Which Best Fit calculation to use on auto circle for an ID

[JohnMark]

Good afternoon,

I would like some help on which Best Fit Type for an ID would be best for replicating a plug gage or a gage pin? It is basically an ID where a straight shaft will be mated with it. We are having a .0003" to .0005" bigger diameter reading on my CMM than what the shop floor gages are being used. The gages have been calibrated and show the correct diameter. I am currently using a Least_Sqr and my form and circularity is consistently measuring .0005" to .0007".

Wanting to know if i can replicate an auto circle or cylinder to that of a Plug Gage.

Thank you,
JM

[JacobCheverie]

Try a maximum inscribed circle instead of least squares - it'll pick up on the highest points of your circle and the form error seems to correlate with the difference in readings. You are essentially averaging out your form error with LSQ.

[JohnMark]

Try a maximum inscribed circle instead of least squares - it'll pick up on the highest points of your circle and the form error seems to correlate with the difference in readings. You are essentially averaging out your form error with LSQ.

Thank you. I knew the background of Least-Sqr, but have never really dove into the other calculations. This helps. I will try it! Thanks again

[JacobCheverie]

In cases where the form error is low, the maximum inscribed/minimum circumscribed fitting algorithms work very well at simulating a functional measurement. If your form error is high you may run into issues because the algorithms are more sensitive to outliers than a traditional least squares algorithm. I hope that works out for you.

[JEFMAN]

In this case, think I would construct a feature set from LS circle, filter it (or construct a filtered circle from it), then construct another feature set from filtered hits and construct the max inscribed circle.
It would avoid outliers...

[JohnMark]

In this case, think I would construct a feature set from LS circle, filter it (or construct a filtered circle from it), then construct another feature set from filtered hits and construct the max inscribed circle.
It would avoid outliers...

OK, them seems great if i knew the background on how to do all that. LOL. Do you have a link explaining how to perform those actions?

[Schlag]

Number of hits is huge when it comes to this. I have found that if your using say 5-7 in a .250 hole,LS will match the pin size the closest. That doesnt always comeout that way though. With a small amount of hits, whats the likelyhood that you find the high and low spot ? Not very good...

[neil.challinor]

Ideally, measure a cylinder using lots of points and at least five levels (use the adaptive circle scan strategy if you have an analogue probe). When you report the hole, include a cylindricity dimension with an analysis view - this will show you the shape of the hole and where the high and low spots are and should make it easier to explain any discrepancy between the plug gauge & CMM.

[Schlag]

I understand what your saying but based on the type of parts your making and the tolerances allowed decisions must be made.. This is a " theory vs. practicality" question. " My CMM program is now 45 minutes instead of 15 " or something similar....There is always a happy medium somewhere.