You are on page 1of 12

Engineering Structures 31 (2009) 23802391

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Modeling short-term tension stiffening in reinforced concrete prisms using a


continuum-based finite element model
H.Q. Wu, R.I. Gilbert
Centre for Infrastructure Engineering and Safety, School of Civil and Environmental Engineering, The University of New South Wales, Sydney, Australia

article

info

Article history:
Received 16 December 2008
Received in revised form
19 May 2009
Accepted 19 May 2009
Available online 13 June 2009
Keywords:
Tension stiffening
Reinforced concrete
Bond
Primary cracks
Finite element method

abstract
Tension stiffening is most often included in models of reinforced concrete by modifying the constitutive
laws of the tensile concrete. In reality, tension stiffening is caused by the bond stress that develops at
the steelconcrete interface between the primary cracks. In this paper, a modified CEBFIP bond model
is incorporated into a non-linear finite element program to accurately model tension stiffening at the
serviceability limit states. The bondslip relationship at any point along the reinforcement bar is modified
to account for the local damage of the surrounding concrete, as well as the level of steel stress. A non-local
analysis is undertaken to adjust the constitutive law of the bond interface element at each load step. The
proposed model is shown to accurately predict the crack spacing, stresses and deformation in axially
loaded tension members at typical in-service load levels.
2009 Elsevier Ltd. All rights reserved.

1. Introduction
Tension stiffening is the contribution to the member stiffness
of the intact concrete between the primary cracks and it plays
a significant role in the deformation of reinforced concrete at
the serviceability limit states, particularly in the case of lightly
reinforced members. Under typical in-service load levels, the
concrete between the primary cracks carries significant tensile
stress and the actual member response is considerably stiffer than
the response of the bare steel bar. To accurately simulate the inservice behavior of reinforced concrete, tension stiffening must be
accurately modeled.
Consider the reinforced concrete prism in uniaxial tension
shown in Fig. 1(a). Prior to cracking, behavior is essentially linearelastic. The first primary crack occurs at the weakest cross-section
of the member at the cracking load Pcr and, at loads P > Pcr ,
the global load-average strain response becomes non-linear, as
shown in Fig. 1(b). Tension stiffening at this global level is often
represented as the difference between the bare bar strain and the
actual member strain (ts ) at any particular load P. Recently,
Fields and Bischoff [1] introduced the concept of a tension
stiffening factor t which is defined as the tension stiffening strain
at load P (ts ) divided by the maximum tension stiffening strain
(ts.max ) which occurs just prior to first cracking at P = Pcr .
Fig. 1(c) shows the relationship between the tension stiffening
factor and the average member strain.

Corresponding author. Tel.: +61 2 9385 6002; fax: +61 2 9313 8341.
E-mail address: i.gilbert@unsw.edu.au (R.I. Gilbert).

0141-0296/$ see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2009.05.012

A detailed explanation of the mechanism of tension stiffening


at different loading stages in a reinforced concrete member (before
yielding of the steel reinforcement) is given by Gilbert and Wu [2].
The decay of tension stiffening with increasing load is mainly
attributed to the formation of new primary cracks during the crack
formation phase and due to degradation of bond during the crack
stabilization phase, as illustrated in Fig. 1(b) and (c). The formation
of each new primary crack causes a loss of concrete tensile stress in
the regions adjacent to the crack. After all the primary cracks have
developed, the crack stabilization phase begins. Cover-controlled
cracks occur under increasing load at the steelconcrete interface
causing a loss of bond and hence a loss of tension stiffening.
The cover-controlled cracks are the internal cracks that radiate
from the bar deformations and are contained within the concrete
cover [3].
Fig. 1(d) shows a typical length of specimen between two
adjacent primary cracks during the crack stabilization stage. At
each crack, the concrete stress is zero. The concrete tensile
stress gradually increases as the distance from the nearest crack
increases, due to bond stress that develops at the steelconcrete
interface, reaching a maximum mid-way between the two cracks.
The intact concrete between the two primary cracks remains
elastic during the entire loading period and the maximum concrete
stress is less than the tensile strength of concrete. Fig. 1(d) also
shows the bond stress and slip distribution at the steelconcrete
interface between the two primary cracks. Bond stress is zero at
the section containing the primary crack, because the concrete and
the steel bar are not in contact. However, the slip at the primary
crack is at a maximum and decreases to zero mid-way between

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391

Notations
The following symbols are used in this paper:
Dc

Damage parameter of concrete elements;

Dc

Spatial averaged damage parameter of concrete


elements;
db
Reinforcing bar diameter;
dmax
Maximum aggregate size;
Ec
Elastic modulus of concrete;
Ecp
Secant modulus at peak uniaxial compressive stress;
Gf
Fracture energy of concrete;
fcp
Compressive strength of the in situ concrete;
fct
Uniaxial tensile strength of concrete;
fc
Concrete compressive strength;
k
Decay factor for post peak response of compressive
stressstrain curve;
lt
Element size;
n
Ratio used in defining the equivalent uniaxial
compressive stressstrain curve (Eq. (11));
Pc
Spatial averaged confinement stress of concrete
elements;
pc
Confinement pressure in the concrete elements
around the reinforcing bar;
R
Radius defining the region of neighborhood concrete element integration points;
r
Distance between the target bond element integration point and its neighborhood concrete element
integration points;
s
Slip;
s1 , s2 , ds3 Values of slip defining CEBFIP bondslip model;

