Here's a general question:

We use a C=Zero sampling system for our incoming inspection, using "associated AQLs". What is the best way to determine which AQL to select for a particular part being inspected under this type of system? I have a Quality Engineer that wants to develop a set of "standard AQL levels" for each type of product being inspected.

For example: Rubber seals would always have an AQL of 1.0, nuts and bolts, 10.0, and so on.

Does anybody within the sound of my voice have any input into how this should be handled? We're trying hard not to be arbitrary.

Thanks.

Norman V. :bigwave:

:confused: :confused:

Sorry, I've got a major disconnect going right now but I'm responding anyway (they say the first thing that goes.......).

I don't see how an AQL has anything to do with a C=0 "sampling plan". By definition, doesn't AQL imply defective/nonconforming parts are OK?

Bill

In the traditional sense of AQL, (MIL-105-E), you are correct, AQL determines (among other things) how many rejects you can accept in a lot. In a C=Zero system, the "associated AQL" merely is used in a table to help you determine how many samples to pull based on lot size.

The AQL levels are "associated", because the Operating Characterisitic (OC) Curves would look about the same in terms of a percent defective for each AQL level, even though the acceptance number is Zero.

My main questions was, what is the best way to decide how to set the AQL level during the Quality Planning process, aside from just picking an AQL arbitrarily?


Thanks,

Norman V.

Since it is incoming inspection, I'd sit down with your internal customer(s) who will use each part and discuss it with them. Discuss what impact a defect will have on them - i.e. if there is 1% defective how would that affect you and downstream operations, etc. Also, what has been the incoming quality level in the past, and was it "acceptable" to them in the past. Also consider time of inspection, etc. Some will rant and rave and say no defect is ever "acceptable", but in the real world, IMO, zero defects is a pipe dream. JHO. Hope this helps.

Hi

Dr Ken Stephens (Applied Acceptance Sampling, available through ASQ) gives the following steps:

1. Determine and list each quality characteristic
2. Determine the product unit.
3. Develop and specify the test method.
4. Determine the criteria for conformity.
5. Determine and list the calssification of nonconforming units or nonconformities and/or groups of nonconformities.
6. Establish an Index Quality Level (AOQL, etc).
7. Determine if sampling/acceptability should be based on lot inspection, on-line (continuous, etc.


I have excerpted here, and he goes on to more specifics. We used this textg for one of my classes at SPSU, and I can say that if you want to find out about acceptance sampling, this is the text.

The class was mind boggling.

Hope this helps.

Craig,

When you say "mind-boggling", do you mean that in a good way or a bad way?


Anyhow, I saw that book in the ASQ catalog, and it does look good. The CD ROM that accompanies it seem interesting as well.

Maybe I'll put it on my Christmas list. :)


Thanks,

Norman V.

Norman:

Mind boggling in both a good and bad way, really.

Before this class, I assumed that Dr. Deming ment that inspection was bad. Not so. What he said, and ment, was that DEPENDANCE on inspection for quality is bad. We still inspect, but our process is such that we don't make bad stuff.

In a bad way, we actually designed acceptance sampling plans - those are some of the major files on the CD. Not for those with less-than-complete statistical backgrounds, IMHO. The class was the most difficult in the SPSU Master's program. We used the draft for Dr. Stephen's book, and he taught the class (online). He is retired now, but his list of publications is impressive, and this particular book is highly recommended.

But the question still has not been answered, unless it is in The BOOK? I believe the act of deciding is a step (#6), but are there any generally accepted parameters in making that decision, OR do we have to read the book?

Douglas:

In short, if you want an optimum answer, read the (a) book, and use it to decide what is best for your situation.

Acceptance sampling is like a lot of other tools - it can be powerful, but only if it is used in a manner that fits the situation. I could give you a "canned" answer, but without knowing the situation, it would likely be worse than the "arbitrary" alternative.

Sorry I can't be of more help, but to give an answer here might do more harm than good.

