Can you Set Inspection Frequencies Scientifically?

[SteveL]

Firstly, apologies if this is the wrong forum but I couldn't find one related specifically to Inspection.

To my main point. As a Quality Engineering Dept have been tasked with devising a methodology for determining Inspection Frequencies of parts we measure on our CMM's. This is proving very difficult to acheive scientifically.

We clearly want to relate inspection frequencies to process performance and therefore we've initally used both Cpk outputs and Est Reject % as our 'drivers' ie Cpk < 1.00 measure 1 in 1

The problem we have is whether or not this rationale is valid and if so is there a recognised scale with which to adopt this approach to
ie Cpk <1.00 = measure 1 in 1, Cpk 1.00 - 1.33 = measure 1 in 2, Cpk 1.33 - 1.67 = measure 1 in 5, etc.

Thanking you in advance for your input.

Steve

 

[Jen Kirley]

Hello Steve, welcome to The Cove!

The method I know of that's used most often is the sampling plan. Are you familiar with them? If not, here's an introduction in the thread Are we using the right sampling plan?

 

[Benjamin28]

I think Jenn has led you down the correct path here. What you're looking for is an acceptable sampling plan. You can read about sampling plans here on the forums and get some good information from a google search. But yes, you can set inspection frequencies (sampling intervals) by using an established sampling plan which will allow you to base your inspections on a statistically valid system.

 

[Jim Wynne]

Good advice from Jennifer and Benjamin so far. Keep in mind that what you're trying to do is find a level of acceptable risk--the risk of not doing inspection as frequently as you do presently. I would forget about trying to correlate Cpk with this sort of risk. If you're currently doing control charting, you should already know what you need to know about the likelihood of NC production. What remains is a rational sampling plan that will allow you to have a reasonable level of confidence in finding defective items, and there's lots of information here on statistical sampling.