Mechanics of Materials 35 (2003) 817830

www.elsevier.com/locate/mechmat

Orthotropic modelling of alkali-aggregate reaction in

concrete structures: numerical simulations
Bruno Capra *, Alain Sellier
Laboratoire de G

enie Civil et Urbanisme, 2 rue A. Einstein Cit

e Descartes, Champs-Sur-Marne 77420, France
Received 17 November 2001; received in revised form 15 March 2002

Abstract
Alkali-aggregate reaction (AAR) is dicult to model due to the random distribution of the reactive sites and the
imperfect knowledge of these chemical reactions. A new approach, based on a probabilistic description of the main
physical parameters of concrete and AAR, allows to simulate orthotropic swelling of concrete subjected to AAR. The
concrete is modelled like a damageable material having elastic and inelastic strains. AAR is modelled by a global
kinetics including temperature and humidity eects. The coupling between mechanics and AAR makes it possible to
simulate tests carried out on concrete specimens. The results, obtained in terms of axial and transverse strains versus
time, for various levels of axial loading, show a good agreement between experimental results and model response. A
nite element implementation of the model is then applied to reinforced concrete beam subjected to dierent gradients
of humidity on its height. The results are achieved in terms of damage elds, non-homogenous swelling and force
displacement curves.
2002 Elsevier Science Ltd. All rights reserved.
Keywords: Orthotropic; Modelling; Alkali-aggregate reaction; Damage; Numerical simulations

1. Introduction
A lot of structures, such as dams or bridges, are
suering from deteriorations induced by alkaliaggregate reaction (AAR) that impair durability
and might also aect the safety of installations.
AAR are internal chemical reactions that appear
in concrete between certain amorphous or not well
crystallised aggregates and the alkaline pore solution. The result is a more or less crystallised silico-

Corresponding author. Tel.: +33-1-6473-0491; fax: +33-16473-0521.

E-mail address: capra@univ-mlv.fr (B. Capra).

alkaline product that can exerts pressures on the

surrounding matrix. AAR induces concrete expansion and generally leads to loss of strength and
cracking. There is an increasing interest in the
structural response to AAR eects. Need for, and
nature of potential intervention must be supported
by knowledge of the spatial distribution and intensity of swelling, stress state resulting from it,
and long term evaluation of the swelling process.
Such a reaction is dicult to model accurately
because of the complexity of the phenomena:
random localisation of the reactive sites, imperfect
knowledge of the reaction mechanisms, etc. In
order to perform structural computations, reactions
have been modelled within a phenomenological

0167-6636/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0167-6636(02)00209-0

818

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

approach framework by taking into account the

following major parameters: the kinetic of the reaction, the temperature, the moisture and the
stress state. These variables will then govern the
raising pressure which represents the mechanical
eects of the reactions within concrete. Furthermore, concrete modelling is based on a physical
description of the crack opening probability. This
probability is related to the damage rate of the
material. The constitutive model allows then to
describe an orthotropic decreasing of the elastic
properties of concrete and the residual swelling,
under both AAR and mechanical loads. Numerical simulations have been performed to compare
the model response to the experimental results of
tests carried out at the Laboratoire Central des
Ponts et Chauss
ees. The behaviour of specimens
subjected to dierent loadings has been properly
reproduced in terms of residual strains and orthotropic swelling. A nite element implementation of the model allows to perform numerical
simulations of a reinforced concrete subjected to
both AAR and mechanical loadings. The results
are achieved in terms of damage index, swelling
elds and forcedisplacement curves.

