Finite-Element Bond-Slip Model for Concrete Columns

under Cyclic Loads

Claire Girard1 and Josée Bastien2

Abstract: The hysteretic behavior of reinforced concrete structures depends in large measure on the transfer mechanism between
reinforcing steel and concrete through bond and the resulting slip. This paper presents a finite-element model developed to investigate the
response of reinforced concrete columns subjected to cyclic loading, considering in an explicit way the relative displacement at the
interface. The novel aspect of the model is its capability of modeling three-dimensional effects as concrete confinement, the softening
response of concrete, and to take into account the gradual deterioration of the bond between reinforcing bars and concrete. Numerical
results are compared to experimental results on rectangular bridge columns with lapped starter bars subjected to cyclic loading. The
numerical model reproduces the physical phenomena observed experimentally, as the gradual stress transfer from one bar to the other. The
numerical results show the effect of damage in concrete on the steel stress distribution, as long as the coupled effect of damage in concrete
and bond-slip constitutive law on the general behavior of the structural element.
DOI: 10.1061/共ASCE兲0733-9445共2002兲128:12共1502兲
CE Database keywords: Bridges; Piers; Finite-element method; Bond stress; Slip; Concrete, reinforced.

Introduction of existing structures. Therefore, some analytical models have

been developed and most studies of the hysteretic behavior of
Bridges, by their simplicity, represent vulnerable structures when reinforced concrete members to date rely on the simplifying as-
subjected to strong seismic excitations 共Priestley et al. 1996兲. A sumption of a perfect bond between steel and concrete. Some
prime concern in the retrofitting of bridge columns is the lack of researchers 共Okamura and Kim 2000; Vecchio 2000兲 have taken
adequate confinement in columns designed in North America be- into account the bond properties in finite-element analysis.
fore the 1971 San Fernando Earthquake. In the design of these tall In this paper, a finite-element model is proposed to investigate
columns, lap splicing of the vertical reinforcement is commonly the response of reinforced concrete columns subjected to cyclic
used with starter bars extending from the footing with a typical loads. Numerical results on modeling of bridge columns with
lap length of 20 times the bar diameter. The behavior of concrete lapped starter bars at their base show good agreement with avail-
columns subjected to strong seismic excitations is greatly influ- able experimental results. Therefore, the proposed model could be
enced by the bond quality of the reinforcing bars anchored in the used for modeling as built bridge columns in order to propose
column base and therefore, lapped splices were identified as a suitable strengthening methods to improve their seismic response.
weak link 共Sun et al. 1993a兲. The behavior of these columns is In the following paragraphs, a review of the related work is pre-
characterized by a rapid strength degradation under cyclic load-
sented followed by a description of the finite-element model used
ings because the lap length is recognized to be insufficient for
herein. Presented afterwards, are the results obtained with the
developing the longitudinal bar yield strength 共Chai et al. 1991兲.
modeling of concrete columns with lap splices and a discussion
Since the 1989 Loma Prieta Earthquake, many efforts were
on a comparison with experimental studies.
made to retrofit the deficient bridge columns by using prestressed
tendons, steel, concrete, or a composite material jacketing. These
strengthening methods enhance the flexural ductility and the shear
resistance of the retrofitted structure 共Priestley et al. 1996兲. Physical Mechanism of Bond Degradation
In the last two decades, many research studies in this field
were based on experiments. However, the finite-element method Modeling of structural elements considering the steel-concrete
is recognized to be a powerful and reliable tool for the evaluation relative displacement requires a deep understanding of the physi-
cal phenomena related to the interface deterioration. It is well
1
PhD, Dept. of Civil Engineering, Université Laval, Quebéc PQ, documented that the steel-concrete interface bond includes three
Canada G1K 7P4. components: chemical adhesion, friction, and mechanical inter-
2
Professor, Dept. of Civil Engineering, Université Laval, Quebéc PQ, locking between the bar lugs and concrete 共Lutz and Gergely
Canada G1K 7P4. 1967兲. In a ribbed reinforcing bar, it is principally the steel lugs
Note. Associate Editor: C. Dale Buckner. Discussion open until May resting against concrete that assure the stress transfer.
1, 2003. Separate discussions must be submitted for individual papers. To In a reinforced concrete element as the one shown in Fig. 1, its
extend the closing date by one month, a written request must be filed with
longitudinal reinforcement is subjected to axial stress which is
the ASCE Managing Editor. The manuscript for this paper was submitted
for review and possible publication on October 31, 2000; approved on progressively transferred to the surrounding concrete. This stress
April 9, 2002. This paper is part of the Journal of Structural Engineer- transfer produces a bond stress field around the reinforcing bar,
ing, Vol. 128, No. 12, December 1, 2002. ©ASCE, ISSN 0733-9445/ due to its relative slip to the surrounding concrete. The develop-
2002/12-1502–1510/$8.00⫹$.50 per page. ment of stress in the concrete beyond its tensile strength initiates

