基于汶川余震记录的强震台场地分类研究.doc

基于汶川余震记录的强震台场地分类研究 Vol.10,No.3EARTHQUAKE ENGINEERINGAND ENGINEERINGVIBRATION September,xxEarthq Eng&Eng Vib (xx)10:325-337DOI:10.1007/s11803-011-0069-xImproved HVSRsite classif i cationmethod for free-f ield strong motion stationsvalidated withWenchuan aftershock recordingsWen Ruizhi1?,Ren Yefei?and ShiDacheng?Institute of Engineering Mechanics,China EarthquakeAdministration,Harbin150080,ChinaAbstract:Local site conditions play an importantrole in the effectiveapplication of strong motion recordings.In the China National Strong Motion Observation Network System(NSMONS)some of the stationsdo notprovide borehole information,and correspondingly,do not assign the site classesyet.In this paper,site classi f i cation methodologies forfree-f ield strong motion stationsare reviewedand the limitations anduncertainties of the horizontal-to-vertical spectral ratio(HVSR)methods are discussed.Then,a newmethod for site classi f i cation based on the entropy weight theory isproposed.The proposedmethod avoidsthe heador tail joggle phenomenonby providingthe objectiveand subjective weights.The method was appliedto aftershock recordings from thexxWenchuan earthquake,and54free-f ield NSMONS stations were selected for site classi f i cation and the mean HVSRs werecalculated.The results show that the improved HVSR methodproposed in this paperhas ahigher suess rate and could be adopted inNSMONS.Keywords:site classi f i cation;strong motion recording;entropy weighttheory;horizontal-to-vertical spectrum ratio;Wenchuan earthquakeaftershock;head-tail joggle1IntroductionLocal siteconditions have a signi f i cant in f l uence on ground motion characteristics andseismic performance of engineeringstructures.Different siteconditions mayinduce variedampli f i cation of the ground motionleading toabnormal earthquake damage phenomenon(Hu et al.,1980;Zhou,1990;China EarthquakeInvestigation GroupAboard toJapan,1995;Li,1996).It wasobserved from the1906San Franciscoearthquake that the mostimportant characteristicsof the strong motionshowed that there was a clearcorrelation between the earthquakeintensity of a siteand itsunderlying geologionditions.Wood (1908)found thatthe variabilityof the surface geologyin theSan FranciscoBay areacontributed the most thesignif i cant change in thestrong motioncharacteristicsduring theirsite investigation.Subsequent earthquakes,such as the1923Great Kantoearthquake(Ohsaki,1969),1976Tangshan earthquake(Gao andHu,1987;Liu andCha,1982),Correspondence to:Wen Ruizhi,Institute of Engineering Mechanics,China EarthquakeAdministration,Harbin150080,ChinaTel:86Fax:86-mail:ruizhiiem.?Professor;?PhD CandidateSupportedby:National KeyTechnology R&D ProgramUnderGrant No.xxBAK55B05;Nonpro f i tIndustry ResearchProject ofCEA UnderGrant No.xx08003;Science Foundation of InstituteofEngineeringMechanics,CEA UnderGrant No.xxC01Received March21,xx;Aepted August2,xx1985Mexico earthquake(Seed et al.,1988),1999Chi-Chi earthquake(Tsai andHuang,2000),and the recentxxWenchuan earthquake(Bo et al.,xx)have allvalidated theimportant effects of the siteconditionson buildingdamage.The conceptof site classi f i cation has been gradually incorporatedin seismic codes inmany countries.The February27,xxChile earthquakesupported thisknowledge of site classi f i cations inan interestingway.An investigationfound thattwo similarbuildings located20m apartnear theLlacolen bridgein downtownConcepcion wereimpacted inpletely differentways:one wasdestroyed and the othersuffered onlyminor damage.Their