ACI Materials Journal/March-April 2005 67

ACI MATERIALS JOURNAL TECHNICAL PAPER

ACI Materials Journal, V. 102, No. 2, March-April 2005.MS No. 03-176 received May 9, 2003, and reviewed under Institute publication policies.

Copyright © 2005, American Concrete Institute. All rights reserved, including the makingof copies unless permission is obtained from the copyright proprietors. Pertinent discussionincluding authors’ closure, if any, will be published in the January-February 2006 ACIMaterials Journal if the discussion is received by October 1, 2005.

A large experimental program about alkali-silica reaction (ASR)was carried out on beam specimens at the Laboratoire Central desPonts et Chaussées (LCPC), with Electricité de France (EDF) as apartner. The aim was to provide data intended to validate a meth-odology of assessment of real structures affected by this expansivereactive mechanism. Three-dimensional deformations weremeasured on five 3 m-long plain or reinforced concrete beamsundergoing partial drying over a period of 14 months. The paperpoints out the effect of a moisture gradient over the depth of thebeams and the influence of reinforcement on the development ofASR-induced expansions. Tests on companion specimens (cylindersand prisms) were carried out to measure the material expansion andmechanical characteristics.

Keywords: alkali-silica reaction; beam; expansion; reinforcement.

INTRODUCTIONAlkali-silica reaction (ASR), observed for the first time in

the U.S. by Stanton,1 has been widely studied all over theworld. The Laboratoire Central des Ponts et Chaussées(LCPC, French Public Works Research Laboratory) hasrecently added its contribution: first, with an experimentalstudy to characterize the mechanical behavior of ASR-affected concrete,2 and, presently, with a large programcarried out on beam specimens, with Electricité de France(EDF) as a partner.3,4 The final objective of this study was tovalidate methodologies to assess the dimensional stabilityand the residual bearing capacity of ASR-affected structures.These methodologies are based on computational models(first-order engineering approach assuming ASR-inducedstrains as imposed strains5,6 or phenomenological models7).Models should assess the structural behavior of ASR-affected structures with a few parameters, namely, thereview of their environmental conditions and data fromresidual expansion tests.6

The effects of water and reinforcement on ASR-inducedexpansion have already been well demonstrated.8-16 Modelshave to be validated using documented large-scale experi-ments, emphasizing the influence of these two parameters.Therefore, in this program, plain and reinforced concretebeams were subjected to a vertical moisture gradient devel-oping between the immersed bottom and the drying upper faceexposed to a 30% relative humidity environment. Twocompanion beams were cast with a reference concrete mixturemade of nonreactive aggregate used to differentiate the normalbehavior of concrete beams under such conditions from theexpansive behavior due to ASR. All of the input data of theassessment computations (mechanical behavior of concrete,potential strains–due to shrinkage, water absorption, andASR–temperature, and moisture saturation within thestructures6) had to be measured during the experiment.Measuring devices were specially developed by LCPC.3,17-19

RESEARCH SIGNIFICANCEThe aim of the research was to validate models developed

to assess the dimensional stability and the residualbearing capacity of ASR affected structures5-7 in a critical andrepresentative situation. In this paper, the numerousmeasurements obtained during the experimental program(variation of water content, relative humidity measurements,local and global deformations, and deflection of the beams)are detailed and analyzed.

EXPERIMENTAL PROCEDURESpecimens

In real structures, moisture gradient occurs between thecore and the external surfaces. The effects of such environ-mental conditions are differential water contents withinstructures. For structures affected by ASR, they cause differ-ential ASR development and expansions. The strains of 3 m-long beam specimens subjected to a moisture gradient weremonitored to analyze the consequences of such gradients. Thebeam specimens had a cross section of 0.25 x 0.50 m (Fig. 1)and were simply supported (with a span of 2.80 m). Thedimensions had been chosen to allow numerous innermeasurements to be performed without disturbing themechanical behavior of the beam specimens. The moisturegradient was unidirectional to perform a simplified one-dimensional analysis.4 The upper face was exposed to air at30% relative humidity (forced convection), and the bottomwas immersed in water for 14 months (Fig. 1). Side faceswere covered by three watertight aluminum sheets 50 µm

Title no. 102-M08

Structural Behavior of Concrete Beams Affected byAlkali-Silica Reactionby S. Multon, J.-F. Seignol, and F. Toutlemonde

Fig. 1—Environmental conditions of the 3 m-long beamspecimens (temperature of beams: 37 ± 1 °C).

ACI Materials Journal/March-April 200568

thick just after the beams were demolded. This aluminumcover was directly applied to the concrete by an acrylic adhe-sive. It avoided moisture leakage20 so as to ensure a verticaldrying process. The temperature within the beams wasconstant at 37 ± 1 °C. To obtain such environmental conditions,they were stored in a large air-conditioned room (30% relativehumidity at 38 °C).

The large dimensions of the beams (3.0 x 0.25 x 0.50 m)and the dimensions of the air-conditioned room allowed fivebeam specimens to be studied (Fig. 1). Two beam specimenswere plain concrete beams (Table 1) to study the effect ofdifferential water supply, and three others were reinforcedwith deformed bars (fY = 500 MPa; Table 1, Fig. 2 and 3) toevaluate the effect of reinforcement.

