測量系統(tǒng)分析-劉曉梅

1、10/17/2022ATL Confidential Information測量系統(tǒng)分析MSA- Measurement System AnalysisPrepare by Liu XM2/8/2020ATLConfidential InformationOutline1.測量是一一個過程程2.測量過程程的變異異3.測量系統(tǒng)統(tǒng)的特性性4.GRR分析5.測量系統(tǒng)統(tǒng)變異性性的影響響Exercise測量是一一個過程程測量測量工具測量方法測試人測量對象正確的數(shù)據(jù)不可控因素樣品的真實值樣品的測量值測量測量是一一個過程程得到的所所有數(shù)據(jù)據(jù)都是測測量值，真實值值只有上上帝知道道測量數(shù)據(jù)據(jù)的變異異有一部部分是

2、測測試造成成的產(chǎn)品的變異（真實值的變異）測量變異總體變異(觀測到的變異)測量過程程的變異異-舉例研究兩個個工序的的能力：根據(jù)上述述數(shù)據(jù)：兩個工工序生產(chǎn)產(chǎn)的產(chǎn)品品是相似似的（相相同的分分布）進(jìn)一步研研究？Line 2Line 1s2 觀測 (相等)測量過程程的變異異-舉例問題：1.這兩條拉拉相同嗎嗎？2.哪個測量量系統(tǒng)對對于發(fā)現(xiàn)現(xiàn)工序變變異有幫幫助？3.對于這兩兩條拉，下一步步改善措措施是什什么？Line 1Line 2總體變異（觀測）測量變異工序變異s2true process + s2MS = s2observed2/8/2020ATLConfidential Information識別過程

3、程以及測測量系統(tǒng)統(tǒng)的變異異，首先先要定義義和區(qū)分分變異的的來源變異的來來源Long-termProcess VariationShort-termProcess VariationVariationwithin samplesActual Process VariationBiasStabilityLinearityVariation dueMeasurement VariationObserved Process VariationRepeatabilityReproducibilityto instrument測量系統(tǒng)統(tǒng)特性測量的系統(tǒng)偏差通過校正解決測量系統(tǒng)統(tǒng)特性分分類-準(zhǔn)確度和和精確度度

4、準(zhǔn)確度精確度ssstotalproductmeasurement222=+totalproductmeasurement測量測系統(tǒng)誤差通過G R&R解決均值方差測量系統(tǒng)統(tǒng)的統(tǒng)計特性準(zhǔn)確度Bias偏倚Linearity線性Stability穩(wěn)定性精確度Repeatability重復(fù)性Reproducibility再現(xiàn)性2/8/2020ATLConfidential Information準(zhǔn)確度(Accuracy)BiasStabilityLinearity穩(wěn)定性和和線性都都是偏倚倚的形式式穩(wěn)定性是是時間維維度上的的偏倚線性是量量程維度度上的偏偏倚2/8/2020ATLConfidential I

5、nformation測量系統(tǒng)統(tǒng)的總變變差對于重復(fù)復(fù)測量設(shè)設(shè)備總變變差的測測量精確度等等于重復(fù)復(fù)性和再再現(xiàn)性的的累加精確度(Precision)2/8/2020ATLConfidential Information測量系統(tǒng)統(tǒng)的5個特性：寬度偏差差：重復(fù)復(fù)性；再再現(xiàn)性；Gage R&R重復(fù)性：由同一評評價人，采用同同一種測測量儀器器，多次次測量同同一零件件的同一一特性時時獲得的的測量值值的變差差。它屬于于系統(tǒng)內(nèi)內(nèi)變差，通常視為為設(shè)備變變差（E.V.-EquipmentVariation）。Repeatability精確度：重復(fù)性性(Repeatability)2/8/2020ATLConfiden

6、tial Information重復(fù)性反應(yīng)測量量裝置的的固有變變異對同樣的的變量在在相似條條件下重重復(fù)測量量時的變變異同樣的部部件同一檢驗驗員同樣的設(shè)設(shè)置相同的環(huán)環(huán)境條件件短期根據(jù)重復(fù)復(fù)測量的的平均標(biāo)標(biāo)準(zhǔn)偏差差評估 s reproducibilityrepeatabilityproducttotal 2 2 2 2 sss + +=精確度：重復(fù)性性(Repeatability)2/8/2020ATLConfidential InformationInspector AMaster ValueInspector BInspector CMachine AMachine BMachine C再現(xiàn)性

7、:由不同測測量人員員用同樣樣儀器對對同一零零件的同同一特性性所得重重復(fù)測量量得平均均值的變變異稱為為測量系系統(tǒng)的再再現(xiàn)性，簡稱再再現(xiàn)性精確度：再現(xiàn)性性(Reproducibility)2/8/2020ATLConfidential Informationsss s 2 2 2 2totalproductrepeatabilityreproducibility=+ +再現(xiàn)性:反映了不不同條件件下同一一測量時時的變異異不同檢驗驗員不同的設(shè)設(shè)置不同的測測試單位位不同的環(huán)環(huán)境條件件長期的測測量變異異根據(jù)不同同測量條條件下的的平均標(biāo)標(biāo)準(zhǔn)偏差差評估精確度：再現(xiàn)性性(Reproducibility)2/8/

