Granite State ASQ 0104 MSA an alternative method for estimating % Tolerance April 18, 2012 Jack Meagher - NHBB

Transcription

1 Granite State ASQ 00 MSA an alternative method for estimating % Tolerance April 8, 0 Jack Meagher - NHBB

2 New Hampshire Ball Bearings Founded in 96 in Peterborough, NH Acquired by Minebea (Japan) in 98 Three U.S. manufacturing facilities Precision Division, Chatsworth, CA Astro Division, Laconia, NH HiTech Division, Peterborough, NH First bearing company to register all of our manufacturing divisions to ISO 9000 and the first company to register at least one division to ISO 900.

3 Before we start Why are you here? Curiosity? Obtain guidelines? Understand the concerns with AIAG methods? Much needed (Gage) R&R?

4 Before we start Why Note: are you here? I am Curiosity? an Engineer NOT - a Math Major or a Statistician. The Obtain guidelines? method proposed here is intended Understand to be of the practical concerns use when with conducting AIAG methods? MSA s. Much needed (Gage) R&R?

5 Before we start Note: Having been trained as a Six Sigma Black Belt, you are taught to let the data lead you. Let s see where the data presented here leads us.

6 Individuals who supported this project: Donald Wheeler, well known Statistician and Author Philip J. Ramsey Ph.D. North Haven Group University of New Hampshire Roy Geiger, NHBB Operations Manager, Peterborough, NH

7 Donald Wheeler, well known Statistician and Author (spcpress.com) Read the article Problems with Gauge R&R Studies Note: JMP is incorporating Wheeler s concepts into their Gauge R&R analysis option.

8 Philip J. Ramsey Ph.D. North Haven Group University of New Hampshire When asked if there was any test in statistics that uses a ratio of standard deviations, the answer was NO. Note: F = σ /σ (ratio of Variances!)

9 Philip Example: J. Ramsey Ph.D. North Haven Group Variances University of New Hampshire F = σ When /σ asked = /0 = /00 = 0.0 if there was any test in Standard statistics Deviations that uses a ratio of standard deviations, the answer F σ was NO. / σ = /0 = 0. Note: F = σ /σ (ratio of Variances!)

10 Before we start A short exercise to set the tone for this presentation Circles in the air.

11 Objectives: Offer an alternative method of obtaining a value for % Tolerance Identify and address specific concerns within the AIAG MSA Fourth Edition relating to % Study Variation and % Tolerance Note: All results shown are from using Minitab to analyze the data

12 Agenda: A preview and questions Guidelines for MSA studies Importance of chart outputs Various sample analyses Variance vs. Standard Deviation Donald Wheeler s studies Forced variation FOR REFERENCE ONLY! Questions/comments

13 A preview of what is to follow. This is where we draw the line Gage R&R %Contribution Source VarComp (of VarComp) Total Gage R&R Repeatability Reproducibility Operator Operator*Part Part-To-Part Total Variation Proportion These are the Honest Gage R&R results as identified by Donald Wheeler! Proportions add up to 00%. Definition of proportion: A part considered in relation to the whole. Process tolerance = Definition of ratio: A part to part relationship, not to the whole. Study Var %Study Var %Tolerance Source StdDev (SD) (6 * SD) (%SV) (SV/Toler) Total Gage R&R Repeatability Reproducibility 0. Ratio Operator Operator*Part Part-To-Part Total Variation Ratios based of values that are square root sum of squares? Note: The value of.88 [(6*Std Dev/(USL LSL)] is the reciprocal of the Cp value! Cp = (USL LSL)/6*Std Dev =.07 (Capability)

14 A few questions What number(s) do you look at when you run an MSA? Have you ever questioned the results? (% Tolerance looks very high) Have you ever modified and/or removed gauging systems based on these results? Have you not found parts that span the entire tolerance?

15 Are these depictions accurate? The area under the curve always equals and gage variation adds to the variation across the board, not just out near the spec limits.

16 Opening statements: You should utilize Gage R&R s for YOUR benefit. Only give the customer(s) what they ask for (just like any audit). Develop the mindset that MSA s are just like capability studies, you need to run them on a regular basis. Start with gage families calipers, micrometers, air gages, etc.

