Precision of tolerance - Test validation

[mboteo]

In order to validate a leak tester the customer (automotive) asked me to perform a "precision of tolerance" calculation.

Basically I need to measure the same part 30-50 times and log the results. The precison of tolerance figure is calculated by dividing 6sigma by the tolerance and expressing it as a %. If less than 10%, the tester is OK.

Would like to know a bit more of this statistic. what is the scope exactly. Why this kind of test instead of a regular crossed Gage RR.

S.

 

[Ravi Khare]

It is in fact one of the R's out of R&R. By repeatedly measuring the part, you are evaluating the Repeatability of the measurement system for the same part. The six sigma spread will give you the inherent repeatability error of the measurement system. This is similar to the bias test conducted under Measurement Systems Analysis.

A leak test often destructive (non replicable)and may not be able to use the same part for repeated measurements. How do you address this issue?

 

[mboteo]

In this case it is a NDT (non-destructive test), using Helium gas as a trace gas sprayed on top of the sealed part. On the bottom the will be vacuum of 7psi, and a gas spectrometer will count the cc/s.

 

[Ravi Khare]

Since you are measuring only one part, the Bias test may perhaps give you a better idea of this aspect. It is described in details in the Rev. 3 of the AIAG MSA Manual. It shows you the method to compute the std. deviation of repeatability, on the basis of repeated measures of the same part.