CMM and GR&R

[Jonathan S]

I have seen discussion on here about GR&R on CMM. What I recall is that it was agreed that CMM will not pass a GR&R. Can you refer me to any credible sources to support this so I can respond to a customer request for GR&R studies on the CMM?

[Paul Sarrach]

A CMM is a great tool for GR&R. I will do a search, but how else would you do a GR&R?

[Underspec]

anything automated is great for GR&R

[Jonathan S]

A GR&R is supposed to use 3 operators. I am the only one here who runs the CMM. I also do not use fixtures.

[Hilton Roberts]

One normally views a Gage R and R as a test to determine the suitability of a gage to reproduce and repeat a set of measurements when used by different operators.

One might have 2 or 3 operators and as many as 10 parts to be measured in 3 trials.

The CMM by virtue of running a program, removes much of the variation.

I don't know how well one could perform a true Gage R and R on a CMM.

I suppose one could have 3 operators who have each written a program to perform a measurement and one could find 10 parts and have the 3 operators measure the parts in random order.

It seems to me that rather than a Gage R and R, you would be better served by using a sample part that you measure over time to determine how well your machine performs.

Just my $0.02 and worth every penny.....

Hilton

[gage guru]

Just had a similar situation - it was for our gear checker. The best way is to actually do a Gage R there is no operator. If you have access to the MSA from the AIAG group check out appendix D. I also have a contact that I have found very valuable as he teaches about GR&R's etc. He is also on the committe for the MSA 3rd edition.

[Hilton Roberts]

I think that is pretty much what I am saying about a sample part that you measure over and over.

The beauty of CMMs is that once a program is written, it executes the same way every time ( mostly ) so the data reported consists of the same outputs for the same inputs that are modified by the influence of the variables present ( vibration, temperature, part clamping etc ). Some of these variables will show more range over time ( long term ) and some will show shifts in averages when data is taken over the short term and again over the long term.

Careful charting can give you a heads up when things are going away on your machine and frequent charting over the short term ( baselining ) can be done both before and after calibration to see if the machine has changed.

Careful baselining can also tell you when you need to calibrate and can provide valuable data supporting interval extension of calibration.

We are planning to buy a new machine soon and we already have the program in place to test this machine at the factory and again when it is first installed here. This will be very interesting. I have all the data on our two Xcel 7-10-7 machines and I am anxious to see how the new machine measures up.

Hilton

[Paul Sarrach]

A GR&R is supposed to use 3 operators. I am the only one here who runs the CMM. I also do not use fixtures.

You can teach anyone to load the part, even your boss , as the other operator. And you can do it with 2 operators. there for chyrsler they accecpt. 2 5 5 2 operators 5 parts 5 times

[Jonathan S]

Thanks.

[Paul Sarrach]

I am in Stampings. The big 3 engineers that I deal with want it on a cmm. What Gage R&R is checking is that the part when put on and taken off will seat the same way, each and everytime. You are not checking a part, you are checking the gage to see if it was designed and built correctly. That is why you first should do a GAGE R 1 operator, 1 Part Checked 10 times. If it fails the GAGE R, It can not pass a GAGE R&R. I think if you had 3 programmers make three different programs, you would be checking to see if the programmers where repeatable not the CMM nor the Gage. The CMM is in my eyes the perfect tool for such a test. Now if there is enough SPC ports across the gage in each axis, that would suffice.

But I have been wrong, way to many times. What other methods are there to do a gage R&R?

[Matthew D. Hoedeman]

Well, I think some people might be missing the original intent (or maybe it's just me).

I think he is asking about doing an R&R ON the CMM to see if IT, as a measureing device, will pass an R&R. If your customer is looking for a study done ON the CMM (not with it), just whip up a program to check all 6 directions on the machine. Locate blocks at various places on the machine (I am sure they will want as much coverage of the 'rock' as is possible), then write the program to check at a minimum 1 point for each major vector.

-OR-

Place 3 or 4 blocks at various places on the rock and at different heights. Myself, I would place one at each corner of the rock and one in the middle (with the middle one as high as I could get it). Create measurements for each axis on each block (you can probably omit the 'from below' point). Use those for the Gage R study. Of course, the next question is, WHAT TOLERANCE to you use for the study? Well, to begin with, you can not do a Gage R&R since there is NO operator 'input' into the study, so you can only do the Gage R study. But still, what tolerance do you use? Well, you have machine specs that the machine SHOULD be within. If those are what the machine SHOULD be within, then those are your starting point. To ensure that the machine is capable of measureing to those specs, I would think it must pass a Gage R using 10 times those specs (based on the reverse of the 10% rule).

[Paul Sarrach]

My Bad

Wow, Matt, I missed the intent again. My bad