2. Mechanical modelling of concrete

Before describing AAR modelling, we are going
to detail the approach used to describe concrete
behaviour. This approach is based on a physical
description of deterioration of concrete by creation
of surface-area discontinuities.
2.1. Eective stresseselastic strains relations
The elastic part of the behaviour law of concrete is based on orthotropic damage theory (Lemaitre and Chaboche, 1988). The state of the
material, in term of stiness, is dependent of a
phenomenological variable d. This variable is
dened as the ratio between the surface of discontinuities (voids, cracks, . . .) and the total surface of the specimen. The damage coecient d
vary between 0 (sound material) and 1 (cracked
material). The use of a damage variable leads directly to the denition of apparent and eective

stresses. The apparent stress is the usual stress directly applied on the whole section of the material
while the eective stress is the one which is applied
on the remaining sound section. The relation between apparent stress rapp , eective stress reff and
damage coecient d is dened by the following
equation:
reff

rapp
1d

Obviously, the previous description of damage is

valid in any space direction represented by a normal vector.
The behaviour law of concrete presented here is
based on the orthotropic formulation of the previous considerations. The elastic strains eei and the
apparent stresses rappi are calculated from the effective stresses reffi as follows (Sellier et al., 2001):
8
>
>
ee 1 r 1 r  t0 r reffk
>
< i Ei effi E0 effi E0 effi


1di
1
1
1
with:
E Pi 1  Pi

Ei
1di
>
0
>
>r
:
r 1  d r 1  d 
appi

effi

effi

In these expressions, the sign or ) denotes

the sign of the principal eective stress ( for
compression, ) for tension). The indices i, j and
k represent the three cartesian directions. As the
law is orthotropic, the damages coecients di may
be calculated for any direction among the three
principal directions in the stress operating space.
E0 and m0 are respectively the Young modulus and
the Poisson ratio for the initial and undamaged
material. Pi lc =li where lc is an internal length
considered in order to avoid the numerical strain
localisation problem and li is the nite element
dimension in the i direction follows (Sellier et al.,
2001). These coecients act on the post-peak part
of the behaviour law by imposing an energy dissipation independent of the characteristic size of
the element calculated in direction i (Barzegar
and Maddipudi, 1997). di mindi ; dipeak where
dipeak is the damage at the peak of the behaviour
law, represented by the curve stress versus strain.
The expression of elastic strain assumed that energetic dissipation is proportional to the volume
considered until the peak and to the surface area of
discontinuities after the peak, in accordance with

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

experimental results (Jansen and Shah, 1997). The

guiding principles of mechanical modelling being
posed, we now will study the modelling of AAR
before establishing the coupling by the use of the
damage coecients. The reader will nd more
details on this orthotropic rotating smeared crack
model in the reference (Sellier et al., 2001).

819

3.2. Gel pressure

3. Modelling of AAR

The pressure generated in the gel Pg is proportional to the volume of gel created and its rate
of connement. Let Vg be the volume of gel created
during the reactions, Veaar the volume of the cracks
occupied by the gel, VP0 the volume of porosity
connected to the reactive sites likely to be used as
an expansion vessel for gel. The pressure developed by the reactions is expressed by:

3.1. Self-stresses generated by AAR

Pg KhVg  VP0  Veaar i

where

When AAR occurs, a gel-like product of the

reactions is generated which takes more space than
the initial reagents. Without any external load, a
concrete specimen, subjected to AAR, shows a
swelling. Thus, an internal self-equilibrated stress
state exists inside the material. Fig. 1 schematically
represents the dierent stresses generated by the
gel inside concrete subjected to AAR.
The self-stress generated by AAR in the direction i, on uncracked concrete, must balance the
pressures generated in the cracks lled with products of reaction. Let Pg be the average pressure
exerted by those products on the cementitous
matrix, and rgi the stress induced on the undamaged material. The equilibrium equation leads to
the following relation, valid for the three principal
directions i:
rgi 

Pg
1  Pfaari

Pfaari represents the probability of cracking due to

AAR in direction i, in extenso, the fraction of
material where the self-stress is greater than the
cohesion of the material. Thus, 1  Pfaari is the
fraction of uncracked material, in direction i, on
which rgi is exerted.