17 Opening statements (continued): Keep the parts you use in Gage studies as standards (if at all possible). Use these standards for evaluating new operators, new gages or just as a yearly audit check on the process. To check new gages, use most consistent operator and label gages as operators.

18 Before we get started - A Gage R&R should NOT be conducted without a prior understanding of your gaging system in regards to: Bias (Accuracy) - difference between the observed average of measurements and the reference value. (NHBB utilizes masters on the shop floor. They must reach equilibrium (temperature) before being used.) Linearity - difference in the bias (accuracy) through the expected range of the gage. Stability - total variation in the measurements over an extended period of time.

19 Possible Sources of Variation Observed process or product variation Actual process or product variation Measurement variation Long-term variation Short-term variation Due to measurement device (Gage) Due to operators total = product + measurement Variable R&R study Calibration program and Gage selection Repeatability Linearity Stability Accuracy Reproducibility

20 Question do you mark where to measure your parts??? Typically, we (NHBB) do mark our parts because we do not want within-part variation to come into play. We measure other characteristics such as roundness, taper, etc. utilizing other measurement devices which will have their own Gage R&R s. However, we have rotated parts to get the average diameter and used that for our response.

21 We measure our product using gages like these Note: Resolution on gage is 0 millionths ( ). Part Master

22 Question do you mark where to measure your parts??? Measuring an OD on a nice round part is not an issue because the measurement should be the same in any location.

23 Question do you mark where to measure your parts??? Too large here If your part is out-of-round, you would now be adding within-part variation into all of the readings. Too small here

24 All components must be at the same temperature! With an expansion rate of 6 millionths per degree per inch and a tolerance of on a 6 diameter, temperature differences become critical! Part Master A temperature difference of just degrees can equate to !

25 Gage R&R Guidelines: Gages should be designed to remove operator influence as much as possible (Reproducibility). Gages should be calibrated and have at least some history in regards to bias, linearity and stability. Try to keep this analysis a blind study. The operators should not know the identification of the parts they are measuring. Do this by randomizing the order of the parts being measured.

26 Gage R&R Guidelines: Randomize operators. Only do one operator for one trial and then move on to the next operator. Try to complete the measurements in a timely manner. Don t drag it out for days. You may have to retake some measurements! Appraisers should be the ones who normally operate the gages. However if you are capable of operating the gage properly, you might consider becoming one of the operators.

27 Gage R&R Guidelines: Observe every operator very carefully to detect the slightest differences that may show up as indications in the results. The study should be observed and/or recorded by an individual who recognizes the importance of the detail required to conduct a reliable study. (DOE lack of execution!) If one operator shows more consistency than the others, find out what they are doing and share that with everyone.

28 Variable Gage R&R Recording Sheet NHBB Gage Repeatability and Reproducibility Report Random Order Recording Format Part Description: Gage Type: Date: Part Number: Gage Number: Performed By: Operator Names: A: Characteristic: B: Specification: C: Enter the overall tolerance (±. =.0): Operator A: Operator B: Operator C: Part # Trial # Measurement Part # Trial # Measurement Part # Trial # Measurement Notes: When running the analysis, go to PB_ENG\Six Sigma\Gage R&R Data Recording Forms\NHBB Random Order GRR.MPJ Open the file and input your information for Operator and Measurement. The Part and Trail columns are already filled in. After you run the analysis, Save Project As a DIFFERENT file name so you do NOT overwrite this file! If using a Master for the gage, include the Gage Setup Check form included in the GRR Forms Minitab Format.xls file.

29 Part Order Sheet Part Order for Gage R&R

30 Minitab Gage R&R Worksheet Operator Part Trial Measurement

31 Guidelines for Acceptability (per AIAG) Guidelines: % Contribution % Study Variation or % Tolerance No Issues < % < 0% Marginal % to 9% 0% to 0% Needs Work > 9% > 0% Guidelines: Distinct Categories No Issues > 0 Marginal to 9 Needs Work <

32 Still the same guidelines!

33 Where to start? Do NOT automatically go to the statistics. Before you even think of looking at the statistics, you should review the Minitab charts for quick visual clues as to where any issues reside. You should identify and correct any obvious issues before running additional analyses to quantify gage performance.

