340

Finite element modeling of concrete structures

reinforced with internal and external fibre-
reinforced polymers1
Ali Nour, Bruno Massicotte, Emre Yildiz, and Viacheslav Koval

Abstract: Externally bonded fibre-reinforced-polymer (FRP) laminates and fabrics have been successfully used for
strengthening damaged or deficient concrete members, whereas internal FRP reinforcements are becoming an efficient
alternative to steel reinforcement, particularly in corrosive environments. Despite the enormous progress that has been
observed in the last decade, further research is still required to consolidate recent developments and expand the scope
of application of FRPs for structural uses. Nonlinear finite element analysis combined with laboratory testing constitutes
an efficient approach for pursuing this objective. The scope of this paper is to illustrate, through a selection of a wide
variety of typical applications, the contribution of a refined three-dimensional (3-D) constitutive model for investigating the
nonlinear response of concrete structures reinforced with internal and external FRPs. The analyses are carried out using a
general and portable constitutive concrete model implemented as a user-defined subroutine at Gauss integration point level
in commercial finite element software. The constitutive law follows a 3-D hypoelastic approach that models the nonlinear
behaviour of concrete using a scalar damage parameter that accounts for the anisotropic behaviour of partially confined
concrete and the inelastic volume expansion upon reaching the peak strength. In tension, the model adopts a macroscopic
approach that is directly integrated into the concrete law. It simulates implicitly the reinforcing bar – concrete interaction
using tension-stiffening factors modified according to the nature of reinforcement that vary as a function of the member
strain. The applications include results of well-known test series published in the literature on beams with external and
internal FRP reinforcement, slabs with internal reinforcements, bond failure analysis of external FRP, and the effect of
confinement on the behaviour in compression of circular and square elements. The paper demonstrates the ability of the
concrete model to correctly simulate the behaviour of structural elements reinforced with FRPs at service load level and
reproduce failure mechanisms and loads that are consistent with the experimental observations.
Key words: constitutive model, nonlinear analysis, finite element, reinforced concrete, glass-fibre-reinforced polymer
(GFRP), carbon-fibre-reinforced polymer (CFRP), strengthening, steel.

Résumé : Les laminés et les tissus en polymères renforcés de fibres (« FRP ») fixés sur les faces externes ont été utilisés
avec succès pour renforcer des éléments de béton endommagés ou défectueux alors que les renforcements « FRP »
internes commencent à remplacer efficacement un renforcement en acier, particulièrement dans les environnements
corrosifs. Malgré les grands progrès réalisés au cours de la dernière décennie, une recherche plus approfondie est
nécessaire afin de faire la synthèse des récents développements et d’élargir l’utilisation des « FRP » dans les structures.
Une analyse non linéaire par éléments finis combinée aux essais en laboratoire constitue un moyen efficace pour atteindre
cet objectif. Cet article illustre, grâce à des exemples provenant d’un très grand nombre d’utilisations typiques diverses,
la contribution d’un modèle constitutif tridimensionnel perfectionné à l’étude de la réponse non linéaire des structures
de béton armé de « FRP » internes et externes. Les analyses ont été réalisées en utilisant un modèle constitutif du
béton général et universel en tant que sous-programme défini par l’usager au point d’intégration Gauss dans un logiciel
commercial d’analyse par éléments finis. La loi constitutive suit une approche hypoélastique tridimensionnelle qui
représente le comportement non linéaire du béton par l’utilisation d’un paramètre de dommages scalaire qui tient compte
du comportement anisotrope du béton lorsque partiellement confiné et de l’expansion inélastique de volume lors de
l’atteinte de la contrainte maximum. En tension, le modèle adopte une approche macroscopique directement intégrée
dans la loi régissant le béton. Il simule implicitement l’interaction tige de renforcement-béton en utilisant des facteurs de
raidissement en tension modifiés selon la nature du renforcement ; ces facteurs varient en fonction de la contrainte dans
les éléments. Les utilisations comprennent les résultats de séries d’essais bien connues, publiées dans la littérature, sur des

Received 25 November 2005. Revision accepted 20 September 2006. Published on the NRC Research Press Web site at cjce.nrc.ca/ on
12 May 2007.

A. Nour, B. Massicotte,2 E. Yildiz, and V. Koval. Département de génie civil, géologique et mines, École Polytechnique de Montréal,
CP 6079, Station Centre-ville, Montréal, QC H3C 3A7, Canada.
Written discussion of this article is welcomed and will be received by the Editor until 31 July 2007.
1
This article is one of a selection of papers in this Special Issue on Intelligent Sensing for Innovative Structures (ISIS Canada).
2
Corresponding author (e-mail: [email protected]).

Can. J. Civ. Eng. 34: 340–354 (2007) doi: 10.1139/L06-140 © 2007 NRC Canada
Nour et al. 341

poutres munies de renforcement en « FRP » externe et interne, des dalles munies de renforcement interne, d’analyse de
la rupture dans les « FRP » fixés sur les faces externes et l’effet du confinement sur le comportement en compression des
éléments circulaires et carrés. Le présent article démontre la capacité du modèle à simuler correctement le comportement
des éléments structuraux renforcés de « FRP » lors d’une utilisation à la charge de service et à reproduire les mécanismes
et les charges de défaillance concordant bien avec les observations expérimentales.
Mots-clés : modèle constitutif, analyse non linéaire, élément fini, béton armé, polymère renforcé de fibres de verre
(« GFRP »), polymère renforcé de fibres de carbone (« CFRP »), renforcement, acier.
[Traduit par la Rédaction]

