Re: MSA 3rd Edition Software for R&R Studies

Posted by Jason Vermeiren on March 25, 2003 at 12:49:16:

In Reply to: Re: MSA 3rd Edition Software for R&R Studies posted by Glenn Black on March 25, 2003 at 12:33:38:

: I would suggest that there are a number of programs, such as SQC Pack that
: are now (or soon will be) compliant. Purchase on the understanding that you
: will get the upgraded software for free if it isn't already compliant.

: See American Society for Quality magazine, Quality Progress, for an
: excellent listing of the major software vendors & their various features.
: For Gauge R&R, I have written my own spreadsheet that does all the
: calculations.

: In addition to the software, the following typical mistakes are made when doing R&R studies:

: 1. Fail to have standardized test method that everybody is trained on & following.

: 2. Randomizing all samples, test subjects, and trials so you don't have confounding of the data. Over 90% of R&R studies are "unacceptable". You will need non-confounded data to figure out why it failed at the next step, so you can fix it.

: 3. Intervention during the R&R study. DO NOT fix, stop an error, change the method, or anything else during an R&R study. You are invisible and a silent observer. Fix & change BETWEEN this R&R and the next trial. If R&R is hopelessly flawed, stop, fix, & re-start all over again.

: 4. Lab vs real world conditions. Do the R&R under reeal world noise, vibration, lighting, & timing. Your testers don't have all day to decide the measurement value. That's not real world.

: 5. Get 10% or less R&R for both your tolerance (decides scrap or OK part), as well as for 10% R&R with reference to your normal process variation (ability to monitor & control your manufacturing process). YOU MUST HAVE BOTH. You are fooling yourself if you only get one of these 10% R&R hurdles passed. DO NOT SKIP THIS CRITICAL PART. This is where an expert can really help you to achieve these mandatory goals.

: 6. Poor statistical data analysis. Fixing R&R's so you get 10% of spec and 10% of process variation is tough work. Use the best statistical tools available & someone who REALLY knows what they are doing with them to maximize the learning from each R&R. This will ensure your problem solving on your poor measurement systems is directed in the right direction, with no error nor waste.

: Hope this helps

: Glenn

Follow Ups: