神经网络在地震反演中的分析与应用

《神经网络在地震反演中的分析与应用》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、’、，A、‘．呻t涞-·噜Subject：NeuralnetworkintheresearchandappHcafionofseismicinversionSpeciality：Name：lnstructor：ABSl’I【AC。I。Seismicinversionisthebasicissueofseismicexploration,itspurposeIStoextractinfornmtiollscontainedbytheundergroundstrataseismicvelocitystructureandthephyslCaIparametersp

2、atialdistributionofgeologicalstructure，toprovidetheimportantbasisfortheresourceexplorationanddevelopment．Sincemostofthegeophysicalinversionsbelongtothetargetfunctionoptimizmionmulti．optima，linearinversionmethodcanleadtothemVerslonresultsinlocaloptimum．Thispaperproposesbasedonthepart

3、icleswarmoptimization,calculation．Inthisway,notonlyhastheneuralnetworktrainingalgorithm(PSO_BP)thatisappliedtononlinearseismicinversionBPneuralnetworksearchperformancebeenimproved，butalsotheseismicinversionofoperationefficiencyandinversionprectslonimprove．Moreover，forseismicwaveimpe

4、danceinversion,itprovidesanewefficientmethod．Thispapercontainingtheresearchachievementsandworkareasfollows：(1)心oundtheexistingproblemsofseismicinversiontechnology,thispaperdiscussesapplicationstatusandtechnologyprincipleintheneuralnetworkandtheseksmlcinVerslon·InresearchBPalgorithma

5、pplication,learningmodeandtheimprovedmethod，thispaperbasesontheproblems，suchasinBPneuralnetworkinseismicinversionprediction,thenetworkstructuredesignbeingnotnimble，weightstogaininsingle，easytofallintothelocalrulesofthemostsup嘶0rlimitations，itproposeswaystotheparticleswarmoptimizatio

6、nalgorRhmtooptimizeneumlnetworkconnectionweightsandthreshold．ItalsodesignsthePSO—BPalgorithmofthetwoalgorithmsmutualinfiltrationandreciprocal·(2)Thispaperin-depthresearchesandsummarizestheprincipleofPSO—BPalgorithm，basicstructurc觚drealizationmodel．Throughthestandardtestfunctionexper

7、iment，itfurtherdiscussestheoperationparametersvaluerange，feasibilityandstabilityoftheparticleswarmoptimizationalgorithm,especiallyengagesinamoredetailedresearchintheconvergenceofPSOalgorithm,andsummarizessomeguidingrules．洲nseismicinversionsimulationexperiments，thispapercomparesthePS

8、O—BPalgorithmwithth