Test Stand R&R - Normally accepted Gage R&R techniques on transmission test equipment

Am having difficulty using normal accepted R&R techniques on transmission test equipment. Many variables, including, temperature range, changing oil viscosity - make it difficult to achieve repeatability during consecutive testing of the same unit.
I am considering combining static R&R of individual gages and combining this with an ANOVA study comparing various sample units, personnel, and test stands.
Has anyone out there experienced anything similar and maybe have a suggestion?
Thanks.

[Jerry Eldred]

I don't have specific experience in that area, but if I interpret your question correctly, it seems that you are having difficulty performing an R&R because the devices under test have so much variability, and not the test equipment. I don't have the QS9000 MSA numbers on the top of my head (I am a metrologist, but don't claim to be so much a QS9000 guru). But it seems that the capability of your transmission test equipment is required to be adequate for the measurement. If you have that large amount of variability in the various parameters, and units are meeting specs, then you need to prove that your test equipment is adequate to assure they meet the specs.

Have you tried taking some sets of numbers (I don't even know what parameters, ranges, specs you are testing, so I am in the dark to that degree)? If indeed your parameters have that wide a set of specs, is it possible that if you were to try taking some data and calculating Cpk, etc, that perhaps you might get acceptable numbers? This is a deviation from my traditional calibration lab instincts, which tell me to look only at the specs of a piece of test equipment and calibrate it. But in the case of GR&R (MSA), I look at the spec limits and control limits of the device to be tested and verify the capability in both repeatability and reproducibility to verify that device meets those specs.

I may well be out of line on this one, but that is my first stab at an answer. We can most certainly take this to more of a QS9000 guru and get more definitive answer in those terms.

Hope I was of some small help.

------------------

Thanks for the reply. I think you are on to something. Combine capability to validate individual test machines and use anova or correlation to compare the 5 machines I must cover. Thanks again.