip (712 kN)
was selected, which is approximately 73% of the measured
Table 1Summary of girder specimens

Specimen name
Age of girder at
Approximate age
release, days of girder at testing, days
Approximate age of
deck at testing, days
Ultimate load
achieved, kip (kN)
Failure mode
SCC-1
270
200
226 (1010)
Strand rupture
SCC-2
900
300
226 (1005)
Strand rupture
HESC
470
100
224 (997)
Strand slip
SCLC-1
16
315
50
290 (1290)
Strand yield followed by deck failure
SCLC-2
16
555
90
276 (1228)
HESLC
16
490
60
260 (1156)
capacity of these girder specimens. The maximum load level

was high enough to exert significant damage while avoiding
failure before the application of a 24-hour load test based on
Chapter 20 of ACI 318-11 (ACI Committee 318 2011). The
number of applied loadsets and their magnitudes was intended
to provide information regarding the behavior of the girders
at different deterioration levels. Prior to loading according
to Fig. 4(a), Girders SCC-2 and SCLC-2 were subjected to
cyclic fatigue loading for 2 million cycles to investigate the
potential effect of simulated field loading on the results.
The lightweight girders had a higher calculated nominal
capacity (Pn equal to 280 kip [1245 kN]) in comparison to the
normalweight specimens due to the modified deck geometry.
Therefore, for SCLC-1 and the HESLC girder, the maximum
test load applied corresponded to approximately 57% of the
nominal capacity. Because significant damage, as judged by
the deviation-from-linearity index, did not occur at this level
of loading, the test load magnitude for GirderSCLC-2 was
increased. For this specimen, Loadsets11 and 12 (at 68% and
80% of Pn, respectively) were added to the loading protocol.
Further descriptions of the loading protocols and results are
provided elsewhere (Barrios 2010; Barrios and Ziehl 2012).
Fig. 3Acoustic emission sensor layout (after Barrios and

Ziehl [2011]).
EXPERIMENTAL RESULTS
The results of the CLT evaluation and the 24-hour load test
in terms of load-versus-displacement behavior are described
in a previous publication (Barrios and Ziehl 2012). The
discussion that follows is focused on the treatment of the AE
data and corresponding results.
Normalweight girder specimens
Calm ratio-versus-load ratio plots for the normalweight prestressed girder specimens are shown with the
damage zone classification quadrants as determined from
the load-versus-displacement behavior (Barrios and Ziehl
2012) in Fig. 7(a). Loadset 3 fell within the minor damage
zone, Loadset 5 within the intermediate damage zone, and
Loadset7 within the heavy damage zone. For purposes of
comparison and to enable discussion, Fig. 7(b) shows the
damage quadrants from an investigation related to reducedscale (maximum span of 23 ft 0 in. [7.0 m]) prestressed
concrete tee-beam specimens (Xu 2008; Xu et al. 2013). The
agreement between the damage quadrants from these two
studies conducted at very different scales is reasonable.
Lightweight girder specimens
Results were similar for the lightweight girder specimens
(Fig. 8(a)). The classification zones shown in Fig. 8(a) were
developed based on the measured load-versus-displacement
behavior as for the normalweight specimens (Barrios 2010;
Barrios and Ziehl 2012). For comparison, Fig. 8(b) again
shows the damage quadrants from an investigation related
to reduced-scale prestressed concrete tee-beam specimens
(Xu et al. 2013). In this case, the agreement between the two
investigations is not as good.
To summarize, the CR and LR quadrants from different
investigations should not be used without modification;
rather, they are specific to the member being tested. Further-
Fig. 4Loading profiles (after Barrios and Ziehl [2011]). (Note: 1 kip = 4.45 kN.)
6
Fig. 5Load versus displacement due to CLT loading

protocol (after Barrios and Ziehl [2012]). (Note: 1 in. =
25.4 mm.)
more, they may vary between similar girders (for example,
Loadset 7 for the lightweight prestressed girder specimens
[Fig. 8(a)]). However, AE data can aid in the classification
of damage for prestressed concrete flexural specimens when
an external parameter, such as CMOD or the deviation-fromlinearity index, is used for calibration of the damage quadrants.
PROPOSED METHODOLOGY
Modified damage classification based on CRversus-LR plot
A modified approach for damage classification using the
CR-versus-LR plot is proposed and described as follows.
The first modification is for convenience: the LR criterion is
replaced with unity minus the load ratio (1.0 LR), referred to
hereafter as the complementary load ratio (CLR), so the curves
increase upward and to the right (Fig. 9). In this figure, the
values of load are labeled where significant changes in slope
occur. This is mentioned for later comparison with the IDLversus-load curves shown in Fig. 10, also referred as structural integrity loops (SIL) (Barrios 2010; Barrios and Ziehl
2012). In the CR-versus-CLR plots, hollow squares represent points that were interpolated using curves obtained for
CR and CLR independently (Barrios 2010). Values were
computed at the ends and at a minimum of two intermediate
positions within each linear segment.
The plots shown in Fig. 9 illustrate a trend in the AE data
with accumulation of damage in these prestressed girder specimens. The curves plotted for the normalweight prestressed
Fig. 6Cracking patterns (after Barrios and Ziehl [2012]).

girder specimens can be divided into three segments based
on changes in slope. The figures indicate that the CR is more
sensitive within very early stages of damage, while the CLR
is more sensitive in later stages of damage. This behavior
continues up to the level of load where cracks no longer
close during unloading. At this point, the CR values begin
to rapidly decrease (Loadsets 11 and 12 for SCLC-2). This
behavior is believed to be specific to prestressed specimens.
It is hypothesized that each change in slope in the CR-versus-CLR plot corresponds to a transition between damage
zones. An external check is required to confirm this hypothesis. Furthermore, to successfully compare and potentially
merge AE evaluation with the deviation-from-linearity index
IDL, a coherent description of the damage process is needed.
To investigate the hypothesis, load values at the transition points on the CR-versus-CLR plots (Fig. 9) are located
on the structural integrity loops (IDL versus load plots
[Fig.10]). The corresponding IDL values are then compared
to those obtained from a previous published investigation
(Barrios and Ziehl 2012), wherein damage is based on loadversus-displacement behavior. The damage classification
zones identified using the CR-versus-CLR plots correspond closely with those found through the load-versusdisplacement behavior. This implies that changes in slope on
the CR-versus-CLR plots represent physical changes in the
specimens and correlate with the damage process.
7
Fig. 7Damage quadrants (normalweight specimens).
Fig. 8Damage quadrants (lightweight specimens).
An interesting effect similar to the Kaiser effect is noticeable in the plots shown in Fig. 10. Girder SCC-1 starts
with an initial slope up to 115 kip (512 kN), where there
is an abrupt positive change. At this point, the IDL begins to
increase more rapidly, crossing from the minor to the intermediate damage zone until it reaches the maximum load
for that loadset at 128 kip (569 kN). On the next loadset,
the member reaches the same load level over a straight line,
and hence, a new higher load must be attained to produce
a positive change in the trend. The same trend occurs later
for SCC-2, where the IDL increases linearly with load up to
135kip (601 kN). At this level, a change in slope occurs
until it reaches the new maximum load of 160 kip (712 kN).
It can also be noted that loadsets applied at similar states of
damage tend to group closely until a new abrupt change in
slope is produced, generating a new cluster.
Lightweight girder specimens, SCLC-1 and HESLC, were
only loaded up to 160 kip (57% of Pn at Loadsets 7 and 9).
Therefore, the AE information gathered was not sufficient
to establish a trend. For SCLC-2 only one damage threshold
(intermediate to heavy at 160 kip [712 kN]) was located
from the information gathered in the CR-versus-CLR plot
(Fig. 9), resulting in the merging of the minor and intermediate zones into one segment.
In flexural members the minor damage region is dominated
by cracking and, therefore, there is a good match between
the damage zones identified with load-versus-displacement
behavior (Barrios 2010; Barrios and Ziehl 2012) and those
obtained from the AE data. It is possible to find more than
three changes in slope in a CR-versus-CLR plot, and the
external parameter of IDL is helpful for discriminating the AE
data into the three levels of damage (minor, intermediate,
and heavy) generally used for structural evaluation.
Proposed AE damage descriptor (arc of damage)

Acoustic-emission-based plots of CR versus CLR offer
useful information that can be related to the damage state
of prestressed flexural members when subjected to the CLT
loading protocol. Due to its extreme sensitivity to crack
growth, the AE method may also have the potential to offer
earlier detection of damage in comparison to methods based
on load-versus-displacement behavior.
To take advantage of this potential for early damage detection in prestressed flexural members, a new AE damage
descriptor is proposed that builds on a previously proposed
damage descriptor developed from load testing of two
building slab systems (Ziehl et al. 2008). In Fig. 11, this
new descriptor, referred to as the arc of damage, is shown
graphically. In this figure, results from Loadsets 3, 5, and
7 of an example specimen are plotted. The point where
the slope of the CR-versus-CLR curve crosses the horizontal axis (located at calm ratio = 0.0, complementary load
ratio = 0.17 in this case) is used as a reference point. The
angular measure (in radians) from the reference vertical line
drawn through this point to any loadset is then used as an
angular measure of damage (i, where the subscript i refers
to a particular loadset); and the linear distance from this
reference point to any loadset is used as a linear measure
of damage di. A numerical value for the arc of damage AD
descriptor is then calculated for each loadset as the product
of these two measures: i di.
To illustrate the influence of these two damage descriptors, values of angular measure i and linear distance di are
plotted separately in Fig. 12(a) and (b) for the normalweight
specimens and in Fig. 13(a) and (b) for the lightweight specimens. The resulting arc of damage AD for both specimen
types is plotted in Fig. 12(c) and 13(c), respectively.
Fig. 9CR-versus-CLR plots. (Note: 1 kip = 4.4 kN.)
Fig. 10Structural integrity loops and damage thresholds (after Barrios and Ziehl [2012]). (Note: 1 kip = 4.4 kN.)
For the normalweight girder specimens, the arc of damage
AD (Fig. 12(c)) provides a more consistent damage representation than either linear distance (Fig. 12(a)) or angular
measure (Fig. 12(b)) alone. The linear descriptor grows
faster at lower levels of damage (between Loadsets 3 and 5)
while the angular measure is more sensitive to damage in the
intermediate and heavy damage zones.
For the lightweight specimens, the linear distance descriptor
from the reference point (Loadset 5) increases rapidly up to
the theoretical minor-intermediate threshold corresponding
to Loadset 7 (Fig 13(a)). At that point, the linear distance
measurement increases at a reduced rate for a wide range of
the load value (62% to 87% of nominal capacity). This leads
to the observation that the linear distance descriptor is more
effective within the minor damage zone, while the angular
measure descriptor is more sensitive to damage within
the intermediate and heavy damage zones. This descriptor
increases steadily after Loadset 7 and follows a closely linear

pattern for the same load range (Fig. 13(b)). The scatter
observed at Loadset 5 for the normalweight specimens and
at Loadset 7 for the lightweight specimens in both the linear
and the angular descriptors is reduced for the resulting AD, as
shown in Fig. 12(c) and 13(c). This supports combining the
linear and angular descriptors, resulting in the proposed AD
descriptor, to increase the reliability of assessment. One key
characteristic of the AD descriptor is that it estimates deterioration of the member through a single numerical parameter, thereby facilitating a comparison of damage levels in
differing members.
Proposed assessment method within minor
damage zone, GIPAE
To take advantage of the sensitivity of AE, a new damage
criterion that is specifically focused on the minor damage
9
Fig. 11Definition of arc of damage.
Fig. 13Arc of damage versus percent of Pn (lightweight

specimens).
where for lightweight concrete

P PO
= 0.001 + 0.145 T
(3)
Pmi PO
and for normalweight concrete

Fig. 12Arc of damage versus percent of Pn (normalweight

specimens).
zone for prestressed flexural elements is proposed. The criterion is based solely on AE data as load-versus-displacement
data may lack the necessary sensitivity for reliable damage
assessment in the transition between the minor to intermediate
damage zone. Classification within the intermediate and heavy
damage zones, on the other hand, can be approached with a
number of different approaches including the arc of damage
(AD), the previously published GIP method (Barrios and Ziehl
2012), or the CLT method (ACI Committee 437 2012).
The proposed evaluation criterion, referred to as GIPAE
(for global integrity parameter based on AE), is based on the
AD and is described as follows

10
GIPAE = AD1 1.0
(2)
P PO
= 0.001 + 0.035 T
(4)
Pmi PO
where AD is the arc of damage for any loadset; PT is the

target load at which the damage criterion should reach unity;
Po is the peak load at the loadset with the lowest CR; and Pmi
is the load at the theoretical minor to intermediate damage
threshold. These equations were derived independently for
each set of girders (lightweight and normalweight) using a
linear interpolation of the average values of AD within the
minor damage region.
The same procedure can be followed to generate similar
equations for prestressed concrete girders with different
span, section geometry, and concrete strength. A minimum
number of two loadsets should be included within the minor
damage zone to calibrate the GIPAE criterion. However, in
a case where structural damage needs to be minimized and
even minor cracking in the members is not permitted, a
factor equal to 0.001 would be a conservative value.
Results from: a) the CLT criteria (ACI Committee 437
2007); b) the GIP criterion (Barrios and Ziehl 2012); and
c) the GIPAE method discussed previously are shown in Fig.
14. For the specimens tested, the GIPAE method provides the
most sensitive damage criterion. This is neither intrinsically
ACI member Paul H. Ziehl is a Professor in the Department of Civil and Environmental Engineering at the University of South Carolina. He is a member
of ACI Committee 437, Strength Evaluation of Existing Concrete Structures.
ACKNOWLEDGMENTS
Portions of this work were sponsored by the South Carolina Department
of Transportation and the ACI Concrete Research Council and their financial
support is greatly appreciated. Portions were performed under the support
of the U.S. Department of Commerce, National Institute of Standards and
Technology, Technology Innovation Program, Cooperative Agreement
Number 70NANB9H9007, and their support is likewise appreciated.
REFERENCES
Fig. 14Values at damage detection.

good nor bad, but in some cases, it may be beneficial and
desirable to avoid damage caused by the load test itself. The
GIPAE method reduces the load required for damage detection to approximately 50% of nominal capacity Pn in the
lightweight girder specimens and 52% of nominal capacity
in the normalweight girder specimens. These values are
comparable to the calculated cracking load (47% of Pn for
the lightweight specimens and 48% of Pn for the normalweight specimens). It is possible that the approach could be
tailored to further reduce load magnitude by decreasing the
AE threshold and/or installing a denser array of sensors.
SUMMARY AND CONCLUSIONS
One new damage descriptorthe arc of damage ADand
one proposed damage criterionGIPAEhave been developed based on data from flexural testing of six full-scale
prestressed girder specimens. Both are based on AE data.
The arc of damage addresses changes in AE behavior that
are related to differing damage states and is applicable to
minor, intermediate, and heavy damage states. The GIPAE
criterion incorporates mechanical properties of the member
and is specifically targeted to the transition between minor
to intermediate damage, thereby using the high sensitivity of
AE for crack initiation and extension. This may be applicable
for prestressed applications, particularly for cases where it is
desired to minimize damage due to the load-testing procedure itself. For further research, results from field testing
of prestressed girders are desirable for refinement of the
methods described.
AUTHOR BIOS
Francisco Barrios is an Assistant Professor in the Department of Civil
Engineering at Universidad del Magdalena, Santa Marta, Colombia. He
received his BS from the Universidad del Norte, Barranquilla, Colombia;
his MS from Tulane University, New Orleans, LA; and his PhD from the
University of South Carolina, Columbia, SC. His research interests include
structural health monitoring, damage diagnosis, load testing, and nonlinear
computer modeling of concrete structures.
ACI Committee 318, 2011, Building Code Requirements for Structural

Concrete (ACI 318-11) and Commentary, American Concrete Institute,
Farmington Hills, MI, 503 pp.
ACI Committee 437, 2007, Load Tests of Concrete Structures: Methods,
Magnitude, Protocols, and Acceptance Criteria (ACI 437.1R-07), American Concrete Institute, Farmington Hills, MI, 38 pp.
ACI Committee 437, 2012, Code Requirements for Load Testing of
Existing Concrete Structures and Commentary (ACI 437-12) (ACI Provisional Standard), American Concrete Institute, Farmington Hills, MI, 34 pp.
ASTM E1316-13d, 2013, Standard Terminology for Nondestructive
Examinations, ASTM International, West Conshohocken, PA, 38 pp.
Barrios, F., 2010, Acoustic Emission Techniques and Cyclic Load
Testing for Integrity Evaluation of Self-Consolidating Normal and Lightweight Prestressed Concrete Girders, PhD dissertation, Department of
Civil Engineering, University of South Carolina, Columbia, SC.
Barrios, F., and Ziehl, P., 2011, Effect of Loading Pattern on the
Acoustic Emission Evaluation of Prestressed Concrete Girders, Journal of
Acoustic Emission, V. 29, pp. 42-56.
Barrios, F., and Ziehl, P., 2012, Cyclic Load Testing for Integrity Evaluation of Prestressed Concrete Girders, ACI Structural Journal, V. 109,
No.5, Sept.-Oct., pp. 615-623.
Colombo, S.; Forde, M.; Main, I.; and Shigeishi, M., 2005, Predicting
the Ultimate Bending Capacity of Concrete Beams from the Relaxation
Ratio Analysis of AE Signals, Construction and Building Materials, V.
19, No. 10, pp. 746-754. doi: 10.1016/j.conbuildmat.2005.06.004
Galati, N.; Nanni, A.; Tumialan, J. G.; and Ziehl, P. H., 2008, In-Situ
Evaluation of Two Concrete Slab Systems. I: Load Determination and
Loading Procedure, Journal of Performance of Constructed Facilities,
V.22, No. 4, pp. 207-216. doi: 10.1061/(ASCE)0887-3828(2008)22:4(207)
JSNDI, 2000, Recommended Practice for In Situ Monitoring of
Concrete Structures by Acoustic Emission, NDIS 2421, Japanese Society
for Nondestructive Inspection, Tokyo, Japan, 6 pp.
Liu, Z., and Ziehl, P., 2009, Evaluation of Reinforced Concrete Beam
Specimens with Acoustic Emission and Cyclic Load Test Methods, ACI
Structural Journal, V. 106, No. 3, May-June, pp. 288-299.
Ohtsu, M.; Uchida, M.; Okamoto, T.; and Yuyama, S., 2002, Damage
Assessment of Reinforced Concrete Beams Qualified by Acoustic Emission, ACI Structural Journal, V. 99, No. 4, July-Aug., pp. 411-417.
Ridge, A., and Ziehl, P., 2006, Nondestructive Evaluation of Strengthened Reinforced Concrete Beams: Cyclic Load Test and Acoustic Emission
Methods, ACI Structural Journal, V. 103, No. 6, Nov.-Dec., pp. 832-841.
Xu, J., 2008, Nondestructive Evaluation of Prestressed Concrete
Structures by Means of Acoustic Emission Monitoring, PhD dissertation,
Department of Civil Engineering, University of Auburn, Auburn, AL.
Xu, J.; Barnes, R.; and Ziehl, P., 2013, Evaluation of Prestressed
Concrete Beams Based on Acoustic Emission Parameters, Materials Evaluation, V. 71, No. 2, pp. 176-185.
Ziehl, P.; Galati, N.; Nanni, A.; and Tumialan, J., 2008, In Situ Evaluation of Two Concrete Slab Systems. II: Evaluation Criteria and Outcomes,
Journal of Performance of Constructed Facilities, ASCE, V. 22, No. 4, pp.
217-227. doi: 10.1061/(ASCE)0887-3828(2008)22:4(217)
Ziehl, P. H.; Engelhardt, M.; Fowler, T. J.; Ulloa, F. V.; Medlock, R.D.;
and Schell, E., 2009a, Design and Field Evaluation of a Hybrid FRP/
Reinforced Concrete Bridge Superstructure System, Journal of Bridge
Engineering, ASCE, V. 14, No. 5, pp. 309-318. doi: 10.1061/(ASCE)
BE.1943-5592.0000002
Ziehl, P.; Rizos, D.; Caicedo, J.; Barrios, F.; Howard, R.; and Colmorgan,
A., 2009b, Investigation of the Performance and Benefits of Lightweight
SCC Prestressed Concrete Bridge Girders and SCC Materials, Final Report
submitted to the South Carolina Department of Transportation, 182 pp.
Ziehl, P.; Rizos, D.; Caicedo, J.; Colmorgan, A.; Howard, R.; and Barrios,
F., 2010, Investigation of the Performance and Benefits of Self-Consolidating Concrete for Prestressed Bridge Girders, Final Report submitted to
the South Carolina Department of Transportation, 212 pp.
11
NOTES:
12
TECHNICAL PAPER
Title No. 112-S02
Analysis of Rectangular Sections Using Transformed

Square Cross Sections of Unit-Length Side
by Girma Zerayohannes
This paper deals with the analytical proof of the equivalence
between the relative biaxial bending resistance of a rectangular
solid reinforced concrete section and the biaxial bending resistance of the transformed solid square section of unit-length side.
Similar proofs are also derived for rectangular hollow sections.
The results of the analytical proof show that the relative biaxial
bending resistance of a rectangular solid section is identical to the
biaxial bending resistance of the transformed square solid section
of unit-length side. These results only occur when the concrete
fiber and reinforcing bar coordinates in the transformed section
are in conformity with the transformation that maps the rectangular section into a square cross section of unit-length side. The
concrete and steel stresses in the transformed section comply with
the resulting stress transformation and the area of reinforcement
in the transformed section must comply with the resulting area
transformation. The proof also shows the equivalence in rectangular hollow sections, provided that similar transformation-related
conditions are met.
Keywords: analysis; biaxial bending; cross section; hollow sections;
homogeneous transformation; solid sections; transformed sections; unitside length.
INTRODUCTION
The results of cross-section analysis1,2 show that the relative
biaxial bending resistance of a rectangular solid section is identical to the biaxial bending resistance of the transformed square
cross section of unit-length side, provided that: 1) the concrete
fiber and reinforcing bar coordinates in the transformed section
are in conformity with the transformation that maps the rectangular section into a square cross section of unit-length side;
2) the concrete and steel stresses in the transformed section
comply with the resulting stress transformation; and 3) the area
of reinforcement in the transformed section comply with the
resulting area transformation. The results of cross-section analysis1,2 also show the equivalence between the relative biaxial
bending resistance of a rectangular hollow section and the
biaxial bending resistance of the transformed square hollow
cross section of unit-length side, provided that similar transformation related conditions are met.
While comparisons of cross-section analysis results have
shown the equivalence, analytical proof for its justification
is hardly available in the literature. More recently, however,
Cedolin et al.3 used the square cross section of unit-length
side to calculate interaction diagrams for load eccentricities
along axes parallel to the axes of symmetry and to a diagonal of a solid rectangular cross section for the derivation of
approximate analytical expressions of the moment contours
based on the ACI 318-05.4 Analytical proof of the equivalence between the dimensionless expressions for the rectanACI Structural Journal/January-February 2015
gular solid section with four-corner reinforcement and the

ultimate bending moments and axial force of an equivalent
square cross section of unit length-side is also provided in
Cedolin et al.3
This research deals with the development of a new
approach for the analytical proof of the equivalence between
the related biaxial bending resistances of rectangular solid
and hollow sections, and biaxial bending resistances of the
transformed square solid and hollow sections of unit-length
sides. The proposed method: 1) covers a wider range of
reinforced concrete sections with arbitrary reinforcement
arrangement; and 2) facilitates the calculations of biaxial
interaction diagrams because it allows the use of a single
valueunityas strength input data for the design strengths
of all classes of concrete, and the use of unit side lengths as
geometric input data representing all rectangular solid and
hollow sections.
Equivalence between relative biaxial bending
resistance of rectangular section and biaxial
bending resistance of transformed square cross
section of unit-length side
Biaxial interaction diagrams for solid rectangular cross
section made of reinforced concrete are presented as load
contours with the design normal force and biaxial bending
resistance expressed in non-dimensional form as

Rd =
Rd y =
Rd z =
N Rd
(1)
f cd b h
M Rd y
f cd b h 2
M Rd z
f cd h b 2
(2)
(3)
where Rd, Rd y, and Rd z are the relative values of the

combined design axial load and biaxial bending resistance
of the rectangular cross section; NRd, MRd y, and MRd z are the
MS No. S-2012-395.R3, doi: 10.14359/51687295, received April 2, 2014, and
reviewed under Institute publication policies. Copyright 2015, American Concrete
Institute. All rights reserved, including the making of copies unless permission is
obtained from the copyright proprietors. Pertinent discussion including authors
closure, if any, will be published ten months from this journals date if the discussion
is received within four months of the papers print publication.
13
combined design axial load and biaxial bending resistance

of the rectangular cross section; fcd is the design compressive
strength of concrete; and b and h are the side lengths of the
rectangular cross section.
Consider a linear homogeneous transformation T represented by the matrix (Eq. (4)), that maps the original
section defined in Eq. (1) to (3) to its image, which will be
referred to as the transformed section. The transformation
constitutes a two-way dilation, of factor k1 along the y-axis
and of factor k2 along the z-axis.

k1
Tk1 , k2 =
0
0
(4)
k2
The relative values of the combined design axial force and

biaxial moment resistances of the transformed section are
determined using the expressions in Eq. (5) to (7).

TRd =
TRd y =
TRd z =
T
N Rd
(5)
f bT hT
T
cd
T
M Rd
y
(6)
(7)
( )
f cdT bT hT
T
M Rd
z
( )
f cdT hT bT
combined design axial force and biaxial bending resistance

T
of the transformed section; f cd is the design compressive
strength of concrete in the transformed section; and bT and hT
are the side lengths of the transformed section.
The proof is carried out by first assuming that the relative values of the combined design axial force and biaxial
bending resistance is invariant under the transformation and
later proving that the assumption is valid.
Thus
Rd =
Rd y =
Rd z =
f cdT bT hT = 1 (11)
f cdT bT (hT ) 2 = 1 (12)
f cdT hT (bT ) 2 = 1 (13)
Equations (11) to (13) represents a system of independent

simultaneous equations in bT, hT, and f cdT . The solutions are
bT = 1, hT = 1, and f cdT = 1
T
N Rd
(8)
f bT hT
T
M Rd
y
( )
f cdT bT hT
M
T
Rd z
( )
f cdT hT bT
(14)
Because the products in Eq. (11) to (13) are dimensionless, it

T
follows that each of f cd , bT, and hT are also dimensionless.
Therefore, the unit values determined as solutions for bT, hT,
T
and f cd in Eq. (14) are constants without dimension. With the
solutions for bT and hT determined, the elements of the transformation matrix can be readily determined as k1 = 1/b and
k2= 1/h, so that the transformation matrix associated with the
special conditions in Eq. (11) to (13) is given by Eq. (15).
1/b 0
T =
(15)
0 1/h
For a more general transformation condition where the unit

values on the right-hand sides of Eq. (11) to (13) are replaced
by constants q1, q2, and q3, respectively (q1, q2, and q3 are real
numbers greater than zero), the solutions are given by Eq. (16)
bT = q3/q1, hT = q2/q1, and f cdT = q13 ( q2 q3 ) (16)
The solutions indicate a linear homogeneous transformation
represented by the transformation matrix in Eq. (4), where
the elements k1 and k2 are given by k1 = (q3/q1) (1/b) and
k2 = (q2/q1) (1/h). The design resistance of the transformed
section is then related to the normalized design resistance
of the original rectangular section through the relationships
shown in Eq. (17) to (19).
T
cd
Rd =
Rd y =
Rd z =
T
N Rd
(17)
q1
(9)
T
M Rd
y
q2
(18)
(10)
It can be seen from Eq. (8) to (10) that the combined design
axial force and biaxial bending resistance of the transformed
T
T
T
section ( N Rd
, M Rd y , M Rd
) satisfying the transformation
z
14
T
T
T
where Rd , Rd y, and Rd z are the relative values of the
T
T
T
of the transformed section; N Rd
, M Rd y , and M Rd
z are the
conditions laid down in Eq. (11) to (13) is equal to the relative values of the combined design axial force and biaxial
bending resistance (Rd, Rd y, Rd z) of the original section.
The conditions state that the results of the products are all
unity and dimensionless.
T
M Rd
z
q3
(19)
Therefore, the equivalent square cross section of unit-length

side is a special case resulting from the special condition q1 = q2
Fig. 1(a) Rectangular solid section; and (b) square solid section of unit-length side.
Fig. 2(a) Rectangular hollow section; and (b) square hollow section of unit-length side.
= q3 = 1. Finally, it can be concluded that the relative values of
the combined design axial force and biaxial bending resistance
of a rectangular section is equal to the combined design axial
force and biaxial bending resistance of a square cross section of
unit-length side that satisfies the transformation requirements
described prevously and associated transformations that will be
described in more detail in the following sections.
Coordinate transformation
Rectangular solid cross sectionsPreviously it was shown
that a homogeneous linear transformation with two-way dilation of factors (1/b) and (1/h) along the y- and z-axes, respectively, transforms a rectangular cross section into an equivalent
square cross section of unit-length side. The transformation was represented by the transformation matrix shown in
Eq.(15). The matrix is referred to as coordinate transformation matrix to emphasize its use in the determination of the
coordinates of any desired point such as corner concrete fibers
and reinforcing bar locations in the transformed sections.
As an example, the transformation matrix is used in
Eq. (20) and (21) to map corner concrete fiber and reinforcing bar coordinates in the first quadrant of the original
cross section (Fig. 1(a)), to the images in the equivalent
square cross section of unit length-side (Fig. 1(b)).

1/b 0 b / 2 0.5
0 1/h h / 2 = 0.5 (20)

1 b 0 (b / 2) b 0.5 (b /b)
0 1 h (h / 2) h = 0.5 (h /h) (21)
The term homogeneous is used to indicate that the

origin is an invariant point under the transformation.
Other invariant variables under the transformation include

geometric reinforcement ratio , mechanical reinforcement
ratio , combined related design axial force and biaxial
bending resistance (Rd, Rd y, Rd z), and strains of corresponding fibers in the original and transformed sections.
The proof of invariance of each variable will be discussed in
subsequent sections.
Rectangular hollow cross sectionsFigures 2(a) and (b)
show the actual rectangular hollow section with uniformly
distributed reinforcement along the edges and the transformed square hollow section of unit-length side, respectively. The latter is determined using the transformation
described in the following.
Biaxial interaction diagrams for hollow rectangular cross
section made of reinforced concrete are presented in nondimensional form as

Rd =
Rd y =
Rd z =
N Rd
(22)
f cd a b h
M Rd y
f cd a b h 2
M Rd z
f cd a h b 2
(23)
(24)
where is the fraction of the solid part of the cross section,

which will be referred to as solidity ratio in short and the
definitions of other variables are as in Eq. (1) to (3).
The relative values of the combined design axial force and
biaxial moment resistance of the transformed section are
determined using the expressions in Eq. (25) to (27).
15
T
N Rd
(25)
f cdT aT bT hT
TRd =
TRd y =
T
Rd z
T
M Rd
y
f cdT aT bT (hT ) 2
=
T
M Rd
z
f cdT hT (bT ) 2
(26)
(27)
where T is the solidity ratio of the transformed section that

will be shown to be invariant under the transformation
that is, T = . The other variables are the same as for solid
crosssections.
Using similar assumptions made in the solid cross sections
regarding invariance under the transformation of the relative values of the combined design axial force and biaxial
bending resistance, Eq. (28) to (30) hold. The validity of the
assumption will be proved subsequently.
Rd =
Rd y =
Rd z =
T
N Rd
(28)
f a bT hT
T
M Rd
y
f cdT a bT (hT ) 2
1/b 0 b / 2 0.5
0 1/h h / 2 = 0.5 (35)
1/b 0 0.4b 0.4

0 1/h 0.3h = 0.3 (36)
T
M Rd
z
f cdT a hT (bT ) 2
(29)
(30)
f cdT a bT hT = 1 (31)
f cdT a bT (hT ) 2 = 1 (32)
f cdT a hT (bT ) 2 = 1 (33)
The solutions are

bT = 1, hT = 1, and f cdT = 1/
The locations of individual reinforcing bar in the square

hollow section of unit-length side are also determined using
the same transformation matrix.
Area and stress transformation
Rectangular solid sectionsIn a linear transformation
T with the matrix given by Eq. (15), the magnitude of the
determinant (determinant = (1/b) (1/h) 0 0 = 1/(b h))
is equal to the ratio of the area of the new shape to the
area of the original shape. Therefore, in a transformation
with two-way dilation of factors (1/b) and (1/h) parallel to
y- and z-axes, respectively, the area of the image undergoes
a dilation of factor (1/b) (1/h) = 1/(b h). As a result, the
transformed area of the compression zone and transformed
area of the reinforcement in the square cross section of unitlength side are given by Eq. (37) and (38), respectively.

T =
(37)
bh
AsT =
As
(38)
bh
(34)
Following the same argument as in solid sections, it follows

T
that each of f cd , bT, and hT are dimensionless. Moreover,
because the solutions for bT and hT remain unchanged, the
derivations leading to and including the transformation
matrix in Eq. (15) apply for hollow cross sections. The argument about the more general transformation conditions can
16
T
cd
It can be seen from Eq. (28) to (30) that the combined design
axial force and biaxial bending resistance of the transformed
T
T
T
section ( N Rd , M Rd y , M Rd
z ) satisfying the transformation
conditions laid down in Eq. (31) to (33) is equal to the relative values of the combined design axial force and biaxial
bending resistance (Rd, Rd y, Rd z) of the original section.
The conditions state that the results of the products are all
unity and dimensionless.
also be directly extended to rectangular hollow sections to

show that the transformed square hollow cross section of
unit-length side is a special case resulting from the special
condition (q1 = q2 = q3 = 1) in Eq. (31) to (33).
Finally, it can be concluded that the relative values of the
of a rectangular hollow section is equal to the combined
design axial force and biaxial bending resistance of a square
hollow cross section of the unit-length side that satisfies
the transformation requirements described previously. The
associated transformation requirements will be described in
more detail in subsequent sections.
As an example, the coordinate transformation matrix is
used in Eq. (35) and (36) to map the outer and inner concrete
fiber coordinates in the first quadrant of the original cross
section (Fig. 2(a)), to the images in the equivalent square
hollow sections of unit-length side (Fig. 2(b)). In the example,
the aspect ratio and relative wall thicknesses of the original
section are arbitrarily chosen as b/h = 2.0 and wb=wh = 0.2h.
where T is the area of the compression zone in the transformed section, and is the area of the compression zone in
the original section.
Similarly, because the transformation conditions in
Eq.(11) to (13) have caused the transformation of the design
compressive strength into unity, the transformation factor
for stresses in concrete and reinforcement is 1/fcd. As a result,
Fig. 3Stress-strain diagrams in original and transformed solid sections: (a) concrete; and (b) reinforcing steel.
the stress-strain relationships of concrete and steel in the
original solid section are transformed into the stress-strain
relationships of concrete and steel in the transformed section
(Fig. 3(a) and (b)). In particular, the transformed design
yield strength of reinforcement, f ydT , is given by Eq. (39)
f ydT =
f yd
f cd
(39)
The stress-strain curves for materials and reduction

factors in the original sections are according to Eurocode 2.5
In Eurocode 2,5 the stress-strain curve for concrete is idealized by a parabolic function followed by a plateau. The
steel stress-strain is idealized as elastic-perfectly plastic.
The design strengths are obtained from the characteristic
(nominal) values using constant reduction factors (partial
factors for concrete, c, and steel, s).
It is possible to derive the transformation factor for stress
resultants after having derived the transformation factor for
stresses, and the dilation factors for areas and length measurements along the y- and z-axes. It is also possible to prove the
assumption made previously with regard to the invariance under
the transformation of the relative values of the combined axial
force and biaxial bending resistance of the original section.
The combined design axial force and biaxial bending
resistance of the transformed section is determined by calculating the stress resultants at the ultimate limit state using
Eq. (40) to (42).

T
N Rd
= Tc d T + AsiT Tsi (40)
T
T T
T
T T T
M Rd
y = c z d + Asi si z si (41)
T
T T
T
T T T
M Rd
z = c y d + Asi si ysi (42)
where T is the area of the compression zone in the transT

formed section; c is the compressive stress on an elemental
area of concrete in the compression zone; AsiT is the area of
steel reinforcement bar i; Tsi is the steel stress in reinforcement bar i; yT and zT are the moment arms of the elemental
area of concrete about z- and y-axes, respectively; and ysiT
and zsi are the moment arms of the reinforcement bar i about
z- and y-axes, respectively.
Because T = /(b h) from Eq. (37), it follows that
dT = (1/(b h))d
(43)
Using the change of variable indicated in Eq. (43) and the

transformation factors described previously, Eq. (40) to (42)
can be rewritten as

T
N Rd
=
1
1
c d + Asi si (44)
i
f cd b h
T
M Rd
y =
1 1 1

c zd + Asi si zsi (45)
i
f cd h b h
T
M Rd
z =
1 1 1

c yd + Asi si ysi (46)
f cd b b h
The quantities in brackets on the right-hand sides of

Eq.(44) to (46) are expressions for the stress resultants in
the original section. Whether or not the stress resultants are
in the ultimate limit state similar to the transformed section
needs further discussion. Figures 4(a) and (b) show the
strain and stress distributions in the original and transformed
sections, respectively. The strain distributions in the transformed section represent an ultimate limit state. The location
of the neutral axis in the original section can be determined
from the intercepts k1y0 and k2z0 in the transformed section
using the dilation factors k1 and k2 along the y- and z-axes,
respectively. The intercepts in the original section are thus
k1y0/k1 = y0, and k2z0/k2 = z0, as shown in Fig. 4(a). Moreover,
it can be verified using geometry that any given fiber parallel
to the neutral axis in the transformed section is an image of a
corresponding fiber parallel to the neutral axis in the original
section. It can also be shown that the ratio of the neutral axis
depth in the transformed section to the neutral axis depth in
the original section can be expressed in terms of k1, k2, y0,
and z0. Let it be designated by k3.
Fiber strains on the neutral axis or other fibers parallel to
the neutral axis in the original section can be determined
from corresponding fiber stresses in the transformed section
using the stress transformation and stress-strain diagrams
17
Fig. 4(a) Strain and stress distribution at ultimate limit state: (a) in original section; and (b) in square cross section of unitlength side.
(Fig. (3a)). As an example, the fiber stresses along the neutral
T
axis in the original section are NA = NA fcd = 0 fcd = 0 and
using the stress-strain diagrams, the strains are NA = 0.
For parallel fibers in the transformed section that are closer
to the neutral axis than the one with strain c2, at reaching
the maximum strengththat is, (1)the stresses are less
than 1 (Fig. 4(b)). As stated previously, the stresses in the
corresponding fibers in the original section can be determined using the stress transformation. Although the stresses
in the two fibers are different, the strains are the same as
shown in the strain stress diagram (Fig. 3(a)). In addition
from geometry, the strain in the most compressed fiber in the
transformed section can be shown to be equal to the strain
in the corresponding fiber in the original section. Therefore,
the strain distribution in the original section is one in the
ultimate limit state and the stress resultants are the design
resistance of the section. The equality of strains holds for
all corresponding fibers, including reinforcing bars, leading
to the conclusion that strains of corresponding fibers are
invariant under the transformation.
Thus, Eq. (44) to (46) can be rewritten as

T
N Rd
=
T
M Rd
y =
T
M Rd
z =
N Rd
= Rd (47)
f cd b h
M Rd y
f cd b h 2
M Rd z
f cd h b 2
= Rd y (48)
= Rd z (49)
Therefore, the transformation factors for the design resistance of the original cross section (NRd, MRd y, MRd z) are
1/(fcd b h), 1/(fcd b h2), and 1/(fcd b2 h), respectively.
Because
TRd y =
TRd z =
T
T
N Rd
N Rd
T
(50)
=
= N Rd
f cdT bT hT 111
T
M Rd
y
f cdT bT (hT ) 2
T
M Rd
z
f cdT hT (bT ) 2
T
M Rd
y
1112
T
M Rd
z
1112
T
= M Rd
y (51)
T
= M Rd
z (52)
it follows that

TRd = Rd (53)
TRd y = Rd y (54)
TRd z = Rd z (55)
Therefore, the assumption that the relative values of the

combined axial force and biaxial bending resistance is
invariant under the transformation is valid, and the derivations based on this assumption are appropriate.
Further, the geometric reinforcement ratio in the transformed section is

18
TRd =
rT =
AsT
(56)
bT hT
T
Substituting bT = hT = 1 and the expression for As from
Eq.(38) in Eq. (56)
A
rT = s = r (57)
bh
Equation (57) indicates that the geometric reinforcement

ratio is invariant under the transformation.
Similarly, the mechanical reinforcement ratio in the transformed section is

T = rT
f ydT
f cdT
(58)
Substituting f cdT = 1

T = rT f ydT (59)
Substituting further for T and f ydT from Eq. (57) and (39)

T = r
f yd
f cd
= (60)
Equation (60) indicates that is also invariant under the

transformation.
Finally, from Eq. (39), (56), (59), and (60)

AsT =
f cd
(61)
f yd
Equation (61) gives the transformed area of steel in the square

cross section of unit-length side in terms of the mechanical
reinforcement ratio , the design compressive strength of
the concrete, and yield strength of the reinforcement in the
original cross section. This same amount of concrete area is
to be deducted if the analysis would be based on net cross
section. Usually, analysis is based on gross cross sections, as
the use of net cross sections does not affect the result significantly. The effect of the displaced amount of concrete on
the cross section capacity may, however, be significant if the
high strength of concrete is used, requiring analysis on the
basis of net cross section.6,7
T
The transformed area of reinforcement, As , can also be
expressed in terms of the transformed design yield strength
of reinforcement, f ydT , as

AsT =
(62)
f ydT
Additional analytical advantage can be gained by setting

f ydT = 1, because it allows the direct substitution of the reinforcement data by the mechanical reinforcement ratio . It is
to be noted that this is not a consequence of the transforma-
tions discussed so far. It is rather an isolated action that

allows the substitution of the amount of reinforcement AsT in
the square cross section of unit-length side by , provided
T
that f yd = 1. The direct use of as reinforcement data can be
used advantageously in the calculation of biaxial interaction
diagrams where it can be systematically varied to cover the
practical range of the mechanical reinforcement ratio.
Rectangular hollow sectionsHollow sections are treated
as the result of two solid components made up of the full
cross section and the hollow part with positive and negative
areas, respectively.
Previously, it was shown that the transformation matrix T
for hollow sections and solid sections are the same. Therefore, the transformed area of the compression zone and
transformed area of the reinforcement in the square hollow
cross section of unit-length side are given by Eq. (37) and
(38), respectively. In Eq. (37), the area of the compression
zone, , has now two components made up of positive and
negative areas associated with the actual solid and hollow
component of the compression zone.
Further, the transformed area of the solid part of the cross
T
section, Ac , can be determined using the area transformation
factor as
AcT =
Ac
a b h
=
= a (63)
bh
bh
The solidity ratio T in the transformed section is

aT =
AcT
a
a
= T T =
= a (64)
T
T
11
b h
b h
Therefore the solidity ratio is invariant under the

transformation.
Similarly, because the transformation conditions in
Eq.(31) to (33) have caused the transformation of the design
compressive strength into 1/, the transformation factor for
stresses in concrete and reinforcement is 1/( fcd). As a
result, the stress-strain relationships of concrete and reinforcing steel in the original hollow section are transformed
into the stress-strain relationships of concrete and reinforcing steel in the transformed section, as shown in Fig. 5(a)
and (b). In particular, the transformed design yield strength
of reinforcement, f ydT , is given by Eq. (65).
f ydT =
f yd
(a fcd )
(65)
Following the same argument that led to Eq. (40) to (42)

in the solid sections and noting the stress transformations in
hollow sections described previously, Eq. (44) to (46) take
the form

T
N Rd
=
1
1
c d + Asi si (66)
i
(a fcd ) b h
19
Fig. 5Stress-strain diagrams in original and transformed hollow sections: (a) concrete; and (b) reinforcing steel.

