MSA测量系统分析(英文版)培训讲学课件.ppt

1、SPC for Test System Gauge R&R有多少個黑點？這些橫線是不是平行的？是靜的還是動的？三個正方形幾何謎題沒有適當的量測工具沒有適當的量測工具就無法管理品質就無法管理品質! 比較式 絕對式待測量轉換訊息標準值顯示量測儀量測之種類量測之種類待測量標準量轉換訊息轉換訊息比測儀顯示轉換放大Basic ConceptsEvery Process produce a “ Product”.Every product possesses Qualities ( Features)Every Quality Feature Can be measuredTotal Variation

2、= Product Variation + Measurement VariationSome Variation Inherent in System Design.Some Variation is due to a Faulty performance of the System(s).Components of Measurement SystemEquipment or GageType of Gage:Attribute: go-no go, vision systems Variable: calipers, probe, coordinate measurement machi

3、nesUnit of Measurement - discriminationAt least 1/10 of 6*process standard deviation (6 6s)Other rules: at least 1/10 of toleranceOperator & Operating InstructionsPart locating or orientation schemea gage must be able to consistently locate the part being measured.Terms You Should Know Measurement s

4、ystem : The collection of operations, procedures,gages and personnel use to obtain measurements of work piece characteristics Gage: A device used to obtain measurements Repeatability : Repeatability (Precision) The variation in measurement obtained with one measurement instrument when used by one op

5、erator to measure.) an identical characteristic on the same part. ( EV : Equipment Variation Reproducibility : The variation in the average of the measurements made by different operators using the same measurement instrument when measuring identical characteristics on the same part. ( AV : Appraise

6、r Variation )Terms You Should KnowBias : The difference between the observed average of measurements and the reference value. Percent R&R : The percentage of process variation related to the measurement system for repeatability and reproducibility Stability : The total variation in the measurements

7、obtained with a measurement system on the same master or parts when measuring single characteristics over an extended timeperiod. Linearity : The difference in the bias values through the expected operating range of the gage.1. AccuracyDifference between the true average and the observed average (tr

8、ue average may be obtained using a more precise measuring tool)2. ReproducibilityVariation in average of measurements made by different operators using the same gage measuring the same part.Appraiser Variation (AV)3. Repeatability (Precision)Random variation in measurements when one operator uses th

9、e same gage to measure the same part several times. Equipment Variation (EV)4. StabilityDifference in the average of at least 2 sets of measurements obtained with a gage as a result of time.Precision and AccuracyAccuracy and StabilityMeasure same parts using a more precise instrument (Accuracy). Mea

10、sure same parts using the same gage at a later date (Stability or Drift).Measurement System VariablesPossible Sources of ProcessVariationSources of Measurement Error Systematic Variations 1. accuracy: improper calibration of equipment 2. reproducibility: different operators using same equipment with

11、 different techniques Random Variations 3. repeatability: (for example, inconsistent part locating) Periodic Variations 4. stability: wear, deterioration, environmental conditionsThe GR&R Design ModelSuppose multiple operators, measure multiple parts several times each. Then: In terms of the gage pe

12、rformance, between-operator reproducibility is due to the operator effect and the operator*part interaction. Within-operator repeatability is from the pure error term.The Components of VarianceMeasurement Error Gage R&RAveragesANOVA analysis begins with the calculation of all relevant averages. Here

13、s an example:Types of Gage Error and AnalysisMethods:Accuracy Gage Accuracy (Bias) Study Gage Linearity Study.Repeatability & Reproducibility Gage R&R (Methods: X-bar/Range or ANOVA) Repeated Measurement StudyRepeatability Static & Dynamic StudyStability Repeated MeasurementsVariance PartitioningThe

14、 total variance of the GR&R data is partitioned into components according to the model.Variances add, so all partitioned pieces sum to the total data variance!Two Way Anova TableGage R&R AcceptanceTo determine acceptability, a ratio is typically formed comparing gage variation to either: Gage R&R %

15、of Tolerance Width Gage R&R % of Total Variation 6s 6s - represents 99.73% of values if normal distribution 5.15 s s - represents 99% of values if normal distributionCriteria for Evaluating Gage R&R Gage R&R % - TV or R&R% - Tolerance if R&R% is: 30% = measurement system needs improvementConducting

16、a Gage R&R Study1. Plan the approach What is nature of operator and measurement interaction? Reproducibility can be negligible sometimes.2. Select 3 well trained operators3. Select 10 parts for test, Select sample parts should represent entire normal production range. ( Prefer cover the distribution

17、 of +/- 4 sigma )4. Repeat measure each parts 3 times. 5. Take measurements in random order.6. Take blind measurement Person measuring doesnt know which piece is being measured, what previous measurements were, or what other appraisers found.Conducting a Gage R&R Study7. Record all measurements on w

18、orksheet.8. Calculate trial range for each piece for each operator and record values, calculate R&R (square root of the sum of the squares of the repeatability and reproducibility).9. Keep the standard parts & its reading for “ Equipment Correlation “ purposeGuidelines For Gauge R&R AcceptanceUnder

19、10% Error : Acceptable10%30% Error : May be Acceptable depending on the importance of the application.Over 30% error : Improvement is neededGauge R&R StudyGauge R&R StudyGauge R&R StudyGauge R&R StudyGauge R&R StudyAnova Method ExampleEquipment Calibration By Simple Linearity Model( Y = mX + b )Sele

20、ction of Standard UnitsSet up about 10 pcs standard units which can cover the overall spread of normal production.Prefer 2 reject unit beyond upper / lower spec.Aging all units Find standard readingRegression Analysis : Equipment vs StandardMeasure Standard unit on each individual test equipment sev

21、eral times.Take the average of measurement data for each unitRegression Analysis ; Equipment Reading vs Standard Reading.If correlation coefficient 90% , then calculate the slope and off-set per test m/c.If correlation coefficient chance and off diagonal frequencies should be chance. Use marginal fr

22、equencies/probabilities to estimate chance agreement. Proportion agreement observed, po= ipii = 1/n iniiProportion agreement expected by chance, pc= ipi. x p.i po - pc 1 - p cKappa, K = po= .53 + .14 .03 = .7pc= .39 + .075 + .01 = .475 .7 - .475 K =1 - .475 = .429 K = 1, perfect agreementK = 0, chan

23、ce agreementK0.75. If Kappa is 0.75, the measurement system must be improved before accepting or rejecting parts.Computing Kappa ( GR&R Test For visual inspector )Example of Computing KappaExample of Computing KappaProportion agreement observedProportion agreement expected by chanceK = P (agreement | no agreement by chance)1-pc = 1- 0.503 = 0.497 of cases where no agreement by chancepo - pc = 0.825 0.503 = 0.322 of cases are those non-chance agreement cases where observers agreed.&