Parameter defining the ascending part of CEB


bondslip curve;
f
Coefficient related to aggregate size;
1 , 2 , 3 Parameters defining the tensile stressstrain relationship of concrete;
0
Ratio between major and minor principal stresses;

Scaling factor for compressive strength;


t
Tension stiffening factor;
1 , 2 Scaling factor for major and minor principal directions;
cp
Concrete compressive strain at peak stress corresponding to fcp ;
tp
Concrete tensile strain at peak stress corresponding
to fct

cp
Concrete compressive strain at peak stress corresponding to fc ;
1
Principal tensile strain;

Strain ratio used in defining equivalent uniaxial


compressive stressstrain curve (Eq. (11));
1 , 2 Parameters in the proposed bond model defining
the effects of cracking and steel stress on bond;
1c , 2c Concrete stresses at principal directions;
s
Steel stress at primary crack;
sy
Steel yield stress;
b
Bond stress;
max
Maximum bond stress or bond strength;
(r )
Non-local weight function;
0 (r ) Normalized weight function;

Ratio between steel stress and yielding stress;

An increment of external load causes a decrease in the bond stress


(Fig. 1(d)) and a widening of existing primary cracks associated
with increasing slip.
The magnitude of the bond stress for a cracked tension member
depends on many factors, some of which vary with the applied
load, and accurate modeling of bond, and hence tension stiffening,
is a complicated task. The local bond stress is not only dependent
on the local slip, but also on the magnitude of stress and strain
in the reinforcing steel, the duration of the applied load and
the level of drying shrinkage. In this paper, the time-dependent
effects of shrinkage are not considered, although restraint to drying
shrinkage greatly affects tension stiffening [4,2,5]. Early shrinkage
causes a reduction in the cracking load and, under sustained service
loads; shrinkage initiates the formation of additional primary
cracks with time and causes a time-dependent degradation of the
steelconcrete bond.
Most of the tension stiffening models in both analytical and
continuum-based finite element methods incorporate tension
stiffening by either modifying the constitutive law of the tensile
concrete [613] or that of the tensile reinforcement [14,15].
However, in reality, at any point in the vicinity of a crack in
the tensile concrete, the concrete stress and strain both reduce
(elastically unload) at first cracking, with the level of reduction
varying with distance from the nearest crack to the point in
question. Tension stiffening is in fact a result of elastic unloading
of the concrete after cracking rather than post-peak softening. The
tensile stress in the concrete between cracks is caused by bond
stress at the steelconcrete interface and it is the deterioration in
bond that causes a reduction in tension stiffening with increasing
load. Tension stiffening is most rationally modeled using a well
calibrated local bond stressslip relationship
In order to better describe this mechanism in the finite element
analysis, a modified CEBFIP bond model is proposed in this paper.
The modified bond model is incorporated into the continuumbased finite element program RECAP [16] to analyze a number
of uniaxially loaded tension members. The numerical results
obtained using the proposed model are compared with both the
experimental results and numerical results obtained assuming the
original CEBFIP bond model [17]. The proposed model is shown to
accurately predict the behavior of the test specimens at all stages
of loading.
2. Concrete model
The biaxial concrete strength envelope proposed by Foster and
Marti [18], and shown in Fig. 2, is used in the finite element model.
In Fig. 2, 1c and 2c are the principal stresses, and fcp is the uniaxial
compressive strength of the in-situ concrete.
For concrete in tension, the bilinear softening curve proposed
by Petersson [19] is used (Fig. 3(a)). Cracking of concrete is
introduced in the element as soon as the principal tensile stress
exceeds the tensile strength of the concrete. In order to deal with
the issue of mesh sensitivity induced by localization of deformation
in the vicinity of a crack in a concrete element, the crack band
theory approach [20] is used to depict the softening branch of the
tensile stressstrain relationship in the fracture zone immediately
adjacent to the crack. The three parameters 1 , 2 , and 3 , which
are defined in Fig. 3(a) and are required to define the tensile
stressstrain relationship, are given as:

1 =
the two cracks, as shown. Bond stress increases rapidly adjacent to
each crack and then decreases to zero mid-way between the cracks.

2381

1
3

2 =

2
9

3 + 1 ;

3 =

18 Ec Gf
5 fct2 lt

(1)

where Ec is the elastic modulus of concrete; lt is the element size; fct


is the uniaxial tensile strength of concrete; Gf is the fracture energy

2382

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391

Fig. 1. Stresses and deformations in an axially loaded tension member.

where is a scaling factor [21] which depends on the stress in


the orthogonal principal direction, as illustrated in Fig. 3(b). The
concrete strain at peak stress is also modified by the same scaling
factor , so that:

cp
= cp .

