Repeatability

[[Cl...]]

I'm performing a type-1 Gage R&R (1-part, 10 replicants without removing/replacing the part in fixture).

The first 7 runs repeat very well. Always the last three runs jump off the rails. I'm confident that the part

is clean and the part/fixture are not moving. It's a simple program checking perpendicularity of a cylinder

to a plane and runout of one cylinder to another. What could I be missing? The base alignment is a simple

plane (Z - Spatial) and cylinder (XY origin). Any ideas?

[[Ri...]]

I would place two indicators on the part to detect movement during the runs, an indicator for each axis the part could move (In your case XY).

Then watch as the program runs for indicator movement.

At the least, this could rule out part movement.

[[Da...]]

Cylinders are tricky, just as a test, run all your parts again but run as "current alignment " this takes a possible alignment, or cylinder analysis out of the results. See attached for Cylinder possibilities. 

Datum Definition When the Datum Is a Cylinder.pdf

[[Cl...]]

Richard. Turns out that sucker was moving ever so slightly. I shored up the fixture and ran it again. This time it repeated very well.

[[Ri...]]

I do this on parts that could have the ability to move due to minimal fixturing or flex issues.

After the Base Alignment and two or more points, in your case, I would add a point that touches the part in X and then one in Y. (Sometimes I use all three axis).

Add these as characteristic in your report. Maybe label them as: Start Detection X and Start Detection Y.

Copy the same points and paste at the end of your features.

Copy the same characteristics and paste the end of your characteristics. Rename to End Detection X and End Detection Y.

Now you will know if the part moved every time you run the program.

This has saved me many headaches.