432

Numerical assessment and prediction method

for the chemico-mechanical deterioration of
ASR-affected concrete structures
Kefei Li and Olivier Coussy

Abstract: The degradation of concrete structures by the alkali–silica reaction (ASR) is highlighted in recent years in
civil engineering. To quantify the deterioration of concrete structures, one has the behavior modeling of affected concrete
proposed by multiple approaches, on one side, and the expansion of the structure measured by the in situ engineers, on
the other side. The task is to determine the progressive structure degradation from these two supports by a systematical
method. Nevertheless, this method should not depend on the employed model so that a better one can always be added
to the assessment. Moreover, one should also be able to analyze the credibility of the assessment result. All these points
differentiate between a laboratory expertise and a structural assessment. This article proposes a time-scaled evaluation
method and formulates it into an inverse problem. With an adopted chemico-mechanical model for ASR expansion and
the analyzed structure observation data, the authors give the numerical solution for the established inverse problem. An
ASR-affected suspension bridge is then evaluated using a finite element code for its durability by the achieved method.
Some further discussions are given in the last section of this article.
Key words: alkali–silica reaction, modeling, inverse problem, assessment, suspension bridge, durability.

Résumé : La dégradation des structures en béton causée par la réaction d’alcali-silce est recemment prononcée en
génie civil. Afin de quantifier la détérioration, on a, d’un côté, des modélisations proposées par multiples approches, de
l’autre, des mesures sur l’expansion de la structure par des ingénieurs in situ. La tâche est de déterminer la dégradation
progressive de la structure, en s’appuyant sur ces deux éléments, par une méthode systématique. Néanmoins, cette
méthode ne devrait pas dépendre du model adopté afin de laisser toujours la possibilité d’adopter un meilleur model. De
plus, on devrait aussi pouvoir quantifier la crédibilité du résultat de l’évaluation. Cet article vise à une telle méthode et la
formule en terme d’un problème inverse. En prenant une modélisation chimico-mécanique et des données in situ analysées,
les auteurs arrivent à résoudre le problème inverse établi par un schéma numérique. Un pont suspendu est ensuite évalué
via un code d’éléments finis pour sa durabilité par la méthode achevée. Des discussions sur l’application de la méthode
sont engagées dans la dernière partie.
Mots clés : réaction d’alcali–silice, modélisation, problème inverse, évaluation, pont suspendu, durabilité.

1. Introduction extensive cracking and to compromise the material durability

and hence the safety of the structure. Since its first reported
The alkali–silica reaction (ASR) happens between the reac- case (Stanton 1940), the ASR pathology has been detected in
tive aggregate containing the poor crystallized silica network numerous engineering structures so that it becomes one univer-
and the high alkaline solution in concrete internal pores (Dent sal concern for concrete structure durability (Swamy 1992).
Glaser and Kataoka 1981; Dron and Brivot 1992). The amor- Indebted to the material and chemical research in the past
phous product from ASR, usually called “gel”, could swell sub- decades, this structure pathology can be greatly reduced by
stantially with the combination of internal water to fill up the the proposed regulations on the concrete fabrication process
connected pores in concrete. Therein an autogeneous pressure (ASTM 1994; RILEM 2000). Therefore, the maintenance of
is exerted on the solid skeleton, which is high enough to create the existed ASR-affected structures becomes the preponderant
concern. To treat these structures, engineers and owners are al-
Received 28 May 2003. Revision accepted 18 January 2004. Pub- ways confronted by such questions as to what extent the ASR
lished on the NRC Research Press Web site at http://cjce.nrc.ca/ on has affected the structure presently, how the degradation will
14 May 2004. evolve during the rest of its service life, and how to justify a
K. Li1,2 and O. Coussy. Laboratoire des Matériaux et des Struc- retrofit scheme. Evidently the answer to these questions is ex-
tures du Génie Civil, Cité Descartes, 2 allée Kepler, Champs-sur- pected to be supported by a quantitative result rather than an
Marne, 77420, France. empirical estimation. As far as the authors’ know, there is little
research dedicated to a systematic, operational, and computer-
Written discussion of this article is welcomed and will be received
based evaluation method for these affected structures. Based on
by the Editor until 31 October 2004.
a comprehensive state-of-the-art study, Léger et al. (1995) pro-
1
Corresponding author (e-mail: kefei.li@oxand.com). pose a numerical simulation method particularly for the ASR-
2 affected concrete dams. Without giving special attention to the
Present Address: OXAND S.A., 36 bis, avenue Franklin
Roosevelt, 77210 Avon, France. detailedASR-expansion modeling, the authors focus on the dam