Fig. 1. Self-stresses generated by AAR.

hX i

X
0

if X > 0
if X < 0

The volume of the cracks occupied by AAR Veaar

is supposed to be proportional ke to the total
volume of created cracks (related to the cubic dilatation tr ), to the quantity of gel likely to penetrate there Vg and to the pressure Pg which will
allow the gel penetration in these cracks:
Veaar ke Pg Vg tr e

Porous volume accessible to AAR VP0 , is supposed to be proportional to the initial porosity of
the concrete P0 :
VP0 kp Pg Vg P0

These expressions make it possible to express the

gel pressure by


Vg
Pg K
6
1 K Vg ke tr e kp P0
where K is a stiness constant of gel, to be determined by inverse analysis.
3.3. AAR global kinetics
Gel pressure depending on gel volume, it is now
necessary to express the evolution of this volume,
that is to say a kinetics of reactions. AAR global
kinetics stands for [Na2 Oeq ], total concentration of
the pore solution in alkaline sodium equivalent
Na2 O 0:658K2 O. Let us regard this concentration as the chemical tracer of AAR and suppose
that the following reaction represents the consumption of alkalis by AAR (Capra, 1997):

820

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

Na2 Oeq ! Na2 Olinked

to gel

Let At being the percentage of reacted alkalis

during time t:
At

Na2 Oeq 0  Na2 Oeq t

Na2 Oeq 0

It is supposed that the volume of gel created during

time t is proportional to the percentage of reacted
alkalis. By making the assumption that rst-order
kinetics is representative of the reactions, and
that an Arrhenius law can apply to AAR, one obtains the following form for the gel pressure Pg t
(Capra et al., 2000):


Pg lim t
Pg t K
8
1 Pg lim tke tr e kp P0
where
Pg lim t Pg1lim RHm

 
Ea
1  exp  k0 exp 
t
RT
9
Pg1lim is the maximum gel pressure which can be
reached in the concrete at an innite time and for a
saturated relative humidity RH 100%. Ea is
the activation energy of the reaction, T the absolute temperature, k0 a global kinetic constant of
the chemical processes related to the reactions. The
eects of the relative humidity (RH) on AAR are
introduced in a global way by a reducing function
RHm of the gel pressure where m 8. The value
of m drags a fast decrease of swelling when RH
is under 100% that is in accordance with experimental data. It is well known that under about
60% of RH, no more AAR are observed. Fig. 2
shows the evolution of this reducing function of
swelling.

Fig. 2. Inuence of relative humidity on swelling.

PfiAAR

1  exp

1

mr

hri rgi i
rur
i

mr 
10

The choice of a Weibulls law for PfiAAR is more

detailed in (Sellier et al., 1999). rur
i is the cohesion
of Weibull for AAR, in the direction i; mr is the
exponent of Weibull for AAR, supposed homogeneous and isotropic.
3.5. Probability of cracking due to tensile and
compressive stresses
In a way identical to the probability of cracking
due to AAR, one can dene a probability of
cracking due to a mechanical loading of tension
Pfit or compression Pfic . The following relations express the fact that the cracks from mechanical origin can have two sources: direct and
indirect. A tensile stress has a direct eect on the
opening of a crack in the loading direction. On the
other hand, a compressive stress generates selfstresses of traction perpendicular to the loading
direction which will crack the concrete. The preceding remarks lead to the following equations:
!

m
1
hr
i
i
Pfit 1  exp 
11
m
rut
i

3.4. Probability of cracking related to AAR

Pfic

The auto-stress rgi , generated by AAR, is superimposed locally on the stresses due to a mechanical loading of concrete, ri . The probability of
cracking due to AAR, in the principal direction
i, PfiAAR , then reads to

0

B 1 B ri
B
B
1  exp B   B
@ m @

q 1m 1
C  2 hrj i 2 hrk i 2 C C
C C
C C
A A
ruc
i

12

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

uc
rut
i , ri are the Weibulls cohesions in direct tension
and in indirect tension (perpendicular compression) respectively. C  is the coecient of variation in compression. m , m are the exponents
of Weibull in tension and compression respectively.

