GR&R by MSA Rules vs. Wheeler & Lyday's EMP Method

[Steve Berian]

We recently did a couple of Gage R&R's on two measuring systems: the first measures Specific Gravity, and the second measures Durometer, or the Hardness of plastics. We got some rather strange results on both, and if believed, it would appear that neither system should be used. We re-did the studies and cranked the calculations through "QI Macros 2012, GR&R to MSA 4th ed".

The Specific Gravity's study is a real head-scratcher because the software admonishes us: "Gage system needs improvement; Gage may need maintenance, redesign, or better clamping; & Not enough Part Variation for Study".
We apparently violate both of these rules:
Rule: The 10 parts should cover the Spec Tolerance (e.g., 0.4995 to 0.5005), otherwise identical parts will cause high %R&R. If NDC (Number of Distinct Categories, F52) is less than 5, there is not enough part variation to enable an accurate study.

We used 3 technicians, each performing 3 trials on 10 different parts.
I can send you the study, if you're interested.

The Durometer Study was done similarly, and we got the following results:
Percent R&R = 100 [ R&R/TV ] 57.3%Percent Equipment Variation = 100 [EV/TV] 53.6%Percent Operator Variation = 100 [AV/TV ] 20.3%Percent Part variation = 100 [ PV/TV ] 81.9%Percent of Tolerance by R&R = 100 [ R&R/Tol ] 172.9%

1- Any clues as to what we are doing incorrectly?
---
2 - Do any of you have experience w/ doing the Wheeler & Lyday method?
3- If yes, would this method produce results different than these MSA results?
4- Will TS-16949 auditors accept Wheeler & Lyday method results, in lieu of the MSA method?

Thanks for your consideration & help.

Steve

 

[statdoug]

Hi Steve,
1) it sounds like your 10 parts were too similar and the noise of the measurement system obscured what differences there were between the parts.
2-3) The Wheeler method will yeild different results, but it won't solve your system problem. It deals with probable error, and Wheeler makes compelling arguments both empiracally and statistically, that the AIAG methods (used as the basis for most software) are incorrect of their assesments especially in terms of % of tolerance and spread. It does not change the fundimental nature of the study of the system.
4) I don't know how auditors will react.

P.S. I would be very interested in looking at your studies.

 

[Steve Berian]

Hi Doug -
Thanks so much for your response. Please advise your email & I will send you copies of our studies. In our business [we make plastic compound & engineered polymers; we ship loose plastic pellets], we strive to make our materials as consistent as possible, w/minimal batch to batch variation. It would seem, by these studies anyway, that we are not only "succeeding", we're "too good"! Somehow, I doubt that.
Best regards,
Steve

 

[statdoug]

I sent my e-mail by private message. I drive up 91 pretty regularly from southern CT to my home in northern NH and probably go close by.

 

[Miner]

For some more information on sample selection and on Wheeler's method, see my blog on MSA(parts 5a, 5b and 7). These posts include files for both methods. Note: even though the file is to the 3rd edition, the methods for R&R did not change.

Be careful about that rule from QI Macros. If you are assessing % Tolerance, that rule will neither help nor hurt you. If you are assessing % Study Variation or ndc the parts MUST represent the actual process variation. Following that so-called rule will usually inflate the study variation and result in artificially good results.

The high results for durometer may be caused by the resolution of the gauge depending on how tight your tolerance is relative to the units. Is this a Shore D scale durometer that measures in one unit increments? What is your specification?

 

[Curtis317]

Just out of curiosity are you doing a "crossed" or "nested" GRR? What you are measuring seems to be something that is damaged with the measurements. The Durometer measurements are a good example. You cannot get the same reading measuring the same spot with a durometer. A Nested study would be more appropiate if you have Minitab.

 

[anandqgp]

Hello friend...This study is interesting. Please share the study with me on [email protected] and I would come back my inputs. Thanking you.

 

[anandqgp]

Hi Steve, I am interested in the case studies. Please do send them on [email protected]. I shall be obliged. Thanking you HS Anand

 

[Steve Berian]

Hi Steve, I am interested in the case studies. Please do send them on [email protected]. I shall be obliged. Thanking you HS Anand

Hi Anand - Thanks for your post & my apologies for not replying sooner. I have been away a long time. In fact, left the firm where I had that problem back in 2Q2016, so, sorry, no longer have access to the data. As I recall, we declared that the gage involved was as good as it gets, and moved on to bigger fish. Ext auditors, at the time, didn't fuss. Be well, & thanks again. Steve