proximityexcludes theeffectsof the soilconditions,and thisabnormal ourrencemay beattributed to the site classi f i cationused indesign.The destroyedbuilding wasdesigned usingSite TypeII,while theother wasdesigned usingSite TypeIII(GEER AssociationTeam,xx).Strong motion recordings arewidely used in investigations and engineering practice,such as seismic zonation,seismic riskanalysis andearthquake resistant design,and responseanalyses ofbuildings,among others.In recentyears,strong motion data havebee moreavailable from a diversearray of organizations andservices.High quality strong motion recordings arenot justa setof qualif i ed aelerograms,but also include stationlocation,earthquake data sources,ground motionparameters,and otherinformation pertainingto the particular recording.Site classif i cation of a strong motion station is oneof the parameters requiredto determine the suitabilityof its recordings forspeci f icapplications.326EARTHQUAKE ENGINEERINGAND ENGINEERINGVIBRATION Vol.10The ChinaNationalStrongMotionObservationNetworkSystem(NSMONS)was deployedinxx.During theM s8.0Great Wenchuan earthquake,more than1,400high-qualitystrong motion recordingswere obtained(Li et al.,xx),and thenmore than2,000sets of3-channel strong motion recordingswere obtained from383aftershocks.After themainshock,another59strong ground motion instrumentswere temporarilyinstalled alongthe LongmenshanFault region,and more than3,250sets of3-channel recordingswere collectedfrom the aftershocks.All thesedata enrichedthe Chinesestrong motion database(Li,xx).In China,standard strong motionrecordingprocessing includesa reviewand processingto reducerandom noisein therecorded signals;it doesnot includesite classif i cation information.For NSMONS,some stationsdo nothave adequateboreholeinformation,so nosite classif i cationhasbeen assignedfor thesestations.Aording to the Chineseseismiode,the boreholepro f i leis usuallyto20m depth and the average shear-wave velocityis calculatedfrom thetop soil layer of20m depth(V20s).The averageshear-wave velocity from thesurface to30m depth(V30s)is nowadopted as an internationalstandard forsite classif i cation.Wen et al. (xx)provided the site classif i cationsof77near-fault stationsby usingthe HVSR method and the response spectral shapes(RSS)method.These classif ications werebased onlyon therecordings fromWenchuan mainshock,and asa result,they lackreliability.In thispaper,an improvedHVSR methodis suggested and appliedto54free-f ield strongmotion stationsconsidering therecordings from the Wenchuan earthquake aftershocks.2Site classif ication schemes2.1Development ofHVSR schemesTheHVSR methodwas fi rst proposedby Nakamura (1989),who used a horizontal-to-vertical Fourierspectrumratioof interestground microtremor to evaluate the sitecharacteristics.Yamazaki andAnsary (1997)extended this method toearthquake ground motionrecordingsto putehorizontal-to-vertical Fourierspectrum ratios.They foundthat itwasastable method,regardless of thegroundshaking level,station-to-source distance,and top soil layerdepth,andcould beauseful toolforsitecondition evaluation.Lee et al. (xx)usedascheme patiblewith the1997UBC provisionsto classify708free-f ield strongmotion stationsites obtainedfrom theTaiwan Strong-motion InstrumentationProgram(TSMIP),in which the RSSmethod andHVSR methodwere bothused forveri f icationpurposes.Their resultshave sincebeen widelycited byresearchers inengineering seismology(Hwang et al.,xx;Sokolov et al.,xx,xx;Liu andTsai,xx;Lin andLee,xx;Roumelioti and