(5)

In the compressioncompression state, is greater than 1.0


and depends on the ratio of major and minor principal stresses
0 = 1c /2c [21]:

2 =
2 =
Fig. 2. Biaxial concrete strength envelope of [18].

of concrete associated with tensile softening, which is here taken


as specified in the CEBFIP model code [17]:
Gf = f

fcp
10

0.7

(fcp in MPa and Gf in N/mm)

for 0 0 0.48

1 0 /2.4

1.15 1 + 5.51
1 + 0 /5.5

(6)


for 0.48 < 0 1.0

(7)

where 2 is the scaling factor accounting for confinement in the


minor principal direction. The confinement scaling factor for the
major principal direction 1 is given by multiplying 2 by 0 :

1 = 0 2 .

(8)

(2)

and f is a coefficient related to the maximum aggregate size (dmax


in mm) and is taken as:

f = (1.25dmax + 10) 103 .

1.0

In the tensioncompression state, the major principal stress


is tensile and it reduces the compressive strength in the minor
principal direction [2224]. The scaling factor developed by
Vecchio and Collins [21] is adopted here:

(3)
1

In this study, the concrete compressive strength in one principal


direction is given by:

fc = fcp

where 1 is the principal tensile strain.

(4)

0.8 + 0.34 1

1.0

cp

(9)

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391

(a) Concrete in tension.

2383

(b) Concrete in compression.


Fig. 3. Stressstrain relationships for concrete.

(a) Cover-controlled cracking at the


steel-concrete interface.

(b) Idealized averaged bond stress.

Fig. 4. Typical configuration of pull-out tests.

The uniaxial compressive stressstrain relationship proposed


by Thorenfeldt et al. [25] is used in this study. That is

c = fcp

(10)

n 1 + nk

in which:
n=

Ec
Ec Ecp

|c |
cp

(11)

cp is the strain corresponding to the peak stress; and Ecp is the sef

cant modulus at the peak of the curve, Ecp = cp . The parameter k in


cp
Eq. (10) is the decay factor associated with the post peak response
and is given by Collins and Porasz [26] as:
k = 1.0

for |c | cp

k = 0.67 +

fcp
62

1.0 for |c | > cp .

(12)
(13)

3. Bond interface model


The bond model must correctly predict the relationship
between local bond stress and slip, including the effects of concrete
cracking and load intensity (i.e. steel stress level). Results of
different pull-out tests (e.g. [2732]) have indicated that the bond
stress is influenced by many parameters, such as concrete strength,
bar diameter, bar spacing, transverse reinforcement, confining
stress, and so on. However, most of the pull-out tests have been
undertaken on short anchorage lengths from 1 db to 6 db (where db
is the reinforcing bar diameter) with variable boundary conditions,
and cracking and steel stress states are usually ignored.
Fig. 4 shows a typical pull-out test configuration. Since the
embedment length is normally shorter than the crack spacing, no
primary cracks can form in the test specimens. The bond stressslip
curves generated in the tests are based on the measured load
applied to the bar and the slip at the free ends, so that the bond

Fig. 5. CEBFIP bond stressslip relationship.

stressslip relationships represent the average response. These


tests take account of the global effects of the localized covercontrolled cracks (shown in Fig. 4(a)) on bond degradation and slip.
The bondslip relationship specified in the CEBFIP code (Fig. 5)
is based on the average bond stressslip relationship measured in
pull-out tests by Eligehausen et al. [30]. The bond stress generally
decreases with increasing steel stress and Shima et al. [33] and
Kankam [34] suggested that the effect of steel strain should be
introduced into the bondslip relationship.
In the following, the CEBFIP bond model is modified by
incorporating the effects of local damage, primary cracking and
concrete confinement pressure. At service load levels, when the
steel stressstrain relationship is linear-elastic (i.e. the steel stress
is less than the yield stress), slip normally does not exceed s1 in
Fig. 5. The basic bondslip relationship (ignoring cracking and steel
stress state) is therefore based on the ascending part of CEBFIP
bond model, which can be expressed as:

b = max

s
s1


(14)

in which b is the bond stress, max is the maximum bond stress


which depends on the size and pattern of deformations on the bar

2384

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391

2r

(a) Concrete element integration points.

(b) Weight function for non-local analysis.