T
M Rd
y =
1
1 1

c zd + Asi si zsi (67)
i
(a fcd ) h b h
T
M Rd
z =
1
1 1

yd + Asi si ysi (68)
(a fcd ) b b h c
Equations (66) to (68) result in Eq. (69) to (71) after

following the same argument as in the solid cross sections.
T
N Rd
=
T
M Rd
y =
T
M Rd
z =
N Rd
= Rd (69)
a f cd b h
Substituting bT = hT = 1 and the expression for AsT from

Eq.(38)
rT =
Equation (76) indicates that the geometric reinforcement

ratio is invariant under the transformation.
Similarly, the mechanical reinforcement ratio in the transformed section is
T = rT
M Rd y
a f cd b h 2
M Rd z
a f cd h b 2
As
= r (76)
a b h
= Rd y (70)
f ydT
f cdT
(77)
Substituting f cdT = 1 a
= Rd z (71)
T = T T f ydT (78)
Substituting further for T and f ydT from Eq. (76) and (65)
Because
TRd
TRd y =
T
Rd z
T
NT
N Rd
= T T Rd T T =
= N T (72)
f cd a b h
(1 a ) a 11 Rd
T
M Rd
y
T
cd
( )
f a b h
T
M Rd
z
T
cd
T 2
T
M Rd
y
(1 a ) a 112
T
M Rd
z
20
rT =
f yd
f cd
= (79)
AsT = a
f cd
(80)
f yd
T
= M Rd
z (74)
it follows that Rd = Rd, Rd y= Rd y, and Rd z = Rd z. Therefore, the assumption that the relative values of the combined
axial force and biaxial bending resistance is invariant under
the transformation is valid, and the derivations based on this
assumption are appropriate.
The geometric reinforcement ratio in the transformed
section is
T
T = r
Equation (79) indicates that is also invariant under the

transformation. Finally, from Eq. (65), (75), (78), and (79)

( ) (1 a) a 11
f a h b
T 2
T
= M Rd
y (73)
AsT
(75)
a bT hT
Equation (80) gives the transformed area of steel in the

square hollow section of unit-length side in terms of the
solidity ratio, mechanical reinforcement ratio , design
compressive strength of the concrete, and yield strength of
the reinforcement in the original rectangular hollow section.
This same amount of concrete area is to be deducted in the
transformed section if the analysis would be based on net
cross section.6,7
T
The transformed area of reinforcement, As , can also be
expressed in terms of the transformed design yield strength
of reinforcement, f ydT , as
AsT =
(81)
f ydT
Thus, the amount of reinforcement in the square hollow

cross section of unit-length side can be replaced by by
setting f ydT = 1.
CONCLUSIONS
The following conclusions can be drawn from this study:
1. The paper deals with the analytical proof of the equivalence between the relative biaxial bending resistance of a
rectangular solid reinforced concrete section, and the biaxial
bending resistance of the transformed square solid sections
of unit-length side. The results of the analytical proof show
that the relative biaxial bending resistance of a rectangular
solid section is identical to the biaxial bending resistance
of the transformed square solid section of unit-length side.
These results only occur when the concrete fiber and reinforcing bar coordinates in the transformed section are in
conformity with the transformation that maps the rectangular section into a square cross section of unit-length side.
The concrete and steel stresses in the transformed section
comply with the resulting stress transformation and the area
of reinforcement in the transformed section must comply
with the resulting area transformation.
2. The results of the analytical proof also show the equivalence between the relative biaxial bending resistance of a rectangular hollow section and the biaxial bending resistance of the
transformed square hollow section of unit-length side, provided
that similar transformation-related conditions aremet.
3. The results of the analytical proof has shown that the
geometric reinforcement ratio , the mechanical reinforcement ratio , the relative values of the combined design
axial force and biaxial bending resistances (Rd, Rd y, Rd z),
and strains of corresponding fibers in the original and transformed sections are invariant under the transformation.
4. Transformed square cross section of unit-length side
fulfilling the aforementioned conditions ((1) and (2)) are
used in the calculation of biaxial interaction diagrams for
rectangular solid and rectangular hollow sections.2
5. Relative values of combined design axial load and
biaxial bending resistances are calculated on the basis of
net cross section for high-strength concrete by deducting the
amount of the transformed area of steel (AsT = fcd/fyd) and
(AsT = fcd/fyd) from the square solid and square hollow
section of unit-length side.2
AUTHOR BIOS
ACI member Girma Zerayohannes is an Associate Professor at Addis
Ababa University, Addis Ababa Institute of Technology, Addis Ababa, Ethiopia. He received his BSc and MSc from Addis Ababa University in 1979
and 1984, respectively, and his PhD from the University of Kaiserslautern,
Germany, in 1995. His research interests include nonlinear analysis of reinforced concrete structures.
ACKNOWLEDGMENTS
Germany. The author wishes to gratefully acknowledge the financial

support by the German Academic Exchange Services (DAAD).
NOTATION
AsT
AsiT
b, h
bT, hT
fcd
fcdT
= area of reinforcement in transformed section

= area of steel reinforcement bar i
= side lengths of rectangular cross section
= side lengths of transformed section
= design compressive strength of concrete
= design compressive strength of concrete in transformed section
fyd, f ydT
= design yield strength of reinforcing steel in original
and transformed sections, respectively
k1, k2
= dilation factors along y- and z-axes, respectively
k3
= ratio of neutral axis depth in transformed section to
that of original section
NRd, MRd y, MRd z =
combined design axial load and biaxial bending
resistance of rectangular cross section
T
T
, M Rd
, T =
combined design axial force and biaxial bending
N Rd
y M Rd z
resistance of transformed section
yT, zT
= moment arms of elemental area of concrete about zand y-axes, respectively
= moment arms of the reinforcement bar i about z- and
ysiT , zsiT
y-axes, respectively
, T
= solidity ratio in original and transformed section,
respectively
c2
= concrete strain at reaching maximum strength fcd and/
or 1
Rd, Rd y, Rd z = relative values of combined design axial load and
biaxial bending resistance of rectangular cross section
TRd , TRd y , TRd z = relative values of combined design axial force and
biaxial bending resistance of transformed section
,
= geometric and mechanical reinforcement ratio,
respectively
T , T
= geometric and mechanical reinforcement ratios in
transformed section, respectively
= compressive stress on elemental area of concrete in
Tc
compression zone
Tsi
= steel stress in reinforcement bar i
T

= area of compression zone in transformed section

= area of compression zone in original section
REFERENCES
1. Busjger, D., and Quast, U., Programmgesteuerte Berechnung beliebiger Massivbauquerschnitte unter zweiachsiger Biegung mit Lngskraft
(Computer Program for the Analysis and Design of Arbitrarily Shaped
Reinforced Concrete Sections under Axial Load and Biaxial Bending),
Deutscher Ausschuss fr Stahlbeton, Heft 415, Beuth Verlag, Berlin,
Germany, 1990, 212 pp. (in German)
2. Zerayohannes, G., Bemmesungsdiagramme fr Schiefe Biegung mit
Lngskraft nach DIN 1045-1: 2001-07 (Interaction Diagrams for Biaxial
Bending with Axial Load on the Basis of the German Code DIN 1045-1:
2001-07), Schriftenreihe der Fachgebiete Baustofftechnologie und Bauschadenanalyse, Massivbau und Baukonstruktion und Stahlbau des Studienganges Bauingenieurwesen der Technischen Universitt Kaiserslautern,
Band 4, 2006, 270 pp. (in German and English)
3. Cedolin, L.; Cusatis, G.; Eccheli, S.; and Rovda, M., Capacity of
Rectangular Cross Sections under Biaxially Eccentric Loads, ACI Structural Journal, V. 105, No. 2, Mar.-Apr. 2008, pp. 215-224.
4. ACI Committee 318, Building Code Requirements for Structural
Concrete (ACI 318-05) and Commentary (ACI 318R-05), American
Concrete Institute, Farmington Hills, MI, 2005, 430 pp.
5. Eurocode 2, Part 1-1: General Rules and Rules for Building, European Committee for Standardization, Brussels, Belgium, 2003, 225 pp.
6. Zilch, K.; Jhring, A.; and Mller, A., Erluterungen zu DIN 1045-1
Explanation on DIN 1045-1), Deutscher Ausschuss fr Stahlbeton, Heft
525, Beuth Verlag, Berlin, Germany, 2003. (in German)
7. DIN 1045-1:2001-07, Tragwerke aus Beton, Stahlbeton und Spannbeton. Teil 1: Bemessung und Konstruktion (Concrete, Reinforced and
Prestressed Concrete Structures Part 1: Design), Normenausschuss
Bauwesen (NABau) im DIN (Deutsches Institut fr Normung) e.V. Beuth
Verlag, Berlin, Germany, July 2001. (in German)
This investigation was conducted at the Institute of Concrete Structure and Building Construction, Technical University of Kaiserslautern,
21
NOTES:
22
TECHNICAL PAPER
Title No. 112-S03
Evaluation of Post-Earthquake Axial Load Capacity of

Circular Bridge Columns
by Vesna Terzic and Bozidar Stojadinovic
Objective evaluation of the capacity of a bridge to carry selfweight and traffic loads after an earthquake is essential for a
safe and timely reopening of the bridge. The ability of a bridge
to function depends directly on the remaining capacity of the
bridge columns to carry gravity and lateral loads. An experimental
study on models of modern circular reinforced concrete bridge
columns was performed to investigate the relationship between
earthquake-induced damage in bridge columns and the capacity
of the columns to carry axial load in a damaged condition. The
earthquake-like damage was induced in the column specimens in
bidirectional, quasi-static, lateral load tests. The damaged column
specimens were then recentered to eliminate the residual drifts
and tested in compression to failure to evaluate their remaining
axial load strength. It was found that well-confined modern bridge
columns lose approximately 20% of their axial load capacity after
sustaining displacement ductility demands of 4.5.
Keywords: axial tests; earthquake; post-earthquake lateral stiffness; quasistatic tests; reinforced concrete.
INTRODUCTION
Modern highway bridges in California designed using
the Caltrans Seismic Design Criteria1 (SDC) are expected
not to collapse during both frequent and rare earthquake
events. Currently, design provisions aimed at preventing
structural collapse are supported by numerous experimental
data points and calibrated computer models.2,3 However,
there is no evidence that the bridge systems were tested for
the remaining traffic load capacity after some damage was
induced under lateral loading. Still, attempts were made
toward analytical evaluation of the ability of a highway
overpass bridge4 or bridge columns5 to carry traffic load
after an earthquake. Due to the lack of the validated quantitative guidelines for estimating the remaining traffic
load-carrying capacity of bridges after an earthquake, bridge
inspectors and maintenance engineers provide an estimate
of the capacity of the bridge to function based on qualitative observations, with each judgment founded on personal
experience. Such subjective evaluation can be significantly
improved if a model to provide a quantitative estimate of the
remaining load-carrying capacity of bridge columns after an
earthquake was developed and calibrated.
A combined experimental and analytical research program
was performed to investigate the relationship between
earthquake-induced damage in reinforced concrete bridge
columns and the capacity of the columns in such damaged
condition.6 This program comprised one axial load test, three
quasi-static cyclic tests, and two hybrid model earthquake
response simulations on scaled models of typical circular
Table 1Test matrix

Specimen designation
Ductility target
Test sequences
Base0
Axial
Base15
1.5
Lateral and axial
Base30
3.0
Lateral and axial
Base45
4.5
Lateral and axial
bridge columns used in modern bridges in California. In

this paper, the outcomes of the quasi-static cyclic part of
the experimental program are presented. In the first stage of
the quasi-static testing procedure, three column specimens
were tested by applying a bidirectional quasi-static incremental lateral displacement protocol with circular orbits of
displacement up to the predetermined displacement ductility
targets of 1.5, 3, and 4.5. In the second stage of the testing
procedure, an undamaged column specimen and the three
damaged specimens with no permanent drifts were subjected
to a monotonically increasing axial force up to failure. The
specimens are listed in Table 1. These results support evaluations of post-earthquake traffic load capacities of bridges
with well-confined reinforced concrete columns.
Reliable evaluation of the capacity of a bridge to carry
self-weight and traffic loads is essential for a safe and timely
re-opening of the bridge after an earthquake. Columns of
modern California bridges are designed to develop significant flexural deformation ductility without shear failure
and prevent bridge collapse. An experimental and analytical
evaluation of earthquake-damaged modern bridge columns
is used to quantify their axial load capacity and to develop
reliable models for objective evaluation of the ability of a
modern bridge to perform as intended after an earthquake:
continue to safely carry traffic load.
EXPERIMENTAL INVESTIGATION
Ketchum et al.7 developed a series of highway overpass
bridges designed in accordance with the Caltrans SDC1
in a recent PEER Center study. Bridge Type 11 (shown in
Fig. 1)typical for tall overpass bridgeswas chosen as a
prototype for this experimental study. The bridge is a fiveACI Structural Journal, V. 112, No. 1, January-February 2015.
MS No. S-2013-075.R1, doi: 10.14359/51687296, received March 5, 2014, and
23
Fig. 1Prototype Caltrans bridge.7

span single-column-bent overpass with 120 ft (36.58 m)
edge spans, 150 ft (45.72 m) inner spans, a 39 ft (11.89 m)
wide deck, and 50 ft (15.24 m) tall circular columns. The two
principal parameters that affect the remaining axial capacity
of bridge columns are the column aspect ratio H/D and the
column shear strength (or transverse reinforcement ratio t).8
Different possible values of these two parameters, bounded
by the provisions of the Caltrans SDC,1 were investigated.6
Values of H/D = 8 and t = 0.75% were selected for the
columns of the prototype bridge. The columns are modeled
with specimens referred to herein as the Base specimens.
The Base specimens are cantilever columns representing the
bottom half of the prototype bridge columns. The specimens
were tested in a single-curvature-bending-using loading
pattern that will induce displacement ductilities observed in
columns of the prototype bridge for the two bridge directions: transverse and longitudinal.
Specimen geometry, reinforcement, and materials
The geometry and the reinforcement of a Base specimen
are detailed in Fig. 2. The specimen has a 73.75 in. (1.875m)
tall, 16 in. (0.4 m) diameter circular column, and a square
(84 x 84 in. [2.13 x 2.13 m]), 24 in. (0.61 m) high foundation block. The effective height of the specimen column,
from its base to the level where lateral load is applied, is
64 in. (1.625m), giving it an aspect ratio of L/D = 4. The
9.75in. (0.25 m) extension accommodates the installation of
the 0.5in. (13 mm) thick, 16 in. (0.4 m) tall steel jacket to
attach the actuators.
The column has 12 longitudinal Grade 60 (nominal
yield stress in tension is 60 ksi [420 MPa]) No. 4 [13]
reinforcing bars placed around its perimeter. The transverse steel reinforcement is a high-strength A82 (nominal
yield stress in tension is 80 ksi [550 MPa]) W3.5 (0.211in.
24
[5.4mm]) continuous spiral with a center-to-center spacing

of 1.25 in. [31.75 mm]. The concrete cover is 0.5 in.
(13mm). The specified unconfined compressive strength of
the concrete was 5 ksi (34.5 MPa). Table 2 shows the specified and the actual strengths of the longitudinal steel, the
spiral steel, and the concrete.
Loading protocol
The loading pattern for the Base specimens was selected
to represent, as closely as possible, the motion experienced by a column of the prototype overpass bridge in an
earthquake. The motion of the bridge column excited by
different three-component ground motions is examined
using a finite element model of the prototype bridge made
in OpenSees.9 Two ground motion suites with 20 records per
suite, representing near-field and far-field ground motions,
were used. The displacement orbits of the tops of the prototype bridge columns were traced during the nonlinear
time history response analyses for the 40 selected ground
motions. Because of the different bridge column boundary
conditions for bending in the longitudinal (fixed-fixed) and
transverse (fixed-free) directions of the bridge, appropriately scaled displacement histories applied to the cantilever
bridge column model would not reproduce the deformation
state of the prototype. To achieve the close correspondence
of deformation states between the model and the prototype,
the displacement history of the prototype bridge column
had to be normalized by its yield displacements, different in
different bridge directions. Therefore, the displacement orbits
applied on the model were expressed in terms of displacement ductility. Although the displacements of the tops of the
bridge columns were larger in the transverse direction than
in the longitudinal bridge direction, the ductility orbits were
proportional in the two bridge directions for most of the
Table 2Bridge column specimen material properties

Specified1, ksi (MPa)
Actual, ksi (MPa)
Material
Yield
Ultimate
Maximum stress
Yield
Ultimate
Test
Maximum stress
Steel, longitudinal
60 (420)
80 (550)
70.7 (487)
120 (830)
Steel, spiral
80 (550)
95 (655)
106 (730)
Concrete
5.0 (34.5)
Base15-L
5.05 (34.82)
Base30-L
4.96 (34.2)
Base45-L
5.09 (35.09)
Base0
5.48 (37.8)
Fig. 2Geometry and reinforcement of Base specimens.
Fig. 3Displacement orbits at top of bridge column: (a)

absolute displacements; and (b) normalized displacements.
ground motions. Normalization of the displacement orbits
for one ground motion is shown in Fig. 3.
A circular loading pattern was selected for this experimental program because this loading pattern imposes larger
sustained displacement ductility demand than was observed
in any of the considered ground motions.6 Given that the
goal of the experimental study is to establish the remaining
load-carrying capacity of damaged bridge columns, imposing

the demand with a circular loading pattern is conservative.
The selected circular loading pattern is defined by two cycles
at each displacement level. In the first cycle, starting from
the initial position O, the specimen control point (the center
of the column cross section at the level of the actuators) is
displaced toward position A, followed by motion that traces
a full circle clockwise until point B (black line in Fig. 4).
The specimen control point is then moved back to the initial
position O to finish the first cycle. The second cycle is a
counterclockwise path O-C-D-O (gray line in Fig. 4).
The maximum displacement ductility demand imposed on
a column specimen in this study was set at 4.5. It is selected
to be slightly larger than Caltrans SDC1 design target
displacement ductility of 4 and is on the conservative side.
Caltrans SDC1 design is based on experimental evidence
that well-confined reinforced concrete circular column can
sustain a displacement ductility demand of 4 without developing significant flexural or shear damage.10-12 Two additional
displacement ductility demand targets of 3.0 and 1.5 were
selected to uniformly sample the demand space and evaluate
the remaining axial capacity of less-damagedspecimens.
25
Table 3Displacement ductility levels of primary

cycles
Fig. 4Bidirectional displacement pattern.

The magnitudes of displacement demand increments for
the quasi-static tests were defined following the recommendations in ACI 374.1-0513 and SAC/BD-00/1014 for
a major far-field earthquake event. For the Base45 specimen, the increments in the magnitude of the displacement
ductility were: 0.08, 0.2, 0.4, 1.0, 1.5, 2.0, 3.0, and 4.5. The
pre-yield displacement levels include: a displacement level
prior to cracking; two levels between cracking and yielding;
and a level approximately corresponding to the first yield
of the longitudinal reinforcement. After the yield level,
the displacement ductility magnitude of each subsequent
primary cycle is 1.25 to 1.5 times larger than its predecessor
to provide data of the damage accumulation. The selected
two-cycle displacement pattern provides data on specimen
strength degradation due to sustained displacement demand.
After yielding, each primary displacement ductility demand
level was followed by a small displacement level equal to
one-third of the previous primary displacement level to evaluate specimen stiffness degradation. The displacement histories for the Base15 and Base30 specimens were obtained by
scaling the displacement history for the Base45 specimen by
0.3 and 0.6, respectively. This way, the number of primary
cycles in the loading history was the same for all tests to
maintain similitude with respect to the duration and number
of excursions imposed by real ground motions. Displacement ductility levels of primary cycles for the three lateral
displacement tests are given in Table 3. After completing the
cycles at the target displacement ductility level, the specimens are cycled through a series of small deformation cycles
decreasing in magnitude to zero to eliminate residual lateral
forces and deformations and recenter the specimens. This
was necessary for the subsequent axial load capacity tests on
damaged specimens.
The lateral deformation tests were conducted with the
column specimen under a constant axial load equal to 10%
of the column nominal axial load capacity. This axial load
magnitude is consistent with typical bridge column gravity
load magnitudes, and slightly larger than the gravity load
magnitude in the columns of the prototype bridge.
26
Cycles
Base15
Base30
Base45
Cycle 1
0.02
0.05
0.08
Cycle 2
0.06
0.10
0.20
Cycle 3
0.12
0.25
0.40
Cycle 4
0.30
0.60
1.00
Cycle 5
0.45
1.00
1.50
Cycle 6
0.60
1.25
2.00
Cycle 7
1.00
1.80
3.00
Cycle 8
1.50
3.00
4.50
Test setup
In the first phase of the test, lateral and axial loads were
applied at the top of the column. The lateral displacement pattern was applied using the two servo-controlled
hydraulic actuators, as shown in Fig. 5. An axial load of
100 kip (445 kN), equal to 10% of the columns nominal
axial load capacity, was applied through a spreader beam
using pressure jacks and post-tensioning rods placed on
each side of the column (Fig. 5). Spherical hinges (threedimensional swivels) were provided at both ends of the rods
to avoid bending of the rods during circular motion of the
column top in the horizontal plane. A hinge connection (twodimensional hinge) was placed between the spreader beam
and the column such that the spreader beam remained horizontal in the plane of the rods during lateral motion of the
column to avoid buckling of the rods. Geometry of the axial
load application apparatus was monitored throughout the test
in order to subtract the horizontal components of the force in
the post-tensioned rods from the forces applied by the actuators and compute the actual lateral resistance of the column.
In the second phase of the test, the three laterally damaged
column specimens and one undamaged column specimen
were compressed axially to induce axial failure in the
columns. A compression-tension axial load machine with a
capacity of 1814 tonnes (4 million lb) and a constant rate of
loading was used to accomplish this (Fig. 6). Longitudinal
reinforcement strain measurements were used to evaluate
presence of bending moment in the specimens during the
axial load test based on which the extent of geometric imperfections was estimated.
ANALYTICAL INVESTIGATION
The experimental results, the hysteretic curves from quasistatic tests, and the axial force-deformation responses from
the compression tests were numerically simulated using the
force-based fiber beam-column element15 of OpenSees.9
The force-based beam-column element is a line element
discretized using the Gauss-Lobatto integration scheme
with the integration points at the ends of the element and
along the element length. Fiber cross sections are assigned
to the integration points. The cross sections of the element
are represented as assemblages of longitudinally oriented,
unidirectional steel and concrete fibers. Each material in the
cross section has a uniaxial stress-strain relation assigned
Fig. 5Lateral test setup: (a) plain view; (b) elevation (A-A); and (c) photo of test setup.
to it. The deformation compatibility of the cross-section
fibers is enforced assuming that plane sections remain plane
afterdeformation.
In a flexibility-based formulation of this element, nodal
loads imposed on the element ends are used to calculate
axial force and moment distribution along the length of the
element. Given the moment and axial load values at each
integration point, the curvature and the axial deformation of
a section are subsequently computed. Because the response
of the cross-section fiber materials may be nonlinear, deformation state determination of the cross section may be iterative. The deformation of the element is finally obtained
through weighted integration of the section deformations
along the length of the member.
A non-shear-critical column with hardening section
behavior was modeled using five integration points16 along
its length. The cross sections of the beam-column element
had 132 fibers (24 for unconfined cover, 96 for confined
core, and 12 for reinforcing steel) distributed nonuniformly16 and arranged as shown in Fig. 7. To model the reinforced concrete section, the fiber section that accounts for
the axial-bending interaction was divided into three parts:
concrete cover; concrete core; and reinforcing steel. Fibers
of the concrete cover (unconfined concrete) and concrete
core (confined concrete) were modeled using the OpenSees
Fig. 6Axial test setup.

Concrete01 uniaxial material that uses the Kent-Scott-Park
model17 to represent the stress-strain relationship of concrete
in compression. Reinforcing steel fibers (longitudinal bars)
were modeled using the OpenSees Steel02 uniaxial mate27
Fig. 7Nonuniform arrangement of fibers in column

section.16
Fig. 9Accuracy in prediction of column axial strength as

function of ratio of initial to post-peak degrading slope of
confined concrete (k).
Table 4Steel02 material model parameters
Material
Reinforcing
steel
fy, ksi (MPa) Es, ksi (GPa)

70.7 (490)*
R0
cR1
cR2
29,000 (200) 0.025 15 0.925 0.15
From coupon tests.
Table 5Concrete01 material model parameters

Material
fc
Concrete cover
fc
Concrete core
fcc
fcu
cu
2fc/Ec
0.0055
2fcc/Ecc
0.2fcc
cu
From test results on concrete cylinders (Table 2).
Fig. 8Calibration of the Concrete01 material using data

from compression test of concrete cylinders.
rial that uses the Giuffre-Manegotto-Pinto uniaxial strainhardening material model.18 Transverse reinforcement was
not modeled directly but its effect was accounted for through
uniaxial stress-strain relationship of the confined concrete
core19 assigned to core fibers.
The parameters of the Steel02 and Concrete01 uniaxial
materials are given in Tables 4 and 5, respectively. The initial
moduli of elasticity used to model plain Ec and confined
concrete Ecc were calibrated from the concrete cylinder
compression tests performed on the on the day of the tests
(for example, Fig. 8). To define the confined concrete model,
the maximum compressive strength fcc was calculated using
Manders equations,19 the strain at the maximum compressive strength 0 was calculated from the initial modulus of
elasticity for the Concrete01 material (Ecc = 2fcc/0), the
concrete crushing strength fcu was set to 0.2fcc1, and the
strain at crushing strength of concrete cu was calculated
using Eq. (1) and is a function of the post-peak degrading
slope of concrete (kEcc). Equation (1) can only be used in
conjunction with Concrete01 material, as it is derived for the
strain at the maximum compressive strength 0 of 2fcc/Ecc

28
e cu =
f cc (1 + 2k ) f cu
(1)
kEcc
Ec, Ecc are initial moduli of elasticity (calibrated to match test results of concrete
cylinders).
Equation from Mander et al.19
Equation (1).
The post-peak degrading slope of concrete fibers has

a significant effect on the accuracy of prediction of the
remaining axial load strength of columns with the earthquake type of damage. If, during a lateral load test simulation, a concrete fiber strain exceeds the strain at the
maximum compressive strength (0), the peak strength that
fiber can attain during the axial loading simulation phase is
the strength that corresponds to the maximum strain reached
during the lateral load test. This strength is smaller than the
peak strength and depends on the post-peak degrading slope
of the concrete fiber stress-strain model. Furthermore, the
post-peak behavior is different for different fibers of the
damaged section. While the post-peak degrading slope for
the fibers of the concrete cover can be calibrated from the
concrete cylinder compression tests (Fig. 8), the post-peak
degrading slope for the fibers of the concrete core can be
calibrated from the axial compression tests of the laterally
damaged column specimens. For the tested column specimens and concrete modeled with the Concrete01 OpenSees
material, the post-peak degrading slope kEcc of 0.014Ecc was
found to provide the best predictions of the axial load capacities (Fig. 9).
The specimen models were cantilevers with displacements restrained to zero at the bottom node. The two horiACI Structural Journal/January-February 2015
Fig. 10Lateral force-deformation response curves in two major directions, and state of specimens after the quasi-static tests:
(a) Base15; (b) Base30; and (c) Base45.
zontal displacements at the top of the specimens matched the
displacements commanded during the test. The vertical force
at the top of the element matched the vertical force applied
and measured during the tests. The moments on the rotation
and torsion degrees of freedom on the top node of the model
were set to zero. The response of the specimen model was
computed using nonlinear analysis, Newton-Raphson integration algorithm, and geometric transformation to account
for P- effect.
RESULTS AND DISCUSSION
Phase 1: Lateral displacement tests
The performance of the Base specimens during the bidirectional quasi-static tests is presented in Fig. 10. Experimental and analytical lateral force-displacement response
curves for the two major directions of loading (X and Y) are
accompanied by the final damage state of specimens for the
three lateral tests. Responses obtained from the analytical
model are in good agreement with the experimental results.
The Base15 specimen was laterally displaced up to the

displacement ductility level of 1.5. Longitudinal reinforcement yielded first during the 1.0 ductility cycle. At the end
of the test, the specimen was only slightly cracked. The horizontal cracks, uniformly distributed along the bottom half
of the column, were less than 1/32 in. (0.8 mm) wide and
approximately 6 in. (152 mm) apart (Fig. 10(a)). A bridge
column in such a light damage state would be classified as
being in Damage State 0 (defined by Mackie et al.20) and
would likely require no repairs.
The Base30 specimen experienced significant yielding
and strain hardening of the longitudinal column reinforcement and initiation of spalling of concrete in the plastic
hinge region (Fig. 10(b)). After the longitudinal reinforcement began to yield (at a nominal displacement ductility of
0.9), the lateral resistance of the specimen slightly increased
due to strain hardening of reinforcing bars, while its stiffness decreased with each subsequent test cycle. In the plastic
hinge region of the column (the bottom 12 in. [305mm]), the
29
distance between the cracks was 3 in. (76 mm) on average

and the maximum width of the cracks during the test was
approximately 1/16 in. (1.6 mm). Outside the plastic hinge
region, the distance between the cracks was 6 in. (152 mm)
on average with the widths of the cracks less than 1/32 in.
(0.8mm). Figure 10(b) shows horizontal cracks, vertical
cracks, and some spalling of concrete at the bottom 8 in.
(203 mm) of the column at the end of the test. Such a
moderate damage state would put the column in Damage
State 1 (defined by Mackie et al.20). The bridge columns
with such earthquake damage are likely to require repairs
such as epoxy injection into the plastic hinge cracks and
coverpatching.
The target displacement ductility demand imposed on specimen Base45 (4.5) slightly exceeds the Caltrans SDC design
target (4.0). This specimen experienced extensive yielding of
the reinforcing steel, spalling of the concrete cover, as well
as a crushing and reduction in volume of the concrete core
in the plastic hinge region. First yielding of a reinforcing
bar occurred at a displacement corresponding to nominal
displacement ductility of 0.75. The specimen response was
highly nonlinear (Fig. 10(c)), with the expected gradual stiffness degradation and gradual strength increase. Based on
the crack distribution along the height of the column during
the test, the column was divided into three regions: 1) the
plastic hinge region (the bottom 12 in. [305 mm] of column);
2) the intermediate region (12 in. [305 mm] of the column
next to the plastic hinge region); and 3) the elastic region
(the top 40 in. [1.02 m] of the column). In the plastic hinge
region, the distance between the cracks was 3 in. (76 mm)
on average, and the maximum width of the cracks during
the test was approximately 1/8 in. (3.2mm). Very extensive spalling of concrete and a reduction in volume of the
concrete core were observed. In the intermediate region, the
distance between the cracks was 4in. (102mm) on average,
with the widths of the cracks less than 1/16 in. (1.6 mm). In
the elastic region, the distance between the cracks was 6 in.
(152 mm) on average, with the widths of the cracks less than
1/32 in. (0.8 mm). Such column damage would be classified
into Damage State 2 (defined by Mackie et al.20), requiring
significant repairs but not requiring replacement.
To analyze bridges for an aftershock, it is important to
know effective stiffness keff of bridge columns after the main
shock. This stiffness is computed using the response data
measured during the small-displacement test cycles that
followed the primary cycles. It represents the tangent slope
at zero force of force-displacement curve for the smalldisplacement test cycles. The effective stiffness at yield
keff,y of tested columns, representing the slope of forcedisplacement curve between origin and the point designating
the first reinforcing bar yield, is used as a reference to measure
stiffness degradation during the quasi-static tests. The ratio
of column effective stiffness over column stiffness at yield
keff/keff,y is given in Table 6 for each specimen. The effective stiffness of the damaged column decreases such that, at
displacement ductility level of 4.5, it is approximately half
that of the effective stiffness of the same column at yield.
30
Table 6Ratio of column effective stiffness to

column stiffness at yield
Ductility
Base15
Base30
Base45
1.0
1.00
1.00
1.00
1.2
0.85
1.5
0.83
0.75
1.8
0.73
2.0
0.63
3.0
0.54
0.52
4.5
0.44
Phase 2: Axial load tests

The experimentally measured and numerically simulated
axial force-deformation curves for one undamaged and
three damaged specimens are shown in Fig. 11. Because the
tests are performed using a force-controlled compression
machine, the axial force-displacement relationships are realistic up to the peak force point. The experimental and analytical axial load strengths, the remaining axial strength after
damage was induced during the Phase 1 of lateral deformation tests, and the errors in predicting the axial strengths are
summarized in Table 7.
Testing of the Base0 column specimen was performed
to establish the axial strength of an undamaged column
specimen: it was 1459 kip (6490 kN). The axial failure
resulted from the formation of the shear failure plane in the
bottom half of the Base0 specimen column (Fig. 11(a)). The
analytical model predicted the axial strength of the undamaged specimen to be 1446 kip (6434 kN) (error is 0.9%).
An equally accurate prediction of the axial strength can be
achieved using Eq. (2)

Po = fcc (Aeff Ast) + fy Ast (2)
if Manders equations19 are used to calculate the area of the

confined core Aeff and the compressive strength of confined
concrete fcc, based on measured strengths of plain concrete,
reinforcing bars, and spiral. Using Eq. (2), the axial strength
of the column specimen, Po, is estimated to be 1455 kip
(6472 kN), resulting in the ratio of estimated to measured
strength (Po/Pm) of 0.997. However, if confinement of the
column is not accounted for and the axial strength is calculated following Eq. (3) (per ACI 31821 and Caltrans BDS22)
Pn = 0.85 [0.85 fc (Ag Ast) + fy Ast] (3)
The estimated axial strength of the column is significantly
smaller than the measured axial strength (Pn/Pm = 0.57).
The remaining axial load strength of the Base15 column
specimen was 1137 kip (5057 kN)78% of the original
axial strength. Longitudinal reinforcement strain measurements during the axial load tests indicated a bending moment
corresponding to a lateral drift of approximately 1%. A posttest inspection of the specimen indicated that the specimen
was not accurately leveled when it was installed for axial
load testing. The resulting second-order bending moment
and the corresponding shear caused a shear crack in the top
half of the Base15 column specimen (Fig. 11(b)). To numerically simulate the axial strength of the damaged column with
the presence of residual drift, the top of the column model
was first laterally displaced following the loading pattern of
the quasi-static test, then laterally displaced to the observed
drift ratio of 1%, and lastly was axially compressed (pushunder analysis) to induce the axial failure. The analytically
predicted axial strength of the Base15 column specimen was
1141 kip (5075 kN)only 0.38% greater than the experimentally measured.
specimen was 1355 kip (6027 kN)93% of the original
axial strength. The specimen was properly leveled before the
axial load tests; lateral drift was not present during this phase
of testing. The axial failure resulted from the formation of
the shear failure plane in the bottom half of the Base30 specimen column (Fig. 11(c)). The analytically predicted axial
strength was 1217 kip (5455 kN)10.2% smaller than the
experimentally measured.
specimen was 1170 kip (5204 kN)80% of the original axial
Table 7Remaining axial strengths following
quasi-static tests: experimental versus analytical
Test
Experiment,
kip (kN)
Analytical,
kip (kN)
Numerical
error, %
P/Po
Base0
1459 (6490)
1446 (6434)
0.9
1.0
Base15
1137 (5058)
1141 (5075)
0.38
0.78
Base30
1355 (6027)
1217 (5413)
10.17
0.93
Base45
1170 (5204)
1173 (5217)
0.24
0.80
strength. The specimen was properly leveled before the axial

load tests but the head of the testing machine was not accurately attached to the specimen: it had a small angle relative
to the specimen. Thus, the force imposed on the column had
two components: a dominant axial component and a small
horizontal component. This initiated a failure slightly earlier
than what would be anticipated if there were no flaws in the
test setup. The axial failure started at the top of the column
and progressed toward the bottom of the column, resulting
in the formation of the shear failure plane along the total
height of the column (Fig. 11(d)). The misalignment of the
test machine head was not measured prior to test and therefore could not be included in analytical simulation of the
axial load test. Although the analytical model provides a
good estimate (1173 kip [5217 kN], 0.24% error), it is hard
to anticipate the magnitude of the error if the flaws in the test
setup were included in the analysis.
Axial load degradation curve with respect to displacement
ductility is developed based on experimental results. It is a
bilinear function (Fig. 12) fitted through experimental data
from compression tests on column specimens with lateral
damage and no geometric imperfections of the damaged
specimens. It includes the result of the Base45 specimen,
thus providing a conservative estimate of the residual axial
strength. It is assumed that there are no losses in the axial
strength in the columns if the lateral displacement ductility
is less than 1.5. This assumption is based on the damage
state of the Base15 specimen after the lateral test (Damage
State0) and on minor loss in axial strength (7%) of the
Base30 specimen. The bilinear fit is
Fig. 11Axial force-displacement relationships and state of specimens after axial load tests: (a) Base0; (b) Base15; (c) Base30;
and (d) Base45.
31
Fig. 12Column axial load capacity degradation data and

a bilinear fit (Eq.(4)).

if < 1.5
P 1
(4)
=
Po 1.09 0.0615 if 1.5
CONCLUSIONS
Tests designed to evaluate the axial load capacity of tall
modern bridge columns damaged in bidirectional quasistatic cyclic tests up to nominal displacement ductility levels
of 1.5, 3, and 4.5 were performed. The following conclusions are drawn:
1. The axial strength and stiffness of a column degrade
with the increase in the amount of damage induced by lateral
displacement of the column. Well-confined modern bridge
columns with no residual post-earthquake lateral drifts
lose approximately 20% of their axial load capacity after
sustaining displacement ductility demand of 4.5, which is
slightly larger than the Caltrans SDC design target displacement ductility demand of 4.0. Therefore, modern bridge
columns designed according to Caltrans SDC1 will not
experience a significant loss of axial load-carrying capacity
after a design-level earthquake. No axial load capacity
loss is expected for displacement ductility demands less or
equal to 1.5. Axial load capacity loss may conservatively
be assumed to vary linearly with increasing displacement
ductility demand. The residual post-earthquake displacements have a significant effect on the axial capacity of the
column: the column that sustained the displacement ductility
demand of 1.5 with no significant local damage but with a
residual lateral drift of 1% experienced a reduction in axial
load capacity of 22%.
2. Damage states observed during bidirectional lateral
displacement tests correspond well to Damage State descriptions defined by Mackie et al.20 Namely, virtually no damage
observed at displacement ductility demand of 1.5 corresponds to Damage State 0; moderate damage characterized
by cover spalling and pronounced yielding of longitudinal
reinforcement at displacement ductility demand of 3.0 corresponds to Damage State 1; significant damage to the cover
and core concrete, very pronounced yielding of longitudinal
and transverse reinforcement, however without any rein-
32
forcement fractures, at ductility demand of 4.5 corresponds

to Damage State 2. These data can be used to calibrate
repair cost and time models for modern reinforced concrete
bridgecolumns.
3. The effective lateral stiffness of a damaged column
decreases with increasing displacement ductility demand.
For example, the effective stiffness of the column that experienced a displacement ductility demand of 4.5 is slightly less
than half of the undamaged column effective stiffness. This
information can be used to estimate the dynamic characteristics of damaged bridges for aftershock response analysis.
4. The circular bidirectional displacement ductility pattern
developed in this study imposes sustained constant displacement ductility demands on the column that very likely
exceed the displacement ductility demands imposed by the
recorded ground motions. Therefore, this load pattern is a
conservative estimate of actual ground motion demands and
could be used to investigate residual post-earthquake capacities or function capabilities of damaged structures.
5. A non-shear-critical bridge column in the range of
strain-hardening response can be modeled using fiber
cross-section force-based beam-column element with
distributed plasticity. Two OpenSees material models are
recommended to model the steel and concrete unconfined
cover and confined core fibers that define a cross section
of the element: the Steel02 material model for reinforcing
steel, and the Concrete01 material model for concrete fibers.
Manders equations19 are recommended to calculate the
compressive strengths of confined concrete. If the material
test data for concrete are available, it is recommended to calibrate the Concrete01 material model to match the test data.
The post-peak degrading slope of the confined concrete core
(kEcc) has a great influence on the accuracy of the response
predictions. For the tested column specimens, the post-peak
degrading slope kEcc of 0.014Ecc was found to provide accurate predictions of the axial load capacities.
The analytical models of the bridge columns presented in
this paper were validated through hybrid simulations of the
seismic response of an entire bridge followed by the truck
load and the axial compression tests of damaged bridge
columns.23 Validated model of a typical bridge was used
in an extensive parametric study to evaluate its post-earthquake truck load capacity. The parametric study examined
the effects of different ground motions and bridge modeling
parameters, including boundary conditions imposed by
the bridge abutments, the location of the truck load on
the bridge, the amount of bridge column damage, and the
amount of bridge column residual drift. Envelopes of bridge
responses developed for ranges of the considered parameters
were used to evaluate bridge safety and ability to function
following an earthquake.6
AUTHOR BIOS
Vesna Terzic is an Assistant Professor at the Department of Civil Engineering and Construction Engineering Management, California State
University, Long Beach, CA. She received her BS from the University of
Belgrade, Belgrade, Serbia; her MS from Saints Cyril and Methodius
University of Skopje, Skopje, Macedonia; and her PhD from the University of California, Berkeley, Berkeley, CA. Her research interests include
performance-based design and evaluation of civil infrastructure.
ACI member Bozidar Stojadinovic is a Professor and Chair of structural

dynamics and earthquake engineering at the Department of Civil, Environmental and Geomatic Engineering at Swiss Federal Institute of Technology
(ETH), Zurich, Switzerland. He received his Dipl. Ing. degree in civil
engineering from the University of Belgrade; his MS in civil engineering
from Carnegie Mellon University, Pittsburgh, PA; and his PhD in civil
engineering from the University of California, Berkeley. He is a member
of ACI Committees 341, Earthquake-Resistant Concrete Structures; 349,
Concrete Nuclear Structures; and 374, Performance-Based Seismic Design
of Concrete Structures; and Joint ACI-ASCE Committee 335, Composite
and Hybrid Structures. His research interests include probabilistic performance-based seismic design of transportation and energy infrastructure.
ACKNOWLEDGMENTS
These data and findings presented herein stem from the work supported by
the California Department of Transportation through Project 04-EQ042 and
the Pacific Earthquake Engineering Research (PEER) Center. This support,
as well as engineering advice from M. Mahan and C. Whitten of Caltrans and
S. Takhirov of the PEER Center, is gratefully acknowledged. Any opinions,
findings, conclusions, or recommendations expressed in this material are
those of the authors and may not be those of the project sponsors.
REFERENCES
1. Caltrans, Seismic Design Criteria, State of California Department of
Transportation, Sacramento, CA, 2006.
2. Fenves, G. L., and Ellery, M., Behavior and Failure Analysis of a
Multiple-Frame Highway Bridge in the 1994 Northridge Earthquake,
Report PEER 98/08, Pacific Earthquake Engineering Research Center,
University of California, Berkeley, Berkeley, CA, 1998.
3. Arici, Y., and Mosalam, K., System Identification of Instrumented
Bride Systems, Earthquake Engineering & Structural Dynamics, V. 32,
No. 7, 2003, pp. 999-1020. doi: 10.1002/eqe.259
4. Mackie, K., and Stojadinovic, B., Fragility Basis for California
Highway Overpass Bridge Decision Making, Report PEER 2005/02,
Pacific Earthquake Engineering Research Center, University of California,
Berkeley, Berkeley, CA, 2005.
5. Saiidi, M. S., and Ardakani, S. M., An Analytical Study of Residual
Displacements in RC Bridge Columns Subjected to Near-Fault Earthquakes, Bridge Structures, V. 8, 2012, pp. 35-45.
6. Terzic, V., and Stojadinovic, B., Post-Earthquake Traffic Capacity of
Modern Bridges in California, Report PEER 2010/103, Pacific Earthquake
Engineering Research Center, University of California, Berkeley, Berkeley,
CA, 2010.
7. Ketchum, M.; Chang, V.; and Shantz, T., Influence of Design
Ground Motion Level on Highway Bridge Costs, Report PEER 6D01,
8. Mackie, K., and Stojadinovic, B., Fragility Basis for California
Highway Overpass Bridge Decision Making, Report PEER 2005/02,
9. McKenna, F., and Fenves, G. L., Open System for Earthquake Engineering Simulation (OpenSees). Pacific Earthquake Engineering Research
Center, University of California, Berkeley, Berkeley, CA, 2004.
10. Lehman, D. E., and Moehle, J. P., Seismic Performance of Well-Confined Concrete Bridge Columns, Report PEER 1998/01, Pacific Earthquake Engineering Research Center, University of California, Berkeley,
Berkeley, CA, 1998.
11. Lehman, D. E., and Moehle, J. P., Behavior of Reinforced Concrete
Bridge Columns Having Varying Aspect Ratios and Varying Lengths of
Confinement, Report PEER 2000/08, Pacific Earthquake Engineering
Research Center, University of California, Berkeley, Berkeley, CA, 2000.
12. Hose, Y., and Seible, F., Performance Evaluation Database for
Concrete Bridge Components and Systems under Simulated Seismic
Loads, Report PEER 1999/11, Pacific Earthquake Engineering Research
13. ACI Committee 374, Acceptance Criteria for Moment Frames
Based on Structural Testing (ACI 374.1-05) and Commentary, American
Concrete Institute, Farmington Hills, MI, 2006, 9 pp.
14. Krawinkler, H.; Gupta, A.; Medina, R.; and Luco, N., Loading
Histories for Seismic Performance Testing SMRF Components and Assemblies, Report No. SAC/BD-00/10, SAC Joint Venture, 2000.
15. Taucer, F. F.; Spacone, E.; and Filippou, F. C., A Fiber BeamColumn Element for Seismic Response Analysis of Reinforced Concrete
Structures, Report No. UCB/EERC-91/17, Pacific Earthquake Engineering
Research Center, University of California, Berkeley, Berkeley, CA, 1991.
16. Berry, M. P., and Eberhard, M. O., Performance Modeling Strategies for Modern Reinforced Concrete Bridge Columns, Report PEER
2007/07, Pacific Earthquake Engineering Research Center, University of
California, Berkeley, Berkeley, CA, 2007.
17. Kent, D. C., and Park, R., Flexural Members with Confined
Concrete, Journal of Structural Engineering, ASCE, V. 97, 1971,
pp.1969-1990.
18. Menegotto, M., and Pinto, P. E., Method of Analysis for Cyclically
Loaded R.C. Plane Frames Including Changes in Geometry and Nonelastic
Behaviour of Elements under Combined Normal Force and Bending,
Proceedings of the Symposium on the Resistance and Ultimate Deformability of Structures Acted on by Well Defined Repeated Loads, International
Association for Bridge and Structural Engineering, Zurich, Switzerland,
1973, pp. 15-22.
19. Mander, J. B.; Priestley, M. J. N.; and Park, R., Theoretical
Stress-Strain Model for Confined Concrete, Journal of Structural Engineering, ASCE, V. 114, No. 8, 1988, pp. 1804-1826. doi: 10.1061/
(ASCE)0733-9445(1988)114:8(1804)
20. Mackie, K.; Wong, J. M.; and Stojadinovic, B., Integrated Probabilistic Performance-Based Evaluation of Benchmark Reinforced Concrete
Bridges, Report PEER 2007/09, Pacific Earthquake Engineering Research
Farmington Hills, MI, 2005, 430 pp.
22. Caltrans, Bridge Design Specifications, State of California Department of Transportation, Sacramento, CA, 2004.
23. Terzic, V., and Stojadinovic, B., Hybrid Simulation of Bridge
Response to Three-Dimensional Earthquake Excitation Followed by a
Truck Load, Journal of Structural Engineering, ASCE, V. 140, 2013. doi:
10.1061/(ASCE)ST.1943-541X.0000913
33
NOTES:
34
TECHNICAL PAPER
Title No. 112-S04
Shear Behavior of Reinforced Concrete Columns with

High-Strength Steel and Concrete
by Yu-Chen Ou and Dimas P. Kurniawan
Eight shear-critical, large-scale, high-strength reinforced
concrete columns under low axial load were tested. The specified
concrete compressive strengths were 70 and 100 MPa (10,000 and
14,500psi). The specified yield strengths were 685 and 785MPa
(100,000 and 114,000 psi) for the longitudinal and transverse
reinforcement, respectively. Test results show that all the columns
exhibited shear failure before longitudinal reinforcement yielding.
The transverse reinforcement did not yield at the peak applied load
for all the columns. In total, test data for 43 high-strength columns
from this study and literature were collected. A comparison to
the ACI 318 shear-strength equations show that the ACI simplified shear-strength equation provided conservative estimation for
all the columns. However, the ACI detailed shear-strength equation yielded unconservative prediction for 19 columns. This study
proposes modifications to the ACI detailed shear-strength equation
to address this issue.
Keywords: column(s); cyclic loading; double curvature; high-strength
concrete; high-strength reinforcement; reinforced concrete.
INTRODUCTION
High-strength steel and concrete have gained increasing
attention recently in reinforced concrete (RC) buildings due
to the need to limit the size of lower-story columns in highrise buildings to increase available floor area. Moreover, the
use of high-strength steel reduces reinforcement congestion
in the plastic hinge regions in seismic design. With recent
advancements in material production technology in Taiwan,
deformed reinforcement SD685 with a specified yield
strength of 685 MPa (100,000 psi) for longitudinal reinforcement (Fig. 1(a)), deformed reinforcement SD785 with specified yield strength of 785 MPa (114,000 psi) for transverse
reinforcement (Fig. 1(b)), and high-strength concrete with
a specified compressive strength of 100 MPa (14,500 psi)
are commercially available. The SD685 reinforcement has
a lower and upper limit on actual yield strength685 and
785 MPa (100,000 and 114,000 psi), respectivelyand a
minimum ratio of actual ultimate strength to actual yield
strength1.25, conforming to the ACI 318 seismic design
provisions.1 It also has a lower limit of 0.014 on the strain
corresponding to a stress equal to the upper limit of yield
strength (ensuring a sufficient yield plateau), and a lower limit
of 0.1 on elongation (Fig. 1(a)). The SD785 has requirements
on minimum yield strength and ultimate strength785and
930 MPa (114,000 and 135,000psi), respectivelyand a
lower limit of 0.08 on elongation (Fig. 1(b)).
Such high-strength materials, when used in columns
for shear design, are not allowed by the ACI 318 Code1
to use their full strengths. The yield strength of deformed
reinforcing bars for shear design is limited to 420 MPa
Fig. 1High-strength reinforcement: (a) SD685; and (b)

SD785. (Note: 1 MPa = 145 psi.)
(60,900psi) (ACI 318-11, 11.4.2). Moreover, concrete
compressive strength for shear design of columns is limited
to 70 MPa (10,000 psi) (ACI 318-11, 11.1.2). Note that this
limit can be removed for beams with the minimum web reinforcement (ACI 318-11, 11.1.2.1), but not for columns.
The purposes of the yield strength limit in the design of
shear reinforcement are to control diagonal crack width
and to ensure bar yielding before shear failure.2 Lee et al.,2
who tested 27 beams with fyt of 379 to 750 MPa (55,000
to 109,000 psi), concluded that maximum crack width and
fyt are not strongly correlated. Test results for 87 beams
MS No. S-2013-134.R1, doi: 10.14359/51686822, was received September 8,
2013, and reviewed under Institute publication policies. Copyright 2015, American
35
Table 1Specimen design
Column
Axial load
ratio, %
Transverse
reinforcing bar
spacing mm (in.)
A-1
Concrete compressive strength,

MPa (ksi)
fcs
fc
70 (10.15)
92.5 (13.41)
100 (14.5)
99.9 (14.49)
70 (10.15)
96.9 (14.05)
100 (14.5)
107.1 (15.53)
70 (10.15)
108.3 (15.71)
100 (14.5)
125.0 (18.13)
70 (10.15)
112.9 (16.37)
100 (14.5)
121.0 (17.55)
Longitudinal reinforcing bar D32

(No. 10) yield strength, MPa (ksi)
Transverse reinforcing bar D13

(No. 4) yield strength, MPa (ksi)
fyls
fyl
fyts
fyt
685 (100)
735 (106.6)
785 (114)
862 (125)
450 (17.72)
A-2
10
A-3
260 (10.24)
A-4
B-1
15
450 (17.72)
B-2
18
B-3
20
B-4
260 (10.24)
with fyt ranging from 484 to 1454 MPa (70,000 to 211,000

psi)2-4 indicated that beams may fail in shear before shear
reinforcement yielding when fyt is very high; for example,
fyt = 1454 MPa (211,000 psi), or when used with normalstrength concrete, for example, fyt 700 MPa (102,000 psi)
with fc< 40 MPa (5800 psi). Test results for 42 columns
with fyt ranging from 846 to 1447 MPa (123,000 to 210,000
psi)3-5 showed that only a few columns had shear failure
after shear reinforcement yielding. This suggests that a stress
limit should be imposed for shear reinforcement in columns.
The limit on fc in the ACI 318 Code is due to a lack
of test data and practical experience with fc 70 MPa
(10,000psi).1 In total, 42 beams specimens3,6-9 without
shear reinforcement and with high-strength concrete with
fc of 72.8 to 104.2 MPa (10,600 to 15,000 psi) were found
in literature. The ACI 318 simplified (Eq. (11-3)) and
detailed (Eq.(11-5)) shear-strength equations overestimated
measured diagonal cracking strength for two and four specimens, respectively. The overestimation was generally within
10%. When one considers measured ultimate shear strength,
only the shear strength of one specimen was overestimated.
In total, 35 column specimens3,5,10,11 that were designed with
high-strength concrete with fc of 73.5 to 130 MPa (10,700to
19,000 psi) and were reported with diagonal cracking
strength were found in literature. Most test results were
not evaluated using the ACI 318 shear-strength equations.
Moreover, most specimens were small. By testing eight
large columns, the objectives of this study are to examine
the following: the shear behavior of the columns under low
axial load with the high-strength steel and concrete materials
mentioned previously; the strains of shear reinforcement at
peak column shear; and the applicability of ACI 318 concrete
shear-strength equations for high-strength columns.
The use of high-strength materials can decrease the size
of lower-story columns in high-rise buildings, increasing
available floor area. Additionally, this can also decrease
the consumption of aggregate and steel, promoting environmental sustainability. By testing eight large-scale, highstrength concrete columns and by examining test data of
43high-strength columns, this research studied the applicability of shear-strength equations in the ACI 318 Code on
36
columns with material strengths higher than the code limits.

