PHILIPS绿带训练教材-MSA测量系统分析

1、Released June 2003测量系统分析 (MSA)绿 带 训 练Green Belt Training, released June 2003BEST 2学习目标 完成本课学习后，你应该能够: 说明测量误差是怎样影响我们评估产品和服务质量和改进流程绩效的. 应用QS9000 推荐的程序开展 量具的 R&R 研究 使用 MINITAB 分析量具的 R&R 研究 的结果并制作图表阐发你的报告Green Belt Training, released June 2003BEST 3测量系统 精度 分辨率 重复性 可再现性Green Belt Training, releas

2、ed June 2003BEST 4不可接受不可接受不可接受不可接受可以接受可以接受不确定性不确定性?不确定性不确定性? LSL USL 质量质量测量误差出现测量误差时，我们怀疑产品或服务的质量Green Belt Training, released June 2003BEST 5决 定接受产品接受产品拒绝产品拒绝产品 产 品 真 实 情 况 优质产品优质产品 劣质产品劣质产品 交付优质产品交付优质产品 优质产品被报废优质产品被报废 交付劣质产品交付劣质产品劣质产品被报废劣质产品被报废测量误差的影响Green Belt Training, released June 2003BEST 6测量

3、误差对过程控制的影响 因为有测量误差，我们可能: 发现不了过程中的变化 发现变化太迟 过程没有变化时以为有变化这往往是业绩变差的一大原因Green Belt Training, released June 2003BEST 7测量误差对过程改进的影响6 57 58 59 51 0 5A re w e g e ttin g h ig h e r yie ld fro m th e n e w p ro c e s s ?O ldN e wY ie ldGreen Belt Training, released June 2003BEST 8精 度 偏差: 实际测量值与真实值的平均差值 怎样应对偏

4、差: 校准Green Belt Training, released June 2003BEST 9分辨率 规则: 测量单位应比允许公差精确一位数字 If T = x 1 If T = x 2.0 分辨率 至少 Punch holes diameter0.810.820.830.840.850.860.870.880.890.900.91147101316192225283134374043464952555861646770737679828588919497100samplenrcmLSLUSLGreen Belt Training, released June 2003BEST 10例四

5、位操作员的测量操作员 直径(mm) 平均值 SD哪一位所得结果最佳?哪一位未来将有更好的业绩?你的建议?Green Belt Training, released June 2003BEST 1119.5 20.0 20.5 21.0 Diameter19.5 20.0 20.5 21.0 Diameter19.5 20.0 20.5 21.0 Diameter19.5 20.0 20.5 21.0 DiameterSmithJonesBrownLee4 位操作员 6 次测量图 示谁最佳?Green Belt Training, released June 2003BEST 12重复性和可再现

6、性 重复性: 测量者在未来还会重复他过去的结果吗？ 可再现性: 其他人还会再作出第一人的测量结果吗？Green Belt Training, released June 2003BEST 13重复性重复性即同一人用同一测量工具重复测量同一个对象得出的测量值变差 重复性常被称为重复性常被称为 “量具变差量具变差”Green Belt Training, released June 2003BEST 14可再现性可再现性即不同的人用同一测量工具重复测量同一个对象得到的其它 测量值变差可再现性常被称为可再现性常被称为 “操作员变差操作员变差”Green Belt Training, released

7、 June 2003BEST 15R & rR & r即不同的人用同一标尺重复测量同一个对象得到的总共 测量值变差R & r 取决于量具变差和取决于量具变差和 操作者变差操作者变差Green Belt Training, released June 2003BEST 16 平均 s-2.575s99%99%的重复测量值位于一个宽s s的范围中。测量值变差测量值变差即99%的重复测量值的覆盖范围Green Belt Training, released June 2003BEST 17R&r 计算 重复性 ssE s E = 某操作员的重复测量值的标准方差 可再现

8、性= 5.15 s A s A = 不同操作员的平均测量值的标准方差 R&r = 5.15 s T s T = 不同操作员的重复测量值的标准方差Green Belt Training, released June 2003BEST 18R&R 有两个成分重复性重复性可再现性可再现性R&R重复性重复性可再现性可再现性R&RR&R = (重复性重复性 + 可再现性可再现性)Green Belt Training, released June 2003BEST 19计算 % R&r R&r 即在允许公差内的产品比例，或产品变差的比例。%R&am