1502 / JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2002

 Fig. 3. Typical bond stress-slip relationship

tion. One can see, in order to maintain equilibrium, the bond

stress must exist and appears as soon as the axial steel stress
varies along the bar.
Bond stress-slip relations have been established experimen-
tally under general excitations 共Eligehausen et al. 1983; Pocha-
nart and Harmon 1989兲 and have become the standard way to
Fig. 1. Dimension and reinforcement of test column with lapped represent the mechanical properties of the steel-concrete inter-
starter bars 关adapted from Sun et al. 共1993a兲兴 face. Therefore, pull out tests are performed to identify the shape
of this relationship, where the bond stress is considered uniform
along the specimen and where the slip is defined as the relative
displacement of the bar with respect to concrete.
primary cracks, propagating to the reinforced element surface. The bond stress-slip behavior is due to the evolution of con-
Once these cracks develop, the bond transfer drops rapidly. This crete microcracking around the deformed bar, as shown in Fig. 3.
phenomenon can be seen in Fig. 2, where the results associated to In the ascending branch, the slip is mainly caused by the crushing
a tensile member test are presented. This figure gives, along the of concrete in front of the lugs. First, the bond resistance is
bar, a good general view of the concrete and steel stress distribu- present by adhesion up to Point A, followed by mechanical inter-
tion, the bond stress distribution, and the cumulated slip distribu- locking between the lugs and concrete. The high pressure on con-
crete in front of the lugs induces tensile stresses around the bar,
creating transverse inclined cracks, therefore modifying the re-
sponse of concrete to loading as its stiffness is reduced 共Point B兲.
Inclined compressive forces spreading from the lugs into concrete
causes splitting longitudinal cracks. At Point C, if no confining
reinforcement is provided, the cracks reach the concrete surface
and the bond resistance will drop to zero. If the concrete is well
confined, the load can be increased further. At the maximum bond
resistance 共Point D兲, shear cracks are initiated in the concrete
keys between ribs. With increasing slip, the bond resistance is
reduced because the shear failure is spreading through the con-
crete. At Point E, the only mechanism remaining is the frictional
resistance of concrete along the failure surface.
Various tensile member tests have shown 共Lutz and Gergely
1967; Mirza and Houde 1979兲 that spacing between primary
cracks is governed by the depth of the concrete cover, while the
relative slip increases with the bar diameter but decreases when
the concrete compressive strength increases. In the case of lapped
splices, the stress transfer from one bar to the other causes radi-
ally outward pressures on concrete and splitting cracks along re-
inforcing bars. These splitting cracks generally initiate at the ends
of the splice where the splitting pressure tends to be larger than
elsewhere along the bars. The guidelines for the seismic design of
reinforced concrete lapped splices are based on the works of Lu-
kose et al. 共1982兲 and Sivakumar et al. 共1983兲. These authors
have observed that under severe cyclic excitations, the steel-
concrete interface around the splice deteriorates rapidly.

Analytical Models
Not much literature exists on the development of nonlinear ana-
Fig. 2. Pull-out test from Tassios and Yannapoulos 共1981兲
lytical models that are capable of describing the complex hyster-

JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2002 / 1503

etic behavior of anchored reinforcing bars in concrete under cy- creasing of its modulus during unloading. Both phenomena result
clic excitations, considering the strength loss under gradual bond from cumulative damage where its evolution is quantified with a
deterioration. Monti et al. 共1997兲 presented a summary of past scalar parameter. The ultimate resistance of concrete in compres-
research activities in the development of such analytical models sion is given by the Willam and Warnke 共1975兲 failure envelope.
and identified some limitations. Often, concentrated bond link In tension, the failure criteria is defined in terms of a damage
elements are introduced at the nodes, connecting concrete and parameter. The tension softening of concrete is modeled with a
steel. The bond link element can be represented by two orthogo- smeared crack approach. The ascending branch is linear while the
nal springs, as proposed originally by Ngo and Scordelis 共1967兲. softening branch after cracking is bilinear. The parameters of the
The models are often limited to the steel and interface elastic softening branch are adjusted depending on the finite-element
constitutive law and are not appropriate to model the behavior mesh refinement, so that the cracking energy by volume is pre-
under cyclic loading. Finally, the existing models are not very served 共ACI 1991兲.
practical because their formulation is not appropriate to incorpo- The postcracking behavior of concrete in compression is more
rate in a general purpose finite-element program. complex than in tension, mainly because crushing of the material
Monti et al. 共1997兲 proposed a finite element based on force comes with an important lateral expansion. The slope of the soft-
instead of displacement interpolation functions that has been ening branch is a function of the ultimate strength in compression
implemented in a general purpose finite-element program based f ⬘c and depends on the degree of confinement.
on the stiffness method of analysis. Sun et al. 共1993b兲 developed Numerical developments were necessary to assure appropriate
a structural analysis program based on a finite fiber element modeling of concrete behavior under cyclic loading as the origi-
method to simulate the overall nonlinear response of reinforced nal model 共Bouzaiene 1995兲 was not developed to perform this
concrete bridge columns. The program was shown to be capable specific task. The results obtained with a confined specimen under
of tracing the complete response envelope 共monotonic static load- cyclic load reversals showed that the model allows an adequate
ing兲 obtained experimentally under imposed lateral loads and dis- transition from a compression stress state to a tensile stress state
placements. and reverse 共Girard 2000兲. The specimen unloaded and reloaded
The model proposed here to investigate the response of rein- in the postpeak region exhibits stable hysteresis loops.
forced concrete columns subjected to cyclic loads is based on the
displacement method of analysis and is incorporated in a general Steel and Bond Modeling
purpose finite-element program. The tridimensional finite-element
model described afterwards considers the concrete confinement The element used herein to model the steel-concrete interface in a
effect, the softening response of concrete, the strength loss under tridimensional geometry is an adaptation of a bidimensional ele-
bond deterioration, and the slip of the reinforcing bar relative to ment developed by Ulm 共1994兲. It consists of taking into account,
the surrounding concrete. These novel aspects make it an ideal in an explicit way, the relative displacement at the interface,
model for applications such as reinforced concrete elements under which leads to the formulation of a finite-element bar with an
cyclic loading, where concrete confinement and bond deteriora- additional degree of freedom at each node, expressing the relative
tion have an important effect on the structural response. slip between steel and concrete. This element represents the be-
havior of both the steel and the interface, as a separate constitu-
tive law is assigned to each.
Description of Finite-Element Model The relative slip (␰ r ) is defined as the difference between the
steel (␰ s ) and concrete (␰ c ) displacement
Two principal aspects of the finite-element model proposed herein
␰ r ⫽␰ s ⫺␰ c (1)
are its capability of modeling three dimensional 共3D兲 effects
through the concrete constitutive law and, at the same time, its The steel strain tensor, based on the assumption of small dis-
capability to take into account the gradual deterioration of the placements, has two components: the concrete elongation (␧ c )
bond between reinforcing bars and concrete via the steel and bond and the relative slip between materials (␧ r )
constitutive law. The main aspects of material modeling are pre- 兵 ␧ s其 ⫽ 兵 ␧ c其 ⫹ 兵 ␧ r其 (2)
sented here.
A curvilinear formulation has been adopted for the element with
Concrete Modeling relative slip. It has the advantage of expressing vector and matrix
components directly in the global coordinate system. The relative
The tridimensional concrete model used in this study was devel- slip vector components are oriented along the tangent vector of
oped by Bouzaiene 共1995兲 for the structural analysis of large- the steel fiber. The tangent vector 兵t其 at a specific point along a
scale reinforced and unreinforced concrete structures under differ- curvilinear element of dimension ds is as described by Salençon
ent loading cases 共Massicotte et al. 1996兲. The concrete 共1988兲
constitutive law proposed by Bouzaiene 共1995兲 consists of a hy-

再 冎 再 冎 冐 再再 冎冎 冐
poelastic model, taking into account the loading path and the ⳵x
evolution of mechanical properties according to the stress history. ⳵x ⳵x ⳵␰ ⳵␰
兵t其⫽ ⫽ ⫽ (3)
Concrete is treated as an orthotropic material with a unique stress- ⳵s ⳵␰ ⳵s ⳵x
strain relationship along the principal strain directions. The stress- ⳵␰
strain curve for each axis is based on the concept of the equiva-
lent uniaxial strain model as proposed originally by Darwin and The vectorial force field relation is expressed as
Pecknold 共1977兲. The model has been validated under various ⳵兵N其
stress conditions and histories and has given satisfactory results ⫹ 兵 f s 其 ⫽0 (4)
⳵s
for proportional and nonproportional loading.
The model represents the anisotropic behavior of concrete where f s ⫽vector of smeared forces acting along an element. Eq.
which is characterized by its volumetric expansion and the de- 共4兲 leads to the following equilibrium requirement:

1504 / JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2002

 ⳵N s
⫹ f t ⫽0 (5)
⳵s
where N s ⫽normal force acting in the tangent direction
N s ⫽ 兵 F 其 • 兵 t 其 ⫽A s ␴ (6)
兵F其⫽vector of forces and ␴⫽stress that is assumed constant over
the bar section.
With the kinematically admissible displacement field ␰ *t , the
variational formulation is established with the equilibrium rela-
tion

W s⫽ 冕 *冉L
␰t
⳵N s
⳵s 冊
⫹ f t ds⫽0 (7)

where L⫽finite-element length and

s 其•兵t其
␰ t* ⫽ 兵 ␰ * (8)
Fig. 4. Interface constitutive law
Using integration by parts, the following relation is obtained from
Eq. 共7兲:

W s ⫽⫺ 冕 L
⳵␰ t*
⳵s
N s ds⫹␰ t* N s 兩 L0 ⫹ 冕 L
␰ t* f t ds (9) ⌬␧ t ⫽
⳵ 兵 ⌬␰ c 其
⳵s
兵t其⫹
⳵⌬␰ r
⳵s
(15)

