Gage R&R for two-operator measurement

[Tim Folkerts]

I'm working on a little project where a measurement is made on a large part. The procedure requires cooperation of two technicians - one on each side. There hasn't been any formal request yet for a GR&R, but I was wondering if anyone has dealt with such a situation.

We could just use three pairs of operators and go ahead with the analysis as if it were three single operators for the purpose of analysis.

On the other hand, I expect that one half of the measurement will have a bigger effect on R&R. I could keep the operator on one side constant and cycle through 3 operators on the other side and do GR&R on that part of the procedure. Then do a second R&R keeping the other half constant.

Or I could come up with some more complicated permutation of operators and then try to determine the best way to separate out the effects of each operator on each side.

Has anyone seen a standard approach for such a system - where multiple operators cooperate to perform a measurement - or do I need to just develop one up myself?


Tim F

 

[Bill Ryan - 2007]

Sorry Tim. I've never come across that scenario and with my "expertise" in statistics I can't suggest anything different than you're two possibilities.

FWIW - I do think, however, that I would keep the operator on one side constant and cycling through all three - then doing the same on the other side.

Sorry I can't be of more help.

 

[Barbara B]

Tim,

are the two operators independent in your opinion or do you assume an interaction between them?

If they are independent, you could simple use a block-design in which one operator on side A is fixed (like you described it). If measurements probably depend on the first operator measurement, you have to randomize all operators to track the single effects.

Sorry, I haven't seen a standard dealing with this issue. But with your DoE-background it should be feasible to develop one, or? I'm interested in your further steps/ideas.

Barbara

 

[Tim Folkerts]

Thanks for the input. I expect I may just have to design my own approach and work out the statistics. Oh, well! It shouldn't be that hard.

The measurement involves aligning two optical devices. Sort of like a surveyor and the assistant who holds the marker at the other end. I expect that one side has a much bigger affect on accuracy and variation than the other. If the person on the critical side does a poor job, then it probably won't much matter what the other person does.

This is actually for a potential consulting job and I am currently doing background work. I specific goal is to improve the actual device, but the more I think about it, the more I need information about the current R&R of the gage. My guess is that it hasn't been done - perhaps that could be an additional consulting job.

Tim F

 

[Jim Wynne]

Thanks for the input. I expect I may just have to design my own approach and work out the statistics. Oh, well! It shouldn't be that hard.

The measurement involves aligning two optical devices. Sort of like a surveyor and the assistant who holds the marker at the other end. I expect that one side has a much bigger affect on accuracy and variation than the other. If the person on the critical side does a poor job, then it probably won't much matter what the other person does.

This is actually for a potential consulting job and I am currently doing background work. I specific goal is to improve the actual device, but the more I think about it, the more I need information about the current R&R of the gage. My guess is that it hasn't been done - perhaps that could be an additional consulting job.

Tim F

It would be interesting to hear your eventual solution. It's a novel problem, and one of those that doesn't lend itself to the conventional GR&R methods. It illustrates the idea that GR&R, as we normally think of it, is a rather limited methodology and one size definitely doesn't fit all. It also serves to point out that there is significant overlap in what we tend to think of as specialized tools; sometimes the best solution requires innovation in setting up experiements, and my hat is off to you for not being a slave to convention.

 

[Atul Khandekar]

Tim, curious to know what was the outcome.

 

[Tim Folkerts]

Atul,

The project has been delayed and I'm still not sure if it will ever actually go forward, so there is nothing new to report. I'm expecting to hear within a week or two (actually I was expecting to hear in the last wekk or two) whether we will be proceeding with the project. If it is a go, I'll get back once we are a little farther along with the project and measurements.

Tim F