Evaluating GRR Scores %Tol vs %TV

[Christy K]

I work for an IATF 16949 registered company. I understand the difference in the %R&R(Tol) and %R&R(TV) scores, according the AIAG's MSA (4th edition). My question is, if a gage is used for part conformance decisions only, is it OK to ignore the %R&R(TV) score when conducting a Gage R&R?

For example, we have conducted a study on a coatings thickness gage. The %R&R(Tol) score is 28.53%. The %R&R(TV) is 87.11%. In our facilities, this gage is only used to determine product conformity (not used to determine control limits or make process adjustments via SPC). Do to the cost of upgrading, we are calling 28.53% acceptable. Can we ignore the 87.11% %R&R(TV)?

 

[Miner]

Your interpretation is correct. There is a section (top of page 123) where it does allow you to evaluate the purpose for which the gage will be used then use the appropriate metric to assess acceptability.

 

[Christy K]

Thank You Very Much! You don't know how long I have been trying to find somehting like that in the MSA book.

 

[Miner]

There are a lot of items similar to that that provide leeway, but they are often placed in sections that are not obvious.

 

[Christy K]

Thanks Again!!

Not only are they not obvious, but the copy that I have has the wrong page numbers listed in the index.

 

[Golfman25]

Good luck. What does your customer require? I have seen several times you must "comply" with 10% R&R requirements without distinction, or otherwise have an act of Congress allowing something else.

 

[WCHorn]

I suggest that the %R&R (TV) is for process control applications (SPC). Imagine if you were keeping a control chart and the gage took up 87% of the variation you see! I believe %R&R (Tol) is for inspection only to determine conformance.

 

[educlos]

If the GRR (TV) is high, it will cause problem of detection with an individual control chart. But if you use an average controle chart, measurement variability will be balanced by the efficiency of the average.

 

[bobdoering]

"The %R&R(Tol) score is 28.53%. The %R&R(TV) is 87.11%. " This makes me a little suspicious that the variation in the samples supplied to the study was very small. If so, it may not represent the variation the process will see over time, and will make the study look much worse. That is why using the process standard deviation instead of the sample variation for the gage R&R calculation is a preferred method. If you do not have enough process history to calculate your process standard deviation, I about to tell you something that no one else will tell you, although it makes incredible sense. If your process is "in control" a good estimate of the total variation should be about 75% of the tolerance. Sure, it could be much better - but for a new process you may not know that for months (different lots, operators, etc.) so, it is a good estimate. Use this as your estimated process standard deviation: ((USL-LSL)*.75)/6). See if the gage r&r output makes more sense.

 

[educlos]

Using the% GRR (TV) with the data of the R&R study is usually a bad idea because the sample of parts used does not cover the entire tolerance range (simply because the process capability is good and user cant select other parts in the production). This apporach is not robust. The good idea is to use the real process long-term sigma to do the calculation. By default, if there is any doubt about the representativity of the parts, it is strongly recommended to only interpret the% GRR (Tol)