Professional Documents
Culture Documents
Construction and Building Materials: I. Yurtdas, D. Chen, D.W. Hu, J.F. Shao
Construction and Building Materials: I. Yurtdas, D. Chen, D.W. Hu, J.F. Shao
Construction and Building Materials: I. Yurtdas, D. Chen, D.W. Hu, J.F. Shao
h i g h l i g h t s
Mechanical and hydraulic properties of ASR affected mortar are studied. ASR process at 60 °C and 95% RH lasts 100 days.
After ASR, the increase in compressive strength of reactive mortar is more significant than that of non-reactive one.
The densification of ITZ could mostly explain the higher increase of reactive mortar strength.
Bending strength and elastic properties are more affected by ASR induced microcracking.
Permeability and porosity measurements can provide useful data for the evaluation of ASR induced microcracking.
a r t i c l e i n f o a b s t r a c t
Article history: This experimental investigation deals with effects of ASR on the mechanical properties of cement-based
Received 29 June 2012 materials. Two selected mortars, only differentiated by the chemical nature of the aggregates used (reac-
Received in revised form 19 April 2013 tive and non-reactive) are investigated. After a maturation period of 28 days, triaxial compression tests
Accepted 28 April 2013
with different confining pressures (0, 5, 15 MPa), bending tests, porosity and permeability measurements
Available online 3 June 2013
are carried out on a part of the samples in order to obtain a reference state as the sound material while
the other part is stored in a controlled environment with temperature of 60 °C and relative humidity of
Keywords:
95% for ASR development. During this period of storage, the axial expansion of samples for the two mor-
Alkali silica reaction
Mortar
tars studied is monitored. After 100 days of storage in the controlled environment, the above cited tests
Microcracking and measurements are performed again in order to evaluate the effects of ASR. In addition, microstruc-
Triaxial compression tural analyses are also carried out in order to better understand various phenomena which occur in mor-
Mechanical properties tars before and after storage of samples in controlled environment. It is found that the increase of
Transport properties mechanical strength under compression is more significant in the reactive mortar than in the non-reac-
Microstructure tive one. This can be attributed mainly to its more compact microstructure in the ITZ, shown by the
microstructural analyses, compared with the non-reactive mortar. However, the ASR process induces a
decrease of bending strength, which is directly related to tensile strength, in reactive mortar. Moreover,
the elastic modulus and Poisson’s ratio of the reactive mortar decrease due to the ASR process. These
results indicate that the tensile strength and elastic properties are more significantly affected by the
ASR expansion related microcracking than the compressive strength. The microcracking due to the ASR
is also detected through the permeability and porosity measurements.
Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction microcracks. The gels fill the microcracks in the aggregates and
paste as well as the voids. The extent of the ASR induced
Alkali silica reaction (ASR) is one of the degradation process and deterioration is influenced in particular by the reactivity (moder-
refers to chemical reaction between the reactive silica phases in ately or highly) [1–3,5,8,9], quantity, size and particle size distribu-
the aggregates and the alkali hydroxides in the pore solution of tion of the reactive aggregates [1–3,6,8,10–14], the amount of
cement-based materials [1–8]. The reaction results in the forma- available moisture [1–4,13,15,16], amount of available alkalis
tion of alkali-silica gels and/or alkali–calcium–silica gels. These [1–3,8,10,12,15], porosity of material and pre-existence of micro-
gels which absorb water and swell can lead to the formation of cracks [1–4,15].
