Statistical approach to clean inspection intervals

[Cesofresco88]

Company produces small stainless steel, "washer" shaped parts that undergo final inspection in a clean room environment. Currently, only final-visual inspection takes place under 20x magnification (no dimensional verification). Procedures are documented and employees visually inspect against 15 criteria that are visible to the trained eye. The company does not have any customer issues (for the past 10+ years) with parts not functioning properly so the inspection process is fairly sound.

A new consultant wants to implement a final dimensional verification in the clean room. We are not against this, but would like opinions as to if it is necessary, and if it is, how to amend the procedure so that the inspectors check only a few (not 100% inspection). What would be the best statistical approach to mathematically validating that a "1-every-50-parts" is sufficient?

Thanks all!

 

[IRstuff]

Lots of dependencies on lot sizes, process variation, degree of confidence, etc. Additionally, you need to develop some infrastructure on what to do if you fail the inspection:
> retest, and how much
> process correction
etc.

The confidence limits ultimately drive the number of samples required.

There are a number of books and references that are available.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[Cesofresco88]

Thank you for the prompt response IRstuff.

I'm assuming you can't recommend literature due to the forum guidelines?

 

[IRstuff]

No, it's just that most of my references are old and military-related:
MIL-HDBK-53
MIL-STD-1235
ASTM E 141



TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[GregLocock]

Introducing SPC for a new process is mainly a case of applied intelligence rather than a cookbook.

First you need to measure the variation in one batch from one machine, then compare machines, then plot trends vs time.

Google led me to Link there are plenty of others.

I think military maths is the same as civilian.

Has your new consultant identified the gains from improving a process that currently has 100% customer acceptance?


Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?