SoilDynamicsandEarthquakeEngineering64(2014)50–62 ContentslistsavailableatScienceDirect Soil Dynamics and Earthquake Engineering journal homepage: www.elsevier.com/locate/soildyn Dynamic analysis of earth dam damaged by the 2011 Off the Pacific Coast of Tohoku Earthquake Bhuddarak Charatpangoona,n, Junji Kiyonob, Aiko Furukawab, Chayanon Hansapinyoc aDepartmentofUrbanManagement,KyotoUniversity,Kyoto615-8530,Japan bGraduateSchoolofGlobalEnvironmentalStudies,KyotoUniversity,Kyoto615-8530,Japan cDepartmentofCivilEngineering,ChiangMaiUniversity,ChiangMai50200,Thailand a r t i c l e i n f o a b s t r a c t Articlehistory: ThisstudyexaminespossiblefailuremechanismsanddynamicbehavioroftheFujinumadam,which Received13August2013 failedfollowingthe2011OffthePacificCoastofTohokuEarthquake.Thestudycomprisestwoparts:an Receivedinrevisedform investigationthroughfieldandlaboratoryexperiments,andanumericalsimulationofthedam.Field 12May2014 measurements and laboratory experiments were conducted to acquire necessary information. Micro- Accepted14May2014 tremorobservationsofthesurvivingportionofthefaileddamwereperformedtoextractdatafromits dynamic characteristics. Triaxial and other laboratory tests provided information required for the Keywords: analysis.Fortheseismicanalysis,acoupledsolid–fluidfiniteelementmethodwasappliedandobserved Earthdam andsimulatedmotionsofthe2011TohokuEarthquakeusedasinput.Mechanicalbehaviorofthedam Earthquake materialwasdescribedusingtheMohr–Coulombfailurecriterion.Frequencyanddynamicanalyseswere Finiteelement performed and dam behavior and possible failure phenomena presented. Furthermore, a comparison Dynamic betweenthesimulationresultsandexistingfactsisdiscussed. Fujinumadam &2014ElsevierLtd.Allrightsreserved. 1. Introduction andblockedroadswithdebris.Sevenbodieswerediscoveredafter searchesbeganatdawnandonepersonwasdeclaredmissing. Damage and loss of life caused by earthquakes are immense. Generally,earthdamsarethemostcommontypeofdambecause This is magnified when accompanied by the collapse of essential oftheircostandtheconvenientsupplyofrawmaterials.Anumberof infrastructure,suchasadamorapowerplant,whichhasthepotential earthdamsexistthroughoutJapanforirrigation,floodmitigation,and ofdestroyingentirecities.Thewaterandpowersuppliedbydamsare hydroelectricpowergenerationpurposes.AsJapanliesinoneofthe essential for the survival of a community, not to mention the other world'smostseismicallyactiveareas,themainissueindammanage- benefitstheybring,suchastourismandfloodcontrol.However,when ment and construction is seismic safety. Therefore, to assure dam adamfails,thedestructionisoftendeadly,causingirreparabledamage safety, proper evaluation of such dams is crucial. Accordingly, the totheland,thepeople,andtotheeconomy. failureoftheFujinumadamcanberegardedasafruitfulresourcefrom Following the devastating earthquake in Japan on 11 March which to gain insight into dam failure mechanisms and to acquire 2011,sevendamsweredamagedandtheFujinumadamcollapsed usefulinformationrelatingtodamsafety. owing to the force generated by the 9.0-magnitude earthquake, In the computation of soil structures, the coupling behavior known as the 2011 Off the Pacific Coast of Tohoku Earthquake betweenthesolidmatrixandporewateriscrucial.Biot[1,2]first (hereafter, the 2011 Tohoku Earthquake). The Fujinuma damwas developed a linear theory of poroelasticity and subsequently, constructedtoserveasawatersupplyforirrigationpurposes.The many researchers in this field have established numerous solid– dam was located on a tributary of the Abukuma River, near fluid coupling formulations based on different assumptions. Sukagawa in Fukushima prefecture, Japan. It failed following the Each technique has its own strengths in solving different pro- 2011 Tohoku Earthquake (Fig.1). The failure caused a flood that blems.Amongthem,owingtoitssimplicity,theu–pformulation washed away 19 houses and damaged others, disabled a bridge, has become well known and applied widely in the field of geotechnics.Finiteelement(FE)codesforporousmediahavebeen developed continuously [3–7]. Recently, as advanced computa- nCorrespondingauthor.Tel.:þ81753833252;fax:þ81753833253. tional techniques in geotechnics have been introduced, it has E-mailaddresses:[email protected], become beneficial for researchers [8–15] to analyze seismic [email protected](B.Charatpangoon), behaviorofdamsbyusingthosetechniques.Consequently,much [email protected](J.Kiyono), researchworkhasbeenconductedintheareaofseismicsafetyon [email protected](A.Furukawa), [email protected](C.Hansapinyo). existingearthdams. http://dx.doi.org/10.1016/j.soildyn.2014.05.002 0267-7261/&2014ElsevierLtd.Allrightsreserved. B.Charatpangoonetal./SoilDynamicsandEarthquakeEngineering64(2014)50–62 51 Fig.1. (a)LocationoftheFujinumadam(Mapdata©2014AutoNavi.Google.SKplanet.ZENRIN).