Many studies have been carried out in order to evaluate the
⇑ Corresponding author. Tel.: +33 326918052. mechanical performances of ASR affected concrete. For this
E-mail address: ismail.yurtdas@univ-reims.fr (I. Yurtdas). purpose, uniaxial compression and tension tests are generally
0950-0618/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.conbuildmat.2013.04.046
166 I. Yurtdas et al. / Construction and Building Materials 47 (2013) 165–174
observe such a diminution due to the ASR related expansion Components Quantity [kg/m3]
[16,22–24] over the period of experimental study. The uniaxial (a)
compressive strength of concrete decreases in particular when Sand (inert or reactive) 1613
the aggregates are highly reactive [5,9,25]. Further, the cores taken Cement CEM I 52.5 R 538
from ASR affected structures can have high residual uniaxial com- De-ionised water 269
Total alkali in mortar 15
pressive strength [20,21], especially at early age of ASR develop-
ment [20]. It seems that the uniaxial compressive strength of Aggregate diameter [mm] Aggregate fraction [%]
(b)
ASR affected concretes continues to increase or does not vary com- 0.08–0.16 25
pared with 28 day compressive strength in particular when the 0.63–1.25 50
aggregates are slowly or moderately reactive [5,9,25]. However, 2.5–4 25
whether the uniaxial compressive strength increases or not during
the development of the alkali silica reaction, a clear degradation of
elastic modulus is observed [2,3,5,9,16,17,22,23,25] and the mate- mortar) was fabricated with an inert crushed limestone sand while the reactive
rial becomes more deformable [9,17,22,24]. The elastic modulus is mortar (designated as R mortar) was fabricated with a reactive crushed siliceous
thus very sensitive to the ASR induced damage, and decreases even limestone sand [6,13]. The mortars were fabricated with specific sand grains com-
posed of only three different grain-size fractions: fine aggregates with an equivalent
if the aggregates are slowly or moderately reactive. This sensitivity
diameter between 0.08 and 0.16 mm; medium aggregates with an equivalent diam-
is also reported for the direct tensile strength, which decreases eter between 0.63 and 1.25 mm and coarse aggregates with an equivalent diameter
drastically with the ASR expansion [5,18,19]. The evolution of the between 2.5 and 4 mm. For each mortar, the sand grains were composed of 25% of
concrete flexural strength also seems to be affected by the ASR fine aggregates, 50% of medium aggregates and 25% of coarse aggregates (Table 1b).
[5,9,19,22,25]. It is also observed that a high strength concrete These fractions were obtained by sieving; the sands were then washed to be dis-
posed of very fine particles and dried before use. CEM I 52.5 R with a high alkali con-
has a better resistance than an ordinary strength concrete against
tent of 1.11% Na2Oeq and demineralised water were used to cast the mortars. The
the harmful effect of the ASR damage [5]. Therefore, these results alkali content was increased to 15 kg/m3 by adding NaOH to the mixed water.
show that the ASR does not influence in the same way all of the Prismatic (40 40 160 mm3) and cylindrical (£ 36 100 mm3) samples
mechanical properties. It is found that the good indicators of the were prepared for this study. The prismatic samples were used for the monitoring
of length variation in order to determine the expansion of material. The bending
quality of the ASR damaged concrete are the elastic modulus and
tests were also performed on prismatic samples. The cylindrical samples were used
the tensile strength (in particular direct tensile strength). Note that in uniaxial and triaxial compression tests, and in permeability and porosity
the simple increase in the alkali content alone leads to a drop of the measurements.
mechanical properties of concrete without reactive aggregates Three days after the casting, the samples are de-moulded and kept in sealed
[26]. This phenomenon should also be taken into account in the bags for 28 days at ambient temperature. During this period of conservation, the
cylindrical samples were sliced to obtain smaller samples for various tests. This
evaluation of the mechanical degradation by ASR.
operation was done under dry conditions in order to avoid any leaching of alkalis
Most concrete structures can also be subjected to multi-axial [13]. After 28 days of maturation, a part of the samples were stored in a tempera-
mechanical loading during their service life. The effect of applied ture and humidity controlled chamber with temperature of 60 °C and relative
stresses on ASR development should be underlined [16,24,27]. humidity (RH) of 95% for ASR development. According to some previous investiga-
For instance, it is reported that the volumetric expansion of con- tions [2,4,8,13,16], this intermediately high temperature can facilitate the begin-
ning and increase the initial kinetics of the ASR process.