Can. J. Civ. Eng. 31: 432–439 (2004) doi: 10.1139/L04-003 © 2004 NRC Canada
Li and Coussy 433

behavior assessment and its applicability using the finite ele- on the affected structure, the determination of κ is actually an
ment method. inverse problem given as
This article aims to propose a time-scaled evaluation method



adapted to all ASR-affected structures. To this end, several steps [2] Find κ : Mu(m; κ) = ū

of study are involved. Firstly, as an elementary part, a mechani- ¯
x∈; t∈T̄
cal model to describe the ASR expansion in concrete is needed
where M represents a mapping operator projecting the behav-
and two radical aspects are involved: the evolution law of gel ¯ u is the structure
ior of the entire structure u(x, t) onto T̄ , ;
expansion and the material response to the expansion. However,
behavior corresponding to the in situ measurement, e.g. the
the modeling itself will not necessarily lead to an assessment of
structure deformation or stress field; and ū is the behavior mea-
affected structures, since the model is not calibrated to the in situ
surement during a monitoring period T̄ and at certain positions
concrete yet. Then comes the calibration of the ASR-expansion ¯
represented by .
model to the in situ concrete by the in situ observations. In
However, the inverse problem in eq. [2] usually turns out to be
other words, the parameters in the ASR expansion modeling
mathematically ill-posed (Mahnken and Stein 1996). Therefore,
are to be determined by the in situ collected data, which can
for its solution, an optimization approach is often retained as
be formulated into a typical engineering inverse problem. The
challenge of this step resides in the solution of inverse problem 1





2
solution as well as in the fact that different structures usually [3] f (κ) = Mu(m; κ) − ū

;
¯ T̄ → min
2
furnish different kinds of in situ measurements. Therefore, for
the sake of a generalized assessment method, one has to nor- where f (κ) is the objective error function. Therefore, the as-
malize the in situ ASR relevant information regardless of the sessment of an affected structure consists of first determining
structure type. Finally, with the model calibrated for the in situ the ASR characteristics through an inverse problem in eq. [3]
concrete, a comprehensive structure analysis can be performed and then a structural computation for a behavior qs by eq. [1].
by the computer method, such as, the finite element program. The solution of eq. [3] needs to detail the ASR characteristics
This way we can reconstitute numerically the structure evolu- κ as well as the structure measurement ū. The precision of κ is
tive degradation for its full service life. It is from this numerical given in the following section through a chemico–mechanical
reconstitution that the structure degradation history could be ASR expansion model, while ū is to be detailed via the analysis
entirely retrieved, its present state could be determined, and its of in situ ASR information.
future behaviors could be predicted.
Based on the above-mentioned methodology, this article is 2.1. Chemico-mechanics of alkali–silica reaction
structured as follows. In Sect. 2, the assessment method is for- expansion
mulated into an inverse problem and a comprehensive structure- Multiple mechanical models have been proposed for the
behavior computation. To complete the assessment method, we ASR expansion in recent years (Moranville 1997). A chemico–
recall the ASR expansion modeling and analyze the in situ ASR mechanical approach characterizes ASR process in concrete by
information. In Sect. 3, a numerical algorithm is proposed for using a normalized reaction extent ξ ∈ [0, 1], where ξ = 0
the solution of the inverse problem with respect to the adopted represents no formation of ASR gel, while ξ = 1 stands for
ASR expansion model and theASR information.A real structure the exhaustion of reactants assessment method such a chemico–
pathology case is treated in the following section by the pro- mechanical model is retained, giving both the constitutive rela-
posed assessment method. Finally, a discussion on the available tion of the affected concrete and the evolution law for ASR gel
assessment method is given in the last section. formation (Li and Coussy 2002)
  
p p

σij = (K − 3 G) kk − kk δij + 2G ij − ij
2
2. Structure assessment method
[4]
 − 3βKξ δij
The purpose of the assessment of an ASR-affected structure 
is to determine one or more of its mechanical behaviors, qs , tc ξ̇ = 1 − ξ
from which the durability and serviceability of this structure p
can be evaluated. The behavior qs , a time and space scaled where σij , ij , and ij stand for the components of the mate-
function, is moreover a function of the ASR characteristics, i.e. rial stress, total strain, and plastic strain tensors, respectively;
ASR expansion modeling parameters K and G are the overall bulk and shear moduli, respectively, β
represents ASR expansion coefficient, and tc is the ASR char-
[1] qs = qs (m, κ) acteristic time. The ASR gel formation characteristic time tc is
identified by a recent campaign of material tests (Larive 1998)
where κ stands for the parameters according to a specific ASR as
expansion modeling and m is the vector of other material pa-
rameters relevant to the solution of qs . If m is assumed as known 1 + exp[−τL /τc ]
[5] tc = τc
for an ASR-affected structure in the assessment, the parameters ξ + exp[−τL /τc ]
represented by κ cannot be assumed likewise because the ASR
characteristics depend strongly on the concrete formulation and where τc and τL are two time parameters. So far, the ASR char-
the available laboratory ASR characteristics cannot represent in acteristics or parameters, κ, are identified as
situ concrete. Therefore, κ can be determined only through the
in situ ASR measurement. If ū is the available measurement [6] κ = (β, τc , τL )

© 2004 NRC Canada

434 Can. J. Civ. Eng. Vol. 31, 2004

The coefficient study (Li et al. 2000) reveals the correlation be- measurements. Accordingly, the problem in eq. [7] can be writ-
tween κ and the ambient temperature θ and humidity h, i.e., ten at structure level as
κ = κ(θ, h). Taking (θ, h) as independent variables trans-
forms eq. [3] into  
fs κ0 = β0 , τc0 , τL0
1





2 1



 
[7] f (κ) = Mu(m; κ0 ; θ, h) − ū

;
¯ T̄ → min [9] = MDs m; κ0 ; θs (x, t), hs (x, t)
2 2


2
where κ0 = (β0 , τc0 , τL0 ) is the ASR parameters at a reference − ds

[t s ,t s ] → min
0 1
thermo-hydrometric condition θ = θ0 , h = h0 .
where fs is the structure-level objective function and Ds is the
2.2. In situ alkali–silica reaction information computed structure-measurement vector. At material level, the
To detail ū, we resort to the available in situ ASR informa- life of extracted cores is composed of two distinct parts in time
tion. Owing to the accumulated engineering experiences, three scale. The first part, 0 ≤ t < t0c , is from the structural mate-
data sources that are supposed to be independent of the structure rial formation (t = 0) till their extraction from the structure.
type are often exploited to quantify the in situ ASR evolution. During this period they are integrated in the structure. And the
To illustrate clearly the ASR information, we refer to a repre- second one, t0c ≤ t ≤ t1c , is from their extraction to the end
sentative structure in Fig. 1. of residual expansion test. During this period they are subjects
of laboratory test. Accordingly their ASR information, includ-
(a) Structure measurement: These measurements are to cap- ing the thermo-hydrometric histories θc (t), hc (t) and expan-
ture the structural response to the internal ASR evolution sion history c (t), is composed also of these two parts. Given n
during an investigation period, noted as [t0s , t1s ]. It can be cores, θc (t), hc (t), and c (t) are denoted by n-dimension vec-
the structure deformation or the concrete surface cracking tors [(·)1c (t) · · · (·)nc (t)]1×n . From the data source (c), the core
(LPC 1997a). Figure 1 illustrates this ASR information by thermo-hydrometric conditions read as
measuring the distance variation between two points C1 and

C2 .  θs (xc , t), hs (xc , t), 0 ≤ t < t0c
(b) Concrete core expansion: At the detection of ASR, con- [10] θc (t), hc (t) =

crete cores are usually extracted from the affected structure θres 1, hres 1, t0c ≤ t ≤ t1c
(illustrated by point A in Fig. 1) at a certain instant, noted as
t0c . Then these cores are subjected to a normalized thermo- where (θ, h)res is the thermo-hydrometric condition in residual
hydrometric condition during a recommended test period expansion test, xc is the core extraction positions in , and
(LPC 1997b), noted as [t0c , t1c ], to evaluate their residual ex- 1 = [1 · · · 1]1×n . However, from the data source Sect. 2.2. (b)
pansions. By the test, we try to capture locally the concrete their expansion histories are only partially known to us
expansion potential, i.e., to characterize ASR expansion ex-