Based on the study results, a modification is proposed for
the detailed shear-strength equation of the ACI 318 Code for
high-strength columns.
EXPERIMENTAL PROGRAM
Specimen design
Eight large-scale columns with a clear height of 1800mm
(70.87 in.) and a square cross section of 600 x 600 mm
(23.62 x 23.62 in.) were tested. Table 1 and Fig. 2 show
column design details. The columns were designed to have
elastic shear failure; that is, shear failure before longitudinal reinforcement yielding. The columns were reinforced
with D32 (No. 10) longitudinal reinforcement and with D13
(No.4) transverse reinforcement. Each transverse reinforcement layer consisted of a closed hoop with 135-degree
hook anchorage and a tie along each principal direction
with 90-degree hook at one end and 135-degree hook at the
other end. The clear cover to the outer edge of the transverse reinforcement was 40 mm (1.57 in.). The SD685
high-strength deformed bars were used for the longitudinal
reinforcement and had specified and actual yield strengths of
685 and 735MPa (100,000 and 106,600 psi), respectively.
The SD785 high-strength deformed bars were used for the
transverse reinforcement and had specified and actual yield
strengths of 785 and 862 MPa (114,000 and 125,000 psi),
respectively. Three study variablesaxial load, concrete
compressive strength, and amount of transverse reinforcementwere examined. Two levels of axial load ratios10%
(Column A series) and 20% (Column B series)were examined. The axial load ratio is the ratio of applied axial load to
fcAg. The value of fc was obtained from the average of three
150 x 300 mm (6 x 12 in.) concrete cylinders. Two levels of
fcs70 and 100 MPa (10,000 and 14,500 psi)were investigated. Two levels of transverse reinforcement spacing
450 and 260 mm (17.72 and 10.24 in.)were studied. Note
that actual axial load ratios applied for Specimens B-1 and
B-2 were 15% and 18%, respectively, due to the inclusion of
relatively low cylinder-strength test results, which were later
eliminated when determining actual concrete compressive
strength.
Fig. 2Specimen design: (a) A-1, A-2, B-1 and B-2; (b) A-3, A-4, B-3 and B-4; and (c) cross section.
Test setup and loading protocol
The columns were tested at the National Center for
Research on Earthquake Engineering (NCREE), Taiwan,
using the Multi-Axial Testing System (MATS) (Fig. 3),
which has maximum axial and lateral load capacities of
60,000 and 7000 kN (13,489 and 1574 kip), respectively.
Lateral force was applied by hydraulic actuators placed at
the bottom of the MATS using a displacement control loading
history (Fig. 4). Axial loading was constant during testing.
TEST RESULTS AND DISCUSSION
Crack pattern and general behavior
Figures 5 and 6 show the lateral force-displacement relationships for Column A series (10% axial load ratio) and
Column B series (20% axial load ratio), respectively. The
P- effect has been removed from the figures. All specimens
exhibited elastic shear failure, where shear failure occurred
without any longitudinal reinforcement yielding, as indicated by strain gauge measurements. The Column B series,
with higher axial loading, was more brittle. Figure7 shows
the crack patterns of specimens at peak applied load. Flexural cracking appeared first at early drift. Shear cracking
initially developed as flexural-shear cracking with an angle
of approximately 45 degrees (relative to a columns longitudinal axis). As the load increased, web-shear cracks appeared.
The crack angles decreased as drift increased (Fig. 8). Each
point in this figure is the average of dominant diagonal
crack angles at that drift. At peak applied load, the average
diagonal crack angles (critical diagonal crack angle) were
27, 26, 27, and 24 degrees for Columns A-1, A-2, A-3, and
A-4, respectively. They were 25, 21, 20, and 21degrees for
Columns B-1, B-2, B-3, and B-4, respectively. The average
values of the critical crack angles were 26 and 22degrees
for Column A and B series, respectively. These angles were
smaller than 30 degrees, meaning the commonly used angle
Fig. 3 Multi-Axial Testing System (MATS).

of 45 degrees is conservative when predicting the shear
capacity of the columns. Explosive sounds were typically
heard during testing when critical cracks appeared. Figure
9 clearly shows several aggregates were cut through by a
crack. This is a typical feature of a failure surface for highstrength concrete. Figure 10 shows the damage distribution
of each specimen at test end. Note that Column A-1 testing
was terminated when lateral force dropped by more than 50%
(Fig. 5(a)), which was earlier than for other specimens, which
were tested until the lateral force dropped to nearly zero.
Column A series showed shear failure in the middle region
of a column, which belongs mainly to the B-region (regions
where conventional beam theory applies) as compared to the
D-region (regions where conventional beam theory does not
apply) at the two ends of a column (extending from each end
37
approximately one member depth). The failure patterns were

similar among Column A series. As the axial load increased,
the size of the failure region increased as observed from
Column B series. Moreover, the failure turned from gradual
spalling of concrete to a more brittle, sudden crushing failure
of concrete. Columns B-2 and B-3 showed signs of inclined
crushing failure (Fig.10(f) and (g)).
Fig. 4Loading protocol.
Shear contribution of steel and concrete

Figure 11 shows the relationships between maximum
stress of transverse reinforcement and column drift for all
columns. Transverse reinforcement stress increased slowly
in the early drift levels until diagonal cracks appeared. The
second column in Table 2 lists the drifts when diagonal crack
appeared; the third column lists the corresponding transverse reinforcement stress determined from the regression
line obtained from regression analysis of the data points
for each column in Fig. 11. The stress levels at diagonal
cracking were very small. After the shear cracks appeared,
stress increased rapidly. The increase rate of reinforcement
stress was higher in columns with high axial load (ColumnB
series) than those with low axial load (Column A series). The
sixth column in Table 2 lists drift at maximum applied load;
the seventh column lists the corresponding transverse reinforcement stress. Transverse reinforcement did not yield at
maximum applied shear for all columns. Note that reinforcement stress increased as drift increased. Because more transverse reinforcement delayed shear failure to a larger drift
(Columns A-3 and A-4 compared to Columns A-1 and A-2,
and similarly for Column B series), a higher level of transverse reinforcement stress developed. This can be observed
by comparing the transverse reinforcement stresses of
Columns A-3 and A-4 to those of Columns A-1 and A-2,
Fig. 5Hysteretic behavior of specimens with 10% axial load ratio: (a) A-1; (b) A-2; (c) A-3; and (d) A-4.
38
Fig. 6Hysteretic behavior of specimens with 20% axial load ratio: (a) B-1; (b) B-2; (c) B-3; and (d) B-4.
Fig. 7Crack pattern at the peak applied load for specimens: (a) A-1; (b) A-2; (c) A-3 (d) A-4 (e) B-1; (f) B-2; (g) B-3; and
(h) B-4.
and by comparing those of Columns B-3 and B-4 to those
of Columns B-1 and B-2. Based on the strain measurement
of longitudinal reinforcement, when longitudinal reinforcement of Column A and B series reached yielding, drifts were
predicted to be approximately 1.6% and 1.85%, respecACI Structural Journal/January-February 2015
tively. Thus, if transverse reinforcement exceeding that in

this study is provided, transverse reinforcement may achieve
yielding at shear failure before longitudinal reinforcement
yielding (Fig. 11).
39
Fig. 8Drift ratio versus crack angle.
Fig. 9Crack cutting through aggregates.
Fig. 10Damage distribution at the end of test: (a) A-1; (b) A-2; (c) A-3 (d) A-4 (e) B-1; (f) B-2; (g) B-3; and (h) B-4.
Experimental shear strength provided by transverse reinforcement (Vs_test) was calculated using Eq. (1). In Eq. (1),
st was determined by the stress-drift relationships (Fig. 11),
and was determined by measuring crack angle (Fig. 8).
Experimental shear strength provided by concrete (Vc_test)
was calculated using Eq. (2). Two conditions were considered when calculating Vs_test and Vc_test: diagonal cracking
and ultimate conditions. The diagonal cracking condition is
defined as when a diagonal shear crack first appears. The
ultimate condition corresponds to peak applied shear.
Vs _ test =
Av st d
cot q (1)
s
Vc_test = Vtest Vs_test (2)
Effect of axial load

Figure 12 shows a representative relationship between
column drift and Vs_test, Vc_test, and Vtest for the effect of axial
load. Test results for Columns A-3 and B-3 with axial load
ratios of 10% and 20%, respectively, are compared. A higher
axial load increased Vc_test and the increase rate for Vs_test.
The higher increase rate of Vs_test is due to a smaller shear
crack angle under a higher axial load. However, the increase
in axial load caused a much more rapid decline in Vc_test after
40
the peak of Vc_test, leading to more brittle behavior. Note that

the peak of Vtest may not coincide with the peak of Vc_test
(Columns A-3 and B-3) (Fig. 12).
Effect of concrete compressive strength
column drift and Vs_test, Vc_test, and Vtest to illustrate the
effect of compressive strength. Test results for Columns
A-3 and A-4 with actual concrete compressive strength of
97 and 107MPa (14,000 and 15,500 psi), respectively, are
compared. Although the two columns were designed with
a difference in concrete compressive strength of 30 MPa
(4400 psi), the actual difference was only 10 MPa (1500 psi)
(Table 1). No significant difference in behavior existed with
such a difference in concrete compressive strength (Fig. 13).
Effect of amount of transverse reinforcement
column drift and Vs_test, Vc_test, and Vtest for the effect of amount
of transverse reinforcement. Test results for Columns A-1
and A-3 with transverse reinforcement spacing of 450 and
260 mm (17.72 and 10.24 in.), respectively, are compared.
The decrease in transverse reinforcement spacing, that is, an
increase in the amount of transverse reinforcement, did not
have a significant effect on the peak of Vc_test. However, due
to the increase in Vs_test for a given drift, Vtest still increased
Table 2Experimental shear strengths contributed by concrete and steel

Diagonal cracking shear strength
Ultimate shear capacity
Column
Drift ratio, %
st, MPa
Vs_test, kN
Vc_test, kN
Drift ratio, %
st, MPa
Vs_test, kN
Vc_test, kN
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
A-1
0.35
19
1255
0.57
243
150
1428
A-2
0.33
1283
0.53
235
150
1488
A-3
0.32
16
13
1266
0.75
359
413
1359
A-4
0.33
14
10
1288
0.79
418
447
1334
B-1
0.45
18
10
1852
0.59
223
165
1913
B-2
0.41
20
11
1996
0.50
183
195
2103
B-3
0.40
16
17
2081
0.54
214
411
2007
B-4
0.42
18
14
2089
0.64
380
522
2006
Notes: 1 MPa = 145 psi; 1 kN = 0.225 kip.
Fig. 11Stress of shear reinforcement.
Fig. 13Drift ratio versus applied shear of Specimens A-3

and A-4.
and (5) are detailed equations for Vc, but should not be taken
greater than Eq. (6). Equation (7) defines Vs.
Nu
Vc = 0.17 1 +
13.8 Ag
Fig. 12Shear strength development of Specimens A-3 and

B-3.
after the peak of Vc_test, thus increasing the peak of Vtest and
the stress in the transverse reinforcement at the peak of Vtest.
EXAMINATION OF ACI 318 SHEARSTRENGTHEQUATIONS
According to the ACI 318 Code,1 nominal shear strength
Vn is contributed by two components: shear strength provided
by concrete (Vc) and shear strength provided by shear reinforcement (Vs). In the ACI 318 Code, Eq. (3) is the simplified
equation for Vc when axial compression exists. Equations (4)
(3)
fcbw d (psi)
V d
Vc = 0.16 fc + 17rw u bw d
Mm
(MPa)
V d
Vc = 1.9 fc + 2500rw u bw d
Mm
(psi)
M m = Mu N u
Vc = 0.29 fcbw d 1 +

Nu
Vc = 2 1 +
2000 Ag
fcbw d (MPa)
(4)
( 4h d ) (5)
8
0.29 N u
Ag
Nu
Vc = 3.5 fcbw d 1 +
500 Ag
(MPa)
(6)
(pssi)
41
Vc =

Nu
ft (max)
bw d 1 +
(8)
F2
ft (max)bw d
Based on beam test data (solid circular dots in Fig. 15),

ft(max)/F2 was set as 0.29fc (MPa) or 3.5fc (psi) (horizontal dashed line in Fig. 15). ft(max) was assumed to be
0.62fc (MPa) or 7.5fc (psi). Thus, Eq. (8) becomes Eq. (9).
Vc = 0.29 fcbw d 1 +
1.6 N u
fcbw d
(MPa)
or equal to
Fig. 14Drift ratio versus applied shear of Specimens A-1
and A-3.
Mu is taken as moment at distance d from the section of
maximum moment when the ratio of shear span to effective
depth is greater than 2, or moment at the center of shear
span when the ratio of shear span to effective depth is less
than 2.12
Vs =
Av f yt d
(7)
The ACI concrete shear-strength equations were derived

based on shear corresponding to diagonal cracking12 even
though they are defined as nominal concrete shear strength
in the ACI Code. The experimental concrete shear strength
at the diagonal cracking and at the ultimate shear condition
(peak applied load) are compared to shear strength predicted
by the ACI simplified (Eq. (3)) and detailed (Eq. (4) to (6))
concrete shear-strength equations (Table 3). In the calculation using detailed equations, the values of Mm in Eq. (5)
for all columns are negative, meaning that normal tensile
stress due to moment effect is small, and diagonal cracking
strength should be governed by Eq. (6). Note that when using
Eq. (3) to (6), the ACI limit on fc ( 70 MPa [10,000 psi])
for concrete shear strength was removed. Results of comparison show that Eq. (3) yields conservative predictions for all
columns. Equation (6) does not yield conservative estimates
of diagonal cracking strength for all columns. Even for ultimate shear strength, Eq. (6) results are not conservative for
all columns. This is discussed further in later paragraphs. For
shear strength provided by steel reinforcement, Eq. (7) does
not yield conservative results for all columns (last column in
Table 3). This is expected, as stress in transverse reinforcement was far from yield at the ultimate condition (Table 2).
As stated previously, if a higher amount of transverse reinforcement than that used in this study is provided, transverse
reinforcement stress at the ultimate condition may be further
increased. Further study is needed to examine transverse
reinforcement stress when the amount of transverse reinforcement is higher than that used in this research.
Equation (6) is based on Eq. (8),12 which was derived
based on principal tensile stress at diagonal cracking with an
assumption that neglects tensile stress due to moment effect.
42
Vc = 3.5 fcbw d 1 +

(9)
0.133 N u
fcbw d
(psi)
For simplicity, a constant value of 0.29 (MPa) (0.002[psi])

replaced 1.6/fc (MPa) (0.133/fc [psi]), which corresponds
to fc equal to approximately 30 MPa (4400 psi). With this
simplification and by approximating bwd as Ag, Eq. (9)
becomes Eq. (6). The simplification that sets fc equal to
30MPa (4400 psi) in 1.6/fc (MPa) (0.133/fc [psi]) overestimates mathematically the shear strength of members
with fc exceeding 30 MPa (4400 psi). The degree of overestimation increases as fc increases and can be as high as
28% for the case of 20% axial load ratio and fc = 100 MPa
(14,500 psi), which are design parameters for Columns B-2
and B-4.
To assess the applicability of Eq. (6) for columns with
fc 70 MPa (10,000 psi), a test database of 43 high-strength
columns from this study and literature3,5,10,11 was established (Table 4). The database includes only columns with fc
70 MPa (10,000 psi) and with data for diagonal cracking
strength available. The values of Mm (Eq. (5)) for all 43
columns are negative. Therefore, the shear strength of the
columns is governed by Eq. (6), not Eq. (4), for the detailed
shear-strength calculation. The eighth column in Table 4
shows the outcome of comparing the measured diagonal
cracking strength to predictions by Eq. (3). Predictions are
conservative for all columns. On the other hand, predictions
by Eq. (6) for 23 columns are not conservative (ninth column
in Table 4). If Eq. (9) is used instead of Eq. (6), the number
of unconservative results is reduced to 6 (tenth column in
Table 4).
To investigate the applicability of using ft(max)/F2 equal
to 0.29fc (MPa) (3.5fc [psi]) in Eq. (8) for high-strength
columns, the values of ft(max)/F2 for all the 43 columns
were back-calculated by substituting the measured diagonal cracking strength for Vc in Eq. (8) and solving for
ft(max)/F2. Figure 15 shows calculation results (data points
bounded by dashed lines). Because the Mm values for all
columns are negative, an infinity value was assumed for
the horizontal coordinate of the data points of the columns
in Fig. 15. Columns from different studies were separated
into different groups in Fig. 15 for comparison. Comparison of the distribution of the column data points and that
of the beam data points against the horizontal dashed line
reveals that ft(max)/F2 equal to 0.29fc (MPa) (3.5fc [psi])
Table 3Ratio of test results to shear-strength prediction using ACI 318 without strength limitation
Diagonal cracking shear strength
Vc _ test
Vc _ test
Vc _ test
Vc _ test
Vs _ test
Column
VEq (3)
VEq (6 )
VEq (3)
VEq (6 )
VEq ( 7 )
(1)
(2)
(3)
(4)
(5)
(6)
A-1
1.63
0.81
1.86
0.93
0.43
A-2
1.51
0.76
1.75
0.88
0.43
A-3
1.58
0.79
1.70
0.85
0.68
A-4
1.47
0.74
1.52
0.76
0.74
B-1
1.72
0.89
1.77
0.92
0.47
B-2
1.41
0.77
1.48
0.82
0.56
B-3
1.55
0.85
1.50
0.82
0.68
B-4
1.44
0.80
1.39
0.77
0.86
provides predictions of diagonal cracking strength for highstrength columns as reasonable as those for the beams. Note
that the original intention of ft(max)/F2 equal to 0.29fc
(MPa) (3.5fc [psi]) is to capture the average behavior of
beam data points in the right side of Fig. 15, rather than to
provide a conservative estimation, because beams in this
category typically have ultimate concrete shear strength
that is significantly higher than diagonal cracking strength.
However, columns carry axial load, which generally delays
diagonal cracking, reducing the difference between diagonal cracking strength and ultimate concrete shear strength
(Table 2). Therefore, it is suggested to lower ft(max)/F2 to
0.25fc (MPa) (3fc[psi])nearly the lower bound of the
column data points in Fig. 15. The new equation is Eq. (10).
The eleventh column in Table 4 shows the predictions using
Eq. (10); conservative results were obtained for all columns
except for Column A-4, which has a ratio of measured to
predicted diagonal cracking strength of 0.99.
Vc = 0.25 fcbw d 1 +
1.6 N u
fcbw d
(MPa)
or equal to
Vc = 3 fcbw d 1 +

Ultimate shear capacity
(10)
0.133N u
fcbw d
(psi)
A comparison of columns from different studies (Table4)

shows that columns tested in this study have lower diagonal
cracking strengths than those in literature. One possible
reason is the size effect. The diameter of the columns tested
in literature ranges from 200 to 400 mm (7.87 to 15.75 in.),
much smaller than that in this study. A large member size
generally has low strength due to the inclusion of more
places of imperfection than in a small member.
CONCLUSIONS
An experimental study on eight large-scale columns
with high-strength steel and concrete was carried out using
double-curvature cyclic loading. Test results are presented
and compared to predictions by the ACI shear-strength equaACI Structural Journal/January-February 2015
Fig. 15Relationship between test data and Eq. (4) and (6).
tions. The conclusions drawn from this study are summarized as follows.
1. All columns tested exhibited shear failure before longitudinal reinforcement yielding. The critical diagonal shear
crack angles with respect to the column longitudinal axis
were, on average, 26 and 22 degrees for columns with 10%
and 20% axial load ratios, respectively. Axial load increased
concrete shear strength in the range of axial load examined,
but also increased brittleness. The amount of transverse reinforcement did not significantly affect concrete shear strength.
Transverse reinforcement stress increased with drift. An
increased amount of transverse reinforcement delayed shear
failure to a higher drift, resulting in a higher level of stress
in the transverse reinforcement at shear failure. With the
amount of transverse reinforcement used in this study, the
transverse reinforcement did not yield at peak applied load
for all columns tested.
2. Predictions by the ACI simplified and detailed concrete
shear-strength equations were compared to test results of
43 columns from this study and the literature. Comparison
results showed that the ACI simplified shear-strength equation yielded a conservative estimate of diagonal cracking
strength for all columns. On the other hand, the ACI detailed
43
Table 4High-strength columns test data
Column
bw, mm
d, mm
fc, MPa
Axial load ratio
a/d
t, %
Vc _ test
Vc _ test
Vc _ test
Vc _ test
VEq (3)
VEq (6 )
VEq ( 9 )
VEq (10 )
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
C13
400
320
93.5
0.39
1.25
0.00
1.83
1.14
1.32
1.54
C23
400
320
93.5
0.39
1.25
0.16
1.67
1.04
1.21
1.40
C43
400
320
77
0.24
1.25
0.40
2.33
1.25
1.39
1.61
C5
400
320
93.5
0.39
1.25
0.40
1.74
1.08
1.26
1.46
C63
400
320
77
0.48
1.25
0.40
1.96
1.22
1.36
1.58
C7
400
320
93.5
0.39
1.25
0.62
1.90
1.18
1.37
1.59
C83
400
320
93.5
0.39
1.25
0.80
1.92
1.20
1.39
1.61
H-0.6-0.15
200
160
128
0.15
1.25
0.60
1.85
1.00
1.23
1.42
H-0.6-0.35
200
160
125
0.3
1.25
0.60
1.64
1.03
1.27
1.48
H-0.6-0.6
200
160
120
0.6
1.25
0.60
1.27
0.98
1.22
1.41
HS-0.6-0.35
200
160
128
0.3
1.25
0.60
1.74
1.09
1.37
1.58
HS-0.6-0.6
200
160
128
0.6
1.25
0.60
1.11
0.88
1.11
1.29
HS-1.2-0.65
200
160
129
0.6
1.25
1.20
1.32
1.04
1.32
1.53
H-0.3-0.6
200
160
128
0.6
1.25
0.30
1.08
0.85
1.08
1.25
H-1.2-0.65
200
160
121
0.6
1.25
1.20
1.50
1.16
1.44
1.67
H-1.8-0.6
200
160
130
0.6
1.25
1.80
1.44
1.14
1.45
1.68
H-0.3-0.35
200
160
130
0.3
1.25
0.30
1.68
1.06
1.33
1.54
H-1.2-0.3
200
160
121
0.3
1.25
1.20
1.65
1.02
1.26
1.46
H-1.8-0.35
200
160
121
0.3
1.25
1.80
2.04
1.27
1.56
1.81
U-0.4-0.65
200
160
130
0.6
1.25
0.40
1.17
0.93
1.18
1.37
U-0.7-0.6
200
160
129
0.6
1.25
0.70
1.25
0.99
1.26
1.46
C6110
400
320
113.8
0.17
1.88
1.19
1.62
0.88
1.05
1.22
10
C62
400
320
113.8
0.17
1.88
0.53
1.73
0.94
1.12
1.30
C6310
400
320
113.8
0.17
1.88
1.19
1.59
0.86
1.03
1.20
10
C31
400
320
113.8
0.33
1.88
1.19
1.66
1.04
1.27
1.47
C3210
400
320
113.8
0.33
1.88
0.53
1.52
0.96
1.16
1.35
10
C33
400
320
113.8
0.33
1.88
1.19
1.60
1.00
1.22
1.42
6-111
300
240
73.5
0.17
1.88
0.53
1.78
0.92
1.00
1.15
6-211
300
240
73.5
0.17
1.88
1.19
2.16
1.11
1.21
1.40
6-311
300
240
73.5
0.17
1.88
0.53
1.78
0.92
1.00
1.15
11
6-4
300
240
73.5
0.17
1.88
1.19
2.16
1.11
1.21
1.40
3-111
300
240
73.5
0.33
1.88
0.53
1.65
0.93
1.02
1.19
11
3-2
300
240
73.5
0.33
1.88
1.19
1.73
0.98
1.08
1.25
3-311
300
240
73.5
0.33
1.88
0.53
1.65
0.93
1.02
1.19
11
3-4
300
240
73.5
0.33
1.88
1.19
1.73
0.98
1.08
1.25
A-1
600
480
93
0.1
1.88
0.16
1.61
0.81
0.91
1.06
A-2
600
480
103
0.1
1.88
0.16
1.49
0.76
0.87
1.01
A-3
600
480
97
0.1
1.88
0.28
1.55
0.79
0.89
1.04
A-4
600
480
107
0.1
1.88
0.28
1.44
0.74
0.85
0.99
B-1
600
480
108
0.15
1.88
0.16
1.69
0.89
1.05
1.22
B-2
600
480
125
0.18
1.88
0.16
1.39
0.77
0.95
1.10
B-3
600
480
113
0.2
1.88
0.28
1.53
0.85
1.02
1.19
B-4
600
480
121
0.2
1.88
0.28
1.42
0.80
0.98
1.13
shear strength upper-bound equation (Eq. (6)) produced

unconservative predictions of diagonal cracking strength
for 23 columns, including all columns tested in this study. If
44
Eq. (9) was used instead of Eq. (6), the number of columns
with unconservative prediction was reduced to 6. Furthermore, because the difference between diagonal cracking
strength and ultimate concrete shear strength for highstrength columns can be insignificant, it is suggested to base
the concrete shear-strength equation on the lower-bound of
column test data and, hence, Eq. (10) is recommended for
high-strength columns.
AUTHOR BIOS
ACI member Yu-Chen Ou is an Associate Professor in the Department
of Civil and Construction Engineering at the National Taiwan University of Science and Technology, Taipei, Taiwan. He received his PhD
from the State University of New York at Buffalo, Buffalo, NY. He is the
Vice President of ACI Taiwan Chapter. His research interests include
reinforced concrete structures, steel-reinforced concrete structures, and
earthquakeengineering.
ACI member Dimas P. Kurniawan is a Research Assistant in the Department of Civil and Construction Engineering at the National Taiwan University of Science and Technology. He received his BS from Bandung Institute
of Technology, Indonesia, and MS from the National Taiwan University of
Science and Technology.
ACKNOWLEDGMENTS
The authors would like to thank National Center for Research on
Earthquake Engineering (NCREE), Taiwan, and the Excellence Research
Program of National Taiwan University of Science and Technology for their
financial support.
NOTATION
Ag
Av
a
bw
d
F2
=
=
=
=
=
=
fc
=
fcs
=
ft(max) =
fu
=
fy
=
fyl
=
fyls
=
fyt
=
fyts
=
M m
=
Mu
=
Nu
=
s
=
Vc
=
gross area of concrete cross section

total cross-sectional area of shear reinforcement
shear span
effective web width
effective depth
ratio of shear stress at cracking point to average shear stress of
effective cross section
concrete compressive strength
specified concrete compressive strength
concrete principal tensile strength
ultimate strength of steel
yield strength of steel
yield strength of longitudinal reinforcement
specified yield strength of longitudinal reinforcement
yield strength of shear reinforcement
specified yield strength of shear reinforcement
applied moment modified to consider effect of axial compression
applied moment
applied axial load (positive in compression)
spacing of shear reinforcement
nominal shear strength provided by concrete
Vc_test
Vn
Vs
Vs_test
Vtest
Vu
u

t
w
st
=
=
=
=
=
=
=
=
=
=
=
experimental shear strength provided by concrete

nominal shear strength
nominal shear strength provided by shear reinforcement
experimental shear strength provided by shear reinforcement
experimental shear strength
applied shear
ultimate strain of steel
shear crack angle to column longitudinal axis, deg
transverse reinforcement ratio
longitudinal tension reinforcement ratio
shear reinforcement stress
REFERENCES
2. Lee, J. Y.; Choi, I. J.; and Kim, S. W., Shear Behavior of Reinforced
Concrete Beams with High-Strength Stirrups, ACI Structural Journal,
V.108, No. 5, Sept.-Oct. 2011, pp. 620-629.
3. Sakaguchi, N.; Yamanobe, K.; Kitada, Y.; Kawachi, T.; and Koda, S.,
Shear Strength of High-Strength Concrete Members, Second International Symposium on High-Strength Concrete, SP-121, W. T. Hester, ed.,
American Concrete Institute, Farmington Hills, MI, 1990, pp. 155-178.
4. Watanabe, F., and Kabeyasawa, T., Shear Strength of RC Members with
High-Strength Concrete, High-Strength Concrete in Seismic Regions, SP-176,
C. W. French and M. E. Kreger, eds., American Concrete Institute, Farmington
Hills, MI, 1998, pp. 379-396.
5. Maruta, M., Shear Capacity of Reinforced Concrete Column Using
High Strength Concrete, Invited Lecture in the 8th International Symposium on Utilization of High-Strength and High-Performance Concrete,
Tokyo, Japan, Oct. 27-29, 2008.
6. Mphonde, A. G., and Frantz, G. C., Shear Tests of High- and
Low-Strength Concrete Beams Without Stirrups, ACI Journal, V. 81,
No.4, July-Aug. 1984, pp. 350-357.
7. Ahmad, S. H.; Khaloo, A. R.; and Poveda, A., Shear Capacity of
Reinforced High-Strength Concrete Beams, ACI Journal, V. 83, No. 2,
Mar.-Apr. 1986, pp. 297-305.
8. Thorenfeldt, E., and Drangsholt, G., Shear Capacity of Reinforced
High-Strength Concrete Beams, Second International Symposium on
High-Strength Concrete, SP-121, W. T. Hester, ed., American Concrete
Institute, Farmington Hills, MI, 1990, pp. 129-154.
9. Xie, Y.; Ahmad, S. H.; Yu, T.; Hino, S.; and Chung, W., Shear Ductility
of Reinforced Concrete Beams of Normal and High-Strength Concrete,
ACI Structural Journal, V. 91, No. 2, Mar.-Apr. 1994, pp.140-149.
10. Kuramoto, H., and Minami, K., Experiments on the Shear Strength
of Ultra-High Strength Reinforced Concrete Columns, Proceedings of the
Tenth World Conference on Earthquake Engineering, Madrid, Spain, July
1992, pp. 3001-3006.
11. Aoyama, H., Design of Modern High-Rise Reinforced Concrete
Structures, Imperial College Press, London, UK, 2001, 391 pp.
12. ACI-ASCE Committee 326, Shear and Diagonal Tension, ACI
Journal Proceedings, V. 59, No. 2, Feb. 1962, pp. 1-124.
45
NOTES:
46
TECHNICAL PAPER
Title No. 112-S05
Three-Parameter Kinematic Theory for Shear Behavior of

Continuous Deep Beams
by Boyan I. Mihaylov, Bradley Hunt, Evan C. Bentz, and Michael P. Collins
This paper presents a theory for predicting shear strengths,
deformations, and crack widths near failure of continuous deep
beams. To describe the deformations in continuous deep beams, the
theory uses a three-degree-of-freedom kinematic model (3PKT),
which is an extension of an earlier two-degree-of-freedom model
(2PKT) for members in single curvature. The extended model is
validated with the help of measured local and global deformations
taken during loading to failure of a large continuous deep beam.
The accuracy of the shear-strength predictions given by the theory
is evaluated using a database of 129 published tests of continuous
deep beams. The theory enables the load distribution and failure
loads of continuous deep beams subject to differential settlement
to be evaluated.
Keywords: deep beams; differential settlement; kinematics; redistribution;
shear strength; ultimate deformations.
INTRODUCTION
While continuous reinforced concrete deep beams, such as
the transfer girder shown in Fig. 1(a), perform more critical
load-carrying functions than slender beams, their safety is
more difficult to assess. Shear forces in such members are
more sensitive to differential settlement of footings and
because longitudinal strains vary nonlinearly over beam
depth, traditional design procedures for slender beams are
not appropriate. The ACI Building Code1 suggests that
either the nonlinear distribution of longitudinal strains be
taken into account or that strut-and-tie models be used.
While finite element programs, such as VecTor2,2 which
was used to produce Fig. 1(b), account both for nonlinear
distributions of strain and nonlinear material response
and can predict both failure loads and deformations, their
use requires considerable engineering time and expertise.
Strut-and-tie models (Fig.1(c)), on the other hand, by
approximating regions of high compressive stress in concrete
and high tensile stress in reinforcement, can usually provide
conservative estimates of strength after a few relatively simple
calculations. As these models concentrate on statics, they do
not provide accurate assessments of deformation patterns
close to failure and post-peak behavior. This also applies
for most analytical models for deep beams in the literature3
with exceptions focusing entirely on deformation capacity.4
Information on ultimate deformations can be critical in, for
example, evaluating the safety of transfer girders damaged
by earthquakes or by large differentialsettlements.
It is the purpose of this paper to present a kinematic
model (Fig. 1(d)) capable of predicting both strength and
deformation patterns near failure of reinforced concrete
continuous deep beams. Within each shear span this model
uses just three parameters: the average tensile strain in the
Fig. 1Modeling of transfer girder.

top longitudinal reinforcement, the average tensile strain in
the bottom reinforcement, and the vertical deformation of the
critical loading zone (CLZ). This three-parameter kinematic
theory (3PKT) is an extension to the two-parameter
kinematic theory (2PKT) proposed by Mihaylov at el.5,6
for simply supported deep beams. With a relatively small
number of hand calculations, the 3PKT can be used to
determine the shear failure load, crack widths, deformed
shape, and support reactions (accounting for differential
settlements) of a continuous deep beam near failure. As part
MS No. S-2013-150.R1, doi: 10.14359/51687180, was received March 24, 2014,
47
Fig. 2Test specimen CDB1 and instrumentation. (Note: 1 mm = 0.0394 in.; 1 kN = 0.225 kip;
1MPa = 145 psi = 0.145 ksi.)
Fig. 3Measuring deformation patterns at LS9.

of the development of the 3PKT, detailed measurements
of deformations were taken on a large beam tested at the
University of Toronto, and results of this test will be used to
introduce deformation patterns of continuous deep beams.
The novel three-parameter kinematic theory presented in
this paper enables engineers to accurately and efficiently
evaluate the shear strength and deformation capacity of
continuous deep beams such as the transfer girders. The effect
of support settlements on the safety of transfer girders can
also be assessed using the method. An example demonstrates
that transfer girders can be sensitive to supportsettlements.
A large continuous reinforced concrete deep beam (refer
to Fig. 2) was loaded to failure to investigate deformation
patterns. The beam, called CDB1, was a one-third-scale
model based on the transfer girder in Fig. 1. It had a
rectangular cross section with a depth of 1200 mm (47.2in.),
48
a width of 300 mm (11.8 in.), and two equal spans of 3475 mm

(11 ft 5 in.). Both top and bottom longitudinal reinforcement
consisted of four 25M bars and one 35M bar. The transverse
reinforcement consisted of No. 3 stirrups spaced at 235 mm
(9.25 in.) that provided a reinforcement ratio of 0.20%. The
compressive strength of the concrete when the beam was
tested 80 days after casting was 29.7 MPa (4300 psi).
Equal point loads, P, were applied near the middle of each
span using a spreader beam which was, in turn, loaded by a
manually controlled testing machine (Fig. 3). The specimen
incorporated short columns to provide realistic loading and
support conditions. The middle support was pinned while the
external supports and the two loading columns were provided
with rollers to allow for free movement of the specimen.
Load cells were placed under the external support columns.
Statics requires that the two external support reactions be
equal and hence the shears, Vext, in the two external shear
spans must also be equal. As the two applied loads, P, are
equal, the shears in the interior spans, Vint, also equal each
other. However, the ratio Vint/P is not dictated by statics but
depends on compatibility of deformations, kinematics.
Figure 2 also shows the layout of linearly variable
differential transformers (LVDTs) attached to the north,
bottom, and top faces of the specimen to measure relative
displacements on the concrete surface, as well as the locations
of electric resistance strain gauges used to monitor the strains
in the longitudinal bars and the stirrups. The south side of the
specimen was equipped with a 300 x 300 mm (11.8 x 11.8in.)
grid of targets for demountable displacement transducers,
which were used to measure displacement patterns at a
number of stages during the test (refer to Fig. 3). At such
load stages, LS, crack patterns were marked and crack widths
measured with crack comparators. After the formation of
major diagonal cracks, crack slip displacements parallel to
the cracks were also measured with crack comparators.
EXPERIMENTAL RESULTS
The measured load-displacement response of the specimen
is shown in Fig. 4. The thick lines show the load P, while the
thin lines plot the shear force Vint. Midspan deflections of the
two spans were measured by LVDTs 2E and 2W with respect
to a steel bar going between the ends of the specimen (refer to
Fig. 2). The test was performed in two phases. First the beam
was loaded until there was clear evidence of shear failure
in one of the spans (refer to Fig. 4(a)). After strengthening
the failed shear span with four external clamps, the second
phase involved reloading until the second span failed (refer
to Fig. 4(b)). Shown in Fig. 4 are four key load stages (LS5,
LS7, LS9, and LS11) where crack widths and deformation
patterns were measured and these are shown in Fig. 5.
In Phase I, first flexural cracking was observed under
the east load at P = 422 kN (95 kip), while at P = 546 kN
(123kip), cracks under the west load and over the central
column were observed. By LS5, the crack pattern resembled
three fans with cracks radiating out from the west load, the
central support, and the east load (refer to Fig. 5). The first
major diagonal crack occurred in the east inner shear span at
P = 1033 kN (232 kip). By LS7 there were diagonal cracks
up to 2.5 mm (0.010 in.) wide across the inner shear spans.
The maximum value of Vint in the east span was reached just
after LS7 and had a value of 813 kN (183 kip) with an east
midspan deflection of 3.70 mm (0.146 in.). At the maximum
shear, the inner shear to applied load ratio, Vint/P, was 0.619,
as opposed to the calculated value of 0.675 from flexural
theory ignoring shear strains. By LS9, the 5% increase in
applied force caused the east displacement to more than
double. While the load P increased, the shear, Vint, decreased
by 7.5%, causing Vext to increase by 25%. Concrete crushing
was visible near the inner faces of the east load and central
support and large displacements (9.28 mm [0.37 in.]) were
measured for LVDTs 11E and 13E, indicating post-peak
behavior of the CLZ regions (refer to Fig. 2). Because of
concerns with irreparably damaging the specimen, the loads
were reduced to approximately 200 kN (45 kip) after LS9
and external clamps were applied to the east inner shearspan.
At the start of Phase II loading, there was a gap of
approximately 1 mm (0.04 in.) between the bottom of the
central column and the support, which closed upon reloading.
The load-deformation response (refer to Fig.4(b)) was nearly
linear up to the peak load of 1604 kN (361 kip) at LS11.
The maximum shear of 916 kN (206 kip) also occurred at
LS11 at a west deflection of 4.23 mm (0.167 in.). Crushing
of the concrete at the top and bottom ends of the critical
crack was observed as deformations increased between
LS11 and LS13. The test was terminated at a deflection of
approximately 18.5 mm (0.73 in.) after a sudden drop of
resistance caused by rupture of stirrups in the west inner
shear span. A photograph of the west side of the beam taken
after the test is shown in Fig. 6. Note the external clamps on
the east inner shear span and the critical diagonal crack in
the west inner shear span.
Based on Vint/P = 0.675 (the elastic value), the moment
over the central support should be 1.17 times the moment
under the load. Figure 7(a) shows the elastic bending moment
diagram for LS5 along with the moment diagrams calculated
Fig. 4Measured load-displacement response. (Note:

1mm= 0.0394 in.; 1 MPa = 145 psi.)
from the observed Vint/P ratios for five selected load stages.
Load redistribution had a significant effect on the magnitude
of the moment over the central support. For example, from
LS5 to LS9, the applied loads increased by 37% but the
moment over the central support reduced by 39%. At the
end of the test, LS13, the moment over the central column
had changed sign because the large deformations in the
west inner shear span had significantly reduced its shear
resistance. Figure 7(b) shows the distribution of strains in
the longitudinal reinforcement for the five load stages. It can
be seen that, due to redistribution of bending moments, the
bottom bars were on the verge of yielding at LS11, the peak
load, and had yielded by LS13. After LS5, the measured
strains in the bottom bars were tensile along the whole length,
while in the top bars they were tensile along almost all of
the inner shear spans. Figure 7(c) shows the distribution of
transverse strains along the length of the beam as measured
by strain gauges on the stirrups and LVDTs mounted across
the beam depth. It can be seen that the strains in the outer
shear spans remained below yield strain. The central regions
of the inner shear spans saw high transverse strains prior to
49
Fig. 5Measured cracks and deformation patterns (maximum deflection equivalent to

1/25 of the span). (Note: 1 mm = 0.0394 in.; 1 kN = 0.225 kip.)
failure but the regions near the loads displayed only small
transversestrains.
Apart from giving information on crack patterns, crack
widths, and crack slips, Fig. 5 also provides detailed
measured displacement patterns at four load stages. These
deformed shapes were calculated from the readings of the
demountable displacement transducers7 and are magnified
so that for each diagram the maximum beam deflection plots
50
as if it was 1/25 of the span. It can be seen that at LS5 the

deformations are primarily flexural, curving down under the
loads and up over the central support. At LS7, significant
shear deformations caused by the diagonal crack can be seen
in the east span near the inner edge of the central support.
By LS9, these shear deformations had greatly increased due
to the opening and slipping of the critical diagonal crack.
As these large displacements developed, the east part of
Fig. 6West span of specimen after failure.

the span rotated counterclockwise to comply with the east
support. This almost-rigid body motion is associated with
a constant slip displacement along the critical diagonal
crack and crack widths that increase linearly from the top
to the bottom of the crack. A confirmation of this kinematic
pattern is provided by the crack measurements at LS9.
The red numbers show that the crack slip was on average
approximately 5 mm (0.197in.) while the maximum crack
width of 8.5 mm (0.335 in.) was measured near the bottom of
the beam. The deformation pattern for LS11, corresponding
to maximum shear in the west inner shear span, is dominated
by the residual deformations in the clamped east shear
span. However, it is evident that at this stage significant
shear deformations are occurring near the top of the critical
diagonal crack in the west inner shear span. Because of
safety concerns, detailed displacement patterns were not
measured for LS13.
KINEMATICS OF SHEAR SPANS UNDER
DOUBLECURVATURE
Predicting deformation patterns such as those shown
in Fig.5 is necessary to understand and predict the shear
behavior of continuous deep beams. 2PKT has already been
presented5,6 for the deformations in external shear spans
subject to only single-curvature bending. This model uses
a superposition of two deformation patterns, each of which
depends on just a single kinematic parameter (degree of
freedom) (refer to Diagrams 1 and 2 in Fig. 8). The first
pattern involves the elongation of the bottom longitudinal
reinforcement or tie (average strain et1,avg) while the second
pattern involves the transverse displacement, Dc, of the
critical loading zone (CLZ). The modeling of the inner shear
spans of continuous girders requires a third pattern, which
involves the elongation of the top longitudinal reinforcement
(average strain et2,avg) (refer to Diagram 3 in Fig. 8). The
complete kinematic model for the general case of double
curvature is shown in Box 1 of Fig. 9. In this figure, the first
and the third patterns from Fig. 8 are combined into a single
deformation pattern related to the elongation of the top and
bottom longitudinal reinforcement. This combined pattern
Fig. 7Measured response at five load stages. (Note: 1 mm

= 0.0394 in.)
consists of two fans of rigid radial struts pinned together at
the loading and support points and connected to the bottom
and top longitudinal ties, respectively. Two main diagonal
cracks appear in this pattern: one runs from the inner edge of
the bottom support to the top load at the location where the
shear force is zero or a minimum. The other crack extends
from the inner edge of the loading element to the bottom of
the section at the support where the shear is zero. Only the
more critical of these two cracks develops the transverse
displacement Dc.
If deformations et1,avg, et2,avg, and Dc are known, the complete
displacement field of the shear span can be obtained from
Eq. (8) to (11). These equations are the same as those of the
2PKT,5,6 except that Eq. (10) and (11) contain an extra term
accounting for the radial cracks above the critical crack. The
51
accuracy of Eq. (8) to (11) is evaluated in Fig.5 by using

10 measured values of the degrees of freedom to predict the
complete deformed shape of the two-span beam at each of
the four load stages. LVDTs 7E/7W to 9E/9W measure the
average strains in the longitudinal ties and LVDTs 11E/11W
and 13E/13W measure the transverse displacements in
the critical loading zones of the inner shear spans (refer
to Fig.2). It can be seen from Fig. 5 that the green circles
indicating the predicted locations of the 90 targets of the grid
match very well the vertices of the triangles representing the
measured location of these targets. Note that at LS5 the major
diagonal cracks going from the loads to the central support
have not yet formed and hence the Dc values are zero and the
predicted deformed shape is flexural. Major diagonal cracks
have formed in both the east and the west spans by LS7 with
the measured Dc values and resulting shear deformations in
the east being approximately twice those in the west. The post
east span shear failure LS9 involved Dc values approximately
Fig. 8Three degrees of freedom of kinematic model.
Fig. 93PKT for deep beams under single- and double-curvature bending.17
52
10 times larger than those at LS7 but the deformation

patterns, which are scaled to the same maximum deflection,
look very similar. At maximum shear in the west inner span,
LS11, the shear deformations near the top face by the load are
prominent. By this stage, the east span had been reinforced
with external clamps, which required removal of the LVDTs
measuring the kinematic parameters required for calculating
the predicted deformation pattern.
From Fig. 5, it can be seen that if the three kinematic
parameters for each shear span are known, then the complete
deformed shape of continuous deep beams can be determined
with reasonable accuracy. The three equations in Box 1 of
Fig. 9 enable the three kinematic parameters for a shear
span to be calculated from the end moments and geometric
properties of the shear span. The transverse displacement of
the critical loading zone at shear failure, Dc, is calculated
as in the 2PKT for single curvature6 but with an additional
coefficient kc. This coefficient accounts for the decreased
deformation capacity of the CLZ due to the tensile strains
et2,min in the top reinforcement. This compression softening
effect8 does not occur in members under single curvature
because in such members the top zone remains uncracked. In
Eq. (3), e1 is the principal tensile strain in the CLZ estimated
from compatibility of deformations.
CALCULATION OF SHEAR STRENGTH
The shear strength of a deep beam can be predicted when
equations for the degrees of freedom of the kinematic model
are combined with equations for geometry of the model,
equations for compatibility of deformations, and constitutive
relationships for the components of shear resistance (refer
to Fig. 9). The derivation of the basic equations has been
discussed in detail elsewhere.6 The procedure will be
illustrated by giving the key steps required to calculate the
shear resistance of beam CDB1 assuming that Vint/P has
the measured value at east span shear failure of 0.619. The
effective length of load, lb1e, transmitting shear to the inner
shear span (refer to Eq. (4)) is 186 mm (7.32 in.) and hence
the angle, a1, of the critical crack going from the inner face
of the support to this zero shear location is 35.8 degrees. The
dowel length, lk, for the 25M bars from Eq. (5) is 220mm
(8.7 in.), while for the 35M bar, it is 260 mm (10.2 in.). The
final geometric term required is the effective area of the
stirrups in the shear span, which from Eq. (7), is 606 mm2
(0.939 in.2) if a weighted average of 233 mm (9.17 in.) is
used for the reinforcing bar dowel length.
Shear-strength calculations commence with an estimate
of the shear failure loadfor example, 850 kN (191 kip).
At this load, M1 = 889 kNm (656 ft-kip) and the average
strain in the bottom longitudinal reinforcement from Eq. (1)
is 1.505 103. The moment M2 = 619 kNm (457 ft-kip) and
the average strain in the top reinforcement from Eq. (2) is
1.048 103. The minimum strains, near the support for the
bottom bars and near the load for the top, are taken as 75% of
these values. From the minimum tensile top bar strain near
the load and the angle of the critical crack, the compression
softening factor, kc, is calculated as 0.840 and then Dc from
Eq. (3) is 2.27 mm (0.089 in.).
From Eq. (14), the shear carried by the critical loading

zone, VCLZ, is calculated as 346 kN (77.9 kip). The shear
carried across the crack interface, Vci, depends on the crack
width, which from Eq. (12) is 2.07 mm (0.081 in.), resulting
in a Vci value from Eq. (16) of 164 kN (36.9 kip). Applying
the dowel action expression, Eq. (15), separately to the four
25M bars and the one 35M bar and summing the resistances
gives a Vd value of 45 kN (10.1 kip). The final shear-strength
component, Vs, depends on the strain in the stirrups, which
is calculated from Eq. (13) to be 3.11 103. As this exceeds
the yield strain, the effective stirrups are at yield stress and so
they resist a shear of 490 x 606, which is 297 kN (66.8kip).
The calculated shear resistance against failure on this critical
crack is thus 346 + 164 + 45 + 297 = 852 kN (192 kip).
Because the calculated resistance agrees closely with the
initial estimate, this value can be taken as converged.
The equations in Fig. 9 have been formulated for the case
where the diagonal crack at the inner edge of the bottom
support is opening at failure. To check whether this crack is
indeed critical, calculations need to be also performed for the
case where the diagonal crack at the inner face of the top load
opens at failure. That is, the CLZ will now be at the bottom
support while the dowel action in the longitudinal bars will
now occur near the top load. The effective length of the support
lb2e resisting the shear is 150 mm (5.91 in.) and the critical
crack thus goes from the center of the supporting column to
the inner face of the upper column, which gives an angle a1
of 36.4 degrees. This small change in slope causes the dowel
length for the 25M bars to change to 216mm (8.50in.) but
does not change the dowel length for the 35M bar. Using the
reduced effective bearing length and the weighted average
dowel length of 231 mm (9.09in.), the calculated value for Av
from Eq. (7) increases to 620mm2(0.961 in.2).
If the initial estimate of the shear resistance is again taken
as 850 kN (191 kip), then the calculated strains in the top
and bottom reinforcing bars will stay the same. However,
as the CLZ is now at the bottom, it is the minimum strain
in the bottom bars, 0.75 1.505 103, which governs the
compression softening factor kc, giving it a value of 0.744.
This smaller value combined with the smaller effective
bearing length means that Dc is reduced to 1.59mm
(0.063in.) and VCLZ is reduced to 254 kN (57kip). The
calculated crack width, 1.43mm (0.056in.), due to the
opening of the critical crack near the inner face of the
top loading area, is a function of Dc and of the minimum
strain in the top bars (0.75 1.048103). The resulting
value of Vci is 221 kN (49.8kip). The dowel action
component, Vd, which has a value of 55 kN (12.3 kip), is
also a function of the minimum strain in the top bars. The
stirrup component, Vs, on the other hand, is a function of
the average strain in the top bars (1.048 103) and has a
value of 266 kN (59.9kip). The calculated shear resistance
against opening of the diagonal crack at the inner face of
the load is thus 254 + 221+55+266 = 796 kN (179kip).
Repeating the calculations for an estimated failure shear of
800 kN (180kip) gives a calculated resistance of 802 kN
(180kip), which can be taken as the converged value. This
is the predicted critical failure shear for the member, as it is
approximately 6% smaller than the shear required to open
53
Fig. 11Deformations in continuous deep girders.
Fig. 10Comparison between 3PKT and ACI shear-strength

predictions for 129 tests of continuous beams.
the diagonal crack at the inner edge of the support. Note that
the measured shear force at first failure for this member was
812 kN (183 kip), giving a Vexp/Vpred ratio of 1.01.
Calculations similar to those described previously were
performed for 129 experiments on continuous deep beams
described in the literature9-15 and the results are summarized
in Fig. 10. In this figure, the Vexp/Vpred ratios are plotted
against a/d ratios of the specimens. Details of specimens and
shear-strength calculations are given in the Appendix* to this
paper. Predicted shear capacity is taken as the larger6,16 of
the deep beam capacity and the sectional shear capacity. All
129 beams were governed by the 3PKT predictions but for
the ACI predictions, 51 beams were governed by sectional
capacity. While the 3PKT method requires somewhat
more computational effort, it results in predictions that
*
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.
54
are significantly more accurate across the entire range of

a/dvalues.
As in the example calculations for beam CDB1, the shear
strength calculations for the 129 experiments assumed that
the load distribution Vint/P in the critical shear span at failure
was equal to that measured in the experiment. If the load
distribution was not reported, then it was calculated based on
linear elastic beam theory. The assumed load distribution had
a relatively small effect on the predicted shear strength of the
test beams. For example, if the shear strength of Beam CDB1
is calculated assuming the elastic value of Vint/P of 0.675,
rather than the 0.619 experimental value, then the predicted
shear-strength changes from 802 to 828 kN (180to 186 kip).
While using the calculated elastic distribution results in a
3% increase in predicted shear strength, the predicted failure
load, P, is reduced from 1296 to 1227 kN (292 to 276 kip)a
5% reduction. For shear-critical transfer girders subjected
to significant differential settlements, the redistribution of
reactions may cause a much more substantial reduction in
failure load. Procedures based on the 3PKT for calculating
load distributions near failure are described in the following.
LOAD DISTRIBUTION IN CONTINUOUS BEAMS
At shear failure of the symmetrical two-span deep beam
shown in Fig. 2 linear elastic theory, assuming uniform
stiffness along the beam, predicts 67.5% of the total load
will be resisted by the central column. The 3PKT can be
used to more accurately access load distribution near failure.
Because of symmetry, the central support will not rotate and
hence each of the two spans acts as a propped cantilever.
As shown in Fig. 11, the deformations near failure in the
propped cantilever can be expressed in terms of just three
parameters: Dc, the transverse displacement of the critical
loading zone in the interior shear span; et2,avg, the average
tensile strain in the top longitudinal bars of the interior shear
span; and et1,avg, the average tensile strain in the bottom
longitudinal bars, which from the procedures in Fig. 9 will

be equal in both shear spans. The downwards displacement
of the tip of the cantilever, Ds, can then be calculated from
Eq. (18), the derivation of which is illustrated in Fig. 11.
D s = (e t ,2 avg a cot a + D c )

+ (e t ,2 avg e t ,1avg )aext cot a e t ,1avg aext cot a ext
(18)
If there is no differential settlement of the supports, then

Ds should be zero. If there is differential settlement, then Ds
is equal to the settlement of the exterior supports minus the
settlement of the central support.
The procedure to calculate the distribution ratio Vint/P
starts with an estimate for this ratio. For example, the elastic
value of 0.675 could be chosen. The predicted shear strength
and the three kinematic parameters for the chosen ratio are
then calculated as explained in the previous section. For the
0.675 ratio, the predicted shear strength is 828 kN (186 kip),
et1,avg is 1.146 103, et2,avg is 1.343 103, and Dc is 1.76 mm
(0.069 in.). Substituting the values of the three kinematic
parameters into Eq. (18) gives
Ds = 4.99 + 0.45 2.85 = 2.59 mm (0.102 in.)
This shows that a Vint/P ratio of 0.675 is appropriate if the
external supports settle by 2.59 mm (0.102 in.) more than the
central support. In this case, the predicted failure value for P
would be 1227 kN (276 kip).
If the initial estimate of the ratio Vint/P is 0.619, then the
predicted shear strength will be 802 kN (180 kip) and the
predicted failure value for P will be 1296 kN (292 kip)
while the three kinematic parameters are as follows: et1,avg
is 1.417 103, et2,avg is 0.987 103, and Dc is 1.63 mm
(0.064in.). From Eq. (18)
Fig. 12Comparison of predicted and observed response.

