Correlation Study

[[Cl...]]

Hey guys! So recently I was asked to make some adjustments to a program to shorten the run time. To do this I only rearranged the order that the points are being taken, and reduced the clearance distances around the part. From my understanding, doing this should not impact the measured results at all so, we were asked to conduct a 30 piece correlation study to verify. The operator who conducted the study had misloaded a handful of parts making it appear as though there was a large deviation in results from the normal program to the reduced time program. We verified the issue was misload by having a different operator load the parts in question and see if the results were the same or if they conformed to the historical results of the parts. My suggestion was to conduct the study again ensuring the parts are loaded correctly but, the QE said this could not be done. Has anyone ever dealt with a situation similar to this one? does anyone have any tips on how to move forward with this or knows where I could find instructions or information on what to do with outliers from a correlation study?

[[SH...]]

Please sign in to view this quote.

As far as in my understand nothing will happened to your program. I daily do R&D in my program, my customer told me don't change the program but I do changes to improve my knowledge.if you did not tried anything new then you can't improve your knowledge.

I don't think changing the order of the point will affect outlier ,cut the point or add it might affect.
You can find more about outlier in manual or outlier ISO standards mentioned in the software.

[[Ke...]]

If the error was introduced due to improper loading, then there is an issue that needs to be resolved. The QE is correct that you can't (shouldn't) just re-do the study with another operator. There needs to be a documented adjustment to your process. You can do a couple of things, such as creating a work instruction for operators to follow, so that they are all loading the parts onto the fixture the same way; or you can "error proof" the fixture, so that it is physically impossible to mis-load the part; also (and I hate to sound like a broken record on the forum, but) are you looping the Base Alignment? If not, try mis-loading a part, so that your results are "no good", add a Loop to your base alignment and see if that corrects the measurement results (looping the alignment won't help if the probes were shanking out, when taking measurements)

[[Me...]]

You can do a GRR study and submit the report to your QE & QM. We usually use 3 operators with 10 pieces.

[[SH...]]

Please sign in to view this quote.

He is not asking about machine tool fixure validation but he asks about CMM program output change related to his amendment in points..

[[De...]]

Please sign in to view this quote.

This is correct, however that in itself is the problem. The reason for the QE stating the results cannot be just re-run with a different operator is the QE is not just checking the program. The QE is taking into account the entire inspection, which by nature will include operator error.

The program likely has not changed, however a problem WAS found in the results which then needs to be addressed.

If the original program had been checked with the same parts misloaded, the problem would have been found in the original program.

As was said - this can be addressed by better fixturing or through some other method, but it does need to be addressed.

[[Cl...]]

Thank you all for your feed back. My suggestion to the QE was to redo the Correlation study without fixture loading then, do a Gage R&R on the new time program with three operators, three trials and five parts. I'm pretty new to Metrology but from my understanding, unlike a Gage R&R with multiple variables (part variation, load variation, operator error, ext.). In a Correlation study we are looking for one variable, "is changing the order of the points being taken going to change the end result". This means that the part should not be loaded and unload when switching from program to program. That way if we do see a large deviation, we know for sure that it's do to the program change and not any other variables. But, the QE wants to treat it like a Gage R&R with multiple variables. The alignment is not looped (customer supplied) so, even if you load and unload the same part, and run it twice on the same program we will still see a deviation. I wanted to know if anyone knows where I could find solid information or documentation that I could present to the QE on how to correctly perform a Correlation study with one variable, not a Gage R&R with multiple variables.