The tangent component of smeared forces acting along an el- The variational formulation 关Eqs. 共11兲 and 共13兲兴 is transformed
ement of length ds at the steel-concrete interface is expressed as in a discrete system for programming purpose. Therefore, the ma-
f t ⫽⫺␾␶ (10) trix system to solve for a linear problem is

where ␾⫽bar perimeter and ␶⫽bond stress along a tangent vec-

tor. Based on the hypothesis of stress vector continuity, one can
兺 关 K T 兴 兵 ⌬␰ s其 ⫽ 兵 F ext⫺F int其 (16)

write the virtual work expression of the concrete and steel bar where the stiffness matrix is the summation of two matrices rep-

W⫽ 冕 具 *典 冕 具 *典
V
E 兵 S 其 dV⫺
V
␰ c 兵 f v 其 dV
resenting either a separate material behavior or the coupling be-
tween the two materials.
The vector of internal nodal forces has therefore two compo-

冕 具 *典
nents
⳵ ␰s
⫺
L
具 *典
⳵s
• 兵 t 其 N s ds⫹ ␰ s • 兵 t 其 N s 兩 L0
兵 F int其 ⫽A s 冕 兵 B c 其 ␴ ds⫹␾ 冕 兵 B r 其 ␶ ds (17)

冕 具 * *典
L L

⫺␾ ␰ s ⫺␰ c • 兵 t 其 ␶ ds⫽0 (11) where the axial stress and the bond stress are expressed for an
L elastic behavior as
where 具 E * 典 ⫽virtual strain tensor; 兵S其⫽Kirchhoff vector of stress;
and ␰ s* and ␰ *
c ⫽, respectively, the virtual steel displacement and
the virtual concrete displacement. The virtual relative slip is de-
␴⫽E s 冕具L
B c 典 兵 ␰ c 其 ds (18)

fined as
具 ␰ r* 典 ⫽ 具 ␰ s* ⫺␰ *c 典 (12)
␶⫽G 冕具
L
B r 典 兵 ␰ r 其 ds (19)

For an increment of axial force acting on the bar, the varia- A bond-slip constitutive law with kinematic hardening is used
tional formulation becomes to determine the bond stress. Such a law is suitable to represent
the interface gradual deterioration. Three parameters are used to
⌬W s ⫽⫺ 冕 具 *典
L
⳵ ␰s
⳵s
• 兵 t 其 ⌬N s ds⫹ 具 ␰ s* 典 • 兵 t 其 ⌬N s 兩 L0
define this constitutive law, as shown in Fig. 4: the initial modulus
G, the bond threshold ␶ 0 , and the hardening modulus H.
The determination of the parameter values is not an easy task
⫺␾ 冕 具 *典 L
␰ r • 兵 t 其 ⌬␶ ds (13)
as little literature is available on the subject. Moreover, the pa-
rameter values vary upon several factors such as the geometry of
the bar, the bar diameter, the concrete strength, and the concrete
where the last term represents the force increment caused by the cover.
development of bond stress at the interface. In Eq. 共13兲, the in- The classic unidimensional plasticity method is used to solve
crement of normal force is defined as the problem. The object is to find the stress increment correspond-
⌬N s ⫽E s A s ⌬␧ t (14) ing to the slip increment with respect to the yield criterion 共f兲

where E s ⫽Young’s modulus; A s ⫽steel bar section; and

⌬␧ t ⫽strain increment along the tangent vector.
⌬␶⫽ 冕 t2

t1
⌬␴ dt⫽ 冕 t1
t2
G ep⌬␰ r dt (20)
Frenet formula allows us to express the strain increment varia-
tion as f ⫽ 兩 ␶⫺H␰ rp 兩 ⫺␶ 0 ⫽0 (21)

JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2002 / 1505

where G ep⫽elastoplastic modulus; ␶⫽actual bond stress, and
␰ rp ⫽plastic slip. The equilibrium in the elastoplastic phase of this
nonlinear problem is obtained with the help of an iterative pro-
cess. Here, the Newton-Raphson method has been used.