(b)Sukagawaandthedamsite(Mapdata©2014AutoNavi.Google. SKplanet.ZENRIN).(c)Fujinumadam. Soralump and Tansupo [16] conducted a dynamic response In Japan, the Aratozawa dam is a rock-fill impervious-core dam analysis of the Srinagarind dam by using 213 records of 35 withthe heightof74.4m,located in Kurihara,Miyagiprefecture. earthquake events and the equivalent linear method for the The Iwate–Miyagi Nairikuearthquake in 2008 caused hugeland- nonlinearbehaviorofdammaterials.Similarly,FallahandWieland slidestooccurontheleftbankofthereservoirfromthedam.This [17] conducted an evaluation of earthquake safety of the Koman causedsettlementofthecorezonebyabout20cm.Therewasno concrete-faced rock-fill dam in Albania, by using a two- evidenceofseveredamagetothedamstructure,butitwastaken dimensionalFEmodelofthemaximumcrosssection.Theirstudy outofoperationbecauseofsafetyconcerns[21]. wasundertakenusingtheequivalentlinearmethod.Thedamwas Following the Fujinuma dam failure, Harder et al. [22] visited checkedforthesafetyevaluationearthquakewithapeakground thesitejustafewweeksaftertheevent.Theprincipalpurposeof acceleration of the horizontal component of 0.45g. Spectrum- thevisitwastoobserveanddocumentthefailureofthemaindam compatible artificially generated accelerograms were used based and the upstream slide in the auxiliary dam. Discussions on the on a site-specific seismic hazard analysis. All of these methods failure mechanism and recommendations on future works were mentionedabovehavebeenappliedtoevaluatethesafetyofearth presented by the authors. In addition, Ono et al. [23] stated that dams.Theimportanceofsuchresearchisnotonlytoexaminethe the failure of the Fujinuma dam was only the second complete behaviorofadamorthelevelofdamageitcansustain,butalsoto failureofsuchadaminJapan'shistory.Fromofficialrecords,there preserveitagainstfutureearthquakes. areabout210,000reservoirsinJapan,ofwhichatleast20,000are Although dam failures are rare, studies have been conducted vulnerable to future earthquakes. Therefore, the contribution of based on such events to understand the causes of those failures. thisreportistoraiseawarenessforreducingthepotentialofdam One example is the failure of the Teton dam, an earthen dam failuresinfuturecatastrophicevents.Tanakaetal.[24]studiedthe locatedinIdaho,intheUnitedStates.Itwas93mhighandhada failure of the Fujinuma dam; thus, comment and discussions on capacity of 355millionm3. The dam failed on 5 June 1976, as it thecauseofthedamfailurehavealreadybeenpresented. wasbeingfilledforthefirsttime,owingtointernalerosionknown This study aims to gain insight into the behavior of the as“piping”.Thefailurecausedahugefloodthatdamagedthecity FujinumadamduringthetimeoftheearthquakebyusingtheFE downstream,whichcostatotalofaboutUS$2billion[18,19].The method. Both the observed motion and the simulated motion Lower San Fernando dam, which was a 40m high hydraulic-fill are used to determine the response of the dam during the earthdamlocatedinSanFernando,California,failedon9February excitation. The soil behavior of the dam materials is described 1971, as the San Fernando earthquake struck southern California. byusingtheMohr–CoulombsoilmodelavailableinthePlaxisFE This seismic shaking triggered liquefaction of the hydraulic-fill code [25]. Dynamic analyses of the model are conducted, and within the upstream slope of the dam. During the earthquake, the overall dam behaviors presented. Discussions and compar- sliding was induced with a loss of strength in the liquefied soils, isons between the simulation results and existing facts are whichledtofailurecausingsignificantdamagedownstream[20]. expressed. 