Beresnev,xx).Zare et al. (1999)provided free-f ield stationclassif ications for the Iranstrongmotionwork.Zhao et al. (xx)used H/V ratiosfor recordsfrom theclassif i ed K- sitesto establisha site classif icationindex usingmean spectralratios over a widerange ofspectral periods.Fukushima etal. (xx)suggested thatHVSR was an effectivemethod whenborehole datawas unavailable.After theChi-Chi earthquake,the Centerfor Researchin EarthquakeEngineering(NCREE)and WeatherBureau(CWB)of Taiwanpleted boreholemeasurement andprovided PSlogging datafor439strongmotion stations from2000toxx.Then,Lee andTsai (xx)bined theGeoxxdrilling databaseof theGeological Survey(CGS)of Taiwanto evaluateV3030s valuesfor eachgrid-point,and pleteda V s mapof Taiwan.The strongmotiondata from twoPingtung,Taiwan,earthquakes thatourred onDecember26,xx,showed thatground motionson soilsites aregenerally largerthan thoseon rocksites.In theNext GenerationAttenuation(NGA)research project,metadata characterizingeach recordingwas developedthat includedthe stationsite classif ication andV30for groundmotion researchands,and the improved dataquality wasmade availableengineeringpractice(Chiou etal.,xx).Ghasemi etal. (xx)improved thepracticality andeff iciency of the HVSR method introducedby Zhao etal. (xx)through Spearmanrank technology.Garnielet al.(xx,xx)improved this method witha waveletanalysis andself-organizing mapmethod tomore eff i cientlydetermihepredominant frequency.2.2HVSR classif ication schemes(Japan RoadAssociation,1980)-Method1Table1lists the site class de f i nitionused in the Japan earthquake resistant design code,together withthe approximatelycorresponding NationalEarthquake HazardReduction Program(NEHRP)site classes.The site natural periodasakey indexof thesite classif ication can be approximatelyobtainedfromthe HVSR curves.Figure1illustrates theempirical HVSR curves plottedby Zhao etal. (xx)for foursite classes.5.0o4.5SCISCIIi ta rl4.0SCIIIaSCIVr t c e3.5p ses3.0n op s2.5er V/H2.01.51.00.11Period(s)Fig.1Mean HVSR plots for different site classes(Zhao etal.,xx).The verticaldashed linesrepresent thebounds of two adjacentsiteclassesprovided in Table1No.3Wen Ruizhietal.:Improved HVSRsite classificationmethod forfree-field strongmotion stationsvalidated withW enchuanaftershock recordings327Table1Site classde f i nitionusedinJapanearthquakeresistantdesigncode and the approximatelycorresponding NEHRP siteclass(Japan RoadAssociation,1980)Site classSC I:(Rock/stiff soil)SC II:(Hard soil)SC III:(Medium soil)Site natural period(s)T G0.2s0.2sT G0.4s0.4sT G600m/s300m/s (xx),who providedan example fromthe Rezvanshahr stationin Iranstrongmotionwork.They pleteda suessfulidenti f ication for SC I,SC II,and SC III,and foundthat only57%,43%and42%,respectively,and the low eff iciency alsomanifested atJapanese K-(Zhao etal.,xx).Zhao etal. (xx)suggested thatthis methodcouldbeappropriate for SC IV and SC III,but wasunreliable for SC I and SC II.For SC IVand SC III,namely softsoil,the longperiod ponentis amplif iedwell.However,forSCIandSC II,it isusually notpossible tochoose the peak periodat highfrequency,so thesiteclass is ambiguous.2.3HVSR classification schemes(Zhao etal.,xx)-Method2Zhao etal. (xx)suggestedanindex classification consideringpeak period and theH/V spectralratios atall periodsto improvethe auracy of theprevious schemeas follows:SI k=2F?abs?ln(?i)?ln(?ki)? (1)ni=1nFollowing a detailed analysis,it canbe statedthatthere are tworeasons thatcontribute to this lowauracy.First,thismethodis essentiallybased on the