Fig. 6. Non-local averaging scheme for the bond elements.

surface (and is taken here to be 2.5 fcp for deformed high-bond

bars and 1.0 fcp for plain round bars); and s is the slip. Eq. (14)
is suitable for modeling the bondslip relationship if cracking is
considered to be smeared throughout the tension zone, but it must
be modified when the problem is considered at a more microscopic
level, to reflect the effect on the local bondslip relationship of
the proximity of discrete primary cracks to each particular bond
interface element.
A number of investigators has proposed to incorporate concrete
cracking or damage into the bond constitutive law [3537],
by reducing the bond stiffness as a result of damage to the
surrounding concrete elements. On the basis of experimental
results, Maekawa et al. [38] proposed a bond model that is related
to the local steel strain.
As shown in Fig. 1, soon after first cracking, the drop in tension
stiffening is substantially attributed to the formation of primary
cracks, where at each crack the bond stress drops to zero and
the slip is substantial. After the formation of the primary cracks
(at the crack stabilization stage), the loss of tension stiffening
can be best modeled by reducing the bond stress with increasing
local steel stress (to model the development of cover controlled
cracking). Two governing parameters 1 and 2 are introduced
here to take into account the local concrete damage at primary
cracks (1 ) and the effect of steel stress on bond degradation (2 ).
The parameter 1 varies from 1.0, when the concrete in the vicinity
of the crack is undamaged and the average principal tensile strain is
less than tp (see Fig. 3(a)), down to 0.0 when the average principal
tensile strain exceeds 3 tp (see Fig. 3(a)) as given by Eq. (15a).
The parameter is a simple model of the phenomena previously
identified by others [3537]. The parameter 2 has been included
to model the observed change in the bondslip relationship at steel
stresses exceeding 250 MPa and varies linearly from 1.0 at a steel
stress of 250 MPa to 0.0 at a steel stress exceeding 500 MPa, as
given by Eq. (15b).

1
2
2
2

= 1 D c , and
= 1.0 when s < 250 MPa;
= 2.0 0.004s when 250 MPa s 500 MPa;
= 0.0 when s > 500 MPa

(15a)

(Dc = 0) and cracked at 1 > tp (0 < Dc 1). A simple


expression to model the transition between undamaged (uncracked) and extensively damaged has been developed for Dc as
follows:
for 1 tp

Dc = 0

where s is the local steel stress, Dc is the non-local averaged


damage parameter of concrete elements based on the concrete
membrane model adopted. The product 1 2 (calculated using
Eqs. (15a) and (15b)) has been calibrated within the finite element
formulation using the experimental data of Bischoff [4] and Gilbert
and Nejadi [39].
According to the basic concept of continuum damage theory,
the magnitude of damage of the concrete can be represented by
a damage variable, Dc . In this study, only the cracking damage in
the major principal direction is considered as the parameter affecting the local bond deterioration. Based on the concrete tensile
stressstrain law (Fig. 3(a)), the concrete remains intact at 1 tp

(16b)
(16c)

In Eq. (14), max and s1 are treated as constants, which


are related to the bond confinement condition. The test data
in [32] shows that the bond strength envelope increases and its
radial deformation decreases with the application of confinement
pressure. This test data is adopted here to alter the peak bond
strength max according to the average confinement stress pc of
concrete around the steel bar elements. To model the test data, a
parameter 3 is selected as a variable power of max as follows:


pc
3 = 1 +
f

(17)

cp

and, consequently, the modified bond stressslip law can be


written as:
3

b = 1 2 max

s1

(18)

In order to incorporate these parameters into the bond


stressslip relationship in the finite element program, a non-local
analysis is undertaken at the end of each load step. As shown in
Fig. 6, at each integration point in the bond interface elements,
the surrounding concrete damage value and confinement stress
within a radius R is averaged by a weighting function (r ), where
r is the distance between the bond element integration point and
the source concrete element integration points within the radius
R (as shown in Fig. 6(a)). Therefore, the weighted average damage
parameter Dc and confinement stress pc can be written as:

Z
(15b)

(16a)

1c
Dc = 1
for tp < 1 3 tp
Ec 1
Dc = 1 for 1 > 3 tp .

Dc =

0 (r ) Dc (r , ) dV

(19a)

0 (r ) pc (r , ) dV

(19b)

Z
pc =
V

where the normalized weight function 0 (r ) is:

0 (r ) = R
V

(r )
.
(r ) dV

(20)

The weight function in this study is the dome shaped surface of


Fig. 6(b) given by:

"   #
2
2r
(r ) = exp
.
R

(21)

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391

2385

Fig. 7. Dimensions and finite element mesh adopted for uniaxial tension specimens.