crete generated by ASR is constant whatever the stress state along
the three directions (axial creep stress and passive restraint),
which means the occurrence of expansion transfer in the less com- 2.2. Testing program and experimental device
pressed direction [24,27]. It is also necessary to understand the
The following tests and measurements are carried out during experimental inves-
influences of ASR on the mechanical properties of materials under
tigation: three point bending tests on prismatic samples (40 40 160 mm3),
such complex loading conditions. It is also useful to characterize uniaxial (Pc = 0 MPa) and triaxial compressive tests (Pc = 5 and 15 MPa) on cylindrical
the evolution of hydraulic properties induced by ASR. The results samples (£ 36 72 mm3), permeability measurement on cylindrical samples (£
obtained will contribute to the improvement of constitutive 36 27 mm3) and porosity measurements on cylindrical samples (£
models for ASR affected materials as well as design tools for the 36 16 mm3). The three point bending tests are performed in order to evaluate ten-
sile strength of material. In the present work, we also propose to investigate the ASR
ASR affected structures. Such results will be also helpful for the effect on compressive mechanical strength. As the compressive strength of cement-
efficiency of the maintenance procedures and repair works. The based material depends on hydrostatic stress, multi-axial compressive tests are
present work is devoted to the experimental study of the evolu- needed. In practice, triaxial compressive tests on cylindrical samples are commonly
tions of mechanical and hydraulic properties of ASR affected mor- carried out to determine failure surface of geomaterials. Such tests are performed
in the present work with two different confining pressures. These tests and measure-
tar. For this purpose, two groups of mortar samples are fabricated
ments were performed respectively after a period of 28 day maturation and after
respectively with reactive and non-reactive aggregates. Triaxial 100 days of conditioning of samples at 60 °C and 95% RH. The results obtained after
compressive tests with different confining pressures (0, 5 and 28 day maturation were taken as the reference values. The change in length of sam-
15 MPa) and bending tests are performed. The expansion, porosity ples was measured during the conditioning of samples. All tests and measurements
and permeability evolution are determined. Moreover, a micro- were performed after a cooling period of samples for at least 6 h at ambient temper-
ature. The length change of the prismatic samples (equipped with stainless steel
structural analysis is also carried out by using scanning electron studs at both end) during time are measured on three samples for each mortar with
microscopy (SEM) together with Energy Dispersive X-ray Spec- a mechanical retractometer.
trometry (EDS). In the following, the experimental program is first Uniaxial and triaxial compression tests were carried out using a hydraulic press
described and then the results obtained are presented and dis- under displacement controlled condition. The axial displacement rate used is 2 lm/
s. The tests were performed with loading-unloading cycles in order to evaluate the
cussed by comparing the behaviours of the two groups of mortar.
variation of elastic properties during mechanical loading. A triaxial cell was used in
order to apply confining pressures by injection of oil into the cell. The axial strain
2. Experimental investigation was measured by LVDT while the radial strain was obtained with the help of a spe-
cial strain ring [28]. However, the radial strains could not be measured in the case of
2.1. Material composition, sample preparation and conditioning uniaxial test because of brittle failure of the samples which seriously destroy the
strain ring. The triaxial compression tests are classically conducted with two load-
The study was undertaken on two groups of mortar whose composition was in- ing phases: increase of hydrostatic pressure until the desired value (5 or 15 MPa)
spired by Poyet et al. [6,13]. The mortar samples had a water to cement ratio of 0.5 and application of the deviatoric stress by keeping the radial stress constant. The
and sand to cement ratio of 3 (Table 1a). The non-reactive mortar (designated as NR deviatoric strength is identified as the maximum (peak) deviatoric stress reached
I. Yurtdas et al. / Construction and Building Materials 47 (2013) 165–174 167
0.45 Note that each mechanical, porosity and permeability test was re-
0.40 peated at least two times in the same conditions in order to be rep-
resentative. However, only the averaged data will be presented
0.35
here for the reason of clarity.
Expansion [%]
Fig. 2. General appearance (polished section, retro-diffused electrons) of NR (a) and R mortars (b) after maturation period: 1 – anhydrous clinker, 2 – aggregate, 3 – matrix in
relatively compact zone, 4 – matrix in relatively micro-porous zone.
168 I. Yurtdas et al. / Construction and Building Materials 47 (2013) 165–174
Fig. 3. Detail of matrices (fresh fracture, secondary electrons) of NR (a) and R (b) mortars after maturation period: 1 – occluded air void, 2 – C-S-Hs, 3 – portlandite, 4 –
ettringite.