52 B.Charatpangoonetal./SoilDynamicsandEarthquakeEngineering64(2014)50–62 2. Sitede scriptionandinputmotions The spectra of both motions are shown in Fig. 4c and d. The predominant period of both motions was 2.71 and 2.96Hz, 2.1. Fujinumadam respectively. The smoothed spectra of both motions were also introduced byapplying a 50-period moving average (50per.Mov. The Fujinuma dam was an earth-fill embankment dam near Avg.) as shown in Fig. 4c and d. The predominant period of both Sukagawa,Fukushimaprefecture,Japan.Itwasestablishedonthe motionswastherefore2.73and4.40Hz,respectively.Inaddition, EbanaRiver,atributaryof theAbukumaRiver,16kmwestof the to verify the effects of the phase of ground motion, the opposite city office of Sukagawa (3711800700N,14011104100E). The dam con- wave forms were also used as additional input motions for struction began in 1937 and it was completed in 1949 after thisstudy. constructionwashaltedduringWorldWarII.Theprimarypurpose ofthedamwaswatersupplyforirrigation.Thecrosssectionofthe 2.3. Dammaterialsandtheirproperties maindamanditslayoutisshowninFigs.2and3.Thedamwasan embankment type,18.5m high and 133m long with a structural volume of 99,000m3 and a crest width of 6m. On the upstream Thematerialpropertiesofthedambodywereobtainedbased bothonstudiesofpreviousresearchersandonexperimentalwork face, the concrete frames and panels were installed in order to done in this study. As with previous investigations, information protect the dam from the erosion. However, there are gaps regardingsoilclassificationwasobtainedmainlyfromthestudyof between these panels so these panels were not served as the Onoetal.[23]andTanakaetal.[24].However,inthisstudy,field impervious layer and still penetrable. In addition, there was an investigationsandlaboratoryexperimentshavebeenconductedto auxiliary dam with a height of about 6m and a length of approximately 60m. The dam was at the head of an 8.8km2 extract the necessary information used for studying the dam drainageareagivingareservoircapacityof1,504,000m3[26]. failure. The data and discussions of dam material obtained from visualinvestigationsandlaboratoryexperimentsareexpressed. 2.2. Groundmotions 2.3.1. Visualinvestigation The2011TohokuEarthquake,ortheso-calledGreatEastJapan Following the dam failure, most of the upper portion of the Earthquake, was a magnitude 9.0 (Mw) undersea megathrust damwaswashedaway;theseveredamageorbreachlocationcan earthquake off the Pacific coast of Japan, which occurred on 11 beobservedatthemaximumcrosssectionneartheleftabutment March 2011. The epicenter was about 70km east of the Oshika fromdownstream.Ontheoppositeside,theremainingdambody Peninsula,inMiyagiprefecture,atadepthofapproximately32km consisted of the compacted fill materials. The foundations of the [27]. It was one of the most powerful earthquakes in the world damwereexposedtogroundlevelonthedownstreamside.From since the establishment of the modern recording system. The visualinspectionofitsproperties,thedambodycanbeclassified ground motion record obtained from Station FKS017, Sukagawa, intothreeparts:upper,middle,andbottom. Fukushima prefecture, Japan, was provided by the Kyoshin net- The upper layer comprises the top 6–8m to the dam crest. work and operated by the National Research Institute for Earth Unfortunately, most of the upper layer was flushed away when the Science and Disaster Prevention [28]. These data were observed dam failed. However, on the left abutment, a portion of the upper about15kmawayfromthedamsite.Furthermore,Hataetal.[29] embankment remains. Visual observation was performed on this also investigated the simulated ground motion at the dam site. section,whichrevealedthatthefillmaterialconsistsmainlyofcoarse They estimated the ground motion for the dam site by micro- sandoflightcolorvaryingfromlightbrowntogray(Photo1). tremor observation by using the site effect substitution method. Themiddleportionisabout7–9mthick.Thislayercanbeseen These motions were used in the dynamic analysis of the dam. clearlyintheremainingportionofthedam.Fromvisualobservationof The peak ground acceleration a was 4.198 and 4.25m/s2 for this portion, this layer appears to have been filled with brownish max theobservedandsimulatedmotions,respectively(Fig.4aandb). cohesivesoilcontainingyellowish-graysilt,loam-typeclay,andhumic Fig.2. Typicalcrosssectionofthemaindam. Fig.3. Maindamlayout. B.Charatpangoonetal./SoilDynamicsandEarthquakeEngineering64(2014)50–62 53 Fig.4. Inputmotions.(a)ObservedmotionatFKS01715kmfromthedamsite[28].(b)Simulatedmotionatthedamsite[29].(c)Spectrumofobservedmotion.