34 Where to start? Reading with outlier Time Series Plot of Reading with outlier Make a run chart of the output to look for outliers data input errors.. Part Project: PART TO PART GRR MODIFIED.MPJ; Worksheet: Worksheet

35 What to look for on the R Chart Gage R&R (ANOVA) for Reading Gage name: Thickness Gage (Gasket) X-9 Date of study : 0//988 Reported by : Tolerance: Misc: J. Meagher Spec MM Gage Ty pe MM Range Chart: This is the first chart you should look at. This is your Repeatability or gage variation. Sample Range Sample Mean A B C Part A B C Part Project: AIAG Data.MPJ; Worksheet: Data Xbar Chart by Operator R Chart by Operator _ UCL= X=0.807 LCL=0.7 UCL=0. _ R=0.08 LCL=0 If you have any range points out-of-control, you should either rerun the entire GR&R, rerun just that operator or rerun just that operator for that part. If the number of zero values on the R Chart exceeds % of the total number of values, discrimination may be inadequate!

36 What to look for on the Xbar Chart Gage R&R (ANOVA) for Reading Gage name: Thickness Gage (Gasket) X-9 Date of study : 0//988 Sample Mean Sample Range Xbar Chart by Operator A B C A B C 9 0 Project: AIAG Data.MPJ; Worksheet: Data Part 6 Part Reported by : Tolerance: Misc: R Chart by Operator 9 0 J. Meagher Spec MM Gage Ty pe MM _ UCL= X=0.807 LCL=0.7 UCL=0. _ R=0.08 LCL=0 Xbar (Average) Chart: You want to see 0% or more of readings outside the control limits! If the gage variation (the control limits based on the range values) overwhelms the part variation, you might not be able to identify differences between parts. You also want to see a repeating pattern for all operators and all on the same plane.

37 What to look for on the Reading by Part Chart Gage R&R (ANOVA) for Reading Gage name: Thickness Gage (Gasket) X-9 Date of study : 0// Reported by : J. Meagher Tolerance: Spec MM Misc: Gage Ty pe MM Reading by Part Reading by Part Chart: You want to see the parts span the tolerance zone (if possible) Project: AIAG Data.MPJ; Worksheet: Data Part If any parts show a lot of scatter in relation to other parts, those parts may be more difficult to measure than the other parts. Try to find out why.

38 What to look for on the Reading by Operator Chart Reading by Operator Chart: Gage R&R (ANOVA) for Reading Gage name: Thickness Gage (Gasket) X-9 Date of study : 0// Reported by : J. Meagher Tolerance: Spec MM Misc: Gage Ty pe MM Reading by Operator You want to see a straight line here. That would indicate all operators have the same overall average for all parts A Project: AIAG Data.MPJ; Worksheet: Data B Operator C If any operators show a lot of scatter in relation to other operators, those operators may be using a different technique. Try to find out why the difference exists.

39 What to look for on the Operator * Part Interaction Chart Gage R&R (ANOVA) for Reading Gage name: Thickness Gage (Gasket) X-9 Date of study : 0//988 Reported by : Tolerance: Misc: J. Meagher Spec MM Gage Ty pe MM Reading the Operator * Part Interaction Chart: Average Part * Operator Interaction Operator A B C You want to all these lines overlap as much as possible. That would indicate all operators read all parts the same way Part Project: AIAG Data.MPJ; Worksheet: Data Any separation would indicate a difference in technique or an issue with certain parts themselves.

40 What to look for on the Components of Variation Chart Gage R&R (ANOVA) for Reading Gage name: Thickness Gage (Gasket) X-9 Date of study : 0// Reported by : Tolerance: Misc: Components of Variation J. Meagher Spec MM Gage Ty pe MM % Contribution % Study Var % Tolerance Components of Variation Chart: You want to see most of your variation in the Part-to-Part columns. Percent Gage R&R Repeat Project: AIAG Data.MPJ; Worksheet: Data Reprod Part-to-Part Also, you can get a visual indication of the relationship between the gage (Repeatability) variation and the operator (Reproducibility) variation.

