Profile callout to three holes that are not a bolt pattern

[LibertyMachine]

I just got a rejection from a customer on a job I've run many times. 13 parts out of 105 failed on a .010" profile callout. I can't share the full print here, but I'll try to describe as best as I can:

3x .197+/-.005 holes. This is Datum B, and Datum A is the top surface. Profile is .010 A and B. The holes are not a bolt pattern, but rather, two that are in line, .950 apart from a common centerline of the overall part (almost a 90 degree arc shaped part) and the third hole is roughly 74 degrees off the vertical. Constructing a circle through these 3 points does not yield a "bolt pattern" that shares the common centerline of the rest of these parts. No matter how I check it, I cannot reproduce their results.
The question I have is; How do I set up PC-DMIS (2017 R1) to properly inspect these parts?

I've attached a screenshot that is the rough idea of the overall part
2019-11-07_06h42_37.png

[Matthew D. Hoedeman]

Start with a basic alignment, level to the plane, rotate to the holes that are in line, then origin on plane, circle.

THEN, do a best-fit alignment, 2D, trans & rotate using the 3 holes.

[Liambo734]

I would do this!!

[Roberto]

I was thinking of creating auto circles with sample hits (3x) for use in an iterative alignment which sets the orientation to their best fit condition.
I would think the customer is looking for the position of the the 3 holes with respect to each other.

[LibertyMachine]

Those three holes are all basic dimensions ( X and Y) with a true position of .014 MMC

[vpt.se]

You should create a set of the three holes and make the set datum B. Then evaluate the profile in Xact (not legacy) to A and B.

[Matthew D. Hoedeman]

I was thinking of creating auto circles with sample hits (3x) for use in an iterative alignment which sets the orientation to their best fit condition.
I would think the customer is looking for the position of the the 3 holes with respect to each other.

The iterative would lock 1 hole to perfect XY and one hole to perfect X (or Y), and put all the position error into the third, I still think 2D trans/rotate bestfit alignment, it will spread the position error over all 3 'equally'

[LibertyMachine]

Okay, I've tried all the above. Results are still showing the part in tolerance. I was talking to someone more familiar with PC-DMIS than I, and they mentioned that there may be a default setting in how Profile is calculated that might be skewing my results. Does that ring a bell with anyone?

[bswagger]

Maybe he's talking about profile using form and location instead of just form?...which would be using legacy reporting instead of xact.

[Peter Fuller]

While I don't know if either of these settings are an issue but it's an avenue to investigate. Below is the information on profile calculation settings in settings editor.

Profile Calculations

• In v4.2 and higher, in the Option section of the PC-DMIS Settings Editor, you can set the UseISOCalculations registry entry to 1. This reports Profile as two times the maximum deviation. This setting only affects Profile (not Flatness). Also, when the GD&T Standard on the Feature Control Frame tab of the GD&T dialog box is set to ISO 1101, the Profile reports two times the max deviation, and Profile form only ignores size.
• In PC-DMIS 2009 and higher, FCF Profile with no datum referenced takes into account size. The UseSizeForProfileDimensions registry entry in the Option section of the PC-DMIS Settings Editor lets you revert to the V42 and earlier behavior for legacy profile dimensions only. This registry entry defaults to 1 (TRUE), so you must use the PC-DMIS Settings Editor to change it to 0 (FALSE). If set to 0, legacy Profile form only ignores size. However, the FCF Profile always takes size into consideration.