The methodology for the assessment of ASR-affectedstructures is based on the potential expansion of the material,6

which directly depends on the concrete mixture and on theenvironmental conditions. The evolution of the concrete’smechanical characteristics was also required for computa-tional validation. Therefore, numerous companion speci-mens were cast to characterize the potential expansion ofconcrete mixtures (160 x 320 mm cylinders, 140 x 140 x280 mm prisms; Table 2) and the mechanical behavior atfour characteristic time-steps from 28 days to 1 year (110 x220 and 160 x 320 mm cylinders). All five beam specimenswere submitted to moisture gradient, while all of thecompanion specimens were kept in uniform external mois-ture conditions. The companion specimens used to charac-terize the potential expansion of reactive concrete werestored at 38 °C in three different environments (in water,sealed in an aluminum cover, and in air at 30% relativehumidity; Table 2). The companion specimens used for themechanical characterization were stored in an aluminumcover at 38 °C.

Concrete mixturesTwo concrete mixtures were studied: one with reactive

coarse aggregate and nonreactive sand and one with nonreac-tive aggregate only. The two mixtures were designed to haveclose rheological and mechanical properties (same slump,compressive strength, and modulus of elasticity). The cementcontent was common for bridge applications (410 kg/m3), andthe water-cement ratio had a value of 0.5. To obtain signifi-cant expansions, potassium hydroxide was dissolved in themixing water of the two concrete mixtures to increase theNa2Oeq content up to 1.25% of the mass of cement, asusually made to carry out experimental programs onASR.2,3,14,15 Moreover, the aim was to study the effect of amoisture gradient; it was thus impossible to expose thebeams to an external accelerating alkali-enriched solutionbecause it would have affected the alkali content in thebottom of the beam and not in the middle part. The differen-tial ASR-induced strains would have been caused by differ-ences in alkali content and not by moisture gradient.

Associated expansion tests oncompanion specimens

Weight and longitudinal deformations were measured onall the companion specimens (160 x 320 mm cylinders and140 x 140 x 280 mm prisms) used for ASR expansion char-acterization according to the methodology recommended byReference 2. Anisotropy of ASR-induced expansions havealready been discussed according to casting and crackingdirections.21,22 It was considered essential data to validatemodels. During the experiment, ASR-induced strains werethus determined both perpendicularly and along the castingdirection. Because deformations were measured along theaxis of companion specimens, expansions perpendicular tothe casting direction were performed on 140 x 140 x 280 mm

S. Multon is a graduate teacher in civil engineering at the Laboratoire Matériaux etDurabilité des Constructions [LMDC (INSA-UPS)], Toulouse, France.

J.-F. Seignol is a research engineer at Laboratiore Central des Ponts et Chaussées(LCPC). His research interests include modeling of concrete nonlinear behavior.

ACI member F. Toutlemonde is the Head of LCPC Structures Laboratory. He isSecretary of the ACI Paris Chapter. His research interests include high-performanceconcrete and ultra-high-performance fiber-reinforced concrete applications forbridges, concrete dynamic behavior and shock strength, fatigue strength, and durabilityand serviceability of reinforced concrete bridge structures.

Table 1—Beam detailsBeams Mixtures Percent area of longitudinal steel

B1 Reactive 0

B2 Nonreactive 0

B3 Reactive 0.45

B4 Reactive 1.80

B5 Nonreactive 0.45

Fig. 2—Elevation and cross section through B3 and B5(d,d′: distances from top fibers of lower and upper steelbars—dimensions in mm).

Fig. 3—Elevation and cross section through B4 (d,d′: distancesfrom top fibers of lower and upper steel bars—dimensionsin mm).

Table 2—Specimens detailsCompanion specimens Quantity Types Review/environment

On cylindersDiameter: 160 mmHeight: 320 mm

6 Reactive 38 °C, sealed

6 Reactive 38 °C, in water

3 Reactive 38 °C, 30% relative humidity

On prisms140 x 140 x 280 mm

6 Reactive 38 °C, sealed

6 Reactive 38 °C, in water

ACI Materials Journal/March-April 2005 69

prisms while expansions along the casting direction weremeasured on 160 x 320 mm cylinders (Table 2, Fig. 4).

Measurements of moisture distribution within beam specimens

The water content of concrete affects the time evolutionand the potential range of the ASR development.8-10 It isthus an essential input data of models5-7 that had to be deter-mined. Precise measuring devices were used to evaluate theconsequences of the environmental conditions (38 °C and30% relative humidity) on the moisture distribution withinthe beam specimens. Capacitive relative humidity sensors(located in precast cavities) measured the inner relativehumidity of the concrete. Weighing equipment assessed themass variations of the beams, called in this paper the “globalmass variation.” Finally, an automated gammadensitometrydevice determined the profile of local mass variations alongthe upper 300 mm of the beams. Indeed, the gammadensito-metry device measured the local mass variation of theconcrete of the beams every 20 mm from the upper facedown to a depth of 300 mm.17

Structural behavior of beam specimensAfter having been sealed in aluminum for 28 days of

curing, the beams were subjected to the moisture gradient for14 months (Fig. 1). The beams were not loaded during thisperiod. The evolutions of strains were only due to dryingfrom the top surface, absorption of water from the bottomside, and, above all, to ASR for the three reactive beams.

