Operating Range - Definition for Gage R&R Study - Micrometers measuring thickness

[Sean Kelley]

Hello to everyone and thanks for all of your help. I am trying to get a consensus on what exactly "operating range" is defined as. We recently found our part-to-part variation to be very high. Frequently as high as 99.98%. This is for micrometers measuring thickness of our steel (we are a steel mill).

We recently found out from a valued cover that his interpretation of this was that we used different parts for our study. This was indeed true. We were under the impression that operating range meant 1)operating range of the measuring device or 2) total operating range of your production. We make .012" to .5" thick steel. So we followed what we thought was the correct method and selected 15 sample from .012" to .5".

My recent enlightenment was that we should have selected the 15 samples from the .012" as an "operating range" and another 15 samples from the .5" and some points in between.

My biggest concern of this is that when I go to the AIAG MSA manual 3rd edition on page 101 it does not appear they have followed this concept. The concept makes sense to me that it would definitely give much more GRR error and if you were under the 10% GRR error they wantyou would definitely have a fantatic measuring system.

Any help on clarifying this is much appreciated.

 

[Bob_M]

Hopefully someone else post a more "exact" explaination but...
-------
BASIC Gage R&R studies (that we do) are 3 people, 10 identical parts (in your case thicknesses), each person measures 10 parts in random order 3 times = 90 readings.

Now if you want get R&R for other thicknesses start over.

This measures the variance of the parts (thickness) and the operators capability to properly use the gages.
-------
I don't have a copy of the manual, but IS there a test that covers mulitple thickness for a single study as Sean Kelley TRIED?

 

[howste]

I believe what it should be is the tolerance range for an individual part. If you are doing an R&R for .012" ± .002 thick steel, you would select parts that cover the range of .010" to .014". This gives you the measurement system variation when you measure that size of part. You would do another study for .5" thick steel.

Also, as Bob stated, usually it's 10 parts, 3 operators, 3 trials each.

 

[Bill Ryan - 2007]

I agree with Bob and Howste.

Regarding the .012 +/-.002, the "best" confidence in the study would be a range of differing thicknesses within the tolerance band - even a piece or two out of tolerance only adds to your confidence level (it's just as important you are confident with an out of tolerance measurement as with one that is in tolerance).

By redoing a study at the .5 specification (with its tolerance), you would have a level of confidence that your micrometer can "handle" the entire spectrum of your materials. I'm not that learned in linearity studies (yet), but it would seem to me that acceptable studies at each end of the spectrum would "prove out" the linearity of the micrometer.

We go through these dual R&Rs fairly often especially with all the transducers and air gaging we have in place.

Hope that helps some.

Bill

 

[chalapathi]

The R&R study is for the Gage and not for the Manufacturing process. You may have many products in the thickness range .012" to .5" as far as the manufacturing process they are different parts. The actual process conditions, setup, rolls etc could be different.
When it comes to measuring the thickness, you are using the same micrometers for several products. The R&R is required for the micrometers and not for rolling mill. The way you have taken several parts (covering different products) is a good and practical way of doing R&R study.
This is the method I followed in the past 12 years.

 

[Jim Wynne]

The R&R study is for the Gage and not for the Manufacturing process.

Not always true; sometimes the part is a component of the measurement system being evaluated. It is true that MSA may be done on "families" of gages where the measurements to be made are more or less generic, but there are also times when it's dangerous to assume that one size fits all, and the study should be part-specific.

 

[chalapathi]

R&R for Families of Products

Not always true; sometimes the part is a component of the measurement system being evaluated. It is true that MSA may be done on "families" of gages where the measurements to be made are more or less generic, but there are also times when it's dangerous to assume that one size fits all, and the study should be part-specific.

Your views are right. I am sharing a point which is not given/discussed in the MSA manual. I have seen auditors and clients are so particular and it is a real pain for the suppliers to repeat G&R of the same instrument for each and every product.
To take care of the concern you are raising, we can look at two ideas:
1. Involve the customer in taking decision.
2. Group the products into 2/3 families and do it for the families. In case of steel sheets - Low thickness and high thickness.

 

[Jim Wynne]

1. Involve the customer in taking decision.
2. Group the products into 2/3 families and do it for the families. In case of steel sheets - Low thickness and high thickness.

Both are good ideas, especially #1. Never surprise the customer. Find out what's expected, make sure it's in writing, and then either do it or offer the customer a better alternative, with the understanding that as far as the customer is concerned, "better" usually means "less expensive."

 

[Atul Khandekar]

....2. Group the products into 2/3 families and do it for the families. In case of steel sheets - Low thickness and high thickness.

I am confused. Do you mean mixing parts with different nominals in one GRR study? If so, you are artificially increasing the Part Variation. This is a sure way of getting good (low) GRR numbers!

 

[Miner]

I am confused. Do you mean mixing parts with different nominals in one GRR study? If so, you are artificially increasing the Part Variation. This is a sure way of getting good (low) GRR numbers!

I agree. This would invalidate the GRR as a percent of process variation.