What's new
What's new

CMM inspection of 3d profile tolerance

thetree

Aluminum
Joined
Apr 27, 2010
Location
mentor, ohio
When doing a profile tolerance of a feature on a CMM, what is the cloud of points necessary? For example, I only need 3 points to define a plane but that would not give me a very good data set. As an exercise: If I have a plate that is 4" X 4" with a profile tolerance on the top surface of .0005, how many points would you take? Is this actually in the ASME standard? Thanks, Dan
 
Depends how the feature is manufactured. If it's a flat plate and it's ground then a few points is probably acceptable. If it's a complex shape that's milled then you'll need substantially more points. You need as many points as it takes to know it's a good part.

Not aware of any rule for number of points to define a profile. Once you get into complex shapes you need to start looking for scanning head cmm.
 
Would you consider it to be "good practice" to make an effort to try to hit any bad or questionable spots in a machined finish? I am being asked to inspect some of my customers parts and some have chatter in spots and I do not want to reject their parts unreasonably but at the same time I do not know that one spot that may fall out of profile tolerance would be grounds for rejection. Any thoughts on this? Thanks, Dan
 
Surface specs apply to the entire surface, not to selected portions thereof.

Any areas which are visibly deviant definitely must be checked.

- Leigh
 
Hi Leigh, Is this opinion or based on ASME standard? I agree with you, just curious because some things are not always so visibly deviant.
 
This is one of those things that just doesn't have a cut-and-dried answer, and it's the sort of thing that makes me crazy when I see parts going through an incoming inspection on a programmed CMM setup, where the specific part characteristics are just ignored, and the minimum number of points is always used for the feature. You end up (sometimes) with parts on the shelf being incorporated into product that cause subtle out-of-spec performance in the final product testing, for no known reason at that point.

In this case of profile tolerance on a surface, especially where you are checking for localized deviations to a fairly close tolerance, "you can't take enough points" or words to that effect. I would recommend setting up a grid pattern of X number of points as a test case, run through the parts, see what you get. Do it again with 4X number of points and see what happens, then analyze your results and talk to whoever is in charge of final decisions on test methods to make sure everyone agrees with the method up front. This situation is one, though, where judgment and experience can play a big role. If you are in some sort of ISO-compliant environment, you may have to standardize the test method as mentioned, and let the statistics do the talking. If this is a less-structured environment, and the checks you are doing are "a customer favor", or to determine whether the parts being checked can then be processed through your manufacturing operation, then you may have more flexibility, and also MORE responsibility for the result. As a side note, this specific type of measurement is one reason optical surface roughness/shape metrology is becoming more popular.
 
......

In this case of profile tolerance on a surface, especially where you are checking for localized deviations to a fairly close tolerance, "you can't take enough points" or words to that effect. .......

I would oh so much agree here.
Profile tolerance checking is something a touch probe CMM does not do well.
Even using 3 points to establish the simple first datum/square plane on a milled surface can get you into trouble fast.
It's hard for a CMM to find the three highest points a part would sit on which is why you must relocate the part differently on each test when you try to find your gauge R&R numbers.

This is a area where scanning probes or optical methods work better as you get thousands of "points".

People tend to "trust" CMMs.
I do not get why. Perhaps since they cost so much?
Many ways and many uses where they do not work so well.

1-D gauges are picky but easy, 2-D gauges make life a bit harder but not that much as you can map it out in the software.
3-D gauges like CMMs are so full of errors that they will drive ya nuts if your job is to write the underlying software that makes them work.
Bob
 
People tend to "trust" CMMs.
I do not get why. Perhaps since they cost so much?
Many ways and many uses where they do not work so well.

1-D gauges are picky but easy, 2-D gauges make life a bit harder but not that much as you can map it out in the software.
3-D gauges like CMMs are so full of errors that they will drive ya nuts if your job is to write the underlying software that makes them work.
Bob

I spend a bit of time programming CMM's. We have scanning head and standard type cmm's. For complex 3D features it's about as good as it gets within reason. If I'm working on a new program it's always a good idea to do a quick sanity check with micrometers or even venier calipers. Blindly trusting a CMM is definitely a mistake, but with a proven program I certainly trust them more than my feel with a micrometer (and not just because they're expensive).

To the OP --> if there's chatter on the surface and you're being held to .0005'' you need to check the highs and lows. If it falls out the part isn't acceptable. Not really sure why anyone would call out a profile tolerance on a flat surface (there's better ways to control that feature). It's definitely good practice to inspect areas that appear deviant.

Also keep in mind if the part receives post processing. The blueprint is usually a finished part. So make sure you're comparing apples to apples. A machined part might not meet the criteria until after finishing. But the way you're describing the chatter it sounds like it isn't going to clean up in a finishing operation. Final point is to understand your parts --> ultimately what makes it a good part is whether it functions as expected without failing prematurely. On a really complex part you could over-inspect the part and probably make just about any of them fall out on one feature of another.
 
