19_转转大师

19_转转大师 CementandConcreteResearch40(2010)477–487 ContentslistsavailableatScienceDirect CementandConcreteResearch journal homepage: Temperature,porepressureandmassvariationofconcretesubjectedtohigh temperature—Experimentalandnumericaldiscussiononspallingrisk Jean-Christophe Mindeguiaa,b,?,PierrePimientaa,AlbertNoumowéc,MulumbaKanemac a CentreScienti,queetTechniqueduBâtiment,84AvenueJeanJaurès,ChampssurMarne,77447MarnelaValléeCedex,France b LaboratoiredesSciencesAppliquéesauGénieCiviletCôtier,UniversitédePauetdesPaysdel'Adour,AlléeduParcMontaury,64600Anglet,France c LaboratoiredeMécaniqueetMatériauxduGénieCivil,UniversitédeCergy-Pontoise,5mailGayLussac,NeuvillesurOise,95031Cergy-Pontoise,France a r t i c l e i n f o a b s t r a c t Articlehistory: Spallingathightemperatureisaphenomenonthatcanbeobservedindifferentmaterialssuchasceramics,rocks Received25June2008 and bricks. For concrete, this phenomenon, considered as a thermal instability of the material, can seriously Accepted8October2009 jeopardize the integrity of a whole structure during ,re and can even constitute a risk for people. Many explanationstothespallingriskexistbutstillnomodelcanaccuratelypredictit.Amongthem,modelsbasedon Keywords: thermo-hydralbehaviourofconcretehavebeenproposedanddevelopedbyseveralauthors.Inparticular,an Temperature(A) importantroleisgiventotheporevapourpressure,consideredbymanyauthorsasthemainmechanismforthe Spalling triggerofsuchathermalinstability.However,porevapourpressureisnoteasytomeasureandnumericalworks Porevapourpressure stillneedmoreexperimentalresultstovalidatetheirassumptionsregardingthespallingrisk.Thispaperpresents Massloss theresultsofanexperimentalstudycarriedouton,vedifferentconcretemixtures.Weusedadeviceintendedfor Firebehaviour measuringtemperature,porevapourpressureandmasslossofconcretespecimens.Theaimofthestudywasto Thermomechanicalbehaviour betterunderstandthethermo-hydralbehaviourofconcreteexposedtohightemperatureandthepossiblelinkto spalling risk. In particular, we focused on the in,uence of matrix compactness on the transfer properties of concreteandwediscussedabouttheimportanceofporevapourpressureonspallingrisk.Moreover,basedonour experimental observations, a numerical analysis of the in,uence of water content on the thermomechanical behaviourofconcreteduringheatingisdone. ? 2009 Elsevier Ltd. All rights reserved. wellexplainedandwhichriskisnotpredictablebymodels.Spalling 1.Introduction riskiscommonlyexplainedbytwodifferentmechanisms: Concrete structure design must take into account the risk of temperatureincrease.Heatingcanbecausedby,res(intunnels,high ? Thermomechanical process: the heating of a concrete element rise buildings, underground parks…) or by accidental situations in involveshightemperaturegradients,particularlyinthe ,rstcenti- nuclear powerplants (e.g. LOCA, loss of coolantaccident or contact metresoftheheatedsurface.Thesegradientscanbeveryimportantin between liquid sodium and the steel liner of the concrete reactor thecaseofarapidheating(e.g.fora,re)andinducehighcompressive vessel).Concretebehaviourathightemperatureisverycomplexand stresses close to the heated surface. These stresses can locally in,uences the global behaviour of a structure during heating. overtaketheconcretestrengthandcausetheejectionofpieces[3–6]. Particularly, previous studies have shown the important risk of ? Thermo-hygralprocess:theheatingofaconcreteelementinvolves thermal instability of concrete, phenomenon commonly called masstransportintotheporousmedium.Fluidsthatarepresentinto spalling[1,2].Concretespallingconsistsinthedetachingoffragments concrete(freewater,watervapour,anddryair)aremovingdueto oftheexposedsurfaceandcanseriouslyjeopardizetheintegrityofthe pressureandmolarconcentrationgradients(DarcyandFicklaws). wholestructure(steelreinforcementdirectlyexposedto,re,increase Particularly,,uidsaremovingthroughtheinnerzonesofconcrete. of the buckling risk of compressed elements, loss of insulating Sincethesezonesarecolder,watervapourstartstocondensateanda properties…).Nowadays,concretespallingisstillaphenomenonnot “moistureclog”isgraduallycreatedclosetotheheatedsurface.This clogisassumedtobearegionofconcretewithhighwatercontent. Sincethisclogactslikearealbarrierto,uid's,ow,porepressuresare increasing [7]. These pressures can locally overtake the tensile strengthofconcreteandinitiatethespalling[8–10]. ? Corresponding author. LaSAGeC?, UFR Sciences et Techniques de la Côte Basque, AlléeduParcMontaury,64600Anglet,France. E-mailaddresses:jean-christophe.mindeguia@univ-pau.fr(J.-C.Mindeguia), Recent progresses in mathematical and numerical tools allow pierre.pimienta@cstb.fr(P.Pimienta),albert.noumowe@u-cergy.fr(A.Noumowé), inokanema@msn.com(M.Kanema). researcherstocouplethetwopreviousprocessesthankstotheuseof 0008-8846/$–seefrontmatter?2009ElsevierLtd.Allrightsreserved. doi:10.1016/j.cemconres.2009.10.011 1 478 J.