standard shape ofH/V ratios.The standardshape isactually thegeometric meaningof selectedstrongmotionrecordings,so the standard deviationshould havebeen included.Second,from Eq. (1),the SI value expressesthe similarityof thespeci ficstation H/V ratioshape tomatch the standardshape.For eachperiod,the contributionto SIis equal.Actually,the peakperiod and its adjacentones shouldhaveahigher contributionto SI.In orderto provideadetailedexplanation,three scenarioHVSRcurvesare constructedand are shown in Fig.2.In Fig.2,the scenarioHVSRcurveA hasan obviouspeak valueat0.3s,and thesegment at0.20.4sisclose to the standardSC IIcurve;thus,the idealresult shouldbe SC II.In Method2,thesiteshould belongto SCI fromTable2.This isattributed toCurve Abeing tooclose to the standardSCI curveat periodsbetween0.7s;this iscalled the head-tail jogglephenomenon.Ideally,the contributionaround thepeak valueat0.20.4s shouldbe improvedand thecontribution attheheadScenario HVSRA ClassI givenby Zhaoetal. (xx)Scenario HVSRB ClassII givenby Zhaoetal. (xx)Scenario HVSRC ClassIII givenby Zhaoetal. (xx)5.5H/V response spectralratio5.04.54.03.53.02.52.01.51.0()where kis thesiteclassnumber,n is the total number of periods,F()is thenormal cumulativedistribution function,?i is the meanH/V ratiofor the ith periodof theinterest site,and?ki is thestandardH/V ratiofor theith periodwith respectto thekth siteclass.The suessrate forSCI,SC IIIandSC IV isimproved butis stillonly30%40%forSCII.Ghasemi etal. (xx)arrived ata similarconclusion withdatafromthe Iranstrongmotionwork.0.11Period(s)Fig.2Scenario HVSRs and standardHVSRs(Zhaoetal.,xx)Table2SIvalue of scenarioHVSRsNo.ABCSuggested classSCIISC ISCIISISC I0.9050.8310.674SCII0.8720.9010.778SC III0.7480.7530.882SCIV0.6380.6620.705NoteHead-tail joggleTail joggleTailjoggle328EARTHQUAKE ENGINEERINGAND ENGINEERINGVIBRATION Vol.10and tailof thecurve shouldbe reducedto obtainthe bestresult.A similarsituation astail joggleours forCurve Band CurveC.For longperiods,the amplification isnotable,which matchesthestandardSCIVcurve;thus,the probabilityof joggleis muchless than for theother standard curves.SCIIcould haveeither headjoggle ortail joggle,resulting inless auracy,which isconsistent withthe conclusionfrom Zhaoetal. (xx).2.4HVSR classificationschemes(Ghasemi etal.,xx)-Method3Ghasemietal. (xx)redesigned SIbased onSpearmans rankcorrelation coefficient(Wolfrom,1999):=n1?6d2SIikn?1) (2)i=1n(2where dfor the ii is therank difference between eachHVSR valueth periodwith respectto thekth siteclass,and nis thetotalnumberofperiods.An SIranging from-1to1is used to measurethe correlationbetween the meanHVSRcurve for thesiteof interestand thestandardcurveswithout consideration the frequencydistribution,while SI=1indicates aperfect positivecorrelation.3Improved HVSR basedon entropy weighttheory-Method43.1Entropy weighttheoryEntropy isa quantitativemeasure ofdisorder ina system.The conceptes fromthermodynamics,which aountsfortheheat energytransfer withina system.Shannon (1948)introduced thismathematical theoryinthemunication field;besides thereare alsosome applicationsin earthquakeengineering(Harte andVere-Jones,xx;Dong etal.,1984;Feng andHong,xx;Main andNaylor,xx).The formulaofEis recognized as thatof entropyand defi nedaording tostatistical mechanics:mE=?p iln p i (3)i=1where pmm is the numbersof possible states,0iis the probabilityof asystem intheithstate,andp i1,pi=1.i=1Weight assignmentde fi nesthe relativeimportance andinfluenceof theinput parametersinthefi naljustif ication.The msets