The local and global deformations of the beams weredetermined using stiffness-free vibrating wire sensors(VWS) specially designed for this study.3,18,19 They wereused in four lengths: 85 mm for the local inner strainmeasurements (these sensors were embedded within thebeams before casting), and 250, 500, and 3000 mm for globalexternal measurement of width, height, and length of thebeam specimens, respectively. Deformations at seven depthlevels for the longitudinal direction (0.08, 0.17, 0.27, and0.37 m for inner VWS, and 0.03, 0.23, and 0.47 m forexternal VWS; Fig. 5 and 6), six depth levels for the trans-verse (0.08, 0.17, 0.27, and 0.37 m for inner VWS, and 0.00and 0.50 m for external VWS; Fig. 5 and 6), and four depthlevels for the vertical direction (0.08, 0.17, 0.27, and 0.37 mwith inner VWS; Fig. 5 and 6) were measured using 54 VWSper beam specimen. The strains at every depth level weremeasured by two or three VWS to obtain statistically repre-sentative measurements (Fig. 5 and 6).

Due to their short length, the measurements of the innerVWS (85 mm in length) were representative of the local

behavior of the surrounding concrete. The presence of reac-tive aggregates close to them could induce more heteroge-neous measured deformations for these sensors than for theouter VWS, which integrated overall deformations of thebeams. The global behavior of the beams was also monitoredusing 10 common displacement sensors to measure thedeflections (five on each side; Fig. 6). This large number ofsensors had been chosen to avoid loss of information and totake into account the heterogeneity of concrete, particularlyof ASR expansions.22 Local and global deformations couldthus be compared to check the consistency of the measure-ments with respect to the structural behavior of the beams.

ANALYSIS OF EXPERIMENTAL RESULTS—INPUT DATA FOR STRUCTURAL ASSESSMENT

Material mechanical propertiesCylindrical standard specimens were tested to characterize

the evolution of the mechanical behavior of the two concretemixtures. The Young’s modulus, compressive strength, andtensile strength at 28, 90, 180, and 365 days were determinedexperimentally on 110 x 220 mm (splitting tensile strength)and 160 x 320 mm (Young’s modulus, compressivestrength) sealed cylinders. Every mechanical property wasmeasured on three specimens for every time-step (Table 3).

Fig. 4—Cylindrical and prismatic companion specimens:casting directions and measurement directions.

Fig. 5—Stiffness-free embedded vibrating wire sensors.

Fig. 6—Stiffness-free external VMS and displacement sensors.

Table 3—Mechanical characteristics of two concrete mixtures studied

Test time-steps 28th day 90th day 6th month 1 year

Reactiveconcrete

fc, MPa 38.4 42.4 41.8 42.1

E, MPa 37,300 37,200 30,100 29,700

ν 0.22 0.24 — —

ft , MPa 3.2 3.4 3.0 3.3

Nonreactive concrete

fc, MPa 35.5 40.6 40.4 41.8

E, MPa 38,700 38,400 37,800 40,700

ft, MPa 3.4 3.8 3.2 3.8

Notes: fc = mean compressive strength on cylinders (diameter: 160 mm; height:320 mm). E = Young’s modulus on cylinders (diameter: 160 mm; height: 320 mm). ν =Poisson’s ratio. ft = mean splitting tensile strength on cylinders (diameter: 110 mm;height: 220 mm).

70 ACI Materials Journal/March-April 2005

The mean Young’s modulus of the reactive concrete,measured on three specimens, decreased from 37,200 to29,700 MPa (approximately 20%) between the 90th and the365th exposure days. The decrease is significant comparedwith the standard deviations of approximately 600 MPa. Theother mechanical properties show a usual evolution forconcrete, with no decrease due to ASR. The increase instrength due to the hydration process appears to be largerthan the losses due to ASR. In previous papers, some authorsreported a decrease of all the mechanical properties,23-25

while others only described the decrease of stiffness.2,15

Therefore, the results of this experimental program areconsistent with the latter results. The differences with thefirst results can be explained by the differences in concretemixture, particularly in the nature of the reactive aggregate.

ASR-induced expansions measured on companion specimens

The ASR-induced expansions were measured for thedetermination of the imposed strains, as the most importantinput data for affected structures assessment.5,6 ASR-induced expansion usually exhibits anisotropy relative to thecasting direction and to the crack patterns.21,22 For the beamspecimens, the horizontal (longitudinal and transverse)expansions were perpendicular to the casting direction (Fig. 1),which corresponds to expansions measured on 140 x 140 x280 mm prisms. The vertical expansion of the beams corre-

sponds to ASR-induced expansions measured along thecasting direction (on 160 x 320 mm cylinders). Verticalexpansion measured on cylinders was approximately twiceas large as horizontal expansion measured on prisms (Fig. 7).It illustrates the anisotropy of ASR expansions measured onstress-free specimens, as already observed in Reference 21.This has to be accounted for in the models. Figure 7 alsoillustrates the differences of the ASR development withrespect to the water supply. Expansions were measured onthe companion specimens kept sealed under aluminum(between 0.05 and 0.10%). Therefore, significant ASRexpansions can occur without an external water supply; theremaining water after cement hydration is sufficient. ASRexpansions, however, were larger with an external watersupply. They were approximately 2.5 times larger forcompanion specimens stored in water than for sealed ones.The aluminum cover was not perfectly watertight (negativemass variations; Fig. 7(a)). It decreased the mass lossessignificantly (between 0.5% and 1%), however, in spite ofsevere drying conditions (the specimens were kept in the30% relative humidity atmosphere). The scattering of ASR-induced expansions was quantified; the difference betweenthe maximum and the minimum value ranges from strains ofapproximately 0.12% for the cylinders and 0.08% for theprisms in water to approximately 0.05% for the sealedcompanion specimens, which namely represents 50% of themean expansion. It is due to the heterogeneity of ASR expan-sions, as already shown in previous experimental programs.It can be explained by the heterogeneous repartition of thereactive silica inside the limestone aggregate.21

These measurements on specimens (Fig. 7(b)) can be usedas input data for models assuming ASR-induced expansionsas imposed strains5,6 or to determine the input parameters ofthe phenomenological model developed by Capra andSellier.7 The scattering of ASR-induced expansions is to beconsidered regarding the expected accuracy of the structuralcalculations that use these strains as input data.

