Bond-Slip Model for Detailed Finite-Element Analysis

of Reinforced Concrete Structures

Juan Murcia-Delso 1 and P. Benson Shing, M.ASCE 2

Abstract: A new interface model to simulate the bond-slip behavior of reinforcing bars is presented. The model adopts a semiempirical
Downloaded from ascelibrary.org by UTEP LIBRARY-SERIALS on 08/18/14. Copyright ASCE. For personal use only; all rights reserved.

law to predict the bond stress-versus-slip relations of bars, accounting for the bond deterioration caused by cyclic slip reversals, the
tensile yielding of the bars, and the splitting of concrete. The wedging action of the ribs is represented by assuming that the normal stress
of the interface is proportional to the bond stress. The model has been implemented in a finite-element analysis program and has been
validated with laboratory experiments that include monotonic and cyclic bond-slip and anchorage tests of bars with different embedment
lengths and a test on an RC column subjected to cyclic lateral loading. The model is easy to calibrate and computationally efficient,
and it accurately predicts the bond-slip behavior of bars embedded in well-confined concrete. It also simulates bond failure attributable to
the splitting of concrete in an approximate manner. DOI: 10.1061/(ASCE)ST.1943-541X.0001070. © 2014 American Society of Civil
Engineers.
Author keywords: Bond slip; Interface model; Reinforced concrete; Finite-element analysis; Reinforcing bars; Bond failure; Analysis and
computation.

Introduction the bond-slip behavior with a steel–concrete interface model in

which a constitutive law is used to capture all of the salient fea-
In the finite-element (FE) analysis of reinforced concrete (RC) tures of bond slip. Interface models allow one to study the in-
structures, modeling the bond-slip behavior of reinforcing bars fluence of bond slip on the inelastic behavior of an RC structure
is desirable to properly simulate the crack spacing and widths and the behavior of a bar in the anchorage zone with good
in reinforced concrete, and the stiffness and deformation capabil- accuracy at a reasonable computational cost.
ity of a structure. Bond-slip models are also used to assess the In an interface model, the interaction between the concrete and a
anchorage capacity of a bar, and to determine the minimum de- reinforcing bar is represented by an equivalent bond stress-versus-
velopment and lap-splice lengths for bars. Different approaches relative displacement (slip) relation in the longitudinal direction
with various degrees of sophistications were proposed to simulate of the bar. Typically, the definition of this constitutive relation re-
bond slip using the FE method. They include simple models that lies on experimental data. The bond stress-versus-slip relations of
account for specific effects of bond slip, such as rotational reinforcing bars have been extensively studied experimentally.
springs inserted at the ends of RC members to mimic end rota- These data were used to develop phenomenological bond-slip
tions caused by bar slips in the joint regions (e.g., Cho and laws such as that proposed by Eligehausen et al. (1983). Given
Pincheira 2006), fiber-section frame element models that account the wedging action of the bar ribs, the stresses and relative displace-
for the slip between the reinforcement and concrete fibers ments in the tangential and normal directions of a steel–concrete
(e.g., Monti and Spacone 2000), and detailed continuum models, interface are coupled. However, only limited experimental data
in which the surface deformation of a reinforcing bar is explicitly (e.g., Gambarova et al. 1989; Malvar 1992) on the interaction
considered and represented in the FE mesh (e.g., Salem and between the stresses and displacements in the two directions are
Maekawa 2004). The detailed approach may be used to investi- available.
gate the basic interaction behavior between a reinforcing bar and A number of researchers (e.g., Cox and Herrmann 1998;
the surrounding concrete but is not suitable or practical for study- Lundgren and Gylltoft 2000; Lowes et al. 2004) proposed inter-
ing the behavior of an entire RC structure. This approach is not face models to represent the bond-slip behavior of reinforcing bars
only computationally very demanding but may not necessarily in two- or three-dimensional FE analysis of reinforced concrete.
yield more reliable results unless all pertinent local damage The model of Cox and Herrmann (1998) adopts an elasto-plastic
mechanisms, such as concrete splitting, concrete shearing, and formulation for the stress-displacement relations at the steel–
crushing between bar ribs, are accurately represented. An attrac- concrete interface. This formulation features a yield function that
tive alternative to the detailed continuum models is to simulate accounts for the influence of the normal stress in the bond stress
and a nonassociated flow rule to represent the shear dilatation
1
Postdoctoral Researcher, Dept. of Structural Engineering, Univ. of caused by the wedging action of the bar ribs. The yield func-
California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0085 (corre- tion and flow rule parameters were calibrated using the experi-
sponding author). E-mail: jmurciad@ucsd.edu mental data from Malvar (1992). Their model is only for bond
2
Professor, Dept. of Structural Engineering, Univ. of California,
slip under monotonic loading, even though an ad hoc extension
San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0085.
Note. This manuscript was submitted on October 25, 2013; approved on
for cyclic loading was proposed in a previous version of the
March 6, 2014; published online on July 14, 2014. Discussion period open model (Herrmann and Cox 1994). The model was validated
until December 14, 2014; separate discussions must be submitted for in- using bar pullout tests from different studies, as shown in Cox
dividual papers. This paper is part of the Journal of Structural Engineer- and Herrmann (1999). Lundgren and Gylltoft (2000) proposed
ing, © ASCE, ISSN 0733-9445/04014125(10)/$25.00. another elasto-plastic bond-slip model for three-dimensional FE

© ASCE 04014125-1 J. Struct. Eng.