41 Dial Caliper Measuring OD of Outer Ring (Centerless Ground) Blueprint Reading Gage R&R Presentation Gage name: Dial C aliper Date of study : 0//06 Sample Mean Percent Sample Range Gage R&R Components of Variation Repeat Reprod R Chart by Operator A B C Xbar Chart by Operator A B C Part-to-Part % Contribution % Study Var UCL=0.069 _ R=0.068 LCL=0 Reported by: Tolerance: Misc: Average J. Meagher Measured OD of Outer Ring Not the right tool for the A Example of improper technique. Calipers were held perpendicular to OD, not parallel. Not measured Measurement by at Part highest points on OD. Part Measurement by Operator B Operator Operator * Part Interaction UCL= job. Calipers read in _ X= increments..8 OD LCL=.87variation.800was C Operator A B C Part

42 Digital Caliper Measuring OD of Outer Ring (Centerless Ground) Blueprint Reading Gage R&R Presentation G age name: Digital C aliper Date of study : 0//06 Percent Components of Variation 00 0 % Contribution % Study Var Differences J. Meagher in technique? One operator read parts Measured OD of Outer Ring different than the other two. A lot of variation for Operator A. Reported by: Tolerance: Misc: Measurement by Part 0 Gage R&R Repeat Reprod R Chart by Operator A B C Part-to-Part.9.00 Part Measurement by Operator Sample Range Sample Mean Xbar Chart by Operator A B C UCL= _ R=0.008 LCL=0 UCL=.9997 _ X=.976 LCL= Average A B Operator Operator * Part Interaction Part C Operator A B C

43 Micrometer Measuring Across Flats of Hex Nut Blueprint Reading Gage R&R Presentation G age name: M icrometer # Date of study : 0//06 Percent 00 0 Components of Variation % Contribution % Study Var Reported by: Tolerance: Misc: This gage may have J. Meagher been inappropriate for Measured measuring F lats on Hex Nut this characteristic. Operators measured Measurement the same by Part parts differently. Sample Range Sample Mean Gage R&R Repeat Reprod R Chart by Operator A B C Xbar Chart by Operator A B C Part-to-Part UCL=0.008 _ R= LCL=0 UCL=0.69 _ X=0.68 LCL= Average Part Measurement by Operator A B Operator Operator * Part Interaction Part C Operator A B C

44 Micrometer Measuring Thickness of Flat Washer Blueprint Reading Gage R&R Presentation G age name: M icrometer # Date of study : 0//06 Percent 00 0 Components of Variation % Contribution % Study Var Reported by: Tolerance: Misc: This gage is acceptable for measuring this characteristic. Very little variation due to gage, J. Meagher Measured Thickness of Flat Washer almost all Measurement in part-to-part. by Part Gage R&R Repeat Reprod R Chart by Operator A B C Part-to-Part Part Measurement by Operator Sample Range Sample Mean Xbar Chart by Operator A B C UCL= _ R= LCL=0 _ X= UCL=0.080 LCL= Average A B Operator Operator C had a Operator * Part Interaction different technique than the other two. Part C Operator A B C

45 Before we review Minitab outputs - Time for a good stretch and a quick mental exercise (--). Bottom line use something of what you learned today in the next week - or - you will lose it

46 Variance is like an area 0 by 0 Square => Total Area = 00 Percentage of by Square to Total Area /00 = 0. = % by Square =

47 Standard Deviation is like a side (square root of area equals side) 0 by 0 Square => Total Area = 00 Percentage of by Square to Total Area /00 = 0. = % Percentage of Side to Total Length /0 = 0.0 = 0% by Square The Square Root of = = Length of one side equals

48 Analysis by Donald Wheeler PV AV TV GRR EV

49 From AIAG FAQ s What is the difference between "% contribution" and "% study" in terms of GRR performance? % Contribution is determined by multiplying by 00 the proportion of the GRR variance to the total study variance. % Study is determined by multiplying by 00 the proportion of the GRR standard deviation to the total study standard deviation. Thus, a level of 0% study is equivalent to a level of % contribution (. x. =.0). Not true in regards to proportions!!!

50 Variance vs. Standard Deviation???

51 AIAG admits the percentages do not equal to 00% - without explanation??

52 Distinct Categories another issue? Note: The Number of Distinct Categories is calculated by:. x (σ p /σ m ) or. x (PV/R&R) Another ratio of standard deviations!