(Note: 1 kN = 0.2248 kip; 1 mm = 0.0394 in.)
Ds = 4.00 0.99 3.01 = 0.00 mm

As Ds is 0, this shows that the predicted value of Vint/P
for the case of no differential settlement is 0.619, which
happens to also be the value measured in the experiment.
The predicted deflection under the load at failure is equal
to the first term in Eq. (18), which in this case is 4.00 mm
(0.158 in.). In the experiment, the measured deflection
under the load at maximum shear force in the east span was
3.70mm (0.146 in.).
Figure 12 demonstrates that the 3PKT can be used to
obtain a reasonably accurate prediction of the entire V-D
response of the member, where V is the shear in the critical
inner shear span and D is the deflection under the load. The
peak shear and corresponding deflection are obtained as
discussed previously. If the peak point is shifted to the left
by Dc and the new point is connected to the origin of the plot,
the predicted response when Dc equals 0 is obtained. The Dc
parameter will be 0 until a major web-shear diagonal crack
forms. The diagonal cracking shear, Vcr, can be calculated
as 520 kN (117 kip) from Eq. (11-7) of the ACICode.1
At low load levels when both M1 and M2 are less than the
flexural cracking moment, Mcr, the response of the beam is
Fig. 13Predicted effect of support settlements on failure

load of deep continuous girders.
well-predicted by traditional linear elastic theory. Assuming
a modulus of rupture of 0.63fc' MPa (7.5fc'psi),1 Mcr
equals 247 kNm (182 ft.kip), which corresponds to a
shear of 258 kN (58 kip) at flexural cracking. The points
55
corresponding to flexural cracking, web-shear cracking,

and the peak load are joined with straight lines to give a
trilinear response up to the peak shear. Finally, the postpeak response is calculated by using the 3PKT equations but
with Dc being increased, which increases the deflection but
decreases the shear resistance. Also shown in Fig. 12 are the
experimentally observed response and the predictions from
the VecTor2 model and the ACI strut-and-tie model.
In the aforementioned discussion, it was demonstrated
that a very small differential settlement of the supports
of the test beam could reduce the predicted failure load
by approximately 5%. Figure 13 has been prepared to
demonstrate how the predicted failure loads change for
a wide range of differential settlements. In addition to the
loads P applied near the middle of each span, a load P
was applied over the central support and loads of P/2 were
applied over each exterior support to better simulate the real
loading situation shown in Fig. 1(a). The additional load over
the central support reduces the effective bearing length, lb2e,
and hence reduces somewhat the predicted strength of the
inner shear span. For different values of the Vint/P ratio, the
failure load P and the associated differential settlement Ds
can be calculated from the 3PKT equations. These equations
have been developed for the case where the deep beam fails
in shear prior to yielding of the longitudinal reinforcement
in flexure. For the beam studied in Fig. 13, the longitudinal
reinforcement is predicted to remain elastic provided that Ds
is less than 10.1 mm (0.40 in.) and greater than 8.7 mm
(0.34 in.). As can be seen in Fig. 13, within this range, the
predicted failure loads go from a low of 870 kN (196 kip)
to a high of 1486 kN (334 kip) and the predicted changes
in failure loads agree closely with the changes predicted by
the VecTor2 model. In evaluating the significance of these
predictions, it is useful to recall that allowable bearing
stresses for spread footings are often based18 on restricting
differential settlements to approximately 20 mm (0.75 in.) at
service loads, which corresponds to approximately 40 mm
(1.5 in.) at structural failure loads. The beam whose response
is calculated in Fig. 13 was approximately a one-third-scale
model of an actual transfer girder and hence the differential
settlements that could be expected at failure for the model
beam would be approximately 40/3 = 13 mm (0.5 in.).
With such differential settlements, the 3PKT and VecTor2
models both predict that significant reductions in member
capacity can occur.
to-predicted-shear-strength ratio of 1.09 with a coefficient

of variation (COV) of 16%, as compared to the ACI code
provisions that resulted in an average ratio of 1.60 and a
COV of 36%. The 3PKT presented in this paper enables an
engineer to evaluate the behavior of a deep transfer girder
subject to differential settlements with a relatively small
number of calculations. For the beam studied in the paper,
the 3PKT model with only three degrees of freedom gave
results very similar to a nonlinear finite element model with
thousands of degrees of freedom.
AUTHOR BIOS
ACI member Boyan I. Mihaylov is an Assistant Professor in the Department
of ArGEnCo at the University of Liege, Liege, Belgium. He received his
PhD from the ROSE School, Pavia, Italy, in 2009.
Bradley Hunt is a PhD Candidate at Carleton University, Ottawa, ON,
Canada. He received his Masters of Engineering degree from the University
of Toronto, Toronto, ON, Canada, in 2012.
Evan C. Bentz, FACI, is an Associate Professor of civil engineering at
the University of Toronto. He is Chair of ACI Committee 365, Service
Life Prediction, and a member of Joint ACI-ASCE Committee 445, Shear
andTorsion.
ACI Honorary Member Michael P. Collins is a University Professor and
the Bahen-Tanenbaum Professor of Civil Engineering at the University
of Toronto. He is a member of Joint ACI-ASCE Committee 445, Shear
andTorsion.
NOTATION
As1
As2
a
aext
ag
d1
d2
db
Es
fc'
fv
fy
fye
fyv
h
k
kc
l0
lb1
lb1e
lb2
lk
M1
CONCLUSIONS
For continuous deep beams, redistribution of forces
caused by differential settlements can significantly reduce
failure loads. The three-parameter kinematic theory (3PKT)
presented in this paper is a valuable tool for assessing such
situations. This approach is based on a kinematic model
which accurately describes the deformed shape of each
shear span using only three degrees of freedom: the average
strains in the top and bottom longitudinal reinforcement
and the transverse displacement of the critical loading zone.
The shear-strength predictions of the 3PKT were validated
against the results from 129 published tests of continuous
deep beams. The 3PKT produced an average experimental56
M2
nb
P
P1/2
smax
V
VCLZ
Vci
Vd
Vs
w
a
a1
Dc
dx
= area of bottom longitudinal reinforcement

= area of top longitudinal reinforcement
= shear span
= external shear span
= maximum size of coarse aggregate
= effective depth of section with respect to bottom reinforcement
= effective depth of section with respect to top reinforcement
= diameter of bottom longitudinal bars
= modulus of elasticity of steel
= concrete cylinder strength
= stress in stirrups
= yield strength of bottom longitudinal bars
= effective yield strength of bottom longitudinal bars
= yield strength of stirrups
= total depth of section
= crack shape factor
= compression softening factor
= length of heavily cracked zone at bottom of critical diagonal
crack
= width of loading plate parallel to longitudinal axis of member
= effective width of loading plate parallel to longitudinal axis of
member
= width of support plate parallel to longitudinal axis of member
=
length of dowels provided by bottom longitudinal
reinforcement
=
absolute value of moment causing tension in bottom
reinforcement
= absolute value of moment causing tension in top reinforcement
= number of bottom longitudinal bars
= applied load
= applied concentrated load/support reaction
= distance between radial cracks along bottom edge of member
= shear force
= shear resisted by critical loading zone
= shear resisted by aggregate interlock
= shear resisted by dowel action
= shear resisted by transverse reinforcement
= crack width
= angle of line extending from inner edge of support plate to far
edge of tributary area of loading plate
= angle of critical diagonal crack
= transverse displacement of critical loading zone
= displacement along x-axis
dz
e1
et1/2,avg
et1/2,min
ev
q
rl
rv
=
=
=
=
=
=
=
=
displacement along z-axis

principal tensile strain in critical loading zone
average strain along bottom/top longitudinal reinforcement
minimum strain along bottom/top longitudinal reinforcement
transverse web strain
angle of diagonal cracks in uniform stress field
ratio of bottom longitudinal reinforcement
ratio of transverse reinforcement
REFERENCES
2. Vecchio, F. J., Disturbed Stress Field Model for Reinforced Concrete:
Formulation, Journal of Structural Engineering, ASCE, V. 127, No. 1,
2001, pp. 12-20. doi: 10.1061/(ASCE)0733-9445(2001)127:1(12)
3. Senturk, A. E., and Higgins, C., Evaluation of Reinforced Concrete
Deck Girder Bridge Bent Caps with 1950s Vintage details: Analytical
methods, ACI Structural Journal, V. 107, No. 5, Sept.-Oct. 2010,
pp.544-553.
4. Hong, S. G.; Hong, N. K.; and Jang, S. K., Deformation Capacity of
Structural Concrete in Disturbed Regions, ACI Structural Journal, V. 108,
No. 3, May-June 2011, pp. 267-276.
5. Mihaylov, B. I.; Bentz, E. C.; and Collins, M. P., A Two Degree
of Freedom Kinematic Model for Predicting the Deformations of Deep
Beams, CSCE 2nd International Engineering Mechanics and Materials
Specialty Conference, Ottawa, ON, Canada, June 2011, 10 pp.
6. Mihaylov, B. I.; Bentz, E. C.; and Collins, M. P., Two-Parameter
Kinematic Theory for Shear Behavior of Deep Beams, ACI Structural
Journal, V. 110, No. 3, May-June 2013, pp. 447-456.
7. Mihaylov, B. I., Behavior of Deep Reinforced Concrete Beams under
Monotonic and Reversed Cyclic Load, doctoral thesis, European School
for Advanced Studies in Reduction of Seismic Risk, Pavia, Italy, 2008,
379pp.
8. 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.
9. Moody, K. G.; Viest, I. M.; Elstner, R. C.; and Hognestad, E., Shear
Strength of Reinforced Concrete Beams Part 2Tests of Restrained Beams
without Web Reinforcement, ACI Journal Proceedings, V. 51, No. 1, Jan.
1955, pp. 417-434.
10. Rogowsky, D. M., and MacGregor, J. G., Tests of Reinforced
Concrete Deep Beams, ACI Journal Proceedings, V. 83, No. 4, July-Aug.
1986, pp. 614-623.
11. Ashour, A. F., Tests of Reinforced Concrete Continuous Deep
Beams, ACI Structural Journal, V. 94, No. 1, Jan.-Feb. 1997, pp. 3-11.
12. Asin, M., The Behaviour of Reinforced Concrete Continuous Deep
Beams, doctoral thesis, Delft University Press, Delft, the Netherlands,
1999, 167 pp.
13. Yang, K.-H.; Chung, H.-S.; and Ashour, A. F., Influence of Shear
Reinforcement on Reinforced Concrete Continuous Deep Beams, ACI
Structural Journal, V. 104, No. 4, July-Aug. 2007, pp. 420-429.
14. Yang, K.-H.; Chung, H.-S.; and Ashour, A. F., Influence of Section
Depth on the Structural Behaviour of Reinforced Concrete Continuous Deep
Beams, Magazine of Concrete Research, V. 59, No. 8, 2007, pp.575-586.
doi: 10.1680/macr.2007.59.8.575
15. Zhang, N., and Tan, K.-H., Effects of Support Settlement on
Continuous Deep Beams and STM Modeling, Engineering Structures,
V.32, No. 2, 2010, pp. 361-372. doi: 10.1016/j.engstruct.2009.09.019
16. AASHTO, AASHTO LRFD Bridge Design Specifications, fourth
edition, American Association of State Highway Officials, Washington,
DC, 2007, 1526 pp.
17. Bentz, E. C.; Vecchio, F. J.; and Collins, M. P., Simplified Modified
Compression Field Theory for Calculating Shear Strength of Reinforced
Concrete Members, ACI Structural Journal, V. 103, No. 4, July-Aug.
2006, pp. 614-624.
18. Craig, R. F., Soil Mechanics, seventh edition, E&FN Spon, London,
UK, 2004, 447 pp.
57
NOTES:
58
TECHNICAL PAPER
Title No. 112-S06
Investigation of Bond Properties of Alternate Anchorage

Schemes for Glass Fiber-Reinforced Polymer Bars
by Lisa Vint and Shamim Sheikh
A study of existing research shows a need for an investigation of
the bond properties of anchorage systems for glass fiber-reinforced
polymer (GFRP) bars including mechanical anchor heads and
bends. In this research program, the standard pullout test procedure was modified, which improved testing efficiency, accommodated bent bar tests, and reduced the variability of concrete properties across test specimens. Based on this experimental work
consisting of a total of 72 specimens, it was concluded that the
surface profile of GFRP bars influences only the post-peak phase
of the bond stress-slip curve. It was also found that GFRP bars with
anchor heads still require a considerable bonded length to develop
the bars full strength. The bend strengths were determined to be
between 58 and 80% of the strength of the straight portion of the
same bar in the specimens from three GFRP manufacturers. The
straight bar embedment length required to develop full strength of
bent bar was found to be approximately five times the bar diameter
for all bar types tested.
Keywords: anchor; bend; bond; glass fiber-reinforced polymer (GFRP)
bars; polymers; reinforced concrete.
INTRODUCTION
Glass fiber-reinforced polymer (GFRP) reinforcing bars
have been introduced as a lightweight, corrosion-resistant
material which offers a viable replacement for traditional
steel reinforcing bars, especially when the structures are
located in aggressive environments such as coastal regions
and those subjected to deicing salts. Extensive experimental
work is needed in order to develop reliable and rational
guidelines for design if GFRP is to be widely accepted as a
practical construction material. One property of importance
is the bond between the GFRP bars and the surrounding
concrete. This property is crucial, as it has a major effect
on the structural performance of a member with regards to
cracking, deformability, internal damping, and instability in
concrete structures (Gambarova et al. 1998).
To develop the required design strength of the bars
in tension, it is common practice to introduce mechanical anchor heads at the ends of straight bars in reinforced
concrete structures when the available space is limited. In
recent years, great improvements have been made in the
manufacturing of GFRP reinforcement with the current
bars having a much higher ultimate strength and stiffness
than the bars of previous generations. With these greater
strengths come much longer required development lengths,
further increasing the need for anchor heads. Due to the lack
of standards for the manufacturing of these anchor heads,
the current Canadian design code for FRP bars, CSA S80612, requires that engineers check the results from research
programs to determine if the anchor heads have been proven

to develop at least 1.67 times the required design strength
within the given length.
Bends in GFRP reinforcement are also used for reducing
the required development length as well as in stirrups used
as shear reinforcement. The strength at these bends has
been reported to be approximately 30 to 60% of the ultimate tensile strength of the straight portion of the bars (Imjai
etal. 2007; Ahmed et al. 2010). This reduced strength is due
to the bearing action of the concrete on the bend inducing
normal stresses in the weak lateral direction of the bars,
and due to the longitudinal fibers becoming kinked along
the interior radius of the bend, reducing their load-carrying
capacity (Ahmed et al. 2010). Results from the research on
the behavior of various anchorage types would allow engineers to be more confident when designing with this relatively new and always evolving reinforcing material.
While there have been a multitude of studies recently
published on the bond properties of GFRP bars, there is a
gap in the literature with respect to the behavior of different
anchorage types, such as headed and bent bars, and their
relative benefits over straight bars. Because no codified
standards for GFRP manufacturing are in place, great variations in the mechanical properties and surface profiles of
the available GFRP bars exist. This research addresses these
issues and presents results from an extensive study on the
bond and anchorage of GFRP bar products from different
manufacturers tested in a similar manner and under the same
conditions. Performance of straight bars, headed bars, and
bent bars is discussed and resistance of each type of anchor
is investigated.
EXPERIMENTAL INVESTIGATION
To minimize the risk of splitting failure in the concrete
cylinders, a slab layout was adopted for the traditional direct
tension pullout test (DTPT) (fib 2000), of anchors. A total of
72 bond specimens of various anchorage types and bonded
lengths were embedded in six concrete slabs each measuring
2400 x 1200 x 320 mm (94.5 x 47.2 x 12.6 in.). The test
parameters included the surface profile based on the type of
MS No. S-2013-180.R3, doi:10.14359/51687042, was received January 16, 2014,
59
Table 1Experimental GFRP reinforcement

coupon test results
Ab,
mm2 (in.2)
fu,
MPa (ksi)
E, MPa (ksi)
Rupture
strain*,
%
197.9 (0.306)
742 (107.6)
49,500 (7180)
1.50
197.9 (0.306)
1277 (185.2)
72,600 (10,530)
1.76
201 (0.312)
1228 (178.1)
59,200 (8590)
2.07
71.3 (0.1105)
833 (120.8)
54,300 (7880)
1.53
B 126.7 (0.1964)
655 (95.0)
42,000 (6090)
1.56
912 (132.3)
57,900 (8400)
1.61
Bar type
Straight
Bent
113.1 (0.1753)
Calculated from ultimate stress fu and stiffness E except for C-Bent, which was
measured.
Fig. 1Concrete compressive stress versus time.

GFRP bar, anchorage type, and bonded length. Each type
of specimen was tested in triplicate. Three surface profiles
were investigated which included helically wrapped bars
with sand coating (Bar Type A), sand-coated (Bar Type B),
and ribbed (Bar Type C). The three anchorage types include
straight bar, headed bar, and bent bar. All bar types were
tested with bonded lengths equal to five and 10 times the bar
diameter, 5db and 10db. Additionally, the headed and bent
bars were tested where the bar was debonded using a polyvinyl chloride (PVC) tube up to the start of the anchorage
mechanism, referred to as 0db, so that only the strength of
the anchor head or bend could be determined.
Concrete materials
The slabs were cast in two batches using ready mixed
concrete provided by a local supplier with specified compressive strength of 25 MPa (3.63 ksi) at 28 days. The concrete
was also specified with no air entrainment, 20 mm (0.79 in.)
maximum aggregate size, and 100 mm (4 in.) slump, which
were each verified at the time of the castings. The concrete
compressive and rupture strengths were found to be similar
for both batches. At the time of testing of bond specimens,
the average of three test cylinders gave concrete compressive strength of 35.8 and 30.2 MPa (5.19 and 4.38 ksi) and
modulus of rupture values of 3.65 and 3.98 MPa (0.53 and
0.58 ksi), for Batches 1 and 2, respectively. Figure 1 shows
the development of compressive strength with age for the
two concretes.
Reinforcing bar materials
All three bar types are manufactured using similar pultrusion processes for the GFRP bar core. The difference is in
the manufacturing process for the bars surface profiles. Bar
Type A is first sand coated and then wrapped helically with
fibers before thermosetting of the resin, creating undulations
in the side of the bar. Bar Type B has a sand-coated surface
that is applied using an in-line coating process (Ahmed et al.
2008) during the pultrusion of the bar. Lastly, Bar Type C
has a ribbed bar surface that is cut into the rod after curing
is completed.
60
Both straight bars and the straight portion of bent bars

with diameters of 16 mm (5/8 in.), and 10 to 12.7 mm (3/8to
1/2 in.), respectively, were tested for their ultimate tensile
strength and Youngs modulus of elasticity. The standard
test method for tensile properties of fiber-reinforced polymer
matrix composite bars (ASTM D7205-06) was followed
when completing these tests and the average results for three
types of bars tested can be found in Table 1. It can be seen in
the table that straight Bar Type A has a lower ultimate tensile
strength and stiffness than Bar Types B and C, whereas bent
Bar Type B has a lower strength and stiffness than the other
two bar types. These properties were calculated using the
nominal bar areas provided by the bar manufacturers as
practiced by designers in most cases. The cross-sectional
areas of the straight bars based on the actual diameters of
each bar, measured by excluding the sand coatings and ribs,
were 195, 241, and 210 mm2 (0.302, 0.374, and 0.326 in.2)
for Bar Types A, B, and C, respectively. These compare with
198, 198, and 201 mm2 (0.307, 0.307, 0.312 in.2) based on
the nominal diameters provided by the manufacturers.
Test setup and instrumentation
In lieu of using the traditional DTPT, where each specimen
is embedded in a single cylinder and undesirable concrete
splitting failures can occur, a slab layout was used. The Standard Test Method for Strength of Anchors in Concrete and
Masonry Elements (ASTM E488-96) was used as a guideline
when designing the slab layout for this research program. A
minimum spacing of 400 mm (16 in.) was provided between
each test anchor, which is greater than the 1.5 times the
embedment length specified in ASTM E488. The bars with
the longest embedment length were placed along the center
line of the slab while the shorter embedment lengths were
alternated along the outer edges (Fig. 2). Both the spacing
and bar placement were designed to minimize the risk of
interaction between adjacent test anchors.
Because one of the objectives of this research program
was to determine the bend strength of the stirrups used in
a large beam test series by Johnson and Sheikh (2013), the
layout of the bent bars was designed to mimic that of a beam.
Therefore, bent Bar Types B and C were the stirrups from
the same batch as the reinforcement for the large beam test
series, and bent Bar Type A was added for the current test
program. Two 16 mm (5/8 in.) Type C bars were used at each
bend to mimic the longitudinal bars used in a beam (Fig. 2).

The 320 mm (12.6 in.) height of the slab in this test series
is one half of the height of the beams in the large beam
testseries.
The design and construction of the reaction bridge was
also in accordance with ASTM E488, where the minimum
spacing between the supports need to be at least four times
the embedment length. The spacing provided between
supports for all the tests performed was 640 mm (25-1/4 in.)
and was four times the longest embedment length of 160 mm
(6-1/4 in.). The reaction bridge was made up of two HSS
102x 102 x 13 mm (4 x 4 x 0.5 in.) and one HSS 178 x 127x
9.5 mm (7 x 5 x 0.38 in.), supported by four steel bearing
plates with a thickness of 44.5 mm (1-3/4 in.) (Fig. 3). In
some cases, the reaction points were asymmetric about the
bond specimen. These reaction points are located at greater
than the minimum distance required by ASTM E488 and
they had no influence on the stresses around the specimens.
A tensile load was applied manually to the coupler
attached to the free of each anchor specimen at a rate of
15kN/min (3.37 kip/min) using a 60 tonne (66 ton) hollow
plunger cylinder which was attached to a hydraulic load

maintainer. The load was measured using a load cell with
a capacity of 1000 kN (225 kip). To ensure the load was
transferred perpendicularly to the slab surface, a spherical
seating was placed on top of the reaction bridge. Custom
steel plates were machined to fit each element, keeping the
bar centered in the test setup throughout the loading cycle.
Three linear voltage displacement transducers (LVDTs) were
placed concentrically around the test bar just below the steel
coupler. Effect of any load eccentricity can be accounted for
by taking the average of the three displacements. Data from
the LVDTs showed minimal eccentricity of loading.
Strain gauges were placed on either side of the test specimen at the loaded end as well as at the midpoint of the
bonded length for the straight and headed bars for one of the
three repeated specimens to evaluate the strain distribution
along the anchor. Each set of strain gauges was protected by
a 50 mm (2 in.) strip of foil tape; this length was considered
to be debonded from the surrounding concrete. To determine
the strain distribution across the bends, strain gauges were
attached parallel to the fibers at the midpoint of the bend in
the same plane as the bend. Figure 4 shows locations of the
strain gauges.
Analysis of measurements
The measurements obtained by the data acquisition
system were used to produce bond-slip relationships for the
three bar types. Although the bond stress does vary along the
bonded length, it is generally accepted to present the bond
stresses as an average shear stress along the bonded length
that is calculated by dividing the applied load by the surface
area of the bonded length (Eq. (1))

Fig. 2Typical slab layout for bent and headed bars in

Batch 1 slabs. (Note: Dimensions in mm.)
N
(1)
pdb lembed
where N is the applied load; db is the nominal bar diameter;

and lembed is the length of the bar bonded to the concrete.
Bar slip at the loaded end can be calculated by subtracting
the elongation of the bar (outside the concrete) between the
LVDT holder and the top surface of the concrete (70 mm
Fig. 3Setup for modified direct tension pullout tests.

61
Fig. 4Strain gauge placement for: (a) headed bars; and (b) bent bars.
Fig. 5Strain distribution along bonded length results from VecTor 2 model of: (a) straight (C-S-10) and headed (C-H-10)
bars; and (b) bent bars.
[2-3/4 in.]) from the average displacement measured by
the three LVDTs. Bar slip at the free end is calculated by
subtracting the elongation of the bar along the debonded
length and the bonded length from the loaded end slip using
the data collected by the strain gauges. Because the strain
gauges only measured the strain at the midpoint of the
bonded length, the strain near the free end had to be extrapolated using the results from an analytical model of the pullout
tests that was created using finite element analysis (FEA)
software (Vecchio and Wong 2002). A detailed description
of the steps used to create this model can be found elsewhere
(Vint 2012). The strain distribution predicted by the model
along the bonded length for the straight, headed and bent
bars can be found in Fig. 5 for Bar Type C with bonded
lengths of 10db. It can be seen that the strain at the free end
of the straight bar is nearly zero, which is consistent with
the existing literature that states free end strain is theoretically zero. The difference between the theoretical and experimental results could be due to the inaccuracy of the strain
gauges. The anchor head does an effective job of transferring
the strains to the free end, producing a more constant strain
distribution along the bonded length. The strain distribution
62
for the bent bars shows that the strains are small in the first
bend and quickly dissipate to zero beyond that point.
Experimental results
In the discussion of the results for all 72 pullout specimens
in this paper, the following nomenclature for the specimens
is used:
The first letter indicates the bar type. The next letter indicates the anchorage type, where S is straight, H is headed,
and B is bent. The first number indicates the ratio of bonded
length to bar diameter of 0, 5, or 10. The next number indicates the test number of the specimens within a group of
three specimens, where the third specimen has the strain
gauges along the bonded length. The last number denotes the
concrete batch number. As an example, A-S-10-1-2 designates a bond test specimen of Bar Type A with a straight
anchorage type and a bonded length of 10 times the bar
diameter. It is the first of the three specimens in that group
and was in the second batch of concrete.
Straight barsResults for the straight bars with no end
anchorage showed that as the embedment length increased
from 5db to 10db the peak average bond stress, 2, decreased.
This trend is consistent with the available literature, indiACI Structural Journal/January-February 2015
Table 2Bond test results for straight bars

Bar type
db, mm (in.)
lembed/db
N2, kN (kip)
f2, MPa (ksi)
f2/fu
2, MPa (ksi)
sl2, mm (in.)
sf2, mm (in.)
15.9
(0.626)
50.6 (11.38)
255 (37.0)
0.346
12.67 (1.838)
1.738 (0.068)
0.420 (0.0165)
10
78.7 (17.69)
396 (57.4)
0.537
9.85 (1.429)
2.99 (0.118)
0.692 (0.027)
15.9
(0.626)
62.3 (14.01)
314 (45.5)
0.246
15.59 (2.26)
1.410 (0.056)
0.656 (0.026)
10
104.8 (23.6)
528 (76.6)
0.413
13.12 (1.903)
1.919 (0.076)
0.352 (0.0139)
50.9 (11.44)
253 (36.7)
0.206
12.65 (1.835)
1.565 (0.062)
0.551 (0.022)
10
90.2 (20.3)
449 (65.1)
0.365
11.22 (1.627)
1.970 (0.078)
0.357 (0.0141)
16
(0.630)
Fig. 7Experimental bond-slip curves for straight bars A,

B, and C.
Fig. 6Failed specimens from left to right: A-S-10-1-2,
B-S-10-1-2, and C-S-10-1-2.
cating an increase in peak load of approximately 70% for all
bar types as the bonded area is doubled. A summary of the
results for the straight bars can be found in Table 2, where
the values given are the average of the three repeated specimens. The sand-coated Bar Type B displayed the largest
bond stresses of the three types of bars at both embedment
lengths. Bar Types A and C gave similar peak average
bond stresses at shorter embedment lengths, but ribbed
BarTypeC outperformed helically wrapped Bar Type A at
a longer embedment length. It should be noted that nominal
dimensions were used in calculating the bond stresses shown
in Table 2. If actual sizes of the bars are used in the analysis,
the bond strength values for all three types of the bars are
almost equal.
All bars failed by bar pullout; however, the shearing
interface is varied depending on the surface profile. When
BarType A was pulled out of the concrete the sand coating
was sheared off while the undulations remain intact,
resulting in a ductile post-peak response. When Bar Type B
is pulled out of the concrete, the sand coating is completely
sheared off. This suggests the shearing interface is between
the core of the bar and the sand coated surface profile. For
Bar Type C, the concrete is sheared between the ribs when
pullout occurs. Thus the shearing interface is between the
ribs of the bar and the concrete. The location of the shearing
interface suggests that the connection between the core
and the surface profile of the three bars is strongest for

BarTypeC, even though Bar Type B was able to sustain the
highest peak average bond stress. This observation indicates
that the thickness of ribs in Bar Type C can be optimized
to increase the bond strength. The shearing interfaces can
be clearly seen in the failed bars found in Fig. 6. Figure 7
shows the average bond stress vs. free end slip for the three
bars with 10db bonded length. It can be seen in the figure
that Bar Type A, although lower in bond strength, is capable
of maintaining higher strength over larger range of slip than
Bar Types B and C.
Headed barsTable 3 shows a summary of results for
two types of headed bars in which each value represents
an average of three readings. Bar Type A headed bars were
not available at the time of these tests. Results show that
the bars can develop a substantial tensile force as a result of
head anchorage alone. Bar Type B headed bars showed an
average peak stress of 605 MPa (87.7 ksi) while Bar Type C
displayed a stress of 507 MPa (73.5 ksi) without any bonded
length. Again, if actual diameters of the bars are used, the
stress level would be almost equal. The diameter of the
heads of the bases in Bar Types B and C bars were 48 and
40 mm (1.89 and 1.57 in.), respectively. The bar stresses that
developed in both types of headed bars with a bonded length
of 10db are approximately 50% higher than the bar stresses
developed in straight bars with bonded length of 10db. When
a bonded length of 5db is added to the anchor heads, the
increase in bar stress is only approximately 100 to 150 MPa
63
Table 3Bond test results for headed bars

Bar type
db, mm (in.)
15.9
(0.626)
16
(0.630)
lembed/db
N2, kN (kip)
f2, MPa (ksi)
f2/fu
2, MPa (ksi)
sl2, mm (in.)
sf2, mm (in.)
120.2 (27.0)
605 (87.7)
0.474
3.33 (0.131)
1.522 (0.060)
141.7 (31.9)
714 (103.6)
0.559
35.5 (5.15)
3.71 (0.146)
1.736 (0.068)
10
161.3 (36.3)
812 (117.8)
0.636
20.2 (2.93)
4.24 (0.167)
1.699 (0.067)
101.9 (22.9)
507 (73.5)
0.413
2.75 (0.108)
0.945 (0.037)
133.2 (29.9)
663 (96.2)
0.540
33.1 (4.80)
3.45 (0.136)
1.152 (0.045)
10
141.4 (31.8)
703 (102.0)
0.573
17.6 (2.55)
4.01 (0.158)
1.249 (0.049)
Fig. 8Typical bar stress-slip relationship for headed bars: (a) Bar Type B; and (b) Bar Type C.
(14.5 to 21.8 ksi) in the two types of bars. A bonded length of
10db only adds approximately 200 MPa (29.0 ksi) bar stress.
While both headed Bar Type B and C were able to
develop similar peak ultimate average bond stresses, their
bond stress-slip reponses varied in the post-peak phase just
like their straight bar behaviors. This is due to the different
manufacturing processes that are used for each bar type.
For Bar Type B, the anchor head is attached to a specially
prepared surface which consists of O-rings spaced every
5mm (0.2 in.) (Drouin 2012). Whereas for Bar Type C,
the anchor head is attached directly to the 16 mm (5/8 in.)
GFRP bar. It can be seen from the bond stress-slip curve
of both headed bar types (Fig. 8) that the behavior of the
anchor head for Bar Type B has multiple peaks and valleys,
while the failure of headed Bar Type C was mostly singular
as the stress drops significantly only once. The same failure
mode was observed for the Bar Type B headed bars for all
bonded lengths. Multiple loud bangs were observed (O-rings
failing) in Bar Type B headed specimens with no bonded
length. The peak and valley behavior seems to be due to the
load sharing mechanism in these specimens and requires
furtherinvestigation.
Due to the high ultimate strength of the bars tested, the
bonded length required to develop the full strength of the
GFRP bars would be relatively large. All headed bars at all
bonded lengths failed by pullout of the bar from the head
connection as seen in Fig. 9, where the unruptured bar pulled
out of the anchor head, which remained intact and inside
64
Fig. 9Failed headed bar specimen in concrete cylinder

with anchor head intact.
the concrete slabs. Note that the photo shown in Fig. 9 is
from a bond test series in which the bars were embedded in
concrete cylinders (Vint 2012). While it was relatively more
manageable to cut cylinders to investigate failed bars, the
failure modes of similar specimens in the two series were
found to be identical. The mechanical anchor heads are
required to develop at least 1.67 times the required level
Table 4Bond test results for bent bars

Bar
type
Fig. 10Experimental and analytical strain variation along

bonded length for specimen (C-H-S-3-1).
Fig. 11Peak loaded end-slip values for straight and

headed bars for all bar types.
of tensile load in the bar (CSA S806-12). This would result
in the design bar stress to be 360 MPa (52.2 ksi) for Bar
Type B, and 304 MPa (44.1 ksi) for Bar Type C based on
the results for anchors only (0db). The tested anchor heads
with reasonable bonded length should thus be adequate for
developing stresses similar to or larger than what would be
used for steel reinforcing bars, which usually have a yield
stress of 400MPa (58.0 ksi).
The experimental results confirmed what was found in
the finite element model, developed in the FEA software,
showing that the anchor head effectively stiffened the bars
near the free end (Fig. 10). The strain gauges at the midpoint
of the bonded length indicated that the strain at this point was
nearly the same as at the loaded end of the bar at peak loads.
This suggests that for the headed bars, the bond stress is low
and evenly distributed along the bonded length. This even
distribution is similar to the response that is often observed
with steel bars, which generally have somewhat higher bond
strength due partly to their higher stiffness (Mosley et al.
2008). The higher loads and constant strain distributions
observed for the headed bars resulted in higher loaded end
slips for these bars when compared to straight bars of the
same bonded length (Fig. 11).
Bent barsThe strength of the bar at bends is lower
than the strength of the straight portion of the bar due to
db, mm
(in.)
9.43
(0.371)
11.93
(0.470)
13.00
(0.512)
rb/db
5.4
3.0
1.75
lembed/ f2, MPa

db
(ksi)
sl2, mm
(in.)
f2/fu,
%
Rupture
location
555
(80.5)
4.58
(0.180)
66.7
Bend
775
(112.4)
4.74
(0.187)
93.1
Bend
10
785
(113.9)
3.63
(0.143)
94.2
Coupler
522
(75.7)
6.19
(0.244)
79.6
Bend
608
(88.2)
4.97
(0.196)
92.9
External
SG
10
612
(88.8)
4.63
(0.182)
93.4
External
SG
531
(77.0)
6.32
(0.249)
58.3
Bend
816
(118.4)
8.50
(0.335)
93.6
Interior
Straight
10
724
(105.0)
6.04
(0.238)
79.4
External
SG
the relatively lower strength in the transverse direction

as well as the kinking of the longitudinal fibers along the
interior bend. To determine this reduced strength, the bent
bars embedded in concrete were debonded to the top of the
loaded end of the bend. The results from these tests showed
that the bend strength is similar for all three bar types with
peak bar stresses of 555, 522, and 531 MPa (80.5, 75.7, and
77.0ksi) for Bar Types A, B, and C, respectively. All bar
types pass the requirement by CSA S807-10 that the bend
strength of a stirrup should be at least 45% of the strength of
the straight portion (Table 4). The difference in bend geometry due to manufacturing constraints should be noted as
it has been shown to have an influence on bend strengths.
BarTypeA produced the greatest bend strength, but it also
has the smallest bar diameter, 10 mm (3/8 in.) and the largest
bend radius, rb, of 5.4db. Whereas Bar Types B and C had
larger bar diameters of 12.7 and 12 mm (0.5 and 0.472 in.)
and bend radii-bar diameter ratios of 3.0 and 1.75, respectively. These bars produced slightly lower bend strengths
than BarType A most likely due to the tighter bend radii.
When the bonded length is increased to five and 10 times
the bar diameter, all bent bars ruptured along the straight
portion of the stirrup indicating that bond was fully developed. This is verified by the fact that when the bonded length
was increased from 5db to 10db the bars did not pick up additional loads, as shown in Fig. 12. For the case of Bar TypeC,
the peak stress actually dropped slightly as the bonded
length increased. This is because the observed failure mode
in the specimen with 10db bonded length was bar rupture at
the location of the external strain gauges where the bar area
had been reduced for the purpose of gauge installation, thus
causing lower failure load. From the bar stress-slip relationships for the bent bars in Fig. 13, it can be observed that
the failure mode was brittle because the bar stress decreased
dramatically after peak stresses were observed. This is
consistent with the rupture failure of GFRP bars. The testing
65
method for the bent bars was designed to mimic the behavior
of a GFRP stirrup in a large beam and all bar types were
found to be fully developed within the given slab height. In
these specimens, the requirement of CSA S806-12, that web
reinforcement should have sufficient development length
to develop its design stress at midheight of the member,
issatisfied.
The rupture location varied between the specimens with
the most common location indicated in Table 4 and Fig. 14.
In many cases, the failure was observed in the bars where
the bar area was reduced as a result of surface preparation
for the strain gauges. In some cases, the test was terminated
prematurely because of the slip of couplers. Three additional bent specimens without strain gauges were tested in
a Batch 2 slab and it was determined that bond could still be
developed within 5db, as the bar ruptured along the straight
portion of the stirrup inside the concrete.
be found in Fig. 15, where the unknown parameters 1, p,

2, and can be solved for using the method described as
follows. The values for peak average bond stress, 2, and the
free end slip at the peak average bond stress, s2, are taken as
averages from the experimental results. The method used to
determine 1 was adopted from Cosenza et al. (1997) where
the area under the experimental curve is equated to the area
under the theoretical curve (Eq. (2) and (3)) for s s2 to
solve for 1.
Analytical modeling of bond-slip relationship

The modified Bertero-Eligehausen-Popov, m-BEP, model
similar to that used by Cosenza et al. (1997), Gambarova et
al. (1998), Focacci et al. (2000), and Kadam (2006) was used
to develop a constitutive bond stress-slip (-s) relationship
for the straight bars (Fig. 15). This model can be divided into
two or three parts: the ascending branch up to the peak stress;
the descending branch after the peak stress; and a constant
stress, 3. The equations used for the different branches can
From the experimental stress-slip curves for the three

bar types shown in Fig. 7, it can be seen that the post-peak
response varies depending on the bar type. As mentioned
previously, this difference is due to each bar types surface
profile. The soft wave shape seen in Fig. 16(a) can be
attributed to the undulations found on the surface of
BarTypeA, which remain as the bar is pulled out. The steep
decrease in bond stress seen in Fig. 16(b) is due to the brittle
bond failure due to the shearing off of the sand coating
on BarType B. Finally, the initial post-peak response of
BarType C was also brittle due to the shearing off of the
concrete between the bars ribs. However, in this case, higher
and more constant residual stresses were observed due to the
more uniform friction plane.
Due to these varying post-peak responses, different
descending branch shapes were used in the analysis for the
three bar types. For Bar Types A and C, a linear decreasing
branch with a slope of p 2/s2 was observed and can be
modeled using (2) in Fig. 15, where p is the unknown
parameter that is solved by using linear regression of the
experimental results. Bar Type B had a nonlinear post-peak
response that can be modeled using (3) in Fig. 15, where
2 is determined using nonlinear regression techniques. The
constant residual stress, 3, was observed for Bar Type C and
was calibrated using a reduction factor, , that was solved
using the experimental results.
The values for known parameters 2 and s2 and unknown
parameters 1, p, 2, and can be found in Table 5. A comparison of the analytical curves and experimental results for each
Fig. 12Peak bar stress with increasing bonded length.
s
s
s
A1 = 0 2 ( s ) ds = 0 2 2
s2
a1 =
a1
ds =
2 s2
(2)
1 + a1
2 s2
1 (3)
A1
Fig. 13Typical bar stress-slip relationship for bent bars with all bonded lengths, for Bar Types (a) A; (b) B; and (c) C.
66
Fig. 14Failed bend specimens with typical rupture locations for Bars A, B, and C, from left to right.
Fig. 15Modified Bertero-Eligenhausen-Popov model for

bond stress-slip relationship (Consenza et al. 1997; Focacci
et al. 2000; Kadam 2006; Quayyum 2010). (Note: 1 mm =
0.0394 in.; 1 MPa = 145 psi.)
Table 5Mean values and coefficients of variation
for unknown parameters
2, MPa
(ksi)
s2, mm
(in.)
11.26
(1.633)
[0.447]
0.556
(0.0219)
[0.1483]
0.0622
[0.785]
0.0131
[0.514]
14.36
(2.08)
[0.312]
0.504
(0.01984)
[0.1109]
0.210
[0.918]
0.255
[0.167]
11.50
0.420
(1.668) (0.01654)
[0.0710] [0.1741]
0.110
[0.778]
0.0296
[0.224]
0.434
[0.286]
Bar type
Note: Values in brackets are coefficients of variation.
Fig. 16Experimental and theoretical curves for straight

