Has anyone done a Gage R&R for Spectrophotometer?

[MasterBB]

Has anyone done a Gage R&R for Spectrophotometer?

If so, how did you do this?
Lessons learned, tips, comments, suggestions, etc.

Thanks

 

[Marc]

Have you read through this discussion? Gage R&R studies for Spectrophotometer & Gloss Meters

 

[Bev D]

Sure have. Treat it like any 'destruct' test approach. I do use the Youden approach and not the AIAG method of course...

And yes, I have found that the operator can matter...

Is there anything specific that you have questions about?

 

[Ron Rompen]

Yes, I have done it several times. The biggest lesson that -=I=- learned in this was to duplicate the conditions of the test as precisely as possible. Same ambient lighting, same procedure for mastering the instrument, same area on the part measured each time.

 

[Bev D]

you might find this tutorial for functional and NDE MSAs of interest when designing your study.

Also you might find this paper on why the Youden approach to MSA is superior to the AIAG method

 

[MasterBB]

Thanks Bev,

Do you have a canned Excel file for the Youden MSA Approach.

Many thanks.

 

[Miner]

I know Bev D has the same material somewhere here in the Cove, but it was easier to locate on QFO. Here is a link to her spreadsheet.

 

[Bev D]

Thanks Miner.
It is also here at the cove (MSA Tools for Elsmar)

 

[MasterBB]

Hi Bev,

When you say Destructive Testing; you mean the regular MSA (Gage R&R Nested)?

Or is it satisfactory to do only ISO plot by Shanin / Youlden plot?

 

[Bev D]

Before trying to do any math it is critical to first plot the data and think about it.
The best plots depend on the nature of the study design.

If you are only performing 2 measurements per 'unit' then a Youden plot is quite helpful. The specifications can be plotted as a square on the Youden for comparison purposes.

If you need to have more than 2 measurements per 'unit', then a multi-vari is the best plot.

Usually I don't calculate the math as it isn't all that helpful - visualizing the measurement error (even when confounded by 'unit' variation as in destruct, NDE and functional tests) vs the specification limits is sufficient to understand your measurement system and it's effect on Acceptance sampling and SPC. In fact this approach is far more insightful than 'blindly' calculating a mathematical summary statistic.

The type of study (nested or crossed) is useful only when determining which mathematical formulas to use...although technically a destruct test MSA is a nested study.

It might be helpful if you could provide the details of your study design a don teh results to demonstrate the various approaches...