53 Semiconductor industry questions AIAG metrics

54 <7 < Distinct Categories can be estimated from the divisions in the Range Chart

55 A preview of what is to follow. This is where we draw the line Gage R&R %Contribution Source VarComp (of VarComp) Total Gage R&R Repeatability Reproducibility Operator Operator*Part Part-To-Part Total Variation These are the Honest Gage R&R results as identified by Donald Wheeler! Proportions add up to 00%. Definition of proportion: A part considered in relation to the whole. Process tolerance = Definition of ratio: A part to part relationship, not to the whole. Study Var %Study Var %Tolerance Source StdDev (SD) (6 * SD) (%SV) (SV/Toler) Total Gage R&R Repeatability Reproducibility Operator Operator*Part Part-To-Part Total Variation Ratios based of values that are square root sum of squares? Note: The value of.88 [(6*Std Dev/(USL LSL)] is the reciprocal of the Cp value! Cp = (USL LSL)/6*Std Dev =.07 (Capability)

56 Standard deviations and normal distributions Note: Tolerance = Density Histogram of Total Gage R&R, Total Variation The ratio between (6*0.66)/ * 00 = 8.8 is the % Tolerance Note: This is the reciprocal of Cp!!! (.07) 0 Data.7 Normal Project: Sample Distributions.MPJ; Worksheet: Data The ratio between (6*0.66)/(6*.770) * 00 = 0.97 is the % Study Variation Variable Total Gage R&R Total Variation Mean StDev N You are still comparing the length of the lines or the width of the distributions - nothing else!!!

57 % Tolerance proportion??

58 Variance vs Standard Deviation From RSD associates web site ( associates.com/mesa.htm)- "Evaluating The Measurement Process" by Donald J. Wheeler and Richard W. Lyday (ISBN X) Wheeler (99) also points to technical deficiencies in the R&R approach. Sources of variation are commonly expressed as a percentage of either tolerance bands or of total variation, but this is done without squaring the values involved, and hence these percentages can be very misleading and will not, of course, add up to 00%. The percentages used in R%R studies can be even more misleading when expressed as a percentage of a tolerance spread.

59 AIAG MSA Manual (Chapter ll Section C on Page 7) States: the sample parts must be selected from the process and represent its entire operating range. If the sample parts DO NOT represent the production process, TV must be ignored in the assessment. Ignoring TV does not affect assessments using tolerance Is this the only option you have left??? What if you have concerns with this metric???

60 So. is there any other way to judge % Tolerance?

61 The following analysis is FOR REFERENCE ONLY!!! This method is no more of a stretch than judging gage capability based on a or 0 pc sample.

62 NHBB Bias & Linearity Chart Linearity Chart for Air Amplifier Gage Nominal Size At every calibration cycle we judge bias and lineraity. From calibtation to calibration we judge stability. If we are using a gage out on the floor, we know it meets our criteria. Variable Minimum Nominal Maximum Actual Standard Size Project: NHBB GAGE LINEARITY.MPJ; Worksheet: Air Amplifier Gage

63 Chart Output Original Data Gage R&R (ANOVA) for Reading Gage name: Date of study: Notch A ngle Open Reported by: Tolerance: Misc: J Meagher Degrees Data as receiv ed Percent Sample Range Sample Mean Gage R&R Repeat Part Reprod R Chart by Operator Karissa Yvette Zeke Components of Variation Xbar Chart by Operator Karissa Yvette Zeke Part Part-to-Part % Contribution % Study Var % Tolerance UCL=0.8 _ R=0.87 LCL=0 UCL=.698 _ X=.069 LCL=. Project: PART TO PART GRR MODIFIED.MPJ; Worksheet: Data Average Karissa Highest.0 Reading by Part Lowest.6 Part Part Reading by Operator Yvette Operator Part * Operator Interaction Zeke Operator Karissa Yvette Zeke

