MSA測(cè)量系統(tǒng)分析報(bào)告

1、 Training Material forMEASUREMENTSYSTEMANALYSISContents:INTRODUCTIONFOR MEASUREMENT SYSTEMANALYSISGENERAL METHODSILLUSTRATION FORMEASUREMENTSYSTEMANALYSISVARIABLEGAGE ANALYSIS METHOD1)THE AVERAGE-RANGEMETHOD2)THE ANOVA METHODATTRIBUTEGAGEANALYSISMETHOD1)SHORTMETHOD2)HYPOTHESISTEST ANALYSIS3)SIGNAL D

2、ETECTIONTHEORY4)LONGMETHODACCEPTABILITY CRITERIACONCLUSIONFOUR METHODSCOMPARISONIntroduction: Basic requirementsbyQS-9000 &TS16949Base on QS9000& TS16949requirements,allmeasurementsystem which werementionedinQualityPlan shouldbeconductedMeasurementSystem Analysis.MSARequirementIntroduction: Thecateg

3、oryofMeasurementSystemMost industrialmeasurementsystem canbedivided twocategories,oneisvariablemeasurementsystem, anotherisattribute measurement system.Anattribute gagecannotindicatehow goodorhow bada partis,butonlyindicatesthatthepartisacceptedorrejected.The mostcommon of these is aGo/No-gogage.Var

4、iableGageAttributeGage(Go/No-goGage)Introduction: WhatisameasurementprocessOperationOutputInputGeneral ProcessMeasurementAnalysisValueDecisionProcess to be ManagedMeasurementProcessMeasurement:Theassignmentofa numericalvaluetomaterialthingstorepresent therelationsamongthemwith respecttoaparticular p

5、rocess.MeasurementProcess:Theprocessofassigningthe numericalvaluetomaterialthings.Introduction: Whatare thevariations of measurement processIntroduction: Whatare thevariations of measurement processMeasurement(Observed)Value=ActualValue+VarianceofTheMeasurementSystem2obs = 2 actual + variance of the

6、 measurement system 2Introduction: Where doesthe variationofmeasurementsystemcomefrom?TheFiveCharacterizationsofMeasurementSystem:1.LocationVariation: Bias; Stability;LinearityBiasisthedifferencebetween theobservedaverage of measurementsandareferencevalue.Bias is often referred to as accuracy.Itisa

7、systematicerrorcomponentofthe measurement systemIntroduction: Where doesthe variationofmeasurementsystemcomefrom?TheFiveCharacterizationsofMeasurementSystem:1.LocationVariation: Bias; Stability;LinearityStability(Alias: Drift):Stabilityisthe total variationinthemeasurementsobtainedwith ameasurements

8、ystem on thesame masterorpartswhen measuringa singlecharacteristicover an extended timeperiod.A stablemeasurementprocessisinstatisticalcontrol withrespecttolocation.StabilityIntroduction: Where doesthe variationofmeasurementsystemcomefrom?TheFiveCharacterizationsofMeasurementSystem:1.LocationVariati

9、on: Bias; Stability;LinearityLinearityisthedifferenceinthebiasvaluesthroughtheexpectedoperating rangeofthemeasurementinstrument.Itisasystematic error componentofthemeasurementsystem.LinearityIntroduction: Where doesthe variationofmeasurementsystemcomefrom?TheFiveCharacterizationsofMeasurementSystem:

10、2.WidthVariation: Repeatability; Reproducibility;GageR&RRepeatabilityisthevariation in measurementsobtainedwith onemeasurementinstrument whenusedseveral times while measuringtheidentical characteristiconthesamepart by an appraiser.ItisaWithin-system variation,commonlyreferredtoasE.V.-Equipment Varia

11、tion.RepeatabilityIntroduction: Where doesthe variationofmeasurementsystemcomefrom?TheFiveCharacterizationsofMeasurementSystem:2.WidthVariation: Repeatability; Reproducibility;GageR&RReproducibilityisthevariation in theaverage of themeasurementsmadebydifferentappraisersusingthesamegage whenmeasuring

