Gage R&R Measurement Location Methodology

[ChiliDon]

This is a wonderful forum to debate what I have to say/ask.
I have been performing Gage (Gauge) R&R for many years. I have created my own Excel Spreadsheets and now use software to do the analysis. I was originally training by the VP of education at Mitutoyo. I have had the opportunity to teach the ASQ Refresher courses where I cover the specifics of the measurement methodology.
I want to point out that I try to be a purist in data gathering which brings me to the debate.
The MSA requires that you use the normal measurement process when taking the measurements. Fundamentally I agree with this however I was taught that when creating the measuring study that you isolate the measurement location and take a single measurement at that location for each trial and each operator. In this case I am measuring painted panels and using a colorimeter for the L, a*, b* values.
The current measurement requirement is to measure the panels in 4 places and report the mean. However I instructed the QE to use only one of the locations measured for the GR&R. Another QE says that I am wrong and the Mean should be used since that is the reoportable measurement requirement. I argued that by using the mean you are smoothing out the measurement data and that is not what you use a GR&R for. It is used for the Measurement system and not the Process Variability in each Painted Panel. Once you have accertained the Measurement Uncertainties and approve the GR&R, the Measuremtn System is available for continued use for that application, Now you can accept the data collected for all 4 data points and you can analyze the variability within each panel. If it is acceptible then the Mean value has validity with Measurement Uncertianty understood.
My QE counterpart argues that by using the Mean value in the GR&R you are usingthe normal measurement process and that Mean value is the only value to qualify the the GR&R For that process.
I would only agree if the dispersion in the panels is very small but in a manual painting process paint coverage on each panel will vary.

Are there any helpful guidance out there to my debate?

 

[Bev D]

well, it depends.

it depends first on the critical nature of the characteristic & object.

If position matters then we shouldn't be averaging the measurements in the first place. If you need all positions to be in spec or in the case of color the user will disconcerted by color variation across the object then your approach is the correct way to perform the R&R

If 'within piece' variation or 'across piece' variation doesn't matter then the R&R for that gage and that characteristic on that object can be performed on the average.

However, I always challenge this as too often people will try to 'overcome' measurement error with averages and/or resort to an 'average' spec because they think they can't reduce the within piece variation because it's inherent in the process/industry or whatever. (I wish I had a dollar for every time I heard this excuse). Too many people do this without really understanding the true need of the final assembly or user. they are doing it for their convenience...