8、2020ATLConfidential Information測量系統(tǒng)統(tǒng)變差:s2meas=s2repeat+s2reprod主要根據(jù)據(jù)Gage R&R的分析結(jié)結(jié)果確定測量量系統(tǒng)是是“好”是“壞壞”，能能否滿足足應(yīng)用的的要求，需要對對測量系系統(tǒng)的變變異與產(chǎn)產(chǎn)品的spec與過程的的變異相相比較Comparingsmeaswith Tolerance:-Precision-to-ToleranceRatio(P/T)Comparingsmeaswith ProcessVariation (P/TV):-%RepeatabilityandReproducibility (%R&R)-Discrimi

9、nation Index計量型測測量系統(tǒng)統(tǒng)分析方方法2/8/2020ATLConfidential InformationPrecisiontoTolerance Ratio說明容差差的多少少百分比比的總散散布是由由測量誤誤差所導(dǎo)導(dǎo)致Best case: 10%Acceptable:10%30%包括重復(fù)復(fù)性和再再現(xiàn)性5.15標(biāo)準(zhǔn)偏差差占測量量系統(tǒng)散散布的996標(biāo)準(zhǔn)偏差差占測量量系統(tǒng)散散布的99.735.15是產(chǎn)業(yè)標(biāo)標(biāo)準(zhǔn)，Minitab14版轉(zhuǎn)換為為6Usually expressedaspercentTolerance=USL-LSLP/TRatio2/8/2020ATLConfidentia

10、l InformationP/TV零件選擇擇時，散散布較小小或較大大的情況況下，由由于分散散的相對對比較緣緣故，會會造成測測量系統(tǒng)統(tǒng)能力的的錯誤性性的判定定P/TV:Bestcase:15樣本的選選擇:Option1:如果過程程的變異異已知，樣本的的選擇應(yīng)應(yīng)代表過過程或產(chǎn)產(chǎn)品的正正常分布布Option2:如果過程程的變異異未知，樣本的的選擇應(yīng)應(yīng)均勻的的涵蓋規(guī)規(guī)格的寬寬度2/8/2020ATLConfidential InformationStep 1:選取10個樣本。另外確確定操作作者，操操作者要要從日常常使用此此設(shè)備的的員工中中選取Step 2:對設(shè)備進(jìn)進(jìn)行校驗驗或確認(rèn)認(rèn)設(shè)備在在上次的的計量

11、周周期內(nèi)Step 3:建立Minitab R&R研究的數(shù)數(shù)據(jù)收集集表Columnheadings:Column1:PartID(1-10)Column2:Operator (1-3)Column3:Trial (1-2)Column4:Measurement(s)Example -GageR&RusingMinitab2/8/2020ATLConfidential InformationStep 4:要求第一一個操作作員按照照隨機(jī)的的順序測測量所有有樣本一一次。Blindsampling：指操作作員不知知道每個個樣本被被定義的的序號，這樣可可以減小小人員的的偏差Step 5:其余操作作員以同同

12、樣的方方式隨機(jī)機(jī)測量同同樣的樣樣本，為為試驗次次數(shù)1Step 6:按照要求求的試驗驗次數(shù)重重復(fù)steps4 &5。如果整整個試驗驗?zāi)軌蛟谠谝欢螘r時間內(nèi)（幾天）完成是是最好的的。Step 7:在Minitab中錄入數(shù)數(shù)據(jù)及規(guī)規(guī)格Stat QualityTools GageR&R StudyStep 8:分析數(shù)據(jù)據(jù)結(jié)果Example -GageR&RusingMinitab2/8/2020ATLConfidential InformationStep 7:在Minitab中錄入數(shù)數(shù)據(jù)及規(guī)規(guī)格Stat QualityTools GageR&R StudyExample -GageR&RusingMi

13、nitabANOVA method is preferred.Enter Gage Info and Options (see next page)2/8/2020ATLConfidential InformationtotalmeastotalmeasTVPssss=22/LSLUSLLSLUSLTPMSMS-=-=ss*15.5*15.5/2Note:P/TV的分母可可根據(jù)樣樣本的研研究或歷歷史數(shù)據(jù)據(jù)評估Example -GageR&RusingMinitab2/8/2020ATLConfidential InformationGage R&RStudy- ANOVA MethodANOV

14、ATableWith Operator*PartInteractionSourceDFSSMSFPParts92.058710.22874539.71780.00000Operators20.048000.0240004.16720.03256Oper*Part180.103670.0057594.45880.00016Repeatability 300.038750.001292Total592.24912Gage R&RSourceVarCompStdDev5.15*SigmaTotalGage R&R0.0044370.0666150.34306Repeatability0.001292