(d)Spectrum ofsimulatedmotion. Photo1. Remainingportionofthefaileddam. silty sand (Photo 1). Moreover, traces of compacted layers can be Table1 identified,eachapproximately20–30cmthick. Soilsamples. The bottom layer is between 4 and 6m thick. This portion Typeoftesting Numberofsamples consists mainly of layers of loam-type clay and humic silty sand ranging from black to dark gray (Photo 1). In addition, previous Permeabilitytest 1 investigators[23,24]haveindicatedthatsomepartsofthebottom TriaxialCUtest 12 layer consist mainly of dark brown sand and gravel. The black (cid:2) Bottomlayer 4 (cid:2) Middlelayer 4 foundationmaterialofthedamconsistsoforganicresidualsmixed (cid:2) Upperlayer 4 withclayandsilt. Total 13 2.3.2. Laboratoryexperiment To facilitate the numerical simulation and study of the dam Foreachembankmentlayer,foursamplesweretaken,whichgave failure, several laboratory experiments have been performed to 12samplesintotalfortheentiredambody.AsshowninFig.5,the obtain the necessary information. Thirteen undisturbed cylinder soilssamplesweretakenfromlocationsa,b,andc,corresponding sampleswerecollectedatthesite.These samplescanbedefined to the upper, middle, and bottom layers, respectively. These based on their origin and testing purpose, as shown in Table 1. samples were tested to acquire the basic soil properties, particle 54 B.Charatpangoonetal./SoilDynamicsandEarthquakeEngineering64(2014)50–62 Fig.5. Soilsamplinglocations(GoogleEarth,image©2013DigitalGlobe). size distribution, and other parameters used in the analysis. The 13th sample, taken from the middle portion, was tested to establishthecoefficientofpermeability. 2.3.2.1. Sieve analysis. Sieve analysis is a practice or procedure usedtoassesstheparticlesizedistributionofagranularmaterial. Inthisstudy,duetotheunavailableofthelaboratoryinstrument at the time of study, the small particles analyses using a hydrometer were not conducted and the sieve number was in therangeof4–200.However,theinformationonthedistribution ofsmallparticleswasobtainedfromthestudyofTanakaetal.[24]. Theresultsshowthatboththemiddleandbottomlayersconsistof ahighpercentageoffineparticlesizes(Fig.6).Itcanbeseenthat Fig.6. Particlesizedistribution. μ for particles smaller than 5 m, the middle and bottom soil samples are found to be about 36% and 16% respectively. Also, 2.3.2.2. Triaxialtest. Atriaxialtestisacommonlaboratorymethod μ the claycontent (particles smaller than 2 m) of the middle and applied widely for obtaining shear strength parameters for a bottom portions is about 30% and 10%, respectively. Whilst, the variety of soil types under drained or undrained conditions. The upper part is consist mainly of sandy materials. The boundaries type of triaxial test applied in this study was the consolidated– for potentially liquefiable soil (b–b) and boundaries for most undrained (cu) test using cylindrical soil specimens of 50mm liquefiable soil (a–a) are shown in Fig. 6 [30,31]. Generally, to diameterand100mmlength.Thetriaxialtestresultsoftheshear judgetheliquefiabilityofsoil,morethanonecriterionisneeded. strength parameters—friction angle, cohesion, and other For clayey soils, Seed and Idriss [32] stated that these soils parameters—are presented in Table 2. Also, each parameter used could be susceptible to liquefaction only when all of these three in the analysis can be described as following. The dry and conditions are met: 1) the soil that contained less than 15% of saturated unit weight defined the mass for the element located 5μmandfinerparticles,2)liquidlimitislessthan35and3)water indryandsaturatedareas,respectively.Permeabilityorhydraulic content/liquid limit is greater than 0.9. Similarly, Andrews and conductivitydescribedthefluidmovementthroughporousmedia. Martin[33]summarizedthatsoilsaresusceptibletoliquefactionif In the elastic range, the modulus of elasticity E indicated the they contain less than 10% finer than 2μm and liquid limit less relationship between stress and strain within the elastic region, υ than 32, soils are not susceptible to liquefaction if they contain andPoisson'sratio providedinformationontheeffectoftheload greaterorequalto10%finerthan2μmandliquidlimitgreateror inonedirectioninrelationtothedeformationinotherdirections. equalto32,andfurtherstudyisrequiredforsoilsthatmeetone, The shear strength parameters cohesion c0 and friction angle butnotboth,ofthesecriteria.AccordingtotheSeedandIdriss[32] ϕ0 wereusedtoformtheyield surface forthis model. Finally,by criteria'sfirstcondition,boththemiddleandbottompartsarenot consideringslightcompressibilityofwater,therateofexcesspore metsothesoilsarenotbeingvulnerabletoliquefaction.Also,by pressurewasdefinedasK /n,inwhichK isthebulkmodulusof w w consideringAndrewsandMartin[33],boththemiddleandbottom