tent at a material level.
 unknown, 0 ≤ t < t0c
(c) Thermo-hydrometric condition: As indicated in the lab- [11] c (t) =
 c
oratory observation and chemico-mechanical modeling, the c (t0 ) + res (t), t0c ≤ t ≤ t1c
thermo-hydrometric condition (θ, h) should be acquired to
characterize the in situ ASR expansion. For a structure, the where res (t) are the core residual expansions. It should be
condition is time-space scaled, i.e., θs (x, t), hs (x, t). To re- noted that res (t) provide the relative ASR expansions com-
constitute the complete thermal and hydrometric histories, pared to c (t0c ). These initial expansions indicate the chemical
we need to collect the meteorological condition, which en- extents already acquired at their extraction, so they are chemico-
close the structure, as well as the local conditions (e.g., stag- mechanical indicators for these cores. Consequenctly, the prob-
nant water and local thermal source) to which the structure lem in eq. [7] can be written for material level as
is subjected.
 
These data sources can be analyzed at two levels: structure [12] fc κ0 = β0 , τc0 , τL0
and material. At structure level, the ASR information concerns
the structure measurements Sect. 2.2. (a), noted as ds (t), t ∈ 1



 

2
= MEc m; κ; θc (t), hc (t) − c

[t c ,t c ] → min
[t0s , t1s ], and the structure thermo-hydrometric condition recon- 2 0 1

stituted by the in situ investigation Sect. 2.2. (c) as

where fc is the objective function at material level and Ec rep-
ds (t), t0s ≤ t ≤ t1s resents the computed core expansions.
[8]
θs (x, t), hs (x, t), 0 ≤ t, x ∈  Therefore, the original problem in eq. [7] is naturally divided
into two subproblems in eq. [9] and eq. [12]. If, furthermore, we
where  is the structure space domain. Using a matrix alge- consider only one structure measurement ds (t) from data source
bra notation, ds (t) can be represented as ds (t) = [ds1 (t), · · · , Sect. 2.2. (a) and the residual expansion for one representative
dsm (t)]1×m where m being the number of effectuated structure concrete core from data source Sect. 2.2. (b), i.e., m = n = 1,

© 2004 NRC Canada

Li and Coussy 435

Fig. 1. Representative ASR-affected structure and in situ ASR information. (See Fig. 2 from LPC (1997b)).

the two subproblems can be simplified as By supposing that Gs ⊆ Gc , the calibration of κ0 can be reduced
to a unique problem
t1s
1
[13a] fs (κ0 ) = [MDs (m; κ0 ; θs (x, t), hs (x, t)) [15] Find κ0 ∈ Gc : fs (κ0 ) → min
2
t0s
For eq. [15] a numerical algorithm can be studied. The chal-
− ds (t)]2 → min lenge here is that both the core residual expansion res (t0c , t1c )
t1c and the structure measurement ds (t0s , t1s ) are the relative mea-
1 surements compared to their unknown initial values. Taking this
[13b] fc (κ0 ) = [MEc (m; κ0 ; θc (t), hc (t))
2 into account, we propose a numerical algorithm at two steps.
t0c

− c (t)]2 → min Material step: Form Gc

Predetermine a value for core initial expansion ε0 = c (t0c )
In other words, the purpose of calibration of κ0 is to find and determine the partial expansion history of core c (t0c , t1c ) =
such a group of ASR parameters that, through the calculation of ε0 + res (t0c , t1c ). With its thermo-hydrometric condition
 θc (t),
concrete core expansion in eq. [13b], could retrieve the residual hc (t), find an optimal group of parameters κ0 = β0 , τc0 , τL0
expansion c (t) and regenerate the distance variation ds (t) by so that eq. [13b] is fulfilled. This procedure is to repeat for a
a structure chemico-mechanical computation in eq. [13a]. series of ε0k , (k = 1, n) so that a solution set Gc is formed
 k 
3. Solution of inverse problem [16] Gc = β0 , τc0 , τL0 , k = 1, n

Even simplified as eq. [13], the calibration of κ0 necessitates Structure step: Find κ
a convergence criterion between the two subproblems. Note, For each group of parameters κ0k in Gc , compute the rela-
Gs and Gc the solution sets of structure and material level sub- tive variation of distance Ds (t0s , t1s ) with the structure thermo-
problem, respectively, are expressed as hydrometric condition θs (x, t), hs (x, t). The optimal group
 
 should observe eq. [13a]. Note this group as
Gs = κ0 → R × R × R
fs (·; κ0 ; ·) → min
[14] 
 jv
κ0 = (β0 , τc0 , τL0 )j v
Gc = κ0 → R × R × R
fc (·; κ0 ; ·) → min