Characterization of local water saturation in tested beam specimens

Measurements of the global (by weighing) and local (bygammadensitometry) mass variations can be directly inter-preted in terms of water movements. The time-evolution of theglobal mass variations of the five beam specimens had beenverified as linear with the square root of time.26 No markeddifference had been noticed either between reactive andnonreactive beams or between plain and reinforced beams.26

Therefore, a mean representative global mass variation perunit of drying area can be defined for the five beams

(1)

where ∆Massglobal is the global mass variation per unit ofdrying area (g/m2) and obeys a diffusion process versus Time(in days) elapsed from the first day of drying.26 This equationrepresents the mean mass variation of the five concrete beamssubmitted to the moisture conditions described previouslyduring the 430 days of exposure. It has to be noted that theintercept at t = 0 is positive. It can be explained by thesuccessive water movements during the exposure. First, theincrease of mass due to water absorption in the lower part ofthe beams was larger than the mass lost by drying in theupper part. Thus, the mass variation was positive during the

∆Massglobal 218 Time– 591+=

Fig. 7—Reactive companion specimens kept in water, sealedunder aluminum, and kept in air at 30% relative humidity(parallel to casting direction—black dots; perpendicular tocasting direction—voided dots).

ACI Materials Journal/March-April 2005 71

first days. Then, the mass losses became the largest, whichexplained the negative slope. The linear representationallows mass variation of the five beams to be represented byonly one equation. The representation of the short-termbehavior, however, is rough.

Gammadensitometry results lead to mean profiles of localmass variation along the upper 300 mm of the beams (Fig. 8).Calculations, using weighing and gammadensitometry data, wereused to estimate a mean sorptivity of approximately0.019 mm.s–1/2. It describes the penetration of water in thelower part of the beams,26 and profiles of local mass variationsalong the 500 mm height of the beams can be established(Fig. 8). ASR-induced expansions are influenced by themoisture conditions.8-10 The experimental determination oflocal saturation can be used to determine the range of theASR-induced strains along the height of the beams.4

Relative humidity measurements using capacitive sensorscomplemented the quantification of moisture transfer. Aftera 14-month-long exposition:

1. At 0.08 m from the drying face, a decrease of approxi-mately 12% relative humidity was observed from a meaninitial relative humidity approximately equal to 97%. Theinternal relative humidity always remained greater than 84%for all of the beams. At this temperature and alkali content,this appears higher than the relative humidity thresholdknown to stop ASR;8,9

2. For the two intermediate sensors (at depths of 0.17 and0.27 m), the relative humidity decreased approximately 3%and stayed above 95%; and

3. The lower sensor, located at a depth of 0.37 m, showedan increase of relative humidity up to 100% in approxi-mately 15 days.26

STRUCTURAL BEHAVIOR OF BEAM SPECIMENSTo validate the capability of models to evaluate the

behavior of the beam specimens affected by ASR, calculationsusing the previous input data might be compared with the realdeformations measured in the beams, detailed as follows.

Longitudinal strains and beam deflectionsImposed differential strains due to the moisture gradient

led to significant deflections. Reinforcement acted bydecreasing the strains along the longitudinal direction, whichcould be explained as a chemical prestressing effect in thereinforced and reactive beams.15,27-29

Beam analysis—For plain concrete Beams B1 and B2, theseven mean longitudinal deformations measured by theVWS fit quite well with a linear distribution along thevertical direction (Fig. 9 and 10). The consistency of theouter and inner measurements is thus illustrated. Themeasurements of the 3 m-long sensors characterized theglobal behavior. They closely fit with a straight line that repre-sents the plane cross sections of the beam. The five beamsshowed the same longitudinal behavior with plane crosssections remaining plane during the experiment. The meanquadratic deviation of deformations measured by the innerVWS relative to this line was lower than 0.02% (200 µm/mobtained for B1). Because the behavior of the beams fulfilledthe Navier-Bernoulli hypothesis (plane sections remainingplane during the deformation), calculations according to thestrength of materials concepts can be performed. The integrationof the rotation of the cross section allows the equation of thedeflection along the span to be established

(2)

where χmeas is the rotation of the cross section measuredwith the 3 m-long VWS (m–1), y(x) is the deflection alongthe abscissa x, and l is the span of the beams (l = 2.8 m).

The deflection measured by the displacement sensors andthe deflection calculated by Eq. (2) have been plotted in Fig. 11for B1. The mean quadratic deviation between the theoreticalparabola and the measurements was evaluated to 0.230 mm forB1 at the last time-step, when the deflection at midspanreached more than 5 mm. The consistency of the measurementby VWS and displacement sensors is thus quantified with adiscrepancy of less than 5%. The behavior of the other beams

y x( )χmeas

2-------------x x l–( )=

Fig. 8—Moisture measurements on beam specimens: meanprofiles of local mass variation at several time-steps.26

Fig. 9—Deformation of cross section of Beam B1.