9、p;r = 100*R&r/(USL-LSL) 评估标准:%R&r %R&r 30%可接受%R&r 30%不可接受Green Belt Training, released June 2003BEST 20 LSL USL LSL USL LSL USL LSL USLR&R = 100%R&R = 30%R&R = 10%R&R = 0%你能证明产品质量吗?Green Belt Training, released June 2003BEST 21量具能力实验选出10个对象对这 10 个对象做好暗记找来数位操作员操作员A 量了1

10、0 个对象操作员A 重新量了10 个对象(不同顺序)操作员B 量了10 个对象操作员B 重新量了10 个对象 (不同顺序)(可以添加第三个操作员)记录这 40 或 60 个测量结果将数据输入 MINITAB制作度量运行图，进行量具的 R&r 研究 分析数据报告结果Green Belt Training, released June 2003BEST 22 重复性 可再现性操作员 操作员和部 件的交互作用变 差我们需要计算40 或 60个结果的总变差并将其细分 总变差 部件的相对变差 测量系统变差 MINITAB 中选择:Stat Quality Tools Gage R&R S

11、tudy (Crossed)Green Belt Training, released June 2003BEST 23例 装架长度装架长度 # 27小组 NR50Jeroen, Theo, Harold, Matt and Peter 每人测量5 个样品各3次 在 options 下添加:Process Tolerance = 3装架长度=Green Belt Training, released June 2003BEST 24R&R 结果3Jeroen van de Laar25 november 1999Mounth Length mmMisc:Tolerance:Report

12、ed by:Date of study:Gage name:026,526,025,525,0TheoPeterMattJeroenHaroldXbar Chart by OperaterSample MeanMean=25,67UCL=25,79LCL=25,5500,50,40,30,20,10,0TheoPeterMattJeroenHaroldR Chart by OperaterSample RangeR=0,116UCL=0,2986LCL=05432126,025,525,0ProductOperaterOperater*Product InteractionAverageHar

13、old Jeroen Matt Peter Theo TheoPeterMattJeroenHarold26,325,825,324,8OperResponse By Operater5432126,325,825,324,8ProductResponse By Product%Contribution %Study Var %Tolerance Part-to-PartReprodRepeatGage R&R100500Components of VariationPercentGage R&R study Mount length group 27Green Belt Tr

14、aining, released June 2003BEST 25视窗显示的文件量具 R&R %Contribution来源 VarComp (of VarComp) 总测量 R&R 0,02252 12,77 重复性 0,01006 5,70 可再现性 0,01246 7,06 操作员 0,01246 7,06 部件的相对变差 0,15391 87,23 总变差 0,17643 100,00 StdDev Study Var %Study Var %Tolerance来源 (SD) (5,15*SD) (%SV) (SV/Toler) 总测量 R&R 0,150077

15、 0,77290 35,73 25,76 重复性 0,100303 0,51656 23,88 17,22 可再现性 0,111636 0,57493 26,58 19,16 操作员 0,111636 0,57493 26,58 19,16 部件的相对变差 0,392310 2,02040 93,40 67,35 总变差 0,420036 2,16319 100,00 72,11R&R 相对于观测到的流程变差R&R 相对于允许公差30% R&R 不可接受Green Belt Training, released June 2003BEST 26测量运行图3Mounth

16、LengthMisc:Tolerance:Reported by:Date of study:Gage name:26.325.825.324.854321ProductMounth LengtHarold Jeroen Matt Peter Theo Runchart of Mounth Lengt by Product, OperaterGreen Belt Training, released June 2003BEST 27术 语% contribution :偏差占总偏差的百分比偏差占总偏差的百分比部件对部件部件对部件 % + R&R % = 100%重复性重复性 % + 再

17、现性再现性 % = R&R % 研究的偏差研究的偏差 :标准方差占总标准方差的百分比标准方差占总标准方差的百分比总总 SD = Sqrt (总变差总变差)%允许公差允许公差:SD) 与总共允许公差的比例与总共允许公差的比例 a %Green Belt Training, released June 2003BEST 28用极差计算 R&R1.计算每位操作员测量每一产品得出的平均值2.计算每位操作员得出的 从X1 到 X3 的平均值3.计算每件产品平均值 Xmean4.计算每位操作员每一产品测量值的极差5. 计算R1 到 R3 的每位测量员得出的极差平均值6.计算个别的Xmean