© 2004 NRC Canada

436 Can. J. Civ. Eng. Vol. 31, 2004

it is thus the solution of the inverse problem. The associated Fig. 2. Core residual expansion and column section variation.
jv
mechanical indicator ε0 is the acquired ASR expansion for the 30 24

core at its extraction. The detailed numerical algorithm is given Section area variation
in Appendix A. 25 Core residual expansion 20
jv

Core residual expansion (×10e-4)

With the ASR parameters κ0 = κ0 calibrated for the in Calculated core expansion

Section area variation (×10e-4)

situ concrete, a specific structure behavior qs (x, t), e.g., the 20 16
stress, strain, or deformation, can be computed considering the
structure thermo-hydrometric condition 15 12
 
[17] κ0 → qs m; κ0 ; θs (x, t), hs (x, t)
10 8
If t¯ is the present instant, a full-life evolution is eventually avail-
able for this specific behavior: qs (0 ≤ t < t¯) is a behavior di-
5 4
agnostic, qs (t = t¯) gives a description of actual structural state,
and qs (t > t¯) furnishes a quantified prediction. Accordingly,
the calibration solution to eq. [13] and the structural assessment 0 0
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
in eq. [17] constitute a numerical, diagnosis-prognosis consis- Year
tent assessment method for ASR-affected structures. By this
method, we are ready to evaluate the ASR-affected structures
and to furnish the quantitative answer for their durability and
security. Fig. 3. Calibration on structure level by the pylon section
measurement.
50
4. Case study: suspension bridge pylon
Structure measurement
The suspension bridge had been constructed during the 1920s, (0.35%,190 d,230 d)
40
destroyed during the second world war and reconstructed during (0.38%,190 d,225 d)
Section area variation (×10e-4)

(0.41%,190 d,220 d)
1949–1952. Since 1980 extensive cracking has been observed at
(0.42%,195 d,215 d)
its reinforced concrete pylons and developed upto the masonry 30 (0.45%,195 d,210 d)
foundation. The ASR product is located in the extracted sam-
ples, and the structural degradation cause was thus attributed
to ASR. Considering this bridge has attained the service life 20
of 50 years, a new bridge is under design with an expected
construction duration of 5 years. To assure the safety and ser-
viceability of this old bridge during this period, an assessment 10

of its mechanical state is expected.

0
4.1. Alkali–silica reaction information and model 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
Year
calibration
At structure level, the section area variation of one pylon col-
umn had been traced from 1990 to 1999, as shown in Fig. 2.
Since no local thermal source and water source are detected on With the ASR information at material level, as shown by eq.
the pylon, the thermal and hydrometric transfers are neglected. [19] and the residual expansion in Fig. 2, the ASR parameter
Accordingly, a constant and uniform thermo-hydrometric con- set Gc is formed by the material step of the proposed algorithm.
dition is attributed to the structural concrete with an annual To compute the pylon section expansion at structure calibration
average temperature of 12 ◦ C and an internal humidity of 95%, step, the column section is modeled as a plan deformation case
i.e. with internal ASR expansion. With help of a finite element code,
[18] θs (x, t) = 12◦ C hs (x, t) = 95% CESAR-LCPC at Laboratoire Central des Ponts et Chaussées
(j )
at Paris, for each κ0 ∈ Gc , the section variation is computed
Therefore, eq. [18] and the section area measurement constitute and compared to its value at 1990. The structure level itera-
the ASR information at structure level. At material level, some tion is illustrated in Fig. 3, and the optimal parameter group is
concrete cores were extracted from one degraded pylon column identified as
and subjected to the residual expansion test from February 1994
to February 1997. The residual expansion of a typical core is [20] κ0 = β0 = 0.41%, τc0 = 190 d, τL0 = 220 d
shown in Fig. 2. So by eq. [18] and the normalized temperature
and humidity conditions in the residual expansion test (LPC with which the core residual expansion is well regenerated, as
1997), the core thermo-hydrometric condition can be reconsti- shown in Fig. 2. The associated mechanical indicator is evalu-
tuted as ated as 0 = 0.31%, implying that in 1994 the ASR expansion
had attained about 75% of its total extent. Moreover, this in-
12◦ C, 95% for 1952–1994 dicator gives the chemical extent for all the pylons due to the
[19] θc (t), hc (t) =
38◦ C, 100% for 1994–1997 constant and uniform thermo-hydrometric condition.