J. Struct. Eng.
 analysis. Their model was validated using monotonic pullout and
cyclic pull–pull tests, and was applied to study the behavior of
lap splices in beams and exterior beam-column joints and bar
anchorages at beam ends (Lundgren and Magnusson 2001). By
incorporating plastic shear dilatation, both of the aforementioned 2 2
models are able to simulate the wedging action of the bar ribs; 1 s1, σ1 s2 ,τ 2 1
s3 ,τ 3
therefore, they may be used to study bond slip under different 3 3
confinement situations. However, because the elasto-plastic stress
update algorithms require iterations, these models are computa-
tionally demanding. The interface model developed by Lowes
et al. (2004) followed a different approach by adopting a phenom-
enological cyclic bond stress-versus-slip law similar to that pro- Interface
1. Normal 2. Longitudinal 3. Transverse
coordinates
posed by Eligehausen et al. (1983). The model has a basic bond
Downloaded from ascelibrary.org by UTEP LIBRARY-SERIALS on 08/18/14. Copyright ASCE. For personal use only; all rights reserved.

stress-slip relation that is modified by factors to account for the Fig. 1. Stresses and relative displacements at the bar-concrete
effect of the normal confining stress, the concrete cover, the bar interface
spacing, the damage in the surrounding concrete, and the strain in
the reinforcing bar. The normal stress induced by the wedging
action of the bar ribs is calculated as a fraction of the tangential
bond stress by assuming that the resultant bond forces have a fixed at the interface is decomposed into one normal, σ1 , and two
angle of inclination with respect to the longitudinal axis of the shear components, τ 2 and τ 3 . Bar slip and bond stress corre-
bar. This constitutive model is more efficient than the models spond to s2 and τ 2 , respectively. The following sections describe
that utilize plasticity formulations in that iterations are not re- the constitutive relations between the stresses and the relative
quired to calculate the stresses at the interface for a given slip displacements.
displacement. However, to calculate the modification factors to
account for the damage in the concrete surrounding the bar and
the axial strain in the steel, the interface model relies on data from Bond Stress
the neighboring concrete and steel elements. This reliance com-
The basic cyclic bond stress-versus-slip law proposed by Murcia-
plicates the implementation of the interface model in a FE analysis
Delso et al. (2013), which utilizes concepts originally proposed
program. by Eligehausen et al. (1983), was adopted. This law was exten-
This paper proposes a new interface model to simulate the sively verified by bond-slip test data on bars embedded in well-
bond-slip behavior of reinforcing bars. The model is an extension confined concrete. In this study, this law was enhanced to account
of the semiempirical cyclic bond stress-versus-slip law proposed for the additional bond deterioration attributable to the tensile
by Murcia-Delso et al. (2013) to simulate the bond-slip behavior yielding of the bar and the splitting of the concrete around
of reinforcing bars embedded in well-confined concrete. The en- the bar, which occurs in reality. To this end, the bond stress,
hancements introduced include the abilities to simulate the nor- τ 2 , is considered to be a function of the slip, s2 , the normal dis-
mal stress introduced by the wedging action of the bar ribs and placement of the steel–concrete interface, s1 , the axial strain in
to account for the influence of the opening of splitting cracks in the reinforcing bar, εs , and the slip history (i.e., the maximum
the surrounding concrete and bar yielding in the bond strength. slips, sþ −
max and smax , attained in the positive and negative direc-
Similar to the model in Lowes et al. (2004), the normal stress is tions, and the cumulative slip, scum ). In the bond-slip law of
defined as a fraction of the bond stress, and the bond stress- Murcia-Delso et al. (2013), the total bond resistance is separated
slip relations are affected by the steel strain and the level of con- into two components. One component is the bearing resistance
finement around the bar. However, this model does not require and the other component is the friction resistance. Given these
information from the neighboring concrete and bar elements, considerations, the bond resistance for initial loading and for
making its implementation in a general-purpose FE analysis pro- loading beyond the maximum slip values attained in the previous
gram more straightforward. The aforementioned extensions allow cycles is expressed as follows:
the model to simulate bar anchorage failure attributable to con-
crete splitting with reasonable accuracy. The interface model was τ 2 ¼ ρn ðρb;s · ρb;c · τ b þ ρf;s · ρf;c · τ f Þ ð1Þ
implemented in the FE analysis program Abaqus and validated
using experimental data. This paper presents the constitutive where τ b and τ f are the full bearing and friction resistances of an
law for the proposed bond-slip model, the validation examples elastic bar subjected to a monotonic pullout action and are func-
consisting of basic monotonic and cyclic bond-slip and anchorage tions of the slip, s2 ; ρn = a reduction factor that accounts for the
tests with bar pullout and concrete splitting failures, and the opening of splitting cracks in concrete and is a function of s1 ; ρb;s
anchorage behavior of column reinforcement extended into an and ρf;s account for the reduction of the bearing and friction re-
RC footing. sistances, respectively, given the yielding of the bar in tension and
are, therefore, functions of the bar strain εs ; and ρb;c and ρf;c
account for the loading-history-dependent bond deterioration
Bond-Slip Interface Model and are functions of sþ −
max , smax , and scum .
Fig. 2(a) plots the monotonic stress-slip curves for the bearing,
Fig. 1 defines the relative displacements and stresses at the interface friction, and total bond resistances. The curves for the bearing and
attributable to the slip between a reinforcing bar and the surround- friction resistances are defined piecewise using polynomial func-
ing concrete. The relative displacement has three components: tions, as shown in Eq. (2), in terms of three governing parameters,
one normal to the bar surface, s1 , and two tangential to the bar namely, the peak bond strength (τ max ) of an elastic bar, the slip at
surface, with s2 in the longitudinal direction of the bar and s3 in which the peak strength is attained (speak ), and the clear spacing
the transverse direction of the bar. Likewise, the stress transfer between the bar ribs (sR )

© ASCE 04014125-2 J. Struct. Eng.

J. Struct. Eng.
 8 τ
>
> 3.0 max s for 0 ≤ s2 < 0.1speak The use of these equations requires the determination of the
>
> speak 2
>
> values of the three governing parameters. The value of sR is a
>
>
  
>
> s2 −speak 4 known geometric property of the bar and is usually between 40
>
> τ 0.75− 0.45 for 0.1speak ≤ s2 < speak
>
< max 0.9s
to 60% of the bar diameter. As discussed in Murcia-Delso et al.
τ b ðs2 Þ ¼
peak (2013), τ max and speak depend on many factors and no theoretical
>
> 0.75τ max for speak ≤ s2 < 1.1speak formulas are available to accurately predict their values. These
>
>

>
> s2 −1.1speak values have to be determined experimentally for each case, if
>
> 0.75τ 1− for 1.1speak ≤ s2 < sR
>
> max possible. When no experimental data are available, the following
>
> sR −1.1speak
>
: empirical formulas proposed in Murcia-Delso et al. (2013) are used
0 for s ≥ sR to estimate the values of these two parameters
ð2aÞ τ max ¼ 1.163fc03=4 ðin MPaÞ ð3aÞ

speak ¼ 0.07db ð3bÞ

Downloaded from ascelibrary.org by UTEP LIBRARY-SERIALS on 08/18/14. Copyright ASCE. For personal use only; all rights reserved.

