MSA study types and questions

[Jayfaas]

Hello all. I have two questions.
1) What are some of your requirements for what types of studies are done and how often? Does the IATF 16949 specify frequency and types of MSAs? I am thinking that our parent company does MSA Type 1 and Type 2/3 annually for each different machine which is very time consuming, albeit maybe required by the standard.

2) When running MSA studies, I would assume they should not be interrupted by production parts, but since some machines run so often and they are difficult to take down, would it be against any rules if we were allowing an operator to run their parts in the middle of our study if it is the exact same part? I could completely understand not wanting to do changeovers in the middle of an MSA study, but if its the same kind of part, I dont see why it would be a problem, but would like to know your thoughts.

 

[Ron Rompen]

1) IATF 16949 does NOT specify frequency or type of MSA studys. There have been several discussions here regarding how often you do an MSA study/why you should have to repeat one. The type of study you perform is based on the overall measurement system you are using (automated/manual/semi-automated), etc.

2) If you need to run production parts in the middle of your MSA study, is it feasible to incorporate those parts INTO the study? Using 'real world' data is (in general) always preferable to using theoretical/customized parts and data

 

[Jayfaas]

1) So it is up to the business or are there specific guidelines based on certain characteristics of the gauge?
2) We could do this as well but usually just grab 10 parts from a current run and have them ready and marked when we start. If there is nothing strictly prohibiting this, then I feel like its a defendable argument that during our study, there was a part or two ran that did not require a changeover and therefore should not have any notable influence on the study.

 

[Bev D]

Any characteristic that is on the control plan, critical characteristics and any final acceptance testing should have an MSA
If the measurement system is a go, no-go or pass/fail type of system then an attributes study is appropriate
If the system provides continuous data (aka variables data) then a continuous study is appropriate

You would only ‘repeat’ a study if a new system is introduced or if there is a change to the characteristic itself.

 

[NotMe]

If your standard operation includes changeovers, then I would highly recommend that you include them in your MSA study, as well. As was stated before: Try to be as close to the real operation as possible.

 

[Jayfaas]

This does not include changeovers for every part. I just meant in the middle of a batch run. I know that if we are running our study and they have to do a changeover to another variant during our study, thats not going to work because then you are moving things around and re-mastering the machine to different values, etc.

 

[Bev D]

So when you say “re-mastering the machine” do you mean re-mastering the machine that is making measurements? Or re-mastering the machine that making parts?

 

[Johnnymo62]

Ford used to used to recommend this parts spread for attribute GRRs. I'm not sure if they still use it. Their point is to check the spec limits because they are areas of risk. The center is: 30% of teh parts in the normal process range.

 

[Ron Rompen]

Ford used to used to recommend this parts spread for attribute GRRs. I'm not sure if they still use it. Their point is to check the spec limits because they are areas of risk. The center is: 30% of teh parts in the normal process range.

View attachment 30935

I remember that diagram all too well And yes, as of 2 years ago, they were still using it. I agree completely with the concept behind it, but complying to it (making parts that are almost but-not-quite outside the tolerance limit can be a REAL challenge.

 

[Bev D]

Yes doing it without creating defects is almost always impossible until production is underway. Marginal characteristics can be made naturally IF are doing real development of specifications and performing real OQ. But most companies don’t do that.