© 2004 NRC Canada

Li and Coussy 437

Fig. 4. Finite element meshes for pylon concrete and Table 1. Chemical and mechanical parameters for the
reinforcement cages by CESAR-LCPC. suspension bridge pylon assessment.
Column stirrup
Parameter Value
Beam stirrup Principal cable service load q (MPa) 8.27
Young’s modulus of concrete Ec (GPa) 9.8
Poisson’s ratio of concrete ν 0.2
Mass density of concrete ρc (Kg/m3 ) 2,400
Compressive strength of concrete Rc (MPa) 19.0
Tensile strength of concrete Rt (MPa) 1.74
Tensile strength of steel Rs (MPa) 240.0
Linking bars Young’s modulus of steel Es (GPa) 200.0
Time coefficient τc0 (d) 190
4997 Elements Time coefficient τL0 (d) 220
3069 Nodes ASR expansion coefficient β0 (%) 0.41
Ambient temperature θatp (◦ C) 12
Humidity of concrete hs (%) 95

4.2. Pylon behavior evaluation

A part of pylon, the upper transbeam, and a part of vertical
column, is retained for studying the impact of progressive ASR situ observation. Their high stress level since 1995 is less re-
expansion on the pylon mechanical behavior. The ASR expan- alistic considering the assumed ideal elasticity. Nevertheless,
sion as well as the structural loads on the pylon are considered the results show that these stirrups are losing their efficiency
in the evaluation. A chemico-elastoplastic behavior is attributed to confine the column section, so a retrofit is necessary for
to the pylon concrete these steel bars.
2 p p Beam stirrups: They are currently working within the elas-
[21] σij = (K − G)(kk −kk )δij +2G(ij −ij )−3βKξ δij tic limit. However, a notable cracking could have also been
3
observed considering the stress level of 160 MPa in these
p
where ij is the plastic strain calculated by an associated elasto- stirrups and the associated material strain. In several years,
plastic law (Willam and Warnke 1975). An ideal elasticity is their stress level (190 MPa in 2005) can approach the tensile
adopted for the reinforcement bars in the pylon body, so their strength. Consequenctly, reinforcement of these stirrups is
stress reads as also expected.

[22] σ = Es  Linking bars: They are designed to strengthen the link be-
tween the reinforcement cages in column and transbeam.
where Es is the Young’s modulus of the steel bars. Moreover, In 1952, a pressure (–50 MPa) was exerted in these bars
a perfect cohesion is supposed between the steel bars and the by the service load from the principal cable. With the pro-
surrounding concrete material. The finite element meshes for gressive ASR expansion of concrete in pylon, their stress
the concrete and the steel bars are shown in Fig. 4 (for the sake state changed successively from pressure to tension. A ten-
of clarity, only one part of these steel bars is shown). Using sion of 50 MPa is observed presently, and there is no need of
the finite element code, CESAR-LCPC, we compute the pylon strengthening the linking bars.
behavior from 1952 to 2005 and the relevant parameters are
given in Table 1.
Amongst the structural responses, we concentrate on the 5. Further discussion
stress evolution of the reinforcement bars present in the column By an inverse problem for the calibration of ASR expansion
and transbeam: the column stirrups, the transbeam stirrups, and model and a comprehensive structural chemico-mechanical com-
the linking bars. From the numerical results presented in Fig. 5 putation, the article achieves a numerical assessment method
the following tendencies can be observed. for the ASR-affected concrete structures. The obtained result
of the bridge pylon assessment, in spite of its simplicity, does
Column stirrups: Horizontally placed in the column, column give the essential quantitative indications for the structure dura-
stirrups were subjected to a tension of 25 MPa from the prin- bility and serviceability. So far, the assessment method remains
cipal cable service load in 1952. Since then with the ASR ex- a static one, which means the calibration of κ0 and the result-
pansion, the concrete confined by these stirrups swelled pro- ing qs (κ0 ) are determined by a structure measurement during a
gressively and exerted on them a superfluous tensile stress. fixed period of [t0s , t1s ].
Accordingly, their tensile stress accumulated such that in
1995 they attained the tensile strength of 240 MPa. It in- [23] ds (t0s , t1s ) → κ0 , qs (κ0 )
dicates that these stirrups have been working in the plastic
phase for about 6 years and an extensive cracking could have This method can be extended easily into a real time one, which
been observed during this period, which coincides with the in updates the κ0 calibration and qs (κ0 ) computation with the con-

© 2004 NRC Canada

438 Can. J. Civ. Eng. Vol. 31, 2004

Fig. 5. Stress evolution of reinforcement bars in the pylon.