Bar Types A, B, and C with an embedment length of 5db.
bar type is shown in Fig. 16. The values for 1 were found to
increase with increasing bar stiffness. These values changed
with relatively high coefficients of variation, primarily due
to the displacement measurements being extremely small
during the pre-peak behavior, which can be difficult to
capture with high precision. The p values correspond to the
post-peak slope, and average values of 0.0131 and 0.0296
were observed for Bar Types A and C, respectively. A higher
p value indicates a steeper descending branch, such as for
Bar Type C, where failure was due to the concrete shearing
off between the ribs. No comparisons can be made for the 2
value, as this post-peak response was unique to Bar Type B;
however, the low COV of 0.1666 indicates that the response
was consistent for this bar type.
CONCLUSIONS
Results are presented from an experimental and analytical program in which 72 bond specimens were tested under
direct pullout. Three different bar types were used. Bond
67
properties of straight bars, headed bars, and bent bars were

investigated. The main purpose of the study was to evaluate
the relative performance of GFRP bars and also compare
these results with the minimum requirements of design
codes. The following conclusions may be made from the
results of this study:
1. The surface profile of the tested GFRP bars was found
to influence the post-peak bond behavior of the bars. Where
Bar Type A is sand-coated and helically wrapped, it produced
a more ductile failure mode with higher residual stresses due
to the high friction forces between the remaining undulations and the surrounding concrete. BarTypes B and C
had less-ductile post-peak responses due to the shearing
that occurred between the bar core and the sand coating,
or between the rib deformations and the concrete for
BarTypesB and C, respectively.
2. The reinforcement bar heads, investigated here, were
able to develop tensile stress between 500 and 600 MPa
(72.5 and 87.0 ksi) in 16 mm (5/8 in.) diameter bars, which
is higher than the tensile strength of traditional steel reinforcing bars. With a bonded length of 10db, the anchor heads
were able to develop peak bar stresses in excess of 800 MPa
(116.0 ksi).
3. The bend strength of the three GFRP stirrups were
found to be similar with strengths of 555, 522, and 531 MPa
(80.5, 75.7, and 77.0 ksi) for Bar Types A, B, and C, respectively. All bent bars met the minimum requirements of CSA
S806-12 for bend strength.
4. The bent bars with 5db bonded length were able to
develop the full tensile strength of the bar as evidenced by
their rupture in the straight portion for all bar types.
5. The theoretical bond-slip models that were derived
using equilibrium equations as well as linear and nonlinear
regression techniques were found to correlate well with the
experimental results. Different surface profiles influenced
the model shape required for the post-peak response.
AUTHOR BIOS
ACI member Lisa Vint is an EIT registered with APEGA, working at
Williams Engineering Canada. She received her BEng in civil engineering
from McGill University, Montreal, QC, Canada, and her MASc in civil
engineering from the University of Toronto, Toronto, ON, Canada. Her
research interests include reinforced concrete structures and the application of glass fiber-reinforced polymer bars as internal reinforcement in civil
engineering structures.
Shamim Sheikh, FACI, is a Professor of civil engineering at the University
of Toronto. He is a member and Past Chair of Joint ACI-ASCE Committee
441, Reinforced Concrete Columns, and a member of ACI Committee 374,
Performance-Based Seismic Design of Concrete Buildings. In 1999, he
received the ACI Structural Research Award. His research interests include
earthquake resistance and seismic upgrade of concrete structures, confinement of concrete, and use of fiber-reinforced polymer in concrete structures.
68
ACKNOWLEDGMENTS
The authors wish to express their gratitude and sincere appreciation to the
sponsors of this research program for their financial and technical support.
These include FACCA Inc., Schoeck Canada Inc., Pultrall Inc., Hughes
Brothers Inc., and Vector Construction Group. Additionally, the financial support provided by the Natural Sciences and Engineering Research
Council of Canada (NSERC) is gratefully acknowledged.
REFERENCES
Ahmed, E. A.; El-Salakawy, E. F.; and Benmokrane, B., 2008, Tensile
Capacity of GFRP Postinstalled Adhesive Anchors in Concrete, Journal
of Composites for Construction, ASCE, V. 12, No. 6, pp. 596-607. doi:
10.1061/(ASCE)1090-0268(2008)12:6(596)
Ahmed, E. A.; El-Sayed, A. K.; El-Salakawy, E.; and Benmokrane, B.,
2010, Bend Strength of FRP Stirrups: Comparison and Evaluation of
Testing Methods, Journal of Composites for Construction, ASCE, V. 14,
No. 1, pp. 3-10. doi: 10.1061/(ASCE)CC.1943-5614.0000050
ASTM D7205-06, 2006, Standard Test Method for Tensile Properties of
Fiber Reinforced Polymer Matrix Composite Bars, ASTM International,
West Conshohocken, PA, 12 pp.
ASTM E488-96, 1996, Standard Test Methods for Strength of Anchors
in Concrete Elements, ASTM International, West Conshohocken, PA, 8pp.
Cosenza, E.; Manfredi, G.; and Realfonzo, R., 1997, Behavior and
Modeling of Bond of FRP Rebars to Concrete, Journal of Composites for Construction, ASCE, V. 1, No. 2, pp. 40-51. doi: 10.1061/
(ASCE)1090-0268(1997)1:2(40)
CSA S806-12, 2012, Design and Construction of Building Structures
with Fibre Reinforced Polymers, Canadian Standards Association, Mississauga, ON, Canada, 206 pp.
CSA S807-10, 2010, Specification for Fibre-Reinforced Polymers,
Canadian Standards Association, Mississauga, ON, Canada, 44 pp.
Drouin, B., personal communication with L. Vint, May 1, 2012.
fib, 2000, Bond of reinforcement in concrete: state-of-the-art report,
Bulletin 10, Fdration Internationale du Bton Lausanne, Switzerland,
434pp.
Focacci, F.; Nanni, A.; and Bakis, C. E., 2000, Local Bond-Slope
Relationship for FRP Reinforcement in Concrete, Journal of Composites for Construction, ASCE, V. 4, No. 1, pp. 24-31. doi: 10.1061/
(ASCE)1090-0268(2000)4:1(24)
Gambarova, P. G.; Rosati, G. P.; and Schumm, C. E., 1998, Bond and
Splitting: A Vexing Question, Bond and Development of ReinforcementA
Tribute to Dr. Peter Gergely, SP-180, R. Leon, ed., American Concrete
Institute, Farmington Hills, MI, pp. 23-43.
Imjai, T.; Guadagnini, M.; and Pilakoutas, K., 2007, Mechanical Performance of Curved FRP RebarsPart I: Experimental Study, Asia-Pacific
Conference on FRP in Structures, S. T. Smith, ed., International Institute for
FRP in Construction, Kingston, ON, Canada, pp. 333-338.
Johnson, D. T. C., and Sheikh, S. A., 2013, Performance of Bent Stirrup
and Headed Glass Fibre Reinforced Polymer Bars in Concrete Structures,
Canadian Journal of Civil Engineering, V. 40, No. 11, pp. 1082-1090. doi:
10.1139/cjce-2012-0522
Kadam, S., 2006, Analytical Investigation of Bond-Slope Relationship
Parameters between Fiber Reinforced Polymer (FRP) Bars and Concrete,
masters thesis, University of Missouri - Kansas City, Kansas City, MO,
193 pp.
Mosley, C. P.; Tureyen, A. K.; and Frosch, R. J., 2008, Bond Strength
of Nonmetallic Reinforcing Bars, ACI Structural Journal, V. 105, No. 5,
Sept.-Oct., pp. 634-642.
Vecchio, F., and Wong, P., 2002, VecTor 2 and FormWorks Manual,
University of Toronto, Toronto, ON, Canada, http://www.civ.utoronto.ca/
vector/ user_manuals.html. (last accessed July 17, 2014)
Vint, L. M., 2012, Investigation of Bond Properties of Glass Fibre Reinforced Polymer (GFRP) Bars in Concrete under Direct Tension, masters
thesis, University of Toronto, Toronto, ON, Canada, 213 pp.
TECHNICAL PAPER
Title No. 112-S07
Stress-Transfer Behavior of Reinforced Concrete Cracks

and Interfaces
by Ali Reza Moradi, Masoud Soltani, and Abbas Ali Tasnimi
This paper aims at determining the stress-transfer capabilities of
reinforced concrete (RC) cracks and interfaces by adopting proper
constitutive laws for aggregate interlock and dowel action mechanisms. The framework of the Contact Density Model, which was
developed in earlier research, is adopted for aggregate interlock.
The original assumptions are enhanced and modified using the
available experimental data and also by introducing four new
parameters for both normal- and high-strength concrete. The
proposed contact density function considers crack roughness
and is probabilistically idealized according to concrete strength,
aggregate size, crack width, and asperity degradation in a unified
manner. Dowel action behavior of bars is simulated by the model,
which was proposed in other research. Stress transfer across RC
cracks and interfaces is determined by a combination of the previously mentioned mechanisms and their consistent interactions. The
systematic experimental verification shows reliability and versatility of the suggested model and assumptions.
Keywords: aggregate interlock; cyclic loading; dowel action; shear.
INTRODUCTION
The local and global behavior of reinforced concrete
(RC) members and structures may considerably be affected
by cracked concrete and the stress-transfer capabilities of
cracks and interfaces. Stress transfer across cracks is a major
problem in seismic assessment and design of RC structures
because the ductility and energy absorption of members are
mainly affected by energy consumption along cracks.
Different approaches and specimens have been designed
to investigate the capabilities of stress transfer across
the cracked concrete experimentally. Some studies used
precracked specimens and the initial crack width was kept
constant during loading.1-3 However, others performed tests
with constant normal stress and measured corresponding
crack width or applied complicated loading paths.4-11
During the past years, extensive empirical and analytical models have been proposed to investigate the shear
transfer behavior of cracked concrete. Baant and Tsubaki12
pointed out the importance of considering the opening of a
crack together with the compressive stress transferred when
the shear displacement occurs.5 Considering the dilatancy,
Baant and Gambarova13 identified a nonlinear elastic model
by optimizing the fits of Paulay and Loebers3 experimental
results, taking into account the influence of the maximum
aggregate and concrete strength. Walraven and Reinhardt6
idealized the crack surface as a set of circular aggregates.
The aggregates and surrounding mortar were modeled
to be rigid-plastic bodies, where the effect of aggregate
grading was considered. Yoshikawa et al.14 proposed a pathindependent model which clearly classified the shear
transfer phenomena into four independent components:
interface shear stiffness, crack dilatancy, frictional contact,

and confining stiffness.
Investigation into the complex nature of stress transfer in
concrete needs an analytical tool that is accurate and suitable to fundamental characteristics of the stress transfer. In
this aspect, Li et al.15 made a great deal of contribution by
thoroughly investigating various models, including those
which use microscopic physical models that simulate stress
transfer based on anisotropic crack surface geometry. They
proposed the original Contact Density (OCD) Model, which
is simple and very successful in dealing with nonlinearity,
shear dilatancy, and path-dependent characteristics of stresstransfer problems. It was proposed that the complicated
asperity of a crack surface can be divided into infinitely
small pieces, defined as contact units, with various global
inclinations. They suggested a simple trigonometric formula
called contact density function (CDF) to represent the directional distribution of the contact units which are supposed
to be independent of size and grading of aggregates.5,15
Based on extensive experimental investigations, Bujadham
and Maekawa9,10 developed a relatively complicated model
called Universal Shear Transfer model for generalized
paths under static and cyclic loading.10 Following the same
approach, Ali and White16 proposed a model that introduced
friction in the CDF. Additionally, the roughness of the interface was correlated with the fracture energy of concrete to
enable the prediction of shear friction (aggregate interlock)
capacity of normal-strength concrete (NSC) as well as highstrength concrete (HSC).16 Based on the experimental observations, Gebreyouhannes11 extended the universal shear
transfer model to capture the behavior of cracked concrete
under long term loading paths.
This paper investigates the stress-transfer capabilities of
RC cracks and interfaces by adopting resisting mechanisms
(such as aggregate interlock and dowel action). The model
proposed by Moradi et al.17 is used to simulate dowel action.
To simulate the stress-transfer behavior of cracked concrete,
the OCD framework is adopted. The original assumptions
and proposals of the OCD are reviewed and the accuracy
of the model is shown through comparison with the corresponding experimental results. It will be concluded that the
CDF, suggested by Li et al.,15 is the main disadvantage of
the OCD model and has a considerable role in the amount of
transferred stress. Then the original CDF will be enhanced by
MS No. S-2013-204.R2, doi: 10.14359/51687297, received May 25, 2014, and
69
component of the local compressive deformation of each

contact unit that can be computed from the compatibility of
crack deformations (, ) (Eq. (7)).

Fig. 1Original assumptions and proposals for OCD suggested

by Li et al.15: (a) definition of contact units; (b) contact stress
direction for contact unit; (c) histogram of contact direction;
and (d) elasto-plastic model for contact compressive stress.
suggesting a consistent CDF which considers crack roughness, concrete compressive strength, aggregate size, the rate
of asperity degradation and the applied loading paths.
The deformational behavior and the ultimate capacity of
RC structures may be considerably affected by weak joints
and cracks and also the corresponding interaction between
concrete and reinforcement. Local and global behavior of
RC members and structures may be affected by cracks and
the capabilities of stress transfer. Constitutive models can be
proper tools to assess, analyze, and design RC members and
structures while considering stress-transfer mechanisms. The
proposed model can predict the shear capacity of an interface
or cracks while considering the stress-transfer mechanisms.
Furthermore, it can simulate the stress-deformation relation
of RC cracks under complicated loading paths.
AGGREGATE INTERLOCK
OCD model
The shape of a crack surface is often rough and when loads
(shear and normal displacements) are applied to the crack, the
two crack surfaces touch each other and the corresponding
normal and shear stresses are transferred (Fig. 1(a)). The
OCD model was developed based on the extensive experimental observations at the University of Tokyo. The first
as well as the main proposal is that the direction of contact
stress Z coincides with the orientation of the initial contact
unit and remains constant during loading (Fig. 1(b)).15 By
measuring the two-dimensional projection of a crack profile
in NSC (Fig. 1(c)), Li et al.15 proposed a simple trigonometric CDF, , that can describe the distribution of contact
units inclinations across the crack surface (Eq. (1)). It was
assumed that the CDF is independent of size and grading of
aggregates Gmax, as well as crack width . Under a certain
applied loading path, some contact units come into contact
while the rest of the crack surfaces have no contribution to
the transferred stresses (Eq. (2)). Figure 1(d) shows that the
contact stress con is assumed to be dependent only on the
70
(q) =
1
cos q (1)
2
= sin cos
(2)
max lim , q > qp

qp =
(3)
0,
q qp
Rs =
f y
lim
(4)
f y = 13.7 f c1/ 3 (5)
lim = 0.04 mm
(6)
con = Rs ( p)
(7)
At =
4
(8)
p
where Rs is the elastic rigidity per length; p is the local

plasticity in direction, which has path-dependent characteristics (Fig. 1(d)); and At is the whole surface area per unit
projected crack area. p can be computed by transforming
the local system defined on each contact unit (Fig. 1(b)). The
effective ratio of contact K() was introduced to take into
account the effect of crack roughness size on the mechanical
behaviors of stress transfer, apart from the effect of crack
shape expressed by the CDF (Eq. (9)). The function for K()
was proposed to be related to the maximum size of coarse
aggregate Gmax and crack width . The CDF (), proposed
to represent the directional distribution of the contact units
, which is independent of the maximum size of aggregate.15

G
K ( ) = 1 exp 1 max 0 (9)
Z = K()At()con (10)
To satisfy the equilibrium condition, the summation of

contact forces Z for all contact units must balance external
shear and normal stresses transferred across a crack as15
p
= 2p Z sin qd q (11)
p
2
p
= Z cos qd q (12)
Feenstra et al.18,19 implemented the five different

crack-dilatancy models in interface elements to model
discrete cracking. Ample attention was given to issues like
formulation of the nonsymmetrical tangential stiffness relation and the stability of the models, including eigenvalues
analyses. They concluded that the OCD model resulted in
the best correlation with experiments at the lowest computational cost. Herein, the OCD is adopted to simulate the
Fig. 2Comparison of OCD model with experimental

results reported by Maekawa et al.5: (a) test setup; (b) shear
stress versus shear displacement under repeated load path;
and (c) shear stress versus shear displacement. (Note: 1 mm
= 0.039 in.; 1 MPa = 0.145 ksi.)
response of cracked concrete. In fact, it is modified and

generalized to capture the behavior of the aggregate interlock mechanism.
To have a better insight into the accuracy of the OCD
model, the results were compared with some experimental
studies under different loading paths. A comparison of the
test and predicted results for two different crack widths
under repeated loading path demonstrates a good agreement, as shown in Fig. 2.15 The OCD can predict the general
trend and the amount of transferred shear. Figure 3 shows
the comparison between the experimental study conducted
by Paulay and Loeber3 and the results of the OCD. It can be
found that the accuracy of the model is reduced by increasing
the corresponding crack width. Increasing the crack width
leads to relatively poor correlation between the model and
the experiment. It seems that the accuracy and the reliability
of the OCD depend on the specific loading paths where the
crack width remains constant or varies little. However, in
some structures, loading paths involve both shear slip and
crack opening at the same time. To investigate stress-transfer
behavior under such deformational paths, Bujadham and
Maekawa9,10 performed a series of stress-transfer experiments on precracked concrete specimens in which crack
deformation could be arbitrarily controlled. Figures 4(a)
and (b) show the applied loading paths and the results of
the corresponding experiments, respectively. Bujadham and
Maekawa9,10 used the step-type loading paths for checking
and examining the applicability of their assumptions. The
analytical results of the OCD model are also shown in the
figure and except for the first loading step, the model predictions tend to be overestimated (Fig. 4(b) and (d)). It can be
concluded the CDF has a considerable role on the accuracy
of the model.
As can be seen, the proposed formula for the CDF by
Li et al.15 just depends on crack unit orientations (Eq. (1)).
But, some comparisons and analyses revealed its drawbacks
(Fig. 3 and 4). In fact, it does not depend on the Gmax, , or
crack asperity degradation , so it can be concluded that the
previously mentioned parameters should be considered in
the proposed generalized formulation (Eq. (13)). Herein, we
assume that the generalized form of the modified CDF can
be written as a normal distribution function with following
standard deviation.
Fig. 3Comparison of OCD model with experimental results reported by Paulay and Loeber3: (a) test specimens; and (b)
shear stress versus shear displacement. (Note: 1 mm = 0.039 in.; 1 MPa = 0.145 ksi.)
71
Fig. 4Comparison of OCD with experimental investigation reported by Maekawa et al.5: (a) step-type loading path; (b) shear
stress versus shear displacement; (c) step-type loading path; and (d) shear stress versus shear displacement. (Note: 1 mm =
0.039 in.; 1 MPa = 0.145 ksi.)

s(, fc, Gmax, , )
= sF1(, fc) F2(Gmax) F3() F4()
(13)
Equation (13) shows the general form of the proposed equation of standard deviation of the modified CDF. The first
term describes the contact units orientations and the contribution of concrete compressive strength to shear strength.
The second term represents the role of the maximum aggregate size on the shear capabilities of cracked concrete. The
last two terms are characterized loading path (crack width)
and crack roughness degradation rate, respectively. Each
term will be described in the next section.
Modified contact density function
Concrete compressive strength and contact units inclinationsLaible et al.4 experimentally simulated shear transfer
capability across a concrete interface, called Interface Shear
Transfer (IST). Experimental observation revealed that crack
surface roughness has a considerable effect on the amount
of transferred shear stress.4 In the past, extensive experimental and analytical investigations have been carried out
to investigate joint roughness, especially in rock mechanics
framework. Some efforts have been devoted to employ
normal distribution (Gaussian), while some others used
gamma distribution to determine joints roughness.20 Also,
some models adopt a non-periodic function similar to a joint
profile to approximate the joint profile by using summation
of some periodic function. This means that the joint profile
in the whole joint length regarded as a periodic function can
be divided into wavelength, amplitude, and phase.21
As it was mentioned, Li et al.15 measured the two-dimensional projection of a crack plane that was experimentally
scanned to get some ideas about the crack surface geometry.
Li et al.15 proposed Eq. (1) to account for the probabilistic
distribution of the contact units orientations. Shear behavior
of HSC was experimentally investigated by Bujadham and
72
Maekawa.9,10 They observed that the crack pattern of HSC is

completely different from NSC because HSC has a smoother
crack surface. Bujadham and Maekawa9,10 probabilistically
idealized the geometry of HSC crack surface by using a
normal distribution function (NDF)

(q) =
5
q
exp 21
p
6
(14)
It seems that any kind of crack roughness can be described

probabilistically by using a proper NDF. In fact, different
types of jointssuch as NSC, HSC, construction joints, and
masonry jointscan be defined uniformly by a simple NDF.
The general form of an NDF can be written as the following
equation (Eq. (15)). It can be defined by two parameters, m
and var
F1 (q, m, f c ) =
1
var
1 q m 2
exp
(15)
2p
2 var
where m is the mean; and var is known as the standard deviation. Li et al.15 showed that the CDF is a zero-mean function,
so in Eq. (16), var is the only unknown parameter. Knowing
var, the roughness probabilistic distribution for any kinds of
joints can be determined
m = 0 F1 (q, f c ) =
1
var
1 q 2
exp
(16)
2p
2 var
Equation (1) can be used to determine var for NSC.

p/2
E q 2 = p / 2 (q)q 2 d q =
p2
2 (17)
4
where E[2] is the mean square value of . Equation (18)

shows the standard deviation of contact units inclinations
forNSC.

var = E q 2 =
p2
2 (18)
4
Knowing var for NSC, Eq. (16) can be described and plotted
easily. Similarly, var for HSC can be derived as follows

(q) =
5
q
exp 21 var = 0.48 (19)
p
6
Equation (16) for NSC, HSC, and the original CDF are plotted
and compared in Fig. 5. The figure reveals that by increasing
the strength of concrete fc, var is decreased. The original
CDF and the proposed equation for NSC (Eq.(16)) are relatively close. The original CDF has higher values, almost
along 90 degrees || 30 degrees, but Eq.(16) has higher
values, between || and 30 degrees. This means that Eq. (16)
has a lower contribution in 90 degrees || 30 degrees
than the original CDF. But Eq. (19) has different values
for all inclinations. It can be seen that the corresponding
curve has higher values between || < 30 degrees. In fact,
the probabilistic contribution of horizontal surfaces is more
than the vertical units. This states that the crack surface
has a smoother surface than NSC because it has a smaller
var (Eq.(19)). It can be concluded that any kind of surface
geometry is described by a simple NDF.
Effect of maximum aggregate sizeDue to the roughness
of the crack faces, stress can be transferred from concrete
to concrete. This mechanism is based upon the fact that in
NSC, the aggregates have a much higher strength than the
matrix material.1 Therefore, a crack runs through the matrix
and along the interface between aggregates and cement
paste. As a consequence, the stiff aggregates cause the crack
surfaces to roughen. Decreasing the maximum aggregate size
makes crack profiles smoother, as in HSC, and subsequently
reduces the shear transfer capability. An experimental study
carried out by Thom22 stated that shear strength provided
by the aggregate interlocking increased to some extent with
increasing maximum aggregate size. Experimental results
reported by Li et al.15 as well as Wattar23 expressed that there
are no noticeable differences in crack profiles for specimens
with aggregate sizes of 15 and 25 mm (0.6 and 1.0 in.).
Thus, it seems that for Gmax < 15 mm (0.6 in.) and NSC, the
original CDF should be modified. Equation (22) expresses
the reduction of stress-transfer capability due to the size of
coarse aggregates

Gmax
(20)
15
exp(a )
( , a ) = 1
(21)
1 + exp ( a ) 1
Fig. 5Comparison between proposed model for distribution of contact units inclinations for NSC and HSC and
CDF suggested by Li et al. 15
Fig. 6Variation of F2(Gmax) with respect to . (Note:

1mm = 0.039 in.)

F2(Gmax) = [1 (, )]
(22)
where controls the contribution of maximum aggregate size in the stress-transfer capabilities of cracked
concrete (Eq. (21) and (22)). In fact, provides a family
of descending curves for F2 depending on the values of
(Fig. 6). A linearly decreasing relationship is obtained in the
particular case of = 0. Figure 6 shows the variation of
F2(Gmax) versus Gmax. For Gmax > 15 mm (0.6 in.), there is
no reduction in F2(Gmax) because the experimental results
showed no differences; but for Gmax < 15 mm (0.6 in.) and
according to , different kinds of variation and reduction
will be obtained. Proposing a unique formulation for
requires the proper and sufficient experimental data.
Effect of crack width (loading path)Experimental
studies have shown that the amount of shear transfer across
cracks directly depends on crack width. Also, applied
normal stress can be regarded as an external restraint to
adjust the corresponding crack width and have a reasonable
effect on shear transfer capability. Figure 7 shows a typical
trend of shear stress versus slip response with the probable
sequence of occurrence of the different mechanisms. The
relative participation of these mechanisms depends on crack
width. Figures7(a) and (b) indicate that for a larger initial
crack width, a lower shear strength will be expected, and
the participation of the first zone in Fig. 7(b) grows consequently. In fact, for a wider crack width, there will be fewer
73
Fig. 7Stress-transfer mechanism across cracked concrete: (a) role of initial crack width in shear capacity of crack; (b)schematic
representation of interface shear; and (c) variation of proposed relation with respect to crack width. (Note: 1 mm = 0.039in.)
contact units to activate and engage in contact with, and,
hence, the shear capacity will be lower.22 According to the
consistency of the crack deformational path, the range of
active contact units can be detected
q = D sin qc cos qc = 0

D
qc = cot 1

(23)
where c defines the lower inclinations of the active contact

units at each loading step (, ). It can be seen that increasing
leads to a decrease in the active contact units range, and
subsequently, the amount of transferred shear is reduced. As
Fig. 3 depicts, the OCD cannot capture the effect of loading
path on the behavior of cracked concrete because it depends
on . Figure 3 states that for = 0.13 mm (0.005 in.), the
OCD model has fair accuracy, but it gradually decreases.
Consequently, a proper formulation to account for the
loading path can be proposed (Fig. 7(c))

1
F3 ( ) = exp
(24)
50
Effect of roughness degradation rateLi et al.15 proposed

the CDF to idealized crack surface roughness and geometry
probabilistically and it is supposed to be independent of
crack asperity deterioration. The OCD is considered a CDF
constant during loading, unloading, and reloading processes.
On the other hand, it was assumed that there is no fracture
and roughness degradation and the surface area is constant.
Tassios and Vintzeleou25 carried out block-type experiments
exploring influence of surface roughness (smooth, sand
blasted, and rough) upon shear strength while the normal
restraint stress was kept constant during loading. The rough-
74
ness of the shear plane was measured before and after the
actual shear test. For the rough interface, the roughness,
defined as half the height of the protruding asperities, was
1.75 mm (0.069 in.) before and 1.45 mm (0.058 in.) after
testing due to the deterioration of the crack face.25 Some
efforts have been made to measure and study asperity degradation for different types of joints.15,22 It can be concluded
that the typical features of rough joint behavior under
different loading paths, such as peak shear strength and
nonlinear dilation, are significantly affected by the degradation of joint asperities.26-29
Some models considered first- and second-order asperities. The first is a Patton-type model consisting of sawtooth
asperity surfaces which degrade, and the second is a sinetooth asperity surface model in which the irregularities are
idealized as a series of continuous sine functions which
degrade (Fig. 8(a) and (b)).26 The joint asperity was formulated as the variation of the initial asperity angle, which
would be evaluated by secant or tangential slope of dilation
curves.26-29 Plesha28 and Dowding et al.29 proposed Eq.(25)
and (26) to represent the degradation of asperity angle
at each load step. In this formula, 0 is the initial asperity
angle, a controls the rate of asperity deterioration, and W is
the work or energy dissipated the frictional sliding. It seems
that the crack width has an effective contribution in joint
shear mechanism as well as shear slip. Equations (27) and
(28) are suggested to account for work dissipation across
crack with respect to the corresponding loading path. As
was stated before, the proposed NDF idealized the crack
surface considering all inclinations; that is, /2 /2.
Joint roughness degradation is expressed by an exponential
formulation which reflects the variation of the standard deviation. var0 is the initial standard deviation; controls the
rate of asperity degradation; and Wcr is the work spent on
fracture processes during loading
Fig. 8Crack roughness definitions: (a) assumptions in rock mechanics framework; (b) crack angle variation during loading
to represent roughness degradation; (c) variation of standard deviation of proposed formulation; and (d) variation of Eq. (28)
to simulate asperity degradation.
t
dW = d D W = dWdt (25)
0
= [1 exp(W)]0 (26)
dW cr = d d + d W cr = dW cr dt (27)
var
= F4 (a ) = exp( a W cr ) (28)
var 0
Equation (28) determines the variation of initial standard deviation based on work dissipation during loading.
Figure8(c) shows the variation of the standard deviation
with respect to Wcr and Fig. 8(d) explains the effect of the
standard deviation variation on the proposed NDF (Eq.(16))
qualitatively. As can be seen, loading causes a reduction in
the initial standard deviation and the corresponding NDF
becomes narrower, which means that the crack surface
asperities deteriorate.
Now, knowing all terms, Eq. (13) can be rewritten as
(q, f c , Gmax , , a ) = var

exp ( a )
1 1
1 + (exp( a ) 1) (29)
1
exp
exp( aW cr )
50
Equation (29) shows the complete and final form of the

proposed modified CDF applied at the framework of the
OCD. The suggested formulation expresses and quantifies

all contributing aspects of stress-transfer capabilities.
DOWEL ACTION
Moradi et al.17 proposed a macroscale model to simulate
dowel behavior of crossing bar across cracks. The proposed
model is established based on the experimental program and
the available experimental results. The model formulation is
based on the beam on elastic foundation (BEF) theory and
extended to the beam on inelastic foundation (BIF) theory
by proposing a consistent formula for subgrade springs. The
effect of concrete cover splitting as well as the effect of axial
stress of deformed bars is considered.17 Herein, the model is
adopted to express the dowel behavior of bars under different
loading paths. More details are available in Reference 17.
STRESS-TRANSFER MECHANISM ACROSS RC
CRACKS AND INTERFACES
To simulate the behavior of different kinds of RC cracks,
interfaces, and construction joints, reliable constitutive
models are necessary. As the shear displacement applies at
the crack plane, roughness of crack surface tends to widen
the crack width (dilatancy). This crack opening increases
the axial stress of the bar Asss, while shear displacement
causes flexure effect in the bar. The overall stress state in
reinforcing bar and surrounding concrete governs the crack
opening and slip, which can control the stress transfer across
the crack plane. The proposed procedure for finding crack
stresses and openings starts by satisfying the equilibrium
(Eq. (30)) normal to the crack (Fig. 9). So for any given
shear displacement, the unknown stresses and crack opening
can be determined in an iterative way (Fig. 10). Figure 9
shows the schematic behavior and stress-transfer mechanism of a single RC crack subjected to shear force V, due to
75
= (, )
s = s ( S , d ) (32)
(31)
where is the transverse displacement of the bar. The

compatibility between the normal and transverse displacement of the concrete and for the reinforcing bar is expressed
by Maekawa et al.5 as

Fig. 9Shear-transfer mechanisms across reinforced

concrete crack: (a) deformational and mechanical characteristics of RC interface; (b) equilibrium condition for bars
perpendicularly crossing crack plane; and (c) equilibrium
condition for inclined bars crossing crack plane.
= 2; = c(2S) (33)
where c is a factor that takes into account the variation in

the crack width from the bar surface to the concrete surface
(Fig. 9(a)).5 Once the displacement path (, ) satisfies the
equilibrium, the constitutive models for concrete shear and
steel s contributions determine the corresponding mechanisms, and the total transferred shear can be computed
asfollows

t = + s (34)
= (, )
(35)
s = s(S, )
(36)
The flowchart for solving (Eq. (30)) across RC cracks is

shown in Fig. 10, starting from an assumed crack opening
and adopting an iterative-based approach until it can satisfy
the equilibrium. The only parameter that should be sought
is the crack opening because shear displacement is known
as input.
Fig. 10Flowchart of computing shear transfer across

crack plane.
which a relative shear displacement, , results. To determine
the shear transfer across the crack plane, the equilibrium of
stresses at a crack can be written as

N
+ r s (30)
Ac
where is the reinforcement ratio; and N is externally

applied force defined positive in compression. In fact, by
means of equilibrium (Eq. (30)), it can be computed as
normal force due to aggregate interlock and the axial bar
stress s .
Normal stress of concrete (due to dilatancy) and bar axial
stress (due to transverse displacement and axial slip) is
asfollows
76
EXPERIMENTAL VERIFICATION
Systematic and extensive experimental verification is
conducted for clarifying the versatility of the proposed
model and assumptions. In first part, the proposed model
for aggregate interlock mechanism is compared with some
experimental works under different loading paths. Then the
stress-transfer behavior of a single crack is examined by
some experimental studies considering aggregate interlock
and dowel action mechanisms. Comparisons are shown in
the framework of shear stress-shear displacement predictions as well as the ultimate shear strength of RC cracks.
Also, the contribution of the aforementioned mechanisms
against the applied shear is determined for different reinforcement ratios and bar diameters.
Aggregate interlock
To verify the proposed and the suggesting assumptions, the
experimental and the computed transferred stress provided
by aggregate interlock mechanisms under different kinds of
loading paths are compared. Paulay and Loeber3 performed
tests on precracked pushoff-type specimens. The upper part
of the specimens could slide along the shear plane of the
lower part, which was fixed. The comparison of the analysis
and the experimental results are shown in Fig. 11(a) for
Fig. 11Comparison of modified contact density model and

test results: (a) Paulay and Loeber3; and (b) Thom.22 (Note:
1 mm = 0.039 in.; 1 MPa = 0.145 ksi.)
different initial crack width. Agreement between the calculated and the experimental values seems reasonable.
Thom22 experimentally simulated the capability of cracked
concrete and the role of the maximum aggregate size in
shear. The comparison of the corresponding shear stressshear displacement curves is given in Fig. 11(b). Both specimens have almost the same compressive strength, while
the maximum size of the aggregate for the second specimen
is half of the first one. Figure 11(b) shows that increasing
Gmax changes the general trend of the curve and also reduces
the shear strength of cracked concrete at a particular shear
displacement (that is, = 0.6 mm [0.02 in.]). The contribution of aggregate size on the amount of transferred shear is
well predicted.
Li et al.15 experimentally investigated the basic proposals
of the OCD under reversed cyclic loading. Figure 12 shows
a reasonable correlation between analytical and experimental results for reversed cyclic shear loading under
constant crack width of 0.3 and 0.5 mm (0.01 and 0.02 in.)
in terms of shear stress-shear displacement and also shear
stress-normalstress.
The step-type loading paths, which were introduced in
the previous section, are used to show the accuracy of the
proposed model (MCD) and are compared with the OCD
model. Figure 13 shows the shear stress-shear displacement
of the first loading path and it can be seen that the MCD
has a fair correlation with the corresponding experimental
results. It indicates the proper assumptions for the suggested
CDF. Good agreement between the model and test results is
confirmed for the second step-type loading path (Fig. 13).
Verification of stress-transfer model
Stress transfer takes place through aggregate interlock and
dowel action across cracked RC surfaces. To simulate this,
it is necessary for each of those mechanisms to have correct
performance. Prediction for the behavior and the entire
response of RC interfaces will be verified in the framework
of shear stress-shear displacement. In addition to comparing
the shear stress-shear displacement relationship, the ultimate
shear strength of precracked as well as uncracked push-off
and pulloff specimens are compared with the results of the
proposed model.
Maekawa and Qureshi30 tested beam-type specimens and
determined the contribution of each mechanism separately
(Fig. 14). Predictions for the cases described in Reference5
are shown in Fig. 14 and are in fair agreement with the
experimental results obtained. Cracks and damages in the
Fig. 12Comparison of numerical and experimental results

reported by Li et al.5 for reversed cyclic loading under constant
crack width. (Note: 1 mm = 0.039 in.; 1 MPa = 0.145 ksi.)
Fig. 13Comparison of behavior under step-type loading

path: (a) applied loading path; (b) transferred shear-stress
responses; (c) applied loading path; and (d) transferred shearstress responses.5 (Note: 1 mm = 0.039 in.; 1 MPa = 0.145 ksi.)
subgrade concrete, reduction in the bond between concrete
and reinforcement, and the localized nonlinearity in the
steel are the main sources of the nonlinear behavior. The
total behavior of the specimens are well simulated by the
proposed model. Besides the comparison between the total
response of the model and the experiment, the contribution
of each mechanism from the models is shown. It can be
concluded that the contribution of dowel action depends on
the reinforcement ratio , concrete strength fc, bar diameter
db, and the roughness of the interface.
Mattock and Hawkins31 investigated the shear transfer
capabilities of RC cracks and interfaces using uncracked
pulloff specimens. The comparison between the ultimate
shear capacity of the specimens and the corresponding
results of the model are shown in Table 1. Also, the details
of specimens and materials are shown in the table. The ratios
of the analysis to the experiment shear capacity show the fair
accuracy of the model to predict the ultimate shear strength
of the specimens where both aggregate interlock and dowel
action are the main shear mechanisms.
Similarly, the comparison with the experimental program
reported by Sagaseta and Vollum32 and the results of the
proposed model are shown in Table 1. In this table, 0 is the
initial crack width of the specimens. They use precracked
77
Table 1Comparison with experimental results reported by Mattock and Hawkins31 and Sagaseta and
Vollum32
Mattock and
Hawkins
Sagaseta and
Vollum
Experiment
Analysis
Specimen No.
db, mm
0, mm
fy, MPa
fc, MPa
fy
u, MPa
u, MPa
Ratio
7.1
9.5
341.3
33.4
2.65
5.87
5.5
0.94
7.2
9.5
341.3
35.3
3.97
6.25
6.9
1.10
7.3
9.5
341.3
34.8
5.3
6.72
7.8
1.16
PL2
8.0
0.132
550
53.1
2.31
4.85
5.14
1.06
PL2b
8.0
0.093
550
53.1
2.31
5.82
5.71
0.98
PL3
8.0
0.123
550
53.1
3.52
5.55
5.52
0.99
PL4
8.0
0.12
550
53.1
4.68
7.1
7.55
1.06
PG2
8.0
0.273
550
31.7
2.31
3.67
4.06
1.11
PG3
8.0
0.081
550
31.7
3.52
4.91
4.67
0.95
Average
1.04
Coefficient of
variation
7.45%
Notes: 1 mm = 0.039 in.; 1 MPa = 0.145 ksi.
strength, surface geometry, maximum aggregate size, the

rate of asperity degradation, and applied loading path. The
NDF was developed to express the distribution of contact
units inclinations based on the compressive strength and
material properties. By using the NDF, the rate of roughness degradation can be considered. Also, the role of aggregate size and loading paths were adopted consistent with the
OCD model framework.
The model proposed by Moradi et al.15 was adopted to
consider the dowel action mechanism across RC cracks and
interfaces. The model formulation was developed by using
the BEF analogy and was extended to the BIF by suggesting
a simple formula for subgrade spring stiffness. By combining
the aforementioned mechanisms and also considering the
nonlinear interaction of reinforcement and surrounding
concrete, the response and behavior of stress transfer across
RC cracks is verified. This model can easily be implemented
for not only discrete but also smeared crack approach to
simulate the behavior of RC members and structures.
Fig. 14Predicted and experimental shear stressassociated
displacement relation at interface (Maekawa and Qureshi30)
and test setup.5 (Note: 1 mm = 0.039 in.; 1 MPa = 0.145 ksi.)
push-off specimens to investigate the influence of aggregate
fracture on shear transfer through cracks. The average ratio
of the analysis ultimate shear to the corresponding value of
the experiment depicts the versatility of the model in determining the shear capacity of the RC cracks and interfaces.
CONCLUSIONS
Stress-transfer capabilities across RC cracks were investigated. The OCD model resulted in fair correlation with
experiments at low computational cost and it is also very
attractive for its simplicity. The original CDF depends on
the contact units inclinations and it was modified to incorporate the effective parameters. It was shown that the modified
CDF can be described based on the concrete compressive
78
AUTHOR BIOS
Ali Reza Moradi received his PhD from the Department of Civil and Environmental Engineering at Tarbiat Modares University, Tehran, Iran. His
research interests include nonlinear analysis and design of concrete structures and development of constitutive models.
Masoud Soltani is an Associate Professor of civil engineering at Tarbiat
Modares University. He received his PhD from the University of Tokyo,
Tokyo, Japan. His research interests include nonlinear mechanics and
constitutive laws of reinforced concrete, numerical modeling of masonry
structures, and seismic response assessment and rehabilitation of structures.
Abbas Ali Tasnimi is a Professor of civil engineering at Tarbiat Modares
University. He received his PhD from the University of Bradford, Bradford, UK. His research interests include nonlinear mechanics and constitutive laws of reinforced concrete, seismic nonlinear analysis, numerical
modeling of reinforced concrete and masonry structures, and seismic
response assessment of structures.
REFERENCES
1. Pruijssers, A. F., Aggregate Interlock and Dowel Action under Monotonic and Cyclic Loading, PhD thesis, Delft University of Technology,
Delft, the Netherlands, 1988, 193 pp.
2. Houde, J., and Mirza, M. S., A Finite Element Analysis of Shear

Strength of Reinforced Concrete Beams, Shear in Reinforced Concrete,
SP-42, American Concrete Institute, Farmington Hills, MI, 1974,
pp.103-128.
3. Paulay, T., and Loeber, P. J., Shear Transfer by Aggregate Interlock,
Shear in Reinforced Concrete, SP-42, American Concrete Institute, Farmington Hills, MI, 1974, pp. 1-16.
4. Laible, J. P.; White, R. N.; and Gergely, P., Experimental Investigation of Seismic Shear Transfer Across Cracks in Concrete Nuclear Containment Vessels, Reinforced Concrete Structures in Seismic Zones, SP-53,
N. M. Hawkins and D. Mitchell, eds., American Concrete Institute, Farmington Hills, MI, 1977, pp. 203-226.
5. Maekawa, K.; Pimanmas, A.; and Okamura, H., Nonlinear Mechanics
of Reinforced Concrete, first edition, SPON Press, London, UK, 2003,
768pp.
6. Walraven, J. C., and Reinhardt, H. W., Theory and Experiments
on the Mechanical Behavior of Cracks in Plain and Reinforced Concrete
Subjected to Shear Loading, HERON, V. 26, 1981, pp. 1-68.
7. Millard, S. G., and Johnson, R. P., Shear Transfer across Cracks in
Reinforced Concrete due to Aggregate Interlock and to Dowel Action,
Magazine of Concrete Research, V. 36, No. 126, 1984, pp. 9-21. doi:
10.1680/macr.1984.36.126.9
8. Divakar, M. P.; Fafitis, A.; and Shah, S. P., Constitutive Model
for Shear Transfer in Cracked Concrete, Journal of Structural Engineering, ASCE, V. 113, No. 5, 1987, pp. 1046-1062. doi: 10.1061/
(ASCE)0733-9445(1987)113:5(1046)
9. Bujadham, B., and Maekawa, K., Qualitative Studies on Mechanisms
of Stress Transfer across Cracks in Concrete, Proceedings of JSCE, V. 17,
No. 451, 1992, pp. 265-275.
10. Bujadham, B., and Maekawa, K., The Universal Model for Stress
Transfer across Cracks in Concrete, Proceedings of JSCE, V. 17, No. 451,
1992, pp. 277-287.
11. Gebreyouhannes, E. F., Local-Contact Damage Based Modeling of
Shear Transfer Fatigue in Cracks and its Application to Fatigue Life Assessment of Reinforced Concrete Structures, PhD thesis, The University of
Tokyo, Tokyo, Japan, 2006, 148 pp.
12. Baant, Z. P., and Tsubaki, T., Concrete Reinforcing Net: Optimum
Slip-Free Limit Design, Journal of the Structural Division, ASCE, V. 105,
No. 2, 1979, pp. 327-346.
13. Baant, Z. P., and Gambarova, P., Rough Cracks in Reinforced
Concrete, Journal of the Structural Division, ASCE, V. 106, No. 4, 1980,
pp. 819-842.
14. Yoshikawa, H.; Wu, Z.; and Tanabe, T., Analytical Model for Shear
Slip of Cracked Concrete, Journal of Structural Engineering, ASCE, V. 115,
No. 4, 1989, pp. 771-788. doi: 10.1061/(ASCE)0733-9445(1989)115:4(771)
15. Li, B.; Maekawa, K.; and Okamura, H., Contact Density Model for
Stress Transfer across Cracks in Concrete, Journal of the Faculty of Engineering, the University of Tokyo, V. 40, No. 1, 1989, pp. 9-52.
16. Ali, M. A., and White, R. N., Enhanced Contact Model for Shear
Friction of Normal and High-Strength Concrete, ACI Structural Journal,
V. 96, No. 3, May-June 1996, pp. 348-361.
17. Moradi, A. R.; Soltani, M. M.; and Tasnimi, A. A., A Simplified
Constitutive Model for Dowel Action across RC Cracks, Journal of
Advanced Concrete Technology, V. 10, No. 8, 2012, pp. 264-277. doi:
10.3151/jact.10.264
18. Feenstra, P.; de Borst, R.; and Rots, J. G., Numerical Study on
Crack Dilatancy Part I: Models and Stability Analysis, Journal of Engineering Mechanics, ASCE, V. 117, No. 4, 1991, pp. 733-753. doi: 10.1061/
(ASCE)0733-9399(1991)117:4(733)
19. Feenstra, P.; de Borst, R.; and Rots, J. G., Numerical Study
on Crack Dilatancy Part II: Applications, Journal of Engineering
Mechanics, ASCE, V. 117, No. 4, 1991, pp. 754-769. doi: 10.1061/
(ASCE)0733-9399(1991)117:4(754)
20. Misra, A., Effect of Asperity Damage on Shear Behavior of Single
Fracture, Engineering Fracture Mechanics, V. 69, No. 17, 2002, pp. 19972014. doi: 10.1016/S0013-7944(02)00073-5
21. Yang, Z.; Taghichian, A.; and Li, W., Effect of Asperity Order on
the Shear Response of Three-Dimensional Joints by Focusing on Damage
Area, International Journal of Rock Mechanics and Mining Sciences,
V.47, No. 6, 2010, pp. 1012-1026. doi: 10.1016/j.ijrmms.2010.05.008
22. Thom, C. M., The Effect of Inelastic Shear on the Seismic Response
of Structures, PhD thesis, University of Auckland, Auckland, New
Zealand, 1988, 203 pp.
23. Wattar, S. W., Aggregate Interlock Behavior of Large Crack Width
Concrete Joints in PCC Airport Pavements, PhD thesis, University of Illinois at Urbana-Champaign, Urbana, IL, 2001, 556 pp.
24. Divakar, M. P., and Fafitis, A., Micromechanics-Based
Constitutive Model for Interface Shear, Journal of Engineering
Mechanics, ASCE, V. 118, No. 7, 1992, pp. 1317-1337. doi: 10.1061/
(ASCE)0733-9399(1992)118:7(1317)
25. Tassios, T. P., and Vintzeleou, E. N., Concrete-to-Concrete Friction,
Journal of Structural Engineering, ASCE, V. 113, No. 4, 1987, pp.832-849.
doi: 10.1061/(ASCE)0733-9445(1987)113:4(832)
26. Lee, H. S.; Park, Y. J.; Cho, T. F.; and You, K. H., Influence of
Asperity Degradation on the Mechanical Behavior of Rough Rock Joints
under Cyclic Shear Loading, International Journal of Rock Mechanics
and Mining Sciences, V. 38, No. 7, 2001, pp. 967-980. doi: 10.1016/
S1365-1609(01)00060-0
27. Hutson, R. W., and Dowding, C. H., Joint Asperity Degradation
during Cyclic Shear, International Journal of Rock Mechanics and Mining
Sciences & Geomechanics Abstracts, V. 27, No. 2, 1990, pp. 109-119. doi:
10.1016/0148-9062(90)94859-R
28. Plesha, M. E., Constitutive Models for Rock Discontinuities with
Dilatancy and Surface Degradation, International Journal for Numerical
and Analytical Methods in Geomechanics, V. 11, No. 4, 1987, pp. 345-362.
doi: 10.1002/nag.1610110404
29. Dowding, C. H.; Zubelewicz, A.; OConnor, K. M.; and Belytschko,
T. B., Explicit Modeling of Dilation, Asperity Degradation and Cyclic
Seating of Rock Joints, Computers and Geotechnics, V. 11, No. 3, 1991,
pp. 209-227. doi: 10.1016/0266-352X(91)90020-G
30. Maekawa, K., and Qureshi, J., Stress Transfer across Interface
in Reinforced Concrete Due to Aggregate Interlock and Dowel Action,
Proceedings of JSCE, V. 34, No. 557, 1997, pp. 159-172.
31. Mattock, A. H., and Hawkins, N. M., Shear Transfer in Reinforced
ConcreteRecent Research, PCI Journal, V. 17, No. 2, 1972, pp. 55-75.
doi: 10.15554/pcij.03011972.55.75
32. Sagasta, J., and Vollum, R. L., Influence of Aggregate Fracture
on Shear Transfer through Cracks in Reinforced Concrete, Magazine of
Concrete Research, V. 63, No. 2, 2011, pp. 119-137.
79
NOTES:
80
TECHNICAL PAPER
Title No. 112-S08
Condition Assessment of Prestressed Concrete Beams

Using Cyclic and Monotonic Load Tests
by Mohamed K. ElBatanouny, Antonio Nanni, Paul H. Ziehl, and Fabio Matta
Eight prestressed T-shaped beams were tested using the cyclic
load test (CLT) method as proposed by ACI 437-12 followed by
the ACI318-11 monotonic (24-hour) load test method. The objective of the study is to assess the ability of these methods to evaluate damage in prestressed concrete (PC) beams. The test matrix
included both pristine beams (subjected to no prior loading) as
well as beams that were cracked and artificially predamaged using
accelerated corrosion techniques, impressed current, and wet/dry
cycles, prior to load testing. Deflections, crack widths, and slipping
of the prestressing strands were recorded during the load tests. The
load at which the monotonic test was conducted was chosen to be
greater than the service load of Class U PC members, which does
not allow cracking. This ensured that at the time of the monotonic
load test the specimens were significantly damaged. However, the
acceptance criteria associated with this test methodology were
still met. Only one index in the CLT acceptance criteria (deviation
from linearity) identified the condition of the specimens. The deviation from linearity index is found to correlate to the opening and
widening of cracks.
Keywords: corrosion; cyclic load test (CLT); monotonic (24-hour) load
test; prestressed concrete (PC).
INTRODUCTION
The economy of developed countries is heavily reliant
on the built infrastructure, and the deterioration of concrete
buildings and bridges is a major concern to both owners and
users. A gap exists between the annual investment needed
to improve the conditions of the U.S. infrastructure and the
amount currently spent.1 Proper assessment of the integrity
of concrete structures is key to help owners to efficiently
prioritize maintenance.
If the integrity of a structure is in question, load tests may
be used for condition assessment.2-4 The American Concrete
Institute (ACI) addresses two methods of load testing: 1)a
monotonic (24-hour) load testing per ACI 318-115; and
2) cyclic load test (CLT) per ACI 437.1R-07.6 Currently, the
CLT method is available as a provisional standard under the
leadership of ACI Committee 437.7 The two documents (that
is, ACI 318-11 and ACI 437-12) have different condition
assessment criteria based on the load-deflection response.
It is noted that the applicability of the monotonic load test
on modern structures may be questioned, as its acceptance
criteria are consistent with design principles and material
properties used in the 1920s.4,6 The CLT method is fairly
recent; therefore, more data is needed to assess the ability
of this method to determine the condition of in-service
structures. Furthermore, most of the research conducted on
both load testing methods dealt with passively reinforced
concrete (RC) structures as opposed to prestressed concrete

