Spread for Gage R&R - Product specific or whole test range of instrument?

[puck1263]

Hello -

My company wants to do a "gage R&R" for a particular test (to qualify a new tester at another plant) and what they've done in the past is to select 6 samples from 5 different products (30 data points). The 5 different products spanned the operating range of the tester.

In this case, the GR&R looks great because of what it is with respect to. To me, this isn't valid...GR&R should be with respect to each product if I'm going to make a decision based on each product. It doesn't tell me anything about resolution with respect to being in control or in spec for a given product.

To me, this is an ok first step if you are qualifying an additional tool to make sure it is reasonably in range of expected, prior to product specific testing.

[Miner]

You are correct to be concerned. Please read my blog entry on Gage R&R, specifically the section on Part Selection. This should answer your questions.

If the new tester is the same design as an existing tester that has an acceptable R&R, you may want to consider comparing the two testers using the same parts. See this blog entry on comparison of two gages.

[puck1263]

Thanks for replying.

Most customers do not ask for P/T or P/V, they ask for %GR&R such as in PPAP package.
Giving a % is ambiguous. No one asks what's in the denominator. Many assume that the part-part is part-part of the same kind (like part #), not part-part of different kinds. We use our whole product spectrum and pass it off as if it were multiple parts of the same product, because that is what customer assumes. All doing a study like this says is "congradulations, you can tell different products apart." but doesn't say much about ability to distinguish within that product part-part.

Does it say anywhere that it should be with respect of part-part of the same product?

[Miner]

%GR&R is typically interpreted as %Contribution, which is specified as the % of process variation. %Tolerance is the P/T Ratio.