机械毕业设计英文外文翻译制造分析进程数使用快速标记技术Word文档下载推荐.docx

机械毕业设计英文外文翻译制造分析进程数使用快速标记技术Word文档下载推荐.docx

《机械毕业设计英文外文翻译制造分析进程数使用快速标记技术Word文档下载推荐.docx》由会员分享，可在线阅读，更多相关《机械毕业设计英文外文翻译制造分析进程数使用快速标记技术Word文档下载推荐.docx（22页珍藏版）》请在冰豆网上搜索。

Author：

D.Clifton

Reviewer：

S.Turner

TableofContents

1ExecutiveSummary4

1.1Introdution4

1.2TechniquesEmployed4

1.3SummaryofResults4

1.4Observations4

2Introdution6

2.1OxfordBioSignalsLimited6

3ExternalReferences7

4Glossary7

5DataDescription8

5.1Datatypes8

5.2PriorExperimentKnowledge8

5.3TestDescription8

6Pre-processing10

6.1RemovalofStart/StopTransients10

6.2RemovalofPowerSupplySignal10

6.3FrequencyTransformation10

7AnalysisI-Visualisation13

7.1VisualisationofHigh-DimensionalData13

7.2Visualising5-DManufacturingProcessData13

7.3AutomaticNoveltyDetection15

7.4ConclusionofAnalysisI-Visualisation16

8AnalysisII-SignatureAnalysis17

8.1ConstructingSignatures17

8.2VisualisingSignatures19

8.3ConclusionofAnalysisII-SignatureAnalysis23

9AnalysisIII-TemplateAnalysis24

9.1ConstructingaTemplateofNormality24

9.2ResultsofNoveltyDetectionUsingTemplateAnalysis25

9.3ConclusionofAnalysisIII-TemplateAnalysis26

10AnalysisIV-None-linearPrediction27

10.1NeuralNetworksforOn-LinePrediction27

10.2ResultsofNoveltyDetectionusingNon-linearPrediction27

10.3ConclusionofAnalysisIV-Non-linearPrediction28

11OverallConclusion29

11.1Methodology29

11.2SummaryofTesults29

11.3FutureWork29

12AppendixA-NeuroScaleVisualisations31

TableofFigures

Figure1-Test90.Fromtoptobottom:

Ax,Ay,Az,AE,SPagainsttimet（s）

Figure2-PowerspectraforTest19afterremovalof50Hzpowersupplycontribution.Thetopplotshowsa3-D“landspace”plotofeachspectrum.Thebottomplotshowsa“contour”plotofthesameinformation,withincreasingsignalpowershownasincreasingcolourfromblacktored

Figure3-PowerspectraforTest19afterremovalofallspectralcomponentsbeneathpowerthreshold

Figure4-Azagainsttime（inseconds）forTest19,beforeremovaloflow-powerfrequencycomponents

Figure5-Azagainsttime（inseconds）forTest19,afterremovaloflow-powerfrequencycomponents

Figure6-SPforanexampletest,showingthreeautomatically-detecrminedstates:

S1-drillingin（showningreen）;

S2-drill-bitbreak-throughandremoval（showninred）;

S3-retraction（showninblue）

Figure7-Examplesignatureofvariable

plottedagainstoperating-point

Figure8-Powerspectrafortest51,frequency（Hz）onthex-axisbetween[0fs/2]

Figure9-Averagesignificantfrequency

Figure10-VisualisationofAEsignaturesforalltests

Figure11-VisualisationofAxbroadbandsignaturesforalltests

Figure12-VisualisationofAxaverage-frequencysignaturesforalltests

Figure13-Noveltydetectionusingatemplatesignature

Figure14-

1

ExecutiveSummary

1.1Introduction

ThepurposeofthisinvestigationconductedbyOxfordBioSignalswastoexamineanddeterminethesuitabilityofitstechniquesinanalyzingdatafromanexamplemanufacturingprocess.ThisreporthasbeensubmittedtoRolls-RoycefortheexpressedofassessingOxfordBioSignals’techniqueswithrespecttomonitoringtheexampleprocess.

TheanalysisconductedbyOxfordBioSignals（OBS）waslimitedtoafixedtimescale,afixedsetofchallengedataforasingleprocess（asprovidedbyRolls-RoyceandAachenuniversityofTechnology）,withnopriordomainknowledge,norinformationofsystemfailure.

1.2TechniquesEmployed

OBSusedanumberofanalysistechniquesgiventhelimitedtimescales:

I-Visualisation,andClusterAnalysis

Thispowerfulmethodallowedtheevolutionofthesystemstate（fusingallavailabledatatypes）tobevisualisedthroughouttheseriesoftests.Thisshowedseveraldistinctmodesofoperationduringtheseries,highlightingmajoreventsobservedwithinthedata,latercorrelatedwithactualchangestothesystem’soperationbydomainexperts.

Clusteranalysisautomaticallydetectswhichoftheseeventsmaybeconsideredtobe“abnormal”,withrespecttopreviouslyobservedsystembehavior.

II-Signaturerepresentseachtestasasinglepointonaplot,allowingchangesbetweenteststobeeasilyidentified.Abnormaltestsareshownasoutlyingpoints,withnormaltestsformingacluster.

Modelingthenormalbehaviorofseveralfeaturesselectedfromtheprovideddata,thismethodshowedthatadvancewarningofsystemfailurecouldbeautomaticallydetectedusingthesefeatures,aswellashighlightingsignificanteventswithinthelifeofthesystem.