8
> τ max where f c0 = compressive strength of concrete and db = bar diameter.
>
> s2 for 0 ≤ s2 < 0.1speak
>
> s The formulas in Eq. (3) utilize experimental data. For a vertically
>
< peak
   cast bar pulled downward, the bond strength and the bond stiffness
τ f ðs2 Þ ¼ s2 − speak 4
>
> τ max 0.25 − 0.15 for 0.1speak ≤ s2 < speak are reduced. Eq. (4) uses the data in Murcia-Delso et al. (2013)
>
> 0.9speak and Lundgren (2000) to describe the monotonic bearing and fric-
>
>
: 0.25τ for s2 ≥ speak tion bond stress-slip relation for a vertically cast bar pulled down-
max
ward. For a bar pulled downward, the slip and bond stress have a
ð2bÞ negative sign

8 τ
>
> −1.6 max js2 j for − 0.15speak ≤ s2 < 0
>
> speak
>
>
 
>
> js2 j − 1.5speak 4
>
> −τ 0.6 − 0.36 for − 1.5speak ≤ s2 < −0.15speak
< max
1.35speak
τ b ðs2 Þ ¼ ð4aÞ
>
> −0.6τ for − 1.6speak ≤ s2 < −1.5speak
>
>
max

>
> js2 j − 1.6speak
>
> −0.6τ 1 − for − sR ≤ s2 < −1.6speak
>
>
max
sR − 1.6speak
:
0 for s2 < −sR

8
>
> 2τ
>
> − max js2 j for − 0.15speak ≤ s2 < 0
>
> 3 speak
<
  
τ f ðs2 Þ ¼ js2 j − speak 4 ð4bÞ
>
> −τ max 0.25 − 0.15 for − 1.5speak ≤ s2 < −0.15speak
>
> 1.35speak
>
>
: −0.25τ for s2 ≤ −1.5speak
max

8
The curves given by Eqs. (2) and (4) were calibrated by Murcia- > 1 for s1 ≤ 0.5hR
<
Delso et al. (2013) for bars embedded in well-confined con- ρn ðs1 Þ ¼ 2ð1 − s1 =hR Þ for 0.5hR < s1 ≤ hR ð5Þ
crete and bond failing from pullout of the bar from the concrete. >
:
However, a bond may also fail prematurely from the splitting of 0 for s1 > hR
the concrete. Bar slip induces a radial compressive stress against
the surrounding concrete given the wedging action of the bar ribs. The dependence on the rib height utilizes the notion that the
In turn, this radial stress induces tensile hoop stresses in the con- loss of bond resistance is related to a reduction in the contact sur-
crete and transverse reinforcement. With insufficient concrete cover face between the ribs and the concrete. For well-confined situa-
and transverse reinforcement, the hoop stress may be severe enough tions, the interface opening may be very small; consequently, ρn
to cause radial tensile splitting cracks in the concrete, which se- will be equal to one. If the interface opening is larger than the
verely jeopardizes the bond resistance. The severity of the splitting rib height, the contact between the ribs and the concrete is lost
cracks is related to the amount of radial dilatation, which may and the bond resistance disappears (ρn ¼ 0). Eq. (5) follows a
be quantified in the model as the degree of opening of the steel– simple rule to establish a smooth transition between these two sit-
concrete interface. The factor ρn in Eq. (1) is used to account for the uations. The selection of 0.5hR as the critical interface opening,
bond deterioration caused by the opening of the splitting cracks. after which the resistance will drop, is an assumption that seems
Given these considerations, this factor is assumed to be a function to provide satisfactory results.
of the normal displacement at the interface, s1 , with respect to the Bond resistance is considerably reduced after a reinforcing bar
rib height, hR , as follows: yields in tension attributable to the plastic contraction of the cross

© ASCE 04014125-3 J. Struct. Eng.

J. Struct. Eng.
 τmax
Total bond resistance

Bond stress

Bond stress
τrev
Bearing
bond resistance
Friction Cyclic response
−τrev
bond resistance Monotonic envelope
0.25 τmax Reduced envelope

O
speak sR
(a) Slip (b) Slip
Downloaded from ascelibrary.org by UTEP LIBRARY-SERIALS on 08/18/14. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Bond stress-versus-slip law: (a) monotonic response; (b) cyclic response

section. In pullout tests carried out by Shima et al. (1987) on bars and sþ −
max and smax = maximum slips in absolute values achieved in
with long embedment lengths, the bond resistance dropped to the positive and negative directions, respectively. As shown in
approximately 25% of the peak bond strength right after the bar Eq. (8), the reduction in the friction resistance is a function of both
yielded in tension, and it continued to decrease gradually as the the absolute maximum slip attained in each loading direction and
inelastic deformation of the bar increased. Two deterioration fac- the cumulative slip
tors, ρb;s and ρf;s , are introduced in Eq. (1) to account for this phe-  þ  
nomenon. These factors utilize the assumption that, once the bar smax þ s− 0.75
ρf;c ¼ 1 − min max
; 1 1 − e−0.45ðscum =sR Þ ð8Þ
yields in tension, the friction resistance remains intact and the bear- sR
ing bond resistance decreases linearly and reaches zero when the
bar strain reaches εsh , which is the point at which strain hardening where scum is considered zero before the slip displacement exceeds
starts. Assuming that εsh is equal to 0.01 is appropriate. For larger the slip at the peak stress, speak , for the first time.
strains, the friction resistance decreases linearly and reaches zero Fig. 2(b) shows the analytical bond stress-versus-slip relations
when the bar strain reaches the ultimate strain of the steel, εu , which for cyclic loading. For loading beyond previously attained slip
is assumed to be 0.15. If the bar remains elastic, no reduction is levels, the bond stress is determined by the expression in Eq. (1).
applied. The Poisson effect of a bar subjected to elastic tensile When the bar slips between previously attained slip levels, the fol-
deformation also causes a small reduction in the cross section; lowing unloading and reloading rules apply. As shown in Fig. 2(b),
however, this elastic contraction is so small that its effect is often right after each slip reversal, unloading follows the initial stiffness
negligible. Regardless, this effect is partially accounted for in the of the curve defined in Eq. (1) until the friction resistance limit
basic bond stress-versus-slip law presented in Eq. (4) because it is τ rev in the opposite direction is reached. If the maximum slip ever
calibrated using bond-slip tests that subjected the bars to elastic achieved exceeds the slip at the peak resistance, speak , the resistance
tensile deformation. Given these considerations, the deterioration τ rev right after slip reversal is equal to a reduced residual friction,
factors are expressed as follows: as shown in Eq. (9a)
8 τ rev ¼ 0.25 · ρn · ρf;y · ρf;c · τ max ð9aÞ
>
> 1 for εs ≤ εy
>
<ε −ε
ρb;s ðεs Þ ¼
sh s
for εy < εs ≤ εsh ð6aÞ Otherwise, it is a fraction of that reduced friction as shown in
>
> ε − ε Eq. (9b)
>
:
sh y
0 for εs > εsh
τ rev ¼ krev · 0.25 · ρn · ρf;y · ρf;c · τ max ð9bÞ

