搜索
37220032JOURNALOFXIANJIAOTONGUNIVERSITYVol.372Feb.2003,(,710049,):,,.,23BP,,.,,,.,,,.:;;;;:TP183;P63118:A:0253n987X(2003)02n0192n04IdentificationofCementQualitybyNeuralNetworkwithNonlinearConnectedWeightsXiaKewen,SongJianping(SchoolofElectronicsandInformationEngineering,XianJiaotongUniversity,Xian710049,China)Abstract:Therearesomeproblemsinpresentcementbondlogofoilcasing2wellengineering,suchasthebigunavoidableerrors,in2correctinterpretationresultbytheamplitude2comparemethodbasedonthetraditionalstatisticstheory.Tosolveaboveproblems,theimprovement2typecompensationsonicinstrumentwiththedualtransmittersanddualreceiversisadoptedtomeasuresonicamplitude,andtheerrorsstemfromthetoolsinclination2eccentricity,sonicsignalattenuationinmud,toolssensitivity,andsoon,canbeelim2inatedbyusingofintegratedheadwaveamplitudevalues.Themethodofartificialneuralnetwork(ANN)withnonliearconnectedweightssuperiortothatofstatisticstheoryisstudied,whichcanreplacethree2layererrorback2propagation(BP)algorithm,sotheimplied2layerremoved,thecalculatingsimplified,andtheoperatedspeedincreased.Inconcreteapplication,firstitmusttobecali2bratedforsonicamplitudevaluessoastoacquireallkindsofsamples,thentheANNisconstructedandtrainedsoastoacquirethenonlinearconnectedweightsandthresholds,finally,thecementqualitycanbeautomaticidentifiedinloggingprocess.Actualapplica2tionexampleshowsthatthemethodofANNcanrealtimetoidentifycementquality,theidentificationaccuracyismuchbetterthanthatofamplitude2comparemethod,andtheapplicationeffectisverynotable.Keywords:artificialneuralnetwork;nonlinearconnectedweights;patternidentification;cementbondlog;cementquality,,().(),,(CBL),n(CBL2VDL).().,,[1].,.,.,[1],20%,;20%30%,;30%,.,.:2002n05n17.:(1965),,,;(),,,.:(2000206n2).©1995-2004TsinghuaTongfangOpticalDiscCo.,Ltd.Allrightsreserved.,,..(ErrorBackPropagation,BP),BP[2],[3].,2(1),4,.(2),23BP[4],BP.1BP33,,,.(1),23BP,,[4].1,Im,Jn.Ixp1,xp2,,xpm,Jdp1,dp2,,dpm.pp,,,f,f(x)=1/(1+e-x)(1)f(x)=f(x)(1-f(x))(2)IiJjwijf(vij),wij,f(vij)(1)1.f1,BP.JjNjNj=mi=1wijf(vij)xi(3)J,jyjyj=f(Nj-j)(4):j.I,BP,.s,E=12sp=1nj=1(dj-yj)2(5):dj.j=-5E5Nj(6)j=(dj-yj)yj(1-yj)(7),wijf(vij),wijvij-5E/5wij-5E/5vij,wij=-5E5wij=--5E5Nj5Nj5wij=j55wijmi=1wijf(vij)xi=(dj-yj)yj(1-yj)f(1-f(vij))xi(8)vij=-5E5vij=--5E5Nj5Nj5vij=j55vijmi=1wijf(vij)xi=(dj-yj)yj(1-yj)wijf(vij)(1-f(vij))xi(9)wijvij,(8)(9)wijvijwij(k+1)=wij(k)-5E5wij(10)vij(k+1)=vij(k)-5E5vij(11)kk,.BP.22.1(1).3912,:©1995-2004TsinghuaTongfangOpticalDiscCo.,Ltd.Allrightsreserved..,,.,..,,1.01,..,,.,,,=+a,E0-b,E00,(12):EE;ab01011.(2).,,4.,,,,,.2.22.2801,,1m,114m,014m.,43,4A11A12A22A21(mV),.(1),.1A11/mVA12/mVA22/mVA21/mV(d1,d2,d3)311542115012541664117114741584129616181722112716281532163517481343155019291942154916551764147016741385179217191222182417201933123612371844115013(1,0,0)(1,0,0)(1,0,0)(0,1,0)(0,1,0)(0,1,0)0,0,1)(0,0,1)(0,0,1)230,A11.9617%,70%.,.3(1),4,4.(2),23BP,,,.,.,.49137©1995-2004TsinghuaTongfangOpticalDiscCo.,Ltd.Allrightsreserved.2A11/mVA12/mVA22/mVA21/mV12345678910111213141580116716571944175517521154177514119147015781839184915321189153716331827122110271833122517371671162819411819112417101344188111681357134318561551115511761512115711283164115481733129011381234122612211828132711261538137115281140172111251611124316A11/mVA12/mVA22/mVA21/mV16171819202122232425262728293018411461574141931026192914561743172817701690155816126132613204151851420193719164121219141119111319111536174513271588141016184111851446197511192152711281957114212271171139011571412718251420319148122011371116312131414171914121710143614421626178819111318312(3),,.,.,.,f(x)=-1/2+1/[+exp(-x)].34,.:[1].[M].:,1989.129146.[2],.[M].:,1991.3245.[3],,.BP[J].,2002,16(2):3032.[4],,.[A]..[C].:,1993.322323.()(138),5145%,8000.IC,.,,.:[1],,,.[M].:,1995.[2].MEMS[D].:,2000.[3]LammerinkTSJ,ElwenspoekM,VanouwerkerkRH,etal.Performanceofthermallyexcitedresonators[J].SensorsandActuators:A,1990,21n23:352356.[4],,,.[J].,2002,36(4):377380.()5912,:©1995-2004TsinghuaTongfangOpticalDiscCo.,Ltd.Allrightsreserved.