Implementation in Finite-Element Program

The finite-element program CLEF 共Desjardins and Fafard 1992兲,
developed at Laval University within the ‘‘Groupe interdiscipli-
naire de recherche en éléments finis’’ 共GIREF兲 has been used
herein for the analysis of concrete columns. To model concrete, a
tridimensional solid element with 20 nodes is used, each node is
associated with three degrees of freedom being, respectively, the Fig. 5. Typical load-displacement input 关adapted from Sun et al.
displacements u, v , and w. This brick element supports the con- 共1993a兲兴
crete constitutive law presented earlier and modified as stated
before.
When a perfect bond between steel and concrete is considered, In the numerical analysis, finite bar elements with relative slip
a classical 3D three-node bar element is used; each node being were used to represent the behavior of the lapped splices where
associated with three degrees of freedom (u, v ,w). An elastoplas- the interface deteriorates rapidly under cyclic loading. An elasto-
tic constitutive law with positive hardening represents the steel plastic constitutive law is used for the interface. The parameters
reinforcing bar behavior. When the relative slip between the ma- of this law have been established from values found in the litera-
terials is considered, a special 3D three-node bar element is used. ture for 19-mm diameter bars. The initial modulus of the interface
As mentioned earlier, this element formulation involves an addi- constitutive law 共G兲 is 20 MPa/mm and the bond stress threshold
tional degree of freedom expressing the steel-concrete relative (␶ 0 ⫽5.6 MPa) was determined so that the elastoplastic phase
slip which is associated with each node (␰ r ). would begin when the lateral displacement corresponds to a duc-
Therefore, the modeling of reinforced concrete structural ele- tility level of one. The hardening modulus was chosen arbitrarily
ments taking into account the interface behavior can be realized as half of the initial modulus (H⫽10 MPa/mm). All material
with an appropriate mesh composed of bar elements with relative characteristics are described in Table 1.
slip, in addition to concrete solid elements. The steel bar elements For visibility and accuracy purposes, the finite-element mesh
are located at the edges of the concrete hexahedron, the coordi- used is subdivided into two distinct portions and presented in Fig.
nates of each element nodes being identical. The difference in the 6. For the numerical simulation, the prototype upper part is su-
number of degrees of freedom for each element presents a com- perposed to the footing shown on the right of Fig. 6, in a way to
patibility problem. This problem can be easily overcome by using line up the longitudinal reinforcement of each part. At the lapped
three linear relations for each pair of nodes having the same co- splice level, on the first 380 mm at the column base, there are two
ordinates. bars placed side by side, both considering relative slip.
The finite-element mesh consists of a total of 776 elements:
408 solid elements for the concrete 共H20兲, 144 bar elements with
Modeling of Concrete Columns with Lap Splices relative slip for the vertical reinforcement 共L3兲, 209 conventional
A concrete column with lapped starter bars was modeled to bar elements for the stirrups, and 15 pressure elements for the
compare its behavior under cyclic excitations with the load- application of the axial load 共Q8兲. A previous study on mesh
displacement curve obtained experimentally by Sun et al.
共1993a兲. The column rectangular cross section at a 40% dimen-
sional scale factor is 730 mm⫻490 mm. The height of the column Table 1. Materials Characteristics
is 3,660 mm and it is set on top of a footing, 460-mm high by
1,680-mm wide. The longitudinal reinforcement of the column Material properties Value
prototype consists of 32 bars with a diameter of 19 mm, uni- Concrete
formly distributed along the sides of the rectangular column cross Compressive strength f c⬘ 35 MPa
section as shown in Fig. 1. All the reinforcing bars start immedi- Tensile strength f t⬘ 3.2 MPa
ately above the top of the column footing, where they are lapped Initial modulus E c 27.23 GPa
with starter bars anchored in the column base with a lap length of Poisson’s coefficient ␯ 0.2
20 bar diameters 共380 mm兲, which is clearly insufficient for seis-
mic design, according to modern standards. Steel
The column prototype is subjected to an uniform vertical com- Modulus of elasticity E s 200 GPa
pressive stress (␴⫽0.15 f ⬘c ) which is representative of the axial Yield stress F y
load expected in single-column bridge piers. On the column’s Longitudinal reinforcement 300 MPa
strong axis, a lateral force and displacement pattern of increasing Stirrups 350 MPa
magnitude, as shown in Fig. 5, is also applied. This load and Constitutive law Kinematic hardening
displacement input history is divided into two phases. The first slope 5655 MPa
phase is in load control up to the theoretical cracking and yield Interface
limit state. At this point, the displacement (⌬ y ) obtained corre-
Initial modulus G 20 MPa/mm
sponds to a ductility level of 1 共␮⫽1兲. The second displacement
Bond threshold ␶ 0 5.6 MPa
history phase is designed to increase the ductility limit states of
Hardening modulus H 10 MPa/mm
the column prototype as multiples of ⌬ y .