8 where
<1 for εs ≤ εsh
ρf;s ðεs Þ ¼ εu − εs ð6bÞ maxðsþ −
max ; smax Þ
: for εs > εsh krev ¼ ≤1 ð9cÞ
εu − εsh speak

where εs = tensile strain of the bar. At slip reversal, the bond stress is equal to τ rev until the maxi-
The reduction factors for the cyclic deterioration of bond, ρb;c mum slip previously attained in that direction (sþ −
max or smax ) is
and ρf;c , utilize the original model of Murcia-Delso et al. (2013). reached. After this point, the bond stress-slip relation is governed
These two factors are defined in Eqs. (7) and (8). For a detailed by the curve defined in Eq. (1), as illustrated by the reloading curve
explanation of their physical meaning and their calibration for in the positive direction in Fig. 2(b). The linear transition from the
cyclic loading, the reader is referred to their paper. As shown in horizontal line (τ ¼ τ rev ) to the loading curve follows the initial
Eq. (7), the reduction in the bearing resistance is a function of the stiffness of the curve in Eq. (1). However, if at the maximum slip
maximum slip attained in each direction previously attained, τ rev is larger than the bond stress given by
Eq. (1), the bond stress is equal to τ rev until the curve in Eq. (1)
0.8
ρb;c ¼ 1.2e−2.7ðs̄max =sR Þ ≤1 ð7aÞ is intersected. This intersection is illustrated by the curve in the
negative direction in Fig. 2(b). The unloading and reloading rules
where proposed utilize the work of Eligehausen et al. (1983), but the val-
ues of τ rev and the unloading and reloading stiffness are calculated
s̄max ¼ 0.75 maxðsþ − þ −
max ; smax Þ þ 0.25ðsmax þ smax Þ ð7bÞ differently to provide better correlation with experimental data.

© ASCE 04014125-4 J. Struct. Eng.

J. Struct. Eng.
 Normal Stress a finite-element model, the interface element should not be smaller
than the bar rib spacing.
To simulate the wedging action of the bar ribs, the resultant bond
force is assumed to have a fixed angle of inclination θ with respect
to the longitudinal axis of the bar. Hence, the normal stress is Model Calibration
proportional to the bond stress with the proportionality constant The model presented requires the calibration of only three param-
determined by the angle θ. This assumption was used in the eters, which are τ max , speak , and sR . Setting the inclination angle of
bond-slip model proposed by Lowes et al. (2004). In addition, a the bond force, θ, to 60° is recommended. Whereas the resultant
penalty stiffness, K pen;1 , which is active only in compression, bond resistance is commonly assumed to have a 45° angle with re-
was introduced to minimize the interpenetration of the steel and spect to the longitudinal axis of the bar (Cairns and Jones 1996),
concrete. Hence, the total normal stress is given by the following Tepfers and Olsson (1992) observed from pullout tests that this
equation: angle varies between 35 and 65°, depending on the rib geometry
σ1 ¼ −jτ 2 j tan θ þ K pen;1 minðs1 ; 0Þ ð10Þ and the intensity of the bond force. The use of a 60° angle is justified
Downloaded from ascelibrary.org by UTEP LIBRARY-SERIALS on 08/18/14. Copyright ASCE. For personal use only; all rights reserved.

on the basis of the validation examples described in the next section.

Transverse Tangential Stress
The rotation of the bar around its longitudinal axis is restrained by Model Validation
a penalty stiffness, K pen;3 , as presented in the following equation:
The ability of the new model to capture the bond-slip behavior
τ 3 ¼ K pen;3 s3 ð11Þ of bars in RC structures in a FE analysis was validated using the
experimental results obtained from various tests reported in the
Interface Element Implementation literature. The experiments include basic monotonic and cyclic
bond-slip and anchorage tests with different embedment lengths
The bond-slip relations presented in the previous sections serve as a
and confinement levels, and a test on an RC column subjected
constitutive model of an interface element that connects truss or
to lateral loading. To simulate the behavior of concrete and steel,
beam elements representing reinforcing bars to the surrounding
constitutive models available in Abaqus were used. Concrete was
concrete, as shown in Fig. 3. To evaluate the stress for a given
modeled with solid elements and a plastic-damage constitutive
relative displacement, the value of the axial strain in the bar is re-
model based on the formulations proposed by Lubliner et al.
quired in Eq. (6). This strain may be calculated inside the interface
(1989) and Lee and Fenves (1998). Steel reinforcement was mod-
element from the displacements parallel to the bar axis at the nodes
eled using truss elements and a rate-independent elasto-plastic law
connected to the bar (nodes A and B in Fig. 3) and the length of
with nonlinear kinematic hardening. The only exception is the lon-
the element, Le , as
gitudinal reinforcement of the RC column, which was modeled
u2B − u2A with beam elements. In this case, a linear kinematic hardening
εs ¼ ð12Þ
Le law was employed because the nonlinear hardening option is not
available for beam elements. The calibration of these constitutive
A new interface element that allows the computation of εs with models is described in Murcia-Delso (2013). Table 1 presents the
Eq. (12) was implemented in the FE analysis program Abaqus in values of the concrete and steel strengths and the input parameters
a user-defined subroutine. The element has four nodes (with two for the bond-slip model (τ max , speak , and sR ) used in the analyses.
nodes connected to the bar and two connected to the concrete), The inclination angle of the bond force, θ, is assumed to be 60°
a linear shape function, and two integration points at its ends. The unless indicated otherwise.
force per unit length of the interface is obtained by multiplying the The accuracy of the basic bond stress-versus-slip law in repre-
interface stresses by the tributary perimeter of the bar that the inter- senting the monotonic and cyclic bond deterioration was already
face element represents. verified at the constitutive level in Murcia-Delso et al. (2013)
The stress-displacement relation is assumed to represent the for elastic bars of different sizes, which were embedded in well-
average response contributed by the bearing and frictional force confined concrete of different strengths. The analyses presented
mechanisms that occur between two consecutive ribs. Hence, in are intended to examine the additional features of the bond-slip
model presented in this paper, including factors that account for
the effect of bar yielding and concrete splitting on the bond resis-
tance, and to examine the capability of the interface model to
accurately capture nonuniform bar slip and bond deterioration in
FE models of RC structures.