-C.Mindeguiaetal./CementandConcreteResearch40(2010)477–487 Thermo-Hydro-Mechanicalmodels(THMmodelling)[11–13].Evenso, thermo-hygral behaviour of the ,ve concretes. Moreover, mass loss spallingriskisstillnoteasilypredictablesincethecriterionofinstability measurementisausefuldataforthevalidationofnumericalmodels. is unknown. On the other hand, we know from the literature that Thebalanceisplacedunderaninsulatingpanelofrockwooltoensure spallingcanbeobservedinverydifferentconditions:atverylowheating that it is not sensitive to heating. Holes are made in the rockwool rate (around 1?C/min), cylindrical concrete samples can violently paneltonotinterferewiththeweighing. explode([14],[15])whereasforhighheatingrate(likea,re),spalling Forsimultaneousporepressureandtemperaturemeasurements, canlooklikeaprogressivepeelingoftheheatedsurface[16]. sampleswereequippedduringcastingwithgauges.Thesegaugesare Among the key parameters that seem to explain spalling, the made of a sintered metal round plate (Ø 12×1mm). The plate is transport properties of concrete (permeability, porosity, water weldedtoathinmetaltube(innerdiameter1.6mm),whichcomes content, and damage) seem to be very important. In particular, we outoftheunheatedfaceofthesample.Thetubeisthenconnectedtoa havetostudyhowvapourpressurescanbuild-upintoconcretepore piezo-electricaltransducerthankstoa,exibletube,lledwithsilicon microstructure.Thisstudyfocusedparticularlyonthein,uenceofthe oil. Secondly, thermocouples (Ø 1.5mm) are introduced into the matrix compactness. We tested ,ve concrete mixtures, with a metaltubesdowntothemetalplate. constantaggregatevolumebutwithdifferentWater/Cementratios. Sampleswereequippedwithsixgauges.Oneofthemconsistsina singleplaintube inwhichone endisplacedat 2mmoftheheated surface for only temperature measurement. In order to measure 2.Experimentaldetails internaltemperatureandpressure,theother,vegaugeswereplaced at 10, 20, 30, 40 and 50mm of the heated surface (Fig. 1). The six 2.1.Concretemixes gaugesareplacedatthecentreoftheheatedsurface,ina10×10cm? square (Fig. 1). By this way, we assume that the pressure and Inordertoobtaindifferentmatrixcompactness,the,veconcrete temperature measurement are not in,uenced by the boundary mixesweredesignedbyvaryingtheW/Cratio.Thestudiedconcrete conditions of the sample. Moreover, in this measurement zone, mixes and some of their characteristics are given in Table 1. The heating,owand,uidmovementareassumedtobeunidirectional. compressive strength was measured at 28days on water stored cylindrical samples. The permeability was assessed according to the CEMBUREAU method [17], based on the Klinkenberg approach [18]. 2.3.Precisionsaboutthepressuremeasurement The initial water content wasassessed by drying cylindrical samples (Ø 150mm×50mm) of the ,ve different concretes at 80?C (during Concrete is a porous material, partially saturated by liquid water. 30days).Thiswatercontentreferstotheinitialfreewaterofthesample. Three phases ,ll the porous media: liquid water (free and adsorbed Otherhightemperaturepropertiesofthesemixeswerealreadyassessed water),vapouranddryair.Theinterfacebetweentheliquidwaterand [11,19].WenotethatthedesignedB450andtheB500concretemixes thegasphase(vapouranddryair)ischaracterizedbyasurfacetension have a compressive strength higher than 60MPa and can then be thatinducesadiscontinuitybetweenthe,uidspressures.Particularly, consideredasHighPerformanceConcretes.Moreover,wenotethatthe thedifferencebetweenliquidwaterpressureandgaspressure(de,ned B325andtheB350concretemixeshaveverysimilarpropertiesatroom asthesumofvapourpressureanddryairpressure)iscalledcapillary temperature(thesamepermeabilityandinitialfreewatercontentand pressure [20]. The capillary pressure mainly depends on the relative almostthesamecompressivestrength).WeassumethattheW/Cratio humidity of concrete pores and can reach high negative values (for oftheB325wastoohightoseeanimportantdifferencewiththeB350. instance, at 20?C and for a relative humidity of 50%, the capillary pressureequalsto?95MPa[21]).Oneimportantresultisthatdryingof concreteinvolvesadecreaseofcapillarypressure,andcanexplainthe 2.2.Setup delayedstrainsofconcrete(suchasdryingshrinkageandcreep)[21]. Ontheotherhand,weassumethattheintroductionofoneofour TheexperimentaldevicedevelopedbyKalifaetal.[9]wasused. gauges (see Section 2.2) into concrete sample modi,es its porous Thetestconsistsinapplyingathermalloadtoonefaceofaprismatic media. Indeed, we assume that a spherical zone of pressure 3)usingaradiantheaterplaced3cmaboveit. sample(30×30×12cmmeasurement(6mmofradius)isformedaroundthesinteredmetal The sample lateral faces are insulated with porous ceramic blocks roundplateof agauge(Fig.2).Weestimatethegasvolumethat is suchasthethermalloadcanbeassumedtobequasi-unidirectional. includedinthissphericalmeasuringzoneat50mm3(10%ofporosity, Thesampleisplacedonabalanceinordertomeasureitsmassloss 50%ofwatersaturationconcrete),whilethefreevolumeinonegauge duringheating.Themasslossofthesampleduringheatingismainly isequalto130mm3.Thesumofthesphericalmeasurementzonegas duetothe,uid'sescapefromconcrete(water,vapouranddryair). volumeandfreevolumeofthegaugeconstitutesthenanimportant