III-TemplateAnalysis

Thismethodallowsinstantaneoussample-by–samplenoveltydetection,suitableforon-lineimplementation.

UsingacomplementaryapproachtoSignatureAnalysis,thismethodalsomodelsnormalsystembehavior.Resultsconfirmedtheobservationmadeusingpreviousmethods.

IV-NeuralnetworkPredictor

Similarlyusefulforon-lineanalysis,thismethodusesanautomatedpredictorofsystembehaviour（aneuralnetworkpredictor）,inwhichpreviouslyidentifiedeventswereconfirmed,andfurthersignificantepisodesweredetected.

1.3SummaryofResults

Earlywarningofsystemfailurewasindependentlyidentifiedbythevariousanalysismethodsemployed.

Severalsignificanteventsduringthelifeoftheprocesswerecorrelatedwithactualknowneventslaterrevealedbysystemexperts.

Changesinsensorconfigurationsareidentified,andperiodsofsystemstability（inwhichtestsaresimilartooneanother）arehighlighted.

Thisreportshallbeusedasthebasisforfurthercorrelationofdetectedeventsagainstactualoccurrenceswithinthelifeofthesystem,tobeperformedbyAachenUniversityofTechnology.

1.4Observations

Basedonthislimitedstudy,OBSareconfidentthattheirtechniquesareapplicabletoconditionmonitoringoftheexamplemanufacturingprocessasfollows:

Evidenceshowsthatautomateddetectionofsystemnoveltyispossible,comparedtoits“normal”operation.

Earlywarningofsystemdistressmaybeprovided,givingadequatetimetotakepreventativemaintenanceactionssuchthatsystemfailuremaybeavoided.

Provision“fleet-wide”analysisispossibleusingthetechniquesconsideredwithinthisinvestigation.

TheinvolvementofdomainknowledgefromsystemexpertsalongsideOBSengineerswillbecrucialindevelopingfutureimplementations.Whilethis“blind”analysisshowedthatOBSmodellingtechniquesareappropriateforprocessmonitoring,itisthecouplingofdomainknowledgewithOBSmodellingtechniquesthatmayprovideoptimaldiagnosticandprognosticanalysis.

2

Introduction

2.1OxfordBioSignalsLimited

ThisdocumentreportsontheinitialanalysisconductedbyOxfordBioSignalsofmanufacturingprocesschallengedataprovidedbyRolls-Royce,inconjunctionwithAachenUniversityofTechnology（AUT）.

OxfordBioSignalsLimited（OBS）isaworld-classproviderofAcquisition,DataFusion,NeuralNetworksandotherAdvancedSignalProcessingtechniquesandsolutionsbrandedunderthecollectivenameQUICKTechnology.Thistechnologynotonlyprovidesforhealthandqualityassurancemonitoringoftheoperationalperformanceofequipmentandplant.

QUICKTechnologyhasbeenextensivelyproveninthefieldofgasturbinemonitoringwithbothon-lineandoff-lineimplementationsatmultiplelevels:

asaresearchtool,atestbedsystem,agroundsupporttool,anon-boardmonitoringsystem,anoff-lineanalysistoolanda“fleet”manager.

ManyofthetechniquesemployedbyOBSmaybedescribedasnoveltydetectionmethods.Thisapproachhasasignificantadvantageovermanytraditionalclassificationtechniquesinthatitisnotnecessarytoprovidefaultdatatothesystemduringdevelopment.Instead,providingasufficientlycomprehensivemodeloftheconditioncanbeidentifiedautomatically.Asinformationisdiscoveredregardingthecausesofthesedeviationsitisthenpossibletomovefromnoveltydetectiontodiagnosis,buttheabilitytoidentifypreviouslyunseenabnormalitiesisretainedatallstages.

3

ExternalReferences

Accompanyingdocumentationprovidingfurtherinformationonthedatasetsisavailableinunnumbereddocuments.

4Glossary

AUT-AachenUniversityofTechnology

GMM-GaussianMixtureModel

MLP-Multi-LayerPerception

OBS-OxfordBioSignalsLtd.

5DataDescription

Thefollowingsectionsgiveabriefoverviewofthedatasetobtainedbyvisualinspectionofthedata.

4.1Datatypes

Thedataprovidedwererecordedoveranumberoftests.Eachtestconsistedofasimilarprocedure,inwhichanautomateddrillunitmovedtowardsastaticmetallicdiskatafixedvelocity（“feed”）,aholewasdrilledinthediskatthatsamefeed-rate.

Thefollowingdatastreamswererecordedduringeachtest,eachsampledatarateof20KHz:

Ax–accelerationofthedisk-mountingunitinthex-plane1,

Ay-accelerationofthedisk-mountingunitinthey-plane1,

Az-accelerationofthedisk-mountingunitinthez-plane1,

AE-RMSacousticemission,50-400KHz2,

SP-powerdeliveredtothedrillspindle3.

Testsconsideredinthisinvestigationusedthreedrill-prices（ofidenticalproductspecification）asshowninTable1.

Table1-ExperimentParametersbyTest

DrillNumber

TestNumbers

DrillRotationRate

FeedRate

1

[12]

1700RPM

80mm/min

2

[3127]

3

[130194]

120mm/min

Notethattests16,54,128,129werenotprovided,thusaseriesof190testsareanalysedinthisinvestigation.These190testsarelabeledasshowninTable2.

Table2–Testindicesusedinthisreportagainstactualtestnumbers

TestIndices

ActualTestNumber

[115]

[1652]

[1753]

[53125]

[55127]