Methods for using CMM data to verify surface profile 1

[score33]

Hello,

can't find a good reference for this, so thought I'd ask...

Once your part is measured by the CMM and you have all your points, what are the accepted/standard ways to verify that the measured points fall in your surface profile tolerance?

 

[ajack1]

It possibly depends on the software you are using but the normal way at least in automotive is to show the error of all points measured against the CAD master and highlight any that fall outside the surface tolerance.

 

[HGMorgan]

Usually, the deviations from nominal are reported as the CMM survey is being performed.
1. Product is aligned per datum call-out.
2.Contour inspection points (point vectors) are taken on the contour. Deviations are reported as +/- distance from nominal, normal to contour.
3.Out of tolerance points are usually noted in Red along with a graph showing direction (+ means excess surface, - means undercut surface.
4.A lot of CMM’s have software that will perform “Best-Fit” on features after survey.

This post reads like the points were taken by CMM, and after that a review is to take place. Not the usual process, however:
The points taken would have to be taken in the “no probe compensation” mode (values reported will be at the center of the probe). Those points can be uploaded to the CAD file of the product and analyzed to the feature. One would need to account for the probe tip radius in the analysis, I.E.; if probe radius is .0600, then the points taken will be .0600 away from the surface, so a value of .0610 means that surface point is .001 excess.

 

[score33]

Thanks ajack1 and HGMorgan.

This is kind of an "after the fact" review. What I'm looking at doig is finding a way to construct a surface out of the measured points and then to compare that what the nominal surface should be.

But it sounds like the easiest way might be to import them into Pro/E to evaluate whether or not the points fall in the tolerance zone.

HGMorgan, do you have any references for the best-fit algorithms used for the measured points?

 

[321GO]

Rhinoceros can also do some of this stuff. It reports the shortest (normal) distance from each point to a selected surface.

Does Pro/E also have that ability?

 

[HGMorgan]

You can make a surface out of a "cloud of points", however, you would have to have software that can do that, and, you could end up with a not very good surface that would not be useful for check of surface to surface (depends on how many and what spacing of points on surface). "Cloud of points" surfaces are best avoided unless there isn't any other way to do it.
Since this is "after the fact", your best option is to import the points into the product CAD file. I don't know Pro/E capabilities, but I would think that it would have “point cloud to surface analysis” and “best fit” functions. Or, as noted, import CAD file & points into software with those functions.
If not, one would have to analyze each point to surface and record deviation. I don’t have any algorithms. Best Fit is basically XYZ Shift and XYZ Rotation. Quick easy example: PTA checks .05 excess & PTB checks .1 excess. Both points are shifted -.05, and then the .05 difference is rotated out. Fairly straight forward trigonometry. But, when there is a complex contour and lots of points….it hurts my head and that’s what the software is for, to keep from hurting my head.




Harold G. Morgan
CATIA, QA, CNC & CMM Programmer