Sincethesampleisunsealed,,uidscanescapefromallthesidesofthe freevolume.Duringheating,weassumethatonlywatervaporization sample. We want to emphasize that the mass loss measurement is is able to ,ll this important free volume, thanks to its important then only speci,c to our testing conditions (sample geometry and volume increase. Indeed, a simple temperature–volume phase heating). However, since the boundary conditions are the same for diagram of water shows that the volume increase of vaporization each test, we can use the mass loss measurement to compare the Table1 Mixtureproportions(kg),compressivestrength,permeabilityandinitialfreewatercontentofthetestedconcretemixtures. B325 B350 B400 B450 B500 CEMI52.5cement 325 350 400 450 500 Siliceous10/20gravel 960 960 960 960 960 Siliceous5/10gravel 89 89 89 89 89 Siliceous0/5sand 740 740 740 740 740 Water 202 194 177 160 143 0 0.35 1.04 1.73 2.43 Superplasticizer 0.55 0.44 0.36 0.29 W/Cratio 0.62 28dayscompressivestrength(MPa) 35 36 53 62 76 1.5?10?16 1.5?10?16 3.9?10?17 1.2?10?18 1.6?10?20 Permeability(m?) Initialfreewatercontent(%) 3.8 3.8 3.5 3.2 2.8 2 J.-C.Mindeguiaetal./CementandConcreteResearch40(2010)477–487 479 Fig.1.Schemeoftheexperimentalset-up(left).Right:positionofthetemperatureandporevapourpressuremeasuringzone(viewoftheheatedsurfaceofthesample). processisquitemoreimportantthanthethermalexpansionofliquid reaches600?C(afteraround5min).Thepowerwasthenmaintained water[22].Ifweneglectthethermalexpansionofdryair,we,nally constantduring6h.Atlast,sampleswerenaturallycooleddown.The assume that our measurement system is theoretically only able to choice of the heating procedure was a compromise: fast enough to measureporevapourpressure.However,wewilldiscusslaterinthis involveporevapourpressurebutslowenoughtonotinducetoomuch paperaboutthecontributionofdryairthermalexpansion. thermal damage to the concrete sample. We present on Fig. 3 the Atlast,sincethemeasuringtubesareplacedinthecolderzonesofthe temperaturemeasuredat2mmoftheheatedsurfaceaswellasthe sample(Fig.1)andsincethe,exibletubesthatare,lledwithsiliconoil heating rate at this depth for one of the concrete mixtures areplacedoutsideofthesample(Fig.1),weassumethatwedonotneed (representativeresultforallthetests).Wealsopresentinthis,gure anycalibrationofthesystemregardingtothetemperatureincrease. thetemperatureat2mminaconcretesampleduringa,retest(ISO 834curve).Wecanseethattheheatingscenariousedinourstudyis relativelyslowincomparisontoamorerealistic,recurve(liketheISO 2.4.Heatingprocedure curve). Twotestswerecarriedoutforeachconcretemixture.Theradiant heater was controlled in such a way that its temperature rapidly 3.Resultsandanalysis 3.1.Experimentalobservations During the test, we observed water drops and vapour ,ow escapingfromthesamples.Moreover,severedamageofthesamples was observed. These two observations had not been reported in previoustestscarriedoutonothertypesofconcrete[9].Wecanseein Fig.4thecrackingandtheaggregatesurfacespallingofasampleafter a test. This importantdamage (i.e. cracking and aggregate spalling) canbeexplainedbytheinstablebehaviourathightemperatureofthe ,intaggregatesusedintheconcretemixtures.Complementaryhigh temperature tests were carried out on ,int aggregates (without cementpaste).Theycon,rmedthefactthatthe,intaggregatesused intheconcretemixturesarethermallyunstablefrom120?Cto200?C. Thisbehaviourisexplainedbythehightemperaturecleavageof,int, whichcanbeduetohighvapourpressuresthatbuildupintotheown laminarmicrostructureoftheaggregate[23]. 3.2.Massloss We present on Table 2 the total water content (free+bonded water)beforeheatingforthedifferentmixturesinpercentage.This valueistheratiobetweenthemassofwaterintroducedduringthe casting and the total mass of the components (cement, aggregates, Fig.2.Illustrationofthesphericalmeasuringzonethatformsaroundapressuregauge. waterand superplasticizer). Table2 also presentsthe proportionof Estimationofthefreevolumeofthemeasuringsystem. 3 480 J.-C.Mindeguiaetal./CementandConcreteResearch40(2010)477–487 Fig.3.Measuredtemperatureandcorrespondingheatingrateat2mmoftheheatedsurface(left).Comparisonwiththetemperaturemeasuredat2mmduringthe,rst60minofan ISOcurve,retest(right). Fig.4.Aggregate(,int)spallingontheheatedsurface(left).Crackingandmoisturemarksonthelateralface(right). lostwaterafterheating(ratiobetweenthemasslossandtheinitial vapour pressure. We can see that the B325 has a weak ratio of massofthesample).Thankstothesetwovalues,wecanassessthe extractedwaterincomparisonwiththeB500.Thiscanbeexplainedby ratioofwaterthathasbeenextractedfromthe,veconcretesduring thefactthatduetoitshighpermeability,theB325concretelosesmore heating.Wecanobservethatthe,veconcreteslosearoundtwothirds waterduringthestoringperiodthanthemorecompactconcretes. oftheirinitialwatercontent.Itmeansthatafterheating,onethirdof Fig.5presentsthemasslossrateforthe,veconcretemixtures.The thewaterthathadbeenintroducedduringcastingisstillpresentinto curves correspond to the average value of two