基于声音信号的轴承故障诊断方法研究

《基于声音信号的轴承故障诊断方法研究》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、大连理工大学硕士学位论文基于声音信号的轴承故障诊断方法研究姓名：卫耀东申请学位级别：硕士专业：电子与通信工程指导教师：余隽20081221大连理上大学专业学位硕士学位论文BearingFaultDiagnosisMethodBasedonAudioSignalAbstractRollingbearingiswidelyusedinlotsofmachineryandplaysfundamentalroleintheindustry．Beatingfaultshavedirectinfluenceontheperformanceofthewholemachine，SOanaly

2、sisoffaultmonitoringandfaultdiagnosisisverysignificant．Audiosignaldiagnosishastheadvantagesofspeedy，convenientsignalacquisitionandeasyhandling．Thisthesisstudiesthediagnosticmethodsofthesinglefaultmodelandthecompoundfaultmodelusingaudiosignalasthefailuresignalsource．Becauseanartificialdiagno

3、sticmethodbasedonthecombinationofwavelettransformandenvelopeanalysiscannotdiagnosesomefaultseffectively，animprovedmethodisproposedthatappliesenvelopeanalysistoeachhighfrequencydetailedsignalsdecomposedandreconstructedbywaveletanalysis．Theexperimentalresultsindicatethatthemodifiedmethodhasth

4、eevidentadvantageinaccuracy；meanwhilethemethodisreal-time，SOitislikelytobeadoptedintheengineering．ThisthesisuseswavelettransformandwaveletpackettransformtoextractenergycharacteristicsandthenappliesBP(BackPropagation)neuralnetworktodiagnosebearingfaultintelligently．Accordingtocomparisonwefin

5、dthatwaveletneuralnetworkhasabetterperformance，andwealsofindtheintelligentdiagnosticmethodisslightlybetterthantheartificialdiagnosticmethod，butintelligentmethodneedssubstantialsamplestobetrainedfirst，whileartificialmethodCandiagnosebyspectrumdiagramdirectly，SOwecanusebothofthemtocompareandt

6、oanalyzeintheengineeringapplication．Incaseofseparatingmix—signal，thispaperusesthefastindependentcomponentanalysisalgorithmtoseparateeveryindependentsourcesignalwhentheoriginalsourcesarestatisticallyindependent，andthenusesthesinglefaultmethodtodiagnosesuccessfully．MethodMSD-ICA—EA(Multiresol

7、utionSubbandDecomposition-IndependentComponentAnalysisandEnvelopeAnalysis)ispresentedtorestrictedsignalswhichstatisticallycorrelativetoeachother．Directlyseparatetheindependentsub-bandanduseenvelopedetectiontodiagnosebearingfaultunderstrongpriorknowledge；