Basic Bond-Slip Tests

One monotonic pullout test and one cyclic pull–pull test conducted
by Murcia-Delso et al. (2013) on No. 14 (43-mm) bars embedded in
well-confined concrete cylinders were analyzed. The test speci-
mens had the bars bonded only in the midheight region of the cyl-
inder over a length of five times the bar diameter, db , to provide a
fairly uniform bond-stress distribution and to prevent the yielding
of the steel to enable the fundamental bond stress-versus-slip rela-
tion to be obtained. However, given the high compressive strength
of the concrete, which was 55 MPa (8 ksi) in the specimens
analyzed, the bars yielded in tension during the monotonic test.
Fig. 3. Interface element
The specimen subjected to cyclic loading did not have bar yielding.

© ASCE 04014125-5 J. Struct. Eng.

J. Struct. Eng.
 Table 1. Material and Bond Properties for Concrete, Steel, and Bond-Slip Models
Bar diameter and material properties Bond properties
Test db mm (in.) f y MPa (ksi) f u MPa (ksi) fc0 MPa (ksi) τ max MPa (ksi) speak mm (in.) sR mm (in.)
a a
Murcia-Delso et al. (2013) 43 (1.69) 493 (71.5) 655 (95) 55.2 (8) 23.8 (3.45) 3.0 (0.12) 24.9b (0.98)
Lundgren (2000) 16 (0.63) 500 (72.5) 800 (116) 36 (5.2) 20.6a (3.0) 1.0c (0.04) 8.0d (0.31)
Plizzari and Metelli (2009) 40 (1.57) 524 (76) 663 (96) 37.6 (5.45) 17.65c (2.56) 2.8c (0.11) 20.0d (0.79)
Shima et al. (1987) 19.5 (0.77) 610 (88.5) 800 (116) 19.6 (2.84) 10.8c (1.57) 1.36c (0.054) 9.75d (0.38)
Viwathanatepa et al. (1979) 25.4 (1.0) 469 (68) 731 (106) 32.6 (4.72) 15.86c (2.3) 1.78c (0.07) 12.7d (0.5)
Lehman and Moehle (2000) (column) 16 (0.63) 472 (68.4) 643 (93.3) 30.3 (4.4) 15.0c (2.18) 1.12c (0.044) 8.0d (0.31)
Lehman and Moehle (2000) (footing) 33.8 (4.9) 16.3c (2.36)
a
Determined from monotonic bond-slip test.
b
Measured.
Downloaded from ascelibrary.org by UTEP LIBRARY-SERIALS on 08/18/14. Copyright ASCE. For personal use only; all rights reserved.

c
Estimated using Eq. (3).
d
Estimated as 0.5db .

Fig. 4(a) shows the FE model of the test specimens. Only one stress is significantly reduced near the loaded end given the pen-
quarter of the specimen was modeled to take advantage of the axial etration of the plastic strain. As shown in Fig. 4(d), good correlation
symmetry of the test specimens. As shown in Fig. 4(b), the FE between the test and analysis results is also obtained for the pull
model is capable of predicting the yielding of the bar and the pull capacity and force-versus-displacement relations for the cyclic test.
capacity in the monotonic test with very different displacements The drop of the bond capacity attributable to cyclic loading, which
(slips) at the pulled (top) end and the free (bottom) end. The dis- prevented bar yielding in this test, is well captured by the model.
placements shown are measured at the top and bottom of the To evaluate the accuracy of the model in simulating the splitting
concrete specimen, including the bar deformation in the unbonded action caused by bar slip, bond-slip tests carried out by Lundgren
region. The differences in the force-versus-displacement curves (2000) and Plizzari and Metelli (2009) were studied using FE
from the test and the analysis are attributed to the absence of a yield models. Lundgren (2000) conducted pullout tests on No. 5
plateau before strain hardening in the steel model. In Fig. 4(c), the (16-mm) bars bonded over a length of 3db in concrete cylinders
FE model shows that before the bar yields in tension, the bond confined by thin steel tubes on the exterior. In these tests, the steel
stress distribution is practically uniform. Once the bar yields, bond tubes provided sufficient confinement to ensure that failure was

pull force
fixed
condition

Displacement (in.)
unbonded 0 0.5 1 1.5
region 800 bar yield strength
bond-slip
elements 150

Pull force (kip)

600
Pull force (kN)

(1/4 bar bonded

perimeter) region
100
400 u (test)
top
unbonded
u (test)
region 200
bottom
50
u (FEA)
top
ubottom (FEA)
0 0
symmetry 0 10 20 30 40
planes 1/4 bar Displacement, u (mm)
(a) (b)
Bond stress (ksi) Displacement (in.)
0 1 2 3 −1 0 1
Distance from the top end (db)

0
−1 500 100
Pull force (kip)
Pull force (kN)

−2
−3 0 0

−4
u =3 mm (prior to bar yielding) u (test)
top
−500
top −100
−5 u =12 mm (after bar yielding)
top
u (FEA)
top

0 5 10 15 20 25 −40 −20 0 20 40
Bond stress (MPa) Displacement, u (mm)
(c) (d)

Fig. 4. FE analysis of bond-slip tests by Murcia-Delso et al. (2013): (a) FE model of a test specimen; (b) force versus displacement for Test 2 of
Series 4; (c) bond stress distributions for Test 2 of Series 4; (d) force versus displacement for Test 4 of Series 4

© ASCE 04014125-6 J. Struct. Eng.