tests for each type of concrete.Indeed,duetotherelativelowvalueoftemperatureintothe concrete.Veryfewdispersionhasbeenobservedbetweenthetwotests samplesduringthetests(lessthan400?Contheexposedsideandless (lessthan5%ofdispersionaroundtheaveragevalueofmasslossrate). than 160?C on the unexposed side), concrete is not totally dried. We note that the rate of water escape increases with the W/C ratio. Particularly,thechemicallylinkedwatercannotbetotallyremovedas Indeed,thelowcompactnessoftheconcretewithhighW/Cratioinduces theentiredehydrationofCSHand Portlanditeisreached forhigher ahigher,uidpermeabilityandthenmakeseasierthewaterescape. temperature.Butsubtractingtheinitialfreewatercontenttothelost water,wecanassessthequantityofchemicallylinkedwaterthatis releasedduringtheheating.ThevaluesarepresentedintheTable2. We observe that the quantity of chemically linked water that is releasedduringheatingisweakregardingtothefreewatercontentof the sample (see Table 1). According to [26], this weak quantity of chemicallylinkedwaterdoesnotallowtoamplifythebuiltupofpore Table2 Totalwatercontent,lostwater,relativelostwater(lostwater/totalwater)andreleased chemicallylinkedwaterforthetestedconcretemixtures. B325 B350 B400 B450 B500 Totalwatercontent(%) 8.72 8.32 7.50 6.67 5.88 Lostwater(%) 5.17 5.43 4.67 4.37 3.89 Relativelostwater(%) 59 65 62 66 66 Releasedchemicallylinkedwater(%) 1.37 1.63 1.17 1.17 1.09 Fig.5.Masslossrateforthetestedconcretemixtures. 4 J.-C.Mindeguiaetal./CementandConcreteResearch40(2010)477–487 481 3.3.Temperature needed to extract vapour molecules from liquid water. In the case of concrete,thesmallertheporeradius(i.e.thedensertheconcrete),the higher the vaporization temperature. The fact that vaporization takes Fig.6presentsthetemperaturesmeasuredintothedifferentconcrete placeataround100?Cinthelessdenseconcrete(B325)indicatesthat samples. For several tests, some gauges were unintentionally clogged capillaryforcesintheseporesarenotimportant,i.e.poreradiusishigh with cement paste during casting. As a consequence, temperature (probablyduetocracking). measurementsforsomedepthsdonotappearinthegraphs. Thermal,owisquitesimilarbetweenthe,veconcretemixtures. Forallconcretemixturesandforeachmeasuringpoint,weobservea Indeed, we observe that the curves of temperature evolution are slight plateau, i.e. a perturbation, in the increase of temperature. This identical from one to another concrete (Fig. 6). For example, the plateauisduetothewaterphasechange(vaporization).Thistransfor- temperatureat10mmattheendofheatingreachesaround350?Cfor mationisendothermicandthenconsumesagreatpartoftheenergythat the,veconcretes,andthetemperatureat50mmreachesaround250?C isbroughtbyheating.Asaconsequence,theheattransferintoconcrete forthe,veconcretes.Thisresultshowsthatconcretethermalproperties sample isslowed down.We want to emphasize thatby thisway, the are not signi,cantly in,uenced by the compactness of the material. watervaporizationcaninduceadditionaltemperaturegradients.Wewill According to the literature, thermal diffusivity of concrete is mainly seeinthelastparagraph,thatthesegradientscouldmodifythestressesin controlledbytheaggregate'snature,becauseoftheirimportantvolume aconcretestructureduring,re.Itwasalsoobservedthatthetemperature inconcrete[24].Sincetheaggregatesareofthesamenatureandthat plateaudependsontheconcretecompactness.Indeed,ittakesplaceat theircontentisverycloseinthe,veconcretes,itcertainlyexplainswhy around 100?C for the B325 and at around 175?C for the B500. This thetemperaturecurvesareveryclosefromonetoanotherconcrete. dependencecanbeexplainedbythecapillaryforcesthatexistinaporeat theinterfacebetweenliquidwater,gasphase(vapouranddryair)and solid(concretematrix).Thesecapillaryforcesin,uencethevaporization 3.4.Porepressure process:underapressureof0.1MPa,andforanin,nite,atsurfaceof water,vaporizationtakesplaceat100?C.Attheopposite,forverysmall Fig.7presentstheporepressuremeasuredintothesamplesforall pores, capillary forces reaches high values and higher temperature is thetestedconcretemixtures.Whencomparingtwosimilartests,we Fig.6.Temperatureasafunctionoftimeforallthetestedconcretemixtures(thenumberafterTisthedistanceinmmfromtheheatedsurface). 5 482 J.-C.Mindeguiaetal./CementandConcreteResearch40(2010)477–487 Fig.7.Pressureasafunctionoftimeforallthetestedconcretemixtures(thenumberafterPisthedistanceinmmfromtheheatedsurface). 6 J.