Gage R&R: Surface Profile

[weisusu]

Traditionally a non GRR is performed using n=10 parts, three operators, and three trials. For a GRR for surface profile, would you use the same set-up as mentioned previously and the response measured would be the maximum or minimum deviation? To calculate % tolerance, the 6-sigma of measurement error will be compared to the surface profile total tolerance. Thoughts?

 

[jitu2387]

most of my experience is related to inspection of heavy machineries as boilers, heat exchanger etc.
It basically included checking of allignment and dimension ann also related themarking and cutting of the job....

 

[PaulJackson]

The thing about gauge R and Rs is not so much what is being measured but rather what percent of the measured variation is due to inconsistant measurement procedures, operator influence, lack of discernable differences in measured values or lack of distinctly measureable intervals within the total tolerace.

GRR's have another hidden problem when they are predicting these influences on geometric measurment conclusions that reflect unilateral tolerance values. The GRR is typically estimated from the data assuming that it exhibits a "normal" distribution where that is typically not the case when one figures the consumed tolerance for each sample in relation to the total tolerance specified rather than examining the individual coordinate values for their +/- displacements from basic.

Fortunately with profile (unlike position or orientation tolerances) when it is applied to non-features-of-size or individual points it can be treated as a +/- value from basic. That actual value when analyzed raw will tend to exhibit more normal behavior than trying to figure the consumed total profile value from that value.

So no matter whether the profile is applied to a spot face surface with an equal-bilateral +/-, unequal-bilateral ++/-, or unilateral +/0 tolerance... analyze and report the measurement results as "[profile|1.0|A|B|C] (+/- 0.5) is -0.1" or "[profile|1.0(U)1|A|B|C] (+1/-0) is +0.8" or "[profile|1.0(U)0.2 |A|B|C] (+0.2/-0.8) is +0.1". Note: the +/- values represent (+) plus material (-) minus material rather than a larger or smaller basic value. You have to recognize that no matter whether the spot face was controlled with a simple limit dimension or with a profile tolerance there are an infinite number of points on that surface and you will choose to abbreviate your inspection by selecting one or more places on that surface to make your measurement. Your GRR may be highly sensative to the location where those measurements are made if the surface is not flat and oriented well to the DRF. More important is that the level of abbreviation is established to protect the customer and function rather than enable a passing GRR.

Unfortunately you cannot analyze an all-around profile applied to a uniform feature-of-size like this so you will have to select points to monitor for those profiles but there are other ways to look at the parameters of the features to help control the process, see the article in the link below.

Paul