基于小波包和EMD相结合的电机轴承故障诊断

太原理工大学硕士学位论文基于小波包和EMD相结合的电机轴承故障诊断姓名：林选申请学位级别：硕士专业：@指导教师：田慕琴20100401IEMDIIEMDIMFEMDIIISTUDYONBEARINGFAULTDIAGNOSISOFASYNCHRONOUSMOTORSBASEDONWAVELETPACKETANDEMDABSTRACTMotorisakindofdrivingmachinerywhichiswidelyusedinmodernsociety.Withtheprogressofmodernizedscienceandtechnologyandconstantdevelopmentofproductionsystem,themotorisplayinganincreasinglyimportantrole.InductionMotorismostlyusedintheproductionandlives.Ithasmanyadvantages,suchassimplestructurelowcosthighreliabilitydurabilityandconvenientmaintenance.Itcannotonlydamagemotorsitselfbutalsoaffectthenormalworkofthewholeproductionsystemandevenendangerspeopleifmotorsdoesn’tworkandareinfaults.Furthermore,thiswillresultinhugeeconomiclossandbadsocialimpact.Thusitisurgentfordiagnosisofmotorfault.Almosthalfofthemotorfaultsarerelatedtothebearingfault,becausethebearingistheoneofelectricalcomponentswhichworksintheworstcondition.Itisusedtowithstandloadandtransferload.Asaresultofloading,installation,lubricationconditionandotherfactors,runningafteraperiodoftime,itwillproduceavarietyofdifferenttypesoffaults.Therefore,motorbearingisaIVrelativelyweaklink,anditsoperationalstatusoftendirectlyaffectstheperformanceoftheentiremachine.Inthispaper,theauthortookmotorbearingastheresearchobject,carriedoutananalysisofinnerfault,ballfaultandcombinedfailureofthetwo,andelaboratedthefaultmechanismofmotor.Theauthoralsoanalyzedthereasonwhichresultsinbreakdownandconcludedfrequencyperformanceofseveralkindsofcommonfaults.Duringthetrial,wecollectedvibrationsignalsofthemotorfaulttoanalyze.Inthefaultdetectiontechnology,signalfaultcharacteristicsofanalysisandextractionarethekeytothefaultdiagnosisanddirectrelationtotheaccuracyoffaultdiagnosis.Motorfaultsignalisunsteady.Fouriertransformcannoteffectivelyextractthecharacteristicsofthefaultmotor,Wavelettransformhasgoodtime-frequencylocalizationandthesignalcanbedividedinanyband,thispapertookthewaveletpackettransformtoextractfaultcharacteristicsinformationofsignal.What’smore,waveletpackethasagoodinhibitoryeffecttothenoisesignalandde-noisingpropertiesareveryobvious,soitwasprovedthatwaveletpackethasthemostsuperioreffecttothesignalde-noisingbycomparingthede-noisingeffectofthewaveletandwaveletpacket.Empiricalmodedecompositionisanewsignalprocessingmethodwhichbasedonthesignaloflocalcharacteristics.Usingit,theauthorcanobtainatime-seriesIntrinsicModeFunction,whichmakesinstantaneousfrequencymeaningful.ItisparticularlysuitablefortheanalysisandprocessingVofnon-linear,non-stationarysignal,andwecanacquirethesignalcharacteristicsofexpressionandinformation.Inthispaper,theauthorcombinestheadvantagesoftheboth,proposesanewmethodforbearingfaultdiagnosiswiththecombinationofwaveletandEMD.Thismethodcanhighlightthecharacteristicsofthedatageneratedbyvibrationsignalwiththemotorbearingfaultconditionandextractthem.ItovercomesthelimitationsofthefastFouriertransform.Furthermore,theauthordealswiththesignalsofinnerfault,ballfaultandcombinedfailureofthetwowhicharemadeduringthetrial,extractsthecharacteristicoffaultandcategorizesthefault.Throughthemethodwecansolvetheproblemoffaultdiagnosisofmotorbearingbetter.KEYWORDS:inductionmotor,bearingfault,waveletpacket,empiricalmodedecomposition11.1[6]:2[12]1.21.2.1MachineConditionDiagnosisTechniqueCDT12341.2.21-11-1Fig.1-1Theflowchartofequipmentdiagnosisprocess31.31.3.14;80%1.3.22060[3](FFT)[4][5][6]:2060(FFT):60SPM:1974D.R.Harting5[3]:2090[7]1.412,34562.1[8]1()0.5mm22()374.:;5.6.[9][10]7.2.22.2.12-1,82-1Fig.2-1Rollingstructurediagram:D:d:1r:2r:α:Z:2.2.2[4]:();[3]1.2.9:(1)(2)(3)cfcfcffnfc±(fn=1,2,3,)1kHz(4)3.4.:5.102.2.31[4]:ρ2424.0Erfb=2-1r-----mρ-----3/mkgE------2N/mMEIannnfr222112)1(+−=π2-2I------4ma------mM-------mkg/N-------20-60kHz112:(1)(2)(3)if(4)of(5)(cf)2-1::)cos(21αππdDffrViii−==2-3:)cos(22αππdDffrVooo+==2-4:DfVVVcoicπ=+=)(5.02-5()])cos1()cos1[(212oioicfDdfDdDVVfααπ++−=+=2-6:)cos1(21αDdfffffiocooc−−=−=2-7:)cos1(21αDdfffffoiciic+−=−=2-812/rd(bcf:)cos1(cos21ααDddDddDdrfficbc−=−==2.912])cos(1[212αDdffdDfoibc−−=2.10oirfff−=:ioirofffff=−==,02:ocZf:rocfDdZZf)cos1(5.0α−=2.11icZf:ricfDdZZf)cos1(5.0α+=2.12(bcf):rbcfDddDf])cos(1[22α−=2.13():rcfDdf)cos1(5.0α−=2.14ocZficZfbcf2-12-1Tab.2-1ThefaultcharacteristicfrequencyformulaofRollingbearing0.5(1cos/)oFZdDfα=−0.5(1cos/)iFZdDfα=+20.5[1(cos/D)]/bFDdfdα=−2.32.3.12-213DEWTTRON1616200kHz±0.059607.5Kw1432DEWETRON2010XYZ2-2Fig.2-2TheSensorarrangementandtestsystem7.5kW50Hz380V13.3A4710r/min2.3.22-21234XYZ1.2-22-2Tab.2-2Thepassagenumberoftestpoint1x-ch10y-ch2z-ch92x-15y-ch11z-ch13x-ch4y-ch5z-ch34x-ch6y-ch7z-ch8141x—10channely—2channelz—9channel2x—15channely—11channelz—1channel3x—4channely—5channelz—3channel4.x—6channely—7channelz—8channelA—12channelB—13channelC—14channel2.630982-36309Fig.2-3Sizechartofbearing6309D=72.5mmd=18.5mm2.3.315630935.3Hz,59.4Hz21.7Hz,2.463091680Y.MeyerJ.MorletA.GrossmanI.DaubechiesS.MallatFourier(Fourier)FFT[11][12][13][1415]3.13.1.1(WaveletAnalysis)2080Morlet[16](wavelet)0(motherwavelet))(tψτa()xt:dtattxaaWTx)(*)(1),(∫∞∞−−=τψτa0(3-1)17:ωωψωπτωωdeaXaaWTjx+∞∞−∫=)(*)(2),((3-2))(ωX