Sampling plan for in-process QC (medical devices)

Mike S.

Happy to be Alive
Trusted Information Resource
Using lot size was a result of negotiations, not statistics. hence why the AQL tables are not statistically justifiable - they only accepted because, well we've always done it and mythology is more powerful than science for too many people.

I will also get on my bigger soap box and remind everyone that AQL stands for ACCEPTABLE quality level, it is the defect rate that you will ship most of the time when it is present. What most people actually want is the RQL or Rejectable Quality Level.

Bev, I'm going to take a big risk here because I think you are more generally knowledgeable than I am in this area, but I'm gonna respectfully disagree that AQL tables "are not statistically justifiable". If what you say is true, no AQL-based sampling plan meets AS9100 requirements.

AQL plans have OC curves associated with them. One can take an AQL plan and calculate AOQL, LQ, ERP, and other quality parameters associated with that sampling plan.

What am I missing?
 

Bev D

Heretical Statistician
Leader
Super Moderator
let me clarify that I mean that the AQL milstd105/ansi TABLES are not statistically justifiable as they use the lot size. If you go to first principles you will see that there is no population size (lot size) used in determining the "effectiveness" or the OC curve of a given sample size. I've had more than one classically trained statistician say to me "you've got to be kidding" when they see these tables.

Beyond that there are many other problems with acceptance sampling in general. Not the least of which is that sampling can actually provide the protection it 'promises'. a great read on this is Donald Wheeler's "The Problem with Acceptance Sampling, parts 1 and 2.
 

Mike S.

Happy to be Alive
Trusted Information Resource
Bev,

Thanks, I do recall having read those articles, but long ago, and I had forgotten that they existed.

Let me ask, forgetting internal production for a moment, what do you do at receiving inspection when you receive a lot of material and want to know whether or not to accept it?
 

Bev D

Heretical Statistician
Leader
Super Moderator
We have very little sampling inspection. Our policy is to have the supplier send us their SPC charts and their final inspection data. We do require an acceptance sampling plan based on the straightforward poisson distribution with c=0 for the RQL. We also require some pretty strict process control approaches to keep the defect rate very close to zero. So sampling inspection - especially attributes based - is pretty useless.

There are times when we know we are receiving lots with defects in large batches and we need to screen. In this case we will typically impose a short term inspection with both AQL and RQL. We also have to have a pretty tight control and mitigation plan to protect the Customer experience. (we sell to the end Customer). If the batches are small enough we do 100% inspection.
 
Top Bottom