搜索
®MaterialsQualityPCSTraining-Rev3May212001Module3:StatisticalProcessControl(SPC)Methodology®MaterialsQualityPCSTraining-Rev3May2120012PCSElementsCreateMeasurementPlanEstablishMonitor(SPC)ImplementResponseFlowChecklist(RFC)Element1Element2Element3®MaterialsQualityPCSTraining-Rev3May2120013Contents•Introduction簡介–WhatisSPC什麼是SPC?–WhatisStability什麼是穩定性?•WhatisaControlChart什麼是管制圖–HowtoSet-upaControlChart如何建立管制圖–TypeofControlChartsAvailable管制圖的種類–HowtoCalculatetheControlLimits如何計算管制界限–SPCTrendRulesSPC法則–WhentoReviseControlLimits何時重新計算管制界限•ProcessCapabilityStudy制程能力研討–SpeclimitsVSControlLimits規格界限vs.管制界限–StabilityVSCapability穩定性vs.能力•ControlChartReduction/Elimination減少管制圖•SPCExpectations®MaterialsQualityPCSTraining-Rev3May2120014WhatisSPC?•Statistical統計–Anythingthatdealswiththecollection,analysis,interpretation&presentationofnumericaldata–關於數據資料的收集,分析,解釋與表現–Gaininginformationformakinginformeddecisions–取得資訊來作有效的決定•Process制程–Combinationofmachines,tools,methods,materials&peopleemployedtoattainprocessspecification–結合機器,治工具,方法,材料與人員來達到制程規格–Asimilarprocedure/eventthatishappeningrepetitively–重覆發生的事件/類似程序•Control管制–Tokeepsomethingwithinadesiredcondition–使某事/物保持在想要的情況–Makesomethingbehavethewaywewantittobehave–使某事/物依我們所想的來執行Theuseofstatisticaltechniquessuchascontrolchartstoanalyzeaprocess,takeappropriateactionstoachieve&maintainastableprocess,&improveprocesscapability.®MaterialsQualityPCSTraining-Rev3May2120015WhatisStability?•AprocessissaidtobeStableifithasthefollowingproperties:•下列特性稱為穩定:–Patternappearsrandom隨機出現–Constantprocessmean平均值一定–Uniformvariabilityovertime變異程度不隨時間改變–Notrends,runs,shifts,erraticups&downs不會偏向一邊•Importantformanyreasons:•穩定性為何重要?–Increasedproductivityofengineering&manufacturingpersonnel–提高生產性–Predictable,repeatableresultswithinaspecifiedrange–結果有重覆性,可預測®MaterialsQualityPCSTraining-Rev3May2120016WhatisaControlChart?•Atrendchartwithcontrollimits•有管制界限的趨勢圖•Graphicalrepresentationofprocessperformance,wheredataiscollectedatregulartimesequenceofproduction•數據依生產順序定時間收集,以圖表表現制程性能•Valuabletoolfordifferentiatingbetweencommoncauseandspecialcausevariation•將一般變異與特殊變異區分開的有用工具•Evaluatingwhetheraprocessisorisnotinastateofstatisticalcontrol•評估制程是否在統計管制中•Itletsthedata‘talk’byitself&basisfordata-drivendecisions•讓數據說話並依數據導向作決定®MaterialsQualityPCSTraining-Rev3May2120017ControlLimitsAtypicalcontrolchartconsistsofthreelines:典型管制圖有三條線:UpperControlLimit(UCL)CenterLine(CL)LowerControlLimit(LCL)CL:Theaverage(measureoflocation)processperformancewhentheprocessisin-controlCL:制程的平均性能UCL&LCL:Therangeof‘usual’processperformancewhentheprocessisstable.Linesdrawn3standarddeviations(3sigma)oneachsideofthecenterline.UCL&LCL:制程穩定情況下,制程性能的範圍®MaterialsQualityPCSTraining-Rev3May2120018ControlChartAssumptions•ProcessStability•制程穩定–Theprocessmustbeinstatisticalcontrol•Normality•常態分布–TheunderlyingprocessdistributionisnormalNote:Iftheassumptionsarenotmet,thecontrollimitscalculatedaremisleading&donotaccuratelyindicate3sigmacontrollimits.Seeyoursitestatisticianforadviceoncalculationmethodswhenassumptionsareviolated.若假設不成立,則管制界限將沒有意義®MaterialsQualityPCSTraining-Rev3May2120019TestforControlChartAssumptions假設•ProcessStability(nooutliers)穩定性–Screenoutoutliersfromthedatabasebeforecomputingfinalcontrollimitsbyusingacontrolchart.Anypointbeyondeithercontrollimitisanoutlier.Reportnumberofoutliersscreened.-計算管制界限前,將超出點排除.所有超出管制界限的點都是outlier•Normality常態性–Plotanormalprobabilityplotofthedataoroverlayanormalcurveoverthehistogram.Normallydistributeddatawillroughlyfallonastraightline.–TestfornormalitybyusingShapiro-WilkWtestinJMP–用JMPWtest來計算常態性®MaterialsQualityPCSTraining-Rev3May21200110•Selectappropriatetypeofcontrolcharttobeused•選擇合適的管制圖型態•Gatherdatatoestablishthecontrolchart.•收集數據建立管制圖–Aminimumof30subgroupsisrequiredoveratimeframeasdeterminedbythesamplingplan.–抽樣計劃至少收集30組數據•PlotthedataintimeorderonaTrendChart•依序在趨勢圖上描點HowtoSet-upaControlChart?(I)®MaterialsQualityPCSTraining-Rev3May21200111•Computethecontrollimits&plotthemonthetrendchart•計算管制界線並畫在圖上•Outliersidentification&exclusion•超出點的確認與排除–ExcludetheOut-ofControl(OOC)pointsoroutliersforwhichthereareverified/confirmedspecialcausesfromthechart–由於顯示是特殊原因造成故將排除超出點–Re-computethecontrollimits,excludingtheOOCpoints–重新計算管制界限–Iftherearefewerthan30pointsremainingatanytime,collectmoredata.It’sveryimportantthatthecontrollimitsarecalculatedusingatleast30subgroups.–若資料點少於30再繼續收集.這是很重要的HowtoSet-upaControlChart?(II)Note:RefertoAppendixAforControlChartsforLimitedProduction,i.e.30subgroups.®MaterialsQualityPCSTraining-Rev3May21200112•Validatethecomputedcontrollimitsagainstdatacollectedbyre-plottingthecontrolchartwithdata&newcontrollimits–Dothelimitsdetectknownproblems?–界限可以查覺已知的問題嗎?–Arethelimitstoosensitive?Wouldtheyflagproblemsyoudonotknowhowtoreactto?–界限是否太敏感?是否有問題你不曉得如何處理?•Usethecontrollimitsestablishedtomonitorthecriticalparameteridentified•使用建立的管制圖來追蹤確認重要參數•Foreachparameter,everymachineshouldhaveaseparatecontrolchartwithseparatelycomputedcontrollimits•對每一參數,每台機器應有一獨立的管制圖與管制界限HowtoSet-upaControlChart?(III)®MaterialsQuali