1. Introduction Laboratory tests on structural elements or on reduced-scale

structures are essential for observing the actual behaviour and
During their service life, civil infrastructures may need struc- failure modes. Testing is, however, expensive and time-
tural repair or strengthening for several reasons, such as design consuming and often limits the pace of research progress.
and construction defaults, climatic or chemical agent damage, Existing structures cannot be tested at ultimate failure, whereas
corrosion, accidental overloading or impact, fires, or earth- because of limitations of testing equipment, scale effects on
quakes. Modifications to the loading level, design forces, or large structures cannot be observed experimentally. This sit-
code specifications may require the strengthening of critical uation has strongly encouraged the development of advanced
load-carrying members. Generally, the complete replacement analytical methods capable of representing the behaviour of
of deficient structures is very expensive, and their repair or concrete structures internally and (or) externally reinforced by
strengthening is often considered as the most economical solu- composite materials under all possible loading conditions. In
tion. this context, recourse to numerical strategies based on the finite
In recent years fibre-reinforced-polymer (FRP) composites element method could enable correct prediction of structural be-
have become an alternative for strengthening and repairing de- haviour up to ultimate failure and capture with fidelity the main
ficient concrete structures. Externally bonded FRP fabric and complexities associated with material nonlinearities, namely
laminates have been successfully used for strengthening re- concrete confinement, lap splice strength, concrete postcrack-
inforced concrete beams, slabs, or columns. They represent ing, and the interaction between concrete and composite or steel
promising means for remedying structural deficiencies and for reinforcements.
enhancing the performance of civil engineering structures Bouzaiene and Massicotte (1997) developed a three-
(Nitereka and Neale 1999). dimensional constitutive model that can simulate the behaviour
Steel has been used for more than a century as an effec- of concrete under multiaxial stress conditions. The model is
tive and cost-efficient internal reinforcement for concrete struc- based on the hypoelastic approach for which the behaviour of
tures. When it is not prone to ion attack, steel reinforcement can concrete is modeled using an equivalent uniaxial stress–strain
last for decades without any noticeable deterioration. However, curve combined with a compression scalar damage parameter.
concrete members in corrosive environments or exposed to de- The model accounts for anisotropy, elastic modulus degradation
icing salts are sensitive to reinforcement corrosion, which often under loading–unloading, and inelastic volume expansion.
leads to significant diminution of their strength and thereby af- The original model was developed as an academic finite el-
fects structure safety. Costs related to repair or replacement of ement software. Within the scope of the Center of Excellence
structures damaged by corrosion is very high and occupies a ISIS Canada project 2.3.4, the model was improved and mod-
large portion of bridge owner budget (El-Salakawy et al. 2003). ified to make it portable to most commercial nonlinear finite
Most solutions adopted to protect reinforcing steel, such as element software. The objective was to provide a powerful con-
galvanization, concrete additives, epoxy coating, or cathodic stitutive model that could be made available to other users.
protection, generally have limited success (Benmokrane et al. Massicotte et al. (2007) introduced a new tension-stiffening
2000). Internal FRP reinforcement is an alternative material model to the original law of Bouzaiene and Massicotte and
to replace conventional steel bars. Fibre-reinforced polymers made the constitutive model portable. The proposed tension-
present outstanding characteristics, such as durability, because stiffening model, inspired by CEB-FIP (1990) and Winkler
of their high corrosion resistance, high strength-to-weight ratio, et al.’s (2004) recommendations, adopts a macroscopic ap-
and advantageous fatigue resistance. In addition, FRP reinforce- proach that modifies the concrete law. The model integrates
ments exceed the strength and can match stiffness properties of variable tension-stiffening factors defined as a function of the
steel. Savings in maintenance usually largely compensate for member strain, as used by many investigators to account for
the higher initial cost of FRPs. steel–concrete interaction (Fields and Bischoff 2004). The post-
In the last decade, significant progress on the use of FRPs cracking modeling of concrete follows the smeared crack ap-
as construction materials has been observed. Recent codes and proach proposed by Rachid (1968), which constitutes an effi-
standards have introduced specifications for the design or the cient representation of the cracked concrete in nonlinear finite
strengthening of members with internal or external FRPs, re- element analysis.
spectively. Despite these significant advances, more research The term portable means that the constitutive model is imple-
is still required to expand the scope of utilisation of FRPs and mented as a user-defined subroutine, at Gauss integration point
relax some limitations imposed on their use owing to the lack level, into finite element software. For the present study, the
of scientific knowledge. model was introduced into each of the two components of the

© 2007 NRC Canada

342 Can. J. Civ. Eng. Vol. 34, 2007

Fig. 1. Concrete model: (a) concrete in compression; (b) concrete in tension.

general-purpose finite element software ABAQUS (Hibbitt et concrete upon approaching ultimate strength. The behaviour
al. 2004) (i.e., ABAQUS standard (STD) and ABAQUS explicit of concrete is modeled using an equivalent uniaxial curve, a
(EXP) (Massicotte et al. 2007)). compression scalar damage parameter λ, and either the five pa-
Experimental results aimed at defining tension-stiffening fac- rameters of Willam and Warnke (1975) or the four parameters of
tors for internal FRP reinforcement are scarce. The high elastic Hseih et al. (1982) failure surfaces. The scalar damage parame-
modulus of carbon-fibre-reinforced polymers (CFRPs) would ter provides an invariant measure of the evolution of irreversible
presumably lead to relationships similar to those of steel re- damage in concrete, which overcomes the difficulties encoun-
inforcement. Reinforcing bars made of glass-fibre-reinforced tered in the application of other hypoelastic models (Darwin
polymers (GFRPs) have stiffness that is lower than steel and and Pecknold 1977; Elwi and Murray 1979). Only an overview
thereby require appropriate formulation for describing the as- is presented in this paper; details on the model are available in
sociated tension-stiffening factors. Bischoff and Paixao (2004) Bouzaiene and Massicotte (1997).
indicated that concrete members reinforced with GFRP bars ex- The damage evolution of concrete in compression is de-
hibit greater tension stiffening than those reinforced with steel scribed by a scalar function in relation to the major compressive
bars. Moreover, concrete members reinforced with GFRP de- stress by means of an equivalent uniaxial curve. A realistic rep-
velop wider and deeper cracks than members reinforced with resentation of the cumulative damage, along any stress path, is
steel. obtained by the following incremental relationship:
This paper presents the main features of the concrete model


1
developed at École Polytechnique (Bouzaiene and Massicotte [1] λ= dλ = d
dε d
1997; Massicotte et al. 2007) and the recent modifications im- εmax
load path
plemented to the model to enable analysis of concrete members
reinforced with FRPs. The paper illustrates the performance of  εeq − σ /Ec
≈ for 0≤λ≤1
the model for capturing and predicting the behaviour of concrete εc − σc /Ec
members and structural elements reinforced with internal and
external FRPs. The examples presented in the paper are taken where dλ stands for the damage parameter increment, εeq is
from well-known tests, published the literature, that have been the uniaxial equivalent strain increment, σ is the compression
specifically selected for illustrating the capabilities and versa- stress increment according to the major axis, εc is the uniaxial
tility of the model for simulating a wide variety of structural equivalent strain at peak load, σc is the major compressive stress
behaviour. at failure, and Ec is the material elastic modulus. Equation [1]
expresses the inelastic stain increment dεd and the total inelastic
d . For an elastic behaviour of concrete, the
strain at failure εmax
2. Material models parameter λ is set equal to zero and equals one at the peak.
The concept of equivalent damage allows the evaluation of the
2.1. Concrete increment and the total equivalent strain for a nonproportional
For simulating the behaviour of concrete under multiaxial loading as follows (Fig. 1a):
loading, Bouzaiene and Massicotte (1997) developed a three-
dimensional hypoelastic constitutive model that accounts for σ − σ̃
the anisotropic behaviour of concrete, elastic modulus degrada- εeq =
[2] Et
tion under loading–unloading, the transition point that εeq = ε̃eq (λ) + εeq
separates brittle and ductile behaviour of concrete under in-
creasing confinement, and the volume increase observed in where σ̃ stands for the effective compression stress obtained

© 2007 NRC Canada

Nour et al. 343

Fig. 2. Constitutive relations for steel and fibre-reinforced-polymer (FRP) reinforcements.

