提取磨削主轴型转子轴承系统在加速度期间振动信号的特征研究外文文献翻译中英文翻译文档格式.docx
《提取磨削主轴型转子轴承系统在加速度期间振动信号的特征研究外文文献翻译中英文翻译文档格式.docx》由会员分享,可在线阅读,更多相关《提取磨削主轴型转子轴承系统在加速度期间振动信号的特征研究外文文献翻译中英文翻译文档格式.docx(9页珍藏版)》请在冰豆网上搜索。
InternationalConferenceonFrontiersofDesignandManufacturing
June19-22,2006,Guangzhou,China
Pages255-260
ASTUDYONVIBRATIONSIGNAL-BASEDFEATUREEXTRACTION
FORGRINDINGSPINDLE-TYPEDROTOR-BEARINGSYSTEM
DURINGACCELERATION
Jong-KweonPark,Bong-SukKim,Soo-HunLeeandJun-YeobSong
IntelligenceandPrecisionMachineryResearchDivision,KIMM,305343,Rep.ofKorea
SchoolofMechanicalEngineering,AjouUniversity,443749,Rep.ofKorea
Abstract:
Thegoalofsystemmonitoringistominimizeeconomicloss,toincreasereliability,tomaximizeproductivity,andtomaintainproductqualityinmanufacturing.Sincevibrationsignalssufficientlycontaintheabundant
runninginformationoftherealsystemandthehiddenfaultsymptoms,thefeatureextractionthroughthosesignalsiswidelyappliedforperformanceevaluationfaultdiagnosticsofrotatingmachineries.
Thispapershowsfeatureextractionfromvibrationsignalsgatheredinthegrindingspindle-typedrotor-bearingsystemduringaccelerationinordertomonitoranabnormalconditionofcurrentsystemlikeshaftcrackbyusingvariouskindsofsignalprocessingmethodssuchastheFastFourierTransform,Short-TimeFourierTransform,Wigner-VilleDistribution,andDiscreteWaveletTransform.Aswell,theresultoffeatureextractioninshaftcrackconditionwascomparedwiththatinnormalcondition.
Keywords:
Featureextraction,Grindingspindle-typedrotor-bearingsystem,Non-stationarysignalprocessingmethod,Acceleratingprocess,Wavelettransform
1.Introduction
Theconditionmonitoringorfaultdiagnosisinrotatingmachineriesandmachiningprocessisacrucialrequirementinordertomaintainreliability,safety,andproductqualityandtopreventfailuresordamages.Comparedwithothermachiningmethods,high-performancegrindingprocessisoneofthemostcomplicatedandimportantcuttingprocessesasfinalmachiningstage;
consequently,themonitoringofgrindingprocessandmachineismuchmorenecessaryinordertosupervisetheprocessandmachineandalsodetectabnormalities.Amongvariouskindsofapproaches,vibrationsignalanalysismethodforfeatureextractionandnondestructivedamageidentificationhasbeenwidelyutilizedduetocapabilitytocarrytheabundantdynamicinformationandtoindicatedetailedmotionofmechanicalsystemsandtodescribesimultaneouslywhenafaultoccursorwhatisitsfrequency.However,sincemostofthevibrationsignalssampledonmechanicalsystemsarenon-stationaryortransientsignalswhichsufficientlycontainadditionalinformationorabnormalsymptom,whichcannotberevealedfromstationarysignal,itisthekeyhowtoaccuratelydrawdominantfeaturecomponentsfromvibrationsignalsbecausenon-stationarysignalismorecomplexthanstationarysignal.Uptodate,forfeatureextractionofrotatingmachinery,manykindsofresearchresultshavemainlybeenfocusedonthestationarysignalprocess;
ontheotherhand,littleresearchhasbeenaccomplishedforthenon-stationarysignalprocesssuchasspeed-upprocess;
especially,thereisalmostnofeatureextractionusingvibrationsignalofspeed-upconditioninthefieldofgrindingprocess.
.Thispaperwasaboutastudytoextractthedominantfeaturesfromvibrationsignalsacquiredinalaboratorygrindingspindle-typedrotor-bearingsystemduringaccelerationbyusingseveralsignalprocessingmethodssuchasTimeDomainAnalysis(TDA),Frequency
DomainAnalysis(FDA),andtheTime-FrequencyAnalysisMethod(TFAM).Modaltesting,whichdetectsdynamiccharacteristicsofthesystemlikenaturalfrequency,wasperformedforthepurposeofdeterminingoperatingrangeforaccelerationintestsetup.Vibrationdatafromthebearinghousingpassingthroughthedistinctiveresonancefrequenciesandfrequencybandinspeed-upprocessweregatheredthroughtheexperimentswithnormalandcrackshaftcondition.Togetprominentsignalsofabnormalityfromacquiredtimedataasafundamentalstagefordiagnosisormonitoringtechnology,theFastFourierTransform(FFT),Short-TimeFourierTransform(STFT),Wigner-Villedistribution(WVD),andWaveletTransform(WT)usingcommercialsoftwarewerecarriedoutandcomparedwitheachresult
2.TheoreticalBackground
2.1.ReviewofSignalAnalysisMethods
Therearetwomajortypesofsignalinthefirstnaturaldivisioncategory:
thestationarysignalandnon-stationarysignal.Stationarysignalsareconstantintheirstatisticalparametersovertime.Moreover,stationarysignalsarefurtherdividedintodeterministicandrandomsignals.
Randomsignalsareunpredictableintheirfrequencycontentandtheiramplitudelevel,buttheystillhaverelativelyuniformstatisticalcharacteristicsovertime.
Ontheotherhand,non-stationarysignalsaredividedintoc