12、 theidenticalcharacteristicsofthe samepart.Itisbetween-systemvariation,commonlyreferredtoasA.V.-Appraiser Variation.ReproducibilityIntroduction: Where doesthe variationofmeasurementsystemcomefrom?TheFiveCharacterizationsofMeasurementSystem:2.WidthVariation: Repeatability; Reproducibility;GageR&RGage

13、 R&RmeansGage repeatabilityand reproducibility,whichcombinedestimateofmeasurementsystemrepeatabilityandreproducibility.This combined measurement error theniscomparedwiththeprocessoutputvariabilitytocompute thegage percentageR&R(%R&R).The%R&Risthebasisformaking ajudgmentofwhetherthemeasurementsystem

14、is goodenoughtomeasure theprocess.AnalysisTechniques:Currentlythereare three techniquesfor variable measurement systemandfourtechniques forattributemeasurementsystem analysis wererecommendedbyAIAG MSAReferenceManual.RangeMethodAverage -RangeMethodANOVAShortMethodLong MethodHypothesis TestAnalysesSig

15、nalDetection TheoryFollowings aresomepracticalexamplestoillustratehowtoperformfourmethods respectively.VariableGageAttributeGageAnalysisTechniques:PreparationbeforeMSATheapproachtobeusedshouldbeplanned.Thenumber of appraisers, numberofsampleparts, andnumberofrepeatreadingsshouldbedeterminedinadvance

16、.Theappraisersshouldbeselectedformthosewhonormallyoperatetheinstrument.Thesample parts mustbeselectedfromtheprocessandrepresent itsentireoperatingrange.Theinstrumentmust haveadiscriminationthat allowsatleastone-tenthoftheexpectedprocessvariationofthe characteristic to be readdirectly.Themeasurementp

17、rocedure shouldbedefined in advancetoensure theconsistent measuringmethod.AnalysisTechniques:VariableGage AnalysisGeneral GageR&R Study:TheAverageandRangeMethodTheANOVAMethodThecommon stepfor conductingGageR&Rstudy:1.Verifycalibrationofmeasurementequipment to be studied.2.Obtainasampleofpartsthatrep

18、resentthe actualorexpectedrangeofprocessvariation.3.Addaconcealedmarktoeach identifying theunitsasnumbers 1through10.Itiscriticalthat youcanidentifywhichunitiswhich.Atthe sametimeitisdetrimentaliftheparticipants in thestudycantelloneunitfrom theother(may biastheirmeasurementshould theyrecall howitme

19、asuredpreviously).4.Request 3appraisers.Refertotheseappraisersasa A, B, andC appraisers.Ifthemeasurementwillbedone repetitivelysuch as in aproductionenvironment,itispreferable to usetheactual appraiserthat willbeperformingthemeasurement.Forextremecases,aminimum of twoappraisers canbeused,butthisisst

20、ronglydiscouragedasa lessaccurateestimateofmeasurementvariation willresult.5.Letappraiser Ameasure10partsina randomorderwhileyourecord thedatanotingthe concealedmarking.LetappraisersB andC measurethe same10partsNote:Donot allow theappraisers to witnesseachotherperforming themeasurement.Thereasonisth

21、e sameaswhy theunit markings areconcealed,TOPREVENT BIAS.6.Repeatthe measurementsforall three appraisers, butthis timepresentthesamples to eachinarandomorderdifferent fromthe original measurements.This is to again helpreduce biasinthe measurements.AnalysisTechniques:VariableGage Analysis10Parts3 App

22、raisers3 TrialsTheAverageandRangeMethod:A range controlchartiscreatedtodetermineifthe measurement processisstableandconsistent.Foreach appraisercalculatethe range of therepeatedmeasurementsforthe samepart.AnalysisTechniques:VariableGage AnalysisAnalysisTechniques:VariableGage AnalysisTheaveragerange

23、foreachoperatoristhen computed.Theaverageofthemeasurements taken by an operator is calculated.A controlchartofranges is created.Thecenterlinerepresents theaverage rangeforall operatorsinthestudy, while theupperandlowercontrollimitconstantsarebasedonthenumber of times eachoperatormeasuredeachpart (tr

24、ials).AnalysisTechniques:VariableGage AnalysisThecenterlineandcontrollimitsare graphedontoa controlchartand thecalculated rangesarethenplotted on thecontrol chart.The range controlchartisexaminedtodeterminemeasurementprocessstability.Ifanyofthe plottedrangesfall outsidethe controllimits themeasureme

