Feature Evaluation

[[Cl...]]

I've been asked to alter a program that was created by someone else. The only real
change is to shorten the primary Datum feature (OD Cylinder) by about half. I can only
assume this is the reason for not passing a type-1 Gage R&R.

The cylinder Ø is .1247" and the new length is .670". One of the things I noticed for this
cylinder that its evaluation is set to LSQ. I'm right on the edge of passing. I'm just looking
for some small change that would get me to <10%.

Would a particular evaluation of a datum feature make any substantial difference in its stability?
I'm running the type-1 again, this time with the evaluation set to MCC. I'd like to see what the
forum thinks about this.

[[Da...]]

Filters and outlier will coax it better than evaluation type.

[[Ow...]]

I could be wrong as I don't deal with Gage R&R much anymore but, from a wingman's perspective, I don't think you'll find anything more repeatable than LSQ.
You are probably aware of all this but, if not here it is.

Speed, the amount of circles measured, points recorded and filters are probably where I would start.

Speed, especially on small diameters, I'd slow it way down and see if repeatability gets better.

The amount of circles, with the shaft length being 3x the diameter, DON"T use anything less than 3 circles. A1/A2 is rarely ever repeatable on Cylinders with just two circles at each end, a 3D-line would be better if only 2 can be measured.

Points recorded, The formula for calculating number of points is based on the UPR Filter you are using multiplied by 7, so 150UPR x 7 = 1,050.

[[Cl...]]

Thanks Owen, Dave. I changed the evaluation to MCE and increased from 2 to 3 circle paths. The result was
much better (still failing). On the subject of filters. Initially the cylinder filter was set to spline and 50 upr. Because
of certain limitations I'm forced to use two 200 degree circle segments to make each circle. Scanning on either side
of the part. So I changed the filter to Gauss and 15 upr (the Ø is < 8mm). Not sure if that's the correct thing to do
as it not a contiguous scan.

[[Ju...]]

I seem to recall reading that L1 evaluation helps when scanning partial arcs, as it is more resistant to outliers.