400

Column Stirrups
350
Transbeam Stirrups
Linking Bars
300

250
Steel bar stress (MPa)

200

150

100

50

-50

-100
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010
Year

tinuation of in situ structure measurement, (Paris, France) for their contribution to the first version of this


paper, Catherine Larive at the Centre d’Etude des Tunnels (Lyon,
[24] ds (t0s , tis )
i>1 → κ0 , qs (κ0 )
i>1 France) for her suggestions for this method. The authors also
express their thanks to Alan Le Mehaute at Laboratoire Ré-
Of course, the more structure measurements we have, the more gional des Ponts et Chaussées (St-Brieuc, France) and Chris-
is the credibility of the calibration and the behavior assessment. tian Tessier at Centres d’Etudes Techniques de l’Equipement
Moreover, in this article the structure assessment is presented to- (Nantes, France) for the discussion on the suspension bridge.
tally in a deterministic way.Accordingly, the sensitivity analysis
of the inverse problem solution is an important supplementary
study to the proposed assessment method. For our problem this
analysis concerns the sensitivity of the final predicted structure References
behavior qs with respect to a certain entry data to the inverse ASTM. 1994. Standard test for potential alkali reactivity of aggregate
problem in eq. [13a]. Mathematically, the sensitivity of qs with (motar-bar, method (C 1260-94). Philadelphia, Pa.
respect to a certain ASR information d can be expressed as Coussy, O. 1995. Mechanics of Porous Continua. Wiley, Chichester,
U.K.
q ∂qs
[25] ,d = Dent Glaser, L.S., and Kataoka, N. 1981. The chemistry of ’alkali-
∂d aggregate’ reaction. Cement and Concrete Research, 11(1): 1–9.
q Dron, R., and Brivot, F. 1992. Thermodynamic and kinetic approach
However, in practice the analytical solution of ,d is rarely avail-
to the alkali–silica reaction. Part 1: Concepts. Cement and Concrete
able as both the behavior qs and the ASR information d, such
Research, 22(5): 941–948.
as the structure measurement ds or the concrete core residual
Larive, C. 1998. Apports combinés de l’expérimentation et la mod-
expansion res , are time-space scaled functions. So the solution
élisation à la compréhension de l’alcali-réaction et de ses effets
of eq. [25] is usually a numerical one. If the sensitivity with
mécaniques (Experimental and modeling contributions to the com-
respect to a certain ASR information, e.g., core residual ex-
prehension of alkali-reaction and its mechanical effects). Mono-
pansion, is rather high, that means the predicted behavior qs is
graph LPC OA28, Laboratoire Central des Ponts et Chaussées,
sensitive to this information. Consequently, during the assess-
Paris, France [partially translated into English].
ment procedure, one should be vigilant to this ASR information
Léger, P., Tinawi, R., and Mounzer, N. 1995. Numerical simulation of
source and render it as credible as possible.
concrete expansion in concrete dams affected by alkali–aggregate
reaction: state-of-the-art. Canadian Journal of Civil Engineering,
Acknowledgements
22(4): 692-713.
The authors would render their sincere thanks to Franz-Josef Li, K., and Coussy, O. 2002. Concrete ASR degradation: from material
Ulm at Massachussets Institute of Technology (Cambridge, modelling to structure assessment. Concrete Science and Engineer-
US), Bruno Godart at Laboratoire Central des Ponts et Chaussées ing, 4: 35–46.

© 2004 NRC Canada

Li and Coussy 439

Li, K., Ulm, F.J., Coussy, O., Larive, C., and Fan, L. 2000. Chemoe- with
lastic modelling of alkali–silica reaction in concrete, Proceedings  l (l−1)
of 11th International Conference on Alkali-Aggregate Reaction in 
 β = β0 + ,β0
 0

Concrete, Québec, Que., June 2000. Edited by M.A. Bérubé, B. 

Fournier, and B. Durand. Centre de Recherche Interuniversitaire 0(m) 0(m−1)
[A.3] τc = τc + ,τc0
sur Béton, Sainte-Foy, Que. pp. 989–998. 