from the equivalent uniaxial curve corresponding to the accu- principal axes, interpreted also as inelastic volume expansion.
mulated damage at the previous step, Et is the stiffness modulus This asymmetrical behaviour suggests the use of an unsymmet-
corresponding to λ, and σ is the major compression stress. One rical Poisson ratio, where for each principal plane the expansion
also notes that the uniaxial stress–strain curve of Saenz (1964) effect is limited to the lowest confined axis. For a triaxial state,
is adopted for its convenience. the effective Poisson ratio, a function of the damage parameter
In compression, the rapid increase of Poisson ratio is related λ and the gradient of the principal stresses, is expressed in the
to the opening of microcracks in the direction of the unconfined following form according to Bouzaiene and Massicotte (1997):

  
υ + λ (1 − υ ) σi − σj for σi < 0 and σi < σj
0 0
[3] υij = σc

υ0 otherwise

where υ0 stands for the initial Poisson ratio and σi and σj take structural response upon mesh refinement and corresponds to a
the values of principal stresses. representative dimension of the mesh size.
In tension, concrete cracks are generally characterized by When the tensile strength criterion is exceeded, a microcrack
the occurrence of microscopic discontinuities that are trans- band perpendicular to the principal direction in tension devel-
formed into discrete cracks with a total loss of resistance after ops. The material gradually loses its integrity, leading to the
reaching a certain strain value. For modeling tensile rupture, degradation of material properties interpreted as tension dam-
the fundamental concept of energy equivalence is adopted. It age. As in compression, tension damage evolution is described
is combined with the smeared crack technique, which assumes by means of a scalar parameter d, which represents the degra-
that cracks are uniformly distributed in the concrete mass. This dation of material properties due mainly to crack propagation.
enables modeling of the crack effect using a stress–strain rela- Thus, once the crack is initiated, the elastic modulus Ec is re-
tion, called a softening curve. As suggested by Feenstra and De duced using the damaged material modulus Ed as follows:
Borst (1996), concrete softening in tension is modeled using a
simple function defined as follows (Fig. 1b): [6] Ed = (1 − d)Ec
  where d = 0 corresponds to undamaged material and d = 1 to
ε − εe
[4] σt = ft exp − completely damaged material. The parameter d is estimated as
εa proposed in Crisfield and Wills (1989). Therefore, with this for-
mulation, the damage parameters in compression λ and in ten-
for which
sion d are totally coupled during the calculation of the degraded
elastic modulus and the residual tensile stress (Massicotte et al.
[5] εa = (Gf /h)ft and εe = ft /Ec
2007).
where Gf is the fracture energy consumed by the crack per
unit of crack surface and h is the equivalent length for which 2.2. Steel and fibre-reinforced-polymer reinforcements
the displacement due to crack opening is uniformly distributed. Steel reinforcement is assumed to behave in an elastic-plastic
As stated by Bažant and Oh (1983) and Oliver (1989), the use manner or as a strain hardening material when applicable
of the equivalent length h in the finite element calculations (Fig. 2a). The yield strength is represented by fys . As shown
leads to results that are insensitive with regard to the global by Fig. 2b, FRP reinforcements are assumed to be linear elastic

© 2007 NRC Canada

344 Can. J. Civ. Eng. Vol. 34, 2007

Fig. 3. Constitutive relations accounting for the tension stiffening action in a global dimension by modifying the stress–strain
effect. relationship of the material, either the reinforcing bar or the
concrete. The tension-stiffening model described and validated
in detail in Massicotte et al. (2007) follows the approach of the
original law of Bouzaiene and Massicotte (1997), which asso-
ciates tension stiffening with the concrete law. This model is
inspired from recommendations by CEB-FIP (1990) and Win-
kler et al. (2004) and integrates the tension-stiffening factors
used by many investigators to account for bar–concrete inter-
action. It is also combined with the smeared crack approach
(Rachid 1968) in the context of nonlinear finite element anal-
ysis. The bond between reinforcing bars and the surrounding
concrete is assumed to be perfect because the local effects asso-
with brittle fracture in tension. The ultimate tensile strength of ciated with the rebar–concrete interface, such as bond slip and
the material is represented by fyFRP , while the corresponding dowel action, are considered implicitly by the tension-stiffening
strain at failure is εuFRP . When FRPs are used as external rein- model. For a given strain, the equivalent concrete stress σc,TSE
forcement, each lamina is considered as an orthotropic layer in attributed to tension stiffening can be described as the difference
a plane stress condition, but the lamina is transversely isotropic, between the average bar stress σbar of the reinforced concrete
where the properties of the FRP are nearly the same in the direc- member and the stress of the bar σbar,II at the crack, determined
tion perpendicular to the fibres. In ABAQUS, FRPs are modeled as follows (Winkler et al. 2004):
as linear elastic materials. 
[7] σc,TSE = ρeff σbar − σbar,II
2.3. Tension-stiffening modeling
for which
The presence of steel or internal FRP reinforcement neces-
sitates the consideration of bar–concrete interaction. The bar– [8] ρeff = Abar /Ac,eff
concrete adherence allows the concrete located between cracks
to resist tensile stresses, thereby reducing the average reinforce- where ρeff is the effective reinforcement ratio, Abar is the bar
ment stress level compared to its magnitude at the crack. This area, and Ac,eff is the concrete area involved in the stiffen-
bar–concrete interaction phenomenon results in a gain in rigid- ing, computed according to CEB-FIP (1990) recommendations.
ity, also called tension stiffening. A simple way to account for Thus, as shown in Fig. 3 the stress–strain relationship is defined
this local phenomenon is to integrate the bar–concrete inter- by Winkler et al. (2004) and Massicotte et al. (2007)



 = ft 1 − [1 − β (ε)]
ε

 for ε < ε1
 ε1
[9] σt (ε) = β(ε)ft   for ε1 < ε < ε2

 εbar,y − ε


= β(ε)ft for ε2 < ε < εbar,y
εbar,y − ε2

where ε1 and ε2 are defined as follows (ε1 < ε2 < εbar,y ): model adopts an equivalent reinforcement ratio, which gener-
  ates the same stiffening contribution as that produced by the
1 1 actual reinforcement, as suggested in Massicotte et al. (1990).
ε1 = [1.3 − β (ε)] ft +
Ebar ρeff Ec The tension-stiffening relationship expressed by eq. [9] and
[10]   proposed by Massicotte et al. (2007) considers the tension-
1 1
ε2 = εbar,y − β (ε) ft + stiffening factor β(ε) as a function of the member strain. This
Ebar ρeff Ec formulation allows the introduction of various empirical models
In eqs. [9] and [10] Ebar is the reinforcing bar modulus of proposed in the literature by proceeding to a direct modulation
elasticity, εbar,y corresponds to the strain at yielding for steel with the various sections of the stress–strain relationship (eq.
reinforcements or the ultimate strain for composite bars, and β [9]). For the case of steel reinforcement, a sensitivity analysis
is the tension-stiffening factor that accounts for bar–concrete was performed to select the most appropriate tension-stiffening
interaction. The tension-stiffening relationship (eq. [9]) is es- factor. It was determined that the Fields and Bischoff (2004)
tablished for the case of cracks orthogonal to the bar direction. model, expressed as follows, provides the best results (Massi-
In real structures cracks have an unspecified orientation with cotte et al. 2007):
the reinforcing bar direction. Therefore, for steel and compos-
[11] βs (ε) = exp [−800 (ε − εe )]
ite bars not orthogonal to the direction normal to the crack, the

© 2007 NRC Canada

Nour et al. 345

Fig. 4. Different empirical models for the tension stiffening factor no experimental evidence is available to quantify this effect.
β(ε). GFRP, glass-fibre-reinforced polymer. Other phenomena such as thermal cracking or alkali–aggregate
reaction could also influence this parameter. It is assumed in
this paper that analyses are carried out for sound concrete.

