CREED
4th April 2007, 09:53 AM
I am trying to find out if anybody nose where i can find the answer to this question. on a gage R&R study the results for GRR is .0005. My Questions is What does this number tell me. I think it says that .00025 from the low limt of the tolerance is suspect and .00025 from the high limit is suspect. Let's say the %GRR is 15.0, My customer wants me to reduce the tolerance by .0005 on each side and i think this is wrong, but i can not find it in balck and white.
Any help would be greatly appreciated.
fireonce
5th April 2007, 03:05 AM
The number tells you that your result of Gr&r is very very excellent.
Atul Khandekar
5th April 2007, 08:32 AM
I think the customer is asking you to do what's known as "guardbanding" the specs. While you take steps to improve the measurement system, this is one way of protecting the customer from the risk of shipping bad products.
CREED
5th April 2007, 08:39 AM
Thats's exactly what the customer want us to do, but my problem is he wants us to cut our tolerance by .0005 on each side of the tolerance. And i think we should cut the tolerance by only .00025 on each side but i can not find this in writing any place. That is really what i'm looking for some place that this is written down in black and white, so i can argue the point, at least internally.
Atul Khandekar
5th April 2007, 08:49 AM
If you have a GRR error of 15% (of tolerance), you would be expected to reduce your spec by 15% on either side: total reduction of 30% !!
I remember there was a paper submitted to the Reading Room forum here that discussed this concept.
Miner
5th April 2007, 10:40 AM
Not having seen your data or analysis, the 0.0005 is either one standard deviation or a multiple of it, such as 3 or 6. Standard deviations are not additive (or subtractive), so you cannot simply subtract it or a multiple/fraction of it from the tolerance.
I recommend that you consult the MSA manual and create a gage performance curve. This will plot the probability of accepting the part at various dimensions compared to the tolerance. You would then establish the guardbands at the dimensions where the Probability of Acceptance drops below your established criteria (such as 95 or 99% or even 100%).
Post your data and tolerance, and I can plot the curve for you.