1506 / JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2002

 Fig. 6. Mesh used for modeling of column and footing

refinement has indicated that this choice of mesh size is appropri-

ate for this type of problem 共Girard 2000兲.
Longitudinal reinforcing bars have been grouped together to Fig. 7. Load versus displacement for test column with lapped starter
reduce the problem size and to assure the objectivity of the solu- bars 关from Sun et al. 共1993a兲兴
tion. Half of the column has been modeled using mirror-
symmetry boundary conditions ( v ⫽0). The footing is considered
perfectly recessed at its base (u⫽ v ⫽w⫽0) as well as the bottom probable that these boundary conditions were more related to a
end of the vertical reinforcement (u⫽ v ⫽w⫽␰ r ⫽0), so to avoid partial restraint on the experimental prototype. The mesh layout
possible abutment slip at the end of the bar. may also have some impact on the numerical model stiffness. In
grouping together some reinforcing bars, the position of the steel
resulting force is slightly different in the model than in the ex-
perimental prototype, thereby inducing a smaller lateral displace-
Comparison with Experimental Studies
ment.
Sun et al. 共1993a兲 have described the general observed experi- The parameter values of the bond-slip constitutive law have an
mental behavior of the column prototype. The first flexural cracks important effect on the response of the reinforced concrete col-
appeared at the bottom section of the test column in the first umn under lateral load. Here, the initial stiffness of the bond-slip
‘‘push cycle’’ and they developed over a wider range of the col- constitutive law is considered constant, for programming simplic-
umn as the applied lateral load was increased. Vertical cracks ity. A polynomial expression as suggested by Mirza and Houde
were identified on the tension side near the bottom column cross 共1979兲 could be used to represent the bond-slip relation and may
section, at a load level of 267 kN, indicating the splitting effect of be more representative of the progressive deterioration of the in-
the lap splices. These vertical cracks further spread over the entire terface. Cyclic loading has an important effect on the interface
lap splice region on the tension side of the column at a ductility behavior. As the number of cycles increases, the bond strength
level of ␮⫽1. decreases considerably. A bond stress-slip relationship with a re-
The recorded lateral load-displacement hysteresis loops of the duced envelope, as suggested by Eligehausen et al. 共1983兲 could
column prototype with lapped starter bars are shown in Fig. 7. be used to represent the behavior of the interface after some re-
The lateral displacement of a specific point at the top of the col- petitive cycles.
umn at ␮⫽1 is 22.7 mm. When the lateral displacement is in-
creased to a ductility level of ␮⫽1.5, the concrete cover at the
corner of the column cross section near the top of the lap splices
started to spall. The recorded load carrying capacity of the col-
umn dropped over 28% from its peak value, due to vertical cracks
and bond strength deterioration in the lap splice region. At a duc-
tility level of ␮⫽3, the concrete cover over the entire lap splice
region was completely separated from the vertical reinforcement
on both sides of the column. At that time, the column prototype
lost most of its load-carrying capacity.
The numerical load-displacement loops are shown in Fig. 8.
This figure shows the lateral load versus the lateral displacement
of the same specific point at the top of the column until a ductility
level of ␮⫽2. The mean displacement obtained for a ductility
level of one is 16 mm, that is, 70% of the displacement obtained
experimentally 共22.7 mm兲. Comparing with the experimental be-
havior, the stiffness of the numerical model is greater. Without
doubts, this difference can be partly explained by the fact that
Fig. 8. Load versus displacement for numerical modeling of column
some construction details are not known. In the numerical model,
with lapped starter bars
we consider the base of the footing totally fixed, however, it is

JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2002 / 1507

Fig. 9. Axial stress distribution in longitudinal reinforcement of
column

Fig. 8 shows that the numerical load-displacement loops are

stable, well centered on the graph and that there is a drop of
nearly 10% in the load-carrying capacity of the column with a
ductility level of 1.5. Experimentally, an important drop in the
load-carrying capacity was also observed 共by 30%兲 at the same
ductility level, when vertical cracks appeared in the lap splice
region. Beyond this ductility level, it is clear that the numerical
model cannot represent the complete destroying of the interface
which happens quickly when the lap length is too short.
The axial stress distributed along the longitudinal reinforce-
ment, at different load levels, is shown in Fig. 9, while the bond Fig. 10. Bond stress distribution along longitudinal reinforcement of
stress distribution is shown in Fig. 10. On these figures, the ab- column
scissa level corresponds to the top of the footing. The values
gathered on those graphs are associated to the state existing at the
cessible, on the general behavior such as hysteresis loops, on the
midlength of each bar element constituting a reinforcing bar; this
damage evolution in concrete as well as on the stress distribution
bar being located the farthest from the neutral axis of the column
within each material.
cross section as shown on the plan view. The progression of dam-
It has been shown that the use of a finite element with an
age in concrete along the column height has also been examined
additional degree of freedom 共DOF兲 at each node expressing the
and is shown in Fig. 11, for different load levels. Each square
relative slip between materials is a simple and reliable method to
represents the fact that a quadrature point in the concrete near the
model structural elements where the progressive deterioration of
longitudinal reinforcement has reached a cracking state. For a
steel-concrete bond link affects the general behavior.
high-load level 共223 kN and more兲, the curves in Figs. 9 and 10
The results obtained with the modeling of a concrete bridge
show a highly nonlinear behavior when the surrounding concrete
column with lapped starter bars show that the use of a finite
is cracked. As shown in Fig. 2, when a primary crack emerges in
the concrete, the steel bar must undertake the total axial stress.
Therefore, concrete stress vanishes and the stress in steel reaches
its highest value for equilibrium purposes.
The bond stress distribution along the longitudinal reinforce-
ment anchored in the footing is shown in the second graphic of
Fig. 10. At the lapped splice level 共from 0 to 380 mm兲, the bond
stress variation is important. The bond stress on the lapped splice
region is in most part negative and the stress transfer from one bar
to the other is done accurately.