Fig. 10—Deformation of cross section of Beam B2.

72 ACI Materials Journal/March-April 2005

was quite the same with a higher discrepancy (30% for B430).Two reasons explained the discrepancy: the measured deflectionwas very small (less than 500 µm) and quite close to theaccuracy of this measurement (due to the assembly of thedisplacement sensors), and due to the closeness of reinforcementbars, the long VWS measurements might locally deviatefrom the mean plane section.

Because plane sections remained plane, it is possible tosynthesize all the longitudinal strains and deflectionsmeasurements by only two data, namely, the mean relativelength variation, quantified by the measurement of the 3 m-long VWS at a depth of 0.23 m (near the uncracked section

neutral axis; Fig. 12(a)), and the deflection at midspan,which characterizes the mean curvature (Fig. 12(b)).

Analysis of plain concrete beams under moisturegradient—As expected for B1, shrinkage was observed inthe upper part, and ASR expansions in the bottom (Fig. 9).The phenomena were not symmetrical, and the line of zerodeformation was very close to the drying face. This provesthe predominance of the ASR-induced strains comparedwith shrinkage. The measurements on the plain concrete andnonreactive Beam B2 showed different behavior (Fig. 10).Shrinkage took place all along the upper 250 mm. Therotations of the plane sections led to midspan deflections of5.4 mm for B1 (that is, a deflection over length ratio ofapproximately 1:500) and approximately 0.9 mm for B2(ratio of approximately 1:3000) (Fig. 12(b)). As expected,the imposed curvature due to drying, water absorption, andASR led to significant effects even without external loading,which constitutes a crucial checkpoint in the process ofmodel validation.

Analysis of reinforced beams under moisture gradient—For nonreactive beams, the reinforcement (B2 versus B5)had no visible influence due to relatively low strains. Thedifference between evolutions of both beams (Fig. 12) onlyconsists in the scattering of long VWS measurements whencarried out on large specimens. For reactive beams, thecomparison of the behavior of the Beams B3 and B4 with B1emphasizes the significant effect of the reinforcement. After400 days, the overall length relative variation at the depth of0.23 m was approximately 0.08% for B3 and 0.04% for B4,compared with 0.11% for B1 (Fig. 12), which illustrates theimportant effect of restraint. The deflection reached only 1.1and 0.3 mm for B3 and B4, respectively, instead of 5.4 mmfor B1. Yet the evolution of the deflection exhibited aspecific mechanism for reinforced beams; at the beginningof the exposure period, the deflection increased, but afterapproximately 75 and 150 days, the deflection decreasedslowly. This can be explained in two stages: first, the ASR-induced expansions appear similarly as for Beam B1, mainlyin the lower part of the beam (due to the water supply). Thereinforcement then restrains these deformations and causescompression in the lower part. An internal negative bendingmoment takes place as with prestressed beams.15,27-29

Transverse strainsTransverse and vertical expansions were measured to

investigate the ASR-anisotropy21,22 and the possibleinduced effects of restraint on the expansions31 when ASRoccurs in simply supported beams.

Plain concrete beams under moisture gradient—Trans-verse strains were quite homogeneous in the major part ofthe reactive beam (B1) (Fig. 13(a)); large expansions tookplace from depths between 0.17 and 0.50 m (between 0.12and 0.20% after 430 days of exposure). For the same depths,transverse deformations of B2 were positive or almost null,which corresponds to water absorption or lack of shrinkage(Fig. 13(b)). At the depth of 0.08 m, the deformations weremuch smaller (less than 0.05%) for B1, while B2 showedrather large shrinkage (approximately 0.03% after 430 daysof exposure). On the drying face of B1, shrinkage reachesapproximately 0.05% (the transverse strain of B2 could notbe measured due to the dysfunction of the correspondingVWS). The large strain gradient occurring along the upper150 mm of B1 should imply large, nonuniform stresses, andthus, internal cracking.

Fig. 11—Comparison between measured deflections(marks) and those obtained by integration of rotation ofcross section of Beam B1 (lines).

Fig. 12—Structural behavior of five beams.

ACI Materials Journal/March-April 2005 73

Analysis of reinforced concrete beams under moisturegradient—As with the longitudinal direction, the reinforce-ment for nonreactive beams (B2 versus B5) had no visibleinfluence. The transverse strains of the nonreactive beamscan thus be summarized by mean evolutions (Fig. 14(b)). Itshows the range of shrinkage and water absorption strainsof beams subjected to such a moisture gradient. The differ-ence between the strains at 0.50 and 0.37 m (approximately30 × 10–6) gives the order of magnitude of the scattering ofmeasurements by VWS.

The transverse deformations measured at all depths werequite the same for the two reactive reinforced concretebeams and B1. As an example, results of all VWS at thedepth of 0.17 m are shown in Fig. 15. Along this direction,ASR expansions were only affected close to the stirrups butnot in the whole beam. Therefore, it is possible to summarizeall of the transverse measurements in only one figure.Figure 14(a) represents the mean deformation of the threereactive beams (B1, B3, and B4) at the six different depthsof transverse measurements.