(PC) structures,4,8,9 a type of RC where the reinforcing steel
is used in active fashion.
This paper describes load tests conducted on eight
prestressed T-shaped beams. The CLT (ACI 437-12)7 was
performed first followed by the ACI 318-115 monotonic
load test. Five specimens were precracked and predamaged
using impressed current or wet/dry cycles to simulate the
behavior of deteriorating structures and the effect of corrosion, common in coastal areas or where deicing salts are
used, on the results of the load tests. The results are used to
assess the sensitivity of monotonic (24-hour) load test and
CLT methods to structural damage. It was shown that the
monotonic load test method failed to identify damage in the
specimens while the deviation from linearity index of the
CLT was more sensitive to damage. In uncracked (pristine)
specimens, the criterion of the deviation from linearity index
is not met when the transition from uncracked to cracked
condition takes place; thus, permanent damage occurs in
the specimen. A modification to the current deviation from
linearity acceptance criterion is proposed for the evaluation
of PC flexural members.
This study aims to evaluate the performance of current load
testing methods when used on PC flexural members. Results
indicate that the monotonic (24-hour) load test prescribed
by ACI 318-115 is not suitable for condition assessment of
PC members. The deviation from linearity index of the CLT
method yields better results; however, the current acceptance limits of this index may be associated with permanent
damage in the tested members. This study proposes a modification to the current CLT acceptance criteria for the case of
PC flexural members.
LOAD TESTING METHODS
Monotonic load test
This method is described in Chapter 20 of ACI 318-115
and has been used for decades. The structure is loaded for
24 hours and deflections are monitored and recorded
continuously or intermittently during the test. The load
magnitude of the test is determined using Section 20.3.2 of
MS No. S-2013-236.R1, doi: 10.14359/51687181, received March 14, 2014, and
81
Fig. 1Schematic of load-versus-deflection curve for: (left) two load cycles (similar to ACI 437.1R-07, Fig. 6.1); and (right)
three load sets (similar to ACI 437.1R-07, Fig. 6.2).
ACI318-11,5 where the applied load is slightly less than the
required strength (80 to 90% of the required strength) and is
considered appropriate compared to load combinations and
strength reduction factors. The structure passes the test if the
measured deflections satisfy either Eq. (1a) or (1b)

D1
lt2
(1a)
20, 000 h
Dr
D1

4
(1b)
In Eq. (1a) and (1b), 1 (in.) is the maximum deflection

recorded during the 24-hour load hold; lt (in.) is the free span
of the member under load test; h (in.) is the overall height of
the member; and r (in.) is the residual deflection measured
following load removal (r is the difference between initial
deflections, measured not more than 1 hour before the test,
and final deflections, after load removal). The final response
of the structure should be measured 24 hours after load
removal. If the measured response does not satisfy either
equation, the test may be repeated no sooner than 72hours
from the removal of the test load. The structure passes the
repeated test if Eq. (1c) is satisfied where r2 (in.) and 2
(in.) are the residual deflection and maximum deflection
measured during the repeated test.5

Dr 2
D2
(1c)
5
The objective of the method is to assess the ability of the

structure to sustain a load level near the required strength.
The test method has three notable drawbacks: 1) the long
duration (24 hours + 24 hours); 2) the significant damage that
can be imparted to the structure; and 3) it serves primarily as
a proof test, with limited ability to assess the condition of the
structure under investigation.
Cyclic load test (CLT)
This method is described in ACI 437.1R-076
and ACI437-12.7 The method was proposed by
ACICommittee437 as an alternative to the monotonic load
82
test.7-13 The test includes a series of load sets where each load
set includes two load cycles with similar load magnitude,
Load CyclesA and B, as illustrated in Fig. 1. The maximum
load is applied in five load steps with a minimum load hold
time at each step. These steps are applied in both the loading
and unloading phases. The evaluation criteria are based on
three indexes: 1) repeatability; 2) permanency ratio; and 3)
deviation from linearity.6 These indexes are calculated using
load and deflection measurements. Figure 1 shows a schematic for calculating the CLT parameters.6
1. RepeatabilityThis parameter represents the ratio
between deflections in two subsequent load cycles (Eq. (2)),
where the residual deflection (D rcycle) is subtracted from the
6
maximum deflection (D cycle
max ) in each cycle. This parameter
6
was included in ACI 437.1R-07 but is not included in
ACI437-12.7 It is included in this paper for completeness.
The repeatability criterion is not met if the index is less
than95%6

Repeatability =
B
D rB
D max
100% (2)
A
D rA
D max
2. Permanency ratioThe calculation of this parameter

was modified in ACI 437-12.7 In Eq. (3a), permanency ratio
Ipr is calculated for each load set using the two load cycles
included in the load set where Ipi (Eq. (3b)) and Ip(i+1)
(Eq.(3c)) are the permanency indexes; D ir and D (ri +1) are the
i +1)
residual deflections; and D imax and D (max
are the maximum
deflections during the i-th and (i + 1)-th cycles, respectively
(first and second load cycles of a particular load set). The
permanency ratio criterion is not met if the index
exceeds50%7

Permanency ratio ( I pr ) =
I p (i +1)
I pi
100% (3a)
D ir
(3b)
D imax
I pi =
I p (i +1) =
D r(i +1)
(3c)
i +1)
D (max
3. Deviation from linearityThis parameter measures

nonlinear response and is assessed in all load cycles. As
shown in Eq. (4), deviation from linearity IDL is calculated
as 1 minus the ratio between the secant line for the loadversus-deflection plot of a particular cycle tan(i), and the
slope of the reference point from the load versus deflection
plot (secant stiffness), tan(ref). The reference point is determined in the first load cycle. The deviation from linearity
criterion is not met if the index exceeds 25%.7

tan(a i )
Deviation from linearity (I DL ) = 1
100 % (4)
tan(a ref )
Test specimens
Eight PC T-section beams were used. Each specimen
was reinforced with two 13 mm (0.5 in.) low-relaxation
prestressing strands. The strands were prestressed to 68% of
their nominal breaking strength (fpu = 1860 MPa [270 ksi]).
The flange contained four 10 (No. 3) bars. The beams had
a length of 4.98 m (16.3 ft) and were designed to fail in
Fig. 2Cross section of specimens showing dimensions and

reinforcement. (Note: 1 mm = 0.0394 in.)
flexure. A schematic of the cross section and reinforcement

layout is shown in Fig. 2. Shear reinforcement was provided
using 10 (No. 3) stirrups with spacing of 240 mm (9.5 in.).
The beams were cast in two groups. The first group
included two beams with a 28-day concrete compressive
strength fc of 29.0 MPa (4200 psi). The water-cement ratio
(w/c) used in the mixture design was 0.4. All strands used
in casting the first group were slightly precorroded with
a uniform mass loss of 33 g/m (0.075 lb/ft) (4% per unit
length). Predamage was attained by immersing the strands
in NaCl solution and leaving them to corrode freely at
ambient temperature. One of the beams in this first set was
exposed to a chloride solution using 3-day wet/4-day dry
cycles to accelerate corrosion. This beam was preloaded
to 80% of the nominal capacity, considering tendon mass
loss resulted from precorrosion, to achieve cracks having a
width of 0.8mm (0.032 in.) to facilitate the penetration of
chlorides. This was achieved in a four-point bending setup,
with a constant moment zone of 0.90 m (3 ft). Plastic inserts
were secured into the cracks to prevent them from complete
closure. The beam was positioned flange down and a 1.22 m
(4 ft) long acrylic dike was built over the midsection of the
beam to force corrosion to occur in the maximum moment
zone, as shown in Fig. 3. A copper plate was immersed in
the chloride solution to form a galvanic cell. The corrosion
phase continued for 140 days.
The second group, identical in prestressing force and
design, included six beams. The concrete had a 28-day
compressive strength of 40.7 MPa (5900 psi) and a w/c
of 0.4. Four of the beams were preloaded to 60% of the
nominal capacity to achieve cracks with a width of 0.4 mm
(0.016in.). The tendons of the preloaded beams underwent
accelerated corrosion using applied current for different
durations to achieve the desired sectional mass and area
losses, as shown in Table 1. The theoretical mass losses were
calculated using Faradays equation, as shown in Eq.(5),
where i is the galvanic current in Amperes and t is the time
in seconds
Mass loss =
i t 55.827
(5)
2 96, 487
Fig. 3Corrosion test setup. (Note: 1 mm = 0.0394 in.)

Table 1Description of specimens
Strand condition prior to casting
Eq. (5)
Eq. (6)
Experimental mass loss, %

(area loss, %)
Corroded (4% per unit length)
0 (0)
U2
Pristine
0 (0)
U3
Pristine
0 (0)
0.4 (0.016)
Pristine
15 (15.4)
5.9 (6.4)
6.3 (6.8)
C2-0.4
0.4 (0.016)
Pristine
15 (15.4)
7.7 (8.2)
10.2 (10.7)
C3-0.4
0.4 (0.016)
Pristine
30 (30.4)
12.3 (12.8)
12.8 (13.3)
Specimen
U1
Compressive strength, Crack width,

MPa (psi)
mm (in.)
Theoretical sectional mass loss, %

(area loss, %)
29.0 (4200)
C1-0.4
40.7 (5900)
C4-0.4
C5-0.8
29.0 (4200)
0.4 (0.016)
Pristine
30 (30.4)
12.5 (13.0)
12.8 (13.3)
0.8 (0.032)
Corroded (4% per unit length)
16 (16.5)
6.9 (7.4)
4.9 (5.4)
83
Fig. 4Overview of test setup: (a) photograph showing

Specimen C5-0.8 in place; and (b) schematic of test setup.
The test matrix is shown in Table 1. An X-Y-Z format
was used, where X indicates the condition of the specimen (U for uncracked uncorroded, and C for cracked
corroded); Y is the specimen number; and Z indicates
the precrack width.
Experimental setup and instrumentation
The specimens were tested in four-point bending, as
shown in Fig. 4. Specially designed roller supports at the top
of each reaction stand were used to ensure simple support
conditions. Each support included (from bottom to top)
a steel plate, steel roller, steel plate, and neoprene pad. A
reaction frame was built to hold a 245 kN (55 kip) MTS
hydraulic actuator. The load was distributed onto the girder
through two roller supports 0.90 m (3 ft) apart. A stiffened
steel spreader beam was used to transfer the load from the
actuator to the roller supports.
Two string potentiometers per specimen were used to
measure the midspan deflection. Two displacement transducers were placed at the supports to measure settlement and
two more displacement transducers (four total) were placed
at either end of each prestressing strand to measure strand
slip. After visual detection of cracking in the CLT, two crack
mouth opening gauges were mounted across selected cracks
to measure crack widths during the remaining part of thetest.
Load intensity
The self-weight of the specimen and the loading apparatus
were determined and subtracted from the applied loads.
The test targeted three loading levels: 1) service load Ps;
2) cracking load Pcr; and 3) required strength (0.9Pn [nominal
capacity]). To obtain these loads, the prestress losses were
calculated according to ACI 318-11 and the effective
prestress fes was determined. The loading protocol was
designed such that the concrete tensile stress was limited with
the intent of minimizing or eliminating tensile cracking. The
service load level Ps was calculated as the load that would
84
cause zero stress at the bottom concrete fiber of the beam.

The cracking load was calculated as the load developing
tensile stress at the bottom concrete fiber equal to ft, the
maximum allowable tensile stress for Class U prestressed
members5 equal to 7.5fc. Both service and cracking loads
were calculated using gross section properties. The nominal
capacity was determined using cracked-section properties.
The loads for uncracked (pristine) specimens were calculated directly using section properties. For the first group
(Specimen U1), the nominal load was calculated to be
103.2kN (23.2 kip), while for the second group (SpecimensU2 and U3), the nominal load was 112.5 kN (25.3kip).
For cracked-corroded specimens, the presence of corrosion
leads to a mass loss in the exposed area of the prestressing
strands, which in turn leads to a localized reduction in
cross-sectional area of the strand and reduces the nominal
capacity of the section. The sectional mass loss was calculated by assuming that the distance of exposure in each
strand is equal to the crack width plus the distance of chloride penetration at either side of the crack. Because Faradays equation is only valid on standalone metals immersed
in solution, a modification to this equation was used to
estimate the sectional mass loss, as shown in Eq.(6).14
The sectional area loss was estimated by assuming that the
density of the corroded section is equal to the density of the
original section, as shown in Table 1.

Mass loss = 0.4651
i t 55.827
0.5624 (6)
2 96, 487
The extent of chloride penetration was estimated based

on a study by Mangat and Molloy.15 More information
regarding the effect of corrosion on the specimens can be
found in ElBatanouny16 and ElBatanouny et al.17
Test procedure
A CLT protocol was calculated for all the specimens with
a loading rate of 0.90 kN/s (0.2 kip/s). The applied loads
in each specimen differed based on the degree of corrosion
present. Tests were composed of a series of load sets, each
containing twin cycles. A typical plot of the CLT loading
protocol is presented in Fig. 5. The maximum load of each
test in Cycle 1 (P1) was equal to 75% of the service load Ps,
where service load was calculated by setting the resultant
stress at the extreme tensile concrete fiber to zero. The load
level in Cycle 2 (P2) was equal to 90% of the load level in
Cycle 1 (0.9P1). These two cycles did not contain any load
steps; therefore, the load was ramped at a constant rate to the
desired level. The hold period at the top of each load cycle
was set to 4 minutes. These cycles are not typical of the CLT
protocol as proposed by ACI 437-127 and are introduced for
the purposes of acoustic emission evaluation described in
ElBatanouny et al.17 The CLT indexes were, however, also
calculated in these cycles.
The maximum load in the second load set, Load Cycles 3
and 4, was equal to the service level load Ps. This load set
contained five load steps at each cycle in both the loading
and unloading phases. The load holds in both cycles (time
between load steps) were equal to 2 minutes at the intermediate load steps and 4 minutes at the peak load for
Cycle3. For Load Cycle 4, the 4-minute hold was employed
following the fourth load step. Changes in load hold times
are not necessarily used in CLT protocol and were inserted
to aid in the acoustic emission evaluation.16,17
The maximum load in the third load set, Load Cycles 5
and 6, was equal to the cracking load Pcr. This was similar to
the first load set, in that the load level in Cycle 6 was equal
to 90% of the load level in Cycle 5 (0.9Pcr). The remainder
of the CLT testing protocol was conducted with a number of
stepped load sets similar to load set 2 (Cycles 3 and 4).
The load test protocol contained seven load sets (14 load
cycles), exceptions being Specimens C5-0.8 and C3-0.4,
where eight load sets (16 load cycles) and six load sets
(12load cycles) were applied, respectively. The load levels
for all the specimens are shown in Table 2 as a percentage
of ultimate experimental capacity Pu of each specimen while
the second row in this table shows the theoretical load level.
A minimum load of 2.2 kN (0.5 kip) was maintained through
the test to keep the actuator engaged. The targeted load value
in the last load set was equal to 90% of the nominal capacity
(0.9Pn). The loading protocol for Specimen U1 is shown
in Fig. 5. All load hold times were 2 minutes except at the
maximum load for the initial cycles and at the fourth step in
the repeated cycles where the hold time is 4 minutes.
Following the completion of the CLT or after a specific

load set, all specimens were loaded using the monotonic
(24-hour) load test criteria per ACI 318-11,5 with the exception of Specimen C3-0.4, which exhibited spalling prior to
commencing the test. The maximum previous load observed
by a specimen during the latest CLT load set (prior to the
monotonic load test) was used as the test load magnitude
for the monotonic load test, as shown in Table 3. After both
tests were completed, the specimens were loaded to failure
to experimentally determine ultimate capacity. It is noted
that the two load tests were applied sequentially, with the
CLT being commenced first, which is not the recommended
procedure per ACI 437-12,7 where only one test is sufficient
for performance assessment.
The experimental mass loss due to corrosion damage
was measured following the failure of the specimens, as
shown in Table 1. From this table, it is clear that the experimental mass loss has a better agreement with the theoretical
sectional mass loss calculated using Eq. (6) as compared to
Eq. (5). The final failure load of the specimens is shown in
Table 3 and the failure modes are shown in Table 4. All the
specimens, except C3-0.4, which failed prematurely due
to concrete spalling, failed in one of two modes: 1) strand
rupture in the first group of specimens; and 2) failure due to
excessive residual deflection (permanent deformation) in the
second group of specimens. Failure as a result of excessive
residual deflection was determined once the residual deflection exceeded that permissible per ACI 318-11.5 The difference in failure mode is attributed to the effect of uniform
corrosion in the strands used in the first group of specimens,
which enhanced the development length and prevented strand
slipping, as discussed in ElBatanouny et al.17 A more detailed
analysis regarding failure modes, strand slipping results, and
effect of corrosion can be found in ElBatanouny et al.17
Cyclic load test (CLT)
The CLT acceptance criteria were also used to assess the
condition of the specimens, as shown in Table 5. Detailed
values for the acceptance criteria of each specimen are
Fig. 5Specimen U1 loading protocol.
Table 2Applied load levels for each specimen as percentage of ultimate capacity Pu
Specimen
Loadset 1
Cycle 1, 2
Loadset 2
Cycle 3,4
Loadset 3
Cycle 5, 6
Loadset 4
Cycle 7, 8
Loadset 5
Cycle 9, 10
Loadset 6
Cycle 11, 12
Loadset 7
Cycle 13, 14
Theoretical load level
0.75Ps
Ps
Pcr
0.60Pn
0.70Pn
0.80Pn
0.90Pn
Final failure load

Pu, kN
U1
0.24
0.32
0.50
0.61
0.71
0.81
0.88
103.7
U2
0.25
0.32
0.52
0.60
0.70
0.80
0.90
113.9
U3
0.24
0.32
0.52
0.61
0.70
0.80
0.89
113.9
C1-0.4
0.20
0.27
0.48
0.52
0.69
0.77
0.87
101.0
C2-0.4
0.20
0.26
0.48
0.52
0.69
0.78
0.87
100.3
C3-0.4
0.26
0.35
0.64
0.69
0.81
0.92
NA
76.5
C4-0.4
0.18
0.24
0.45
0.48
0.72
0.81
0.90
89.8
C5-0.8*
0.29
0.33
0.43
0.49
0.54
0.78
0.89
89.9
Specimen had additional load set (load set 8) with load = 0.95Pu.
Note: 1 kip = 4.448 kN; NA is not available.
85
Table 3Results of monotonic load test

Specimen
Maximum load experienced
Test load magnitude
Dl, mm, (Eq. (1a))
Dr, mm (Eq. (1b))
Performance
Final failure load Pu, kN
U1
0.88Pu
0.88Pu
25.8 > 2.9
3.9 < 6.4
Pass
103.7
U2
0.80Pu
0.80Pu
45.4 > 2.9
7.5 < 11.4
Pass
113.9
U3
0.80Pu
0.80Pu
66.9 > 2.9
14.6 < 16.7
Pass
113.9
C1-0.4
0.69Pu
0.69Pu
18.6 > 2.9
1.4 < 4.7
Pass
101.0
C2-0.4
0.78Pu
0.78Pu
42.5 > 2.9
5.6 < 10.6
Pass
100.3
C3-0.4
NA
76.5
C4-0.4
0.72Pu
0.72Pu
20.2 > 2.9
1.9 < 5.1
Pass
89.8
C5-0.8
0.54Pu
0.70Pu
27.0 > 2.9
1.6 < 6.7
Pass
89.9
Notes: 1 in. = 25.4 mm; 1 kip = 4.448 kN; NA is not available.
Table 4Failure mode

Specimen
Maximum deflec- Residual deflection, mm (in.)

tion, mm (in.)
Failure mode
U1
46.5 (1.83)
NA
Strand rupture
U2
103.6 (4.08)
46.2 (1.82)
Excessive deflection
U3
93.2 (3.67)
38.6 (1.52)
C1-0.4
86.1 (3.39)
40.9 (1.61)
C2-0.4
78.8 (3.09)
39.1 (1.54)
C3-0.4
90.9 (3.58)
NA
Concrete spalling
C4-0.4
66.5 (2.62)
30.9 (1.22)
C5-0.8
48.5 (1.91)
NA
Strand rupture
Note: NA is not available.
available in ElBatanouny.16 The maximum load achieved

during the test in all specimens was near 0.9Pu (Table 2).
As shown in Table 5, the repeatability index did not fail for
any of the specimens. For all specimens, 88% of the load
sets conducted had a repeatability index exceeding 100%,
which indicates that they are far from failing the index (only
seven load sets of the 56 had a value between 95 and 100%,
and none had a value below 95%). Four specimens did not
meet the permanency ratio criterion toward the end of the
test at a load of 77% of ultimate capacity Pu, which shows
that the acceptance criterion of this parameter is not sensitive to damage. The only index that was capable of assessing
damage in all the specimens was deviation from linearity. By
definition, deviation from linearity measures the deviation
in the load-deflection slope from a reference slope calculated at the beginning of the test in the linear portion of the
load-deflection plot. Therefore, the index can capture significant changes in the slope attributed to damage such as that
caused by yielding, slippage of the reinforcement, and crack
opening or widening.
The different initial conditions, cracked or uncracked,
of the specimens affected the load at which the specimens
did not meet the deviation from linearity criterion. For the
uncracked specimens (U1, U2, and U3), the acceptance
criterion was not met when cracking occurred at a load
magnitude equal to 70% of ultimate capacity (Pu). This is
highlighted in Fig. 6, showing the load-deflection relation of
Specimen U1, which represents a typical plot for uncracked
(pristine) specimens. During the last step in Cycle 9, the
86
deflection increased significantly, almost doubling its value,

resulting in a considerable decrease in the load-deflection
slope in this cycle as compared to the reference slope.
For the cracked specimens, deviation from linearity criterion was not met at a load magnitude ranging from 0.45 to 0.52
of the ultimate capacity, excluding Specimen C3-0.4, which
failed prematurely due to concrete spalling. If compared to
uncracked (pristine) specimens, cracked specimens did not
meet this criterion at a lower percentage of the ultimate
load. This is attributed to the presence of cracks in these
specimens, which caused nonlinear behavior to initiate at a
lower level of load compared to uncracked (pristine) specimens. This is opposite to the case of passively RC members,
where it is expected that the deviation from linearity index
will not be met at a higher level of load in cracked specimens if compared to uncracked specimens. Figure 7 shows
the load-deflection relationship for Specimen C1-0.4, representing the typical behavior of cracked specimens. The specimen exhibited signs of significant damage at Cycles 5 and
6; however, deviation from linearity criterion was not met
at Cycles 7 and 8. Further deviation from the initial slope
showed that nonlinearity increased with the increase of load
magnitude through the remainder of the cycles.
Monotonic load test
In all specimens, the CLT was performed first followed
by the monotonic (24-hour) load test. The residual deflection was measured prior to commencing the monotonic test
and the residual deflection criterion per ACI 437-127 was
checked; then the load was applied for 24 hours. The residual
deflection prior to the monotonic load test was not included
in the calculation of the acceptance criteria to simulate field
conditions where the residual deflection is not necessarily
known. The final residual deflection was measured 24 hours
after removal of the load. A summary of the results from the
monotonic load test is shown in Table 3. The load magnitude varied in the specimens with a range of 69 to 88%
of measured ultimate capacity Pu. All specimens did not
meet the first evaluation criterion (Eq. (1a)) and passed the
second evaluation criterion (Eq. (1b)). Therefore, the performance of all specimens was found to be satisfactory per
ACI 318-11.5 It is noted that the load magnitude applied
during the 24-hour load test significantly exceeded the
service load level for PC members, Class U.5 During the
monotonic load test the specimens were noticeably cracked
Table 5Load at which CLT criteria was not met

Specimen
Repeatability (Eq. (2))
Permanency ratio (Eq. (3a))
Deviation from linearity (Eq. (4))
Load
U1
All cycles passed
Was not met at load set 7 (Cycles 13 and 14)
Was not met at load set 5 (Cycle 9)
0.70Pu
U2
All cycles passed
All load sets passed
0.70Pu
U3
All cycles passed
0.70Pu
C1-0.4
All cycles passed
0.52Pu
C2-0.4
All cycles passed
0.52Pu
C3-0.4
All cycles passed
0.60Pu
C4-0.4
All cycles passed
0.45Pu
C5-0.8
All cycles passed
0.50Pu
Fig. 6Load-deflection relation Specimen U1 with slopes

for even cycles shown. (Note: 1 mm = 0.0394 in.)
Fig. 7Load-deflection relation Specimen C1-0.4 with

slopes for even cycles shown. (Note: 1 mm = 0.0394 in.)
with crack widths exceeding 1 mm (0.04 in.). Because significant cracking is not allowed in Class U PC members, all
specimens were considered damaged during the monotonic
load test. Therefore, the monotonic load test failed to assess
the condition of the specimens and served only as a proof
test for ultimate capacity. For PC members, a reduced monotonic test load magnitude and revisions to both the deflection
and recovery criteria should be considered.
Discussion
An acceptance criterion that did identify damage in all
specimens was deviation from linearity index, which is a
CLT-based damage index. This was true regardless of the
initial condition of each specimen, which only affected the
Fig. 8Crack widths and load versus time; Specimen

C1-0.4. (Note: 1 mm = 0.0394 in.)
level of load at which damage was assessed. In uncracked
(pristine) specimens, deviation from linearity criterion was
not met when cracks occurred. This agrees with the design
procedure of Class U PC members, where the presence
of cracks is not allowed. It is noted that upon failing the
deviation from linearity index, the cracks measured in the
uncracked (pristine) specimens had a width greater than
0.33 mm (0.013 in.). Ideally, the load test should be halted to
prevent undesirable permanent damage.
In cracked specimens, the deviation from linearity likewise identified damage as a result of the opening of existing
cracks. For this purpose, a crack comparator tool was used.18
This tool was used as a guideline for identifying the condition of the specimens tested in this study. Based on the crack
comparator tool standards, the cracks are divided into three
categories: 1) Level 1 (hairline cracks), visible cracks with
width less than 0.33 mm (0.013 in.); 2) Level 2, cracks with
widths between 0.33 and 0.63 mm (0.013 and 0.025 in.);
and 3) Level 3, cracks with widths greater than 0.63 mm
(0.025in.). Figure 8 shows measured crack widths and load
magnitude versus time for Specimen C1-0.4. The crack
widths were measured through Cycles 1 to 8 and at Cycle 10.
It can be seen that the specimens did not meet the deviation
from linearity criterion when the cracks obtained a width
greater than 0.33 mm (0.013 in.). These crack widths fall
within Level 2 of the crack comparator tool. The nonlinear
behavior in the specimen increased when the crack widths
exceeded 0.63 mm (0.025 in.) (Level 3 cracks) at Cycles9
and 10 (Fig. 8). This behavior was observed in all speci87
mens that were precracked with crack widths of 0.4 mm

(0.016in.). For the heavily cracked specimen (C5-0.8),
crack widths exceeded 0.33 mm (0.013 in.) in Cycle 3.
Bearing in mind that the specimen was initially cracked, in
the precracking test, to crack widths of 0.8 mm (0.032 in.);
the reference load-deflection slope, calculated at Cycle 1,
was smaller than the reference slope in the slightly cracked
specimens (crack widths equal to 0.4 mm [0.016 in.]). The
aforementioned results highlight the ability of deviation
from linearity index to evaluate damage in cracked specimens based on initial crack widths with reasonable comparison to the established crack comparator tool. This shows
promise toward the further work of calibrating the deviation
from linearity index.
Results of the monotonic load test showed that the acceptance criteria were always met regardless of the level of
damage. This is notable because in some cases the specimens
were loaded to 88% of the measured ultimate capacity prior
to conducting the monotonic load test and were significantly
cracked. Compliance with the criteria was in all cases due
to the insensitivity of the recovery criterion. Chapter 20 of
ACI318-115 states that excessive cracking may be considered as evidence of failure. This provision relies heavily on
engineering judgment, contradicting the purpose of load
tests that should standardize performance assessment. This
is significant in that this study is aimed at evaluating load
test methods and acceptance criteria, the scope of this work
does not consider proof testing in which the ability of a structure to withstand ultimate loading is considered. The final
decision regarding suitability of the structure is made by the
licensed design professional. However, the load test evaluation criteria should provide as much useful information
as possible to aid in this decision. Given the passing of all
monotonic load tests as described previously, it is suggested
that the acceptance criteria may therefore be revisited for
PCelements.
This study included the load testing of individual members
with simple supports and the interpretation of results. The
behavior of complex systems will differ due to the presence of alternative load paths and other mechanisms such
as moment redistribution. However, the acceptance criteria
of load tests should be applicable to simple supports, fixed
supports, and other conditions.
Recommendations
Although the deviation from linearity index provides
a warning for damage in the specimens, this warning is
provided after undue damage results in the uncracked specimens. As such, the specimen in good condition prior to testing
is damaged beyond design parameters (uncracked [pristine]
specimens) as a result of the test. In the case of pretensioned
beams such as those tested herein, a cracked condition may
result in excessive deflections and/or an avenue for chloride penetration. As a result of the load test itself, which
is ideally intended to evaluate the suitability of a structure
without imparting significant damage, damage to the structure may occur. Therefore, a modification to the acceptance
limit of the deviation from linearity index is proposed for PC
members for the case of pristine (uncracked) specimens. The
88
results indicate that the deviation from linearity index can be

given a passing grade when this index is less than 10%. This
limit has been determined by considering the results at and
beyond the theoretical cracking load. The application of this
limit will cause the test to stop prior to imposing permanent
damage. The load limit at such point, in this test, is equal to
the theoretical cracking load, which exceeds the maximum
allowable test load magnitude for the CLT method7 for the
case of Class U PC members. This is different than the
case of passively RC beams where cracked conditions are
expected under service conditions.
For the cracked specimens, the deviation from linearity
criterion was not met at the theoretical cracking load. Therefore, modifications to the current practice are not proposed
for the case of prestressed members other than Class U
(deviation from linearity can be given a passing grade if
less than 25%). The proposed limits were based on results
from beams that failed in flexure due to strand rupture or
excessive slipping as discussed in ElBatanouny.16 Therefore,
further investigation of these limits should be undertaken for
specimens with different failure modes.
Both monotonic and cyclic load testing were conducted
on eight PC beams. The specimens had different characteristics in terms of presence and width of cracks, as well as
corrosion damage. The measured load-deflection response
was used to evaluate the condition of the specimens. The
findings are summarized as follows:
1. The monotonic load test acceptance criteria gave all
tested specimens a pass grade regardless of the level of
damage. Although the beams were loaded beyond the design
criteria (Class U PC members) and were heavily cracked,
all specimens passed the recovery portion of the acceptance
criteria, thereby passing the load test.
2. Considering the passing of all monotonic load tests in
the simplest of conditions, the evaluation of the specimen (or
structure) is then delegated to the judgment of the Engineer
of Record without having sufficient information from the
testa subjective opinion.
3. The repeatability index failed to determine the condition
of the specimens and was insensitive to damage. Therefore,
this parameter appears to be unsuitable for PC members.
4. The permanency ratio index criterion was not met in
half the specimens toward the end of the test. This indicates that the limit of this index needs to be redefined for
the case of PC members. The permanency index relies on
residual or permanent deformation, which is affected by the
prestressing. The current criterion may be more sensitive for
passively RC members.
5. The deviation from linearity index succeeded in identifying damage in all the specimens. The critical crack width
at which the criterion was not met in both pristine and
slightly cracked specimens (crack width equal to 0.4 mm
[0.016 in.]), was found to be 0.33 mm (0.013 in.), which is
considered Level 2 cracks according to the crack comparator
tool used in this research. However, the pristine specimens
were damaged at the point of test failure due to the formation
of cracks.
6. The proposed acceptance limit for deviation from

linearity enables one to determine a point at which the
acceptance criteria is not met in pristine specimens prior to
the occurrence of permanent damage (cracking). The new
limit is suitable for application in uncracked PC members,
Class U, as the load level associated with this limit exceeds
the service load of such members. This suggests that the test
load magnitude associated with the CLT method should be
revised for PC members such that the maximum test load
magnitude is equal to service level load.
7. Overall, the CLT method proposed by ACI Committee
437 provides better assessment than the monotonic (24-hour)
load test. The findings are limited to PC structures with flexural failure modes and should not be extended to structures
that may exhibit shear failure without further investigation.
AUTHOR BIOS
ACI member Mohamed K. ElBatanouny is a Postdoctoral Fellow in the
Department of Civil and Environmental Engineering at the University of
South Carolina, Columbia, SC. He received his BS from Helwan University,
Cairo, Egypt, and his PhD from the University of South Carolina in 2008
and 2012, respectively. He is a member of ACI Committee 444, Structural
Health Monitoring and Instrumentation. His research interests include
structural health monitoring using acoustic emission and load testing.
Antonio Nanni, FACI, is a Professor and Chair in the Department of Civil,
Architectural, and Environmental Engineering at the University of Miami,
Coral Gables, FL, and a Professor of structural engineering at the University of NaplesFederico II, Naples, Italy. He is a member of ACICommittees 437, Strength Evaluation of Existing Concrete Structures; 440,
Fiber-Reinforced Polymer Reinforcement; 544, Fiber-Reinforced Concrete;
549, Thin Reinforced Cementitious Products and Ferrocement; and 562,
Evaluation, Repair, and Rehabilitation of Concrete Buildings. His research
interests include construction materials and their structural performance
and fieldapplication.
ACI member Paul H. Ziehl is a Professor in the Department of Civil
and Environmental Engineering at the University of South Carolina. He
received his PhD from the University of Texas at Austin, Austin, TX. He is
a member of ACI Committee 437, Strength Evaluation of Existing Concrete
Structures, and Joint ACI-ASCE Committee 335, Composite and Hybrid
Structures. His research interests include structural health monitoring and
load testing.
ACI member Fabio Matta is an Assistant Professor in the Department of
Civil and Environmental Engineering at the University of South Carolina.
He received his PhD from Missouri S&T, Rolla, MO. He is a member of
ACI Committee 440, Fiber-Reinforced Polymer Reinforcement, and Joint
ACI-ASCE Committee 446, Fracture Mechanics of Concrete. His research
interests include monitoring and assessment of constructed facilities.
ACKNOWLEDGMENTS
Special thanks are extended to the personnel of the University of South
Carolina Structures and Materials Laboratory, in particular to W. Velez,
B. Zarate, and W. McIntosh, and to Mistras Group, particularly M.Gonzalez,
for the technical support provided. E. Deaver of Holcim Cement is thanked
for support and technical input.
REFERENCES
1. American Society of Civil Engineers (ASCE), Report Card for Americas Infrastructure, Reston, VA, 2009.
2. ElBatanouny, M.; Mangual, J.; Barrios, F.; Ziehl, P.; and Matta,F.,
Acoustic Emission and Cyclic Load Test Criteria Development for
Prestressed Girders, Structural Faults and Repair 2012, Edinburgh, Scotland, UK, 2012, 9 pp.
3. Ziehl, P. H.; Galati, N.; Nanni, A.; and Tumialan, J. G., In Situ Evaluation of Two Concrete Slab Systems. II: Evaluation Criteria and Outcomes,
Journal of Performance of Constructed Facilities, ASCE, V. 22, No. 4,
2008, pp. 217-227. doi: 10.1061/(ASCE)0887-3828(2008)22:4(217)
4. Galati, N.; Nanni, A.; Tumialan, J. G.; and Ziehl, P. H., In Situ Evaluation of Two Concrete Slab Systems. Part I: Load Deformation and Loading
Procedure, Journal of Performance of Constructed Facilities, ASCE, V. 22,
No. 4, 2008, pp. 207-216. doi: 10.1061/(ASCE)0887-3828(2008)22:4(207)
6. ACI Committee 437, Test Load Magnitude, Protocol, and Acceptance
Criteria (ACI 437.1R-07), American Concrete Institute, Farmington Hills,
MI, 2007, 38 pp.
7. ACI Committee 437, Code Requirements For Load Testing of
Existing Concrete Structures (ACI 437-12), American Concrete Institute,
8. Galati, N.; Casadei, P.; Lopez, A.; and Nanni, A., Load Test Evaluation of Augspurger Ramp Parking Garage in Buffalo, N.Y., RB2C Report,
University of Missouri-Rolla, Rolla, MO, 2004.
9. Casadei, P., and Parretti, R., Nanni, A.; and Heinze, T., In-Situ
Load Testing of Parking Garage RC Slabs: Comparison Between 24-h
and Cyclic Load Testing, Practice Periodical on Structural Design
and Construction, ASCE, V.10, No. 1, 2005, pp. 40-48. doi: 10.1061/
(ASCE)1084-0680(2005)10:1(40)
10. Gold, W. J., and Nanni, A., In-Situ Load Testing to Evaluate New
Repair Techniques, Proceedings, NIST Workshop on Standards Development for the Use of Fiber Reinforced Polymers for the Rehabilitation of
Concrete and Masonry Structures, National Institute of Standards and Technology, Gaithersburg, MD, 1998, pp. 102-112.
11. Nanni, A., and Gold, W., Evaluating CFRP Strengthening Systems
In-situ, Concrete Repair Bulletin, V. 11, No. 1, Jan.-Feb. 1998, pp. 12-14.
12. Nanni, A., and Gold, W. J., Strength Assessment of External FRP
Reinforcement, Concrete International, V. 20, No. 6, June 1998, pp.39-42.
13. Mettemeyer, M., and Nanni, A., Guidelines for Rapid Load Testing
of Concrete Structural Members, Report No. 99-5, Center for Infrastructure Engineering Studies, University of Missouri-Rolla, Rolla, MO, 1999.
14. Auyeung, Y.; Balaguru, B.; and Chung, L., Bond Behavior of
Corroded Reinforcement Bars, ACI Materials Journal, V. 97, No. 2,
Mar.-Apr. 2000, pp. 214-220.
15. Mangat, P. S., and Molloy, B. T., Prediction of Long Term Chloride
Concentration in Concrete, Materials and Structures, V. 27, No. 6, 1994,
pp. 338-346. doi: 10.1007/BF02473426
16. ElBatanouny, M. K., Implementation of Acoustic Emission as a
Non-Destructive Evaluation Method for Concrete Structures, PhD dissertation, Department of Civil and Environmental Engineering, University of
South Carolina, Columbia, SC, 2012, 184 pp.
17. ElBatanouny, M. K.; Ziehl, P.; Larosche, A.; Mangual, J.; Matta,F.;
and Nanni, A., Acoustic Emission Monitoring for Assessment of
Prestressed Concrete Beams, Construction and Building Materials, V. 58,
2014, pp. 46-53. doi: 10.1016/j.conbuildmat.2014.01.100
18. Lovejoy, S., Acoustic Emission Testing of Beams to Simulate
SHM of Vintage Reinforced Concrete Deck Girder Highway Bridges,
Structural Health Monitoring, V. 7, No. 4, 2008, pp. 327-346. doi:
10.1177/1475921708090567
89
NOTES:
90
TECHNICAL PAPER
Title No. 112-S09
Crack Distribution in Fibrous Reinforced Concrete Tensile

Prismatic Bar
by Yuri S. Karinski, Avraham N. Dancygier, and Amnon Katz
The behavior of fibrous concrete containing conventional steel
reinforcement under axial tension is analyzed. The current study
reveals that, while distances between cracks in plain concrete are
equal, this is not the case for fibrous concrete. It is shown that
the crack patterns in conventionally reinforced concrete with and
without fibers are qualitatively different, even when distribution
of fibers is uniform. The paper proposes a model for the behavior
of a reinforced fibrous concrete bar subjected to increasing axial
tension load. The model was verified against experimental results
from two different sources. Based on the proposed model, an algorithm is presented to calculate the tensile forces that cause cracking
and to determine the intervals between the cracks.
Keywords: cracking pattern; fibrous reinforced concrete; tensile behavior.
INTRODUCTION
The use of fibers in reinforced concrete (RC) elements has
become more widespread in recent decades. Furthermore,
in the past decade, consideration has been paid in modern
codes to the mechanical properties of fibrous concrete used
in structural design.1,2 Fibers have been studied for their use
in structural membersfor example, as part of the shear
reinforcement.3-5 Under tension, they affect crack width,
spacing, and pattern.6 Therefore, tension stiffening, which is
an essential characteristic of reinforced concrete, is enhanced
by the presence of fibers that can bridge cracked cross
sections.7 Fiber action of this kind leads to additional advantages of using them in concrete mixturesnamely, improved
crack control8 and increased material toughness.9-11
Improved crack control has been observed in tension
elements containing both steel fibers and conventional
steel reinforcing bars at service load, at which thin and
closely spaced cracks were reported.12-14 Researchers also
noted that, interestingly, at ultimate load, only one or two
cracks widened more than the others. A similar phenomenon of localization was also observed in flexural tests of
beam specimens.15,16 One reason for this phenomenon is
fiber distribution,17,18 which causes a crack pattern in tensile
reinforced fiber-reinforced concrete (R/FRC7) elements8 that
is different from the relatively uniform pattern in plain RC
tension bars.19 Any attempt to explain this phenomenon of
non-uniform crack distribution should be based on a proper
model of the tensile behavior of fibrous reinforced concrete.
This paper presents such a model.
The behavior of an RC bar subjected to axial tension is
controlled by the steel-concrete bond, the reinforcement
ratio, and the concrete tensile strength. The behavior of
R/FRC tensile bars depends also on the fiber type and content,
especially in the post-peak range.1,20 Cracks develop at locaACI Structural Journal/January-February 2015
tions where the concrete stress reaches its tensile strength. In

plain RC (without fibers), as the tensile force increases, new
cracks, of similar width, develop midway between existing
cracks (subject to strength variation along the concrete
bar). Furthermore, all uncracked segments are of the same
length.19 This results from the fact that, at relatively small
tensile strains, the residual tensile stresses in the cracked
concrete are relatively small. In fibrous concrete, however,
fibers cause considerable residual stresses across the cracked
cross sections, which affect the cracking process.7,21
The bond between the conventional steel and fibrous
concrete is another important factor in the cracking process.
Some of the existing models assume constant bond stress
along the reinforcing bar,22 which lead to a simplified analytical solution. Yet, the bond-slip relation is more complex
and is commonly assumed to be nonlinear. Models that
apply such a nonlinear relation require the development
of a numerical crack analysis procedure.6,7,19,21 For small
steel-concrete slip values (as in the initial phase of cracking
when transverse cracks are already developed but are still
relatively narrow), the bond stress may be assumed to be a
linear function, which enables an analytical solution.
This paper proposes a simplified model for the behavior
of a reinforced fibrous concrete bar while increasing axial
tension load. The model predicts the loads that cause new
cracks to develop as well as the distances between the cracks
at various stages of loading. Note that this phase of the
cracking process (crack initiation) occurs within the elastic
range of the material behavior. Initiation of a crack can be
predicted by using a tensile strength criterion, while further
propagation of existing cracks can be handled with fracture
mechanics criteria. The current paper deals with the basic
problem that refers to the first phase of crack initiation that
precedes the phase of crack propagation, which is out of the
scope of this paper.
A simplified model is proposed for the assessment of
crack distribution in an RC tensile prismatic bar containing
fibers. The model refers to the cracking phase, during which
cracks are relatively narrowthat is, up to the yield of
reinforcing bar in any one of the cracks. An analytical soluACI Structural Journal, V. 112, No. 1, January-February 2015.
MS No. S-2013-290.R2, doi: 10.14359/51687298, received May 25, 2014, and
91
c = c0, distributed over the entire concrete cross section

Ac such that the resultant tension forces for both the actual
and equivalent stress distributions are equal
Fig. 1Problem definition: tensile reinforced concrete bar.
Ac c ( r ) dAc
*
c 0 Ac
Ac f ( r ) dAc
Ac
(1)
The coefficient may be different for various types of bars,

depending on the concrete cover (that is, on the expected
distribution of the transverse stress).
Fig. 2Stress distribution in uncracked concrete.
DIFFERENTIAL EQUATION OF REINFORCING

BAR/CONCRETE SLIP
The method of obtaining the differential equation of the
reinforcing bar/concrete slip is well documented7,21,23 and is
presented herein only for the sake of clarity of subsequent
derivations.
Equilibrium of a differential segment of the considered
specimen of length dx (Fig. 3) yields the following system
of equations
Ac d c dx = 0

Fig. 3Equilibrium in differential segment.

tion is enabled using a linear bond-slip relation during this
phase. This simplified approach enables one to calculate the
entire distribution of distances between cracks that develop
in the tensile prismatic bar, while most models refer only to
the average, minimum, or maximum crack spacing. In spite
of the simplified approach, the model yields predictions that
are in good agreement with experimental results.
PROBLEM DEFINITION
Consider a R/FRC specimen, a long bar of RC that
contains fibers that is subjected to tension force T applied
at the ends of the reinforcing bar (refer to Fig. 1). This work
refers to the cracking phase, during which crack widths are
relatively smallthat is, before yielding of the reinforcing
bar in any of the cracks. The cross-sectional areas of the
concrete and steel reinforcing bar are Ac and As, respectively,
and the total perimeter of the steel-concrete interface is
(note that in the case of a single reinforcing bar with a diameter , As is equal to 2/4 and = ). This study deals with
conventional reinforcement ratiosthat is, As << Ac. It is
also assumed that the fiber content is not very high and, thus,
the fibrous concrete is characterized by post-peak softening
after cement matrix cracking.1
Due to the strain distribution along the concrete cross
section, the tension stress in the uncracked concrete is generally non-uniformthat is, c*(r) = c0 f(r), where r is the
distance from the longitudinal x-axis and c0 is the stress at
the reinforcing bar/concrete interface (refer to Fig. 2).
Similar to the approach presented in fib Bulletin 10,23 this
research used in the following derivation a uniform stress,
92
(2)
rAc d s + dx = 0
where is the reinforcement ratio (As/Ac); is the bond

stress; and s is the steel tension stress. The following relations are valid for small deformations

c = Ec
duc
du
; s = nEc s (3)
dx
dx
where us and uc are, respectively, the steel and concrete

displacements at the steel-concrete interface; Ec is the
Youngs modulus of fibrous concrete; and n = Es/Ec (where
Es is the Youngs modulus of steel). Additionally, for small
values of steel-concrete slip w (as in the initial phase of
cracking when transverse cracks are already developed but
are still relatively narrow), bond stress is assumed to be the
following linear function

= Aw (4)
where

w = us uc (5)
The minus sign in Eq. (4) conforms to the sign agreement

according to Fig. 3. Substituting Eq. (3) into Eq. (2) and
summing the result yields

d 2 uc
d 2 us
+
n
r
= 0 (6)
dx 2
dx 2
Equations (5) and (6) yield the following relations between

the second derivatives of the displacements and w
d 2 uc
nr d 2 w
=
nr + 1 dx 2
dx 2
d 2 us
1 d 2w
=+
2
nr + 1 dx 2
dx
(7)
Substituting Eq. (3) into Eq. (2) and subtracting the

result yields

A E d 2u
d 2u
= c c 2c nr 2s (8)
2 dx
dx
Substituting the expressions from Eq. (7) and (4) into Eq. (8)
yields the following differential equation for the slip w

d 2w
2 w = 0 (9)
dx 2
where is given by

2 =
A
Ac Ec
1
1 + nr (10)
CRACK INITIATION
Cracks develop at cross sections at which the stress in
the concrete (either plain or fibrous) reaches the concretes
tensile strength fct. According to the present model, concrete
stress can be represented by either of the following two ways
(Fig. 2): 1) stress closest to the steel bar, c0 (Eq. (11) and
(12)); or 2) equivalent stress c (Eq. (13)). The first way may
be written in the following form

c0 =
c
= f ct (11)
which yields

c = fct (12)
The second way suggests that

c = fct (13)
The only difference between these two ways is in the choice

of the cracking criterionthat is, in choosing the critical
value of the calculated stress c. According to the second
way, this value is equal to the concrete tensile strength
(Eq.(13)). However, when the first way is applied, this critical value is smaller than the tensile strength and is obtained
by multiplying the concrete tensile strength by a coefficient
(smaller than 1.0 [Eq. (12)]). The choice of this criterion
depends on the expected distribution of the transverse stress
in the concrete (this choice is out of the scope of this paper).
In the following derivation, the cracking criterion has the
form c = fct, where fct is an equivalent tensile strength,
which is given by either Eq. (12) or (13).
Fig. 4Equilibrium
adjacentcracks).
in
segment
(between
two
Consider an element of length Lc located between two

cracks (Fig. 4). The origin of the local coordinate system
is set at the center of the element and the reinforcing bar/
concrete slip within the element (Lc/2 < x < Lc/2) is
governed by Eq. (9).
Bond analysis of an RC specimen that does not contain
fibers commonly assumes that concrete stresses at the crack
are zero. In fibrous concrete, however, the fibers provide
a residual tensile stress, denoted herein as f. The current
model refers to the initiation of cracks, where this phase
occurs within the elastic range of the steel behavior. Furthermore, in concrete reinforced with hooked-end steel fibers,
the stress across the crack remains relatively constant for
rather large crack opening.24 Thus, the residual stress in this
phase may be considered constant and can be obtained from
standard tests for determination of the mechanical properties of fibrous concrete (for example, RILEM25). This is an
empirically calibrated parameter that represents the overall
action of fibers that bridge a crack, including the effects of
their quantity, elastic modulus, aspect ratio, bond properties,
orientation, and other characteristics of the fibers and the
surrounding matrix.20 Note that due to uneven fiber distribution, residual stress f may vary among different cracks.
As a result, steel stress at each crack may also vary to maintain equilibrium along the specimen (the total axial force at
each cross section is equal to external force T). The forces
in the steel and in the fibrous concrete at the cracked cross
sections located at the edges of the element are given by the
following expressions
Es As
dus
dx
= T f Ac
x = Lc / 2
du
Ec Ac c
dx
(14)
= f Ac
x = Lc / 2
where f = f(x = Lc/2) is the residual stress at the lefthand and right-hand edges of the element (refer to Fig. 4).
Due to the longitudinal strain distribution along the crack,
residual stress is generally non-uniform. Similar to the stress
distribution in the uncracked cross section (Fig. 2), a coefficient similar to (Eq. (1)) allows one to use an equivalent
uniform residual stress, represented herein by the term f.
Hence, f is applied in the calculations similar to fct, either
according to the criterion stated in Eq. (11) and (12) or that
stated in Eq. (13).
Extracting the expression for
93
dw
dus duc
=
dx x = Lc / 2 dx
dx
x = Lc / 2
duc
dx
dus
dx
from Eq. (14) yields the following boundary conditions for