Summary and Conclusions

The finite-element model developed to investigate the response of

reinforced concrete columns subjected to cyclic loads, takes into
account the principal characteristics of the nonlinear behavior of
each material: concrete, steel, and interface. A simple analysis on
Fig. 11. Damage in concrete near longitudinal reinforcement
any structural element provides a lot of information, quickly ac-

1508 / JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2002

element with slip DOF is appropriate to represent the general f c⬘ ⫽ concrete compressive strength;
behavior due to an important deterioration of the steel-concrete 兵 f s 其 ⫽ vector of smeared forces on surface;
bond at the base of the column. The numerical hysteresis loops f t ⫽ tangent component of smeared forces;
obtained until a ductility level of two are stable, well centered on f ⬘t ⫽ concrete tensile strength;
the graph, showing that the model responds well under cyclic 兵 f v 其 ⫽ vector of smeared forces on volume;
loading. In the presence of a lapped splice, the stress transfer is G ⫽ interface law initial modulus;
done gradually from one bar to the other. The effect of damage in G ep ⫽ interface law elastoplastic modulus;
concrete on the steel stress distribution has been demonstrated, as H ⫽ interface hardening modulus;
long as the coupled effect of damage in concrete and bond-slip 关 K T 兴 ⫽ stiffness matrix;
constitutive law on the general behavior of the structural element. L ⫽ finite-element length;
One of the main advantages of the tridimensional concrete N s ⫽ normal force acting in tangent direction;
model is its capability of representing multiple cracks in a struc- 兵S其 ⫽ Kirchhoff vector of stress;
tural element because of the modeling of the softening response 兵t其 ⫽ tangent vector;
of concrete. The numerical model can treat problems under cyclic V ⫽ volume;
loading and in the same time, take into account the gradual dete- W ⫽ variational formulation;
rioration of the steel-concrete bond. The concrete constitutive law W s ⫽ steel variational formulation;
represents well any triaxial stress state and indirectly, the confine- ⌬ ⫽ increment;
ment effect brought by the presence of lateral strirrups in a con- ⌬ y ⫽ displacement corresponding at ductility
crete column. ␮⫽1;
One of the difficulties lies in the choice of the parameter val- ␧ c ⫽ concrete strain;
ues for the interface constitutive law. Very little literature exists ␧ r ⫽ relative slip strain;
on bond-slip law obtained from specific testing on lapped splice, ␧ s ⫽ steel strain;
with different bar diameter and with different concrete cover. ␧ t ⫽ strain along tangent vector;
There is a need in the elaboration of such a testing program. Some ␮ ⫽ ductility level;
additional developments are needed in the definition of the finite- ␯ ⫽ Poisson’s coefficient;
element bar with relative slip such as the implementation of re-
␰ c ⫽ concrete displacement;
duced envelopes in the interface constitutive law. As for the steel
␰ r ⫽ relative slip between steel and concrete;
constitutive law, the model could also represent the hardening
␰ s ⫽ steel displacement;
behavior of the material. Doing so, the overall response of rein-
具 ␰ *c 典 ⫽ virtual concrete strain tensor;
forced concrete structures modeled under cyclic loading will un-
␰ rp ⫽ plastic slip strain;
doubtedly come closer to the experimental behavior.
具 ␰ r 典 ⫽ virtual relative slip strain tensor;
*
具 ␰ s* 典 ⫽ virtual steel strain tensor;
␰* t ⫽ virtual displacements;
Acknowledgments
␴ ⫽ axial stress;
␶ ⫽ bond stress;
The writers wish to thank Professor Mario Fafard for discussions
␶ 0 ⫽ interface law bond threshold; and
and suggestions regarding the formulation and numerical imple-
␾ ⫽ bar perimeter.
mentation of the elements formulation. Support from the National
Sciences and Engineering Research Council of Canada 共NSERC兲
and from FCAR 共‘‘Fonds pour la Formation de Chercheurs et
l’Aide à la Recherche’’兲 are gratefully acknowledged. The writers References
wish to thank the ‘‘Groupe de recherche interdisciplinaire en
éléments finis’’ 共GIREF兲 from ‘‘Université Laval’’ for technical American Concrete Institute 共ACI兲 Committee 446. 共1991兲. ‘‘Fracture
support. mechanics of concrete: Concepts, models and determination of mate-
rial fracture properties.’’ ACI 446.1R-91, Detroit.
Bouzaiene, A. H. 共1995兲. ‘‘Caractérisation et modélisation du comporte-
ment non linéaire d’éléments en béton armé et non armé sous sollici-
Notation tations multiaxiales.’’ PhD thesis, École Polytechnique de Montréal,
Quebec 共in French兲.
The following symbols are used in this paper: Chai, Y. H., Priestley, M. J. N., and Seible, F. 共1991兲. ‘‘Seismic retrofit of
A s ⫽ steel bar section; circular bridge columns for enhanced flexural performance.’’ ACI J.,
兵 B c 其 ⫽ vector relating strain to nodal concrete dis- 88共5兲, 572–584.
placement; Darwin, D., and Pecknold, D. A. 共1977兲. ‘‘Nonlinear biaxial stress-strain
兵 B r 其 ⫽ vector relating strain to nodal relative slip law for concrete.’’ J. Eng. Mech., 103共2兲, 229–241.
displacement; Desjardins, R., and Fafard, M. 共1992兲. ‘‘Développement d’un logiciel
ds ⫽ length of curvilinear element; pour l’analyse des structures par éléments finis utilisant l’approche de
E c ⫽ concrete modulus; la programmation orientée objets.’’ Rep. No. GCT-92-05, Dept. of
Civil Engineering, Université Laval, Quebec.
E s ⫽ Young’s modulus;
Eligehausen, R., Popov, E. P., and Bertero, V. V. 共1983兲. ‘‘Local bond
具 E * 典 ⫽ virtual strain tensor; stress-slip relationships of deformed bars under generalized excita-
兵F其 ⫽ vector of forces; tions.’’ Rep. No. UCB/EERC-83/23, Earthquake Engineering Research
兵 F ext其 ⫽ vector of external forces; Center, Univ. of California, Berkeley, Calif.
兵 F int其 ⫽ vector of internal forces; Girard, C. 共2000兲. ‘‘Modélisation de colonnes de pont en béton armé sous
F y ⫽ yield stress; sollicitations cycliques.’’ PhD thesis, Dept. of Civil Engineering, Uni-
f ⫽ yield criterion; versité Laval, Quebec 共in French兲.

JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2002 / 1509

Lukose, K., Gergely, P., and White, R. N. 共1982兲. ‘‘Behavior of reinforced Salençon, J. 共1988兲. Mécanique des milieux continus. Tome II: Élasticité-
concrete lapped splices for inelastic cyclic loading.’’ ACI J., 79共5兲, Milieux curvilignes. Ellipses, Paris 共in French兲.
355–365. Sivakumar, B., Gergely, P., and White, R. N. 共1983兲. ‘‘Suggestions for the
Lutz, L. A., and Gergely, P. 共1967兲. ‘‘Mechanics of bond and slip of design of R/C lapped splices for seismic loading.’’ Concr. Int.: Des.
deformed bars in concrete.’’ ACI J., 64共11兲, 711–721. Constr., 5共2兲, 46 –50.
Sun, Z., Seible, F., and Priestley, M. J. N. 共1993a兲. ‘‘Flexural retrofit of
Massicotte, B., Bouzaiene, A. H., and Loulou, R. 共1996兲. ‘‘Analyse par
rectangular reinforced concrete bridge columns by steel jacketing. Ex-
éléments finis du comportement en service et à l’ultime d’un pont en perimental studies.’’ Rep. No. SSRP-93/01, Dept. of Applied Mechan-
béton armé.’’ Rep. No. EPM/GCS-1996-08, Dept. of Civil Engineer- ics and Engineering Sciences, Univ. of California, San Diego.
ing, École Polytechnique de Montréal, Quebec 共in French兲. Sun, Z., Seible, F., and Priestley, M. J. N. 共1993b兲. ‘‘Diagnostics and
Mirza, S. M., and Houde, J. 共1979兲. ‘‘Study of bond stress-slip relation- retrofit of rectangular bridge columns for seismic loads. Theoretical
ships in reinforced concrete.’’ ACI J., 76共2兲, 19– 46. studies.’’ Rep. No. SSRP-93/07, Dept. of Applied Mechanics and En-
Monti, G., Filippou, F., and Spacone, E. 共1997兲. ‘‘Finite element for an- gineering Sciences, Univ. of California, San Diego.
chored bars under cyclic load reversals.’’ J. Struct. Eng., 123共5兲, 614 – Tassios, T. P., and Yannapoulos, P. J. 共1981兲. ‘‘Études analytiques sur
623. l’adhérence acier-béton et la fissuration du béton armé sous charges
Ngo, D., and Scordelis, A. C. 共1967兲. ‘‘Finite element analysis of rein- cycliques.’’ Annales de l’Institut technique du bâtiment et des travaux
publics, 393, 46 – 62 共in French兲.
forced concrete beams.’’ ACI J., 64, 152–163.
Ulm, F.-J. 共1994兲. ‘‘Un modèle d’endommagement plastique: application
Okamura, H., and Kim, I. H. 共2000兲. ‘‘Seismic performance check and
aux bétons de structure.’’ PhD thesis, École Nationale des Ponts et
size effect FEM analysis of reinforced concrete.’’ Eng. Fract. Mech., Chaussées, Paris 共in French兲.
65共2兲, 369–389. Vecchio, F. J. 共2000兲. ‘‘Disturbed stress field model for reinforced con-
Pochanart, S., and Harmon, T. 共1989兲. ‘‘Bond-slip model for generalized crete: Formulation.’’ J. Struct. Eng., 126共9兲, 1070–1077.
excitations including fatigue.’’ ACI Mater. J., 86共5兲, 465– 474. Willam, K. J., and Warnke, E. P. 共1975兲. ‘‘Constitutive model for triaxial
Priestley, M. J. N., Seible F., and Calvi, G. M. 共1996兲. Seismic design and behavior of concrete.’’ IABSE Seminar on Concrete Structures Sub-
retrofit of bridges, Wiley, New York. jected to Triaxial Stresses, 19, Bergamo, Italy.

1510 / JOURNAL OF STRUCTURAL ENGINEERING / DECEMBER 2002