J. Struct. Eng.
 Slip (in.)
0 0.1 0.2 0.3 0.4
60 60
Test
FEA (θ=45 deg.)
FEA (θ=60 deg.) 10 10

Force (kip)
Force (kN)

Force (kip)
40 40

Force (kN)
20 5 20 5

Test FEA (θ=45 deg.) FEA (θ=60 deg.)

0 0 0 0
0 2 4 6 8 10 0 0.5 1 1.5
(a) Slip (mm) (b) Strain x 10
−3

Fig. 5. FE analysis of a bond-slip test by Lundgren (2000): (a) force-slip relations; (b) hoop strain in steel casing
Downloaded from ascelibrary.org by UTEP LIBRARY-SERIALS on 08/18/14. Copyright ASCE. For personal use only; all rights reserved.

the result of the pullout of the bars from the concrete, and no split- increases the chance of producing a splitting failure and, although
ting cracks were observed. To quantify the splitting force intro- it leads to a slightly worse result for the test of Lundgren (2000),
duced by bar slip, the hoop strains in the tubes were measured the numerical result is still acceptable.
during the tests. Plizzari and Metelli (2009) conducted pullout tests
on large-diameter bars embedded in unreinforced cubic concrete
blocks of side dimensions equal to 10db with bonded lengths equal Anchorage Tests
to 5db . The test specimens in this study failed abruptly given the Shima et al. (1987) conducted pullout tests on 19.5-mm (0.77-in.)
splitting failure of the concrete. Fig. 5 compares the FE analysis bars with an embedment length equal to 50db to study the bond
results with the experimental results for one of the tests conducted characteristics of bars in the post-yield regime. A FE analysis was
by Lundgren (2000). As shown, the FE model well captures the conducted on one of their test specimens. The model is presented
force–displacement relation at the pulled end of the bar. The in Fig. 7(a). The long embedment of the bar enables a coarser mesh
numerical and experimental values for the hoop strain in the steel to be used in the lower portion of the embedment region, in which
tube are in good agreement when the inclination angle of the bond smaller slip variations are expected. Fig. 7(b) compares the numeri-
forces, θ, is 45°. If θ is equal to 60°, the force–displacement relation cal and experimental results on the force-displacement relation at
remains the same, but the hoop strain in the steel tube increases the loaded end of the bar. The numerical result matches the exper-
significantly. Although larger hoop strains are observed in the con- imental data well in terms of initial stiffness, displacement at bar
crete, the steel tube restrains splitting cracks from opening and the yield, and post-yield stiffness. In addition, good correlation was
bond resistance is not affected. Fig. 6 compares the analysis and also obtained for the strain distribution along the anchorage length
experimental results for the tests conducted by Plizzari and Metelli of the bar, as shown in Fig. 7(c). The small differences in the post-
(2009) on 40-mm (1.57-in.) diameter bars. In these tests, bond fail- yield stiffness, as shown in Fig. 7(b), and in the strains near the
ures were caused by the splitting of the concrete because of the lack pulled end of the bar, as shown in Fig. 7(c), are attributed to the
of transverse reinforcement. As shown in Fig. 6(a), the sudden absence of a yield plateau in the steel model. Some differences also
anchorage failure caused by the splitting of the concrete is captured exist between the bond stress distribution obtained from the FE
in the analysis when θ is equal to 60°. Fig. 6(b) plots the maximum model and that estimated by Shima et al. (1987) from experimental
principal strain distribution in the concrete at the peak load in the strain measurements, as shown in Fig. 7(d). They calculated the
analysis. The vertical bands of large strains in this plot indicate that bond stresses from the axial stress variation along the bar using
a splitting crack has formed. However, if θ is 45°, a higher bond strain gage readings and an analytical stress-strain relation. Both
strength is obtained and the bond fails from the pullout of the bar the FE model and the test results show that the bond stress distri-
from the concrete, as shown in Fig. 6(a). In conclusion, the model is bution along the bar is strongly nonuniform. Before bar yielding,
capable of approximately reproducing the radial dilatation pressure the peak bond stress is developed at the loaded end of the bar, and
in the steel–concrete interface. On the basis of these results, θ is the bond resistance drops to practically zero at the unloaded end.
recommended to be equal to 60°. This conservative assumption As plastic strains penetrate into the embedment zone, the bond

Displacement (in.)
0 0.05 0.1 0.15 0.2 0.25 0.3
25
bar pull−out 3
Bond stress (MPa)

20
Bond stress (ksi)

15
2
10 splitting of concrete
Tests 1
5 FEA (θ=45 deg.)
FEA (θ=60 deg.)
0 0
0 2 4 6 8
(a) Displacement (mm) (b)

Fig. 6. FE analysis of a bond-slip test by Plizzari and Metelli (2009): (a) bond stress versus slip at unloaded end; (b) maximum principal strains in
concrete at peak load (θ ¼ 60°)

© ASCE 04014125-7 J. Struct. Eng.

J. Struct. Eng.
 symmetry
planes

unbonded
pulled end Displacement (in.)
length
0 0.05 0.1 0.15 0.2
1/4 bar 800
100

Bar stress (MPa)

600

Bar stress (ksi)

80
embedment
Test
bond-slip elements length 400 60
FEA
(1/4 bar perimeter) 40
200
20
Downloaded from ascelibrary.org by UTEP LIBRARY-SERIALS on 08/18/14. Copyright ASCE. For personal use only; all rights reserved.