of the schemeincluding nindicators areusedtoassemble the assignment array Rbasedonentropy weighttheory.The elementr representsthe evaluation grade forthe jth indicator of ij ofarray Rtheith scheme.R=(r ij)mn(i=1,2,m;j=1,2,n) (4)If theexcellent valueis r*j,i.e.r*=?maxr?iij,more excellent,larger asjth indexj?minir ij,more excellent,smaller asjth index (5)The normalized assignment array B will beB=(b ij)mn(i=1,2,m;j=1,2,n) (6)where theelement b ij=r*ij/r jand it isevident that0b ij1.Aording to this procedure,the entropyof jthindicator is de fi nedE j=?mf ijln fij (7)where fiji=1ij=bmb iji=1In statistics,if anindicator representsgreater discrepancybetween eachrespective scheme,it willmake moreofacontribution inthe evaluation system,anditscorresponding weightwill behigher.If the discrepancy tendsto zero,so willthe weight.The entropymeasures the uncertainty ofa distributionand reachesa maximumwhen theprobabilities areuniform.The normalizedentropy measurementof the jthindicator is:E jE j1me j=E=? (8)maxlnm lnmfijln fiji=1Then the objective weightof jthindicator canbe givenas:h1?e j1?e jj=n=n (9)(1?ej)n?e jj=1j=1nwhere0h j1,h j=1jEquation (9)states thatthe indicatorswith less=1entropy valueshave higherlevels ofinformation content,and ahigher weightis assignedto them.The decisionmaking alsoneeds toinclude empiricalexperience,for whichthe subjectiveweight shouldbe considered.Equation (10)is appliedto biheobjective weight h j withthe subjectiveone wthindicator parameter.j toevaluatetheintegrated importanceof thej?+w jhj+w jj=hjn=(h+n (10)jw j)1+w jj=1=j13.2Procedure ofentropy weightevaluationWhile the entropy weightof each indicatorisinvolved,theassignmentarray Bwill betransferred intoA,as shown in Eq. (11)A=(a ij)mn(i=1,2,m;j=1,2,n) (11)where theelement a ij=b ijj,and theindicator vectorof theith schemeof theA:No.3Wen Ruizhietal.:Improved HVSRsite classificationmethod forfree-field strongmotion stationsvalidated withW enchuanaftershockrecordings329A i=(a i1,ai2,a in) (12)Calculating the expected assignmentof the indicators asP=(p1,p2,p j,p n) (13)where p j denotesthe maximumvalueofjth columnin array A,as shownp j=maxa iji=1,2,?,m(j=1,2,n) (14)iThen the similarity C i between vector A i and vector P isP?A iP?PT?A i?PTC i=P P?PT(p)?(a2jj=1j=1jn nij?pj)=1?(aj=1nj=1nij?pj) (15)2j(p)j=1n2(p)In theend,the minimumC imeans thebest decisionwith respectto theith schemefor allof thepossiblestates.3.3Test casestudy Atest caseto illustratetheimprovedHVSRmethodin solvingthe“head-tailjoggle”issue inMethods2and3is discussedin this section.Suppose thattherearefour typesofsite classificationand fi ve indicators,i.e.*m=4and n=5,r j=maxr ijwillbeset astheexpectedi excellentvalue.To showthe differencebetween theevaluated values,the followingparameters areset:r4jr1jr2jr3j,then rj*=r4j(j=1,2,5),b4j=1.And,b3j=x,b1j,b2j andb3j are as listed in assignmentarray Be,see Table3.The entropyof eachindictors ej(j=1,2,5)is calculated anditwas foundthat ej isa functionwhose valueincreases asthe variablex increases,and the variation of eachindicatordecreases.When xreaches maximum,theentropyis the maximum.Meanwhile,the variationofeachindicator tendsto minimum,especially forj=1;when x=1,theindicatorsvalue willbe uniform,thevariationis zero,andtheentropy weightis zeroas well.33In addition,for jthindicator,(b4j?bij)=3?b iji=1i=1was usedto showthediscrepancybetween theother indicatorvalues andthe maximumone.As shown in Table3,the differencedecreases fromj=1to j=5;however,Fig.3shows that for agiven x,theentropygra