waterandnisthesoilporosity. arenotsusceptibletoliquefaction.However,furtherstudyshould The parameters of cohesion and friction angle show that the be conducted in order to obtain more accurate results including bottom and middle portions of the dam are classed as cohesive liquidlimit. soil, but that the upper portion is classed as loose material. B.Charatpangoonetal./SoilDynamicsandEarthquakeEngineering64(2014)50–62 55 Table2 Soilproperties. Mater ials Layer γdry(kN/m3) γsat(kN/m3) kyffikx/4(m/s) E(MPa) Kw/n(GPa) υ c0(kPa) ϕ‘(deg) Dambody Bottom 16.0 18.00 5.5E(cid:3)7 50.0 1.87 0.3 18.4 31 Middle 14.0 16.00 5.5E(cid:3)7 30.0 1.12 0.3 7.80 32 Upper 16.0 18.00 5.5E(cid:3)7 17.5 0.65 0.3 0.0 37 Foundation – 30.0 – – 300 – 0.2 – – Fig.7. Originaldammodel. Fig.8. Remainingdammodel. Furthermore,themodulusofelasticityrevealsthatthedeeperthe Inaddition, soillayer,thehigheritsmodulusbecomes.Thetopportionofthe (cid:4) (cid:5) r¼r0þmp;m¼ 1 1 0 T ð4Þ dam is shown to have a relatively smaller modulus of elasticity thannormallyexpectedforfillmaterials. v^¼(cid:3)kðgradp(cid:3)ρ ðb(cid:3)u€ÞÞ=ρ g ð5Þ Itisshouldbenotedthatbytakingsamplesfromtheremaining w w of a collapsed dam represents a difficult task and raises many in which the relevant parameters are effective stress r0, pore questions on the reliability of the representativeness of the pressurep,totalstressr,coefficientofpermeability k,densityof ρ material properties of the retrieved samples. Therefore, the ana- water ,andgravitationalaccelerationg. w lyzedresultsobtainedusingthesedataandthefurtherusedofthis In this study, 956 triangular 15-noded elements with three- study,shouldberecognizedandawareofthislimitation. point Gaussian integration, a second-order polynomial interpola- tionforthe displacements, andafirst-orderinterpolationforthe pore water pressure were used. Two dam models have been 3. FEmethod proposedforthisstudy.Thefirstmodelisasimplifiedversionof the original shape of the Fujinuma dam's typical section and in 3.1. FEmodel this study the effect of the concrete frames and panels, and the interactionamongthemanddamisnottakenintoaccount(Fig.7), The FE analysis was performed under plane strain conditions. andthelatterisbasedontheremainsofthefaileddamfollowing The effect between the solid and fluid phases was carried out the site investigation (Fig. 8). The first model was used for through a u–p formulation; the governing equation of motion of conducting a dynamic analysis to acquire the dynamic behavior coupledsolid–fluidproblemcanbewrittenasfollows: and failure mechanism. The second dam model was used to LTr(cid:3)ρbþρu€¼0 ð1Þ determine the dam mode shape and natural frequency, and for verification against the microtremor observations. The modal (cid:1) (cid:3) analysis was conducted using ANSYS code [34]. The foundation mTε_¼divv^þ n p_ ð2Þ wasmodeledbyextendinga10m-thicklayer100meithersideof K w the center. The boundary conditions were restrained in the wherethetotalstressesr,thebodyloadsb,andtheaccelerationu€ horizontal and vertical directions at the bottom of the model. aretimedependentwiththedifferentialoperatorLgivenas For both edges, the boundary conditions were fixed only in the " # ∂=∂x 0 ∂=∂y lateral direction and were free in the vertical direction. To avoid LT¼ ð3Þ the effects of wave reflection on the boundary, an absorbent 0 ∂=∂y ∂=∂x boundary was applied at both ends and at the bottom of the 56 B.Charatpangoonetal./SoilDynamicsandEarthquakeEngineering64(2014)50–62 boundary. The dam's body was classified into three portions: upper,middle,andbottom.TheYoungmodulusisassumedtobe constantinanyofthethreeportionsofthedam.Inthisstudy,for thedynamicanalysis,thedamwasanalyzedunderthemaximum reservoir condition. The effect of reservoir pressure on the upstream face was applied as static pressure acting on the upstreamslopeinadirectionperpendiculartotheupstreamface. To acquire the initial stress condition, the static and seepage analyses were conducted in advance. Dynamic analysis was performedusingthePlaxisFEcode[25]. 