MSA测量系统分析培训教程

WONGKJ:MSATRAINING,APR05MEASUREMETSYSTEMANALYSISWONGKJ:MSATRAINING,APR05Title:MeasurementSystemAnalysisObjective:ToequipyouwithadetailedunderstandingonMeasurementSystemAnalysis(MSA)andthebasicstatisticaltoolsnecessarytoperformtheanalysisusingMinitabsoftware.Thisiscrucialtoensuringagoodmeasurementsystemisavailableandtobeusedforprocesscontrolandimprovementactivities.WONGKJ:MSATRAINING,APR05Thiscoursewillcover:1.What’sMSA?2.ElementsinMSA•Accuracy/Bias•Linearity(Regressionmethod)•Stability•GRR•CasestudiesinimprovingAccuracyandLinearity3.BasicIntroductiontoMinitabforMSAAnalysis4.VariableMeasurementSystem•NumbersofDistinctCategories(ndc)•AverageandRangeMethod•ANOVAmethod•ChartInterpretation(GRRContributionbyAppraisers,Parts,InteractionbetweenAppraisersandParts)•CasestudiesinimprovingGRRandndc5.DestructiveGRR(Nested)•CasestudiesinimprovingGRR,ndc.6.AttributeMeasurementSystem•LinkingAttributeMeasurementSystemwithOperator’scertification,RisktoCustomer(under-kill)andYield(over-kill)7.MeasurementProblemAnalysisWONGKJ:MSATRAINING,APR05ObservedTotalProcessvariationMeasurementInstrumentvariationAppraiser/UservariationBetweensamplevariationWithinsamplevariationMeasurementvariation(MS)Actualprocessvariation(Process)σ2processσ2MSσ2totalWONGKJ:MSATRAINING,APR05KeyAspectsofMeasurementSystem•Accuracy/bias•Stability•Linearity•GRR•Repeatability•Reproducibility•ndcWONGKJ:MSATRAINING,APR05WONGKJ:MSATRAINING,APR05ACCURACY/BIAS•Deviationoftheaveragemeasurementfromthetruevalue/refvalue.•MeasurethesystematicerrorofMS.•Possiblecausesincludepoorcalibration,worninstrument,damagedMaster.BIASMSAVERAGEREFVALUEWONGKJ:MSATRAINING,APR05Accuracy/BIAS•Obtainonesamplewithitsrefvalue•Measurethesamplen10timesandrecordeachresult•Calculatethebiasofeachindividualreading•UsingMinitab:•UseOnesampleTtest,•ifp0.05,acceptHo(nobiasissue)•ifp0.05,rejectHo(Biasissueexists)•UseRegressionModel,checkforRsq(adj)•ifRsq(adj)0.80,regressionmodelisgood/valid(biasissueexists)•ifRsq(adj)0.80,regressionmodelisinvalid(nobiasissues)•InAnova•ifp0.05,regressionmodelisinvalid(nobiasissues)•ifp0.05,regressionmodelisgood/valid(biasissueexists)BIASMSAVERAGEREFVALUEWONGKJ:MSATRAINING,APR05ZeroBiaslineObservedRefValuePositiveBiasObserved-refRefValueZeroBiaslineNegativeBiasZeroBiaslineObservedRefValueConstantBiasNonConstantBiasPositiveBiasObserved-refRefValueZeroBiaslineNegativeBiasWONGKJ:MSATRAINING,APR05BIAS-Example-refvalue=6,measured15times-plskeyinReading’sdataonlyMeasureReadingBias15.8-0.225.7-0.335.9-0.145.9-0.156.00.066.10.176.00.086.10.196.40.4106.30.3116.00.0126.10.1136.20.2145.6-0.4156.00.0WONGKJ:MSATRAINING,APR05MeasureBias16141210864200.40.30.20.10.0-0.1-0.2-0.3-0.4-0.5S0.210313R-Sq8.6%R-Sq(adj)1.6%FittedLinePlotBias=-0.1048+0.01393MeasureRegressionAnalysis:BiasversusMeasureTheregressionequationisBias=-0.1048+0.01393MeasureS=0.210313R-Sq=8.6%R-Sq(adj)=1.6%AnalysisofVarianceSourceDFSSMSFPRegression10.0543210.05432141.230.288Error130.5750120.0442317Total140.629333One-SampleT:BiasTestofmu=0vsnot=0VariableNMeanStDevSEMean95%CITPBias150.0066670.2120200.054743(-0.110746,0.124079)0.120.905BIASAnalysis–MinitabRegressionandOneSampleTTestWONGKJ:MSATRAINING,APR05WONGKJ:MSATRAINING,APR05STABILITY•Totalvariationinthemeasurementsonasinglecharacteristicofamaster/part,obtainedwiththesameMSandoveranextendedperiodoftime.•Todeterminewhetherthere’sanychangeinBIASovertime.•Possiblecausesincludeinstrumentaging,instrumentneedscalibrationorenvironmentaldrift.TIMEWONGKJ:MSATRAINING,APR05STABILITY•Obtainsampleandestablishitsrefvalue•Measurethesample3-5timesperiodically(daily/weekly)•Plotthedataonthexbar-Rchartorxbar-Schart•Establishcontrollimitsandevaluateforoutofcontrolorunstablecondition(WECOrules,outliers,+/-3sigma)STABLEWHENTHEREISNOOBVIOUSSPECIALCAUSEIDENTIFIEDTIMEWONGKJ:MSATRAINING,APR05DayDensity4030201002.5102.5052.5002.4952.4904030201002.5102.5052.5002.4952.490123ScatterplotofDensityvsDayPanelvariable:OperatorThemeasurementbyeachappraiserspreadaboutthestddensityof2.5.TheredoesnotappeartobeaneffectofoperatorSampleSampleMean454035302520151052.5102.5052.5002.4952.490__X=2.5UCL=2.50949LCL=2.49051SampleSampleRange454035302520151050.0240.0180.0120.0060.000_R=0.00927UCL=0.02387LCL=0Xbar-RChartofDensityTheXbarchartindicatesthattheMSisstableandontargetscatteredrandomlyabouthistoricalmeanofmeasurementbyeachappraiserspread2.5Apieceoffilmwithaknowndensityof2.5unit,wasmeasured3timeseachmorningfor46days.Thedensitometerwascalibratedonlyonceatthebeginningoftheexperiment.Themeasurementistakenby3differentappraisers,Stability–CaseStudy1WONGKJ:MSATRAINING,APR05Youhaveadevicedesignedtomeasure%ofacertainchemicalcontainedinasmallplasticblock.Youdecidedtoconductagagestabilityexperimenttoevaluatethestabilityofthegage.Thedeviceobtainsameasurementbyevaporatingthechemical,soasingleblockcanonlybeusedonce.Youmustassumethatallthepartsmeasuredfortheexperimentarefairlyidentical.Tokeepthevariabilitybetweenpartstoaminimum,youselect26blocksfromthesamebatch.Theblocksaredesignedtocontain3.4%ofthechemical.Eachday,th