基于矢量化反射率法的多波叠前联合AVA反演

(整期优先)网络出版时间:2018-03-13
/ 1
MostcurrentprestackAVAjointinversionmethodsarebasedontheexactZoeppritzequationanditsvariousapproximations.However,theseequationsonlyreflecttherelationbetweenreflectioncoefficients,incidenceangles,andelasticparametersoneithersideoftheinterface,whichmeansthatwave-propagationeffects,suchassphericalspreading,attenuation,transmissionloss,multiples,andeventmismatchingofP-andS-waves,arenotconsideredandcannotaccuratelydescribethetruepropagationcharacteristicsofseismicwaves.ConventionalAVAinversionmethodsrequirethatthesewave-propagationeffectshavebeenfullycorrectedorattenuatedbeforeinversionbuttheserequirementscanhardlybesatisfiedinpractice.Usingaone-dimensional(1D)earthmodel,thereflectivitymethodcansimulatethefullwavefieldresponseofseismicwaves.Therefore,weproposeanonlinearmulticomponentprestackAVAjointinversionmethodbasedonthevectorizedreflectivitymethod,whichusesafastnondominatedsortinggeneticalgorithm(NSGAII)tooptimizethenonlinearmultiobjectivefunctiontoestimatemultipleparameters,suchasP-wavevelocity,S-wavevelocity,anddensity.Thisapproachisrobustbecauseitcansimultaneouslycopewithmorethanoneobjectivefunctionwithoutintroducingweightcoefficients.Modeltestsprovetheeffectivenessoftheproposedinversionmethod.Basedontheinversionresults,wefindthatthenonlinearprestackAVAjointinversionusingthereflectivitymethodyieldsmoreaccurateinversionresultsthantheinversionbyusingtheexactZoeppritzequationwhenthewave-propagationeffectsoftransmissionlossandinternalmultiplesarenotcompletelycorrected.