Fig. 4. (a) Appearance of the gel in an occluded air void in a R sample after conditioning at 60 °C and 95% RH (fresh fracture, secondary electrons): 1 – gel, 2 – cementitious
matrix, (b) detail of (a): 1 – C-S-Hs more or less impregnated by the gel, 2 – smooth gel, 3 – smooth gel, 4 – re-crystallisation of the gel.
impregnates more or less the C-S-Hs in the vicinity of the occluded to form C-S-Hs. The occurrence of this reaction and the impregna-
air wall, which leads to a decrease of the micro-porosity (1). Then tion of C-S-Hs by gels would lead to the compact facies of C-S-Hs.
the gel forms a deposit in the occluded air void where it has Overall Figs. 4–7 seem to indicate that there is a densification of
smooth facies (2). Afterwards, a last generation of gel forms a the microstructure, particularly a densification of the ITZ [32] by
deposit on the previous one (3). There is a crystallisation on the the presence of ASR gels (filling of the ITZ porosity, some impreg-
surface (4) that remains free. nation of C-S-Hs and possible reaction of the silica of gels with
Fig. 5 shows the general appearance of two materials on pol- portlandite to form the new C-S-Hs), in the R mortar.
ished section (retro-diffused electrons). The R sample presents a
relatively more compact microstructure than the NR sample. The
interfacial transition zone (ITZ) of many aggregates of NR sample 3.3. Effect of ASR on mechanical behaviour
presents a micro-porous structure (dark zones around aggregates)
whereas that of R sample is very often compact. Moreover, micro- 3.3.1. Multi-axial compression
cracks also appear in the R sample due to ASR expansion (a part of An example of the results obtained is presented in Fig. 8. This
these microcracks could be linked to the sample preparation figure shows the evolution of the axial and radial strains versus
process). However, the occurrence and the intensity of these deviatoric stress during triaxial compression tests with a confining
microcracks are not always the same for each tested sample (not pressure of 15 MPa. It can be seen that after the maturation period,
shown here). the mechanical behaviour of the NR mortar and R mortar are very
On the other hand, Fig. 6 indicates that the NR sample has a rel- close with a slightly higher deviatoric strength for the later one.
atively micro-porous matrix near an air void (fresh fracture, sec- However, important differences appear after the conditioning of
ondary electrons) while the R sample has a relatively compact samples at 60 °C and 95% RH. For the NR mortar, the deviatoric
matrix. The details of these matrices near an aggregate are shown strength increases slightly due to cement hydration and the mate-
in Fig. 7. It can be seen from Figs. 6 and 7 that the C-S-Hs appear rial behaviour becomes more brittle compared with the behaviour
often in granular and fibrous form in the NR sample and in a rather after maturation period. For the R mortar, the increase in deviatoric
compact form in the R sample. Notice that the portlandite content strength is much more important, which indicates a much more in-
in the R sample seems lower than that in the NR sample. This could tense chemical activity. Its behaviour becomes also more brittle
be due to the fact that the silica of gels reacts with the portlandite compared with the behaviour after maturation period.
I. Yurtdas et al. / Construction and Building Materials 47 (2013) 165–174 169
Fig. 5. General appearance (polished section, retro-diffused electrons) of two mortars after conditioning at 60 °C and 95% RH: (a) NR sample: 1 – anhydrous clinker, 2 –
aggregate, 3 – matrix in relatively compact zone, 4 – micro-porous transition zone, (b) R sample: 1 – anhydrous clinker, 2 – aggregate, 3 – matrix in relatively compact zone, 4
– compact transition zone, 5 – microcrack.
Fig. 6. Appearance (fresh fracture, secondary electrons) of matrices of two mortars after conditioning at 60 °C and 95% RH: (a) NR sample with relatively micro porous matrix:
1 – granular C-S-Hs, 2 – portlandite, 3 – occluded air void, (b) R sample with relatively compact matrix: 1 – dense C-S-Hs, 2 – ettringite in an air void.
Fig. 7. Detail of matrices of two mortars near an aggregate after conditioning at 60 °C and 95% RH: (a) NR sample: 1 – granular and fibrous C-S-Hs, 2 – aggregate, (b) R sample:
1 – compact C-S-Hs, 2 – aggregate.
