Minitab常用操作说明.pptx

ConfidentialMinitab 常用操作說明Revision : APrepared by : 蔡培奪Date : 12/1 2006 Confidential內容：1. 產品尺寸評估方式說明2. 客訴報告中常用操作(2-sample t）3. Gauge R&R評估說明Confidential1. 以Boxplot 確認量測值內是否有離散值2. 以Normality Test 確認量測值是否呈常態分配3. 以1-sample T 檢定量測值是否與目標值相符4. 以Capability Analysis 來確認Cpk & Sigma Level1. 產品尺寸評估方式說明Confidential1.1. Boxplot 離散值確認M M e ea as su ur re e3. 1623. 1603. 1583. 1563. 1543. 1523. 150B Bo ox xp pl lo ot t o of f M M e ea as su ur re em m e ea as su ur re e4. 2304. 2254. 2204. 2154. 2104. 205B Bo ox xp pl lo ot t o of f m m e ea as su ur re e有離散值無離散值 GraghBoxplotOne Y/SimpleConfidential1.1. Boxplot 離散值確認從Boxplot 確認是否有Outlier, 若有的話需再確認其是否已out spec (有無異常).Confidential1.2. 量測值是否呈常態分配確認 StatBasic StatisticNormality testConfidentialM M e ea as su ur re eP Pe er rc ce en nt t3. 1643. 1623. 1603. 1583. 1563. 1543. 1523. 150999590807060504030201051M ean0. 1253. 156St Dev0. 002893N32AD0. 574P- Val ueP Pr ro ob ba ab bi il li it ty y P Pl lo ot t o of f M M e ea as su ur re eNorm al m m e ea as su ur re eP Pe er rc ce en nt t4. 2304. 2254. 2204. 2154. 2104. 2054. 200999590807060504030201051M ean 0.05, 量測值呈“常態分配”P-value 0.05, 若其0.05 即表示其接受虛無假設Ho, 也就是量測值等於目標值; 若P-value 0.05 即表示拒絕虛無假設而接受對立假設, 也就是量測值不等於目標值.Confidential1.4. 以Capability Analysis 來確認Cpk & Sigma Level首先需先確認其是否為常態分配(同1.2量測值是否呈常態分配確認）M M e ea as su ur re eP Pe er rc ce en nt t3. 1643. 1623. 1603. 1583. 1563. 1543. 1523. 150999590807060504030201051M ean0. 1253. 156St D ev0. 002893N32A D0. 574P- V al ueP Pr ro ob ba ab bi il li it ty y P Pl lo ot t o of f M M e ea as su ur re eNorm al m m e ea as su ur re eP Pe er rc ce en nt t4. 2304. 2254. 2204. 2154. 2104. 2054. 200999590807060504030201051M ean0. 0054. 217StD ev0. 005622N32A D1. 572P-V alueP Pr ro ob ba ab bi il li it ty y P Pl lo ot t o of f m m e ea as su ur re eNorm al Confidential1.4.1. 當量測值為常態分配 StatQuality ToolsCapability Analysis NormalConfidential由於我們為一個lot 收集32 個量測值而非不同lot 進行收集, 因此於Subgroup size 部分填入32.分別填入既定的尺寸上下限值ConfidentialConfidential0. 7200. 7080. 6960. 6840. 6720. 6600. 648LSLUSLProcess Dat aSam pl e N32St Dev(W i t hi n)0. 00895598St Dev(Overal l )0. 00895598LSL0. 64Target*USL0. 72Sam pl e M ean0. 683906Pot ent i al (W i t hi n) Capabi l i t yOveral l Capabi l i t yZ. Bench4. 03Z. LSL4. 90Z. USL4. 03Ppk1. 34CpmZ. Bench*4. 03Z. LSL4. 90Z. USL4. 03Cpk1. 34Observed Perform ancePPM USL0. 00PPM Tot al0. 00Exp. W i t hi n Perform ancePPM USL27. 87PPM Tot al28. 35Exp. Overal l Perform ancePPM USL27. 87PPM Tot al28. 35W ithinOverallP Pr ro oc ce es ss s C Ca ap pa ab bi il li it ty y o of f C C1 1可分別得到Cpk & Z Bench 若數據為短期：CPK=1.34 sigma lever=Z.Bench=4.93 若數據為長期：CPK=1.34 sigma lever=Z.Bench+1.5=6.43ConfidentialConfidential1.4.2. 當量測值為非常態分配填入組內量測值數量 StatControl ChartsBox-Cox TransformticConfidentialL La am m b bd da aS St tD D e ev v5. 