Ford - Type 1 Gage Study MSA (Measurement Systems Analysis) acceptable?

[gmantyk]

:truce: Have a fully automated end of line tester - performing through MiniTab a Type 1 gage study. Performing 45 tests on one sample. 100% confident what is being done is correct.

Problem being this study (which I've always referred to as Gage R) is not being accepted. Told "The Gage R is based on a European standard (and I use the word standard very loosely) which does not meet Ford or AIAG requirements." Which leads to my question - where is the evidence to support a Gage R / Type 1 gage study in either MDS or Ford acceptable method for validation on an automated system?

The method the Ford person is using is 30 samples x 5 test x 1 operator and calculating the number though MiniTab Gage RR. I know this isn't correct and doesn't work for the type of tester but stuck at an impasse. Any help would be greatly appreciated. Glinda

 

[bobdoering]

Re: Ford - Type 1 Gage Study MSA acceptable?

According to the AIAG MSA rollout, Gage R for an automated gage is acceptable if accompanied by another test, usually stability (which ends up being doing gage R on a different day or shift). That response was from the Ford representative Patrice White-Johnson on the AIAG panel, and it does meet the AIAG requirement (page 205 of MSA 4th ed) if accompanied by the other studies. (...and they think we don't pay attention at those roll-outs!)

Any time the customer disagrees with your methodology, they are responsible to provide you with a valid alternative. Often, tehy are at a loss for that, too.

 

[gmantyk]

Re: Ford - Type 1 Gage Study MSA acceptable?

Thank you! Would of helped if I had been using the 4th edition

Back on track - the line This method CANNOT be used for final gage acceptance without other more complete and detailed MSA method. Any knowledge what other MSA methods they're looking for? It's pretty open ended. Glinda

 

[bobdoering]

Re: Ford - Type 1 Gage Study MSA acceptable?

What I mentioned above was specifically the answer Patrice gave to your specific question (because everyone else there had the same question in mind). She indicated stability would be the other additional appropriate tool.

However, if there are separate nests or fixtures that present the part to the test unit, you might consider them the same as "operators" and look for the effect of different nests as you "between" (AV) variation.

 

[gmantyk]

Thank you for your response, it was a huge help.

 

[michiganmike]

Intersting that many call it the AIAG MSA Requirements Manual. Look at the Foreword in the front of the book. "This Reference Manual..."

Also on that page: "It is not intended to limit the evolution of analysis methods suited to particular processes or commodities". I take that to mean that the methods in the book are one way, but not necessarily the only way, to assess the suitability of a measurement system.

 

[NumberCruncher]

Intersting that many call it the AIAG MSA Requirements Manual. Look at the Foreword in the front of the book. "This Reference Manual..."

Also on that page: "It is not intended to limit the evolution of analysis methods suited to particular processes or commodities". I take that to mean that the methods in the book are one way, but not necessarily the only way, to assess the suitability of a measurement system.

Hi michiganmike

That's correct. Auditors and companies seem to treat the AIAG manual as a standard. It's not. Extremely informative? Yes. Best Practice? Perhaps. A requirement if you want to work in automotive manufacture? Probably.

But it's not an ISO standard. Nor does it cover all eventualities.

For example, I am currently working on an attribute study where I could have anything from 0 to 10 different possible faults on a single part. The attribute study in the AIAG manual implicitly assumes you are only looking at a single fault. How do you assess the overall performance of inspection when you have different faults, with different levels of obviousness? It seems unreasonable to treat all faults as the same. They are not. Some faults are more serious, some faults are more obvious.

Do I generate 10 different kappa studies (p136)? The manual doesn't cover such a situation explicitly. I'm still trying to work out the best way of doing this that will satisfy
a) the auditors
b) management
c) the staff who do the inspection

I'm now trying to "evolve" a suitable measure.

NC

 

[supreecha]

Type I Error & Isoplot