3.3.1.1. Multi-axial compressive strength. Fig. 9 presents the evolu- uniaxial compression, triaxial compression with 5 MPa confining
tion of deviatoric strength versus confining pressure for the two pressure and triaxial compression with 15 MPa confining pressure
sets of mortar after maturation period and after conditioning per- for the NR mortar. For the R mortar, the strength increases by
iod at 60 °C and 95% RH for the development of the ASR expansion. about 42%, 38% and 26% respectively in uniaxial compression, tri-
Depending on confining pressure, the mechanical strength in- axial compression with 5 MPa confining pressure and triaxial com-
creases during the conditioning period of samples. For instance, pression with 15 MPa confining pressure. On the other hand, the
the strength increases by about 14%, 12% and 6% respectively in results issued from the tests after the maturation period show that
170 I. Yurtdas et al. / Construction and Building Materials 47 (2013) 165–174
Table 2
Effect of ASR on the evolution of uniaxial compressive strength (fc) and Young’s modulus (E).
for the first and slightly increases, by 5%, for the second one. Conse- mortar becomes much lower than those of the NR mortar by
quently, it can be said that the ASR influences mainly the cohesion 8–17% and 37–39% respectively. Thus, the difference between the
of reactive material, at least in case of the present study. Moreover, elastic coefficients of the two materials is increased after the con-
the friction coefficient of the R mortar is equal to that of the NR ditioning of samples.
mortar while its cohesion is higher (by 8%) after the maturation per- Thus, in spite of an important increase in multi-axial compres-
iod. The conditioning of the samples at 60 °C and 95% RH leads to an sive strength, the Young’s modulus of the R mortar decreases
improvement of friction coefficient (6%) and cohesion (26%) of R slightly after ASR process. As underlined already by several
mortar with respect to NR mortar. authors, this shows a more important sensitivity of the Young’s
modulus to the development of ASR which leads to induced
3.3.1.2. Elastic properties. The evolutions of elastic properties were microcracking. This is the reason that a reduction in the Young
also investigated during multi-axial compression tests and com- modulus of reactive concrete is obtained whereas its compressive
pared between R and NR mortars. The initial values of elastic coef- strength continues to increase. Moreover, it seems that there is no
ficients are determined on the third cycle of loading-unloading of significant influence of the confining pressure up to 15 MPa on
each test. Moreover, as indicated earlier, the Poisson’s ratio could the decrease in Young modulus. A higher confining pressure
not be determined in the case of uniaxial compressive test (see would be necessary to observe a possible closure of the ASR in-
Section 2.2). The evolution of the initial Young’s modulus and Pois- duced microcracks. The reduction in the Young’s modulus of reac-
son’s ratio versus confining pressure are presented on Figs. 11 and tive concrete reported by Giaccio et al. [9], Larive [16], Multon
12 respectively. As for the deviatoric strength, the elastic coeffi- [24], Marzouk and Langdon [5] and Ahmed et al. [25] were 18%,
cients also depend on confining pressure. Both the elastic modulus 22%, 24%, 31% and 67% respectively (see also Table 2). Ahmed
and Poisson’s ratio increase with confining pressure. Regarding the et al. [25] even obtained a drop up to 92% for their highly reactive
conditioning effects, the initial Young’s modulus of NR mortar concrete (see also Table 2). On the other hand, even if the results
nearly does not change while its Poisson’s ratio increases by 12– of the present study indicate a sensitivity of Poisson’s ratio to the
13% after the conditioning of the samples at 60 °C and 95% RH. ASR induced damage, such a sensitivity is not observed by Larive
For the R mortar, the initial Young’s modulus decreases by 5–8% [16] and Multon [24]. However, a clear decrease of Poisson’s ratio
and the Poisson’s ratio by 20% after the conditioning of the samples is also reported by Giaccio et al. [9] on their highly reactive
at 60 °C and 95% RH. On the other hand, the tests after the matura- concrete.
tion period show that the Young’s modulus and Poisson’s ratio of
the R mortar are lower than that of the NR mortar by 0–11% and 3.3.2. Bending strength
12–13% respectively. After the conditioning of samples at 60 °C Table 4 gives the values of average bending strength of two
and 95% RH, the Young’s modulus and Poisson’s ratio of the R mortars after the maturation period at ambient temperature and
172 I. Yurtdas et al. / Construction and Building Materials 47 (2013) 165–174
Table 5 in the present work have clearly shown that the tensile strength
Average porosity of mortars after 28 day maturation and after conditioning at 60 °C and elastic modulus (and also Poisson’s ratio) were more signifi-
and 95% RH for ASR development.