3. Nonlinear finite element analysis

The constitutive model is implemented as a user-defined sub-
routine, at Gauss integration point level. For the results pre-
sented in this paper, the model has been introduced in the
two components of the general purpose finite element software
ABAQUS (Hibbitt et al. 2004), i.e., ABAQUS/Standard (STD)
and ABAQUS/Explicit (EXP). The formulation, element types,
and solution strategies for solving nonlinear problems are quite
different in these two software components. The concrete model
is written in FORTRAN and is included in the UMAT subrou-
tine for the analyses with STD and the VUMAT subroutine for
the analyses with EXP. In this paper, all analyses were carried
out using STD only, the comparison with EXP solutions being
Tension stiffening for FRP rebars with an elastic modulus in beyond the scope of the paper. More details on that matter are
the range of steel could probably be modeled satisfactorily with provided in Massicotte et al. (2007).
eq. [11], although, to the authors’ knowledge, no experimental Analysis with STD represents the classical version of the
data of direct tension tests are available. In the case of GFRP re- general finite element software and uses a tangential stiffness
inforcing bars, limited information is available in the scientific approach that is unconditionally stable. The two principal meth-
literature regarding the tension-stiffening factor β(ε). Bischoff ods available for the resolution of nonlinear problems are the
and Paixao (2004) observed that GFRP-reinforced concrete traditional Newton–Raphson method (Static General), and the
members exhibit greater tension stiffening than equivalent steel- Riks arc length method (Static Riks).
reinforced concrete elements for any given axial strain value. STD calls the subroutine UMAT at each increment and iter-
Moreover, concrete members reinforced with GFRPs develop ation of the Newton–Raphson solving process, for each Gauss
wider and deeper cracks than members reinforced with steel integration point of all elements of the mesh. The finite ele-
or CRFP rebars. They proposed the following expression for ment software proves the concrete model subroutine with the
the tension-stiffening factor applicable to GFRP or any type of updated strain tensor. With these data, the subroutine updates
reinforcement: the stress tensor and the solution-dependent state variables that
contain the damage history for each integration point of the fi-
[12] βG (ε) = exp [−1100 (ε − εe ) (Ebar /200)]
nite element model. The subroutine also returns the Jacobean
where Ebar is the modulus of elasticity of the composite bar in matrix ∂σ /∂ε at the Gauss integration level of the current it-
eration, which is used to assemble the tangent stiffness matrix.
gigapascals.
At each Gauss integration point, the concrete models uses two
Bischoff and Paixao (2004), Fields and Bischoff (2004), and
damage vectors, WA01 and WA02. The first contains to the up-
Bischoff (2001, 2005) have demonstrated that bar finish does
dated information at the end of the last converged step, whereas
not influence tension stiffening when member response is con-
the second includes the damage variables at the current itera-
sidered. They showed that the bar elastic modulus is the main
tion. For damage in each of the three principal directions, the
parameter that affects the member response in term of force-
model uses a damage index called NCR, defined according to
elongation properties, as is considered in the formulation.
the following:
Furthermore, they indicated that the effect of the surface char-
acteristics (sand coated, ribbed, etc.) have an effect on crack • NCR= 0, undamaged material
spacing and crack width only. They concluded that their for-
mulation (eq. [12]) is independent of the surface characteristics • NCR= 1, damaged material in tension with partial crack
of GFRP bars. However, they also indicated that further tests opening
are needed to confirm their findings. For this reason, with the • NCR= 2, damaged material in tension with reduced tension
adopted formulation, the bond is implicitly considered by the stiffening (or softening), but the crack still active
tension-stiffening factor β(ε) and does not appear in the stress–
strain relationship. • NCR= 3, damaged material in tension and the crack do not
Figure 4 synthesizes the tension-stiffening factors β(ε) for transmit any tensile stress
both steel- and GFRP-reinforced members. It is well depicted
that GFRP-reinforced concrete exhibits greater tension stiffen- • NCR= –1, compression material failure
ing than the steel reinforced concrete. In the subsequent • NCR= –2, compression material failure, but is in compres-
analyses, eq. [11] is used for simulating the steel–concrete in- sion softening
teraction, and eq. [12] is used for GFRP reinforcement.
It is worthwhile to mention that the influence of concrete • NCR= –3, compression material failure with a total loss of
age on β is an interesting concern. To the authors’ knowledge resistance

© 2007 NRC Canada

346 Can. J. Civ. Eng. Vol. 34, 2007

Fig. 5. Description of M’Bazaa (1995) and Chicoine (1997) beams (adapted and revised from M’Bazaa (1995) and Chicoine (1997)).
CFRP, carbon-fibre-reinforced polymer.

4. Applications Table 1. Material properties for M’Bazaa (1995) and Chicoine

(1997) beams. CFRP, carbon-fibre-reinforced polymer.
4.1. General
Three-dimensional finite element analyses are carried out to Properties
examine several aspects of the complex behaviour of reinforced
Concrete
concrete (RC) structural elements internally or externally rein-
fc (MPa) Ec (MPa) εc
forced with FRPs. The constitutive model is validated by means
45 30 200 –0.0035
of a variety of well-known tests, taken from the literature, for
demonstrating the reliability and versatility of the proposed con- Steel
crete model. Basically, the model requires only the compressive Es (MPa) fys (MPa)
strength of concrete and the yield strength of steel or the elastic 200 000 440
modulus and ultimate strength of FRPs. For concrete, actual Composite (CFRP)
values can be provided, but the model computes default prop- E11 (MPa) E22 (MPa) G12 (MPa) ν12 fyCFRP (MPa) εuCFRP
erties in the absence of data. For all the applications presented 82 000 34 750 3 720 0.25 1378 0.0168
in this paper, only default material data were used. It is worth-
while to note that the constitutive concrete model was already
validated by Bouzaiene and Massicotte (1997) and by Massi- Fig. 6. Load–deflection curves, validation.
cotte et al. (2007) for plain concrete and reinforced concrete
members, respectively.
Concrete is modeled using 8-node 3-D solid elements. The
internal reinforcements by steel and FRP bars are modeled using
2-node embedded bar formulation in the isoparametric concrete
elements. With this feature of ABAQUS, the reinforcing bars
are treated as integral parts of the concrete element to determine
the total internal resisting forces that are directly added to those
of concrete. External FRP reinforcements are modeled using
multilayered 4-node shell elements. It is worthwhile to mention
that in ABAQUS STD analyses are performed using 8-Gauss
integration points for solid elements and 4-Gauss integration
points for shell elements, with 2-Gauss integration points over
the thickness. For capturing the post-peak behaviour of concrete
via the STD solution strategy, all analyses were performed using
the Riks arc length method.

