GR&R training

GaugeRepeatability&ReproducibilityStudyPreparedby:QA/ChujunZhu(Ext:8110)Dec,2004Contents•Whytodomeasurementsystemanalysis.•Measurementsystemasprocessview•DefinitionofTerms•PDSAcyclesforGR&R•GR&Rforclassificationdata•QuestionsinpracticeMeasurementsystemAnalysis---Why???•Measurementdataareusedmoreoftenandinmorewaysthaneverbefore.–Product/Processmonitorandcontrol–Product/Processimprovement.•Wemustmakesurethedataisaccuratebeforeweuseit.•“…statisticalcontroloftheprocessofmeasurementisvital;otherwisethereisnomeaningfulmeasurement.”Dr.W.E.Deming(1986,p.332)MeasurementasaProcessObservedVariationinQualityCharacteristicMaterialsMethodsPeopleManagementEquipmentMeasurementMaterialsMethodsPeopleObservedVariationinQualityCharacteristicManagementEquipmentVariationinProductorProcessMeasurementProcessDefinitionofTerms•Bias(Accuracy):Differencebetweenaverageofindividualmeasurementsandacceptedstandardvalue.•Precision:Thedegreeofvariationinindividualmeasurements.–Repeatability–ReproducibilityUnderstandingMeasurementVariationMaterialsMethodsPeopleObservedVariationinQualityCharacteristicManagementEquipmentVariationinProductorProcessMeasurementProcess1.Bias2.Precision(σprocess)²+(σmeasurement)²=(σobserved)²^^^FromRepeatMeasuresUnknownWhatweobserveGR&RDefinitionRepeatability:Variationinmeasurementsobtainedwithonemeasurementsystemwhenusedseveraltimesbyoneappraiser(operator)whilemeasuringtheidenticalcharacteristiconthesamepart.Reproducibility:Variationintheaverageofthemeasurementsmadebydifferentappraisers(operators)usingthesamemeasuringsystemwhenmeasuringtheidenticalcharacteristiconthesamepart.Sample:Arepresentativegroupselectedfromapopulation.Thesampleisusedtodeterminethepropertiesofthepopulation.StandardDeviation(s):Ameasureofthespreadoftheprocessoutput.Tolerance:(UpperSpecificationLimit–LowerSpecificationLimit):Thepermittedrangeofvariationaboutanominalvalue.Thepermittedtoleranceisthedifferencebetweentheupperandthelowerspecificationlimit.Variation:Theinevitabledifferenceamongindividualoutputsofaprocess.StudyVariation(%StudyVar”inMinitab)isforcontrollingaprocessandisequalto100*(R&R/TotalStudyVariation)%ToleranceVariation(%ToleranceinMinitab)isfortestingtheproducttocustomerspecificationandisequalto100*(R&R/Tolerance)Discrimination:Themeasurementsystemsabilitytodetectandcontinuallyindicatethechangesinthemeasurementcharacteristic.ifthemeasurementsystemdoesnothaveadequatediscrimination,itmaynotbeanappropriatesystemtoidentifytheprocessvariationorquantifypartcharacteristics.TwoMethodsforGR&RStudyANOVAMethod:•ismostaccurate,givesRepeatability&Reproducibilityandpossibleinteractions.•Thecalculationsaremorecomplicated,butarebesthandledbysoftware(Minitab)X-BARRMethod:•BreaksdownRepeatability&Reproducibility,butdoesnotincludethepossibleinteractionbetweentheoperatorandtheparts..•CalculationFormulaRepeatability---EquipmentVariation:EV=5.15*Rbarbar*k1Rbarbar=(RbarA+RbarB+RbarC)/3Reproducibility---AppraiserVariation:AV=Sqrt(sq(Xbardiff*K2)-sq(EV)/(nr))where:n=parts,r=trialsTotalGRR=SQRT(SQ(EV)+SQ(AV))TotalTV=Rbar(observation)*K3%GRR=GRR/TV*100%or%GRR=5.15*GRR/ToleranceWhatisthedifferencebetweenthe2methods?TrialsK1Partsk320.886220.707130.590830.5231AppraisersK240.446720.707150.40330.523160.3742OverallVariationOverallvariationParttoPartVariationMeasurementSystemVariationVariationduetogageVarationduetooperatorsOperatorOperatorbyPartReproducibility+Repeatabiltiy+Bias+Discrimination+linearity+...VarationwithinpartProcessVariationAssumption:0GR&RStudyPDSACycle---Planning11/14/2002PLANGR&RHavecalibratedDeterminenumberofoperatorsIsGaugecalibrated?DeterminenumberofrepeatsperoperatorDeterminenumberofpartsReviewtestingprocedure&criteriaNoProcedure&criteriaareclear&understood?YesAuthoritytoclarifyprocedureandinstructoperatorsNoWillrepeatinganypartsinprocedureaffecttheresults?EvaluateandblockforeffectYesObtainpartsfortesting(coveringtestrange)SuitablyprotectpartsforrepeateduseModifypartsasrequiredforrepeatedtestingWatchprocedurebeingdonebyoperatorstoensureconsistencyYesAlloperatorsconsistentlyfollowingprocedure?NoVerifydatagatheringScheduletimefortestingProceedtoDOStartPLANLabelparts(blindtooperator,yetsystematicforanalysis)YESNOGR&RPDSACycle---PlanningTheplaninclude:•Numberofsamples:5-10•Numberofoperators:2-3•Numberofrepeats:Usually2to3timeseachpart,eachoperator•Ensurethenumberofsamplestimesthenumberofoperators15•Ensurethegaugeiscalibrated•Ensurethetestingprocessisclear•Ensuretheoperatoriscertified.•Ensurethetestingprocessisstable.Sampleselection:•Mustbefromtheprocesstobestudied.•Covertheoperatingrange,theminimumiscovering80%oftheoperatingrange.•Ifdatadistributionisavailable,examinefortypicalrangeandchooserepresentativeparts,fromaboveandbelowspecifappropriate.Tolerancedefinition:Definethetoleranceforunilaterallimitspec:a)usingdatadistribution;b)usingdesigndata3)…GR&RPDSACycle---Doing11/14/2002DOGR&ROperator1testseachpart(repeat1)StartDoOperator1testseachpart(repeat2)Operator1testseachpart(repeat3ifrequired)Randomizeparts&labelsRandomizeparts&labelsOperator2t