cantly affected by the ASR induced microcracking than the com-
Material Porosity after maturation (%) Porosity after conditioning (%) pressive strength. Our results confirmed those reported in some
NR mortar 15.6 ± 0.5 15.5 ± 0.8 previous works. However, in the present study, we have further
R mortar 15.9 ± 0.8 17.2 ± 0.7 studied the evolution of microcracks induced by ASR process by
the permeability and porosity measurements. In particular, the
measurement of permeability provides a good way to evaluate
the ASR induced microcracking. Finally, microstructural analyses
have been well correlated with mechanical tests, permeability
Table 6
Average permeability of mortars after 28 day maturation and after conditioning at and porosity measurements. Note that it will be very useful to
60 °C and 95% RH for ASR development. investigate the effects of ASR at very long term. This investigation
will enable the influence of an additional ASR induced damage on
Material Permeability after Permeability after
maturation (10–16 m2) conditioning (10-16 m2) the evolution of mechanical properties, in particular compressive
strength, and hydraulic properties of R mortar to be evaluated.
NR mortar 5.6 ± 0.1 0.9 ± 0.1
R mortar 5.9 ± 0.1 1.4 ± 0.1
Acknowledgements
international conference on alkali-aggregate reaction in concrete, Quebec City; [25] Ahmed T, Burley E, Rigden S, Abu-Tair AI. The effect of alkali reactivity on the
2000, p. 593–02. mechanical properties of concrete. Constr Build Mat 2003;17:123–44.
[19] Clayton N. Structural performance of ASR affected concrete. In: Okada K, [26] Smaoui N, Bérubé MA, Fournier B, Bissonnette B, Durand B. Effect of alkali
Nishibayashi S, Kawamura M, editors. Proceedings of the 8th international addition on the mechanical properties and durability of concrete. Cem Concr
conference on alkali-aggregate reaction in concrete, Kyoto; 1989, p. 671–76. Res 2005;35:203–12.
[20] Swamy RN. Structural implication of alkali silica reaction. In: Okada K, [27] Multon S, Toutlemonde F. Effect of applied stresses on alkali-silica reaction-
Nishibayashi S, Kawamura M, editors. Proceedings of the 8th international induced expansions. Cem Concr Res 2006;36:912–20.
conference on alkali-aggregate reaction in concrete, Kyoto; 1989, p. 683–89. [28] Secq J. Collier de mesure de la déformation latérale d’une éprouvette lors
[21] Abe M, Kikuta S, Masuda Y, Tomozawa F. Experimental study on mechanical d’essais de compression, notamment uniaxiale ou triaxiale collier à lame,
behaviour of reinforced concrete members affected by alkali-aggregate Patent N 0505204; 2006.
reaction. In: Okada K, Nishibayashi S, Kawamura M, editors. Proceedings of [29] Loosveldt H, Lafhaj Z, Skoczylas F. Experimental study of gas and liquid
the 8th international conference on alkali-aggregate reaction in concrete, permeability of a mortar. Cem Concr Res 2002;32:1357–63.
Kyoto; 1989, p. 691–96. [30] Yurtdas I, Burlion N, Skoczylas F. Comportement différé des mortiers soumis
[22] Monette L, Gardner J, Grattan-Bellew P. Structural effects of the alkali-silica au séchage et effet induit sur la perméabilité. Eur J Environ Civ Eng
reaction on non-loaded and loaded reinforced concrete beams. In: Bérubé MA, 2008;12:701–21 [in French].
Fournier B, Durand B editors. Proceedings of the 11th international conference [31] Burlion N, Skoczylas F, Dubois T. Induced anisotropic permeability due to
on alkali-aggregate reaction in concrete, Quebec City; 2000, p. 999–1008. drying of concrete. Cem Concr Res 2003;33:679–87.
[23] Saint-Pierre F, Rivard P, Ballivy G. Measurement of alkali-silica reaction [32] Rivard P, Ollivier J-P, Ballivy G. Characterization of the ASR rim Application to
progression by ultrasonic waves attenuation. Cem Concr Res 2007;37:948–56. the Potsdam sandstone. Cem Concr Res 2002;32:1259–67.
[24] Multon S. Evaluation expérimentale et théorique des effets mécaniques de
l’alcali-réaction sur des structures modèles. Etudes et recherches des
Laboratoire des Ponts et Chaussées, OA46, Paris; 2004 [in French].