Variable-GRR-Training

VariablesGR&R:ProceduresandInterpretationWWP/SQEDellConfidentialSupplierQualityEngineeringDell-InternalUse-ConfidentialMeasurementSystemsEvaluation(MSE)Ameasurementsystemincludestheoperators,gauges(orequipment),andproceduresforthemeasurementprocessDellConfidential2AcompleteMSErequiresanalysisof:•Discrimination:Smallestscaleunitofmeasureforthegauge•Accuracy(Bias):Thedifferencebetweenanobservedaverageofmeasurementsandareferencevalue•Stability:Thechangeinbiasovertime(abilityofameasurementsystemtoproducethesamevaluesovertimewhenmeasuringthesamesample)•Linearity:Thechangeinbiasovertheoperatingrange•Precision(GRR):–Reproducibility:Variationintheaverageofmeasurementstakenbymultipleoperatorswhenusingasinglegaugetomeasureasinglecharacteristicofasinglepart(AppraiserVariability)–Repeatability:Variationinmeasurementsobtainedafterrepeateduseofsinglegaugebyasingleoperatoronasinglepart(Equipmentvariability)SupplierQualityEngineeringDell-InternalUse-ConfidentialGRRStudy:ObjectiveDellConfidential3ByconductingaGRRstudy,wehopetounderstandandlimitthecontributionofmeasurementsystemvariabilitytototalobservedvariability•We’reinterestedinactualvariability𝜎𝑂𝑏𝑠𝑒𝑟𝑣𝑒𝑑2=𝜎𝐴𝑐𝑡𝑢𝑎𝑙2+𝜎𝑀𝑆2𝜎𝑂𝑏𝑠𝑒𝑟𝑣𝑒𝑑2=𝜎𝐴𝑐𝑡𝑢𝑎𝑙2+𝜎𝑅𝑒𝑝𝑒𝑎𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦2+𝜎𝑅𝑒𝑝𝑟𝑜𝑑𝑢𝑐𝑖𝑏𝑖𝑙𝑖𝑡𝑦2(Note:Variancesareadditive;standarddeviationsarenot.)SupplierQualityEngineeringDell-InternalUse-ConfidentialGRRStudy:ProcedureDellConfidential4SetupoftheGRRstudyshouldbebasedonfactorsthatinclude:•Actualnumberofoperatorsinthemeasurementsystem•Actualnumberandtype(s)ofgaugesinthemeasurementsystem•Automaticmeasurementsystems;inmanycasesreproducibilitycanbesafelyignored•Criticalityofthemeasuredparameter.Conductingthestudy:•Appraisersshouldbeselectedatrandomfromthepoolofactualappraisers•Appraisersshouldbekeptunawareoftheirpreviousresultsonthesamepartandtheresultsarrivedatbyotherappraisers•Measurementsshouldbetakeninrandomorder•Measurementproceduresshouldbedefinedandfollowedduringthestudy•Partsmustnotbeselectedatrandom;theyshouldbeselectedfromproductionpartsandrepresenttheentirerangeofprocessvariability,spanthefulltolerancerange•ThemeasurementsystembeingassessedmustbeproperlycalibratedSupplierQualityEngineeringDell-InternalUse-ConfidentialGRRStudy:MetricsDellConfidential5AcceptabilityofGRRstudyresultsshouldbebasedontheappropriatemetricForProductControl•WherethecontroltechniqueisInspection(either100%orsampling),and•Theresultingmeasurementsarecomparedtospecificationstodeterminetheacceptabilityoftheproduct,use:ForProcessControl•Wherethecontroltechniqueisstatisticalprocesscontrol(SPC),and•Theresultingmeasurementsareusedtoassessprocessstabilityi.e.whetherornotspecialcausevariationispresentintheobserveddata,or•Themeasurementsarebeingusedtosupportprocessimprovementstudies,use:PrecisiontoTolerance(%PT):6.0𝜎𝑚𝑠(𝑈𝑆𝐿−𝐿𝑆𝐿)x100PrecisiontoTotalVariation(%TV):x100𝜎𝑚𝑠𝜎𝑇𝑜𝑡𝑎𝑙SupplierQualityEngineeringDell-InternalUse-ConfidentialGRR:VarianceModelsDellConfidential6Nested&Crossedmodels:-Whentwoormorefactorsarepresentinanexperimentaldesign,theymaybecrossedornesteddependingonthestructureofthemodel.Ingeneral,foratwofactormodel:-IfeachleveloffactorAoccurswitheachleveloffactorB,thetwofactorsarecrossed-IfeachleveloffactorBoccurswithinonlyoneleveloffactorA,thenfactorBisnestedwithinfactorAEx:FullycrossedDesign-3Operators,1gauge,10parts-Eachoperatorusesthesinglegaugetomeasureeachpart3timesEx:NestedthenCrossed-3Operators,3gauges,10parts-Eachoperatorusesasingle(different)gaugetomeasureeachpart3times21103456789CA1BCA2B2110345678913NestedCrossedSupplierQualityEngineeringDell-InternalUse-ConfidentialGRR:MethodologiesDellConfidential7Average&Range+Controlcharts-Assumptions:-Onlytwofactors(appraisersandparts)plusmeasurementsystemrepeatabilityaresignificant-Theeffectofvariabilitywithineachpartisnegligible-Therearenooperator-gaugeoroperator-partinteractions-Theprocedurewasdevelopedbeforeeveryonehadcomputers-Interactionsareignoredbecausethey’redifficulttomanagewithahandcalculator!-Almostalwaysusedinconjunctionwiththesamesimple,crossedmodel;verylimitedflexibility-Typicallyimplementedin(unverified)spreadsheettemplates-Shouldbediscontinuednow!21103456789CA1BOperatorPartGaugeSupplierQualityEngineeringDell-InternalUse-ConfidentialGRR:MethodologiesDellConfidential8AnalysisofVariance(ANOVA)+ControlCharts-RecommendedmethodforallGRRstudies-‘Dude–You’vegotaDell,ItsEasy!’-Canaccommodateawiderangeofmodels-InJMP,upto5factorscanbeincluded-Includesanalysisofallinteractioneffects(operator-gauge,operator-part,etc.)(Note:Interactioneffects;A*Betc.,areonlydefinedinmodelsthatincludecrossedfactors)SupplierQualityEngineeringDell-InternalUse-ConfidentialInClassGRRStudyDellConfidential9ExperimentDesign-Threeoperators-5parts(duetotimeconstraints)-Threereplicatemeasurementsbyeachoperatorusingeachgauge(45measurementsperoperator)-Ensuredataloggingprocessisestablishedandcontrolled-SharedatawithallstudentsviaspreadsheetforJMPanalysis-Gauges:-Two6-inchanalogcalipers(differentmodels)-One4-inchdigitalc