0 0
0 1 2 3 4 5 6
(a) (b) Displacement, u (mm)
fixed base
0.05 12 Test (u = 4.3 mm, after yield)
FEA (u = 0.75 mm, before yield) 1.5
0.04 10 FEA (u = 4.3 mm, after yield)

Bond stress (MPa)

Bond stress (ksi)

0.03 8
Strain

Test 1
FEA 6
0.02
4
0.01 0.5
2
0
0 10 20 30 40 50 0 0
Distance from loaded end (d b) 0 10 20 30 40 50
(c) (d) Distance from loaded end (d b)

Fig. 7. FE analysis of an anchorage test by Shima et al. (1987): (a) FE model of the test; (b) bar stress versus displacement at the loaded end; (c) strain
distribution along the bar at a displacement of 4.3 mm (0.17 in.); (d) bond stress distributions along the bar

symmetry
plane
bond-slip elements
(1/2 bar perimeter)
Displacement (in.)
0 0.2 0.4 0.6 0.8 1 1.2
800 tensile strength of steel

free end 100

Bar stress (MPa)

600
Bar stress (ksi)
pulled end 80
bar pull−out

400 60
1/2 bar
40
200
Test 20
FEA
0 0
0 5 10 15 20 25 30
Displacement (mm)
(a) (b)
Displacement (in.)
0 0.02 0.04 0.06 0.08 0.1 0.12
800 tensile strength of steel
100
Bar stress (MPa)

600
Bar stress (ksi)

80
bar pull−out

400 60
40
200
Test 20
FEA
0 0
0 0.5 1 1.5 2 2.5 3
(c) Displacement (mm)

Fig. 8. FE analysis of an anchorage test by Viwathanatepa et al. (1979): (a) FE model of the test; (b) bar stress versus displacement at the loaded end;
(c) bar stress versus displacement at the unloaded end

© ASCE 04014125-8 J. Struct. Eng.

J. Struct. Eng.
 becomes weaker and the peak bond stress moves away from the this solution is not entirely satisfactory in that the cohesive strength
loaded end. The FE model predicts a higher peak bond stress of the concrete is completely ignored at the contact interface, the
and more rapid bond decay in the middle portion of the embedment effect is not significant because cracks are expected to develop
length than the experimental bond stress distribution. early at the column base.
Fig. 8(a) shows the FE model of a pullout test conducted by As shown in Fig. 9(b), the FE model provides a good prediction
Viwathanatepa et al. (1979) on a No. 8 (25-mm) bar anchored of the lateral load capacity and the lateral force-versus-drift relation
in a concrete block over a length equal to 25db This test was part obtained in the test. The hysteretic behavior is fairly well captured,
of an experimental program to characterize the bond-slip behavior even though the reloading branches are stiffer in the model. This
of bars in beam-column joints. The bars were expected to yield difference is caused by the early resumption of contact in crack
during these tests. Fig. 8(b) plots the bar stress versus the displace- closing. This problem is partially mitigated by the introduction
ment at the loaded end of the bar obtained from the analysis and the of the contact condition at the column base but not at other loca-
test. These two curves show a very good match. The model repro- tions in the column. The load decay observed at the end of the test
duces the pullout failure of the bar observed during the test and was caused by the buckling and subsequent fracture of the vertical
Downloaded from ascelibrary.org by UTEP LIBRARY-SERIALS on 08/18/14. Copyright ASCE. For personal use only; all rights reserved.

accurately predicts the maximum tensile stress developed in the bars at the base of the column, at which a plastic hinge formed.
bar, which was 97% of its ultimate tensile capacity. Fig. 8(c) com- The model predicts the formation of the plastic hinge at the col-
pares the bar slip measured at the free end of the bar during the test umn base but does not simulate bar buckling and fracture. Hence,
and that predicted by the analytical model. Given the small slip the model does not capture the load drop observed in the test.
experienced at this end of the bar, the accuracy of the numerical Fig. 9(c) shows that the FE model provides a good prediction of
prediction is satisfactory. the strain penetration along the development length inside the foot-
ing, indicating that the bond-slip behavior of these bars is well
captured. However, the steel strains at the column-footing interface
RC Column Subjected to Cyclic Lateral Loading are overestimated. These differences are considered acceptable
The bond-slip model was used in the FE analysis of an RC column knowing that a small difference in bar stress produces a large varia-
tested by Lehman and Moehle (2000) using quasi-static lateral tion in strain in the post-yield regime, and the post-yield stress-
loading. Fig. 9(a) shows the FE model of the column. Bond slip strain relation is approximated using a straight line in the model.
is considered in the longitudinal reinforcement, which is modeled A second FE analysis was conducted on the same column as-
using beam elements. A perfect bond is considered for the trans- suming a perfect bond between the longitudinal reinforcement
verse reinforcement, which is modeled using truss elements. As and the concrete. In this case, the bar elements are embedded in
discussed in Murcia-Delso (2013), the concrete model in Abaqus the concrete elements. The two end nodes of a longitudinal steel
is incapable of simulating the closure of cracks in a realistic man- element that crosses the column-footing interface are embedded
ner, leading to a large residual crack opening on unloading. To in concrete elements on the opposite sides of the contact surface.
circumvent this problem, a contact interface was introduced to re- As shown in Fig. 9(b), this change does not affect the lateral load
present cracking at the base of the column in a discrete manner, capacity and the envelope of the lateral force-versus-drift relation
which is where large cracks are expected to occur. Even though for the column. However, with a perfect bond, the hysteresis

symmetry plane
lateral load Top displacement (in.)
−6 −4 −2 0 2 4 6
400 µ=1 µ=3
column
Lateral load (kip)
Lateral load (kN)

200 50

0 0

footing −200 −50

Test FEA FEA (perfect bond)
−400
−20 −10 0 10 20
(a) (b) Top displacement (cm)
Distance from column base (db)

−10

µ=0.88 (test) µ=0.88 (FEA)

−20 µ=1 (test) µ=1 (FEA)
µ=3 (test) µ=3 (FEA)

−30
µ = displacement ductility demand

0 0.005 0.01 0.015 0.02 0.025 0.03

(c) Strain

Fig. 9. FE model of an RC column tested by Lehman and Moehle (2000): (a) FE model of the column; (b) lateral load versus top displacement;
(c) strain of vertical reinforcement embedded in footing

© ASCE 04014125-9 J. Struct. Eng.