Vertical strainsAnalysis of plain concrete beams under moisture

gradient—Assuming a vertical moisture gradient, the rela-tive humidity depends only on depth. Thus, the potentialexpansion can be considered as constant with the depth. Thevertical deformations can be considered to be totally free ofrestraint for the nonreinforced beams. Therefore, no stressshould have appeared along this direction except due toASR-expansion heterogeneity. For B1, the concrete that wasthe closest to the water supply showed the larger ASR-

induced expansions (Fig. 16(a)). As expected, the amplitudeof strains at the depth of 0.37 m was close to expansionsmeasured on cylinders kept in water (Fig. 17(a)). Largedeformations were obtained between the depths of 0.17 and0.37 m (approximately 0.18 and 0.24% after 430 days ofexposure). In the same conditions, the deformationsmeasured along the height of B2 were similar to transverseones: positive in the bottom, from depths between 0.50 and0.25 m, and negative in the upper part (Fig. 16(b)).

At the depth of 0.08 m, corresponding to a zone between0.04 and 0.12 m due to the 85 mm length of the inner VWS

Fig. 13—Transverse strains along height of plain concretebeam specimens. Fig. 14—Mean evolution of transverse strains at six depth

levels of measurements.

Fig. 15—Comparison of transverse strains of three reactiveBeams—B1, B3, and B4—at depth of 0.17 m (three measure-ment locations for each beam).

74 ACI Materials Journal/March-April 2005

(Fig. 5), vertical expansions reached more than 0.08% afterthe 430 days of exposure for B1. Thus, even if drying graduallyoccurred at this depth, it was not fast enough to stop the ASRdevelopment (Fig. 17(a)). Strains measured at 0.08 m werequite close to expansions measured on sealed cylindricalspecimens (Fig. 17(a)). This result is consistent because thesame mass loss of approximately 1% was measured at 0.08 mon beam specimens (Fig. 8) and on sealed companion speci-mens (Fig. 7). Therefore, vertical measurements of the plainreactive beam showed expected strain ranges in comparisonwith the profile of mass variations and the relativehumidity measurements. Computational analysis of theseresults will bring a precise relation between the moisturedistribution of the ASR-affected concrete and the ASRexpansion development.

Reinforced concrete beams—As with the two otherdirections, no influence of the reinforcement for nonreactivebeams (B2 and B5) was noted30 (the mean evolution of verticalstrains of the nonreactive beams is plotted in Fig. 17(c)).Along the vertical direction, the expansions measured on thetwo reactive and reinforced beams (B3 and B4) were closefor the four depths of measurements and showed no differencebetween the two reinforcement ratios. The different measurescan be summarized by mean strains (Fig. 17(b)). In comparisonwith B1, the mean deformations of B3 and B4, after the 430 daysof exposure, are always smaller: approximately 15% at 0.37 m,25% at 0.27 m, and approximately 50% at the depths of 0.17 and0.08 m (Fig. 17(a) and (b)). The reductions can be explained bya local effect of the stirrups (which depends on the distancebetween the VWS and the reinforcement bar). Moreover,

transverse cracks observed on the drying face of the reinforcedbeams (Fig. 18(c) and (d)) imply a different behavior of concretein the upper parts of B3 and B4 than in B1. They modify the ASRgel development and can cause an apparent decrease of the ASR-induced expansions. Therefore, it is not possible to summarizethe vertical measurements in only one mean evolution as for thetransverse ones. This can be done, however, for the reactive andreinforced Beams B3 and B4. Thus, all of the vertical data can besummarized by Fig. 17: the deformations of B1 (Fig. 17(a)),mean evolutions for B3 and B4 (Fig. 17(b)), and meanevolutions for B2 and B5 (Fig. 17(c)).

Fig. 16—Vertical strains along height of plain concretebeam specimens.

Fig. 17—Vertical strains at four depth levels of measurements.

ACI Materials Journal/March-April 2005 75

CrackingThe two nonreactive beams did not exhibit any cracking,

even on the drying face. The ASR development causedcracks in Beams B1, B3, and B4 on the immersed and on theupper faces. Before cracking, the behavior may be assumedas elastic, but after the onset of the first cracks, models haveto take into consideration irreversible strains.4

Cracking of the bottom face of B1 appeared between the130th and 160th day of exposure with a common map-cracking (as described in Reference 12 and Fig. 18(a)). Thefirst cracks, however, appeared transversely. For B3, theyappeared around the 90th day (Fig. 18(b)), and for B4,between the 190th and 210th day of exposure (Fig. 18(d)).Visual inspection of the immersed face was quite difficultdue to the water tank, and it was only possible when thebeams were lifted for the weighing. Thus, the cracks couldnot be noticed as soon as they occurred, but only at thefollowing weighing time-step, which explained the 20 to30 days of uncertainty.

The cracks of the two reinforced beams appear to be morelongitudinal, following the reinforcement (as observed inReferences 12, 26, and 29). For the three reactive beams,perfectly transverse cracks were observed on the upper faceafter approximately 200 days. B1 only exhibited three trans-verse cracks at approximately 0.65 m from the ends andspaced at approximately 0.85 m. Their occurrence can beexplained by the increase of nonuniform stresses due tolongitudinal restrained shrinkage in the upper part. For thereinforced concrete beams, crack spacing was approximately0.35 and 0.20 m for B3 and B4, respectively, which wasclose to the shear reinforcement spacing (Fig. 18(c) and (e)).