4. Damage of concrete: coupling AAR/mechanics

4.1. Damage coecients
The damage coecients in tension di and in
compression di used in Eq. (1) are calculated according to the probabilities of cracking in the
following way:
di 1  1  Pfit 1  Pfic 1  PfiAAR
di 1  1  Pfjc wj 1  Pfjt wj 1  PfjAAR wj
1  Pfkc

wk

1  Pfkt

wk

1  PfkAAR

wk

13
These damage coecients are deduced from the
dierent probabilities previously calculated according to the weakest link theory. These coecients express the dierent sources of damage:
direct and indirect tension, damage due to AAR.
The weighing functions w allow possible crack
reclosing in case of cyclic loadings.

821

comes signicant, this phenomenon is modelled by

the following equation:
eAAR
eAAR
i
0

PfiAAR
1  PfiAAR

15

is a material parameter to be deterwhere eAAR

0
mined. The use of the relations between probability of cracking and damage on the one hand,
probabilities of cracking and swelling on the other
hand, makes it possible to link AAR swelling and
tension damage:
eAAR
eAAR
i
0

di
1  di

16

Experimental results (ISE, 1992; Larive, 1997)

show a decrease of the mechanical characteristics
of concrete according to the swelling rate. Fig. 3
presents the experimental data used to determine
eAAR
by the use of Eq. (16). The tting leads to:
0
0:3%
eAAR
0

17

This parameter seems to be relatively constant

even for tests carried out on dierent concrete
mixes.
4.4. Anelastic strains of mechanical origin
In this section, we assume that the anelastic
strains of mechanical origin are expressed by


epl
i ei ei

18

4.2. Total strains

The total strains ei are supposed to be the sum
of the elastic strain eei , the mechanical anelastic
strains epl
i (without AAR) and of the strains due to
AAR eAAR
:
i
AAR
ei eei epl
i ei

14

We assume that the mechanical anelastic strains

are independent of AAR. The relations between
elastic strains and eective stresses have been
previously detailed in Eq. (1).
4.3. Anelastic strains due to AAR
The anelastic opening of cracks due to AAR
evolves in an asymptotic way when cracking be-

Fig. 3. Evolution of the mechanical characteristics versus AAR

swelling rate.

822

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

Indice and  stands for compression and

tension respectively. As for elastic behaviour,
probabilities of cracking are introduced, leading
to the following expressions of the anelastic
strains:

e
i e0 Pi wi


e
i e0

Pfi  Pfi
1  Pfi

Pfj
Pfi
w  tp
w
 i
1  Pfi
1  Pfj j
!!
Pfk

kw
1  Pfk k

19

20

The signication of the various terms is not described here but the reader will nd more details
in (Sellier et al., 1999). Figs. 4 and 5 compare experimental results and numerical simulations of
biaxial behaviour of concrete subjected to an uniaxial loading.
An example of the stress/strain relationship for
cycling tension/compression loading path is presented in Figs. 6 and 7.
The reclosure function used in the behaviour
law allows to progressively close cracks that have
been created in tension and to recover the undamaged stiness in compression. In the same way,
when unloading is applied in compression, cracks
re-open them and recover the previous damaged
stiness and stress peak in tension. The progressive
damage of the material can be seen by the decrease
of the Youngs modulus during the cycles in compression.

Fig. 4. Experimental results of the behaviour of concrete in

pure compression.

Fig. 5. Numerical simulations of the behaviour of concrete in

pure compression.

Fig. 6. Cycling loading path.

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

823

Fig. 7. Stressstrain curve obtained in uniaxial loading path (m 1:5, C 0:5, ru 2:03 MPa, C 6:94, E0 30,000, m0 0:2,
Rc 38 MPa: compressive strength of concrete).

