Program Validation

[zoolander303]

There's a customer we deal with who has some pretty odd shaped sheet metal parts, I created a program ran it, everything looked pretty normal, tolerances weren't on the high end especially for profile call outs, I proved it out 3 times, I don't see an issue with the program. This said customer is stating when they run the same parts they get better results, than what were giving them, I asked okay are you using the same setup we are, the answer was yes. Beyond that I could think of a plethora of reasons that would have results vary, model of cmm, what does the cmm calibrate at compared to ours, what tips are they using, are you aligning per the datum's so on and so on. Funny thing is when we ask these questions they don't want to give answers or show their reports to us, so even though everything is well in tolerance I figured I'd ask you guys who have been doing this a while what variances can their be in a situation like this for this to happen other than what I stated? First time ever dealing with a issue like this, the only thing is the part has a slight angle for the primary plane so maybe they constructed a perfect theoretical plane for their primary datum, or the software I currently have is 2015 dmis, I know they are on 2020? I could see that maybe helping a little, but looking for your guy's thoughts on this?

Thank you.

[panelbeater1]

No program can be truly validated unless you run a verification program and Gauge R+R study.

Our customer requires a production and verification program each wrote by a different person and the results to be 10% to one another i.e 10±0.05 can have a 0.01mm deviation between the 2 programs

We also do a 5 time run on the program under a dynamic study i.e take part off fixture and back on. Again these results will have to be within 10% to 20% depending on critical features.

There are so many variables that can cause different results, environment, operator error, program robustness, loading error etc

Does your customer run an artefact on their CMM on a weekly basis or do you?

[zoolander303]

I don't know if they do, I just started here and have not what do you run for this weekly?

[dwade]

Program validation can be done using alternative measurement methods (including using another type of CMM software). Been through this with several OEM medical device quality requirements. Dimensional verification should is our first priority on what we call Critical to Quality Characteristics, then a GR should be done to prove that our programs are repeatable. The last thing is to have another set of eyes go over the report and make sure nominals and tolerances are correct.

If you would like a copy of the procedure that I wrote send me a PM with your email address and I'll send it to anyone that wants it.

[A-machine-insp]

Honestly @zoolander303 if they won't show proof then I would tell them to STFU (in much more a polite professional and politically correct way). If they won't show proof then how does it help you as a programmer? It doesn't. It just sends you on a wild goose chase. We have had customers do this to us before and we no longer make parts for them (our choice).

Now, we have some parts that we outsource for various reasons and the suppliers provide us with a FAI report. Being an aerospace/defense shop with our butts on the line, we perform our own full FAI and submit ours to the customer. If we get discrepancies between our report and theirs we will share limited information. By limited information I mean information pertaining only to the dimension(s) in question.

Most times, if there are discrepancies but the part is still good, we stay quiet. If they say it is good and we say it is not, we will show them how we achieved our result and they share how they achieved theirs.

It angers me when shops do what you are saying. It's arrogance and I don't deal with that too well. Being correct and showing how you are correct is one thing. Boasting about how they got a better dimension than you and not explaining is just arrogant ant not helpful in any way. Sharing some specific information is a good way to help the customer/supplier relationship and not doing it can be harmful.

Sorry for the rant but I see stuff like this (not limited to inspection or machining in general) all the time and it ticks me off. Help each other out and everybody will benefit.

[A-machine-insp]

Another thought I just had after rereading your post, profile could very well be down to how it is being dimensioned, legacy vs. Xact. I run Xact but my boss wants a reference dimension in legacy and I see differences between the two every time.

[zoolander303]

ya, I'm using xact, I was going to write a program and use legacy, the weird thing is the a plane im offsetting points for targets A3-A5 to match targets A1,A2 height then constructing my plane least squared. Another thing is the secondary datum is a pattern of 3 holes I'm just constructing a set as my secondary datum and using that as my x,y. Any advice on maybe a 2d best fit to the pattern? or 3d best fit no rotation? No worries on the rant it's frustrating me as well, just can't think of how I could get it to show better results.

[A-machine-insp]

zoolander303 Are you allowed to post the print or a section of it? That would help.

[JEFMAN]

Maybe the answer is at the end of your first post...
"I currently have is 2015 dmis, I know they are on 2020" : datums construction since 2018R2 follow ISO / ASME rules.
If you construct datums usinf least square, and them usng L2 constrained, you can obtain differences around half of flatness (if the datum is a plane !).
If there's a secondary, then, old versions doesn't construct them perp to the first.
You can take a look at this thread, which give some solutions on it... https://www.pcdmisforum.com/forum/pc...oup-to-a-plane
Good luck !