The weighting function is largest at the nearest integration


point and decreases as the distance r increases. The weighted
averaged parameters Dc and pc are then substituted into Eqs. (15)
and (17) to give the governing parameters 1 and 3 . In this study,
only the concrete within a distance of 1.5 db from the surface of
the bar is assumed to influence the bondslip relationship, i.e. R =
2 db in Fig. 6. This assumption has been found to give reasonable
agreement with the available test data.
4. Finite element modeling of uniaxial tension tests
The 2-D finite element program RECAP [16,40,18], incorporating the bondslip relationship of Eq. (18), was used to model a
series of short-term uniaxial tension tests conducted by Wu and
Gilbert [41]. Each test specimen consisted of a concrete prism of
square cross-section (100 mm by 100 mm) and 1100 mm long, containing a single hot-rolled deformed reinforcing bar running longitudinally through the centroid of each cross-section, as shown in
Fig. 7. The steel bar was Grade 500 normal ductility steel, i.e. the
nominal characteristic yield stress was 500 MPa. The tensile axial
load was applied to the ends of the reinforcing bar protruding from
each end of the concrete prism. The experimental and numerical
results from two prisms (STN12 and STN16) are presented here.
The diameters of the deformed reinforcing bar in STN12 and STN16
were 12 mm and 16 mm, respectively. Each prism was moist cured
prior to testing, so that the magnitude of shrinkage in the concrete
at the time of testing was small (about 25 ) and is ignored here.
The finite element mesh used in the analyses is also shown in Fig. 7.
Taking advantage of symmetry, only one quarter of the test specimen was analyzed.
An element size of 5 mm by 5 mm was adopted, so as to reliably
model the local stress redistribution between adjacent primary
cracks. The prisms are modeled using 1332 nodes, 1100 concrete
elements, 110 two node steel reinforcement truss elements and
110 zero width bond interface elements. The concrete elements are

isoparametric quadrilateral plane stress elements with numerical


integration performed using 2 2 gauss quadrature. The steel
elements are connected to the concrete elements with overlapping
nodes by the zero-width bond interface elements (also shown in
Fig. 7). The finite element solution is dependent on the mesh size.
Relatively small elements are necessary to accurately model the
discrete nature of cracking and bondslip at the steelconcrete
interface. However, for the specimen sizes considered here, a
further reduction in the element size resulted in relatively little
increase in accuracy.
During the experiments, as the applied load increased, primary
cracks formed at reasonably regular centers. At the crack
stabilization stage, 5 primary cracks were observed along both
STN12 and STN16, with average crack spacing of 195 mm and
165 mm, respectively. In the finite element modeling, randomized
tensile strengths are generated in concrete elements with a 2%
maximum difference being assumed. Therefore, first cracking
was initiated in the concrete element with the lowest assigned
tensile strength and the tensile stress in surrounding elements
in the vicinity of the first crack dropped sharply. As the load
was increased, additional cracking occurred subsequently in
elements at distances far enough away from the existing cracks
for concrete stress levels not to be significantly affected. As the
distance from the nearest crack increases, the bond stress transfers
the tensile force from steel bar to the undamaged surrounding
concrete.
The concrete and steel properties adopted in the study are
the properties measured and reported by Wu and Gilbert [41]:
fct = 2.0 MPa (with 2% randomization), fcp = 21.6 MPa, Ec =
22,400 MPa, fsy = 540 MPa, and Es = 200,000 MPa. The basic
parameters for the CEBFIP bond stressslip relationship are in
accordance with the code for good confinement conditions, i.e. =
0.4, max = 11.6 MPa and s1 = 1.00 mm and typical values
for the parameters that define the tensile stress strain curve for
concrete (Fig. 3(a)) as recommended by Peterson [19] are assumed,
i.e. 1 = 0.333, 2 = 50, 3 = 230.

2386

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391

(a) STN12.

(b) STN16.
Fig. 8. Experimental and numerical load versus average strain curves.

(a) STN12.

(b) STN16.
Fig. 9. Tension stiffening factor versus average strainexperimental vs. numerical.

The Modified Newton Raphson solution scheme [40] is used


in the analysis to capture the nonlinear behavior of the specimen
and to save computational time. The tangent stiffness matrix is
formed at the beginning of each load step. Because this study is
concerned with the short-term tension stiffening effect under inservice conditions (i.e. when s is less than about 400 MPa), the
analysis was terminated when the steel stress reached 500 MPa.
5. Finite element analysis results
Fig. 8 compares the numerical and experimental load versus
average strain graphs for STN12 and STN16. The average strain
is calculated from the finite element model by dividing the axial
deformation between the two monitored nodes A and B, as shown
in Fig. 7, by the distance between them. The proposed bond model
provides an excellent correlation with the experimental results,
especially at the cover-control crack stage. With increasing loads
(i.e. with increasing steel stress during the crack stabilization
stage), the finite element model incorporating the proposed bond
model predicts the gradual reduction of tension stiffening that was
observed in the test specimens. By contrast, the CEB bond model
tends to overestimate tension stiffening under increasing loads.
Fig. 9 shows the tension stiffening factor versus average strain
for STN12 and STN16. The tension stiffening factor predicted using
the CEBFIP bond model tends to increase with increasing strain
after all the cracks have formed. Furthermore, this overestimation
of stiffness is more significant in the specimen with the smaller
reinforcement ratio (STN12). In reality, this is not the case. The
tension stiffening factor obtained from the experiments declines