LPC. 1997a. Détermination de l’indice de fissuration d’un parement 
 0(n) 0(n−1)
de béton (Determination of cracking index on a concrete sur- τL = τ L + ,τL0
face). Méthode d’Essai LPC 44, Laboratoire Central des Ponts et
0(0) 0(min)
Chaussées, Paris, France. [In French]. where l, m, n ≥ 1 and β00 = β0min , τc,L = τc,L as
LPC. 1997b. Alcali-réaction du béton: Essai d’expansion résiduelle l, m, n = 1.
sur béton durci (Alkali reaction: Residual expansion test on the
hardened concrete). Méthode d’Essai LPC 47, Laboratoire Central (d) Evaluate the function fck in eq. [13b], with the ASR pa-
 0(m) 0(n) 
des Ponts et Chaussées, Paris, France. [In French]. rameters κ0 = β0l , τc , τL
Mahnken. M, and Stein, E. 1996. A unified approach for parameter

identification of inelastic models in the frame of the finite element  N   i  2
 | res + ε0k − Eci |
fc = 
method. Computer Methods Applied in Mechanical Engineering. k
136: 225–258. [A.4] (i)
i=1 |res |
Moranville, M. 1997. Modelling of Expansion Induced by ASR —
New Approaches. Cement and Concrete Composites, 19: 415–425.
where N is the number of recordings in the residual expan-
RILEM Technical Committees. 2000. Recommendations of RILEM
sion test.
TC 106-AAR: Alkali–aggregate reaction. Materials and Structures,
33(229): 283–293. (e) Iterate (c) and (d) to find a group of parameters
Stanton, T.E. 1940. Expansion of concrete through reaction between (k)
κ0 = β0l , τc , τL by minimizing fck until
0(m) 0(n)
cement and aggregate. Proceedings of ASCE, 66(10): 1781–1811.
Swamy, R.N. 1992. The alkali–aggregate reaction in concrete. Van
Nostrand Reinhold, New York, N.Y. [A.5] fck ≤ ηc
Ulm, F.J., Coussy, O., Li, K., and Larive, C. 2000. Thermo-chemo-
mechanics of ASR expansion in concrete structures. Journal of En- where the tolerance ηc is assumed to be 10−2 . Note that
 lk 0(mk ) 0(nk ) 
gineering Mechanics, ASCE, 126(3): 233–242. β0 , τc , τL are the found ASR parameters.
Willam, K.J., and Warnke, E.P. 1975. Constitutive models for the triax-
ial behaviour of concrete. Proceedings of International Association (f) If k → k + 1, return to (b) until ε0k = ε0max . A set of κ0k is
of Bridge and Structure Engineering, 19: 1–30. thus formed



(k)
[A.6] Gc = κ0k
fc (κ0 ) ≤ η , ε0min ≤ ε0k ≤ ε0max
Appendix A. Algorithm
A1. Material step  
(a) Predetermine the search intervals (·)min , (·)max and the A2. Structure step
 
increments ,(·) for ASR parameters κ0 = β0 , τc0 , τL0 j
(g) Choose κ0 from Gc in [A.6]
and the initial expansion ε0 , respectively
 j l 0(mj ) 0(nj )
 β0min ≤ β0 ≤ β0max with ,β0 [A.7] κ0 = β0j , τc , τL j ≥1





0,min
τc,L 0 ≤ τ 0,max with ,τ 0
≤ τc,L j
[A.1]
 c,L c,L (h) Compute the function fs in eq. [13] taking into account

 the structure thermo-hydrometric condition θs , hs


 min
ε0 ≤ ε0 ≤ ε0max with ,ε0 
K  2
 |(Dsi − Ds0 ) − dsi |
[A.8] fs = 
j
(b) Compute the value of ε0 for the k th iteration (i)
|ds |
i=1
(k−1)
[A.2] ε0k = ε0 + ,ε0 , (k ≥ 1) where K is the total recording number at the in situ mea-
surement.
where
(0)
(i) If j → j + 1, iterate (g) and (h) to find an optimal group
ε0 = ε0min when k = 1 (j v)
κ0 by the criterion

(c) Choose one combination of the parameters [A.9] fsj v ≤ ηs

β0l , τc0(m) , τL where the tolerance ηs is assumed to be 10−2 .

0(n)

© 2004 NRC Canada