Re: Within Part variation
The goal is to assess the measurement system, not just the gage. The part and the operator are both part of the system. The measurements taken during the gage study should replicate what's being done, or what is intended to be done, in production. Masking within-part variation during the study defeats part of the purpose of doing the study. If the parts are to be measured at a certain position, or as held in a jig or fixture in production, that's the way it should be done for the MSA.
It is true you want to see the results of the
gage system - the interaction of the operator, gage, and part geometry. THat is why we use a part and not a gage pin or gage block. However, variation within-in the part should be analyzed in another venue - process capability, control, etc. Adding every possible error to the gage R&R will never allow you to do what the test intended to do - answer the question if the
gage the correct gage for the measurement. I recommend measuring in the same spot to answer that question. It includes the factors of the
total variance equation that are related to the gage, and not the process.
However, if there is significant within-part variation, then when you measure the part in production you can not measure "one spot" - it will not represent the geometry of the feature being analyzed. You need to measure about the feature and report the high and low value. The only time you can measure one point if it is know that the within- part variation is statistically insignificant (generally in relation to the tolerance, or control limits, if charting is employed.)
Gage R&R is to determine
gage system error. Using a good gage (passed gage R&R) and not determining if the feature has within-part variation as a process measurement is using a
perfectly good gage the wrong way. That is
measurement error. Different thing.