64 Chart Output Original Data Have Gage R&R (ANOVA) we painted for Reading Gage name: Date of study: Notch A ngle Open Reported by: Tolerance: Misc: ourselves into a corner? J Meagher Degrees Data as receiv ed Components of Variation Reading by Part What if we had parts 00 % Contribution.8 % Study Var Highest % Tolerance.0 0. that spanned the tolerance? Lowest.6 0. Gage R&R Repeat Reprod Part-to-Part Part R Chart by Operator Can we get there from here? Karissa Yvette Zeke Reading by Operator UCL= _ R= LCL=0 Karissa Part Karissa..0 Karissa.96 Yvette.9696 Zeke Operator Karissa Xbar Chart.09by Operator Karissa Karissa Yvette. Zeke.6 Part *-0.6 Operator Interaction UCL=.698 Karissa _ Karissa...6 X= LCL=.. Part Part Percent Sample Range Operator Part Trial Reading Overall Part Average High/Low Limit Difference from Limit Forced Variation Sample Mean Project: PART TO PART GRR MODIFIED.MPJ; Worksheet: Data Average Operator Karissa Yvette Zeke

65 Chart Output Forced Variation Gage R&R (ANOVA) for Forced Variaion Gage name: Date of study: Notch A ngle Open Reported by: Tolerance: Misc: J Meagher Degrees Forced Variation Percent Sample Range Sample Mean Gage R&R Repeat Part Reprod R Chart by Operator Karissa Yvette Zeke Components of Variation Xbar Chart by Operator Karissa Yvette Zeke Part Part-to-Part % Contribution % Study Var % Tolerance UCL=0.8 _ R=0.87 LCL=0 _ UCL=.897 X=.706 LCL=. Project: PART TO PART GRR MODIFIED.MPJ; Worksheet: Data Average Karissa Forced Variaion by Part Part Part Forced Variaion by Operator Yvette Operator Part * Operator Interaction Zeke Operator Karissa Yvette Zeke

66 Session Output Comparison Gage R&R - Original Data Gage R&R - Forced Variation %Contribution %Contribution Source VarComp (of VarComp) Source VarComp (of VarComp) Total Gage R&R Total Gage R&R Repeatability Repeatability Reproducibility Reproducibility Operator Operator Operator*Part Operator*Part Part-To-Part Part-To-Part Total Variation Total Variation Note: The ONLY change to the analysis is the Part-to- Part Variation! Since the part variation now equals the tolerance spread, the Total Gage R&R value given estimates the % Tolerance.

67 Session Output Comparison Gage R&R %Contribution Source VarComp (of VarComp) Total Gage R&R Repeatability Reproducibility Operator Operator*Part Part-To-Part Total Variation Forced Variation Original Data Gage R&R %Contribution Source VarComp (of VarComp) Total Gage R&R Repeatability Reproducibility Operator Operator*Part Part-To-Part Total Variation Process tolerance = Study Var %Study Var %Tolerance Source StdDev (SD) (6 * SD) (%SV) (SV/Toler) Total Gage R&R Repeatability Reproducibility Operator Operator*Part Part-To-Part Total Variation Number of Distinct Categories = 6 Process tolerance = Study Var %Study Var %Tolerance Source StdDev (SD) (6 * SD) (%SV) (SV/Toler) Total Gage R&R Repeatability Reproducibility Operator Operator*Part Part-To-Part Total Variation Number of Distinct Categories =

68 Chart Output Original Data Gage R&R (ANOVA) for Measurement Gage name: ADCAMS C/A Gage Date of study : 0/9/009 Reported by : Tolerance: Misc: A. Dorchies to 9 = 6 Degrees C ontact A ngle K Percent 00 0 Components of Variation % Contribution % Study Var % Tolerance Measurement by Part Sample Range Gage R&R Repeat 6 7 Reprod R Chart by Operator A. Dorchies L. Morin H. Oki Part-to-Part UCL=0.970 _ R=0.77 LCL= Part Measurement by Operator 9 0 Sample Mean Part Xbar Chart by Operator A. Dorchies L. Morin H. Oki _ UCL=.70 X=. LCL=.969 Average A. Dorchies L. Morin Operator Part * Operator Interaction H. Oki Operator A. Dorchies L. Morin H. O ki Part Part Project: CONTACT ANGLE FORCED VARIATION.MPJ; Worksheet: 0 Part Random Order