Craig,

Good information.

Kevin

In the traditional sense of AQL, (MIL-105-E), you are correct, AQL determines (among other things) how many rejects you can accept in a lot. In a C=Zero system, the "associated AQL" merely is used in a table to help you determine how many samples to pull based on lot size.

The AQL levels are "associated", because the Operating Characterisitic (OC) Curves would look about the same in terms of a percent defective for each AQL level, even though the acceptance number is Zero.



Please, clear this to me!
I think my case is similar. We do a sample based electrical testing of electronic components and the requirement is zero defects!
We are using Mil 105-E only to set up the sample sizes.
What is this C=Zero system? Please explain!

Thanks!

Brutas,

Using MIL-105_E, you have both an acceptance number and a rejection number for each sampling you pull, based on the lot size and AQL value. For instance if the acceptance number is 4 and the rejection number is 5, when you did your sampling you would accept the lot if it had 4 or fewer nonconformances, and reject it if it had 5 or more.

The "acceptance number" is also know as the "C" value. A "C=0" sampling plan means that you accept the lot ONLY if there are zero nonconformances found in the sampling. ONE nonconformance indicates cause for rejection of the entire lot.

It's all a bit more complicated than that, but that's a pretty good summary.

Hope this helps,

Norman V.

dear friends
now a days all are talking about six sigma,zero defect,lean manufacturing.Still i found those companies are using sampling plans for IQC & OQC.What is the reality? Really zero defect is there in any manufacturing company?

Sanjay,

Great question!

Even with the continual pursuit of Six Sigma and Zero Defects, incoming (and outgoing) inspection is still sometimes required, usually when some doubt is cast on the capability to produce products to specs.

When we do perform it, the C=0 sampling plan is used to assure that the entire lot will be rejected if any nonconformances are found.

Hope that helps! It's an imperfect answer, but we live in an imperfect world. (job security for us QA people, I suppose.)

Norman V.

dear friends
now a days all are talking about six sigma,zero defect,lean manufacturing.Still i found those companies are using sampling plans for IQC & OQC.What is the reality? Really zero defect is there in any manufacturing company?

Not that I know of. I can't fathom a company of any size being truly zero defect for any length of time. Even if you manage to ship zero defects for a time, which is incredibly hard, there will always be defects internally that are "caught". It has always been so, and IMO it will always be so because fallible humans are involved. But despite this, we can and will continue to do great things.

In the traditional sense of AQL, (MIL-105-E), you are correct, AQL determines (among other things) how many rejects you can accept in a lot. In a C=Zero system, the "associated AQL" merely is used in a table to help you determine how many samples to pull based on lot size.

The AQL levels are "associated", because the Operating Characterisitic (OC) Curves would look about the same in terms of a percent defective for each AQL level, even though the acceptance number is Zero.

My main questions was, what is the best way to decide how to set the AQL level during the Quality Planning process, aside from just picking an AQL arbitrarily?


Thanks,

Norman V.

What was the quoted process stability of that process? They should also have an OQC with specified sampling plan. If it is an end product, the field return % should give the clue. Ultimately, your customer is doing 100% inspection.

What was the quoted process stability of that process? They should also have an OQC with specified sampling plan. If it is an end product, the field return % should give the clue. Ultimately, your customer is doing 100% inspection.

Michael,

Good question!

The C=0 plans have "Associated AQL" values that correspond to the old MIL-105-E standard. On pages 10-12 of Squeglia's book, ("Zero Acceptance Number Sampling Plans", 4th ed, ISBN 0-87389-305-0), he shows that in most cases, the A.O.Q.L value is better using the C=0 plan than it is using the corresponding MIL-105-E plan.

Sample sizes are higher with the C=0 plan, and when planning which index value ("associated AQL") to use, we consider process capability and the characteristic class of the of the part in question. Critical characteristics are usually inspected to tighter index values (larger sample size)

Norman V.