4.2. Rectangular beam externally strengthened with

carbon-fibre-reinforced polymer
This section presents the modeled behaviour of rectangular (1997), followed by a numerical investigation presented by
beams strengthened in flexure with bonded unidirectional car- Nitereka and Neale (1999).The geometry and loading condi-
bon fibre laminates. The experimental work was carried out tions reported by Chicoine (1997) are shown in Fig. 5, and
at Sherbrooke University by M’Bazaa (1995) and Chicoine the different material properties are summarized in Table 1.

© 2007 NRC Canada

Nour et al. 347

Fig. 7. Cards showing details of the finite elements model: (a) material failure indices for Chicoine beam; (b) stress card for the
composite; (c) stress card for the longitudinal steel bars and the stirrups.

The beam was strengthened with three CFRP layers for a total Fig. 8. Load–deflection curves, prediction.
thickness of 0.9 mm.
In M’Bazaa (1995) the beam exhibited an early FRP delam-
ination failure mode, a problem eliminated by Chicoine (1997)
by applying U-shaped composite anchors at the ends of the
beam. The anchorage was not modeled, but its presence allows
the assumption in the analysis of a perfect bond between the
reinforced concrete beam and the composite. Using symmetry,
only half of the beam is analyzed.
Figure 6 shows the load–deflection curves of the control rein-
forced concrete beam without composite (M’Bazaa 1995) and
the strengthened beam with CFRPs (Chicoine 1997). In both
tests failure was caused by concrete crushing. For the refer-
ence beam, failure occurred after significant steel reinforcement
yielding, while the second beam exhibited an important increase
of its load-carrying capacity with less ductility compared to the
reference beam without CFRP debonding.
The modeled responses of the two beams reproduce accu-
rately the global experimental measurements and observed fail-
ure mechanisms. First, the significant load-carrying capacity
and additional stiffness provided by the CFRP laminated were ing (in dark gray and light gray in the print version of this article;
accurately simulated by the numerical analysis. At early loading in red and yellow in the Web version) is predominant in the cen-
stages, the same response was obtained for the two beams, an tral portion of the beam, whereas a local compression failure
indication of the small contribution of the composite. However, by crushing appears in the vicinity of the applied loads. This
as cracking progressed and yielding of the steel reinforcement is in accordance with the experimental observation reported by
occurred upon increasing the applied load, the contribution of Chicoine (1997).Figure 7b shows the longitudinal stress card
the composite became more important and its effect was accu- for the composite. Also, it is possible to capture with fidelity the
rately depicted by the analytical model. effect of steel stirrups on the performance of external strength-
Figure 7a shows the material failure indices at failure. Crack- ening. In ABAQUS, as for the longitudinal bars, the stirrups

© 2007 NRC Canada

348 Can. J. Civ. Eng. Vol. 34, 2007

Fig. 9. Description of Grace et al. (1998) beams (adapted and revised from Grace et al. (1998)). GFRP, glass-fibre-reinforced polymer.

are rigorously modeled using truss elements embedded in the Fig. 10. Load–deflection response.
concrete mass (Fig. 7c).
Figure 8 depicts the predicted load–deflection curves for the
additional cases of 1 and 2 layers of CFRP, presented along with
the model response of the two beams of Fig. 6. The analyses
illustrate the effect of the number of layers on the load-carrying
capacity, which corresponds to a decrease of the midspan de-
flection attributed mainly to the additional stiffness brought by
the composite. In the case of two CFRP layers, the numerical
model predicted the same failure mode as for three layers, but
with only one layer the failure is caused by the rupture of the
composite.

4.3. Rectangular beam internally reinforced with

glass-fibre-reinforced polymer
This section compares the model prediction and the exper-
imental observations for a rectangular beam internally rein-
forced with GFRP bars selected from the testing program car-
ried out by Grace et al. (1998). The geometry and loading con- of the GFRP stirrups. Furthermore, they reported that after steel
ditions are illustrated in Fig. 9. The analysis of two beams is reinforcement yielding, failure developed in the control beam
presented: a control beam reinforced with conventional steel with extensive propagation of flexural cracks in the central por-
bars and a composite beam that used GFRP bars for both flex- tion of the span. In spite of the presence of steel stirrups, the
ural reinforcement and stirrups (specimens sb-st and gb-gt in diagonal cracks propagate up to supports, leading to a flexu-
the original paper, respectively). ral failure mode. In the composite beam, the failure developed
The control specimen used high-strength steel reinforcement with the extensive propagation of flexural cracks in the central
with 650 MPa yielding strength and modulus of elasticity of portion of the span, accompanied by a large shear deformation
205 GPa. The mean GFRP bar elastic modulus was 41.8 GPa, attributed to the low elastic modulus of the GFRP stirrups.
and the mean tensile strength was 1100 MPa. For both beams, the model reproduced with fidelity the ob-
Symmetry was used to analyze only half the beam. For ten- served failure mechanisms. Since GFRP reinforcement has no
sion stiffening, eq. [11] is used for the steel reinforcements for yielding point, the behaviour of a beam with composite bars was
modeling the bar–concrete interaction, whereas eq. [12] was quite different than conventionally reinforced concrete mem-
adopted for the analysis of the beam reinforced with GFRP. bers. First, in the vicinity of the loading point, the compres-
Figure 10 illustrates the load–deflection curves correspond- sion stress of concrete at the end of the analysis exceeded at
ing to the reference and composite beams at midspan. In both several Gauss points the compressive strength of concrete, in-
cases, numerical results agree satisfactory with the experimen- dicating that the failure mechanism is of a shear-compression
tal measurements. Grace et al. (1998) observed large deflec- type, which is in accordance with experimental observations.
tions for the composite beam due to the low stiffness of the Furthermore, the GFRP bars and stirrups did not reach their
GFRP bars in comparison with the control beam. The authors tensile strength (fyGFRP = 1100 MPa). The maximum tensile
attributed this difference in deflection to the large deformation stress in the longitudinal bars at failure predicted by the model is

© 2007 NRC Canada

Nour et al. 349

Fig. 11. Description of slab specimens: (a) Benmokrane et al. (2004) slabs; (b) El-Sayed et al. (2005) slabs (adapted and revised from
Benmokrane et al. (2004) and El-Sayed et al. (2005)). GFRP, glass-fibre-reinforced polymer.

about 920 MPa, while the corresponding maximum value in the slabs was selected from the experimental program carried out
stirrups is 260 MPa. The high tensile strains in the GFRP bars by Benmokrane et al. (2004) and the second from the El-Sayed
associated with these stress levels lead to nearly zero tension- et al. (2005) experimental program. The geometry and loading
stiffening stress in the concrete, indicating that concrete is not conditions of the selected slabs are illustrated in Fig. 11.Result
contributing to the resistance mechanism that leads to a shear presentation adopts the same notation for the slabs as reported
failure. in the original papers. For the first group, the model predic-
tions are compared with the experimental results for two slabs:
4.4. One-way concrete slabs reinforced with glass-fibre- specimen S-GGB, reinforced with two double mats of GFRP
reinforced-polymer bars bars, and control specimen S-ST, reinforced with conventional
This section presents the results of tests of two series of one- steel bars. For the second series, the selected slabs SG1, SG2,
way concrete slabs internally reinforced with GFRP bars; these and SG3 refer to specimens having the same GFRP reinforcing
tests were performed at Sherbrooke University. The first set of bars in all directions except the bottom GFRP reinforcement in

© 2007 NRC Canada

350 Can. J. Civ. Eng. Vol. 34, 2007

Table 2. Reinforcing bar properties of Benmokrane et al. Fig. 12. (a) Moment–deflection curves. (b) Load–deflection
(2004) slabs. GFRP, glass-fibre-reinforced polymer. curves.