5. Numerical simulations of AAR

5.1. Swelling under uniaxial loading
Tests of swelling under uniaxial loading, with
measurements of axial and transverse strains, were
carried out at the Laboratoire Central des Ponts et
Chauss
ees (Larive, 1997). The model presented
previously was applied to these tests. Figs. 811
present the response of the model compared with
the experimental points for various loadings: free
swelling, 5, 10 and 20 MPa of uniaxial compression. The model shows a good agreement with the
tests except for transverse swelling under 10 MPa.
It is to note that the model is not only capable to
represent the axial behaviour of concrete aected
by AAR but also the transversal one for dierent
loading cases.
5.2. Numerical simulation of a beam subjected to
AAR
5.2.1. Data of the simulation
The previous model of concrete subjected to
internal swelling has been implemented in the nite
element code CASTEM 2000 developed by the CEA
(Commissariat 
a lEnergie Atomique) in France. In

Fig. 8. Comparison between experimental points and numerical modelling (free swelling r 0 MPa).

order to test the model, the structures of Figs. 12

and 13 has been computed.
The parameters of the problem are the following ones:
L 3 m h 0:5 m
b 0:25 m d 0:45 m
Re-inforcement: 2 HA 25 9:81 cm2

824

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

Fig. 12. Re-inforced beam subjected to AAR.

Fig. 9. Comparison between experimental points and numerical modelling (compressive loading: 5 MPa).

Fig. 13. Typical cross-section of the beam.

Poissons coecient: ms 0:3

Volumetric mass: qs 7850 kg/m3
Fig. 10. Comparison between experimental points and numerical modelling (compressive loading: 10 MPa).

Fig. 11. Comparison between experimental points and numerical modelling (compressive loading: 20 MPa).

The principal parameters of the behaviour of

the two materials are:
Steel bars:
Elastic behaviour
Youngs modulus: Es 210,000 MPa

Concrete:
Non-linear orthotropic damageable behaviour
Initial Youngs modulus: Ec 42,000 MPa
Poissons coecient: mc 0:2
Volumetric mass: qc 2500 kg/m3
5.2.2. Loading and humidity gradient
The dead-weight of concrete is taken into account. The beam is subjected to no external mechanical loading, except the dead-weight. The
concrete of the beam is subjected to AAR. Fig. 14
shows the evolution of the pressure developed by
the gel created versus time, following Eq. (9), for
two RH (60% and 100%) which represent the
maximum amplitude on swelling. Numerical values have been estimated using experimental tests
carried out by Larive (Larive, 1997).
At the present time, there are not a lot of studies
concerning the eects of RH on swelling due to
AAR, even the threshold of about 50% or 60%
under which RH is not important enough to generate swelling is approximate. Nevertheless, it appears that like temperature gradient, humidity

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

825

Fig. 16. Finite element mesh and boundary conditions (four

nodes elements).

Fig. 14. Evolution of the gel pressure versus time for dierent
RH of concrete.

gradient can be prejudicial by generating dierential swelling.

Fig. 15 shows the gradient of humidity that
have been applied to the beam.
5.2.3. Finite element simulation of the beam
The nite element mesh used for the simulations
in plane stresses of the half-beam and the boundary conditions are represented in Fig. 16.
Fig. 17 show the deformed mesh of the half
beam at 400 days.
The previous gure shows that without any
external loading, AAR is able to create important
deection. Without gradient of humidity (RH
100% on all the beam), the curvature of the beam
would have been reverse because of the steel bars
that prevent concrete to swell. The gradient of
humidity, i.e. the gradient of swelling, is able to

Fig. 15. Gradient of humidity applied to the beam (top: 60%,

bottom: 100%).

Fig. 17. Deformed shape of the beam at 400 days (amplitude:

300).

change the curvature of the beam, it is an important environmental factor that highly modify the
way a structure will be damaged. Such a phenomenon is often observed on real structures and
can be reproduced in laboratory (Swamy and AlAsali, 1989).
The modelling of AAR takes into account an
other major environmental parameter which is the
stress state or more generally the connement. Fig.
18 show the evolution of the damage coecients
dxx and dyy due to AAR on the beam for a constant
humidity of 100%.
The more damaged part of the beam is on the
top left corner (dxx eld) because it is the least
enclosed part of the beam. On the contrary, the
concrete near the end of the steel bar is the more
conned and then, show a minimum of swelling.
In the middle of the beam, far enough from the
support, the classical distribution of normal
stresses leads to a damage more important at the
bottom of the beam (tensile stress) than at the top
(connement by a compressive stress). More generally, we can note that the damage eld, or in the