18、s: Rp 的极差7.确定从 R1 到 R3: Rmm 的平均值8.确定从 X1 到 X3: X diff 的极差9.确定重复性 (量具变差) EV:10.确定可再现性 (测量员变差) AV:11.确定 R&R 值:12.确定产品分布 PV:13.确定总分布 TV:14. R&R 值除以 TV (允许公差的宽度), 确定 % R&R例见下页和 1 03a M 用标准偏差代替极差，使用同样原则计算MSE， 这样更复杂困难，但更有效， 同样可以用于计算产品和人员的交互作用。1*KREVmmrnEVKXAVDiff*/*22222&AVEVRR3*KRPVp22&am

19、p;PVRRTVPVRRRRiationprocess&%varTVRRRRidthTolerancew&%Green Belt Training, released June 2003BEST 29Measurement System Analysis (FORD)Characteristic:Focus DistanceDate:12-3-03Department:Process ControlMeasurement Unit:0.01 mmMeasurement device:MitutoyaNominal Value:58Controlplan number:A123T

20、olerance:14Remarks:Final VersionEvaluation number:4Product NumberX Kaart12345678910Operator 1Trial 157.457.957.957.758.258.458.458.458.258.3Trial 257.457.957.757.858.158.458.458.458.358.5Trial 357.45857.957.858.358.458.458.558.358.5meanXmmRangeRmm X157.4057.9357.8357.7758.2058.4058.4058.4358.2758.43

21、58.11 R10.000.100.200.100.200.000.000.100.100.200.10Operator 2Trial 157.35857.657.858.358.458.258.458.258.5Trial 257.457.857.857.758.158.458.358.358.158.5Trial 357.45857.857.758.258.358.358.458.258.3 X257.3757.9357.7357.7358.2058.3758.2758.3758.1758.4358.06 R20.100.200.200.100.200.100.100.100.100.20

22、0.14Operator 3Trial 157.55857.957.858.358.458.358.458.358.5Trial 257.557.957.857.858.358.558.458.458.358.5Trial 357.457.957.857.858.358.558.458.458.358.5 X357.4757.9357.8357.8058.3058.4758.3758.4058.3058.5058.14 R30.100.100.100.000.000.100.100.000.000.000.0558.100.10Xmean57.4157.9357.8057.7758.2358.

23、4158.3458.4058.2458.461.04Upper Control Limit mean (UCLx) =58.20Xmm + A2 * Rmmtrials ( r )A2D3D4trialsK1OperatorK2Lower Control Limit mean (LCLx) =58.00Xmm - A2 * Rmm21.8803.26724.5623.6531.02302.57533.0532.7Upper Control Limit range (UCLr) =0.25Rmm * D440.72902.282Lower Control Limit range (LCLr) =

24、0.00Rmm * D350.57702.115Parts(p)K323.65Measurement average operator 1 (X1) =58.11Measure range operator 1 (R1,mean) =0.1032.7Measurement average operator 2 (X2) =58.06Measure range operator 2 (R2,mean) =0.1442.3Measurement average operator 3 (X3) =58.14Measure range operator 3 (R3,mean) =0.0552.08Ra

25、nge of X1 till X3 (Xdiff) =0.08Average of R1 till R30.1061.93Range of measurement average per Product Number (Rp) =1.0471.8281.74Equipment Variation (EV) or Repeatability =0.29Rmm * K191.67Appraiser Variation (AV) or Reproducibility =0.21SQRT(Xdiff * K2)2 - ( (EV)2 / (n * r) ) 101.62Repeatability &a

26、mp; Reproducibility (R&R) =0.36SQRT(EV2 + AV2)sigmaR&R (sR&R)0.07(R&R)/5.15Part Variation (PV) =1.69Rp * K3sigmap(sp)0.33(Rp * K3)/5.15Signal/Noise(S/N)4.68sp / sR&R%R&R (100 * R&R / TV )=21Nr. Distinct Product Categories6.601.41 * sp / sR&R, eis 5%R&R (100 * R&am

27、p;R / Tolerance width ) =3Total Variation (TV) =1.73SQRT(R&R2 +PV2)The Measurement System is evaluated as Acceptablealle berekeningen zijn gebaseerd op 5.15 s (99%) van het oppervlak onder de normaal curveK2 = 5.15 / d2, d2(m = # operators, g = 1 (1 range berekening)K1 = 5.15 / d2, d2(m = # trials, g = # nr. parts * # operators, verondersteld 15)K3 = 5.15 / d2, d2(m = # parts, g = 1 (1