Reinforcement Modulus of elasticity Tensile strength

bar type (GPa) (MPa)
GFRP 45 775
Steel 200 465 (yielding)
Table 3. Reinforcing bars properties of El-Sayed et al.
(2005) slabs. GFRP, glass-fibre-reinforced polymer.

Reinforcement Modulus of elasticity Tensile strength

bar (GPa) (MPa)
GFRP No. 16 40 597
GFRP No. 22 40 540

the main direction, as shown by Fig. 11b. In both test series,

all slabs were 3100 mm long and 1000 mm wide, and speci-
men thickness of the first and the second groups were 150 and
200 mm, respectively. The main properties of the reinforcing
bars are summarized in Tables 2 and 3. Owing to symmetry,
only one quarter of the slabs is analyzed.
Results obtained from the numerical analysis are presented
in terms of moment–deflection (Fig. 12a) for Benmokrane et
al. (2004) slabs, and in terms of of load–deflection (Fig. 12b)
for El-Sayed et al. (2005) slabs.
For all cases the numerical solutions are globally in good
agreement with the experimental values. For the first group
(Benmokrane et al. 2004) shown in Fig. 12a, the experimental
moment–deflection curve for the control slab (S-ST) is trilin-
ear with a constant strain ascending branch attributed to the
reinforcement strain hardening. The authors observed that after
steel reinforcement yielding, failure developed with extensive
propagation of flexural cracks in the central portion of the span
and towards the supports. For the composite slab (S-GGB), the
experimental moment–deflection curve is bilinear. The authors
indicated that the GFRP slab presented an appreciably higher
load-carrying capacity compared with the control specimen.
Furthermore, they attributed the 20% larger deflection of the For this test series the numerical model correctly predicted
GFRP slab observed at ultimate to the low stiffness of the com- the failure modes of the three slabs. For slab SG1, fc was not ex-
posite material. ceeded at any Gauss points in compression, while in the central
For the composite slab, the model predicted a failure gov- portion of the slab several Gauss points exceeded their tensile
erned by concrete crushing, as several Gauss points in the vicin- strength and were unable to transmit any tensile stress (NCR =
ity of the applied load exhibited compression failure (NCR = 3), indicating that the failure is of diagonal tension. In this case,
–3), while the GFRP bars were still elastic. This prediction is the predicted maximum tensile strain in the GFRP bars is ap-
in accordance with the concrete crushing reported by the au- proximately 1.15%. For slabs SG2 and SG3, fc was exceeded
thors. One notes that the transition between the uncracked and at several Gauss points, as NCR = –3 in the vicinity of the
the fully cracked states and the stiffness in the post-cracking loading point at ultimate, which indicates a failure by concrete
condition are well reproduced by the tension-stiffening models crushing while GFRP was still active. The tensile strains in the
adopted for steel and GRFP reinforcements. longitudinal bars at failure in the model are 0.95% and 0.79%
In Fig. 12b related to the second test series, the measured ex- for SG2 and SG3 slabs, respectively.
perimental load–deflection curves for SG1, SG2, and SG3 slabs
are bilinear. As reported by the authors (El-Sayed et al. 2005), 4.5. Analysis of damaged beams externally reinforced
the flexural stiffness of the slabs increases with the GFRP rein- with carbon-fibre-reinforced polymer
forcement percentage. For the slab SG1, failure develops with This section presents the analysis of damaged beams exter-
extensive propagation of flexural cracks in the central portion nally strengthened in flexure with bonded unidirectional CFRP
of the slab up to the supports. On the other hand, the slabs SG2 laminates. The tests were carried out by Koval and Massicotte
and SG3 exhibited the same behaviour as slab S-GGB (Ben- (2007) at École Polytechnique de Montréal in a project aimed
mokrane et al. 2004), and the observed failure was caused by at studying the improvement of the concrete–CFRP interface as
concrete crushing in the vicinity of load application. well as investigating the performance of repair techniques for

© 2007 NRC Canada

Nour et al. 351

Fig. 13. Description of damaged beams externally reinforced by CFRP (adapted and revised from Koval and Massicotte 2007): (a)
schematic drawing; (b) test set-up for MC1O specimen; (c) damaged zone for MC1O specimen; (d) damaged zone for MB1O specimen;
(e) damage repaired zone.

damaged beams. This study follows a previous experimental in- damaged beam MA1O is considered as the reference. Beams
vestigation (Folcher et al. 2003) in which premature debonding MB1O and MC1O have initial imperfections reproducing the
due to the flexional rigidity discontinuity between the damaged effect of damage that could have been caused by an impact
zone and the sound concrete was observed. The main scope of or corrosion of prestressing strands. To simulate the damaged
this numerical example is to examine the ability of the model zone, the formwork was fabricated to keep the central bottom
to capture local phenomena such as the influence of local re- part of the beams without concrete.The longitudinal T13 strands
pair of damaged zones on the global behaviour of the beams. were cut in the centre to simulate a 150 mm discontinuity. Con-
The geometry and loading conditions are shown in Fig. 13, and crete surface in the repaired zone was roughened according to
material properties are summarized in Table 4. construction practice. Figure 13 illustrates the dimensions of
In this application, three RC beams are analyzed. As shown in the damaged zone. Two repair methods were considered. In
Fig. 13, these beams are externally strengthened by two sheets beam MB1O additional reinforcement was added in the dam-
of CFRP laminate (2900 mm × 50 mm × 1.2 mm). The un- age zone to eliminate the flexional rigidity discontinuity. Each

© 2007 NRC Canada

352 Can. J. Civ. Eng. Vol. 34, 2007

Table 4. Material properties for Koval and Massicotte Fig. 14. Load–deflection curves.
(2007) beams.

