提高机械诊断质量的若干方法

1.2.3.4.5.6.7.8.9.SeveralwaystoimprovethequalityofmachineryfaultdiagnosisThesisSubmittedtoXi’anJiaotongUniversityInpartialfulfillmentoftherequirementforthedegreeofMaster(MechanicalEngineering)ByWang,LizhongThesisSupervisor:Prof.Qu,LiangshengMarch,2002-1-Morlet8K4BP-1-Subject:SeveralwaystoimprovethequalityofmachineryfaultdiagnosisSpeciality:MechanicalEngineeringName:Wang,Lizhong(Signature)Instructor:Qu,Liangsheng(Signature)AbstractInthisthesis,theresearchandanalysisfromsignalcollectiontosignalprocessingisdoneinordertoimprovethequalityofmachineryfaultdiagnosis.Firstly,themeasuresofsignalcollectionarereviewedandsummarized.Itisconsideredthattheconditionandenvironmentoffacilitieswillmoreorlesspollutethesignals.So,thecorrespondingsensorsshouldbeselectedaccordingtothedifferentconditions,andsomespecialmethodsshouldbeusedtoavoidbeingcontaminatedbynoise.Theaimistogethighqualitysignals.Secondly,thethesisisfollowedbysomeresearchworkinnoisesuppression,informationdecomposition,informationintegration,featureextractionandneuralintelligentdiagnosis.Theeffectivenessofnoisesuppressionbymeansofauto-correlationprocessingisproventheoreticallyandpracticallyinthisthesis.Itcanbeseenthatthenoisecanbesuppressedsomewhatcompletelyforgivenexamples.Forinformationdecomposition,thewavelettransformationisanalyzedusingmulti-resolutionwaveletandcontinuouswavelet.Throughcomparisonofdifferentprocessingmethodsusingsimulatingsignalsandpracticalsignals,aconclusionisachievedthatthecontinuouswavelet,especiallyMorletmotherwavelet,ismoresuitableforthefaultdiagnosisofmechanicalfacilities.Featureextractionofsignalsisalwaysthetargetfromthebeginningtotheendinsignalanalysis.Goodfeaturecorrespondstothehigherrateofdiagnostic-2-accuracy.Sothegoodfeaturescanalwaysimprovethequalityofsignals.Inthisthesis,normalbearingandthreekindsoffaultybearingsareinstanced,andbearingsignalsarepickedunderdifferentoperatingconditions.Bothaccelerationsignalsandacousticsignalsaresampled,andwithstatisticsimulationtheireightfeaturesarecalculatedandthedistributionsofthesefeaturesaredrawn.Afterbeinganalyzed,severalfeaturesofaccelerationsignalsandacousticsignalsarefoundtobemoresensitiveinfaultdiagnosis.Intheendofthischapter,somesyntheticfeaturesthatderivefromthefusionofthesefeaturesaremadetobemorepowerfulthananysolofeatures.Also,thisisanexampleofinformationintegration.Intheendofthisthesis,kurtosisandKfactor,whicharecalculatedfromaccelerationsignals,andcrestfactorandskewness,whicharecalculatedfromacousticsignals,areputintoaBPneuralnetwork.Theweightsofnetworkaretrainedforanautomaticdiagnosissystem.Finallythissystemistestedandprovedtobeeffectiveforrollingbearingfaultdiagnosis.Keywords:DatacollectionAuto-correlationWaveletTime-domainfeaturesArtificialneuralnetworksThesis:ApplicationFundamentals-1-11.11.1.11.1.21.1.31.21.2.11.2.21.31.422.12.22.2.12.2.22.2.32.2.42.32.3.12.3.22.4........................................................................................................1................................................................................................1...................................................................................1...........................................................................2.......................................................................3...............................................................................4............................................................................................4.......................................................................................5..................................................6...................................................................................7...............................................................8...................................................................................8...................................................................................9...................................................................................9.................................................................................10.....................................................................................10.................................................................................10.................................................................................11.................................................................................11..............................................................................................11......................................................................................................15-2-33.13.1.13.1.23.23.2.13.2.23.2.33.33.444.14.1.14.1.24.24.2.14.2.24.355.15.1.15.1.25.2.........................................................................16.....................................................................16.....................................................................................16....................................................................