826

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

Fig. 18. Damage elds dxx and dyy due to AAR (t 400 days).

same way the swelling due to AAR, are heterogeneous. On the dyy eld, the level of connement is
less important than in x direction except near the
support of the beam, then the damage is more
homogenous. Further simulations need to be performed in order to take into account the eect of a
temperature eld coupled with a gradient of humidity and higher gel pressure.
Fig. 19 presents the evolution of tensile strength
in steel bars Nst along the bar during time.
At the beginning, the tensile strength is given by
the dead weight of the beam; when AAR occurs,
this strength evolves with gel pressure. On real
structure, when the swelling is very important,
AAR are sometimes able to reach and pass the
limit of elasticity of steel. Although the simulation
takes into account an elastic behaviour for steel,
the pressure reached would not be able to plastify
steel bars.
5.2.4. AAR inuence on the mechanical response of
the beam
The previous simulation concerned a beam
without any external mechanical loading. In order
to measure the inuence of AAR on the mechanical stiness of the beam, the following (Fig. 20)
simulations were performed.

For the reactive concrete, AAR is simulated

during 400 days and then the beam is submitted to
a 3 points bending test. In addition, two dierent
humidity elds have been applied: the rst one is
constant and equals to 100%, the second one is the
gradient of 60% at the top to 100% at the bottom
of the beam as previously shown.
Fig. 21 present the principal damage eld on the
beam for the three simulations (sound concrete,
gradient of humidity and constant humidity).
The gures show that AAR has an inuence on
the crack pattern that appears with loading. The
beam without AAR shows a classical pattern of
cracks which developed from the middle to the
edge of the beam. The spacing between cracks is
dependent on the material properties of steel and
concrete. The simulation of a beam subjected to
AAR with a gradient of humidity shows a nonclassical result because the rst crack appears in
the middle of the beam but jumps to the neighbouring element and then continues to propagates.
Another dierence compared to the previous simulation is the spacing between cracks. As the mechanical properties of concrete are aected by
AAR (loss of tensile strength and Youngs modulus), the crack spacing is also changed. This observation is more evident on the third simulation

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

827

Fig. 19. Evolution of tensile strength along steel bar during time.

Fig. 20. Numerical simulations of a 3 points bending test with and without AAR.

where the whole beam is aected by AAR whereas

in the second, due to the power law for humidity,
only the bottom of the beam is signicantly deteriorated by AAR. The dierence of crack spacing
is maximum between the sound beam and the
whole aected. We can note that the behaviour law
of concrete used is able to reproduce crack rota-

tion due to shear eects. Fig. 22 presents the

progressive crack apparition along the beam for
the reactive beam with constant humidity. We can
see a jump into the steel strength when cracking
occurs. With the progressive loading from F 0
to F Fmax , cracks appear progressively from the
middle to the edge of the beam. The presence of

828

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

Fig. 21. Principal damage elds for the simulations (sound concrete, AAR with and without humidity gradient).

the crack leads to an increase of steel strength at

this place.
Now, if we consider the global responses of the
beams in terms of load/deection, the dierences
between the beams are less signicant than the
crack pattern showed it. Fig. 23 compares the
load/displacement curves for the three beams.
The global responses of the beams are very
close to each other. Before the rst crack apparition, the curves are practically identical because
the level of degradation reached by AAR is not
sucient to be seen. When cracks appear, the difference between the mechanical characteristics of
the beams, aected or not by AAR, is able to produce dierent behaviours. Nevertheless, it seems
that the level of degradation, deduced from experimental data and generated by AAR during 400
days, is a little too weak. Another explanation is
the level of reinforcement used to simulate the
beam. Indeed, this level leads to a beam that is
under reinforced, then the global behaviour of the
beam is controlled by steel. Once cracks appeared,
steel is the governing parameter of the behaviour
and the little decrease of mechanical characteristics
of concrete is less signicant. The next step will be
the test of a more reactive concrete, in order to

damage it more signicantly, and the simulation of

both under reinforced and over re-inforced concrete beams.