3.2. Materialproperties In fact, the soil behavior is non-linear and to accurately reproduce the stress–strain relationship during dynamic load, it required the advanced constitutive law along with the finite element formulation at large deformation. As it is a preliminary study, the simple elastic-perfectly plastic model was applied and laterwhenthenonlinearfiniteelementcodehasbeenimplemen- Fig. 9. The cyclic soil behavior using the simple elastic-perfectly plastic constitutivelaw. ted by the authors, the further study will be conducted and comparedwithcurrentresults.Thebehaviorofthedammaterials is described using the Mohr–Coulomb soil model, which is an In dynamic analysis, this model produced elastic behavior elastic-perfectly plastic model with a yield surface whose elastic without the hysteretic damping. In order to include the soil's behaviorisdefinedbyisotropicelasticitythroughalinearYoung's damping characteristics, the Rayleigh damping is applied there- modulus E and Poisson ratio υ. In this study, the representative fore.Biggs [35] suggested that for geological materials, the mini- modulusofelasticitywasappliedusingtheinitialstiffnessofthe mumcriticaldampingratiocommonlyfallsintherangeof2–5%. actual soil's stress–strain curve obtained from the triaxial test. Also, many researchers [36–38] conducted the dynamic analyses Actually,thesmallervalueoftheprimarilymodulusofelasticityat ofanearthdamandappliedthedampingratiosofabout5%to6%. larger strain can be selected however as the samples have been Also,byusingtheMohr–Coulombwhichisincludedtheplasticity takenfromthefaileddamthatmightencounteredthedegradation model, when the failure surface is reached the energy is subse- of the modulus so the initial stiffness was applied in this study. quentlydissipatedthoughplasticdeformation.Therefore,Rayleigh The model has yield surfaces defined by cohesion c and friction dampingwasassumedinthedynamicanalysisbyconsidering5% angleϕ.The generalizationof theCoulomb frictionfailurelawis damping in the frequency range of 1.64–2.45Hz. However, by thendefinedby using the simple constitutive model Mohr–Coulomb, it must be notedthatthecomplexdynamicsoilbehaviorcannotbecaptured τ¼c(cid:3)s tan ϕ ð6Þ n comprehensively particularly the strain dependent behavior of whereτisthemagnitudeoftheshearingstress,s isthenormal stiffnessanddampingcannotbemodeled. n stress, c is cohesion, and ϕ is the angle of internal friction. The The parameters of the dam material were obtained mainly Mohr–Coulomb failure surface is often expressed in terms of the fromthelaboratoryexperiments,asmentionedearlierinrelation invariantsp,q,andrasfollows: to Table 2, whereas the properties of the foundation were pffiffiffi assumed. Each parameter used in the analysis was described ½ð1= 3Þsinðθþπ=3Þ(cid:3)ð1=3Þtan ϕ cosðθþπ=3Þ(cid:5)q(cid:3)p tan ϕ¼c earlierasstatedinSection2.3.2.2. ð7Þ where 9 4. Resultsanddiscussions rpqθ¼¼¼¼dðsðs1eii=it=iss33;Þj;ja(cid:3)rcscjoiss½iðjÞr==2q;Þ3(cid:5)>>>>=>>>>; ð8Þ 4.1.MModoadlalananalaylsyissisis conducted to extract the natural frequency anditscorrespondingmodeshape.Inthisstudy,twodammodels Theplasticflowruleusedinthisstudywascarriedoutusingthe were analyzed. Both cases were analyzed under the empty associateflowruleinwhichthedilatancyangleisexactlythesame reservoir condition. In Fig. 10, the results of the modal analysis valueasthefrictionangle,suchthatthesamesurfaceisappliedfor areshown.Thenaturalfrequencyis1.64and2.45Hzforthefirst thenormalizedplasticrule. two modes, respectively. For the remaining dam model, the The cyclic soil behavior of the simple elastic-perfectly plastic is natural frequency is 2.35Hz for the translational mode in the illustratedinFig.9.Itcanbeseenthatforstressstatesthatfallwithin horizontaldirection. the yield surface, the behavior is purely elastic and all strains are Fig.11 shows the amplification ratio that was taken from the reversible as showed at points A and B in Fig. 9. Thereafter, if the top of the remaining dam totwolocations near the dam base of system was loaded further until it reached or exceeded the yield the upstream and downstream slopes, defined as 2/1 and 2/3, surface,theplasticstrainorirreversiblestrainisoccurredasdefinedas respectively. By comparing the numerical results with microtre- pointsBandCinFig.9.Atthisstage,thetotalstrainisacombination mortestresults,itcanbeseenthatinthelatitudinaldirection,the betweenelasticandplasticstrainswhichisdefinedaspointCinFig.9. amplification ratio is within the range of 2–5Hz. The results Andifthesystemisunloaded,theelasticstrainisthenreducedand obtained from the modal analysis of the remaining dam agree finally vanished, left only plastic strain when the system is at rest well with field observations in relation to the vibration in the condition (Fig. 9C and D). Similarly, when reloaded the system, the latitudinal direction. Therefore, the