points to capture on a 3D measurement

[OVieira]

How do you define the number of points needed for a 3D measurement using a CMM?

As a practical situation:
There is a Profile of a Surface call out on the print. In order to have a good valid dimensional report, how many points should I capture?
Is there a standardized criteria to be used based on the surface size and tolerance?

 

[IRstuff]

Shouldn't this depend on the shape of the surface? A flat surface is easier to characterize than round surface, which is easier to characterize than a randomly irregular surface.

How is the profile defined? An equation? A polynomial would require as many points as there are roots for a minimal correlation

 

[3DDave]

The closer the actual surface variation is to the allowed surface variation tolerance limits the more points will be required to accept the feature.

Some relaxation can be had if the form of the actual surface is very uniform compared to the variation tolerance. For example, if the location of a surface has a large location tolerance and the surface is precisely made then only a few points might be required.

In short, the plans for inspection of actual surfaces requires knowing a significant amount about the history of the manufacture of the surfaces.

 

[OVieira]

I had seen a lot of tribal knowledge and people defining the amount of points based on experience.
I understand more points taken means a better reporting of the measurement.

I was looking for criteria that could be used as a tabulation corresponding to the feature size or extension with the tolerance.
Per mathematical definition, 3 points define a plane, the standard BS recommends to take 9 points distributed on three lines (but to correlation on how big this plane is).

How do you define the inspection plan how many points to be taken with CMM?

 

[IRstuff]

How do you define the inspection plan how many points to be taken with CMM?

That would likely depend on the exact requirement and the consequence of failure. To wit, an optical surface would have the equivalent of hundreds of thousands points of measurements.