MSA Manual - Page 60 states that "If the analysis is based on tolerance...

[Sam]

Having completed our QS9000 audit and now having idle time on my hands(sic)I have gone back to re-visit the MSA manual.

Page 60 states that "If the analysis is based on tolerance . . . then substitute tolerance for total variation."
I substituted the tolerance,.4mm, into the equation, the result 59% R&R. I would think the two results would have been closer.

Are these typical results when using both options? What would I look for in the analysis to fix the problem?

 

[Marc]

Long, detailed responses are appreciated! The more folks help, the more these forums mean to each of us!

[This message has been edited by Marc Smith (edited 27 March 2000).]

 

[mbruner]

You didn't state what the %GRR results were based on total variation but I guess from the context that the total variation %GRR was less than the %GRR based on tolerance.

First, regarding agreement between the two methods:They don't have to agree and they typically don't. Total variation is a combination of part variation & measurement system variation. Tolerance is the acceptable limits for the part value which may or may not have some relation to part variation.

Generally you would hope that total variation is less than tolerance and that measurement system variation is much less than part variation.

In this case %GRR based on tolerance would be less than %GRR based on total variation.
In my experience at my company this is typical.

What to look for in the analysis?
Make sure you use the correct value for tolerance.

If GRR% tolerance is greater than GRR% total variation the first item I check is the value for tolerance. GRR as percentage of tolerance uses the tolerance span. If the tolerance is +/- 0.4mm then the value that should be used in the calculation is 0.8mm not 0.4mm(0.8mm/sqrt(EV^2+AV^2).

If the tolerance is a single sided spec, then the value for tolerance span is calculated by abs(2*(PartAverage-tolerance)).
In this case the average measured value for the part influences the GRR result. The closer the part average is to the limit(tolerance) the smaller the tolerance span will be and thus makes tighter constraints on acceptable measurement system variation.
If you can't reduce measurement system variation to an acceptable level in this case, check to see if the part average can be brought closer to nominal.

I didn't mean to post such a long response. I don't profess to be an expert. I guess I'm just therapeutically bearing the scars of past mistakes.