25、ntprocess isnotstable, andfurtheranalysisshouldnot takeplace.However,itiscommontohavetheparticularoperatorre-measure theparticular processoutput again andusethatdata if it isin-control.AnalysisTechniques:VariableGage AnalysisRepeatability -Equipment Variation(E.V.)Theconstantd2*isbasedonthenumber of

26、 measurementsused to computetheindividualranges(n)ortrials, thenumberofpartsinthe study,andthe numberofdifferentconditionsunderstudy.The constant K1isbasedonthenumber of timesa partwas repeatedlymeasured(trials).Theequipment variationisoftencomparedtotheprocessoutputtoleranceorprocess outputvariatio

27、ntodetermine apercentequipmentvariation (%EV).AnalysisTechniques:VariableGage AnalysisReproducibility-AppraiserVariation(A.V.)Xdiffisthedifferencebetween thelargest averagereadingbyanoperatorandthesmallestaveragereading by an operator.The constant K2isbasedonthenumberofdifferentconditionsanalyzed.Th

28、e appraiservariationisoftencomparedtotheprocess outputtoleranceorprocessoutputvariation to determineapercentappraiservariation (%AV). AnalysisTechniques:VariableGage AnalysisRepeatability andReproducibility( GageR&R)Thegageerror(R&R)iscomparedtotheprocessoutputtolerance to estimate theprecisiontotol

29、erance ratio (P/Tratio).Thisisimportanttodetermine if themeasurementsystemcan discriminatebetween goodand badoutput.Thebasicinterestofstudyingthe measurement processistodetermine if themeasurementsystemiscapableofmeasuringaprocess outputcharacteristicwithitsown uniquevariability.ThisisknowasthePerce

30、ntR&R(P/Pratio,%R&R),andcalculatedasfollows:AnalysisTechniques:VariableGage AnalysisProcess or Total Variation:Iftheprocessoutputvariation (m) is notknown,the total variationcanbeestimatedusingthe datainthe study.Firstthe partvariation is determined:Rpistherangeofthe partaverages, while K3isa consta

31、nt based on thenumberofpartsinthestudy.Thetotalvariation (TV)isjustthesquare rootofthe sumofthesquaresofR&Randthe partvariationAnalysisTechniques:VariableGage AnalysisTheANOVAMethod:Aweaknesswith theAverage-Range methodofusingtherangetodeterminegage R&Risthat it doesnot consider thevariationintroduc

32、edinto ameasurementthrough theinteractionbetween differentconditions (appraiser) andthegage.Consequently, to accountfor thisvariation an analysis of variance method(ANOVA)isutilized.In addition,when thesamplesizeincreases, useoftherangetoestimatethevariation in notvery precise.Furthermore,withsoftwa

33、repackagesreadilyavailable, theANOVAmethodisaviablechoice.Thetotalvariation in an individualmeasurementequals:Theparttopart variationisestimatedbyp2; theoperatorvariationisestimated byo2; theinteractioneffectisestimated byop2; while repeatabilityisestimated byr2AnalysisTechniques:VariableGage Analys

34、isPart:Operator:Interaction:Repeatability:AnalysisTechniques:VariableGage AnalysisTotal:ThegageR&Rstatisticsarethencalculated as follows:MeasurementError:Part:Operator:Interaction:Reproducibility:Repeatability:MeasurementError:Total:AnalysisTechniques:VariableGage AnalysisAcceptability Criteria:Theg

35、agerepeatability andreproducibility: %R&R(P/Pratio:%totaloftotalvariance;P/T ration:% total of tolerance):Less than10%Outstanding10%to20%Capable20%to30%MarginallyCapableGreater than30%NOTCAPABLEForthe P/Pratioandthe P/Tratio,either or bothapproachescanbetakendependingonthe intended useofthemeasureme

36、ntsystem andthedesiresofthecustomer.Generally,Ifthemeasurementsystem is onlygoingtobeuse to inspectiftheproductmeetsthespecs, thenweshould usethe%R&Rbase on thetolerance(P/T ratio).If themeasurementsystemisgoingtobeuse forprocess optimization/characterizationanalysis,thenweshoulduse the%R&R baseonto