to almost zero as the bare bar steel stress approaches the yield
stress of the steel. The tension stiffening factor predicted using the
proposed bond model shows good correlation with testing results,
with the tension stiffening factors decreasing as the applied load
increases.
Figs. 10 and 11 show the major principal strain contours
predicted by the finite element model for each specimen using the
proposed bond model, together with the variation of steel force,
bond stress and slip along each member at different load levels
during the crack stabilization phase. Slip is a vector quantity, with
direction either positive or negative in the direction of the bar. Also
shown in Figs. 10 and 11, are the measured variations of steel force
in each test specimen at the same two load levels.
In the numerical solution, three primary cracks developed in
the half-length of STN12 at the locations indicated in Fig. 10(a)
and four cracks in the half-length of STN16 as shown in Fig. 11(a).
The numerical model provided excellent agreement with the
experiments in terms of both the number of cracks and the
spacing between them. The principal tensile strains at the crack
locations are much greater than the strain corresponding to the
onset of cracking (about 0.0001). In between the cracks; all the
concrete strains remain in the elastic range (<0.0001) when using
the proposed bond model and the concrete remains uncracked
(intact).
The steel force distributions determined numerically for STN12
and STN16 are shown in Fgs. 10(b) and 11(b), respectively, while
the measured values are given in Figs. 10(c) and 11(c), respectively.
In the experiments, strain gauges were fixed onto the surface of the
deformed bars at 25 mm centers to measure the variation of steel

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391

2387

(a) Tensile strain contours on half sample (FEM) when P = 50 kN.

(b) Steel forcehalf sample (FEM).

(c) Steel forcemiddle 600 mm of sample (experiment).

(d) Sliphalf sample (FEM).

(e) Bond stresshalf sample (FEM).


Fig. 10. Local response of STN12 using the proposed bond model.

Table 1
Measured and calculated maximum crack widths (mm).
STN12
Load (kN)

40.0
50.0

STN16
Steel stress at crack (MPa)

354
442

Maximum crack width (mm)


Experiment

FEM

0.30
0.375

0.26
0.35

strain as loading progressed. The steel force shown in the figures


is obtained directly from the measured strains by multiplying the
local strain by the elastic modulus and the cross section area of
steel bar. To minimize the impact of the strain gauges on the
bond between the steel bar and the concrete, the gauges were
carefully attached to the longitudinal ridge that runs longitudinally
along the bar and any contact between the gauge adhesive and the
transverse bar deformations was avoided.
As the figure illustrates, at the cracks the steel force reaches its
maximum and mid-way between cracks it is at a minimum. The
graphs show that with increasing load the steel force tends to be

Load (kN)

Steel stress at crack (MPa)

Maximum crack width (mm)


Experiment

FEM

76.0
90.0

378
448

0.325
0.375

0.25
0.34

more uniform along the bar (due to the degradation of bond), as


the tension stiffening effect decreases. The variation of steel force
measured during the tests is not smooth (as is predicted by the
numerical model) because the breakdown in bond is quite local at
the location of cover-controlled cracks and this is reflected by the
relatively large differences in strains measured at adjacent strain
gauges. Nevertheless, the variation of steel forces predicted by the
numerical model is in reasonable agreement with the measured
variation. The proposed bond model reduces the average bond
stress with increasing steel stress, without capturing the very
local effect of damage on the bond caused by cover-controlled
cracking.

2388

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391

(a) Tensile strain contours on half sample (FEM) when P = 90 kN.

(b) Steel forcehalf sample (FEM).

(c) Steel forcemiddle 600 mm of sample (experiment).

(d) Sliphalf sample (FEM).

(e) Bond stresshalf sample (FEM).

Fig. 11. Local response of STN16 using the proposed bond model.

With regard to the slip at the steelconcrete interface and the


average local bond stress (Fig. 10(d) and (e) for STN12 and Fig. 11(d)
and (e) for STN16), the proposed bond model gives rise to reduced
bond stress with increasing slip as the load is increased within the
service load range. The maximum bond stress decreases by around
1.77 MPa for STN12 as the load P is increased from 40 kN to 50 kN
and by about 1.37 MPa for STN16 as P increases from 76 kN to
90 kN. The bond stress distribution given by the proposed model
also illustrates a rapid built-up of bond stress from zero at the crack
to its maximum value near the cracks, dropping to zero mid-way
between the cracks.
The concrete tensile stress distributions along STN12 and STN16
calculated using the proposed bond model at different transverse
distances z from the steel bar (at the bar surfacez = 0; at
the concrete surfacez = 50 mm; and mid-way betweenz =
25 mm) at two different load levels are shown in Figs. 12 and 13,
respectively. The tensile concrete stress at all locations reduces as
the applied load increases, i.e. the intact concrete is unloading as
the load increases during the crack stabilization stage. The crack
stabilization stage usually occurs at typical service load levels, with
steel stresses at the primary cracks typically in the range from

Table 2
Average primary crack spacing (mm).
STN12

STN16

Experiment

FEM

Experiment

FEM

195

181

165

145

250 MPa to 400 MPa. For STN12, the maximum tensile stress midway between two primary cracks at the level of the steel bar drops
from 1.9 MPa at an applied load of P = 40 kN (when the steel stress
at a crack s = 354 MPa) to approximately 1.3 MPa at P = 50 kN
(when s = 442 MPa). For STN16, the corresponding tensile stress
drops from 1.35 MPa to 0.61 MPa as the applied load increases from
76 kN (s = 378 MPa) to 90 kN (s = 448 MPa).
Tables 1 and 2 summarize the calculated and experimental
maximum cracks width and average crack spacing. The maximum
crack width was measured at the concrete surface. The crack
width predicted by the numerical model at the steel level can
be obtained simply by adding up the two maximum slips either
side of the primary crack. The slight increase in crack width with
distance z from the steel bar can be obtained by integrating the

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391

2389

(a) At P = 40 kN.