-C.Mindeguiaetal./CementandConcreteResearch40(2010)477–487 483 can see that the pressure measurement is scattered. It is a local measurement(seeSection2.2)andthepressurevaluedependsonthe tube location (into cement paste, close to an aggregate or in an air void).However,wecanclearlyseeinFig.8thatthehighertheW/C ratio, the lowerthe pore pressure (in this ,gure, we have assumed that the more representative value of pressure for each type of concreteisthehighestmeasuredvalue).Thisunderlinestheroleof the concrete transport properties (porosity, permeability) on the thermo-hygral behaviour of the material, and in particular on the build-upofporepressures[13]. WepresentinFig.9theevolutionoftemperatureandpressureasa function of time that have been measured at 40mm of the exposed surface for one of the ,ve concretes (the same behaviour has been observed for all the concretes). We can see that the peak of the measured pore pressure is reached during the temperature range of vaporization (slight plateau on the temperature curve). This result Fig.9.Evolutionwithtimeofthetemperatureandporepressuremeasuredat40mmof indicatesthatwatervaporizationintoconcreteiswellassumedtobe theexposedsurfaceforoneofthe,veconcretes.Itmustbenoticedthatthescalesfor responsibleforthebuild-upoftheporepressurethatismeasuredbyour porepressurearedifferentinthegraph. experimental device. The diffusion of heat into concrete, and the complex hygral behaviour which is induced, may create a saturated On the other hand, since we assume that during heating, water is zone close to the exposed surface (the so-called moisture clog). moving totheinnerpartofthesample(transportthatmay create a Accordingtonumericalresults[11,13],thisclogisgenerallycreatedin moistureclog),itcanexplainwhytheoverpressureismoreimportantin a zone from around 1cm to 6cm from the exposed surface and its thedeepestzonesfromtheheatedsurfaceofthesample(seeFig.10).In position depends on the transport properties of the concrete. As the Fig.10,wealsoobservethatP10thatwasmeasuredinB450showsa vapour ,ow cannot go through the clog, it only goes towards the differentbehaviourthantheothersensors.Itisattributedtotheoriginal exposed surface (i.e. in the opposite direction of the heat ,ow). evolutionofthetemperaturethatwasmeasuredat10mmintheB450 Moreover,theconcretelocatedbetweentheexposedsurfaceandthe (seeFig.6).Indeed,weobservearound175?C(i.e.forthetemperature moistureclogcontinuestodryanddehydrate.Assoonastherateof correspondingtotheP10highestvalue),adecreaseofthetemperature vapourescapingapore(duetothe,owtowardstheexposedsurface)is (around 20?C of cooling). This phenomenon can be due to a very higher than the rate of vapour ,lling the pore (coming from drying, important consummation of energy by the phase transformation of dehydration and transport), the pressure starts decreasing. Simulta- water.Thisconsummationisassumedtoberesponsibleforthecooling neously,bycreatingmoreavailablevolumeinthematrix,andespecially ofaconcretezoneclosetothesensor. byincreasingthepermeabilityofthematerialtovapourescape[27], Furthermore, it was observed that the measured pore pressures crackingcanalsocontributetotheporepressuredecrease[3].Thesetwo vary approximately from 0.2MPa (B325) to 1MPa (B500). These phenomenacanexplainthebellshapeofthepressurecurves. pressures can be considered as low values compared to those of In Fig. 10, the evolution of the pore pressure as a function of previous studies carried out on concrete with similar compactness temperature is presented. Experimental results of pressure (P) are [26].Thismaybeexplainedbytheimportantdamage(i.e.cracking)of comparedwiththesaturatingvapourpressure curve (Pvs).Formany thesamplesduetotheunstablebehaviourofthe,intaggregatesused in the concrete mixtures (see Section 3.1). Indeed, there is an cases,itcanbeobservedthatmeasuredpressuresfollowthePvs curve importantlinkbetweenthepermeabilityandthedamageofconcrete duringtheascendingbranch.Thisresultseemstocon,rmthatthepore [27]. In particular, high damage strongly modi,es the transport pressure that is measured by our experimental device is vapour propertiesof concrete,making easier by this way the movement of pressure.However,forsomeresults,themeasuredpressuresarehigher ,uids.ThiscanbeclearlyseeninFig.11wherewecomparethemass thanthePvs.Sinceitistheoreticallyimpossible,vapourpressureisnot lossrateoftheB350andB400concreteswiththemasslossrateof responsibleforthisoverpressure.Thisoverpressureisoftenattributed similarconcretesmadewiththermallystableaggregates(calcareous tothepartialpressureofthedryairthatisenclosedintheporousmedia aggregates)[26].Indeed,themaximalmasslossrateishigherforthe [9,25].Thepartialpressureofairinaporeisstronglydependentonits cracked samples and the peak of mass loss (time for reaching the liquidwatersaturation:thehigherthewatersaturation,theloweristhe maximal rate) appears sooner. One of the consequences of this free volume available to the air to expand during heating. As a crackinginducedbyaggregateinstabilityisthefactthatitdoesnot consequence,forporeswithhighliquidwatercontent,thecontribution allow the build-up of important pressure. As a ,rst conclusion, ofthermalexpansionofdryairtothetotalpressurewillbeimportant. crackingmustbetakenintoaccountinordertocorrectlydealwiththe thermo-hygral behaviour of concrete at high temperature. Particu- larly,incaseofrapidheating(e.g.fora,re),concretecanbestrongly damaged and pore vapour pressure should then be signi,cantly reduced[26]. Inthisstudy,wedidnotmeasurehighporevapourpressureinthe ,veconcretemixturesduetothecrackinginducedbytheaggregate's thermalinstability.Ontheotherhand,previousheatingtestsshowed thattheriskofspallingexistsforthese,vemixtures,andcanevenbe veryimportantforthedensestones[11].Fromthisobservationand fromtheresultspresentedinthispaper,itcanbededucedthatpore pressures are not the only origin of spalling. Recent experimental studiesfromtheauthors[26]andfromaSwedishlaboratory[28,32] seem to con,rm this assumption. However, despite of the fact that spallingriskcannotbeexplainedonlybyporepressures,itmustbe emphasised that the moisture content and the overall moisture Fig.8.MaximumpressureasafunctionofW/Cratioforalltests. 