Properties
Old concrete
fc (MPa) Ec (MPa) εc ν ft (MPa)
52 34 887 –0.003 0.23 2.1
Reparation concrete
fc (MPa) Ec (MPa) εc ν ft (MPa)
54 31 000 –0.0035 0.22 2.3
10M steel bars
Es (MPa) fys (MPa)
200 000 400
T13 strand
Es (MPa) fys (MPa)
205 600 1860
Composite (CFRP)
E (MPa) ν fyCFRP (MPa) εuCFRP
207 000 0.25 2400 0.016
concrete properties were defined using eq. [8], which accounts
for the presence of strands in both specimens and additional
cut T13 strand was overlapped on a length of 150 mm by one reinforcement for specimen MB1O.
10M reinforcing steel bar. Moreover, a 50 mm × 50 mm steel The results presented in Fig. 14 illustrate the comparison be-
wire mesh with 9 mm2 wires was placed around the overlapping tween the two repair techniques. For the reference beam MA1O,
bars. For comparison with the Folcher et al. (2003) study, the the measured load-carrying capacity was 90 kN, while for the
repair concrete in the damaged zone of MC1O specimen was repaired beams MB1O and MC1O, the measured strength was
not reinforced. 71 and 50.3 kN, respectively. The tests clearly showed that the
The observed failure mode for the two repaired beams was repairing technique adopted for specimen MB1O is more ef-
initiated by the debonding of the repair concrete along the in- ficient compared with that of specimen MC1O, which exhib-
clined faces of the repaired zones (Fig. 13a), which was then ited premature failure similar to that observed by Folcher et al.
followed by the composite debonding. For the reference speci- (2003).
men, failure was governed by the CFRP debonding. Therefore, Using symmetry, only half of the beams were modeled. For
it is important to correctly model the concrete–composite as the three beams, the numerical solutions are in good agreement
well as old-to-new concrete contacts. with the experimental results for predicting the maximum load
In the model, all surface contacts are modeled withABAQUS/ and failure mechanism. The analytical obtained results showed
Standard, which considers contact between two bodies in terms that the model captured with fidelity the influence of local repair
of two surfaces that may interact. The order in which the two sur- of the damaged zones on the global behaviour of the beams.
faces are specified defines the slave and master surfaces. Two al-
gorithms are available: the small-sliding contact algorithm and 4.6. Concrete columns confined with fibre composite
the finite-sliding contact algorithm. For small-sliding contact sheets
problems the contact area is calculated from the undeformed The scope of this section is to examine the model perfor-
shape of the model and is thus kept constant throughout the mance for predicting the increase in strength and the associated
analysis. In this formulation, contact pressures are calculated ductility of concrete columns confined with fibre composite
according to an invariant contact area. The modeled behaviour sheets. Two test series on circular and square concrete columns
is different from that in finite-sliding contact problems, where strengthened with FRP are analyzed.
the contact area and contact pressures are calculated according The first test series deals with circular column specimens
to the deformed shape of the model. tested by Demers and Neale (1994, 1999) and the associated nu-
For this application the analyses performed clearly indicated merically investigation by Deniaud and Neale (2006). The spec-
that the finite sliding formulation is the most appropriate for imens consisted of 300 mm long and 150 mm diameter plain
reproducing the observed behaviour. In the Interaction module concrete cylinders strengthened with CFRP sheets with an av-
of ABAQUS, the strategy adopted for the concrete–composite erage thickness of 0.34 mm and a tensile strength of 380 N/mm,
contact modeling uses the beam as the master surface and the with a corresponding ultimate strain of 1.6%.
composite as the the slave surface. Similarly, for the concrete– The results of the analyses are compared in Fig. 15 to test
concrete contact, the old concrete is selected as the master measurements corresponding to two types of concrete. In all
surface and the repair concrete as the slave surface. Contact cases the numerical solution agrees well with the experimental
properties are specified for tangential and normal behaviour. data. Based on the comparison of numerical predictions using
Another important aspect of modeling the damaged beams various concrete models reported by Deniaud and Neale (2006),
is related to the definition of the tension-stiffening properties the results obtained herein are consistent with the experiment
of the repair concrete. For beams MB1O and MC1O, different for all types of concrete and compare advantageously with the

© 2007 NRC Canada

Nour et al. 353

Fig. 15. Load–strain curves response for plain concrete circular Fig. 16. Load–strain curves for square reinforced concrete column
columns: (a) 32 MPa concrete; (b) 44 MPa concrete. CFRP, specimens S1 and S3.
carbon-fibre-reinforced polymer.

5. Conclusions
The scope of this paper was to investigate by finite elements
the nonlinear response of concrete structures reinforced with
internal and external FRP. A general and portable constitutive
concrete model is used. It is implemented as a user-defined
subroutine at Gauss integration point level into the general-
purpose finite element software ABAQUS.
This paper demonstrated the ability of the proposed concrete
model to correctly simulate different kind of composite rein-
forcements (internal and (or) external), and its reliability was
confirmed by means of several tests covering various types of
structural elements taken from the literature. The numerical
predictions obtained for load–displacement responses and fail-
ure mechanisms agreed well with the experimental values and
observations. This study is evidence that this concrete model
can be used with confidence to simulate and estimate load-
carrying capacity of reinforced concrete structures, reinforced
or not with FRPs. The model can be used in conjunction with
experimental testing in research environments or as a predictive
tool for actual structures.

elastoplastic model predictions recommended by Deniaud and Acknowledgements

Neale regarding the specimen with three carbon sheets. The authors thank the Network of Centres of Excellence ISIS-
In the second test series, RC square columns specimens S1 Canada and the Natural Sciences and Engineering Research
and S3 tested by Pessiki et al. (2001) were selected. Both speci- Council of Canada (NSERC) for their financial support.
mens are 1830 mm high and 457 mm wide, reinforced with eight
22 mm longitudinal steel bars and 9.5 mm steel stirrups spaced
at 356 mm. Specimen S1 had no composite and is considered References
as the control column, whereas specimen S3 was jacketed with Bažant, Z.P., and Oh, B.H. 1983. Crack band theory for fracture of
three plies of GFRP composite. Each GFRP ply had a tensile concrete. Materials and Structures, RILEM, 16: 155–177.
strength of 383 N/mm, an elastic modulus of 21.6 kN/mm, and
Benmokrane, B., Rahman, H., Mukhopadhyaya, P., Masmoudi, R.,
an ultimate strain of 1.9%. Concrete had a compressive strength
Chekired, M., Nicole, J.F., and El-Safty, A. 2000. Use of fibre rein-
of fc = 26.4 MPa with εc = 0.0021.
forced polymer reinforcement integrated with fibre optic sensors for
Figure 16 illustrates that the model predictions are consis- concrete bridge deck slab construction. Canadian Journal of Civil
tent with the experimental data. The model reproduced cor- Engineering, 27: 928–940.
rectly the increase in strength caused by the presence the three Benmokrane, B., El-Salakawy, E., Cherrak, Z., and Wiseman, A. 2004.
GFRP plies. Failure in column specimen S3 was initiated by Fibre reinforced polymer composite bars for the structural concrete
the rupture of the composite in tension; this was also correctly slabs of a Public Works and Government Services Canada parking
predicted by the model. garage. Canadian Journal of Civil Engineering, 31: 732–748.

