Please help with the construction of OC curves

R

rilakuma1121

hello all,
I have to construct a composite oc curve using the switching rule from iso 2859.
I know the ASNI Z1.4 contains the scheme oc curve, however it does not include the consideration of discontinue inspection.

I have to assume after 5 lot rejected while on tightened inspection, discontinue inspection is imposed.

Thank you
 

David-D

Involved In Discussions
I'd first comment that I'm a bit concerned with the underlying premise. If I have a supplier so disinterested in root cause and corrective action that they blindly plug onwards (at a constant, bad quality level) with lot failure after lot failure to the extent that they get to the point of rejecting five lots in a row then you want to be rid of them as quickly as possible.

I'd question the usefulness of such an OC curve as having that constant level of devects over time isn't really realistic (you'd hope for some level of corrective action and improvement). Hopefully the only time you'd run into a situation like that would be if you had A supplier with lots of WIP and a process with no monitoring or control; in that case you're gambling anway. Ideally, you'd never select them as a supplier or dump them as quickly as possible. I guess, for comparative purposes it might be an interesting exercise.

That being said, I think this is an easy opportunity for a Monte Carlo simulation (I'd personally use Excel). Build whatever underlying OC curve evaluator you want for each of the sampling plans (Reduced, Normal, and Tightened). Personally,I'd probably just cheat/simplify and assume binomial as there are enough other more important sources of error. Then I'd simulate a bunch of runs and let the accept/reject drive me where it may. Alternately, maybe run until termination length.Tally up the results and assign the value for each point on the OC curve.

Hopefully not too hard or particularly time consuming although your spreadsheet is likely to be pretty large so you'll want to turn off the auto-cLculate function of excel.

My thoughts at 6 am on a phone, good luck,

David
 
R

rilakuma1121

Thank you very much David. Actually, you are right. It is a interesting exercise than a real case study. We assume the producer is producing lots that can meet up to AQL. We are kind of investigating the advantages of the switching rule and would like to try to use this simulation to show that misuse of switching rule may lead to a significant increase in the producer risk.

However, is it also possible to include the discontinue inspection effect in to the simulation? According to the ISO 2891, 5 times reject remains on the tightened inspection will lead to discontinue inspection. Based on this, we assume the after discontinue inspection is imposed, the next ten lots will be rejected and subjected to 100% inspection. After that, the inspection will resume to tightened.

Once again, I really appreciate your help!!:thanx::thanx:
 

David-D

Involved In Discussions
The real question to me stems from what you do once you discontinue inspection. Just going to a period of 100% inspection solves nothing as the process isn't being improved so as soon as you cease the screening, you'll just return to the same problem you started. Furthermore, on all the previous rejected lots the supplier was probably performing 100% inspection after the lots failed to then try to resell them; now shifting to 100% inspection is just cutting out the effort of failing the AQL sample.

For any usefullness, some sort of corrective action should be implemented coming out of the shutdown (up to and including getting a new supplier). Perhapse in your example you could have an improvement in process capability upon resumption. If you do nothing and just resume production, you just reenter the same failure loop you were in in the exact same OC curve. I don't know how or why you would further model that. Unfortunately I'm not sure that helps you with your question.

David
 
R

rilakuma1121

Can we see like this , I have read some journal regarding the studies of the performance of the Switching rule , they normally will assume psychological pressure act on the supplier will force him to increase the quality .
 

Bev D

Heretical Statistician
Leader
Super Moderator
Are you looking for the OC curve or the average outgoing quality curve?
The OC curve just provides the probability of detection given a defect rate and sample size. There is no OC curve for 100% inspection.


Yes the switching rules assume the producer will be motivated to fix the problem (otherwise its just a vicious cycle of roullette - going from a sample to 100% based on the 'whim' of a non-homogenous process). But why rely on that? if you knwo the process is incapable why don't you just fix it? is your management clueless about the cost or do they not believe that Problems can be fixed?
 
Top Bottom