Reporting Size Dimensions with a CMM

In this Question Line video, Jason answers questions about reporting size dimensions when using a CMM. He briefly explains Rule #1 of GD&T before diving into the criteria needed in order to accept or reject a feature of size. He also discusses manual vs CMM inspection and the reporting requirements of the ASME Y14.45 standard (Measurement Data Reporting standard).
Interested in learning more about Print Reading or GD&T? Check out more of our free GD&T resources here: download.gdand...

[@ericschertz9110]

I'd like to throw in my two cents in the hope that someone reads it and offers a counter opinion. I asked this same question in the GD&T Basics Inspection course and I agree that in an ideal world, anytime the UAME or local size reports out of tolerance, then the part should be rejected. But in my (limited) CMM experience, the UAME and local size rarely report within tolerance on a CMM when dealing with a tolerance that tight unless you have an analog scanning probe or take an extremely dense number of hits. And I'm talking like a point every couple of millimeters (who has time for that?). And even then, it can sometimes report slightly out of tolerance. I'm sure it has something to do with the math involved, which is beyond my knowledge. I've come across this more times than I can count when measuring drill holes for example where the UAME reports too small, yet the plug gauge always fits. But we don't have functional gauges for every single feature that we machine. Sure we have bore gauges and micrometers, but those just give a two point dimension. So I have no other choice but to rely on the average size (or least squares in CMM terms). It's also a lot more repeatable. Do I like it? No, I hate it. But realistically, if I were to fail every part where the UAME or the local size reports out of tolerance on the CMM, then I'd be rejecting about 80% of the parts that we make. And that doesn't help me, the company, or the customer. Again, I want to stress that I get much better results with an analog scanning probe vs taking hit points, but even that has its limits and not everyone has one of those. That also comes with other things that need to be taken into consideration, like using the correct filter for instance. But that's another topic entirely. So in my view, when dealing with tolerances as tight as the one in the video, the average size is often a necessary evil when the UAME or the local size reports out of tolerance. If anyone else experiences the same thing or has a different view entirely, please chime in because I'm still relatively new myself.

[@jefffrodermann5348]

Eric, I truly appreciate you sharing your experience on this relevant topic. It sounds like you have a great understanding of the manufacturing process and some of the common challenges that are faced when we deal with evaluation methods (least squared, envelope, etc.).
I appreciate that you also mentioned being open to more knowledge on the topic. I would say this is the crux of the issue. Having a seasoned metrologist and CMM programmer involved in the measurement process is the key. A common misconception is that it takes more time to scan at high density then to take single points. While this may have been the case years ago, it no longer is. I can capture 8,000 points on my Zeiss prismo CMM in the same time it takes to do a hit pattern of eight single points and 360°. Because of this, scanning always gives us the best chance of accurately reflecting true geometry of the part.
The evaluation algorithm, i.e. least squared versus envelope, is determined by the drawing and customer specification. Understanding the end use of the part is key. I agree that it is really the case where both least squared and envelope measure repeatable and with intolerance. However, there are times where envelope is definitely the method to use. This is definitely a topic. I would love to have more conversation about it.
Thanks again for your thoughts.

[@ericschertz9110]

@@jefffrodermann5348 I wish I could always rely on the envelope results from the CMM to pass or fail parts. From a theoretical stand point, it makes sense since it should determine if parts would physically mate together regardless of form error, or conversely if there is not enough contact between parts. But in practice, I would only be passing somewhere between 10-20% of the parts we make if I relied on the UAME and local size results from the CMM because the vast majority of the time, one or both report out of tolerance. I can put a bore gauge or a micrometer on it all day long in a hundred different spots and never get anything close to the UAME or local size, but it will almost always match the least squares (average) results. Even on features for which we have a fixed functional gauge, like a plug gauge for example, the "GO" will always fit, yet the CMM will say otherwise. So I'm thinking that it might be something to do with the math used by the software to evaluate the point cloud or scan data. I think it's called Constrained Least Squares or CL2 Math for short. Or maybe I'm just plain doing something wrong.
I'll admit I've only been programming for a little over a year. But I've had over 100 hours of formal classroom training and I scrutinize my measurement routines to death. If I'm doing something wrong, I don't know what it is. At the end of the day, the shop wants to get paid, the customer wants their parts, and I want to have a job. So my hands are tied. I have to go with what gives me the most repeatable results and what agrees the most with the hard gauges. And unfortunately, that's almost always the average least squares result. Will it come back to bite me at some point? Maybe. But I'm almost certain that if I reject every part that fails the CMM's UAME or local size results, I will be shown straight to the door.
Thanks for listening. It's nice to vent after a stressful day at work.

[@jefffrodermann5348]

@@ericschertz9110 you have no idea how much I appreciate your comments on this. I've thought about posting an article on LinkedIn about something similar. Envelope is not a one-size-fits-all solution, although it is often communicated as such.