J. Struct. Eng.
 loops are less pinched and the maximum resistance in the opposite Cox, J. V., and Herrmann, L. R. (1999). “Validation of a plasticity bond
direction is engaged much sooner when the loading is reversed model for steel reinforcement.” Mech. Cohesive-Frict. Mater., 4(4),
compared with the case with bond slip and the test result. This phe- 361–389.
nomenon is caused by the smaller crack opening and, thereby, Eligehausen, R., Popov, E. P., and Bertero, V. V. (1983). “Local bond
stress—slip relationships of deformed bars under generalized excita-
the earlier closing of the crack at the column base given the inability
tions.” UCB/EERC-83/23, Earthquake Engineering Research Center,
of the longitudinal bars to slip. Univ. of California, Berkeley, CA.
Gambarova, P., Rosati, G. P., and Zasso, B. (1989). “Steel-to-concrete bond
after concrete splitting: Test results.” Mater. Struct., 22(1), 35–47.
Conclusions Herrmann, L. R., and Cox, J. V. (1994). Development of a plasticity bond
model for reinforced concrete, CR 94-001, Naval Facilities Engineering
A new easy-to-calibrate and computationally efficient bond-slip in- Service Center, Port Hueneme, CA.
terface model for the finite-element analysis of RC structures was Lee, J., and Fenves, G. L. (1998). “Plastic-damage model for cyclic loading
presented. This model adopts a semiempirical bond stress-versus- of concrete structures.” J. Eng. Mech., 10.1061/(ASCE)0733-9399
Downloaded from ascelibrary.org by UTEP LIBRARY-SERIALS on 08/18/14. Copyright ASCE. For personal use only; all rights reserved.

slip law that accounts for the bond deterioration caused by cyclic (1998)124:8(892), 892–900.
bar-slip reversals, the tensile yielding of the bar, and the splitting of Lehman, D. E., and Moehle, J. P. (2000). “Seismic performance of well-
confined concrete bridge columns.” PEER Report 1998/01, Pacific
the concrete. Coupling between the stresses and relative displace-
Earthquake Engineering Research Center, Univ. of California, Berkeley,
ments in the tangential and normal direction of the interface attrib- CA.
utable to the wedging action of the ribs was modeled in a simple Lowes, L. N., Moehle, J. P., and Govindjee, S. (2004). “Concrete-steel
manner. The new model was implemented using an interface bond model for use in finite element modeling of reinforced concrete
element in a finite element analysis program. The ability of the structures.” ACI Struct. J., 101(4), 501–511.
model to capture the bond-slip behavior in the finite element analy- Lubliner, J., Oliver, J., Oller, S., and Oñate, E. (1989). “A plastic-damage
ses of RC structures was validated by a number of experimental model for concrete.” Int. J. Solids Struct., 25(3), 229–326.
tests, including bond-slip and bar anchorage tests with different em- Lundgren, K. (2000). “Pull-out tests of steel-encased specimens subjected
bedment lengths and with and without concrete splitting failure, to reversed cyclic loading.” Mater. Struct., 33(7), 450–456.
and a large-scale test of an RC column. The comparison between Lundgren, K., and Gylltoft, K. (2000). “A model for the bond between
concrete and reinforcement.” Mag. Concr. Res., 52(1), 53–63.
the numerical and experimental results showed that the model is
Lundgren, K., and Magnusson, J. (2001). “Three-dimensional modeling
able to accurately predict the bond-slip behavior of bars embedded of anchorage zones in reinforced concrete.” J. Eng. Mech., 10.1061/
in well-confined concrete and capture bond failures caused by the (ASCE)0733-9399(2001)127:7(693), 693–699.
splitting of the concrete. Malvar, J. (1992). “Bond of reinforcement under controlled confinement.”
ACI Mater. J., 89(6), 593–601.
Monti, G., and Spacone, E. (2000). “Reinforced concrete fiber beam
Acknowledgments element with bond-slip.” J. Struct. Eng., 10.1061/(ASCE)0733-9445
(2000)126:6(654), 654–661.
Funding for the research presented in this paper was provided Murcia-Delso, J. (2013). “Bond-slip behavior and development of bridge
by the California Department of Transportation (Caltrans) under column longitudinal reinforcing bars in enlarged pile shafts.” Ph.D.
Contract No. 59A0710. The authors would also like to thank dissertation, Dept. of Structural Engineering, Univ. of California,
San Diego.
Fundacio La Caixa and Fundacion Caja Madrid for the graduate
Murcia-Delso, J., Stavridis, A., and Shing, P. B. (2013). “Bond strength and
fellowships provided to the first author. However, opinions cyclic bond deterioration of large-diameter bars.” ACI Struct. J., 110(4),
expressed in this paper are those of the authors and do not neces- 659–669.
sarily reflect those of the sponsors. Plizzari, G., and Metelli, G. (2009). “Experimental study on the bond
behavior of large bars.” Technical report, Dept. of Civil Engineering,
Architecture and Environment, Univ. of Brescia, Italy.
References Salem, H. M., and Maekawa, K. (2004). “Pre- and post-yield finite element
method simulation of bond of ribbed reinforcing bars.” J. Struct. Eng.,
Abaqus V. 6.10 [Computer software]. Dassault Systemes Simulia Corp., 10.1061/(ASCE)0733-9445(2004)130:4(671), 671–680.
Providence, RI. Shima, H., Chou, L., and Okamura, H. (1987). “Bond characteristics in
Cairns, J., and Jones, K. (1996). “An evaluation of the bond-splitting action post-yield range of deformed bars.” Proc. JSCE, 6(387), 113–124.
of ribbed bars.” ACI Mater. J., 93(1), 10–19. Tepfers, R., and Olsson, P. (1992). “Ring test for evaluation of bond
Cho, J. Y., and Pincheira, J. A. (2006). “Inelastic analysis of reinforced properties of reinforcing bars.” Bond in Concrete—From Research
concrete columns with short lap splices subjected to reversed cyclic to Practice: Int. Conf., Riga, Latvia.
loads.” ACI Struct. J., 103(2), 280–290. Viwathanatepa, S., Popov, E. P., and Bertero, V. V. (1979). “Effects of gen-
Cox, J. V., and Herrmann, L. R. (1998). “Development of a plasticity bond eralized loadings on bond of reinforcing bars embedded in confined
model for steel reinforcement.” Mech. Cohesive-Frict. Mater., 3(2), concrete blocks.” Rep. No. UCB/EERC-79/22, Earthquake Engineering
155–180. Research Center, Univ. of California, Berkeley, CA.

© ASCE 04014125-10 J. Struct. Eng.

J. Struct. Eng.