37、talvariation(P/P ratio).AnalysisTechniques:VariableGage AnalysisAcceptability Criteria:ForaGage deemedtobeINCAPABLEfor itsapplication. Theteam mustreviewthedesign of thegage to improveitsintendedapplicationanditsability tomeasure critical measurementscorrectly. Also, if are-calibrationisrequired, pl

38、easefollowcaliberation steps.Ifrepeatability is large compared to reproducibility,the reasonsmightbe:1)theinstrumentneedsmaintenance,the gageshould be redesigned2)thelocationforgagingneedstobeimproved3)thereisexcessivewithin-partvariation.Ifreproducibilityislargecomparedtorepeatability,thenthepossib

39、lecausescouldbe:1)inadequate training on thegage,2)calibrationsare noteffective,3)a fixturemay be neededtohelp usethegagemore consistently.AnalysisTechniques:VariableGage AnalysisTheMeasurementBias:Usinga certifiedsample, anda controlchartofrepeatedmeasurements,thebiasofa measurement processcan be d

40、etermined.Bias is thedifference betweenthe knownvalueandthe averageofrepeatedmeasurementofthe known sample.Biasissometimescalled accuracy.Process Variation= 6SigmaRangePercent Bias =BIASProcess VariationAnalysisTechniques:VariableGage AnalysisLinearity:Linearityofameasurementprocess is thedifference

41、 in thebias or precisionvaluesthrough theexpectedoperatingrangeofthe gauge.Toevaluatelinearity,a graphcomparingthe biasorprecision to theexpectedoperatingrangeiscreated.A problemwithlinearityexists if thegraphexhibitsdifferentbiasorprecisionfordifferentexpectedoperating ranges.Byusingthefollowing pr

42、ocedure,linearity canbedetermined.1)Selectfivepartswhosemeasurementscoverthe operatingrangeofthegage.2)Verifythe truemeasurements of eachpart.3)Have eachpartberandomlymeasured12timesonthegagebyoneoperator.4)Calculatethe partaverageandthe biasfor eachpart.5)Plot thebias andthereference values.6)Calcu

43、latethe linearregression linethatbest fitsthesepoints.AnalysisTechniques:VariableGage Analysis7)Calculatethe goodness of fitstatistic:AnalysisTechniques:VariableGage Analysis8)Determinelinearity andpercent linearity:Linearity=Slopex Processvariation(m)%Linearity =100linearity/Process VariationTheacc

44、eptabilitycriteriaofBias,LinearitydependonQualityControl Plan,characteristicbeingmeasuredandgagespeciality,suggestedcriteriaofESG isasfollowing:Under5%- acceptable5%to15%-maybeacceptablebasedupon importanceofapplication, costofmeasurementdevice, costofrepairs,etc.,Over 15%- Considerednot acceptable-

45、everyeffort shouldbemade to improvethesystemThestabilityisdetermined throughthe useofa controlchart. It is importanttonote that, whenusingcontrolcharts, onemust notonly watch forpointsthatfalloutside of thecontrol limits,but alsocareotherspecial cause signalssuchastrendsandcenterlinehugging.Guidelin

46、eforthe detectionofsuch signalscan be found inmany publicationsonSPC.AnalysisTechniques:AttributeGageStudyShortMethod:A Short Methodexample forbattery lengthgo/no-gogage study:TheShortmethodneedtobeconduct by selecting20partswhichhave beenmeasuredbya variable gageinadvance,someofthepartsare slightly

47、 below andabovebothspecification limits. Twoappraisers thenmeasureallpartstwicerandomly.MeasurementResulttable1AnalysisTechniques:AttributeGageStudyAcceptability criteria:Ifallmeasurementresults(fourperpart)agree, thegageisacceptable.Ifthemeasurementresultsdonotagree, thegage cannotbeaccepted,itmust

48、 be improved andre-evaluated.Conclusion:Because table 1listed measurement resultsare notwholeagreement,atpart 15#and17#,appraisersdecisionsare notagree.sothe batterylength gagecannot be usedand mustbeimprovedand re-evaluated.AnalysisTechniques:AttributeGageStudyHypothesis TestAnalysis:Shortmethodsho

