基于小波包系数与隐马尔科夫模型的刀具磨损监测.pdf

《基于小波包系数与隐马尔科夫模型的刀具磨损监测.pdf》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、机床与液压Jun．2014HydromechatronicsEngineeringVo1．42No．12DOI：10．3969／j．issn．1001—3881．2014．12．008Tool1wemarmonmitorin~basedonwaveletpackKetcoefHfiicilentandohiddenMarkovmodelYingQIU，Feng—yunXIESchoolofMechanicalandElectronicalEngineering，EastChinaJiaotongUniversity

2、，Nanchang330013，ChinaAbstract：Inordertopreventtoolfailuresduringtheautomationmachiningprocess，thetoolwearmonitoringbecomesveryimpo~ant．However．thestaterecognitionofthetoolwearisnotaneasytask．Inthispaper，anapproachbasedonwaveletpacketcoeficientandhiddenMarkovmo

3、del(HMM)fortoolwearmonitoringisproposed．Therootmeansquare(RMS)ofthewave-letpacketcoeficientsatdiferentscalesaretakenasthefeatureobsewationsvector．TheaD-proachofHMMpa~ernrecognitionisusedtorecognizethestatesoftoolwear．Theexperimentalresultshaveshownthatthepropo

4、sedmethodhasagoodrecognitionperformance．Keywords：Too1wear，Waveletpacketcoeficient，HiddenMarkovmodeI，RootmeansquareToolwearmonitoringiscrucialinordertopre-taforthelearningalgorithm．Otherwise，itwillre—venttoolfailuresduringtheautomationmachiningducetherecognitio

5、nrateofthetoolwear．process．However，theon-linetoolwearmonitoringisInthispaper，anapproachbasedonwaveletnotaneasytaskduetothecomplexityoftheprocess．packetcoeficientandHMMfortoolwearmonitoringFormanyyears，lotsofscholarshavestudiedtoolisproposed．Inordertomonitorthe

6、toolwearstateswearmonitoringbyvariousmethods．Thereareim-inmachiningprocess，thedynamometerisusedforportantcontributionspresentedforconditionmonito·dataacquisition．Thewaveletpacketdecompositionisring，forinstance，on—linetoolmonitoringbyusingadoptedfordataprocessi

7、ng．TherootmeansquareArtificialintelligencewaspresentedbyVallejo[1]，a(RMS)ofthewaveletpacketcoeficientsatdifferentmethodofstaterecognitionsbasedonwaveletandscalesaretakenasthefeatureobservationsvector．hiddenMarkovmodel(HMM)waspresentedbyXieTheHMMisusedtorecogni

8、zethestatesoftoolwear．[2]．On—lineconditionmonitoringbasedonempiri—Theresultsshowthattheproposedmethodhasarela—calmodedecompositionandneuralnetworkwaspro—tivelyhighrecognitionrate．p