(b) At P = 50 kN.
Fig. 12. Concrete tensile stress distribution calculated using proposed bond model (STN12).

reduction in elastic tensile strain in the concrete (from that at


the level of the bar) over the half crack spacing on either side
of the crack. The proposed bond model provides a reasonable
prediction of maximum crack width and average crack spacing;
slightly underestimating both the observed maximum crack width
and the average crack spacing in both test specimens.
The CEBFIP bond model, therefore, fails to correctly predict
tension stiffening at short-term service loads in both global and
local scale, because it fails to include the effects of concrete damage
and increasing steel stress. The bond stressslip relationship in
practice must be correlated with either concrete damage or steel
stress evolution if an accurate model of tension stiffening is
required over the full service load range.
6. Summary and conclusions
The loss of tension stiffening in reinforced concrete members
under monotonically increasing load is caused by the formation
of primary cracks and the gradually decreasing bond stress under
increasing loads after all the primary cracks have formed. The
bond stressslip relationship at the interface between the concrete
and the steel reinforcement is therefore of paramount importance
when modeling tension stiffening in reinforced concrete. A bond
stressslip relationship based on the CEBFIP bond model suitably
modified to include the effects of concrete damage due to primary

cracking and bond degradation under increasing steel stress has


been proposed and incorporated into an existing finite element
model of reinforced concrete. Two uniaxially loaded tension
members have been analyzed numerically using both the proposed
bond model and the CEBFIP bond model and the numerical results
have been compared with the experimental observations.
The proposed bond model gives good correlation with the
test results at service loads (when steel stresses remain in the
elastic range), while the CEBFIP bond model gives a significantly
stiffer response and tends to overestimate tension stiffening. The
experiments indicate that the tension stiffening factor t decreases
with increasing load after the primary cracks have developed and
this is predicted well using the proposed bond model. In contrast,
the CEBFIP bond model does not predict the reduction of t
with increasing load after the formation of the primary cracks.
The proposed bond model provides reasonable predictions of the
variation of steel stresses, bond stresses and slip between the
concrete and the steel. In addition, the model provides a good
estimation of the width of the primary cracks, as well as the spacing
between them.
This paper has not considered the variation of tension stiffening
with time, which is an important consideration in engineering
practice. Future research will concentrate on the effects of creep
and shrinkage on tension stiffening and the inclusion of these
effects in the proposed bond model.

2390

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391

(a) At P = 76 kN.

(b) At P = 90 kN.
Fig. 13. Concrete tensile stress distribution calculated using proposed bond model (STN16).

References
[1] Fields K, Bischoff PH. Tension stiffening and cracking of high-strength
reinforced concrete tension members. ACI Struct J 2004;101(4):44756.
[2] Gilbert RI, Wu HQ. Time-dependent stiffness of cracked reinforced concrete
elements. In: ASEC 2008 Australian structural engineering conference.
Engineers Australia. Paper No.028; 2008.
[3] Goto Y. Cracks formed in concrete around deformed tension bars. ACI J 1971;
68(4):24451.
[4] Bischoff PH. Effects of shrinkage on tension stiffening and cracking in
reinforced concrete. Can J Civil Eng 2001;28(3):36374.
[5] Kaklauskas G, Gribniak V, Bacinskas D, Vainiunas P. Shrinkage influence on
tension stiffening in concrete members. Eng Struct 2009;31(6):130512.
[6] Lin CS, Scordelis AC. Nonlinear analysis of RC shells of general form. ASCE J
Struct Eng 1975;101(3):52338.
[7] Prakhya GKV, Morley CT. Tensionstiffening and momentcurvature relations
of reinforced concrete elements. ACI Struct J 1990;87(5):597605.
[8] Barros M, et al. Tension stiffening model with increasing damage for reinforced
concrete. Eng Comput (Swansea, Wales) 2001;18(56):75985.
[9] Soltani M, An X, Maekawa K. Computational model for post-cracking analysis
of RC membrane elements based on local stressstrain characteristics.
Eng Struct 2003;25(8):9931007.
[10] Kwak H-G, Kim D-Y. Material nonlinear analysis of RC shear walls subject to
monotonic loads. Eng Struct 2004;26(11):151733.
[11] Ebead UA, Marzouk H. Tension-stiffening model for FRP-strengthened RC
concrete two-way slabs. Materials and Structures/Materiaux et Constructions
2005;38(276):193200.
[12] Stramandinoli RSB, La Rovere HL. An efficient tension stiffening model for nonlinear analysis of reinforced concrete. Eng Struct 2008;30(7):206980.
[13] Yankelevsky DZ, Jabareen M, Abutbul AD. One-dimensional analysis of tension
stiffening in reinforced concrete with discrete cracks. Eng Struct 2008;30(1):
20617.
[14] Gilbert RI, Warner RF. Tension stiffening in reinforced concrete slabs. ASCE J
Struct Eng 1978;104(12):1885900.
[15] Kwak H-G, Song J-Y. Cracking analysis of RC members using polynomial strain
distribution functions. Eng Struct 2004;24(4):45568.