CONCLUSIONSThis experimental study confirms important results about

ASR expansions measured on the stress-free companionspecimens:

1. The anisotropy of ASR-induced expansions has beenverified with vertical swellings twice as large as horizontalones;

2. Significant ASR expansions (0.10%) can occur withoutan external water supply and lead to significant structuraldegradation;

3. An increase in ASR expansions with an external watersupply has been quantified. Detailed results have beenreported concerning plain and reinforced concrete beamspecimens subjected to shrinkage and ASR for 14 months,which caused significant strains and deflections;

4. This paper provides data for the validation of ASRmodels5-7 relative to the studied concrete: input data (moisturedistribution and potential expansions) and validation data(strains and deflections measured on the beam specimens). Ithas been checked that the beam specimens fulfilled theNavier-Bernoulli assumption (cross sections remained planeduring the deformation). Moreover, the longitudinal rotationof the cross sections was consistent with the deflectionsmeasured on the beams. The structural behavior of the fivebeams thus satisfied the strength of materials assumptions;

5. Transverse and vertical measurements of the nonreactivebeams showed expected strain ranges compared with the profileof mass variations and the relative humidity measurements,which validates moisture input data for the analysis;

6. For the reactive beams, the water distribution inducedlarge expansions from 0.17 to 0.50 m along the transverseand vertical directions, smaller expansions at the depth ofFig. 18—Cracking of reactive beams.

ACI Materials Journal/March-April 200576

0.08 m, and shrinkage on the drying face. The dominantinfluence of water supply on ASR-induced expansion wasthus confirmed and quantified on beam specimens submittedto local intense drying; and

7. The effect of reinforcement on ASR-induced strainswas quantified on beams; it was highly significant along thelongitudinal direction, with a large decrease of strains anddeflections for the reinforced concrete beams (B3 and B4)compared with the plain beam (B1), while the local effectsof stirrups were hardly significant on vertical and transversedeformations.

Once models will be validated in the framework of this well-controlled study, they will be used to assess the dimensionalstability and the residual bearing capacity of real ASR-affectedstructures (for example, bridges and dams). Assuming the roleof water is close for all ASR-affected concrete, the differenceswith this study will be mainly the reactivity of aggregate and itseffects on the range of ASR expansions. For models actuallystudied by LCPC5,6 and EDF,7 this parameter can bedetermined from a stress-free ASR expansion test. Thus, thestructural behavior of ASR-affected structures will bedetermined from the review of their environmental conditionsand residual expansion tests.

ACKNOWLEDGMENTSThe authors wish to thank the teams of technicians at LCPC that partici-

pated in the measurements reported in this paper (F.-X. Barin, R. Bernot, J.Billo, S. Dubroca, M. Estivin, L. Lauvin, A. Laplaud, G. Leclainche, and A.Maloula). They would also like to thank S. Prené, H. Tournier, E. Bourdarot,A. Jeanpierre, and D. Chauvel (EDF) for their analysis and financial supportand C. Larive (CETU, Lyon, France) for instigating this project.

REFERENCES1. Stanton, T. E., “Expansion of Concrete through Reaction between Cement

and Aggregate,” Proceedings, ASCE, V. 66, No. 10, 1940, pp. 1781-1811.2. Larive, C., “Apports combinés de l’expérimentation et de la

modélisation à la compréhension de l’alcali-réaction et de ses effetsmécaniques,” OA 28, Laboratoire Central des Ponts et Chaussées, ed.,Paris, 1998. (in French)

3. Larive, C.; Toutlemonde, F.; Joly, M.; Laplaud, A.; Derkx, F.; Merliot, E.;Multon, S.; Bourdarot, E.; Prené, S.; and Jeanpierre, A., “Structural Effectsof ASR in France on Real and Laboratory Structures,” Proceedings of the11th International Conference on Alkali-Aggregate Reaction, Quebec,Canada, 2000, pp. 979-988.

4. Multon, S.; Seignol, J.-F.; and Toutlemonde, F., “Large GirdersSubjected to Alkali-Silica Reaction,” Durability of Concrete, Proceedingsof the Sixth CANMET/ACI International Conference, SP-212, V. M. Malhotra,ed., American Concrete Institute, Farmington Hills, Mich., 2003, pp. 299-318.

5. Ulm, F.-J.; Coussy, O.; Li, K.; and Larive, C., “Thermo-Chemo-Mechanics of ASR Expansion in Concrete Structures,” Journal ofEngineering Mechanics, ASCE, V. 126, No. 3, 2000, pp. 233-242.

6. Li, K., and Coussy, O., “Concrete ASR Degradation: From MaterialModeling to Structure Assessment,” Concrete Science and Engineering,V. 4, 2002, pp. 35-46.

7. Capra, B., and Sellier, A., “Orthotropic Modelling of Alkali-AggregateReaction in Concrete Structures: Numerical Simulations,” Mechanics ofMaterials, V. 35, 2003, pp. 817-830.

8. Olafsson, H., “The Effect of Relative Humidity and Temperature onAlkali Expansion of Mortar Bars,” Proceedings of the 7th InternationalConference on Alkali-Aggregate Reaction, Ottawa, Canada, 1986, pp. 461-465.

9. Kurihara, T., and Katawaki, K., “Effects of Moisture Control and Inhi-bition on Alkali Silica Reaction,” Proceedings of the 8th InternationalConference on Alkali-Aggregate Reaction, Kyoto, Japan, 1989, pp. 629-634.

10. Larive, C.; Laplaud, A.; and Coussy, O., “The Role of Water inAlkali-Silica Reaction,” Proceedings of the 11th International Conference

on Alkali-Aggregate Reaction, Quebec, Canada, 2000, pp. 61-70.11. Swamy, R. N., and Al-Asali, M. M., “Control of Alkali-Silica Reaction

in Reinforced Concrete Beams,” ACI Materials Journal, V. 87, No. 1, Jan.-Feb. 1990, pp. 38-46.