Hi Leigh, Is this opinion or based on ASME standard? I agree with you, just curious because some things are not always so visibly deviant.
My statement is simply based on the definition of "flatness", which applies to every point on the surface, not to a select few.

Suppose for example that a piece was "flat" except for an inch-tall "mountain" in the middle.

Using statistical methods, one could claim it to be "in spec" simply by not measuring the mountain.

That's obviously nonsense, but it illustrates the concept quite well.

- Leigh
 
Thanks Helter, Spec The part I am being asked to inspect is not actually flat, but a 2" section of 16.00" radius on a smaller part. I do not get the greatest results checking such a short chord length of a large radius, especially when their finish is not that great. I think I will ask them to find someone with an optical system or a scanning CMM.

Leigh, I understand flatness, thank you. I agree with you completely, simply playing devils advocate here. The question was regarding your "visibly deviant" statement. If a part had a warp on one corner and an inspector simply did not see it, the part would be inspected incorrectly. Another inspector may see it, resulting in differing results. Is this just a matter of separating the men from the boys or is there something in the spec that specifies how much of the surface needs to be inspected? It seems like the spec should try to limit judgement in as many cases as possible, limiting the head butting we have all over the industry. Perhaps even specifying that profile tolerances need to be checked by scanning or optics.
 
Thanks Helter, Spec The part I am being asked to inspect is not actually flat, but a 2" section of 16.00" radius on a smaller part. I do not get the greatest results checking such a short chord length of a large radius, especially when their finish is not that great. I think I will ask them to find someone with an optical system or a scanning CMM.

Leigh, I understand flatness, thank you. I agree with you completely, simply playing devils advocate here. The question was regarding your "visibly deviant" statement. If a part had a warp on one corner and an inspector simply did not see it, the part would be inspected incorrectly. Another inspector may see it, resulting in differing results. Is this just a matter of separating the men from the boys or is there something in the spec that specifies how much of the surface needs to be inspected? It seems like the spec should try to limit judgement in as many cases as possible, limiting the head butting we have all over the industry. Perhaps even specifying that profile tolerances need to be checked by scanning or optics.

Is the .0005'' profile tolerance held to a datum or allowed to float? Short chord and large radius is definitely not ideal. A scanning head CMM is probably the only way you're going to measure it (optical system will not be accurate enough). Is it a 2'' section that was cut from a 16'' ring? Parts like that will often spring when cut from the ring.
 
No it is actually a piece milled from aluminum with 2 16" radii opposite each other creating like a wing cross section. My tolerance is actually .004 and held to 3 datums. I suppose I over simplified my question trying not to confuse things. Thanks for your input.
 
Leigh, I understand flatness, thank you. I agree with you completely, simply playing devils advocate here. The question was regarding your "visibly deviant" statement. If a part had a warp on one corner and an inspector simply did not see it, the part would be inspected incorrectly. Another inspector may see it, resulting in differing results. Is this just a matter of separating the men from the boys or is there something in the spec that specifies how much of the surface needs to be inspected? It seems like the spec should try to limit judgement in as many cases as possible, limiting the head butting we have all over the industry. Perhaps even specifying that profile tolerances need to be checked by scanning or optics.
My "visibly deviant" comment was just a first-pass evaluation, looking for anything that might be obvious, like tooling marks or material issues.
This is just normal inspection procedure in my experience.

Obvious things can get overlooked.
I once had to put a large number of holes in a 10-foot long narrow steel bar. Tolerances were tight. It took two days to finish.
When it got to final inspection, the bar was 1" too long. I had forgotten to trim the length. That was supposed to be the final step.

Stuff happens, sometimes obvious, sometimes not.

- Leigh
 
For a CMM inspection on what sounds like a similar problem we checked a number of points, each as a go/no go on the point. It was approximately a 16" long part that was a 18 degree segment of a 20" radius cylinder. From other features the centerline of the cylinder was constructed within the CMM software. Then points were taken on the surface. IIRC the profile of the surface was .005 total, and we were able to easily show the surface was within tolerance.

If the surface was probed and the CMM allowed to best fit the cylinder, the CMM constructed a cylinder that was to large.
 
If the surface was probed and the CMM allowed to best fit the cylinder, the CMM constructed a cylinder that was to large.

Not sure this would be feasible. My chord length is so short that even with .003 tolerance, I have over an inch of tolerance on a best fit cylinder diameter.
 
thetree,

In my experience with 3D cmm's I was able to override the cylinder or spherical fit with a radius measurement. You're left with the best fit profile to your nominal radius. I generally try to get measurements at least 45 degrees apart before trusting the fit to come up with the correct radius.
 








 
Back
Top