[16] Foster SJ, Gilbert RI. Non-linear finite element model for reinforced concrete
deep beams and panels. UNICIV Report no.R-275. Kensington: School of Civil
Engineering, University of New South Wales; December 1990. p. 113.
[17] Comite Euro-International du Beton. CEB/FIP model code (design code).
Thomas Telford Ltd; 1993.
[18] Foster SJ, Marti P. Cracked membrane model: Finite element implementation.
ASCE J Struct Eng 2003;129(9):115563.
[19] Petersson PE. Crack growth and development of fracture zone in plain concrete
and similar material. Report no.:TVBM-1006. Lund (Sweden): Lund Institute of
Technology; 1981.
[20] Baant ZP, Oh BH. Crack band theory for fracture of concrete. Materiaux et
Constructions, Materials and Structures 1983;16(93):15577.
[21] Vecchio FJ, Collins MP. Modified compression-field theory for reinforced
concrete elements subjected to shear. ACI J 1986;83(2):21931.
[22] Vecchio FJ, Collins MP. The response of reinforced concrete to in-plane shear
and normal stresses. Toronto: Department of Civil Engineering, University of
Toronto; 1982.
[23] Miyakawa T, et al. Nonlinear behaviour of cracked reinforced concrete plate
element under uniaxial compression. JSCE Proc 1987;378:24958.
[24] Belarbi A, Hsu TCC. Constitutive laws of reinforced concrete in biaxial
tension compression. Report no.:UHCEE 91-2. Houston: Department of Civil
Engineering, University of Houston; 1991.
[25] Thorenfeldt E. et al. Mechanical properties of high strength concrete and
application in design. In: Proceedings, international symposium on utilization
of high strength concrete. 1987.
[26] Collins MP, Porasz A. Shear strength for high strength concrete. Bull. No. 193.
Design aspects of high strength concrete. Comite Euro-International du Beton;
1989. p. 7583.
[27] Untrauer RE, Henry RL. Influence of normal pressure on bond strength. ACI J
Proc 1965;62(5):57786.
[28] Tepfers R. Cracking of concrete cover along anchored deformed reinforcing
bars. Mag Concr Res 1979;31(106):312.
[29] Edwards AD, Yannopoulos PJ. Local bond-stress to slip relationships for hot
rolled deformed bars and mild steel plain bars. ACI J 1979;76(3):40520.
[30] Eligehausen R. et al. Local bond stress-slip relationship of deformed bars
under generalised excitations. Report no. UCB/EERC-83/23. Berkeley (CA):
Earthquake Engineering Research Centre, College of Engineering, University
of California; 1983.

H.Q. Wu, R.I. Gilbert / Engineering Structures 31 (2009) 23802391


[31] Giuriani E, et al. Role of stirrups and residual tensile strength of cracked
concrete on bond. ASCE J Struct Eng 1991;117(1):118.
[32] Malvar LJ. Bond of reinforcement under controlled confinement. ACI Mater J
1992;89(6):593601.
[33] Shima H, et al. Micro and macro models for bond in reinforced concrete. J Fac
Eng Univ Tokyo Series B 1987;39(2):13394.
[34] Kankam CK. Relationship of bond stress, steel stress, and slip in reinforced
concrete. J Struct Eng 1997;123(1):7985.
[35] Bolander J, et al. Bond degradation near developing cracks in reinforced
concrete structures. Mem Fac Eng Kyushu Univ 1992;52(4):37995.
[36] Lowes LN, Moehle JP, Goundjee S. Concretesteel bond model for use in finite
element modeling of reinforced concrete structures. ACI Struct J 2004;101(4):
50111.

2391

[37] Soh CK, et al. Damage model based reinforced-concrete element. J Mater Civil
Eng 2003;15(4):37180.
[38] Maekawa K, et al. Nonlinear mechanics of reinforced concrete. New York: Spon
Press; 2003.
[39] Gilbert RI, Nejadi S. An experimental study of flexural cracking in reinforced
concrete members under short term loads. UNICIV Report no. R-434. Sydney:
School of Civil and Environmental Engineering, The University of New South
Wales; 2004. p. 47.
[40] Foster SJ. An application of the arc length method involving concrete cracking.
Int J Numer Methods Eng 1992;33(2):26985.
[41] Wu HQ, Gilbert RI. An experimental study of tension stiffening in reinforced
concrete members under short-term and long-term loads. UNICIV Report R449. University of NSW; 2008.

You might also like