Determining Statistical Confidence and Reliability Levels with Sampling Plans

Hello,

I have been tasked with writing a validation protocol for a plastic injection molding process and it is not clear to me how to determine the statistcal rational for a sampling plan.

For example, a Performance Qualification may be one run of 200 parts and as part of the validation, we want to ensure a 95% confidence level that 95% or 99% of the parts are conforming.

How do I determine the sampling plan for this?

Please let me know if more detail is needed to answer this question.

Thank you very much.
//Ed

Re: Determining Confidence and Reliability levels with sampling plans

BigEd09;

Just a few thoughts:

The AIAG PPAP manual requires a total a minimum of 300 consecutive parts (pg. 3, 4th ed.) and page 7 Note 5 requires "For those characteristics that can be studied using X-bar R charts...a minimum of 25 subgroups containing at least 100 readings from consecutive parts."

If 300 parts are checked without failure over an attribute fixture this translates into an approximation of 1.33 CP/Cpk at 99.73% confidence.

Your customer may accept fewer parts using published statistically valid sampling plans like ANSI/ASQC Z1.9 'Sampling Procedures and Tables for Inpsection by Variables for Percent Nonconforming' or ANSI/ASQC Z1.4 'Sampling Procedures and Tables for Inpection by Attributes'.

Z1.9 was found particularly useful for destructive testing and smaller sample sizes based on whether the standard deviation is known or unknown.

On the other hand, the late Hans Bajaria asked a roomful of 6-sigma Black Belt hopefuls "What should the ideal sample size be?". The responses were many and varied.

Hans then said, "the sample size is to test the potential of personal injury endured when landing on cement after leaping from a 5-story window! Do you think a sample size of 1 would suffice?". His point was that no sampling is a substitute for deep engineering knowledge and process understanding.

Best of luck!

Re: Determining Confidence and Reliability levels with sampling plans

Since you used the term "performance qualification" (PQ) I have to ask if you are working in the medical industry?

You can look at the following thread for an attachment I made for (simple) sample size calculations for V&V (as oppposed to acceptance sampling.
https://elsmar.com/Forums/showthread.php?t=36663&highlight=validation+sample+size&page=2

I am currently in the medical device industry and I have used these formulas and other more complicated ones and experimental structures for V&V.

Re: Determining Confidence and Reliability levels with sampling plans

bev, in your posting from way back in 2009 you said that you had put together a table for calculation of sample size for V&V but i don't know where to look for it. I think that this may be just what i have been searching for. is this something you can repost so i can check it out.