estimated predominant fre- behaviorisrepeatedfromA–DinFig.9asmentionedearlier. quenciesoftheoriginaldamare1.64and2.45Hz.Thecomparison B.Charatpangoonetal./SoilDynamicsandEarthquakeEngineering64(2014)50–62 57 betwee nthenaturalfrequencyoftheremainsdamobtainedfrom expressed through acceleration, deformation, shear strain, the microtremor observation and the modal analysis of the excesspore pressureandtensioncrack. originaldammodelshowsthatthenaturalfrequencyoftheactual dam model is less than the microtremor test. This is due to the 4.2.1. Acceleration geometryoftheoriginaldaminwhichthedam'sheightishigher Results show that larger maximum crest acceleration can be whiletheremainsdamisplumpinshape.Itshouldbenotedalso obtained using inverse phase of the observed and simulated thatasthesoilspossiblysubjecttothelargestrainatthetimeof motions. Therefore, the acceleration contours at the time when the dam failure that may cause degraded values of the shear the maximum acceleration reached the dam crest are shown in modulus. As a result, the natural frequency of both microtremor Figs.12 and 13 for inverse phase of the observed and simulated testoftheremainsdamandthemodalanalysisresultoftheactual motions, respectively. By using the observed motion, the max- dammodelmaybedifferentfromthecurrentresultsiftheactual imumcrestaccelerationisdeterminedtobe7.42m/s2and8.56m/s2 materialpropertiescanbeacquired. accelerating towards the upstream direction at the time of 104.45s and 111.60s for normal and inverse phase motion (Fig.12),respectively.Forthesimulatedearthquake,themaximum 4.2. Dynamicanalysis crestaccelerationis7.60m/s2actingtowardstheupstreamdirec- tionatthetimeof72.78s.Similarly,foritsinversephase,thecrest A dynamic analysis is performed to establish the possible experienced the maximum acceleration moving towards the behavior of the dam during an earthquake. The results were upstream side with a magnitude of about 7.65m/s2 at 71.24s (Fig.13). The observed motion contains high-amplitude waves for a longer duration than does the simulated motion. This generates a great response throughout the dam body resulting from sig- nificant inertial force acting on the dam. This causes a great amount of plastic deformation and eventually the dam failure owing to the loss of freeboard. In addition, regarding to the spectrumoftheobservedmotion,itcanbeseenthatthismotion is characterized by a wide band. Therefore, it might possible induced resonance in the dam and caused a higher response. In contrast, the simulated motion contains high-amplitude waves only for a comparatively short period; thus, this does not con- tributeasufficienteffecttocauselargemovementandsettlement ofthedam. Fig.10. Natural frequency. (a) 1st mode: 1.64Hz (horizontal direction). (b) 2nd For both motions, it can be seen that the acceleration at the mode:2.45Hz(verticaldirection). dam crest is amplified significantly. The maximum amplification ratio taken from the crest to the base is 1.76 and 1.79 for the observedandsimulatedmotions,respectively.Toverifytheeffects ofthephaseofgroundmotion,theoppositewaveformswereused as input motions and amplification ratios of 2.03 and 1.80 were obtainedfortheobservedandsimulatedgroundmotions,respec- tively. This indicates that the response is very sensitive to input groundmotion. 4.2.2. Deformation Deformationcanbeusedfortheevaluationofthesafetyofthe dam due to freeboard loss. The results are expressed through its deformedshapeandthedisplacementcurvewithrespecttotime atvariousheightsalongthedamcenterline. Fig. 14 shows the deformed mesh at 104.19s and its corre- Fig.11. Amplificationratio. sponding displacement curves when subjected to the observed Fig.12. Accelerationcontours(inversephaseofobservedmotion). 58 B.Charatpangoonetal./SoilDynamicsandEarthquakeEngineering64(2014)50–62 Fig.13. Accelerationcontours(inversephaseofsimulatedmotion). Fig.14. Deformedmeshat104.19susingobservedmotion. Fig.15. Deformedmeshatendoftheanalysisusinginversephaseofobservedmotion. motion.Itcanbeseenthattheentiredambodywasmovedinthe motion, respectively.As a result,byconsideringovertopping fail- downstream direction. As time increases, plastic deformations ure, the dam was able to withstand safely the simulated input were accumulated. This caused the permanent horizontal displa- motion except for the case of using the inverse phase of the cement U of 2.95m, and crest settlement U reached 1.80m at observedmotion,forwhichoverflowisexpectedtooccur. x