12. Koyanagi, W.; Rokugo, K.; and Uchida, Y., “Mechanical Propertiesof Concrete Deteriorated by Alkali Aggregate Reaction under VariousReinforcement Ratios,” Proceedings of the 9th International Conferenceon Alkali-Aggregate Reaction, London, 1992, pp. 556-563.

13. Fan, S., and Hanson J. M., “Length Expansion and Cracking of Plainand Reinforced Concrete Prisms Due to Alkali-Silica Reaction,” ACIStructural Journal, V. 95, No. 4, July-Aug. 1998, pp. 480-487.

14. Ballivy, G.; Khayat, K.; Gravel, C.; and Houle, D., “Influence ofReinforcement Steel on the Expansion of Concrete Affected by Alkali-Aggregate Reaction,” Proceedings of the 11th International Conference onAlkali-Aggregate Reaction, Quebec, Canada, 2000, pp. 919-928.

15. Monette, L. J.; Gardner, N. J.; and Grattan-Bellew, P. E., “ResidualStrength of Reinforced Concrete Beams Damaged by Alkali-Silica Reaction—Examination of Damage Rating Index Method,” ACI Materials Journal, V. 99,No. 1, Jan.-Feb. 2002, pp. 42-50.

16. Mohammed, T. U.; Hamada, H.; and Yamaji, T., “Alkali-Silica Reaction-Induced Strains over Concrete Surface and Steel Bars in Concrete,” ACIMaterials Journal, V. 100, No. 2, Mar.-Apr. 2003, pp. 133-142.

17. Multon, S.; Merliot, E.; Joly, M.; and Toutlemonde, F., “WaterDistribution in Beams Damaged by Alkali-Silica Reaction: Global Weighingand Local Gammadensitometry Measurements,” Materials and Structures,V. 37, No. 269, 2004, pp. 282-288.

18. Joly, M., “Capteur à fil vibrant notamment destiné à mesurer lesdéformations d’une structure,” Patent Application No.: 9915522 (Dec. 9,1999), Publication No.: 2802299 (June 15, 2001).

19. Bourquin, F., and Joly, M., “A Magnet-Based Vibrating Wire Sensor,”Proceedings of the International Conference on Smart Technology Demon-stration and Devices, Edinburgh, Scotland, 2001.

20. Toutlemonde, F., and Le Maou, F., “Protection des éprouvettes debéton vis-à-vis de la dessiccation—Le point sur quelques techniquesde laboratoire,” Bulletin des LPC, 1996, V. 203, pp. 105-119. (in French)

21. Larive, C.; Joly, M.; and Coussy, O., “Heterogeneity and Anisotropyin ASR-Affected Concrete—Consequences for Structural Assessment,”Proceedings of the 11th International Conference on Alkali-AggregateReaction, Quebec, Canada, 2000, pp. 969-978.

22. Clark, L. A., “Modeling the Structural Effects of Alkali-AggregateReaction on Reinforced Concrete,” ACI Materials Journal, V. 88, No. 3,May-June 1991, pp. 271-277.

23. Pleau, R.; Bérubé, M. A.; Pigeon M.; Fournier, B.; and Raphael, S.,“Mechanical Behavior of Concrete Affected by ASR,” Proceedings of the8th International Conference on Alkali-Aggregate Reaction, Kyoto, Japan,1989, pp. 721-726.

24. Inoue, S.; Fuji, M.; Kobayashi, K.; and Nakano K., “StructuralBehaviors of Reinforced Beams Affected by ASR,” Proceedings of the8th International Conference on Alkali-Aggregate Reaction, Kyoto, Japan,1989, pp. 727-732.

25. Fan, S., and Hanson, J. M., “Effect of Alkali-Silica ReactionExpansion and Cracking on Structural Behavior of Reinforced ConcreteBeams,” ACI Structural Journal, V. 95, No. 5, Sept.-Oct. 1998, pp. 498-505.

26. Multon, S., and Toutlemonde, F., “Water Distribution in ConcreteBeams,” Materials and Structures, V. 37, No. 270, 2004, pp. 378-386.

27. Bach, F.; Thorsen, T.; and Nielsen, M. P., “Load Carrying Capacityof Structural Members Subjected to Alkali Silica Reaction,” Proceedingsof the 9th International Conference on Alkali-Aggregate Reaction, London,1992, pp. 9-21.

28. Jones, A. E. K., and Clark, L. A., “The Effects of Restraint on ASRExpansion of Reinforced Concrete,” Magazine of Concrete Research,V. 48, No. 174, 1996, pp. 1-13.

29. Ahmed, T.; Burley, E.; and Rigden, S., “The Static and FatigueStrength of Reinforced Concrete Beams Affected by Alkali-Silica Reaction,”ACI Structural Journal, V. 95, No. 4, July-Aug. 1998, pp. 376-388.

30. Multon S., “Evaluation expérimentale et théorique des effets del’alcali-réaction sur des structures modèles,” PhD thesis, Université deMarne la Vallée, France, 2003. (in French)

31. Larive, C.; Laplaud, A.; and Joly, M., “Behavior of AAR-AffectedConcrete,” Proceedings of the 10th International Conference on Alkali-Aggregate Reaction, Melbourne, Australia, 1996, pp. 670-677.