02. 50. 0-2. 5-5. 00. 0059750. 0059500. 0059250. 0059000. 0058750. 005850Lam bda- 5. 00( usi ng 95. 0% confi dence)Est i m at e- 5. 00Low er CL*U pper CL*Rounded V al ueB Bo ox x- -C Co ox x P Pl lo ot t o of f C C1 1正態轉換係數Lambda = -5.00Confidential將之前得到的正態轉換係數填入即可, 其他步驟則與量測值為常態分配時一樣 StatQuality ToolsCapability Analysis NormalBox CoxConfidential0. 04680. 04560. 04440. 04320. 04200. 0408LSL*USL*t ransform ed dat aProcess Dat aSam pl e N32St Dev(W i t hi n)0. 00591975St Dev(Overal l )0. 00591975Aft er Transform at i onLSL*0. 0474145Target *LSL*USL*0. 0403861Sam pl e M ean*0. 0430389St Dev(W i t hi n)*0. 0006708St Dev(Overal l )*0. 00067081. 84Target*USL1. 9Sam pl e M ean1. 87603Pot ent i al (W i t hi n) Capabi l i t yOveral l Capabi l i t yZ. Bench3. 95Z. LSL6. 52Z. USL3. 95Ppk1. 32CpmZ. Bench*3. 95Z. LSL6. 52Z. USL3. 95Cpk1. 32Observed Perform ancePPM USL 0. 00PPM Tot al0. 00Exp. W i t hi n Perform ancePPM LSL*0. 00PPM LSL*0. 00PPM USL* 38. 33PPM Tot al38. 33W i t hi nOveral lP Pr ro oc ce es ss s C Ca ap pa ab bi il li it ty y o of f M M e ea aUsi ng Box-Cox Transform at i on W i t h Lam bda = -5 若數據為短期：CPK=1.32 sigma lever=Z.Bench=3.95 若數據為長期：CPK=1.32 sigma lever=Z.Bench+1.5=5.45Confidential分別填入Cpk & Sigma Level (Overall Capability Z Bench + 1.5)最後將mini tab 分析結果做綜合判斷Confidential2. CAR 報告中操作說明.2.1 1-sample的數據1-sample的數據分析同item 1 Boxplot 離散性確認 Normal test CPK確認Confidential2.2 2-sample的數據2-sample的數據分析如下 Boxplot 離散性確認 Normal test標准差檢定標准差檢定差異分析Confidential2.2.1 sample size確定Confidential2.2.2 Boxplot 離散性確認M M e ea as su ur re e3. 1623. 1603. 1583. 1563. 1543. 1523. 150B Bo ox xp pl lo ot t o of f M M e ea as su ur re em m e ea as su ur re e4. 2304. 2254. 2204. 2154. 2104. 205B Bo ox xp pl lo ot t o of f m m e ea as su ur re e有離散值無離散值 GraghBoxplotOne Y/with groupConfidential2.2.2 Normal Test(同item 1.2) StatBasic StatisticNormality testConfidential2.2.3標准差檢定ConfidentialConfidential2.2.4平均值檢定ConfidentialConfidential3 Gauge R&R-計數型計數型3.1基本概念基本概念Confidential3.2 准確性及線結評估准確性及線結評估ConfidentialConfidentialConfidential3.3 重復性及再現性評估重復性及再現性評估ConfidentialConfidentialConfidentialConfidentialConfidentialKey message: 量測人員之間無明顯量測差異,不同sample之間有明顯差異.顯示量測系統無異常.3.4 Gauge R&R 實例實例例: A,B,C 3個量測員分別針對10個產品進行量測，量測設備為QV，重復量測次數2次，所得數據如下，針對量測系統進行分析:3.4.1 准確性及線性評估Confidential3.4.2 Gage R&R-重復性及再現性分析StaticQuality toolsGauge studyGauge R&R(crossed)ConfidentialPART P-value小於0.05, 顯示樣品為影響量測變異的顯著因子.量測系統造成之變異占總變異的3.29% 5此量測系統可接受可接受%SV為18.13%30%, 判定為可可接受接受.3.4.2 Gage R&R-重復性及再現性分析Confidential3.4.2 Gage R&R-重復性及再現性分析從Gage R&R和Part to Part所占比例可以看出 主要變異來自樣品本身.從 Chart 可以發現部份樣品在不同的Operator重復量 測下差異較不大不大,由此顯示量測沒有產生變異X軸:樣品,Y軸:結果從 Chart顯示量測員對于樣品量測呈現一致性呈現一致性從X-Bar chart by operator可以發現的數據都落在管制界線之外, 顯示大部分的變異均來自於Part-to-Part的變異的變異.Confidential