69 Chart Output Original Data Gage R&R (ANOVA) for Measurement Gage name: ADCAMS C/A Gage Date of study : 0/9/009 Reported by : Tolerance: Misc: A. Dorchies to 9 = 6 Degrees C ontact A ngle K Percent Sample Range % Contribution % Study Var % Tolerance Gage R&R Repeat Reprod Part-to-Part Part R Chart by Operator A. Dorchies L. Morin H. Oki Measurement by Operator.0 UCL= _ R= LCL= Part A. Dorchies.7 L. Morin 9.0 H. Oki Operator. Xbar Chart -.9 by Operator A. Dorchies L. Morin -.9 H. Oki Part * Operator.7 Interaction Operator _ UCL= A. Dorchies. L. Morin X= H. O ki.0 LCL= Part Part Sample Mean Components of Variation Average Measurement by Part P/N Current Offset Forced Part P/N Current Offset Forced Part Project: CONTACT ANGLE FORCED VARIATION.MPJ; Worksheet: 0 Part Random Order

70 Chart Output Forced Variation Gage R&R (ANOVA) for Forced Variation Reported by : A. Dorchies Gage name: ADCAMS C/A Gage Tolerance: to 9 = 6 Degrees Date of study : 0/9/009 Misc: C ontact A ngle K For Reference Only! Sample Range Percent Gage R&R Components of Variation Repeat Reprod R Chart by Operator A. Dorchies L. Morin H. Oki Part-to-Part % Contribution % Study Var % Tolerance UCL=0.970 _ R=0.77 LCL= Forced Variation by Part Part Forced Variation by Operator 9 0 Sample Mean Part Xbar Chart by Operator A. Dorchies L. Morin H. Oki _ X=.8 UCL=.8 LCL=.0 Average 8 6 A. Dorchies L. Morin Operator Part * Operator Interaction H. Oki Operator A. Dorchies L. Morin H. O ki Part Part Project: CONTACT ANGLE FORCED VARIATION.MPJ; Worksheet: 0 Part Random Order

71 Session Output Comparison Gage R&R Original Data Gage R&R Forced Variation %Contribution Source VarComp (of VarComp) Total Gage R&R Repeatability Reproducibility Operator Part-To-Part Total Variation %Contribution Source VarComp (of VarComp) Total Gage R&R Repeatability Reproducibility Operator Part-To-Part Total Variation Note: The ONLY change to the analysis is the Part-to- Part Variation! Since the part variation now equals the tolerance spread, the Total Gage R&R value given estimates the % Tolerance.

72 Session Output Comparison Gage R&R - Original Data Study Var %Study Var %Tolerance Source StdDev (SD) (6 * SD) (%SV) (SV/Toler) Total Gage R&R Repeatability Reproducibility Operator Part-To-Part Total Variation Number of Distinct Categories = Gage R&R Forced Variation Study Var %Study Var %Tolerance Source StdDev (SD) (6 * SD) (%SV) (SV/Toler) Total Gage R&R Repeatability Reproducibility Operator Part-To-Part Total Variation Number of Distinct Categories = 0

73 In Conclusion: If you have not altered the gage Repeatability (gage variation) or the Reproducibility (operator variation) and you have only altered the Part-to- Part Variation - you can now judge your gage variation against the total tolerance.

74 In Conclusion: Boxplot of % GRR comparing % Contribution and % Tolerance % GRR A P-Value of for this analysis indicates that % Contribution after forcing variation is significantly lower than either the initial % Contribution or % Tolerance. This was analyzed using nonparametric methods (nonnormal data). Values shown are the medians for each group % Contribution Initial % Contribution Forced Metric Project: COMPARISON OF % TOLERANCES.MPJ; Worksheet: Data % Tolerance

75 In Conclusion: Fitted Line Plot % Tolerance = % Contribution Forced S 8.8 R-Sq 86.0% R-Sq(adj) 8.% % Tolerance % Contribution Forced 0 Note: A 0% Contribution (Forced Variation) equates to a % Tolerance! % Tolerance is times greater than % Contribution when forcing part variation to match the full tolerance. Project: Comparison of % Tolerances.MPJ; Worksheet: Data

76 Forced Variation Comments This procedure should not be used blindly. You must know your product and your gaging system and judge when this method is appropriate. If you consider using this method, you should record what the original results are and then document what the forced variation results are so you can explain the analysis and the reasoning behind the analysis to your customer(s).

77 Closing Comments Any and all comments would be greatly appreciated. An article is being authored and will be submitted to either Quality Progress (ASQ) or Quality Digest to allow the ideas expressed in this presentation to be debated in an open forum.

78 Thank you for coming. Comments/questions? Contact information: