基于數(shù)據(jù)深度的變點(diǎn)識別方法研究.pdf

1、天津大學(xué)碩士學(xué)位論文基于數(shù)據(jù)深度的變點(diǎn)識別方法研究姓名：廖丹申請學(xué)位級別：碩士專業(yè)：管理科學(xué)與工程指導(dǎo)教師：聶斌201112ABSTRACTInindustrialproductionprocess，TheimportanceofstatisticalprocesscontroI(SPC)hasbeenwidelyrecognizedinqualitymamgemeracontrolmodelsinSPCisimportanttools

2、formonitoringwhetheramanufacturingprocessisstatisticallyincontrolornotWecandivideSPCintwostage：PhaseI，analysisthehistoricaldata，identifycorrectbaselinedatawhichmeetthequalityrequirements；PhaseII，Parameterestimatbnestabli

3、shcontrolchartsandcontinuerealtimemonitoringthecontroIchartswillalarmwhentheprocessdataisoutofcontrolthenweshouldanalysisthereasonandtakeappropriaten℃asurestoadjustInthestageofPhaseI。itisveryimportanttoidentifychangepoin

4、taccuratelyButmostofthetraditionalmethodisbasedontheassumptionofnorrmldistr酗utionOurnewmethodproposesanideaofchangepointidentificationbasedontheproximityofthedatacenterTheproposemethodtransferindividualobservationssequen

5、cesintopointsinmultidimensionalspace，andthevariableisconstructedbasedondatadepththenachangepointlocationruleisdevelopedTheproposedmethodistestedonseveralsimulationexperimentsfromaliterate，andcomparedwiththetraditionalmet

6、hod(LRTAHC)，itisprovedevenbettertoidentifythebaselineperiodwithoutthenorrmldismbutionassumptionFinallythispapergivesasummaryaboutthemethod’sadvantagesand1硼fl婦tionsdrawsomeusefulconchsbmandmakeprospectsforfuturedevelopmen