6. Conclusion
The three-dimensional modelling of AAR presented here makes it possible to perform mechanical/chemistry coupled simulations. It requires, on
the one hand, a powerful behaviour law of concrete
capable to represent the orthotropic behaviour of
the material in terms of strains and damage. In
addition, this modelling of AAR is based on a
global description of the chemical kinetics of the
reactions and on a physical description of the
swelling mechanisms. The method employed to
take into account the coupling between chemical
and mechanical phenomena leads to a model which
integrates the various possible sources of concrete
damage. The numerical simulations performed
show a good agreement between experiments and
the model responses in term of three-dimensional
swelling under dierent loading levels. These results highlight the need for having a realistic description of the various physical mechanisms

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

Fig. 22. Evolution of tensile strength in steel bar during mechanical loading.

Fig. 23. Load/displacement curves for the three beams (sound, reactive with or without constant humidity).

829

830

B. Capra, A. Sellier / Mechanics of Materials 35 (2003) 817830

concerned in order to obtain a coherent response of

the model. The numerical implementation of this
model in a nite element code has been done. The
rst simulations of a beam subjected to AAR show
some interesting results which reproduce some
phenomenon observed on real structures. The next
part of this work will be to simulate another type of
beam, especially with regard to re-inforcement and
AAR rate, in order the determine the loss of stiness and bending capacity. Besides, in order to see
if our modelling is in a good agreement with experiments, we need to take into account beams with
shear re-inforcement because some experiments
were carried out with it.
Acknowledgements
EDF-CIH (Electrict
e de France) is gratefully
acknowledged for its nancial support towards
these studies.
References
Barzegar, F., Maddipudi, S., 1997. Three-dimensional modelling of concrete structure I: plain concrete. Journal of
Structural Engineering 23 (10), 13391346.

Capra, B., 1997. Mod
elisation des eets mecaniques induits par

les Reactions Alcalis-Granulats, PhD thesis, ENS Cachan.
Capra, B., Sellier, A., Bourdarot, E., 2000. Mechanical modelling of alkali-aggregate reaction in concrete. 11th International Conference on Alkali-Aggregate Reaction, Qu
ebec,
1116 June B
erub
e, Fournier, Durand editors.
Institution of Structural Engineers, 1992. In: Structural eects
of alkali-silica reaction, technical guidance on the appraisal
of existing structures. Published by the Institution of
Structural Engineers, London.
Jansen, D.C., Shah, S.P., 1997. Eect of length on compressive
strain softening of concrete. Journal of Engineering Mechanics 25 (1), 2535.
Larive, C., 1997. Apport combin
es de lexp
erimentation et de la
mod
elisation a la compr
ehension de 1alcali-r
eaction et ses
eets m
ecaniques, PhD thesis, LCPC Paris.
Lemaitre, J., Chaboche, J.L., 1988. M
ecanique des mat
eriaux
solides. Dunod, Paris.
Sellier, A., Capra, B., M
ebarki, A., 1999. Concrete behavior: a
probabilistic damage model. In: 8th International Conference on Applications of Statistics and Probability, Sydney,
1215 December 1999. Balkema, Rotterdam, pp. 331
335.
Sellier, A., Bary, B., Capra, B., 2001. Rotating smeared
crack and orthotropic damage modelling for concrete. In:
De Borst et al. (Eds.), Proceedings of the Fourth International Conference on Fracture Mechanics of Concrete
Structures, Balkema, 28 May1 June. Cachan, France, pp.
629636.
Swamy, R.N., Al-Asali, M.M., 1989. Eect of alkali-silica
reaction on the structural behavior of reinforced concrete
beams. ACI Structural Journal 86 (4), 451459.