Eq. (9)
f
dw
T 1 1
1
=
+
(15)
dx x = Lc / 2 Ec Ac nr nr Ec
Stresses in concrete and steel

The solution of Eq. (9) with boundary conditions (Eq.(15))
takes the following form
w=
C1 cosh ( x ) C2 sinh ( x )
(16)
+
Lc
Lc
sinh
cosh
2
2

C2 =
+f f 1
+ 1
2 Ec nr
(17)
c = Ec

(19)
dus
1 ( f f ) sinh(x)
=
dx nr 2 Ec
L
sinh c
2
1 f + f
T
1 cosh(x)
+ D2
nr 2 Ec
Ec Ac nr + 1
Lc
cosh
2
where D1 and D2 are constants of integration. According to

Eq. (14), the boundary conditions for the left-hand side of
the segment are
94
(20)
nrEc Ac
T
1

Ec Ac ( nr + 1)
(21)
duc ( f f ) sinh(x)
=
2 Ec
dx
L
sinh c
2
+f + f T
1 cosh(x)
1
T
+
+

L
r
1
r
+ 1)
E
A
n
A
(
n
+
2
c
cosh
c
c
c
2
(22)
+f + f T
1
cosh(x)
T
n
+

Ac nr + 1
L Ac (nr + 1)
2
cosh c
2
cosh ( x )
sinh ( x )
d w
= C1
+ C2
(18)
2
L
dx
L
sinh c
cosh c
2
2
+f + f
T
1 cosh(x)
+ D1
+
Ec Ac nr + 1
Lc
2 Ec
cosh
2
T f Ac
dus 1 1 ( f f ) sinh(x)
s = nEc
=
dx r r
2
L
sinh c
2
duc ( f f ) sinh(x)
=
2 Ec
dx
L
sinh c
2
Ec
These constants also satisfy the boundary conditions for the

right-hand side (refer to Eq. (14) for x =+Lc/2).
Equations (19) and (20) yield the concrete and steel
stresses as follows
T 1 f +f 1
+ 1
2 Ec nr
Ec Ac nr
Substituting Eq. (17) and (18) into Eq. (7) and integrating
once yields the concrete and steel strains
=
x = Lc / 2
D1 = D2 =
The second derivative of the slip w (Eq. (16)) is given by

x = Lc / 2
Hence, D1 and D2 are given by the following expression
where is given in Eq. (10), and C1 and C2 are the following

constants
C1 =
It should be noted that under external tension force T,

both concrete and steel stresses at the ends of the element
(x = Lc/2) must be non-negative. This leads to the evident
condition on the residual stresses at the cracked cross
sections, 0 f T/Ac. In addition, residual stresses at the
cracked cross sections, f, must be lower than the tensile
strength of the fibrous concrete fct, yielding the following,
final requirement

0 f min , f ct (23)
A
c
CRACKING PROCESS
For an idealized representation of the material, as the
external load T increases, the maximum stress within the
concrete, c,MAX, increases as well and a crack develops
when (and if) c,MAX reaches the concrete tensile strength
fct. The RC bar specimen is divided by the first crack into
two uncracked segments (Fig. 5(a)) and with the increase
in T, this cracking process continues within each of the
two uncracked segments, whereby each occurrence k of
c,MAX=fct corresponds to 2k uncracked segments and 2k 1
cracks (for example, refer to Fig. 5(b) for k = 2). Higher
Fig. 5Cracking process (first two stages) in tensile reinforced concrete bar.
external loads (up to the yield load of the steel bar) correspond to larger number of shorter, uncracked segments. A
new crack develops in an uncracked segment only if the
maximum value of c (Eq. (22)) equals the value of fct within
the segment (Fig. 6).
For plain concrete, in which concrete stress at the cracks
drops to zero, the crack spacing is uniform (that is, all
uncracked segments have the same length). This, however,
is not the case for fibrous concrete, as will be shown below.
Zeroing the first derivative of c (Eq. (22)) leads to the
following expression for the location of the maximum
concrete stress x* (according to the local coordinate system
[Fig. 4 and 6])
1
B
(24)
x* ( Lc ) = arctanh
Lc
tanh 2
where
B=
+f f
(25)
2T
+f f
Ac ( nr + 1)
Fig. 6Stress distribution in concrete within uncracked

segment upon commencement of new crack.
+f f sinh x* ( Lc )
T
1
+
Ac ( nr + 1)
2
L
sinh c
2
*
+f + f
T
1 cosh x ( Lc )
= f ct
+
2
Ac ( nr + 1)
L
cosh c
2
(27)
Note that the condition dc/dx = Ec(d2uc/dx2) = 0 is satisfied

for cross sections for which x = x*. This condition, together
with Eq. (6) and (8), imply that for cross section x = x*, the
bond stress between the concrete and the steel is zero.
Therefore, there is no slip in this cross sectionthat is, w(x*)
= 0, and the strains in the concrete and in the steel are equal.
The well-known rule of mixtures26 may therefore be applied,
and the concrete stress at x = x* is equal to (T/Ac)(1/(nr + 1)).
It follows that a crack can exist only when

T Ac(n + 1)fct (28)
It further follows that Eq. (23) may be rewritten as follows

If |x (Lc)| > Lc/2, the maximum of the function of c
(Eq.(22)) occurs outside the segment and the function
inside the segment is monotonic and less than fct (Eq. (23)).
Hence, only Lc/2 x*(Lc) Lc/2 is relevant for the case of
further cracking. This range of x*(Lc) and Eq. (24) yield the
following condition for the segment length for which further
cracking is possible
*
Lc
2
arctanh
( B ) (26)
It is noted that the condition (Eq. (26)) is required but not

sufficient for cracking.
The condition c(x*) = fct (Eq. (22)) yields the following
equation for the force T that causes the initiation of a new
crack within the segment (of length Lc)
0 f fct (29)
Substituting x* from Eq. (24) into Eq. (27) yields the

following implicit equation for T at cracking

L
L
tanh 2 c B 2 = C sinh c (30)
2
2
where B is given in Eq. (25) and C is given by
2T
2 f ct
Ac ( nr + 1)
C=
(31)
2T
+f f
Ac ( nr + 1)
Equation (30) always has a solution, as ensured by Condition (Eq.) (26), which represents a physical state of cracking
and guarantees that the expression under the square root
95
in Eq. (30) is non-negative. According to Conditions (Eq.)

(28) and (29), C is always positive and so Eq. (30) can be
squared, yielding the following biquadratic equation
L
L
tanh 4 c 1 + B 2 C 2 tanh 2 c + B 2 = 0 (32)
2
2
Considering the definitions of B and C (Eq. (25) and (31))

and Conditions (Eq.) (28) and (29), it can be shown that
all four roots of this equation are real. While two of those
solutions are negative, and therefore irrelevant to the current
physical problem, the positive solutions are given by
2
( Lc )1,2 = arctanh
1
1 + B2 C 2
2
(1 + B
C2
4 B 2 (33)
It can be shown, based on Conditions (28) and (29), that the

discriminant in Eq. (33) is always positive. It can be further
shown that only the larger root of Eq. (33) satisfies Condition (26) and so the correct relation between the length of the
segment, Lc, and the force T that initiates a new crack in the
segment is given by
Lc =
2
1
1 + B2 C 2 +
arctanh
2
(1 + B
C2
4 B 2 (34)
Equation (34), with B and C given by Eq. (25) and (31),

yields the following two solutions for T
Ac (1 + nr)
2
(35a)
L
2
2 Lc
2
sinh
cossh c
+D
+
2
2
f + f +
L
c
2
2
2
c
sinh
cosh
2
2
T=
where = f+ f, and = 2fct (f+ + f).

The condition that T must go to infinity as Lc approaches
zero (that is, lim (T ) = ) leads to the choice of a minus sign
Lc 0
in the denominator of Eq. (35a); that is

T=
Ac (1 + nr)
2
(35b)
L
L
2 sinh 2 c + D 2
cossh c
+
2
2
f + f +
Lc
2
2
2 Lc
sinh
cosh
2
2
When T reaches this value, two new segments appear

within the segment in question and their lengths are equal
to Lc/2+x* and Lc/2 x* (left- and right-hand segments,
respectively [Fig. 6]).
When f+ > f, x* is positive (Eq. (24) and (25)) and the
left-hand segment is larger, and vice versa. In the symmetric
case of f+ = f = f, B is equal to zero and the maximum
concrete stress (as well as the minimum steel stress) occurs
96
Fig. 7Concrete stress distribution at onset of first crack

(coordinates along the bar and concrete stress are normalized with respect to the bars length and fct, respectively).
at the middle of the segment (x* = 0). In this case, the value
of the force T is given by
L
f f ct cosh c
2
T = (1 + nr) Ac
(36)
Lc
1 cosh
2
In a plain RC tension bar in which f = 0, cracks will appear

at equal intervals (because x* = 0 in all segments). In a fibrous
RC bar, on the other hand, the stresses f in the cracked cross
sections are greater than zero and are equal to zero only at
the bar edges. Residual stresses in the segments located near
the left-hand edge of a tension bar are, therefore, f = 0 and
f+ = f > 0; whereas near the right-hand edge of a tension
bar, residual stresses are f+ = 0 and f = f > 0. For these
edge segments, x* 0 and the cracking force must be calculated using Eq. (35b). As a result, crack patterns in conventionally reinforced concrete with and without fibers will be
qualitatively different (even if fiber distribution is uniform
that is, f+ = f in all internal segments).
The aforementioned residual stress is distributed uniformly
along the cracked cross section (Fig. 4) and it represents an
average value of the tractions resulting from the local action
of the fibers. Because it is reasonable to assume that the
average fiber distribution in well-mixed concrete mixtures
is essentially the same for all cross sections, the assumption in the following derivation is that residual stresses in
all internal, cracked cross sections are equal (f+ = f = f).
Thus, the location x* of the maximum concrete stress in an
internal uncracked segment, which is bound by two existing
cracks, is equal to zero (Eq. (24) and (25)). Furthermore, for
the case of an internal segment like this, the cracking tension
force is given by Eq. (36) (unlike the case of edge segments,
in which f+ f and Eq. (35b) applies).
The above model refers to a condition of constant concrete
tensile strength fct in each of the cross sections. This condition is based on the understanding that for common sizes of
RC members, the distribution of concrete properties across
each cross section is similar. Therefore, the variation in fct
(which is essentially the mean value of concrete tensile
strength across a section) along the length of the bar is very
limited. In relatively long bars, it is, however, important to
take into account the distribution of fct along the bar. This
point is illustrated in Fig. 7, which shows the concrete stress
distribution c(x) (Eq. (22)) for a force Tc1 that initiates the
Fig. 8Schematic cracking process of fibrous reinforced concrete bar under axial tension.
first crack in the segment (Eq. (36) with f = 0) for two
lengths Lc of uncracked bars: a relatively short bar and a bar
three times longer (denoted L and 3L in Fig. 7).
Figure 7 shows that as the length of the bar increases, the
stress distribution along the center of the segment becomes
more uniform and extends over a longer portion of the
segmentthat is, the gradient of c is relatively small. In
such cases, variation of fct along the length of the bar can
lead to a crack, which is not necessarily located at the calculated point of the mean concrete tensile strength.
ALGORITHM FOR CALCULATING CRACKING
PROCESS
The general procedure for calculating tensile forces that
cause cracking and the intervals between the cracks is as
follows:
Stage 1. The first cracking force Tc1 is calculated using
Eq. (36), with f = 0 and the length of the entire bar substituted for Lc. The first crack is set at the center of the bar
(x* = 0 [Fig. 5(a)]).
The following steps refer to the left-hand half of the bar,
based on a theoretical symmetric cracking pattern (Fig. 8(a)).
Stage 2. The second cracking force Tc2 is calculated using

Eq. (35b), with f = 0, f+ = f > 0, and half of the bar length
substituted for Lc. Parameter B is calculated from Eq. (25),
with Tc2, and is then is used in Eq. (24) to calculate x*. The
lengths of the two new segments, L2left and L2right (Fig. 8(b)),
are given by

Lileft = Lc/2 + x*; Liright = Lc/2 x*; i = 2, 3, 4... (37)
Note that in this case (f+ > f), x* > 0 and, therefore, L2left>
L2right. Hence, the following crack is expected to develop
within L2left, which is the (left-hand) end-segment of the bar.
Stage 3. The third cracking force Tc3 and the corresponding
crack location (x*) are calculated as in Stage 2, with Lc = L2left
(f = 0, f+ = f > 0). The lengths of the two new segments,
L3left > L3right (Fig. 8(c)), are again given by Eq. (37).
At this stage, each half of the bar is subdivided into three
segments of different lengths. (On the left hand, they are
L3leftt L3right L2right [Fig. 8(c)]).
The subsequent stages are calculated as follows:
Stage i (i 4). 1. Find the largest internal segment from
the previous stages. Substitute the length of this segment,
Lint,max, for Lc in the calculation of cracking force Tci,1 using
97
Fig. 9Comparison between model and experimental

results27 (measured maximum value of 198 mm at strain of
0.00049 is not shown). (Note: 1 mm = 0.039 in.)
Eq. (36), with f 0 (f is equal to the residual stress of
fibrous concrete).
2. If Lint,max is greater than the length of the edge segment,
then this will be the next cracked segment, Tci=Tci,1, and the
crack will be set at the center of the segment (x* = 0 and Lileft
= Liright [Fig. 8(d)]).
3. Else, calculate the cracking force Tci,2 for the edge
segment and its corresponding crack location x2* (Eq. (24)),
as in Stage 2 with f = 0, f+ = f > 0.
The next cracking force Tci = min{Tci,1, Tci,2}. If Tci = Tci,2,
then the lengths of the two new segments, Lileft > Liright, are
given by Eq. (37), with x* = x2*. Else, x* = 0 and Lileft = Liright.
Note that this procedure is not valid when T exceeds the
yield force of the steel reinforcing bar.
A flow chart for the cracking process is given in
theAppendix.
COMPARISON WITH EXPERIMENTAL RESULTS
The above approach is compared both with experimental
results reported by Leutbecher and Fehling27 and with results
obtained by the authors at the Technions National Building
Research Institute (NBRI).
Comparison with results obtained by Leutbecher
and Fehling
Leutbecher and Fehling reported on experiments with
ultra-high-performance concrete in which 1300 mm
(51.2in.) long concrete bars were subjected to tension.
The bars, which had a 70 x 220 mm (2.8 x 8.7 in.) cross
section, were reinforced with four 12 mm (0.5 in.) steel bars
(St1470/1620) and contained 0.9% 17 mm (0.7 in.) steel
fibers. These authors reported minimum, maximum, and
average values of crack spacings for three different levels of
reinforcing bar strain (Fig. 9).
The calculation algorithm described in the previous section
was performed with the following data, according to the
values reported by Leutbecher and Fehling: Ac = 15,400mm2
(23.9 in.2); As = 452 mm2 (0.7 in.2); = 151 mm; (5.9 in.);
tensile strengths are 9 MPa (1305 psi); residual strength is
8MPa (1160 psi); and Ec = 50,000 MPa (7252 ksi) (based on
a reported compression strength of 160 to 190 MPa [23.2 to
98
Fig. 10Effect of parameters A and . (Note: 1 mm =

0.039 in.)
27.6 ksi]). Additionally, the values of the following parameters were completed from Leutbecher28: A = 665 N/mm3
(2450 kip/in.3), and Es = 205,000 MPa (29,733 ksi).
Here we have set a value of 0.5 for and used the criterion specified by Eq. (11) and (12). This value of was
calibrated to best fit a single experimental result (at steel
strain of 0.0045) and then the same value was applied for
all other values of steel strain. These data yield n = 4.1,
= 0.029, =0.035 mm1 (0.884 in.1), fct = 0.5 9 = 4.5MPa
(653psi), f = 0.5 8 = 4 MPa (580 psi).
Figure 9 presents a comparison between the measured
and calculated values of uncracked segment lengths for
the earlier parameters. The figure shows good agreement
between the average values of the measured and calculated
results (mean relative error and standard deviation of 0.15
and 0.25, respectively), especially for strains greater than
0.0013 (mean relative error and standard deviation of 0.05
and 0.18, respectively). A fair agreement was obtained for
the maximum values of uncracked segment lengths (mean
relative error and standard deviation of 0.26 and 0.12,
respectively), although there is poor agreement for the
minimum values.
Figure 10 shows a parametric study of the sensitivity of
the average crack spacing to A and . For clarity, the figure
also shows the average values that correspond to the various
experiments.27 It can be seen that the effect of parameter
is pronounced at small steel strains whereas it is negligible
at relatively larger strains (see two graphs of A = 340 N/mm3
[1253 kip/in.3] in Fig. 10). An opposite effect was found for
the bond-slip coefficient A, which has a pronounced effect
only at relatively larger strains (see = 0.7 in Fig. 10).
Comparison with results obtained at NBRI
Experimental procedureFour types of concrete were
tested: normal-strength concrete (NSC) and high-strength
concrete (HSC), with or without fibers. The tests considered
herein are of those specimens that contained steel fibers: two
of HSC and one NSC (a second NSC specimen exhibited
technical problems). Table 1 presents the mixture composition
of these mixtures together with their mechanical properties.
Concrete prisms, 75 x 75 x 670 mm (3 x 3 x 26.4 in.),
were prepared with one 12 mm (0.5 in.) deformed steel reinforcing bar located along the center of each prism with both
Table 1Mixture composition (kg/m3) and

mechanical properties of tested fibrous concretes
N-1
H-1
190
160
Cement
255
500
Silica fume
40
Fibers
60
60
1170
880
725
860
Water
*
Coarse aggregate
Fine aggregate
Compressive strength at 28 days, MPa
30.7 (1.3)
115 (1.3)
Flexural strength at 28 days, MPa||
5.9 (0.9)
13.1 (0.7)
Splitting strength at 28 days, MPa
3.7 (0.3)
13.1 (2.5)
CEM I 52.5N, complies with EN 196-1.
Steel fibers, RC-65/35-BN.
Numbers in parentheses denote standard deviations.
Measured on 100 mm (3.93 in.) cubes that were wet-cured for 7 days.
||
Measured on 70 x 70 x 280 mm (2.76 x 2.76 x 11.02 in.) prisms.
Notes: 1 kg/m3 = 0.0624 lb/ft3; 1 MPa = 145 psi.
ends protruding from the concrete to enable it to be mounted

on the testing machine, as depicted in Fig. 1. The bar was
isolated from the concrete along 60 mm (2.4 in.) using bond
breakers at each end of the specimen to avoid stress concentration29 (so that the net length of the RC prism specimen
was 550 mm [21.6 in.]). Load was applied gradually to the
protruding ends of the steel, and development of cracks
along the concrete prism was recorded. Although the tests
were performed up to the ultimate state of the specimens,
only the service phase of the load, as mentioned earlier (refer
to Problem definition), is referenced.
The calculation algorithm described earlier was
performed using the following data: Ac = 5625 mm2
(8.7in.2); As=113mm2 (0.2 in.2); and = 38 mm (1.5 in.).
The modulus of elasticity of the steel bar Es = 206,800 MPa
[29,994 ksi] was measured directly.
High-strength concrete specimensThe modulus of elasticity Ec of the HSC specimens was taken as 47,000 MPa
(6817 ksi) based on the measured concrete compressive
strength (Table 1) and its corresponding modulus of elasticity, proposed in Table 5.1.7 of the Model Code.1 Tensile
strength was set to 7.1 MPa (1030 psi) based on the measured
flexural strengths (Table 1) and the ratio of mean flexural
strength to axial tensile strength of concrete also proposed
in the Model Code, which according to the data in Table 1 is
equal to 0.54. A residual strength of 5 MPa (725 psi) was used
based on the residual-maximum strength ratio reported by
Leutbecher and Fehling,27 and was corrected proportionally
to the lower fiber volume ratio of 0.75% (refer to Table1).
In addition, the value of A = 200 N/mm3 (737 kip/in.3) was
obtained from bond-slip measurements of direct and flexural
pullout tests.30 The value of was set to 0.7 and the criterion specified by Eq. (11) and (12) was used. This value of
was calibrated to best fit a single experimental result (tensile
force at first cracking) and the same value was then applied
to all other values of steel strain. These data yielded n = 4.4,

results of HSC specimens (NBRI tests). (Note: 1 mm =
0.039in.; 1 kN = 0.225 kip.)
= 0.02, = 0.02236 mm1 (0.56794 in.1), fct = 0.7 7.1 =
4.97 MPa (721 psi), and f = 0.7 5 = 3.5MPa (508psi).
Figure 11(a) presents a comparison between the results
measured for each of the two specimens and the calculated
prediction of the average crack spacing. Although the figure
reveals good agreement between the measured and calculated results, it also shows a considerable scatter in the
experimental results for the two specimens (refer to H-1-1
versus H-1-2 in Fig. 11(a)). The scatter in concrete properties also caused these results to diverge from the expected
symmetric cracking pattern. It is therefore reasonable to
consider average values of (actual) corresponding segments
with respect to the axis of symmetry of the entire bar
(Fig.12) for both specimens.
Figure 11(b) presents a comparison between these values
and the calculated values for the minimum and maximum
segment lengths, as well as average lengths of all segments.
It is evident that this way of considering the scatter in the
results yields better agreement between the experimental
and calculated data. The mean relative errors for minimum,
average, and maximum values of uncracked segment lengths
are 0.02, 0.06 and 0.07, respectively (with corresponding
standard deviations of 0.15, 0.10, and 0.06). These values,
as well as Fig. 11(b), indicate a very good agreement for all
the tested parameters.
99
Table 2Crack spacing at increasing tensile load

(NBRI tests)
Crack spacing, mm
Load,
kN
First
Second Third
segment segment segment Minimum Average Maximum
Calculated
Fig. 12Corresponding segments in Specimen H-1-1.

(Note: The two right-hand cracks developed simultaneously
and are considered as one.)
17.0
275
275
275.0
275
22.7
153
122
122
137.5
153
41.87
86
67
122
67
91.7
122
Measured*
17.6
275
275
275.0
275
20.7
158
117
117
137.5
158
43.6
87
71
117
71
91.7
117
3.4%
0.0%
0.0%
0.0%
0.0%
9.7%
3.2%
4.3%
4.3%
0.0%
3.2%
4.0%
1.1%
5.6%
4.3%
5.6%
0.0%
4.3%
Relative error (Exp Cal)/Exp (%)
Average values of corresponding left- and right-hand segments with respect to axis
of symmetry of entire bar (Fig. 12).
Notes: 1 kN = 0.225 kip; 1 mm = 0.0394 in.

results of NSC specimens (NBRI tests). (Note: 1 mm =
0.039in.; 1 kN = 0.225 kip.)
Normal-strength concrete specimensLike in the case
of the HSC specimens, the following parameters were
used for the NSC specimen calculations: Ec = 28,000 MPa
(4061ksi); tensile and residual strengths of 3.5 and 1.5 MPa
(508 and 218 psi), respectively; A = 80 N/mm3 (295 kip/in.3);
and = 0.7. These data yielded n = 7.4, = 0.02,
=0.0122mm1 (0.3099 in.1); fct = 0.7 3.5 = 2.45 MPa
(355 psi) and f = 0.7 1.5 = 1.05 MPa (152 psi).
Figure 13 presents a comparison between the measured
lengths of the uncracked segments (minimum, maximum,
and average) and the calculated predictions. Note that there
are some differences in the measured lengths of corresponding left- and right-hand segments. Therefore, each
experimental point in Fig. 13 represents the average of these
two values. The mean relative errors for minimum, average,
and maximum values of uncracked segment lengths are
0.03, 0.02, and 0.02, respectively (with corresponding standard deviations of 0.15, 0.10, and 0.07). These values, as
well as Fig. 13, indicate a very good agreement for all the
tested parameters. This agreement is demonstrated also in
Table 2, which describes the crack distribution at increasing
tensile loads, as well as the corresponding relative errors.
CONCLUSIONS
This paper proposes a simplified model for the behavior
of a reinforced fibrous concrete bar when subjected to
increasing axial tension load. It refers to the cracking phase
during which cracks are relatively narrowthat is, up to the
yield of reinforcing bar in any one of the cracks. The model
100
predicts the loads that cause new cracks to develop as well as

the distances between the cracks at various stages of loading.
Phenomena that occur at the ultimate state, which is related
to the cracking pattern, originate from the crack spacing
that commences at the service state. The proposed approach
was verified against experimental results from two different
sources. Based on this model, an algorithm is presented to
calculate the tensile forces that cause cracking as well as the
intervals between the cracks.
The current study shows that, while distances between
cracks in plain concrete are equal, this is not the case for
fibrous concrete. It is shown that crack patterns in conventionally reinforced concrete with and without fibers are qualitatively different, even when fiber distribution is uniform. This
phenomenon is a result of the stress-free edges of the concrete.
The models predictions of the minimum, average, and
maximum of crack spacing (lengths of uncracked segments)
were compared with available experimental results. It is
evident from these comparisons that there is fair to good
agreement, especially for the average values (mean relative error and standard deviation do not exceed 0.06 and
0.18,respectively).
A parametric study shows that the effect of the stress
distribution in the concrete (indicated by the parameter ) is
pronounced at small steel strains whereas it is negligible at
relatively larger strains. An opposite effect was found for the
bond-slip coefficient A, which has a pronounced effect only
at relatively larger strains.
AUTHOR BIOS
Yuri S. Karinski is a Senior Researcher at the National Building Research
Institute, Technion Israel Institute of Technology (the Technion), Haifa,
Israel. He received his BSc, MSc, and PhD at the Faculty of Mathematics
and Mechanics, Moscow State University, Moscow, Russia. His research
interests include mathematical simulations of the dynamic behavior of
elastic-plastic structures and of the dynamic response of buried structures.
ACI member Avraham N. Dancygier is an Associate Professor at the

Faculty of Civil and Environmental Engineering and a Researcher at the
National Building Research Institute at the Technion. His research interests
include static and dynamic behavior of reinforced concrete structures, highstrength/high-performance concrete structures, and the behavior of buried
structures under static and dynamic loads.
ACI member Amnon Katz is an Associate Professor at the Faculty of
Civil and Environmental Engineering at the Technion. He received his BSc,
MSc, and PhD at the Technion. He is a member of ACI Committees 440,
Fiber-Reinforced Polymer Reinforcement, and 555, Concrete with Recycled
Materials. His research interests include advanced concrete technology and
environmental aspects of concrete structures.
ACKNOWLEDGMENTS
This work was supported by a joint grant from the Centre for Absorption in Science of the Ministry of Immigrant Absorption and the Committee
for Planning and Budgeting of the Council for Higher Education under
the framework of the KAMEA Program and by the Israeli Ministry of
Construction and Housing. The research grants are greatly appreciated. The
authors would like to thank also S. Engel, V. Eisenberg, and G. Ashuah for
their useful advice and support.
REFERENCES
1. fib Bulletins 65, 66, Model Code 2010Final Draft, Volumes 1, 2,

Federation internationale du bton (fib), Lausanne, Switzerland, 2012, 720 pp.
3. Casanova, P.; Rossi, P.; and Schaller, I., Can Steel Fibers Replace
Transverse Reinforcements in Reinforced Concrete Beams? ACI Materials Journal, V. 94, No. 5, Sept.-Oct. 1997, pp. 341-354.
4. Foster, J. F., The Application of Steel-Fibres as Concrete Reinforcement in Australia: From Material to Structure, Materials and Structures,
V. 42, No. 9, 2009, pp. 1209-1220. doi: 10.1617/s11527-009-9542-7
5. Voo, Y. L.; Poon, W. K.; and Foster, S. J., Shear Strength of Steel
Fiber-Reinforced Ultrahigh-Performance Concrete Beams without Stirrups, Journal of Structural Engineering, ASCE, V. 136, No. 11, 2010,
pp.1393-1400. doi: 10.1061/(ASCE)ST.1943-541X.0000234
6. Fantilli, A. P.; Ferretti, D.; Iori, I.; and Vallini, P., Behaviour of R/C
Elements in Bending and Tension: The Problem of Minimum Reinforcement Ratio, Minimum Reinforcement in Concrete Members, A. Carpinteri,
ed., Elsevier, Oxford, UK, 1999, pp. 99-125.
7. Lee, S.-C.; Cho, J.-Y.; and Vecchio, F. J., Tension-Stiffening Model
for Steel Fiber-Reinforced Concrete Containing Conventional Reinforcement, ACI Structural Journal, V. 110, No. 4, July-Aug. 2013, pp. 639-648.
8. Vandewalle, L., Cracking Behaviour of Concrete Beams Reinforced
with a Combination of Ordinary Reinforcement and Steel Fibers, Materials
and Structures, V. 33, No. 3, 2000, pp. 164-170. doi: 10.1007/BF02479410
9. Wafa, F. F., and Ashour, S. A., Mechanical Properties of HighStrength Fiber Reinforced Concrete, ACI Materials Journal, V. 89, No. 5,
Sept.-Oct. 1992, pp. 449-455.
10. Balendran, R. V.; Zhou, F. P.; Nadeem, A.; and Leung, A. Y. T.,
Influence of Steel Fibres on Strength and Ductility of Normal and Lightweight High Strength Concrete, Building and Environment, V. 37, No. 12,
2002, pp. 1361-1367. doi: 10.1016/S0360-1323(01)00109-3
11. Balaguru, P. N., and Shah, S. P., Fiber Reinforced Cement Composites, McGraw-Hill, Inc., New York, 1992, 530 pp.
12. Redaelli, D., Testing of Reinforced High Performance Fibre

Concrete Members in Tension, Proceedings of the 6th International Ph.D.
Symposium in Civil Engineering, Zurich, Switzerland, 2006, pp. 122-123.
13. Deluce, J. R., and Vecchio, F. J., Cracking of SFRC Members
Containing Conventional Reinforcement, ACI Structural Journal, V. 110,
No. 3, May-June 2013, pp. 481-490.
14. Redaelli, D., and Muttoni, A., Tensile Behaviour of Reinforced
Ultra-High Performance Fiber Reinforced Concrete Element, Proceedings of fib Symposium: Concrete Structures: Stimulators of Development,
Dubrovnik, Croatia, 2007, pp. 267-274.
15. Dancygier, A. N., and Berkover, E., Behavior of Fibre-Reinforced
Concrete Beams with Different Reinforcement Ratios, Proceedings of fib
Symposium: Concrete Engineering for Excellence and Efficiency, Prague,
Czech Republic, 2011, pp. 1133-1136.
16. Dancygier, A. N., and Savir, Z., Flexural Behavior of HSFRC with
Low Reinforcement Ratios, Engineering Structures, V. 28, No. 11, 2006,
pp. 1503-1512. doi: 10.1016/j.engstruct.2006.02.005
17. Barragn, B. E.; Gettu, R.; Martn, M. A.; and Zerbino, R. L.,
Uniaxial Tension Test for Steel Fibre Reinforced ConcreteA Parametric
Study, Cement and Concrete Composites, V. 25, No. 7, 2003, pp. 767-777.
doi: 10.1016/S0958-9465(02)00096-3
18. Gettu, R.; Gardner, D. R.; Saldivar, H.; and Barragfin, B. E., Study
of the Distribution and Orientation of Fibers in SFRC Specimens, Materials and Structures, V. 38, Jan.-Feb. 2005, pp. 31-37.
19. Yankelevsky, D. Z.; Jabareen, M.; and Abutbul, A. D., One-Dimensional 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
20. Bentur, A., and Mindess, S., Fibre Reinforced Cementitious Composites, second edition, Taylor & Francis, Oxford, UK, 2007, 624 pp.
21. Chiaia, B.; Fantilli, A. P.; and Vallini, P., Evaluation of Crack Width
in FRC Structures and Application to Tunnel Linings, Materials and Structures, V. 42, No. 3, 2009, pp. 339-351. doi: 10.1617/s11527-008-9385-7
22. Yuguang, Y.; Walraven, J. C.; and den Uijl, J. A., Combined Effect
of Fibers and Steel Rebars in High Performance Concrete, Heron, V. 54,
No. 2/3, 2009, pp. 205-224.
23. fib Bulletin 10, Bond of Reinforcement in Concrete, Fdration
internationale du bton (fib), Lausanne, Switzerland, 2000, 434 pp.
24. ASTM C1609/C1609M-05, Standard Test Method for Flexural
Performance of Fiber-Reinforced Concrete (Using Beam With Third-Point
Loading), ASTM International, West Conshohocken, PA, 2005, 8 pp.
25. RILEM TC 162-TDF, RecommendationsBending Test, Materials and Structures, V. 33, Jan.-Feb. 2000, pp. 3-5.
26. Mindess, S.; Young, J. F.; and Darwin, D., Concrete, second edition,
Prentice Hall, Upper Saddle River, NJ, 2003, 644 pp.
27. Leutbecher, T., and Fehling, E., Tensile Behavior of Ultra-HighPerformance Concrete Reinforced with Reinforcing Bars and Fibers: Minimizing Fiber Content, ACI Structural Journal, V. 109, No. 2, Mar.-Apr.
2012, pp. 253-264.
28. Leutbecher, T., Rissbildung und Zugtragverhalten von mit Stabstahl und Fasern bewehrtem Ultrahochfesten Beton (UHPC), PhD thesis,
University of Kassel, Kassel, Germany, 2007, 264 pp.
29. Rilem/CEB/FIP, Bond Test for Reinforcing Steel: 2. Pullout Test,
Materials and Structures, V. 3, No. 15, 1970, pp. 175-178.
30. Dancygier, A. N.; Katz, A.; and Wexler, U., Bond between Deformed
Reinforcement and Normal and High Strength Concrete with and without
Fibers, Materials and Structures, V. 43, No. 6, 2010, pp. 839-856. doi:
10.1617/s11527-009-9551-6
101
APPENDIXFLOWCHART FOR CRACKING PROCESS
102
TECHNICAL PAPER
Title No. 112-S10
Glass Fiber-Reinforced Polymer-Reinforced Circular

Columns under Simulated Seismic Loads
by Arjang Tavassoli, James Liu, and Shamim Sheikh
This paper presents the experimental results of nine large-scale
circular concrete columns reinforced with longitudinal and transverse glass fiber-reinforced polymer (GFRP) bars. These specimens
were tested under lateral cyclic quasi-static loading while simultaneously subjected to constant axial load. Based on the measured
hysteretic loops of moment-versus-curvature and shear-versus-tip
deflection relationships, a series of parameters related to ductility,
energy dissipation capacity, and flexural strength are used to evaluate the seismic behavior of each column. The results showed
that concrete columns reinforced with GFRP bars and spirals
can behave in a manner that has stable post-peak response and
achieve high levels of deformability. The results indicate that, as
a relatively new material with excellent corrosion resistance and
high strength-weight ratio, GFRP bars can be successfully used as
internal reinforcement in ductile concrete columns.
Keywords: concrete column; ductility; experiment; GFRP reinforcement;
seismic resistance; strength.
INTRODUCTION
A considerable amount of work has been done on the
behavior of steel-reinforced concrete columns. Appropriately designed lateral confinement, such as the use of closely
spaced transverse steel reinforcement, externally bonded
steel jackets, or fiber-reinforced polymer (FRP) wrapping,
have proven to significantly improve the ductility, energy
dissipation capacity and flexural strength of steel-reinforced
concrete columns under seismic loading.
The corrosion of steelespecially the lateral reinforcement
in structureshas cost billions of dollars in infrastructure repair
in North America. It is estimated that $3.6 trillion are needed
by 2020 to alleviate potential problems in civil infrastructure.1
Approximately one in nine bridges in the United States are rated
as structurally deficient, requiring about $20.5 billion annually
to eliminate the bridge deficient backlog by 2028. As a relatively
new material with excellent corrosion resistance and a high
strength-weight ratio, internal glass fiber-reinforced polymer
(GFRP) reinforcement is considered a feasible and sustainable
alternative to steel reinforcement for future infrastructure. A
number of studies have been carried out on GFRP-reinforced
concrete members subjected to flexure and shear. However,
only a few experimental studies have been reported on GFRPreinforced concrete columns. Alsayed et al.2 reported that
replacing longitudinal steel reinforcement with GFRP bars
of the same volumetric ratio resulted in a 13% reduction in
the axial load capacity of columns. De Luca et al.3 reported
that at low longitudinal reinforcement ratios, the response
of GFRP-reinforced columns is very similar to that of steelreinforced columns and the contribution of GFRP bars can
be neglected in the axial load capacity of columns. They

also stated that the relatively lower compressive strength and
stiffness of GFRP bars will make FRP-reinforced concrete
columns susceptible to instability. Based on tests on eight
columns under concentric axial loads, of which five were
longitudinally and transversely reinforced with GFRP,
Tobbi et al.4 concluded that GFRP bars can be used as main
reinforcement in columns provided that closely spaced
transverse reinforcement is used. Based on the sectional
analysis of GFRP-reinforced columns with a longitudinal
reinforcement ratio no less than 3%, Choo et al.5 concluded
that no balance point exists in the axial load-moment interaction curves and that the flexural strength tends to increase
monotonically with the decrease of axial load. Sharbatdar
and Saatcioglu6 tested square columns reinforced with FRP
bars under axial and lateral load. They concluded that FRPreinforced columns under 30% of their axial capacity can
develop 2 to 3% lateral drift ratios. Aside from this study, experimental work on GFRP-reinforced circular columns subjected
to combined axial, shear and flexural loads, especially under
cyclic loads simulating seismic forces, is almostnonexistent.
The experimental study reported here is part of an
ongoing comprehensive research program at the University
of Toronto, in which full-scale square and circular concrete
columns are tested under simulated seismic load in the same
manner to provide comparable results to investigate different
variables and design parameters (refer to References 7 and
8). The variables include column types, steel configuration, concrete strength, axial load level, and amounts of
steel and FRP confinement. This paper describes the results
from nine large-scale circular concrete columns internally
reinforced with longitudinal GFRP bars and transverse GFRP
spirals. A series of parameters related to curvature ductility,
displacement ductility, energy dissipation capacity, and
flexural strength are used to evaluate the seismic behavior
ofcolumns.
Available research on the behavior of columns internally
reinforced with longitudinal and lateral GFRP reinforcement is very limited. Information on their response under
large lateral cyclic displacements is almost nonexistent.
MS No. S-2013-309.R1, doi: 10.14359/51687227, received April 25, 2014, and
103
Table 1Test results of GFRP-reinforced circular columns

Specimen
Spiral diameter, mm (in.) Spiral pitch, mm (in.) Axial load level P/Po Mmax, kN.m (kip.ft) Vmax, kN (kip) Plastic hinge length, mm (in.)
P28-C-12-50
12 (1/2)
50 (2.0)
0.28
224 (165)
70.0 (15.7)
260 (10.2)
P28-C-12-160
12 (1/2)
160 (6.3)
0.28
152 (112)
71.1(16.0)
320 (12.6)
P28-C-16-160
16 (5/8)
160 (6.3)
0.28
123 (91)
59.4 (13.4)
260 (10.2)
P28-B-12-50
12 (1/2)
50 (2.0)
0.28
227 (167)
78.2 (17.6)
270 (10.6)
P42-C-12-50
12 (1/2)
50 (2.0)
0.42
219 (161)
74.9 (16.8)
250 (9.8)
P42-C-12-160
12 (1/2)
160 (6.3)
0.42
162 (119)
58.8 (13.2)
300 (11.8)
P42-B-12-160
12 (1/2)
160 (6.3)
0.42
160 (118)
63.1 (14.2)
270 (10.6)
P42-B-16-160
16 (5/8)
160 (6.3)
0.42
187 (137)
70.0 (15.7)
240 (9.4)
P42-B-16-275
16 (5/8)
275 (10.8)
0.42
154 (113)
69.8 (15.7)
370 (14.6)
Fig. 1Details of test specimen. (Note: 1 mm = 0.039 in.)

This research provides results from large-scale GFRPreinforced concrete columns tested under simulated earthquake loading. The measured ductility and deformability,
energy dissipation capacity, and strength of the columns
provide valuable parameters in understanding the seismic
behavior of these columns. The comparison between similarly tested GFRP- and steel-reinforced columns can also help
clarify and further develop some of the existing design provisions that are overly conservative and in some cases vague.
Specimens
Nine circular columns were constructed and tested in this
study. Each specimen consisted of a 356 mm (14 in.) diameter and 1473 mm (58 in.) long column cast monolithically
with a 508 x 762 x 813 mm (20 x 30 x 32 in.) stub (Fig.1).
The column part of the specimen represents the column
between the sections of maximum moment and zero moment
in a structure while the stub represents a footing or a joint.
Both the longitudinal and transverse reinforcement in each
column were made out of GFRP. The main variables investigated were the axial load level, the type of GFRP, and size
and spacing of GFRP spirals.
104
The specimens were tested under constant axial load

and lateral cyclic displacement excursions. The axial
load was either 0.28Po or 0.42Po, where Po is the nominal
axial load capacity of the column, determined by Po =
0.85fc(Ag AGFRP) + cEGFRPAGFRP; Ag is the gross cross-sectional area of column; fc is the compressive strength of
unconfined concrete; c is the concrete strain at peak
strength; and AGFRP and EGFRP are the total cross-sectional
area and the modulus of elasticity of the longitudinal GFRP
reinforcement in compression, respectively. The specimen
details are provided in Table 1. Each specimens label has
four parts indicating the level of axial load (P28 or P42),
the type of GFRP material (B or C), the spiral diameter
(12or 16 mm [0.47 or 0.63 in.]) and the spiral pitch in mm
in the 740 mm (29 in.) long test region of columns adjacent
to the column-stub interface. The length of the test region
was decided based on the previous tests in which failure
was observed in that region. To ensure that failure occurs in
the test region of columns, a closely spaced steel spiral was
used outside the test region which was additionally confined
by externally bonded FRP wrapping. Figure 2 shows the
column cage and the FRP-wrapped column before the test.
Fig. 3GFRP material used in this study: 12 mm (0.47 in.)

Type B-SP spiral (left); 25 mm (1 in.) Type B-ST longitudinal
bar (top right); and 25 mm (1 in.) Type C-ST longitudinal
bar (bottom right).
Fig. 2Column cage reinforcement (left) and column before
test (right).
Concrete
The specimens were cast vertically by using two trucks of
ready mixed concrete with the maximum aggregate size of
14 mm (0.55 in.) and a slump of around 150 mm (6 in.). One
truck was used for the nine stubs and another for the nine
columns. The development of concrete strength with age was
monitored in accordance with ASTM C39/C39M9 by tests of
150 x 300 mm (6 x 12 in.) concrete cylinders that were cured
adjacent to the column specimens under similar conditions.
There were at least three cylinders in each group which were
tested at 7, 14, and 28 days after casting and throughout the
column testing period. Testing of the nine columns started
on the 70th day and was completed on the 125th day after
casting, during which the concrete strength was measured
between 34.1 and 35.4 MPa (4.95 and 5.13 ksi). Hence, fc
was taken as 35 MPa (5.08 ksi) for all columns.
Glass fiber-reinforced polymers (GFRP)
Each column contained six longitudinal GFRP bars of
25mm (1 in.) diameter, which were embedded completely
to near the end of stubs (Fig. 1 and 2). The transverse
reinforcement in the test region of the column consisted
of 12 or 16 mm (0.47 or 0.63 in.) diameter GFRP spirals
at specified spacing. The clear concrete cover was 24 mm
(0.95in.) to the outmost surface of spirals and the area ratio
of concrete core, measured to the centerline of spirals, to the
gross cross-section of the column was approximately 72%.
Two types of GFRP reinforcement from two different
manufacturers were used in this program, as shown in Fig.3.
Five columns were reinforced with GFRP longitudinal bars
and spirals from manufacturer C, while the other four
columns contained GFRP reinforcement from the manufacturer indicated as B. The mechanical properties of the
GFRP bars were obtained by testing three coupons for each
bar type. Tensile tests were carried on both longitudinal and
transverse GFRP bars, while compressive tests were done on
longitudinal GFRP bars only.
For the tensile tests of GFRP bars, steel couplers were
used on each end of GFRP samples and the gap between
the GFRP bar and the coupler was filled with an expansive cement mortar and capped. Figure 4 shows the GFRP
samples during the tension test. The strain was measured
using a detachable gauge attached to the bar at midheight
with a gauge length of 25 mm (1 in.). To avoid damage to the
gauge, it was removed at about a third of the ultimate load.
Elastic behavior was assumed to extrapolate the strain data
until the ultimate stress. Table 2 summarizes the results for
all the GFRP coupon tension tests.
The behavior of GFRP bars in compression is still not as
well-studied as the tensile response. For GFRP bars with
L/d < 6, where L is the free length and d is the diameter
of bars, it was reported that their compressive strength can
be 10to 50% of their ultimate tensile strength, depending
on fiber content, the manufacturing procedure, and the
resin quality.10 Based on tests on FRP bars with L/d < 6,
it was reported that the ultimate compressive strength
is approximately half of the tensile strength, while the
modulus of elasticity was found to be approximately equal
in both compression and tension.11 In this study, 15 GFRP
bar samples were tested in compression. The test setup is
shown in Fig. 5, in which the axial compression was directly
applied to the samples through steel cylindrical caps at both
ends. A different free length of GFRP bars between the two
steel caps was chosen based on the spiral pitch in different
columns. Two strain gauges were installed at the midheight
of each GFRP coupon on opposite sides to measure the
longitudinal compressive strain. The modulus of elasticity
was obtained by averaging the two strains during the initial
stages of the test before the two strain values deviated from
each other due to buckling effect. Table 3 summarizes the
results obtained from these tests.
Steel
Two types of deformed steel bars were used in specimens. In each column outside the 740 mm (29 in.) long test
region, the transverse reinforcement consisted of Grade 60
U.S. No.3 steel spirals at 50 mm (2 in.) pitch (volumetric
ratio to the concrete core sh = 1.93%). The stubs were reinforced with steel cages made of 10M (area of cross section
= 100mm2 [0.155 in.2]) stirrups spaced at 64 mm (2.5 in.)
in both horizontal and vertical directions. The mechanical
105
Table 2Mechanical properties of GFRP bars in tension

Reinforcing
bar type*
Nominal diameter, mm (in.)
Actual diameter,
mm (in.)
Modulus of elasticity,
MPa (ksi)
Ultimate stress,
MPa (ksi)
Ultimate strain
C - SP
12 (0.472)
12.62 (0.497)
C - SP
16 (0.630)
16.05 (0.632)
58,399 (8470)
1454 (211)
0.0249
51,224 (7429)
1069 (155)
0.0209
0.0165
C - ST
25 (0.984)
25.11 (0.989)
65,779 (9540)
1087 (158)
B - SP
12.7 (0.5)
12.7 (0.5)
58,948 (8550)
1243 (180)
0.0211
B - SP
15.87 (0.625)
16.01 (0.630)
54,567 (7914)
1159 (168)
0.0213
B - ST
25.4 (1)
28.44 (1.120)
74,270 (10772)
1338 (194)
0.0180
The first part indicates type of GFRP material and second part identifies whether straight sample was taken from material used for straight bars (ST) or spirals (SP).
These values do not represent ultimate stress and strain. The test was terminated due to slippage of bars in coupler.
Table 3Mechanical properties of GFRP bars in compression

Reinforcing
bar type
Free length,
mm (in.)
Nominal area,
mm2 (in.2)
Actual area,
mm2 (in.2)
491 (0.761)
495 (0.768)
50 (2)
C
Modulus of elasticity,
MPa (ksi)
Ultimate stress,
MPa (ksi)
Ultimate strain
55,569 (8060)
619 (90)
0.01132
160 (6.3)
56,357 (8174)
602 (87)
0.01066
50 (2)
71,018 (10300)
864 (125)
0.0104
72,165 (10467)
873 (127)
0.00932
72,701 (10544)
759 (110)
0.00901
160 (6.3)
507 (0.786)
275 (10.8)
635 (0.985)
Fig. 4GFRP coupon test (tension), left to right: 12 mm (0.47 in.) Type B-SP before and failure; and 12 mm (0.47 in.) Type
C-SP before and failure, left to right: Bar Type B with free length of 275 mm (10.8 in.) before and after failure; and Bar Type
C with free length of 160 mm (6.3 in.) before and after failure.
Fig. 5GFRP coupon test (compression).