7 484 J.-C.Mindeguiaetal./CementandConcreteResearch40(2010)477–487 Fig.10.Pressureasafunctionoftemperatureforthetestedconcretemixtures,plottedtogetherwiththesaturatingvapourpressurecurve(Pvs ). movementcanplayanimportantroleforspalling.Inparticular,ithas 4.Numericaldiscussion been observed in this study that the moisture content modi,es the heat transfer into concrete and then induces additional thermal Following the experimental observations (see Section 3.3), the gradients. By using numerical simulations, we will see in the next in,uence of the water vaporization on the behaviour of concrete paragraphhowthesethermalgradientscanin,uencethemechanical duringheatinghastobeanalysed.Particularly,wewanttoseehow behaviourofconcreteduringheating. thethermomechanicalbehaviour(elasticbehaviour)ofconcretecan Fig.11.ComparisonofthemasslossratebetweentheB350andtheB400mixturesandsimilarconcretesmadewiththermallystableaggregates(B40andB55,see[21]).Thearrow underlinesthetime-lagofmaximalmasslossratebetweenthetwotypesofconcrete. 8 J.-C.Mindeguiaetal./CementandConcreteResearch40(2010)477–487 485 Table3 assessed[29].Thewatercontentofthewall,andthusitspotentialfor Boundary conditions for the thermomechanical simulation of a concrete wall during watervaporization,isimplicitlytakenintoaccountbyvaryingthepeak ISO-,re. of speci,c heat between 100?C and 250?C. This approach takes inspirationfromtheEurocode2,Part1.2[30].Wesimulatedtheelastic thermomechanicalbehaviouroftheconcretewallwiththreedifferent levels of water content (i.e. three different levels of vaporization potential):constantspeci,cheat(novaporizationpotential),apeakof 5000Jkg?1K?1 (middle vaporization potential) and a peak of 10,000Jkg?1K?1 (high vaporization potential). The three different evolutionsofspeci,cheataregiveninFig.12. Fig.13presentstheresultsofthethermalsimulations.Inthis,gure, wenotethatthehigherthewatercontent(i.e.thehigherthepeakof speci,cheat),themoreimportanttheslowingdownoftheheat,ow. Side Variables Valuesandcoef,cients This result is in good accordance with the experimental results ?,? Displacementuy y=0 upresentedinSection3.3.Aconsequenceofthethermalslowingdown Thermal,owqT qT=0 istheincreaseofthethermalgradientsclosetotheheatedsurface(i.e.in TemperatureT Convective–radiativewithhc=25(W/m2K),ε=0.7 ? T(t)accordingtothestandardISO-,re834 thezoneofconcretespalling).Thisisclearlyshowninthesecondplotof ? ux=0 x DisplacementutheFig.13whereweobservethatthedifferencebetweenthesurface Convective–radiativewithhc=4(W/m2K),ε=0.7 TemperatureT temperatureandthetemperatureat10mmreachesthehighestvalues T=293K forthehigherwatercontentsimulation.Inparticular,thedifferenceof temperatureisparticularlyhighduringtheperiodwhenonecanassume thatconcretespallingoccursduringanISO-,re. Table4 Fig.14presentstheresultsofthethermomechanicalsimulations. Concreteparametersusedinthethermomechanicalsimulation. Foreachcaseofspeci,cheat,weobservethatacompressiveclogis Parameters Values progressively created close to the heated surface. The formation of thisclogisduetotheprogressivedegradationoftheconcretecloseto Mechanical Modulusof 52GPaatroomtemperature,seeFig.12forthe elasticity evolutionwithtemperature theheatedsurface(duetothedecreaseofthemodulusofelasticity Poissoncoef,cient 0.25 withheating)andmaybeapossiblecauseforconcretefractureclose 1.510?5K?1 Thermal totheheatedsurface,i.e.apossiblecauseforconcretespalling.The expansion in,uenceofthewatercontent,andparticularlythein,uenceofthe coef,cient vaporization process, is analysed on the Fig. 15. In this ,gure, we Thermal ?1.K?1 1W.mThermal conductivity observethatthehigherthevaporizationpotential,theclosertothe Bulkdensity 2300kg.m?3 heatedsurfacethecompressiveclog.Thisresultindicatesthatforhigh Speci, cheat SeeFig.12fortheevolutionwithtemperature water content concretes, the ,re-induced compressive zone is thinner. Inspired by [31], if we assume that spalling can be caused by the buckling of a concrete band close to the heated surface, our simulationsshowthattheconcretebucklingrisk(i.e.spallingrisk)is higherforhighwatercontentconcrete.Thisassumptioniscurrently bemodi,edbecauseofthephasetransformationofwater.Thiswill observedduringexperiments[1,2,26]anditallowsustoemphasize allow us to give new tracks of investigation to assess the possible thepossiblelinkbetweenconcretespallingandthethermomechani- causesofthermalspallingofconcrete. calprocess.In otherwords,oursimulationshowsthat notonly the In this chapter, thermomechanical behaviour of