StatisticalProcessControl(SPC)

®MaterialsQualityPCSTraining-Rev2Feb232001Module3:StatisticalProcessControl(SPC)Methodology®MaterialsQualityPCSTraining-Rev2Feb2320012PCSElementsCreateMeasurementPlanEstablishMonitor(SPC)ImplementResponseFlowChecklist(RFC)Element1Element2Element3®MaterialsQualityPCSTraining-Rev2Feb2320013Contents•Introduction–WhatisSPC–WhatisStability•WhatisaControlChart–HowtoSet-upaControlChart–TypeofControlChartsAvailable–HowtoCalculatetheControlLimits–SPCTrendRules–WhentoReviseControlLimits•ProcessCapabilityStudy–SpeclimitsVSControlLimits–StabilityVSCapability•ControlChartReduction/Elimination•SPCExpectations®MaterialsQualityPCSTraining-Rev2Feb2320014WhatisSPC?•Statistical–Anythingthatdealswiththecollection,analysis,interpretation&presentationofnumericaldata–Gaininginformationformakinginformeddecisions•Process–Combinationofmachines,tools,methods,materials&peopleemployedtoattainprocessspecification–Asimilarprocedure/eventthatishappeningrepetitively•Control–Tokeepsomethingwithinadesiredcondition–MakesomethingbehavethewaywewantittobehaveTheuseofstatisticaltechniquessuchascontrolchartstoanalyzeaprocess,takeappropriateactionstoachieve&maintainastableprocess,&improveprocesscapability.®MaterialsQualityPCSTraining-Rev2Feb2320015WhatisStability?•AprocessissaidtobeStableifithasthefollowingproperties:–Patternappearsrandom–Constantprocessmean–Uniformvariabilityovertime–Notrends,runs,shifts,erraticups&downs•Importantformanyreasons:–Increasedproductivityofengineering&manufacturingpersonnel–Predictable,repeatableresultswithinaspecifiedrange®MaterialsQualityPCSTraining-Rev2Feb2320016WhatisaControlChart?•Atrendchartwithcontrollimits•Graphicalrepresentationofprocessperformance,wheredataiscollectedatregulartimesequenceofproduction•Valuabletoolfordifferentiatingbetweencommoncauseandspecialcausevariation•Evaluatingwhetheraprocessisorisnotinastateofstatisticalcontrol•Itletsthedata‘talk’byitself&basisfordata-drivendecisions®MaterialsQualityPCSTraining-Rev2Feb2320017ControlLimitsAtypicalcontrolchartconsistsofthreelines:-UpperControlLimit(UCL)CenterLine(CL)LowerControlLimit(LCL)CL:Theaverage(measureoflocation)processperformancewhentheprocessisin-controlUCL&LCL:Therangeof‘usual’processperformancewhentheprocessisstable.Linesdrawn3standarddeviations(3sigma)oneachsideofthecenterline.®MaterialsQualityPCSTraining-Rev2Feb2320018ControlChartAssumptions•ProcessStability–Theprocessmustbeinstatisticalcontrol•Normality–TheunderlyingprocessdistributionisnormalNote:Iftheassumptionsarenotmet,thecontrollimitscalculatedaremisleading&donotaccuratelyindicate3sigmacontrollimits.Seeyoursitestatisticianforadviceoncalculationmethodswhenassumptionsareviolated.®MaterialsQualityPCSTraining-Rev2Feb2320019TestforControlChartAssumptions•ProcessStability(nooutliers)–Screenoutoutliersfromthedatabasebeforecomputingfinalcontrollimitsbyusingacontrolchart.Anypointbeyondeithercontrollimitisanoutlier.Reportnumberofoutliersscreened.•Normality–Plotanormalprobabilityplotofthedataoroverlayanormalcurveoverthehistogram.Normallydistributeddatawillroughlyfallonastraightline.–TestfornormalitybyusingShapiro-WilkWtestinJMP®MaterialsQualityPCSTraining-Rev2Feb23200110•Selectappropriatetypeofcontrolcharttobeused•Gatherdatatoestablishthecontrolchart.–Aminimumof30subgroupsisrequiredoveratimeframeasdeterminedbythesamplingplan.•PlotthedataintimeorderonaTrendChartHowtoSet-upaControlChart?®MaterialsQualityPCSTraining-Rev2Feb23200111•Computethecontrollimits&plotthemonthetrendchart•Outliersidentification&exclusion–ExcludetheOut-ofControl(OOC)pointsoroutliersforwhichthereareverified/confirmedspecialcausesfromthechart–Re-computethecontrollimits,excludingtheOOCpoints–Iftherearefewerthan30pointsremainingatanytime,collectmoredata.It’sveryimportantthatthecontrollimitsarecalculatedusingatleast30subgroups.HowtoSet-upaControlChart?Note:RefertoAppendixAforControlChartsforLimitedProduction,i.e.30subgroups.®MaterialsQualityPCSTraining-Rev2Feb23200112•Validatethecomputedcontrollimitsagainstdatacollectedbyre-plottingthecontrolchartwithdata&newcontrollimits–Dothelimitsdetectknownproblems?–Arethelimitstoosensitive?Wouldtheyflagproblemsyoudonotknowhowtoreactto?•Usethecontrollimitsestablishedtomonitorthecriticalparameteridentified•Foreachparameter,everymachineshouldhaveaseparatecontrolchartwithseparatelycomputedcontrollimitsHowtoSet-upaControlChart?®MaterialsQualityPCSTraining-Rev2Feb23200113ControlChartClassificationsClassificationsofcontrolchartsaredependingonthetypeofdata•Variablesdata–Acharacteristicmeasuredonacontinuousscaleresultinginanumericalvalue–Examples:VoidSize,BondPullStrength,Coplanarity,BallHeight,etc.•Attributesdata–Acharacteristicmeasuredby#ofconforming&non-conformingtoaspecification.Outputisclassifiedaspass/failoraccept/reject.–E.g.BrokenWire,LiftedBond,FM,Chipping,BentLead,etc.–Canbeexpressedintermsoffraction,percentage,countorDPM®MaterialsQualityPCSTraining-Rev2Feb23200114ControlChartWhentoUse(Guidelinesonly)?X-R-whensubgroupingofsamplesor(Mean-Range)Chartmeasurementsisapplicable-n10X-whensubgroupin