49、uld knowthe variable referencevalueofsamples in advance.However,insome situationsitishardtorealize to getallsamplesvariablereferencevalue. So in thiscase,Hypothesistest analysis shall be appliedfor gagestudy.IIII TargetIIIIIUSLLSLAnalysisTechniques:AttributeGageStudyHypothesis testanalysisdependsonc

50、rosstabulation methodwhichneedstotakea randomsampleof50 parts fromthe presentprocessanduse 3appraiserswhomake 3measurements on eachpartanddecide if thepart is acceptableornot.Appraisers measurethe parts andifthepartiswithinlimits theygive“1”and if notthey give“0” andwritethoseresults in atable. In o

51、rder to eliminateanybiasproduced,the labeledsamplesaremixedbefore givingtoappraisers foridentificationineach trails.Followingtable2listedfiller gagemeasuring resultsfor thebattery welding gap:50Samples3 Appraisers3 TrialsTable2Filler gagemeasuring resultTable2Filler gagemeasuring resultTable2Filler

52、gagemeasuring resultAnalysisTechniques:AttributeGageStudyInordertodeterminethe level of agreementamongtheappraisers,weappliedCohens Kappawhichisused to assessinter-raterreliabilitywhen observingorotherwisecoding qualitative/categorical variables.Itcan measurethe agreementbetween theevaluationsoftwor

53、aters whenbotharerating thesame object.Step 1. Organize thescoreinto acontingencytable. Since thevariablebeingratedhastwo categories, thecontingencytablewill be a2*2 table:Table3AnalysisTechniques:AttributeGageStudyA*BCross-TabulationTable 3AnalysisTechniques:AttributeGageStudyStep 2. Computethe row

54、totals(sumacrossthe valuesonthesamerow) andcolumntotals of theobservedfrequencies.Step 3Computethe overalltotal(showinthe table 3).Asa computationalcheck,besure thatthe rowtotalsand thecolumntotals sumtothesamevaluefortheoverall total,andthe overalltotalmatchesthenumber of cases in theoriginaldata s

55、et.Step 4Computethetotalnumber of agreementsbysummingthevalues in thediagonalcellsofthetable.a=53+89=142Step 5Computetheexpectedfrequency forthenumber of agreementsthatwouldhave beenexpectedbychance foreach codingcategory.ef=21.6Repeatthe formulafor other cell, we gototherexpectedcount(showinthetabl

56、e3).rowtotal*coltotaloverall total59* 55150Step 6Computethesum of theexpectedfrequenciesofagreementbychance.ef= 21.6+57.6= 79.2Step 7ComputeKappaK =0.89Step 8EvaluateKappa- Ageneralrule of thumb is thatvalues of kappa greaterthan0.75 indicategood to excellentagreement; valuesless than0.4 indicate po

57、oragreement.Repeatabovestep,wecan gotfollowingkappameasuresfor theappraisers:Table4AnalysisTechniques:AttributeGageStudya-efN-ef142-79.2150-79.2Table4Usingthesamestepstocalculated thekappameasure to determinetheagreement of eachappraiser to thereferencedecision: Table 5Totalsummary on Table 6:Analys

58、isTechniques:AttributeGageStudyTable5AnalysisTechniques:AttributeGageStudyAnalysisTechniques:AttributeGageStudyTheAIAGMSAreference manualedition 3providesacceptabilitycriteriaforeach appraisersresults:Definition:FalseAlarm Thenumberoftimesofwhichthe operator (s)identifya goodsampleasabadone.Miss The

59、 numberoftimesofwhichtheoperators identify abad sampleasagood one.Number of correct decisionsTotal opportunities for a decisionEffectiveness = AnalysisTechniques:AttributeGageStudyNumber of False AlarmTotal opportunities for a decisionFalse Alarm Rate = Number of False AlarmTotal opportunities for a

60、 decisionMiss Rate = Sosummarizingallthe information of theexample withthistable:Table7AnalysisTechniques:AttributeGageStudyConclusion:Themeasurementsystem wasacceptable withappraiserB,marginalwith appraiserA,andunacceptable forC.Soweshalldetermineifthereisa misunderstandingwithappraiserCthat requir