a concrete wall watercontentofconcreteplayaroleinthethermo-hygralprocessbut exposedtoISO-,rewillbeanalysed.A,nite-elementsoftwarehasbeen alsointhethermomechanicalone. used(Cast3m).Theboundaryconditionsareschematicallypresentedin Table 3 and concrete properties are given in Table 4. Among the parameters, the assumed evolution of the modulus of elasticity with 5.Conclusions temperature is given in Fig. 12. Modulus of elasticity is assumed to decrease with heating in order to implicitly take into account the Samplesof,veconcretemixturesweretestedinordertostudythe thermal damage of concrete. This behaviour is often experimentally in,uence of the matrix compactness on the concrete behaviour at Fig.12.Assumedevolutionofconcreteparameterswithtemperature:modulusofelasticity(left)andspeci,cheat(right).Watervaporizationistakenintoaccountbyvaryingthe peakofspeci,cheatbetween100?Cand250?C. 9 486 J.-C.Mindeguiaetal./CementandConcreteResearch40(2010)477–487 Fig.13.Evolutionofthetemperatureat10mmfromtheexposedsurfacedependingonthepeakofspeci,cheat(left).Evolutionwithtimeofthedifferencebetweenthetemperature atthesurfaceandthetemperatureat10mmdependingonthepeakofspeci,cheat(right).Concretespallingisassumedtooccurbetweenthe10thandthe30thminofISO-,re. Fig. 14. Pro,le of the axial stress (σ ) at different times of heating: a—no water xx vaporization(nopeak),b—watervaporization(peakof5000J.kg?1.K?1)andc—water xx)at3differenttimesofheatingdependingon Fig.15.Comparisonoftheaxialstress(σvaporization(peakof10,000J.kg?1.K?1). thepeakofspeci, cheat. 10 J.-C.Mindeguiaetal./CementandConcreteResearch40(2010)477–487 487 [3] Z.P.Ba?ant,Analysisofporepressure,thermalstressandfractureinrapidlyheated high temperature. The study focused on the temperature and pore concrete,ProceedingsoftheInternationalworkshopon,reperformanceofhigh- vapour pressure ,elds that develop into concrete during heating as strengthconcrete,NIST,Gaithesburg,USA,1997. well as on the samples mass loss. The tests presented have been [4] F.J.Ulm,P.Acker,M.Lévy,TheChunnel,re.II:analysisofconcretedamage,Journal ofEngineeringMechanics125(1999)283–289. carriedoutataheatingratelowerthantheoneprovidedbytheISO [5] Y.Msaad,G.Bonnet,Analysisofheatedconcretespallingduetorestrainedthermal curve (see Fig. 3). The results allow discussing about the physical dilation: application to the Chunnel ,re, Journal of Engineering Mechanics 132 originsofspalling.Thefollowingconclusionscanbedrawn. (2006)1124–1132. The thermal instability of ,int aggregates involves an important [6] J.Sercombe,C.Gallé,S.Durand,P.Bouniol,Ontheimportanceofthermalgradients in the spalling of high-strength concrete, Proceedings of EM 2000: Fourteenth damageoftheconcretesamples.Somelocalspallingontheexposed EngineeringMechanicsconference,AustinTexas,USA,2000. surfacewasobserved. [7] T.Z. Harmathy, Moisture in materials in relation to ,re tests, Special Technical Thepermeability(ormatrixcompactness)oftheconcretestrongly Publication,no385,ASTM,1964. in,uencesthewaterescapeduringheating.Loweristhepermeability [8] Y.Anderberg,SpallingphenomenaofHPCandOC,ProceedingsoftheInternational workshopon,reperformanceofhigh-strengthconcrete,NIST,Gaithesburg,USA, andsloweristhemasstransferintoconcrete.Thisisingoodagreement 1997. withtheoreticalconsiderationssuchastheDarcyandFicklaws. [9] P.Kalifa,F.D.Menneteau,D.Quenard,SpallingandporepressureinHPCathigh Porepressureisbuilt-upintoconcreteduetowatervaporization. temperature,CementandConcreteResearch30(2000)1915–1927. Themaximalpressurestronglydependsontheconcretecompactness. [10] S. Dal Pont, H. Colina, A. Dupas, A. Ehrlacher, An experimental relationship betweencompleteliquidsaturationandviolentdamageinconcretesubmittedto Inparticular,lowpermeabilityinvolveshighbuild-upofporepressure. hightemperature,MagazineofConcreteResearch57(no8)(2005)455–461. The coupled heat and mass transfers into concrete involve the [11] M.Kanema,M.V.G.DeMorais,A.Noumowe,J.L.Gallias,R.Cabrillac,Experimental developmentofasaturationfront(alsocalledmoistureclog).Inzones andnumericalstudiesofthermo-hydroustransfersinconcreteexposedtohigh ofquasiwater-,lledpores,thethermalexpansionofdryairinduces temperature,HeatandMassTransfer44(N?2)(2007). [12] R.T. Tenchev, J.A. Purkiss, L.Y. Li, Numerical analysis of thermal spalling in a someoverpressure. concrete column, Proceedings of the 9th National congress on theoretical and appliedmechanics,Varna,Bulgaria,2001. Incomparisontopreviousstudies,ithasbeenobservedthatthe [13] D.Gawin,F.Pesavento,B.A.Schre,er,Towardspredictionofthethermalspalling concrete cracking (caused by the thermal instability of aggregates) riskthroughamultiphaseporousmediamodelofconcrete,ComputerMethodsin AppliedMechanicsandEngineering(n?195)(2006)5707–5729. stronglylimitsthebuild-upofpressure.However,anotherstudyhas [14] L.T. Phan, J.R. Lawson, F.L. Davis, Effects of elevated temperature exposure on shown that the ,ve tested concretes present an important risk of heating characteristics, spalling, and residual properties of high