y 104.19s;thisisconsideredasthepointofdamfailureduetothe According to the facts, the dam experienced overflow and sub- loss of freeboard. Settlements at other observationpoints yield a sequentbreach.Theonepossiblecauseisthatthesignificantsettle- similar tendency, but the settlements were less with decreasing mentofthedambodyledtothelossoffreeboard.Thus,theobserved damheight. motiondoesreflecttheactualfailuremechanismofthedam,whereas In addition, the deformed meshes and deformation curves in the simulatedmotiondoesnot.However,allcasesshowthatsettle- bothdirections,attheendoftheshakingusingtheinversephase mentishigh,especiallyintheupperportionofthedamowingtoits of the observed and simulated ground motion, are shown in softness. For the changed phase, the observed motions show the Figs.15 and 16. These indicate that both horizontal and vertical significant effect of creating the displacement obtained from the displacements fluctuated during the earthquake and settled con- normalandinversephase,tobeinoppositedirections. tinuouslyatarapidrate,andthenremainedconstantuntiltheend of the excitation. The permanent horizontal displacement of the dam was 3.40m (upstream side), 0.30m, and 0.42m for the inverse phase of the observed, simulated, and inverse phase of 4.2.3. Shearstrain simulated motion, respectively. The vertical displacement at the Shear strainprovides information for understanding the loca- dam crest reached 2.90m, 0.63m, and 0.60m for the inverse tionwithinthedambodythatmightbedamagedseverelyduring phaseoftheobserved,simulated,andinversephaseofsimulated theearthquakeexcitation. B.Charatpangoonetal./SoilDynamicsandEarthquakeEngineering64(2014)50–62 59 Fig.16. Deformedmeshatendoftheanalysisusingsimulatedmotion. Fig.17. Shearstraincontoursat40sandat104.19susingobservedmotion. Fig.18. Shearstraincontoursat40sandattheendofmotionusingobservedmotion(inversephase). Fig.19. Shearstraincontoursat40sandattheendofmotionusingsimulatedmotion. Shear strain contours for various times obtained using the motionsalsoyieldgoodagreement,butonlywithasmallervalue observed motion are shown in Fig. 17. Most of the dam body ofshearstrain. experiencedaninsignificantrateofshearstrain.Itcanbeseenthat initially, large shear strains occurred in the upper portion, espe- cially on the upstream side. Thereafter, the occurrences of large 4.2.4. Excessporepressure shear strain can be observed clearly on both sides of the dam in The distribution of the excess pore pressure within the dam the middle and bottom portions of the downstream slope. This bodyandtherecordsoftheobservationpointsAandBat104.19s excessivelevelofshearstrainmayindicateapossiblecauseforthe usingtheobservedmotionareshowninFig.20.Itcanbeseenthat dam failure. Similarly, shear strain contours obtained from the the excess pore pressure developed mostly at the downstream remainingmotionsyieldalmostthesamepatternastheprevious side.Themaximumexcessporepressurewasabout90kPaandit motion(Figs.18and19).However,theexceptioniswhenusingan can be found near the dam's base close to the dam center line. inverse phase of the observed motion, for which large shear Besides, the suction or the positive excess pore pressure can be strains can be seen on the upstream slope instead of on the found on the upstream face and within dam's body on the downstreamslope. upstream side. It can be seen that from the time history curves The sequence of dam failure has been reported by previous oftheexcessporepressureattheobservationpointsAandBthat researchers. Initially, the dam experienced excessive deformation the build-up pore pressure was fluctuated at the early period of or a slide on the upper portion of the upstream slope. This, theshakingandthenkeptconstantuntiltheendoftheexcitation. together with a subsequent large slide that occurred on the Themaximumexcessporepressureobtainedfromtheobservation downstream face, yielded a loss of freeboard and triggered the pointsAandBwasabout48kPaand18kPa,respectively. overflowthatresultedinthebreachingofthedam.Mostnumer- In case of using inversed phase of the observed motion, the icalresultsshowgoodagreementwiththisscenario,exceptforthe distribution of the excess pore pressure shows that the build-up caseofusingtheinverseoftheobservedmotion,whichexhibited pore pressure was mostly concentrated on the upstream side. large strain appearing mainlyon the upstream slope. The simulated The maximum excess porepressurewas about 90kPawhich can