© 2007 NRC Canada

354 Can. J. Civ. Eng. Vol. 34, 2007

Bischoff, P.H. 2001. Effects of shrinkage on tension stiffening and Folcher, D., Massicotte, B., and Labossière, P. 2003. Étude du renforce-
cracking in reinforced concrete. Canadian Journal of Civil Engi- ment passif d’une poutre en béton précontraint. Rapport EPM/GCS-
neering, 28: 363–374. 2003-07, Département des génies civil, géologique et mines, École
Bischoff, P.H. 2005. Reevaluation of deflection prediction for concrete Polytechnique de Montréal, Montréal, Que.
beams reinforced with steel and fiber reinforced polymer bars. Jour- Grace, N.F., Soliman, A.K., Abdel-Sayed, G., and Saleh, K.R. 1998.
nal of Structural Engineering, ASCE, 131(5): 752–767. Behaviour and ductility of simple and continuous FRP reinforced
Bischoff, P.H., and Paixao, R. 2004. Tension stiffening and cracking beams. Journal of Composites for Construction, ASCE, 2(4): 186–
of concrete reinforced with glass fiber reinforced polymer (GFRP) 194.
bars. Canadian Journal of Civil Engineering, 31: 579–588. Hibbitt, H.D., Karlsson, B.I., and Sorensen, E.P. 2004.ABAQUS user’s
Bouzaiene, A., and Massicotte, B. 1997. Hypoelastic tridimensional manual, version 6.5. Hibbitt, Karlsson, & Sorensen, Inc., Pawtucket,
model for non proportional loading of plain concrete. Journal of R.I.
Engineering Mechanics, ASCE, 123(11): 1111–1120. Hsieh, S.S., Ting, E.C., and Chen, W.F. 1982. A plasticity-fracture
CEB-FIP. 1993. CEB-FIP model code 1990 (MC-90), design code. model for concrete. International Journal of Solids and Structures,
Comité Euro-International du Béton (CEB), Thomas Telford Ser- 18: 181–197.
vices Ltd., London, UK. Koval, V., and Massicotte, B. 2007. Utilisation des matériaux compos-
Chicoine, T. 1997. Conception et analyse des ancrages sur les poutres ites pour le renforcement de poutres de ponts précontraintes endom-
renforcées en flexion à l’aide des matériaux composites. M.A.Sc. magées. Département des génies civil, géologique et mines, École
thesis, Département de génie civil, Université de Sherbrooke, Sher- Polytechnique de Montréal, Montréal, Que. Rapport: In press.
brooke, Que. Massicotte, B., Elwi, A.E., and MacGregor, J.G. 1990. Tension-
Crisfield, M.A., and Wills, J. 1989. Analysis of R/C panels using dif- stiffening model for planar reinforced concrete members. Journal
ferent concrete models. Journal of Engineering Mechanics, ASCE, of Structural Engineering, ASCE, 116: 3039–3058.
115: 578–597. Massicotte, B., Nour, A., Ben Ftima, M., and Yildiz, E. 2007. EPM3D
Darwin, D., and Pecknold, D.A. 1977. Nonlinear biaxial stress–strain – A user-supplied constitutive model for the nonlinear finite element
law for concrete. Journal of the Engineering Mechanics Division, analysis of reinforced concrete structures. École Polytechnique de
ASCE, 103: 229–241. Montréal, Montréal, Que. Report SR07-10.
Demers, M., and Neale, K.W. 1994. Strengthening of concrete columns M’Bazaa, I. 1995. Reinforcement en flexion des poutres en béton armé
with unidirectional composite sheets. In Developments in Short and à l’aide des lamelles en matériaux composites : optimisation de la
Medium Span Bridge Engineering: Proceedings of the 4th Interna- longueur des lamelles. M.A.Sc. thesis, Département de génie civil,
tional Conference on Short and Medium Span Bridges, Halifax, Université de Sherbrooke, Sherbrooke, Que.
N.S., 8–11 August 1994. Edited by A.A. Mufti, B. Bakht, and L.G. Nitereka, C., and Neale, K.W. 1999. Analysis of reinforced concrete
Jaeger. Canadian Society for Civil Engineering, Montréal, Que. pp. beams strengthened in flexure with composite laminates. Canadian
895–905. Journal of Civil Engineering, 26: 646–654.
Demers, M., and Neale, K.W. 1999. Confinement of reinforced con- Oliver, J. 1989. A consistent characteristic length for smeared cracking
crete columns with fibre-reinforced composite sheets – an experi- models. International Journal for Numerical Methods in Engineer-
mental study. Canadian Journal of Civil Engineering, 26: 226–241. ing, 28: 461–474.
Deniaud, C., and Neale, K.W. 2006. An assessment of constitutive Pessiki, S., Harries, K.A., Krestner, J.T., Sause, R., and Rices, J.M.
models for concrete columns confined with fibre composite sheets. 2001.Axial behaviour of reinforced concrete columns confined with
Composite Structures, 73(3): 318–330. FRP jackets. Journal of Composites for Construction, ASCE, 5(4):
El-Salakawy, E., Benmokrane, B., and Desgagné, G. 2003. Fibre- 237–245.
reinforced polymer composite bars for the concrete deck slab of Rachid, Y.R. 1968. Ultimate strength analysis of reinforced concrete
Wotton Bridge. Canadian Journal of Civil Engineering, 30: 861– pressure vessels. Nuclear Engineering and Design, 7: 334–355.
870. Saenz, L.P. 1964. Discussion of “equation for the stress-strain curve
El-Sayed, A., El-Salakawy, E., and Benmokrane, B. 2005. Shear of concrete” by P. Desayi and S. Krishnan. Journal of the American
strength of one-way concrete slabs reinforced with fibre-reinforced Concrete Institute, 61: 1229–1235.
polymer composite bars. Journal of Composites for Construction, Willam, K.J., and Warnke, E.P. 1975. Constitutive model for triaxial
ASCE, 9(2): 147–157. behaviour of concrete. Proceedings of the IABSE Seminar on Con-
Elwi, A.E., and Murray, D.W. 1979. A 3D hypoelastic concrete consti- crete Structures Subjected to Triaxial Stresses, ISMES, Bergamo,
tutive relationship. Journal of the Engineering Mechanics Division, Italy. International Association for Bridge and Structural Engineer-
ASCE, 106: 1383–1403. ing (IABSE), Zürich, Switzerland. Rapports des commissions de
Feenstra, P.H., and De Borst, R. 1996. A composite plasticity model travail, 19: 1–30.
for concrete. International Journal of Solids and Structures, 33(5): Winkler, B., Hofstetter, G., and Lehar, H. 2004. Application of a con-
707–730. stitutive model for concrete to the analysis of a precast segmental
Fields, K., and Bischoff, P.H. 2004. Tension stiffening and cracking of tunnel lining. International Journal for Numerical and Analytical
high-strength reinforced concrete tension members. ACI Structural Methods in Geomechanics, 28: 797–819.
Journal, 101(4): 447–456.

© 2007 NRC Canada