performance spalling.It appearsthen thatthepore vapourpressuresarenot the concrete,MaterialsandStructures34(2001). only cause to the concrete spalling risk. Another explanation for [15] A. Noumowe, H. Carré, A. Daoud, H. Toutanji, High-strength self-compacting concreteexposedto,retest,AmericanSocietyofCivilEngineeringPublications, spalling risk can come from thermomechanical assumptions. In ASCEMaterialsJournal18(N?6)(2006). particular we observed in this study that the water vaporization [16] P.Kalifa,D.Pardon,F.D.Menneteau,C.Gallé,G.Chené,P.Pimienta,Comportement modi,estheheattransferandthetemperaturepro,leintoconcrete. à haute température des bétons à hautes performances: de l'éclatement à la Inparticular,itcaninduceadecreaseofinternalheatingrateinvolving microstructure,CahiersduCSTB,no3154,1999,(inFrench). [17] J.J. Kollek, Mesure de la perméabilité du béton à l'oxygène par la méthode additionalthermalgradients.Thisphenomenonhasbeenemphasised CEMBUREAU,Ciments,Plâtres,Chaux,vol.778,pp169–173,1989. by a numerical simulation. The simulation showed that the water [18] L.J.Klinkenberg,Thepermeabilityofporousmediatoliquidsandgases,Drilling content ofconcrete can modifythethermomechanical behaviour of andProductionPractice,1941,pp.200–231. [19] M. Kanema, A. Noumowé, J.L. Gallias, R. Cabrillac, Propriétés mécaniques et concreteduringheating.Inparticular,thebucklingriskofaconcrete perméabilité résiduelles de bétons exposés à une température élevée, Revue band close to the heated surface can be increased in high water EuropéennedeGénieCivil10(no10)(2006)(inFrench). contentconcretes.Accordingtopreviousstudies,sincethisbuckling [20] B.A. Schre,er, G.A. Khoury, D. Gawin, C.E. Majorana, Thermo-hydro-mechanical risk may be relied to spalling risk, our simulations showed the modelling of high performance concrete at high temperatures, Engineering Computations19(7–8)(2002)787–819. possible link between concrete spalling and the thermomechanical [21] F. Benboudjema, F. Meftah, J.M. Torrenti, A viscoelastic approach for the process. assessment of the drying shrinkage behaviour of cementitious materials, MaterialsandStructures40(2)(2007)163–174. More experimental tests will be carried out to con,rm and to [22] CRC Handbook of Chemistry and Physics, 90th Edition, David Lide, National complete the different assumptions about the physical origins of InstituteofStandards&Technology,2009. spalling risk. Other ,re tests (with ISO and Hydrocarbon Modi,ed [23] C.Meyer-Ottens,Thequestionofspallingofconcretestructuralelementsunder ,reloading,PhDThesis,TechnicalUniversityofBraunschweig,Germany,1972. Curve) with pore vapour pressure measurement have also been [24] Z.P.Bazant,M.F.Kaplan,ConcreteatHighTemperatures:MaterialPropertiesand carriedoutonothertypesofconcretes[32].The,rstresultsseemto MathematicalModels,Longman,Harlow,1996516pp. con,rmthatporevapourpressurebuild-upisnottheonlycausefor [25] O. Kontani, S.P. Shah, Pore pressure in sealed concrete at sustained high temperatures, Proceedings of the International Conference On Concrete Under spalling. Severeconditions,CONSEC'95Sapporo(Japan)2(1995)1151–1162. [26] J.C. Mindeguia, P. Pimienta, C. La Borderie, H. Carré, Experimental study of ,re behaviourofdifferentconcretes—thermo-hygralandspallinganalysis,Proceedings Acknowledgements oftheFibworkshopFiredesignofconcretestructures,Coimbra(Portugal),2007. [27] M. Choinska, A. Khelidj, G. Chatzigeorgiou, G. Pijaudier-Cabot, Effects and Moreoftheresultsobtainedinthispaperwereobtainedwiththe interactions of temperature and stress-level related damage on permeability of concrete,CementandConcreteResearch37(n1)(January2007)79–88. device developed by Dr Pierre Kalifa and Mr François-Dominique [28] R.Jansson,L.Boström,Thein,uenceofpressureintheporesystemon,respalling MenneteauofCSTB.Theauthorswouldliketoacknowledgethemfor ofconcrete,Proceedingsofthe,fthinternationalconference“StructuresinFire”, theirhelp.SpecialsthanksarealsogiventoRobertJanssonfromSP, Singapore,May28–302008,pp.418–429. Sweden,forthequalityofourdiscussionsaboutspallingmechanisms. [29] I.Hager,P.Pimienta,MechnicalpropertiesofHPCathightemperature.Fibtask group4.3“Firedesignofconcretestructures”.Milan,Italy,2–4December2004. [30] EN1992-1-2,Eurocode2: Designof concrete structures, Part 1.2 Structural ,re References design,December2004. [31] D.Gawin,F.Pesavento,B.A.Schre,er,Towardspredictionofthethermalspalling [1] V.K.R.Kodur,Spallinginhighstrengthconcreteexposedto,re—concerns,causes, riskthroughamulti-phaseporousmediamodelofconcrete,ComputerMethodsin critical parameters and cures, Proceedings of Structures Congress, Advanced AppliedMechanicsandEngineering195(41–43)(2006)5707–5729. TechnologyinStructuralEngineering,Philadelphia,USA,May8–102000. [32] R.Jansson,L.Bostrom,Thein,uenceofpressureintheporesystemon,respalling [2] L.Bostrom,R.Jansson,Spallingofselfcompactingconcrete,ProceedingsofSIF'06 ofconcrete,FireTechnology,SpecialissueontheSIF'08conference,March2009. workshop,StructuresinFire,Aveiro,Portugal,May2006. 11