106
Table 4Mechanical properties of reinforcing steels

Reinforcing bar
size
Area,
mm2 (in.2)
Yield stress fy,

MPa (ksi)
Yield strain
y
Modulus of elasticity
E, MPa (ksi)
Start of strainhardening sh
Ultimate stress fu,

MPa (ksi)
Ultimate
strain u
10M
100 (0.16)
420 (60.9)
0.0023
187,105 (27,137)
0.0251
542 (78.6)
0.1960
US No. 3
71 (0.11)
485 (70.3)
0.0025
191,570 (27,785)
0.0273
598 (86.7)
0.1632
Fig. 6Location of strain gauges on GFRP bars and spiral. (Note: 1 mm = 0.039 in.)
properties obtained from the tests on a minimum of three
samples of each type of steel bar are presented in Table 4.
Instrumentation
To monitor the deformation of GFRP reinforcement
in each specimen during testing, 18 strain gauges were
installed on the longitudinal bars and six on the spirals
three on each of the two turns adjacent to the stub faceas
shown schematically in Fig. 6. Ten linear variable differential transformers (LVDTs) were installed on one side of the
column and light-emitting diode (LED) targets were used
on the other side to measure deformation of the concrete
core in the potential plastic-hinge region. The LVDTs were
mounted on the threaded rods installed inside the columns
before concrete casting to measure the inelastic deformation
of core concrete. Three LEDs were mounted on a stationary
location and were used as reference, while 14 targets were
placed on each specimen. In addition to the linear strains, the
LED targets provided three-dimensional movements at each
location. The lateral deflection along each specimen was
measured by six LVDTs. The instrumentation is displayed
in Fig. 7.
Testing procedure
Each column was tested under a constant axial load and
quasi-static lateral cyclic displacement excursions. The
axial load was applied by a hydraulic jack with a capacity
of 10,000 kN (2250 kip), while the cyclic lateral loading
was applied using an actuator with a 1000 kN (225 kip) load
capacity and approximately 100 mm (4 in.) stroke capacity.
The test frame is shown in Fig. 8. Two hinges permitted each

end of the specimen to rotate freely in plane and kept the
axial load path constant throughout testing. Even though the
cantilever length of concrete columns was 1473 mm (58in.),
this test setup resulted in an actual shear span of 1841 mm
(72.5 in.) measured from the center of the right hinge to
the column-stub interfacethat is, from the zero moment
section to the maximum moment section of columns.
Each specimen was strictly aligned before testing, so that
its center line coincided with the action line of axial load. At
the beginning of each test, the predetermined axial load was
firstly applied to the specimen and kept constant throughout
testing. The lateral cyclic excursions were then applied in a
displacement-control mode following the specified deflection regime shown in Fig. 9 until the column collapsed under
the constant axial load.
TEST OBSERVATIONS
All columns behaved almost elastically during the first
two lateral load cycles. Flexural cracks appeared on top
and bottom faces of the column in the testing region during
the third cycle. For columns under an axial load of 0.42Po,
surface cracks in the longitudinal direction appeared prior
to flexural cracks close to the column-stub interface. Cover
spalling for the majority of specimens initiated during the
fourth lateral cycle. For well-confined columns, cover
spalling was delayed by a few cycles, while for columns with
a higher axial load, complete cover deterioration occurred
before the eighth cycle. In general, columns showed a very
stable response with large deformability (Table 5). For
107
Fig. 7Instrumentation: (a) strain gauges on bars and spirals in the test region (left), horizontal LVDTs on one side of the
column (top right), and LED targets on the other side of the column (bottom right); and (b) location of LVDTs. (Note: 1 mm =
0.039 in.)
Fig. 9Lateral loading protocol. (Note: 1 mm = 0.039 in.)

Fig. 8Test setup.
instance, specimen P28-B-12-50 underwent 34 cycles of
lateral displacement excursions and still maintained applied
axial load. Two additional cycles at the maximum stroke
limit of the actuator caused crushing of the longitudinal
bars in compression and a drop in the axial load. Specimen
108
P28-C-16-160 showed lower ductility and strength than

expected due to the honeycombing in the test region of the
column, which led to premature failure.
The failure of all columns was gradual and mainly due to
the crushing and buckling of the longitudinal GFRP bars in
compression accompanied by damage of the core concrete.
Even after the collapse of the columns, no spiral damage or
rupture was observed in any of the tests. It was observed that
where the inclination of the stub-column interface is
Table 5Ductility parameters

Specimen
,
%
N80
W80
Vmax,
kN (kip)
Mmax,
kN.m
(kip.ft)
P28-C>10.0 6.8 7.3

12-50
84
210 159 456
70.0
(15.7)
224
(165)
P28-C12-160
>9.0
3.2 3.0
17
22
99
160
71.1
(16.0)
152
(112)
P28-C16-160
>5.3
2.8 2.6
11
10
86
134
59.4
(13.4)
123 (91)
260 178 671
78.2
(17.6)
227
(167)
219
(161)
P28-B>15.6 9.2 9.0

12-50
94
P42-C>11.8 5.8 4.4

12-50
30
65
142 395
74.9
(16.8)
P42-C12-160
>7.5
3.7 3.0
16
28
50
93
58.8
(13.2)
162
(119)
P42-B12-160
>9.5
2.3 1.9
70
96
63.1
(14.2)
160
(118)
P42-B>10.7 4.0 3.3

16-160
18
34
119 239
70.0
(15.7)
187
(137)
P42-B16-275
56
69.8
(15.7)
154
(113)
>8.8
2.5 2.1
60
the failure of each column occurred within the test region.

Due to the additional confinement provided by the concrete
stubs, the most damaged section of columns was pushed
away from the stub-column interface where the moment
is the largest. Photos of the failed columns are shown in
Fig.10, while the measured lengths of the most damaged
regions are presented in Table 1.
Test results
The columns were tested horizontally; therefore, certain
geometric adjustments are needed to convert the measured
test data according to the cantilever column model shown
in the schematic drawing of the specimen in Fig. 11. The tip
displacement is the tangential deviation of the contraflexural Point B, calculated as

D = dL
a+b
(1)
a
V = PL
a
(2)
a+b
As discussed by Liu and Sheikh,8 the shear force V at the

base of the cantilever column differs from the applied lateral
force V at Point B. As a result, the base shear force V can
be determined from the components of the axial load P and
lateral force V, as shown in Eq. (3)

V = Vcos Psin
D
(4)
a+b
Moment at the most damaged section is found by summing

the moments caused by both the lateral and axial load. The
following expression is used to calculate the moment in the
most damaged section

M = V (L Dmd) + P
(5)
where L = H + c = 1841 mm (72.5 in.) is the shear span of

the column; and Dmd is the distance from the column-stub
interface to the most damaged section.
As mentioned earlier, 10M bars were precast in the testing
region of the columns to measure core deformations after
cover spalling and to obtain curvature. However, as a result
of high deformation of the column and considerable concrete
crushing in the damaged region, the measured curvature
values from these rods were not found to be accurate at large
displacements. Thus the curvature results reported here were
obtained using strains from the strain gauges installed on the
longitudinal bars. In cases where the strain gauges on the
outer bars in the most damaged region stopped functioning
before the failure of columns, curvature was obtained from
the strain gauges located on the inner bars to extend the
moment curvature (M-) behavior. This extension is shown
in dotted line in Fig. 12, while the horizontal dashed lines
in each graph in Fig. 12 represent the nominal moment
capacity of the column section (Mn). To evaluate the value
of Mn by sectional analysis, the stress-strain relationship of
unconfined concrete was used with the ultimate strain of
0.0035. Meanwhile, linearly elastic stress-strain relationship was assumed for longitudinal GFRP reinforcement
with ultimate tensile and compressive strengths provided in
Tables 2 and 3, respectively. In addition to the M- response
of the most damaged section, the shear-versus-tip deflection
relationship (V-) is also provided for each column. The
dashed lines on the V- curves represent the nominal shear
capacity Vn with a decreasing slope caused by secondary
effects. Vn is calculated using the following expression

where L is the deflection at the point of application of lateral

load PL.
The lateral force V at the right hinge B can be determined
from the applied force PL as
q=
Vn =
M n PD y
L
(6)
where P is the applied axial load; y is the yield displacement; and L is the shear span. Although GFRP-reinforced
columns do not undergo yielding, the procedure used by
Liu and Sheikh8 for steel-reinforced columns was also used
in this study to define a hypothetical yield displacement y
to evaluate the displacement ductility factor. Displacement
y is determined corresponding to the nominal lateral load
capacity Vn along a straight line joining the origin and a
point of 65% Vn on the ascending branch of the lateral shearversus-tip deflection curve, which is determined for a
specific axial load applied on the column.
(3)
109
Fig. 10Plastic hinge regions at end of testing.
Fig. 11Schematic drawing of test specimen.
110
Ductility parameters
Several ductility parameters have been used to explain the
behavior of reinforced concrete sections in the literature.7
Curvature and displacement ductility factors (, ), drift
ratio (), cumulative displacement ductility ratio (N80), and
work damage indicator (W80) are used in this study to quantify the deformability of the specimens.6 Figure 13 provides
a graphical representation of how the displacement-related
ductility parameters are calculated while parameters based
on curvature can be defined in a similar manner. The results
on ductility parameters are summarized in Table 5. The
curvature at the most damaged section is obtained from the
strain gauges on the longitudinal bars. Because the strain
gauges stopped functioning before the failure of columns,
the ultimate curvature ductility factors will be higher in most
cases than the values reported here.
Fig. 12Moment-curvature and shear-deflection responses of columns. (Note: 1 mm = 0.039 in.; 1 kN = 0.225 kip; 1 kNm =
0.738 kip.ft; 1 rad/km = 305106 rad/ft.)
111
Fig. 13Definitions of ductility parameters.

The values of N80 and W80 are respectively the cumulative displacement ductility ratio and the work damage indicator over a number of cycles until the shear capacity of the
column drops to 80% of the peak load. As shown in Fig. 12,
the column loses its stiffness (shown by Ki in Fig. 13) as
the test progresses and becomes negative for most columns
during the last few cycles of testing. Due to the presence
of the axial load on a highly deformed column during the
last cycles, the shear force has to switch direction to maintain equilibrium. To avoid adding negative energies to the
work damage indicator, only cycles with a positive Ki are
included in the calculation of N and W. Although this does
not capture the full response of the column, the results show
a similar trend to the reliable values of N80 and W80.
Effect of axial load level
Two levels of column axial load (0.27Po and 0.42Po) were
used in this study. In steel-reinforced columns, it has been
found that increasing the axial load would result in faster
deterioration of the core concrete and initiation of longitudinal bar buckling.7 Similar behavior was observed for the
GFRP-reinforced concrete columns tested here. The plastic
hinge length (most damaged zone) shown in Table 1 is greater
for columns with higher axial load indicating wider spread
of the damaged zone. The effect of axial load on column
ductility is more visible on well-confined specimens. For
example, Columns P28-C-12-50 sustained an axial load of
0.28Po and achieved a displacement ductility factor of
6.8 while the similar Column P40-C-12-50 carrying an axial
load of 0.42Po showed of 5.8. The same observation can
also be found by comparing the test results of other columns
shown in Table 5 and Fig. 12.
An increase of axial load from 0.28Po to 0.42Po did not
significantly affect the flexural strength of GFRP-reinforced
columns. For example, the measured flexural strengths of
columns P28-C-12-50 and P42-C-12-50 were almost identical. This experimental observation confirms the similarity
between nominal moment capacities of specimens under
112
Fig. 14Hysteresis response of steel-confined versus

GFRP-reinforced column. (Note: 1 mm = 0.039 in.;
1 kN = 0.225 kip.)
different axial loads evaluated using sectional analysis.
Similar results were also reported by Choo et al.5
Effect of GFRP bar type
In general, there was not a significant difference between
the performances of the columns reinforced with two
different types of GFRP bars. The maximum moment
capacity at the most damaged section, as shown in Table 1,
is slightly higher for columns reinforced with GFRP TypeB.
This can be attributed to the fact that both longitudinal
bars and spirals of Type B have considerably larger actual
areas compared with the nominal areas than those of GFRP
TypeC. The length of the damaged region is very similar for
columns in pairs such as P28-C-12-50 and P28-B-12-50 or
P42-C-12-160 and P42-B-12-160, in which the only difference
was the type of GFRP bar. The latter pair of columns showed
a similar deflected shape at all stages and failure for both
columns occurred at almost the same lateral displacement.
Effect of spiral pitch and reinforcement ratio
Research on steel-reinforced columns has shown that
increasing the transverse reinforcement ratio and decreasing
the spiral/tie pitch delay column failure by confining the

concrete core prevents premature buckling of the longitudinal bars.7,8 Consequently, columns with lower spiral
spacing have higher strength and ductility.
Similar conclusions can be made for GFRP-reinforced
columns using the results obtained in this experimental study.
The moment capacity of the column considerably increased
as the spiral spacing reduced from 160 to 50 mm (6.2 to
2in.). The maximum shear capacity is approximately similar
for all columns. This is expected because the maximum shear
occurs in the first few cycles and is not affected by the spiral
configuration. Table 5 shows that the absorbed energy is
significantly higher for columns with smaller spiral spacing
and the resulting higher transverse reinforcement. Closer
spiral spacing resulted in better confinement of the concrete
core and delayed the buckling of the longitudinal bars. For
instance, the lateral drift ratio for column P28-C-12-50 is
more than twice that of column P28-C-12-160. Reducing the
spiral spacing from 160 to 50 mm (6.2 to 2 in.) resulted in a
50% increase in the moment capacity of the column.
The transverse reinforcement ratio can be adjusted by
changing the size of the spiral as well. However, the effect
on column response is not as noticeable as adjusting the
spiral spacing. The effect of changing the spiral size can
be observed by comparing columns P42-B-12-160 and
P42-B-16-160. Increasing the spiral size from 12 to 16 mm
(0.47 to 0.63 in.) resulted in a 13% increase in the moment
capacity of the column and doubled the dissipated energy.
Comparison of GFRP-reinforced and steelreinforced columns
GFRP-reinforced concrete columns generally show a
softer ascending branch of shear-versus-deflection behavior
than steel-reinforced concrete columns due to the lower
modulus of elasticity of GFRP. The difference in stiffness
is significant even though reinforcement represents only a
small portion of the columns properties. A comparison of
the shear-versus-lateral tip deflection relationships of two
columns is shown in Fig. 14. Column P27-NF-2 had 9.5 mm
(No. 3) steel spiral with 100 mm (4 in.) pitch which resulted
in sh = 0.9%. The longitudinal steel reinforcement ratio
for this column was 3.01%. Both these values were similar
to those of column P28-C-12-160, as were the column
dimensions and testing conditions. The concrete strength
for column P-27-NF-2 was 40 MPa while this value for
column P28-C-12-160 was 35 MPa (5080 psi). The hysteretic response shows that the steel-reinforced column absorbs
more energy in each lateral deflection loop as a result of the
yielding and Bauschinger effect of the longitudinal steel
reinforcement. It should be noted that a lack of yield plateau
in GFRP bars results in a much lower residual deflection when the load is removed. The stiffness of the steelreinforced column is higher than that of the GFRP-reinforced
column and the higher shear capacity of column P27-NF-2 is
due to the higher modulus of elasticity of the steel reinforcement. The capacity gap could simply be decreased by using
a larger amount of the GFRP reinforcement to balance its
low stiffness which may reduce deformability of the column.
Table 6Average spiral strain in most

damagedregion
Maximum
spiral strain
at 80% peak
fh80%,
Maximum
spiral strain
fh,
Ultimate
spiral strain
ufh,
fh/ufh
P28-C12-50
4145
4145
24,900
0.167
P28-C12-160
3612
4031
24,900
0.162
P28-C16-160
4450
5605
19,900
0.282
P28-B12-50
9306
9779
21,100
0.463
P42-C12-50
2276
6440
24,900
0.259
P42-C12-160
7835
11,882
24,900
0.477
P42-B12-160
5010
7581
21,100
0.359
P42-B16-160
6597
13,054
21,300
0.613
P42-B16-275
5902
9723
21,300
0.456
Average
B
6704
10,034
21,200
0.473
Average
C
4464
6421
23,900
0.269
Specimen
Both columns showed stable post-peak descending

branches and high ductility. However, it can be seen that the
GFRP-reinforced column has a longer post-peak descending
branch before final failure than the steel-reinforced column.
This is mainly due to the fact that the steel bars have very
low tangent modulus after yielding and therefore are more
susceptible to buckling under compression than GFRP bars
which maintain their modulus of elasticity throughout the
entire duration of loading.
GFRP reinforcement does not experience yielding and
results in a nearly linear elastic moment-versus-curvature
relationship of columns with no post-peak decline, as shown
in Fig. 12, unlike the steel-reinforced columns in which the
steel yields and the moment curvature response displays a
descending branch. The GFRP transverse reinforcement
despite being softer than steel at small strains, continue to
provide increasing confinement until the column failure.
Strain effectiveness of GFRP spirals
The maximum strain in the GFRP spiral, sp,80%, was
recorded for each column at the stage when shear capacity
dropped to 80% of the peak shear and also just before
failure. The results are summarized in Table 6. As expected,
in columns with high axial load, a greater hoop strain is
observed. The average value of sp,80% for all nine columns
is 0.00546, which is close to the recommended Canadian
code12 design value of 0.006. Maximum recorded spiral
strains before column failure indicate that GFRP spirals were
able to provide increasing confinement to the core while the
113
maximum strain is significantly less than the rupture strain.

The average maximum spiral strain for the nine columns
is about four times the steel yield strain before failure.
The ratio between the maximum spiral strain and ultimate
GFRP strainspiral efficiencyis also listed for all the
columns in Table 6. The increase in the moment capacity
of the columns, especially those reinforced with a 16 mm
(0.63in.) spiral, over the last few cycles indicates the effective confinement provided by the GFRP spiral. This strength
gain over the last cycles can clearly be seen in the momentversus-curvature response of columns P28-C-16-160,
P42-B-16-160, and even P42-B-16-275.
Comparison with code requirements for confinement
The transverse GFRP spirals in specimens P28-C-12-50
and P28-B-12-50 were designed for a targeted lateral drift
ratio of 4% according to CAN/CSA-S806-12,12 while specimen P42-C-12-50 was designed to achieve 2.5% drift
ratio, and the rest of specimens had only substandard transverse reinforcement in terms of confinement spiral spacing
required by this design code.
Both columns P28-C-12-50 and P28-B-12-50 outperformed the design expectations, with measured drift ratios
of 7.3 and 9.0%, respectively. Column P42-C-12-50 also
achieved a lateral drift ratio of 4.4%, which exceeded the
designed target of 2.5%. In fact, only columns P42-B-16-275
and P42-B-12-160 had a lateral drift ratio lower than 2.5%.
The widely spaced spirals in P42-B-16-275 led to the premature buckling of longitudinal reinforcement and limited
deformability. Specimen P42-B-12-160 showed a lateral drift
ratio of only 1.9% due to the presence of honeycomb regions
in the testing region. The experimental results indicated that,
if designed according to CSA-S806-12,12 GFRP-reinforced
columns can achieve much higher lateral drift capacity than
design expectations.
To understand the behavior of concrete columns reinforced
longitudinally with GFRP bars and transversely with GFRP
spirals, nine large-scale specimens were tested under lateral
displacement excursions and constant axial load. Experimental results in the form of moment-versus-curvature
and shear-versus-tip deflection hysteretic responses and
various ductility parameters are presented. The following
conclusions can be drawn from this study:
1. The crushing strength of GFRP bars in compression is
approximately half of their ultimate tensile strength, and the
modulus of elasticity in compression for a GFRP bar was
found to be similar to that under tension.
2. Columns reinforced with GFRP bars and spirals showed
a stable response, and the type of GFRP material used did not
cause a significant change in the behavior of the columns.
3. Columns P28-B-12-50 and P28-C-12-50 were designed for
a lateral drift ratio of 4% and they achieved a lateral drift ratio in
excess of 7%. Column P42-C-12-50 achieved a lateral drift ratio
of 4.4%, which was 1.9% higher than the design value.
4. Columns that were subjected to a higher axial load
sustained more damage and displayed lower levels of
ductility and deformability. The amount and detailing of
114
transverse reinforcement is more critical for high axially

loaded columns.
5. GFRP bars, due to their larger stiffness at larger strains,
performed in a more stable manner than steel bars.
6. The transverse steel reinforcement provides effective
confinement to the core at early stages; however, as the steel
starts to yield, the confinement is less effective, allowing the
concrete core to expand. GFRP spirals, on the other hand,
provide an increasing level of confinement with increased
deformation which delays crushing of the core concrete.
AUTHOR BIOS
Arjang Tavassoli is a Project Associate at Parsons Brinckerhoff Halsall

Inc. in Toronto, ON, Canada. He received his BASc and MASc in civil engineering from the University of Toronto, Toronto, ON, Canada, in 2011 and
2013, respectively.
James Liu is a Structural Engineer with Brown and Company Engineering Ltd. in Toronto, ON, Canada. He received his BS in civil engineering from Tongji University, Shanghai, China, and his PhD from the
University of Toronto. His research interests include analysis and retrofit
of concretestructures.
Shamim Sheikh, FACI, is a Professor of civil engineering at the University of Toronto. He is Chair of ACI Subcommittee 441-E, Columns with
Multi-Spiral Reinforcement. He is a former Chair and member of
JointACI-ASCE Committee 441, Reinforced Concrete Columns, and a
member of ACICommittee 374, Performance-Based Seismic Design of
Concrete Buildings. In 1999, he received the ACI Chester Paul Seiss Award
for Excellence in Structural Research. His research interests include earthquake resistance and design of concrete structures, confinement of concrete,
use of fiber-reinforced polymer for sustainable concrete structures.
REFERENCES
1. ASCE, 2013 Report Card for Americas Infrastructure, American Society of Civil Engineers, www.infrastructurereportcard.org. (last
accessed Oct. 15, 2014)
2. Alsayed, S. H.; Al-Salloum, Y. A.; Almusallam, T. H.; and Amjad,
M. A., Concrete Columns Reinforced by Glass Fiber Reinforced Polymer
Rods, 4th International SymposiumFiber Reinforced Polymer Reinforcement for Reinforced Concrete Structures, SP-188, C. W. Dolan,
S. H. Rizkalla, and A. Nanni, eds., American Concrete Institute, Farmington
Hills, MI, 1999, pp. 103-112.
3. De Luca, A.; Matta, F.; and Nanni, A., Behavior of Full-Scale Glass
Fiber-Reinforced PolymerReinforced Concrete Columns under Axial
Load, ACI Structural Journal, V. 107, No. 4, July-Aug. 2010, pp. 589-596.
4. Tobbi, H.; Farghaly, A. S.; and Benmokrane, B., Concrete Columns
Reinforced Longitudinally and Transversally with Glass Fiber-Reinforced
Polymer Bars, ACI Structural Journal, V. 109, No. 4, July-Aug. 2012,
pp. 551-558.
5. Choo, C. C.; Harik, I. E.; and Gesund, H., Strength of Rectangular
Concrete Columns Reinforced with Fiber-Reinforced Polymer Bars, ACI
Structural Journal, V. 103, No. 3, May-June 2006, pp. 452-459.
6. Sharbatdar, M. K., and Saatcioglu, M., Seismic Design of FRP Reinforced Concrete Structures, Asian Journal of Applied Sciences, V. 2, No. 3,
2009, pp. 211-222. doi: 10.3923/ajaps.2009.211.222
7. Sheikh, S. A., and Khoury, S. S., Confined Concrete Columns with
Stubs, ACI Structural Journal, V. 90, No. 4, July-Aug. 1993, pp. 414-431.
8. Liu, J., and Sheikh, S. A., Fiber-Reinforced Polymer-Confined
Circular Columns under Simulated Seismic Loads, ACI Structural
Journal, V. 110, No. 6, Nov.-Dec. 2013, pp. 941-952.
9. ASTM C39/C39M-12, Standard Test Method for Compressive
Strength of Cylindrical Concrete Specimens, ASTM International, West
Conshohocken, PA, 2012, 7 pp.
10. ACI Committee 440, Guide for the Design and Construction of
Structural Concrete Reinforced with FRP Bars (ACI 440.1R-06), American Concrete Institute, Farmington Hills, MI, 2006, 44 pp.
11. Deitz, D.; Harik, I.; and Gesund, H., Physical Properties of Glass
Fiber Reinforced Polymer Rebars in Compression, Journal of Composites for Construction, ASCE, V. 7, No. 4, 2003, pp. 363-366. doi: 10.1061/
(ASCE)1090-0268(2003)7:4(363)
12. CAN/CSA-S806-12, Design and Construction of Building Components with Fiber-Reinforced Polymers, Canadian Standards Association,
Mississauga, ON, Canada, 2012, 198 pp.
DISCUSSION
Discussion 111-S22/From the March-April 2014 ACI Structural Journal, p. 257
Bond Strength of Spliced Fiber-Reinforced Polymer Reinforcement. Paper by Ali Cihan Pay, Erdem Canbay,
and Robert J. Frosch
Discussion by Jos R. Mart-Vargas
Professor, ICITECH, Institute of Concrete Science and Technology, Universitat Politcnica de Valncia, Valncia, Spain
The discussed paper presents an interesting experimental study on the bond behavior of unconfined tension
lap-spliced reinforcement. Steel-reinforced concrete beams
and reinforced beams with ber-reinforced polymer (FRP)
barsglass FRP and carbon FRPwere tested to provide
additional experimental data for a better understanding of
the bond strength between FRP and concrete.
Variables such as splice length, surface condition, modulus
of elasticity, axial rigidity, and bar casting position on bond
strength were considered. The authors should be complimented
for producing a detailed paper with comprehensive information. This is acknowledged by the discusser, who would like to
offer the following comments and questions for their consideration and response, mainly about some aspects included in
the Bond Strength section and the corresponding conclusions.
The authors conclude that bond strength depends on splice
length, the modulus of elasticity of reinforcement, and the axial
rigidity of reinforcement, among other factors. These conclusions are supported by Fig. 6 to 10, as follows: (a) Fig. 6 (for
No. 5 bars) and 7 (for No. 8 bars) depict the effect of splice
length on bond strength; (b) Fig. 8 depicts the effect of modulus
of elasticity on bond strength; and (c) Fig. 9 (for No. 5 bars)
and 10 (for No. 8 bars) depict the effect of axial rigidity on bond
strength. As stated by the authors, bond strength rises nonlinearly with increasing splice length, and linearly as the modulus
of elasticity and/or the axial rigidity increase. However, despite
the values being normalized by the fourth root of concrete
compressive strength to eliminate the effect of variations in
concrete strength, these figures present the computed reinforcement stress reached at failure ftest instead of average bond stress
avg. Therefore, it seems that some conclusions should correspond to reinforcement stress rather than to bond strength.
By way of example, for the glass-sand coated case in
Fig. 6, normalized reinforcement stresses approximately
range from 30 to 50 ksi (207 to 345 MPa), whereas splice
length ranges from 12 to 54 in. (305 to 1372 mm). Then
normalized reinforcement stresses increase by 67%, whereas
splice length increases by 350%. It is true that the effect is
nonlinear. However, as bar size does not vary and the hypothesis of uniform distribution of bond stresses along splice
length is assumed, this implies that the average bond stress
is lower for the longer splice length case. In other words,
normalized reinforcement stress at failure increases when
splice length becomes longer, but bond strength decreases. If
the discusser is right, this conclusion is the opposite of that
reached by the authors. The same interpretation can be made
for the remaining cases included in Fig. 6 and 7.
Regarding the effect of the modulus of elasticity of the
reinforcement on bond strength (Fig. 8, No. 5 bars), one can
interpret that for one same deformation, a greater reinforcement stress results for a higher modulus of elasticity. By way
of example for the fabric texture case in Fig. 8, normalized
reinforcement stresses approximately range from 47 to 62 ksi
(324 to 427 MPa), whereas modulus of elasticity ranges

from 7300 to 18,500 ksi (50.3 to 127.6 GPa). Then, normalized reinforcement stresses increase by 32%, whereas
modulus of elasticity increases by 154%. It is true that the
effect is linear in terms of the normalized reinforcement
stresses. However, as bar size and splice length do not vary
and the hypothesis of uniform distribution of bond stresses
along splice length is assumed, additional information based
on reinforcement deformation is needed to know the effect
on average bond stress. This is because one same normalized reinforcement stress can be offered from two different
moduli of elasticity if different bar deformations are developed which, in turn, may generate distinct bond-stress relationships based on distinct reinforcement-to-concrete slips.
Regarding the effect of the axial rigidity of reinforcement on
bond strength, the discusser believes that the same reasoning
provided in the previous paragraph based on modulus of
elasticity is applicable, which is included in axial rigidity
together with the nominal cross-sectional area of the bar, and
is more complex in this case given the potential differences
in normalized reinforcement stresses and deformations due to
both modulus of elasticity and area parameters.
AUTHORS CLOSURE
The authors would like to thank the discusser for his
thoughtful comments. Unfortunately, it appears that there
was confusion regarding the use of the term bond strength.
The authors used the term bond strength consistent with
the terminology commonly used in the field and consistent
with that used by Joint ACI-ASCE Committee 408, Bond and
Development of Steel Reinforcement. In ACI 408R-03, Bond
and Development of Straight Reinforcing Bars in Tension, it
is stated that the term bond strength represents the maximum
bond force that may be sustained by a bar. The term bond
force represents the force that tends to move a reinforcing bar
parallel to its length with respect to the surrounding concrete.
Therefore, in the paper, when it was commented that the
bond strength increased, it was referring to the total force
or total stress resisted by the bar, not the average bond stress.
As discussed for Fig. 6, the normalized reinforcement stress
at failure increases as the splice length increases. Therefore,
considering the aforementioned terminology, it was concluded
that the bond strength increases as splice length increases. It
is agreed that the average bond stress (Table 2 provides the
values) decreases as the splice length increases due to the
nonlinear relationship of bond force to splice length.
As discussed for Fig. 8 through 10, the bond strength
increases linearly with increasing modulus of elasticity
and increasing axial rigidity. In these cases, both the bond
strength (bar stress or force) as well as the average bond
stress increase considering that the bar surface area and
splice length are identical for the cases compared. Even in
the case of the steel bar with the hole, the surface area of the
115
bar resisting bond remains the same. While the paper did
not discuss average bond stress, the average bond stress for
these cases would be directly computed from the bar force
divided by the surface area of the bar (pdbls), which is independent of the bar deformation or surface treatment. It is not
clear why the discussion indicates that additional information based on reinforcement deformation is needed to know
the effect on average bond stress. Regardless, bond strength
was found to be essentially independent of the surface deformation for the bars tested as discussed in the paper.
Overall, average bond stress is not a good measure to

evaluate the bond strength of reinforcing bars. As indicated
by the nonlinear relationship with splice length shown in
Fig. 6 and 7, bond stresses are nonlinear over the length of
the reinforcement. This nonlinearity results in significant
variation of the maximum average bond stress, as indicated
by the test results in Table 2. For this reason, the conclusions
were presented based on the bond strength (maximum bar
force or bar stress) rather than average bond stress.
Discussion 111-S25/From the March-April 2014 ACI Structural Journal, p. 291
Behavior of Epoxy-Injected Diagonally Cracked Full-Scale Reinforced Concrete Girders. Paper by MatthewT.
Smith, Daniel A. Howell, Mary Ann T. Triska, and Christopher Higgins
Discussion by William L. Gamble
FACI, Professor Emeritus, University of Illinois, Urbana, IL
This paper is an important contribution to the literature on

repair of reinforced concrete girders. However, I am not sure
that the test specimens were very representative of 1950s
details. The shear reinforcement is quite heavy compared to
at least some examples.
Gamble (1984) describes a reinforced concrete deck girder
(RCDG) bridge that was built in 1957 and suffered major distress
in 1967, followed by partial girder replacement and epoxy
injection. The bridge had been designed by the 1953 AASHTO
Spec (AASHTO 1953) and the provisions of that specification
are at least partially responsible for the distress.
The allowable service load shear stresses under
the 1953 AASHTO specification were extremely high,
compared to any current standard. The allowable value of
vc was equal to 0.04fc', with no limit on concrete strength.
The specified fc' = 4000 psi (27.6 MPa), so the service load
shear stress assigned to the concrete was 120 psi (0.83 MPa).
Shear reinforcement was required only when this stress was
exceeded, and the maximum shear stresses in this bridge
were slightly less than the allowable value. Any current
Code will assign a shear stress somewhat less than this
value, applied to factored loads rather than service loads,
and this change combined with other restrictions will lead to
shear reinforcement in nearly all reinforced concrete beams.
In addition, the 1953 AASHTO specification had almost no
minimum limits on shear reinforcement. There was no minimum
shear steel ratio Av/bws. There was no maximum spacing limit
such as the 24 in. (600 mm) now commonly observed, and the
d/2 limit in current codes was invoked only when shear reinforcement was required. ACI 318-51 (ACI Committee 318 1951) was
essentially similar with respect to shear design.
These defects were addressed in the first AASHTO Strength
Design Provisions, but were not addressed in the Working
Stress Design provisions until the 1974 Interim Specs.
ACI318-56 (ACI Committee 318 1956) had a partial fix to
this set of problems, with a major revision in the 1963 Code.
The test specimens had a shear steel ratio of approximately 0.0024. The lowest shear steel ratio in the structure
in Gamble (1984), which did not require shear reinforcement
by 1953 AASHTO requirements, was approximately 0.00035,
and the highest approximately 0.0007. ACI 318-56 required
a minimum shear steel ratio of 0.0015 (when web reinforcement was required), and current ACI Codes and the last
AASHTO non-LRFD spec require a minimum of 0.00125 for
Grade 40 steel and 0.00083 for Grade 60 steel. The LRFD
116
minimum shear reinforcement ratios are somewhat larger but

still much smaller than that used in the test specimens.
A lower shear steel ratio would have led to larger, less wellcontrolled cracks and would have been more typical of the 1950s.
Stirrups spaced at d/2 would also have been more representative.
The bridge in Gamble (1984) needs a little additional
comment herein. It had a significant problem with concrete
strength in the span that failed, but had significant shear
cracking in other spans with adequate concrete strengths.
It was designed as a rigid frame structure with integral
abutments, and as such apparently developed significant axial tensions due to restrained shrinkage in spite of
expansion joints in two locations. The bridge was one of
approximately 85 similar designs (with different numbers of
spans, locations of expansion joints, and foundation details),
and approximately half of these had experienced significant
cracking requiring repairs by 2002 (personal communication; Gamble [2002]). None of the other similar structures
had significantly deficient concrete strengths, so it must be
concluded that the combination of very high shear stresses
assigned to the concrete in conjunction with axial tensions
from restrained shrinkage led to the distress.
REFERENCES
AASHTO, 1953, Standard Specifications for Highway Bridges, sixth

edition, American Association of State Highway Officials, Washington,
DC, 328 pp.
ACI Committee 318, 1951, Building Code Requirements for Reinforced
Concrete (ACI 318-51), American Concrete Institute, Farmington Hills, MI, 63 pp.
ACI Committee 318, 1956, Building Code Requirements for Reinforced
Concrete (ACI 318-56), American Concrete Institute, Farmington Hills, MI, 77 pp.
Gamble, W. L., 1984, Bridge Evaluation Yields Valuable Lessons,
Concrete International, V. 6, No. 6, June, pp. 68-74.
Personal communication from W. D. Gamble, deceased, 2002.
AUTHORS CLOSURE
The authors would like to thank the discusser for his
thoughtful comments and for reading our paper. These allow
us to provide additional detail regarding the design considerations for our specimens that could not be reported in the
original paper due to space constraints.
The issues related to what many would consider a shear
problem for 1950s-era RCDGs are several. We respectfully
submit Table 3, which highlights the relevant changes in the
AASHTO 1944, 1949, and 1953 standards from the period
considered. In addition, the AASHTO allowable stress for the
transverse reinforcing steel changed from 16 ksi (110 MPa)
in 1949 to 20 ksi (138 MPa) in 1953. The maximum spacing

of transverse steel, when required, was prescribed as threequarters the section depth. While the prescribed allowable
concrete stresses for shear are different than those described
by the discusser, they do not change the premise of his discussion. We strongly agree that RCDG bridges of the 1950s are
under-reinforced for shear by modern standards and that they
represent a disproportionately large number of bridges with
reported diagonal-tension cracking. However, it is not only the
concrete allowable stress and transverse reinforcement which
contributes to their deficiencies. Bridge designers now widely
recognize the interaction between flexure and shear on the
diagonal-tension behavior of reinforced concrete girders and
beams. Therefore, one must consider the nexus of the many
different changes that were taking place in the design specifications (AASHTO and ACI) as well as in material standardization (ASTM) at the time. One of the most important changes
that took place was the standardization of reinforcing steel bars
that occurred in 1950 with the adoption of ASTM305-50T
(1950). The newly standardized deformation patterns substantially relaxed the anchorage and bond requirements and
design practice changed rapidly from the past as according
to ACI208-58: Acceptance by the ACI Building Code
committee of the ASTM A305 definition of a deformed bar
produced an immediate drastic change in both structural and
general practice. Where designers would previously have
required hooks and bends to anchor or transition reinforcing
steel in flexural tension zones, they could now terminate the
steel with straight-bar cutoffs. This was an era of economy of
materials and designers worked hard to produce highly efficient designs. As a result, RCDGs designed during the 1950s
and early 1960s typically contain poorly detailed flexural reinforcing steel in addition to being lightly reinforced with transverse reinforcing steel. These are the population of bridges that
commonly contain diagonal cracks and which require redress.
Our research has focused on assessing and remediating
shear-moment deficiencies in the population of RCDG bridges
from around the 1950s. To do this, we have attempted to
mimic the original features of the girders and cross beams
in overall geometry, concrete and reinforcing materials and
details, as well as loading geometry to control the moment-toshear ratios in the specimens to those that reflect the in-place
conditions. When we began this work over a decade ago, my
colleagues and I developed a database of the relevant geometry and details (Higgins et al. 2004). We considered RCDG
bridges constructed from 1947 to 1962 that were identified as
cracked within the inventory contained by the Oregon Department of Transportation, which maintains a large population
of this type of bridge. There were 442 bridges in the database
with 1487 separate spans. Structural drawings for each span
were reviewed and parameters corresponding to overall bridge
geometry (span length, indeterminacy, skew, girder spacing,
number of lanes, number of girder lines, size, location, and

spacing of diaphragms, as well as deck thickness); material
properties (concrete compressive strength and reinforcing steel
properties); and member proportions and reinforcement details
(cross-section type, web width, member height, haunch, taper,
span-depth ratios, flexural reinforcing steel areas in positive
and negative regions, stirrup sizes and spacing in high and low
shear zones, for example) were recorded in the database. The
database was queried to provide summary details for individual
parameters and relationships between parameters. Further,
dead load magnitudes and live load capacities were developed
for comparison with specified load models, weigh-in-motion
service-level loads, and permit tables. Using information
extracted from the population of cracked RCDG bridge spans
contained in the database, laboratory specimens were dimensioned and individual bridges were identified for field investigation. The most common details in high-shear regions for
positive- and negative-moment regions (average results which
corresponded closely to the mode of the data) were selected as
representative. Also using this data set, we conducted experiments at the practical bounds, such as no stirrups and heavy stirrups (equivalent transverse reinforcing ratios of 0.0 and 0.0048,
respectively), with and without cutoff flexural steel, the results
of which are reported by Higgins et al. (2004, 2007).
Based on the detailed investigation of the many different
parameters that coalesce around the problem space and in-depth
study of large numbers of actual designs, we feel our specimens are indeed representative of high-shear negative moment
regions in RCDGs of the 1950s. They were not intended to
represent the minimum case, and based on our review of
vintage designs, find that the negative-moment region near
continuous supports contained higher-than-minimum stirrups
(even by modern standards).We acknowledge that there are
girders with more and less transverse steel and we further feel
that our findings would hold for girders with at least minimum
stirrups (as defined by the modern design standards).
REFERENCES
AASHTO, 1944, Standard Specifications for Highway Bridges, fourth

edition, American Association of State Highway Officials, Washington, DC.
AASHTO, 1949, Standard Specifications for Highway Bridges, fifth
edition, American Association of State Highway Officials, Washington, DC.
ACI Committee 208, 1964, Test Procedure to Determine Relative Bond
value of Reinforcing Bars (ACI 208-58), American Concrete Institute,
Farmington Hills, MI.
ASTM A305-50T, 1950, Tentative Specifications for Minimum
Requirements for the Deformations of Deformed Steel Bars Concrete
Reinforcement, ASTM International, West Conshohocken, PA.
Higgins, C.; Yim, S.; Miller, T.; Robelo, M.; and Potisuk, T., 2004,
Remaining Life of Reinforced Concrete Beams with Diagonal-Tension
Cracks, Report No. FHWA-OR-RD-04-12, Research Unit, Oregon
Department of Transportation, Salem, OR, 124 pp.
Higgins, C.; Potisuk, T.; Farrow III, W. C.; Robelo, M. J.; McAuliffe, T.K.;
and Nicholas, B. S., 2007, Tests of RC Deck Girders with 1950s Vintage
Details, Journal of Bridge Engineering, ASCE, V. 12, No. 5, pp. 621-631.
Table 3Comparison of relevant AASHTO-specified concrete shear and bond stresses, noting changes
occurring around 1950
Allowable shear stress in concrete, psi
With web reinforcement
Without web reinforcement
Longitudinal bars not Longitudinal bars Longitudinal bars not Longitudinal bars
Design specification and year
anchored
anchored
anchored
anchored
AASHTO 1944
AASHTO 1949
AASHTO 1953
0.046fc
0.046fc
0.075fc
0.06fc
0.06fc
0.075fc
0.02fc
0.02fc
0.02fc
0.03fc
0.03fc
0.03fc
Allowable bond stress (straight), psi

Structural or intermediate grade
0.033fc(max 100)
0.05fc(max 150)
0.10fc(max 350)
Note: 1 psi = 6.89 kPa.
117
NOTES:
118
2015 ACI Membership Application

American Concrete Institute 38800 Country Club Drive Farmington Hills, MI 48331 USA
Phone: +1.248.848.3800 Fax: +1.248.848.3801 Web: www.concrete.org
Please print or type all information requested below:

Mr. Mrs. Ms. Dr.
Gender: Female Male
FIRST NAME_______________________________________ MIDDLE INITIAL____________ LAST NAME (SURNAME)_______________________________________________________
EMPLOYER__________________________________________________________________ CORPORATE TITLE___________________________________________________________
E-MAIL ADDRESS_____________________________________________________________ BIRTHDATE__________________________________________________________________
ADDRESS________________________________________________________________________________________________________________________________________________
CITY______________________________________________ STATE____________________ ZIP___________________ COUNTRY ____________________________________________
TELEPHONE_________________________________________________________________ FAX_________________________________________________________________________
Categories of membership
Please select the desired category of membership and submit the appropriate dues described below.
ORGANIZATIONAL $1020 PLUS APPLICABLE SHIPPING FEES

Includes one set of the Manual of Concrete Practice (MCP), a wall plaque, and a subscription to Concrete
International, Concrete Repair Bulletin, and both ACI Materials and Structural Journals.
MANUAL OF CONCRETE PRACTICE: Select a format and include applicable shipping fee:
Seven-volume Set U.S. add $30 shipping fee; Canada and Mexico add $130 shipping
fee; and all others, add $220 shipping fee

CD no additional shipping costs
Online Subscription no additional shipping costs
INDIVIDUAL $230/year
Individuals 28 years old or above residing worldwide.
YOUNG PROFESSIONAL $130/year
Individuals under the age of 28 who do not qualify for student membership.
E-STUDENT FREE Join at www.concrete.org/Membership/StudentMembership.aspx.
STUDENT (U.S. AND CANADA) $40/year
STUDENT (Outside U.S. and Canada) $80/year
Individuals under the age of 28 who are registered full-time students at an educational institution.
Full-time students age 28 and above may be granted Student Membership when the request is endorsed by the students advisor.
SUSTAINING MEMBERSHIP ACI Sustaining Members receive all membership benefits of
Organizational Members plus a free copy of every new ACI publication and increased corporate exposure, positioning them as a leader in the concrete industry, and much more. For complete details or to
join, visit www.sustaining.concrete.org or call +1.248.848.3800.
Profile Information (Select all that apply.)

Markets
Occupations
Design
Materials
Production
Construction
Testing
Repair
Owner
Management
Consultant
Engineer
Architect
Contractor
Technical Specialist
Quality Control
Inspector
Craftsman
Sales & Marketing
Association
Employee
Government
Employee
Included Subscriptions
Your ACI membership includes a hard copy and
online subscription to Concrete International and
your choice of one of the following:
Digital access to the ACI Materials Journal
and ACI Structural Journal
Concrete Repair Bulletin, hard copy
Additional Periodical Options
To add subscriptions in hard copy: please check a

circle below and include an additional $47 ($52 for
individuals outside U.S. & Canada, $30 for
students, and $42 for students outside of the U.S. &
Canada) per title.
ACI Structural Journal, hard copy
ACI Materials Journal, hard copy
Please allow up to 2 months for delivery outside the U.S.
Optional Online Subscriptions

Bundle your ACI membership with these optional
online subscriptions for additional ACI resources:
Symposium Papers Subscription, $98
Manual of Concrete Practice, $409
Researcher
Educator
Student
Other __________
Payment
Fees
CHECK NUMBER
Membership dues
CHECK payable to American Concrete Institute
Additional subscriptions
CREDIT CARD Visa MasterCard American Express
Organizational six volume set

shipping fees
For Journals, Canadian residents

add 5% GST (#126213149RT)
ACCOUNT NUMBER
EXPIRATION DATE
NAME (AS APPEARS ON CREDIT CARD)
TOTAL (U.S. Funds Only)

SIGNATURE
Code: SJ
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
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.

Aci Structural Journal Jan.-Feb. 2015 v. 112 No. 01

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aci Structural Journal Jan.-Feb. 2015 v. 112 No. 01

Uploaded by

Copyright:

Available Formats

V. 112, NO.

A JOURNAL OF THE AMERICAN CONCRETE INSTITUTE

ACI Structural Journal

January-February 2015, V. 112, No. 1

Technical Activities Committee

a journal of the american concrete institute

Bond Strength of Spliced Fiber-Reinforced Polymer Reinforcement. Paper by

Ali Cihan Pay, Erdem Canbay, and Robert J. Frosch

ACI Structural Journal/January-February 2015

20-22National Concrete Masonry

18-192015 ICON-Xchange, San Antonio,

UPCOMING ACI CONVENTIONS

ON COVER: 112-S03, p. 27, Fig. 6Axial test setup.

The ACI Structural Journal is an open

ACI Structural Journal/January-February 2015

ACI STRUCTURAL JOURNAL

Title No. 112-S01

A Global Integrity Parameter with Acoustic Emission for

materials and techniques combined with a need for rapid and

where tan(i) is the secant stiffness of any point i on the

Fig. 1Deviation from linearity (after ACI Committee 437

assistance of additional data, such as crack mouth opening

displacement response; at 35%, IDL cracks started to remain

Fig. 2: Photograph and cross-section details (after Barrios

(27.6 MPa) concrete deck was cast in the laboratory prior to

Table 1Summary of girder specimens

Strand yield followed by deck failure

Strand yield followed by deck failure

Strand yield followed by deck failure

ACI Structural Journal/January-February 2015

capacity of these girder specimens. The maximum load level

Fig. 3Acoustic emission sensor layout (after Barrios and

ACI Structural Journal/January-February 2015

Fig. 5Load versus displacement due to CLT loading

Fig. 6Cracking patterns (after Barrios and Ziehl [2012]).

Fig. 7Damage quadrants (normalweight specimens).

Fig. 8Damage quadrants (lightweight specimens).

Proposed AE damage descriptor (arc of damage)

ACI Structural Journal/January-February 2015

Fig. 9CR-versus-CLR plots. (Note: 1 kip = 4.4 kN.)

increases steadily after Loadset 7 and follows a closely linear

Fig. 11Definition of arc of damage.

Fig. 13Arc of damage versus percent of Pn (lightweight

and for normalweight concrete

Fig. 12Arc of damage versus percent of Pn (normalweight

GIPAE = AD1 1.0

where AD is the arc of damage for any loadset; PT is the

Fig. 14Values at damage detection.

ACI Structural Journal/January-February 2015

ACI Committee 318, 2011, Building Code Requirements for Structural

ACI Structural Journal/January-February 2015

ACI STRUCTURAL JOURNAL

Title No. 112-S02

Analysis of Rectangular Sections Using Transformed

gular solid section with four-corner reinforcement and the

where Rd, Rd y, and Rd z are the relative values of the

combined design axial load and biaxial bending resistance

The relative values of the combined design axial force and

combined design axial force and biaxial bending resistance

f cdT bT (hT ) 2 = 1 (12)

f cdT hT (bT ) 2 = 1 (13)

Equations (11) to (13) represents a system of independent

Because the products in Eq. (11) to (13) are dimensionless, it

ACI Structural Journal/January-February 2015

ACI STRUCTURAL JOURNAL

ACI STRUCTURAL JOURNAL

ACI STRUCTURAL JOURNAL