15、0.0359400.18509Reproducibility0.0031460.0560880.28885Operator0.0009120.0302000.15553Oper*Part0.0022340.0472630.24340Part-To-Part0.0371640.1927810.99282TotalVariation0.0416020.2039651.05042Source%Contribution%StudyVar%ToleranceTotalGage R&R10.6732.6622.87Repeatability3.1017.6212.34Reproducibility7.56

16、27.5019.26Operator2.1914.8110.37Oper*Part5.3723.1716.23Part-To-Part89.3394.5266.19TotalVariation100.00100.0070.03NumberofDistinctCategories= 4Session WindowOutput:GraphicalOutput:Gage R&ROutput2/8/2020ATLConfidential InformationGage R&RStudy- ANOVA MethodANOVATableWith Operator*PartInteractionSource

17、DFSSMSFp-valueParts92.058710.22874539.71790.00000Operators20.048000.0240004.16720.03256Oper*Part180.103670.0057594.45880.00016Repeatability300.038750.001292Total592.24913零件與操操作者是是統(tǒng)計上上顯著的的Gage R&R,ANOVATable0.052/8/2020ATLConfidential InformationGaugeR&RStudy-ANOVAMethodSourceVar CompStdDev95%ConfInt

18、5.15*SigmaTotalGaugeR&R0.0044370.066615(0.0597,0.2250)0.34306Repeatability0.0012920.035940(0.0287,0.0480)0.18509Reproducibility0.0031460.056088(*,*)0.28885Operator0.0009120.030200(0.0000,0.2169)0.15553Oper*Part0.0022340.047263(0.0254,0.0800)0.24340Part-To-Part0.0371640.192781(0.1098,0.3345)0.99282To

19、talVariation0.0416020.2039651.05042Varianceduetothe measurement system(broken downintorepeatability andreproducibility)TotalvarianceVarianceduetothe partsStandarddeviationfor eachvariancecomponent99%ofthe variationrepresentedGage R&R,Variation Components2/8/2020ATLConfidential InformationSourceVarCo

20、mp5.15*SigmaTotalGaugeR&R0.0044370.34306Repeatability0.0012920.18509Reproducibility0.0031460.28885Operator0.0009120.15553Oper*Part0.0022340.24340Part-To-Part0.0371640.99282TotalVariation0.0416021.05042326600504134300. StudyVarTV/Ptotalmeas=ss=Source%Contrib%StudyVar%ToleranceTotalGage R&R10.6732.662

21、2.87Repeatability3.1017.6212.34Reproducibility7.5627.5019.26Operator2.1914.8110.37Oper*Part5.3723.1716.23Part-To-Part89.3394.5266.19TotalVariation100.00100.0070.032287.05.13430.0 *15.5/=-=TolLSLUSLTPMSs1067.0041602.0004437.0 22=totalMSContributionssGage R&R,Results2/8/2020ATLConfidential Information

22、Gage R&R-Number of Distinct Categories識別力與與ContributionDistinctCategories=類別2345678Contribution33.318.211.17.45.33.93.02/8/2020ATLConfidential InformationGage R&R-Number of Distinct Categories識別力的的意義及及判定Automobile Industry ActionGroup(AIAG)建議：類別數(shù)2=2=35備注系統(tǒng)無法法區(qū)分兩兩個部品品系統(tǒng)只能能將數(shù)據(jù)據(jù)分為兩兩組，比比如：高高和低系統(tǒng)只能能將數(shù)據(jù)據(jù)分

23、為三三組，比比如：高高、中、低系統(tǒng)基本本可接受受2/8/2020ATLConfidential InformationGage R&R,ComponentsofVariation“整個散布布中，Gage R&R占據(jù)的比比重是否否足夠小??？”Gage R&R,Repeat,Reprod的高度越越接近0越好2/8/2020ATLConfidential InformationGage R&R,Xbar&R極差圖(重復(fù)性)用于確定定過程是是否受控控確定與重重復(fù)性有有關(guān)的統(tǒng)統(tǒng)計控制制確定過程程中評價價人之間間對每個個零件的的一致性性R chart的點超出出控制線線的話，需調(diào)查查其原因因后再測測量作業(yè)者之之間反復(fù)復(fù)測量值值是否穩(wěn)穩(wěn)定？2/8/2020ATLConfidential InformationGage R&R,Xbar&R均值圖(再現(xiàn)性)提供了測測量系統(tǒng)統(tǒng)的“可可用性”指示：50的點落落在控制制線外測量系系統(tǒng)應(yīng)該該能夠充充分探測測零件間間的變差差測量系系統(tǒng)能夠夠提供過過程分析析和過程程控制有有用的信信息如果=5 (10)2/8/2020ATLConfidential Information應(yīng)該各關(guān)關(guān)注哪項項？測量系統(tǒng)統(tǒng)還是