CMM program qualification criteria

[KDM0811]

Does anyone have any tips on qualifying a CMM program for a particular part. I am NOT talking about the CMM hardware, or the CMM's computational software. I am more interested in something like a Gage R&R analysis. We currently have good repeatable results in measuring gage blocks, but poor results in measurement repeatability with a given part. I am looking for some criterial whereby the QA Department must qualify their CMM programs before using them on production parts.

 

[dporte17]

KDM0811:

What kind of parts are you measuring? We perform repeatability studies on our parts whenever we get new fixtures. We have found that problems with repeatability when running the programs tend to be issues with the fixtures and not the programs. We work with sheet metal, so there is always a balance between not enough constraint to have a repeatable measurement process and too much constraint so that variation is not seen.

That answer your question or are you asking more about the specifics of the process?

Dave

 

[PSE]

AIAG has a publication on Measurement Systems Analysis that is a good reference (MSA-3 price is $27 US).

http://www.aiag.org/

Their site is a little challenging when it comes to finding the right publication (download the PDF listing). The publication should help you to determine which type of Gage R&R study would be appropriate. Keep in mind that a Gage R&R study result on a part does not qualify the equipment as being capable of accurately measuring all varieties of parts. It applies to that setup only.

Regards

 

[KDM0811]

We are measuring machined parts and machined castings. I think that our problem stems from the fact that our CMM program may take as few as 3 "hits" off of a datum surface that may have a 125 inch surface finish. We typically use a .062" diameter probe tip. We then "attempt" to measure the true position of a dowel pin hole to the datum within .0003". The datum's surface finish has a major effect on everything measured against the datum. How many "hits" should be used to establish (precisely) where the plane of the datum surface is?