Hello,

I am currently using the AQL system for my work but I think there is a problem with this system.

This system gives acceptance or rejection thresholds for major and minor defects but sometimes it can lead to an incherence.

Let me explain this problem with an exemple:

If we have for instance:
-Minor defects: acceptance 7, rejection 8
-Major defects: acceptance 2, rejection 3

If we find 8 minor defects and 0 major defects we reject the production.
If we find 7 minor defects and 1 or even 2 major defects we accept de production.

I think this is a real problem because the second case is more serious than the first one and we accept it while we reject the first one.

The proof of this problem is that we can easily put in order the different cases bellow, from de least serious to the most serious:

8 minor defects and 0 major defects < 7 minor defects and 1 major defect < 7 minor defects and 2 major defects.


Can someone help me to understand why AQL works like that?

Thanks a lot.

Damien

It works that way because your organization has agreed to accept a certain level of defects (the Acceptable Quality Level). If you really can't accept that level of defects, you need to change and use a plan with a lower AQL. You might also choose to use a "C=0" sampling plan which will never accept a lot with known defectives.

First, thanks for you answer.

What I say is wathever the levels of defects chosen by the organization, this system can allow the organization to accept a production worse than a production it would reject (see my example above).

And "mathematically", this is due to the separation of major and minor defects.

I find your reasoning to be without logic.

If your acceptance criteria is not more than 7 minor and not more than 2 major, then one minor is equivalent to 2/7 major.

Therefore, Conditions for acceptance should be:
not more than 7 minor
not more than 2 major
not more than 1 major + 3 minor
anything more - reject.

But this is just looking from the logical point of view.

The sampling tables work well in ideal situations.They are pure theory, basesd on probability.We can neither ignore the sampling tables nor strictly adhere to them. The quality inspector has to use lot of his experiance and discretion in using the tables.The Quality plan /control plan specification on AQL has to be fixed carefully, because Control Plan is mandatory.
V.J.Brahmaiah

Hello,

I am currently using the AQL system for my work but I think there is a problem with this system.

This system gives acceptance or rejection thresholds for major and minor defects but sometimes it can lead to an incherence.

Let me explain this problem with an exemple:

If we have for instance:
-Minor defects: acceptance 7, rejection 8
-Major defects: acceptance 2, rejection 3

If we find 8 minor defects and 0 major defects we reject the production.
If we find 7 minor defects and 1 or even 2 major defects we accept de production.

I think this is a real problem because the second case is more serious than the first one and we accept it while we reject the first one.

The proof of this problem is that we can easily put in order the different cases bellow, from de least serious to the most serious:

8 minor defects and 0 major defects < 7 minor defects and 1 major defect < 7 minor defects and 2 major defects.


Can someone help me to understand why AQL works like that?

Thanks a lot.

Damien


When using attribute inspection, and an AQL, one has sevral choices to make. One can determine what is the acceptable AQL for major defects and apply this sampling plan. One can also determine what is the acceptable AQL for minor defects, and apply this plan. These are two separate actions, and do not "add" together, so the mathematics of the situation is irrelevent. If you wanted to have a quality level score that included major and minor defects together, you could decide that, one major defect was equivalent to, say 3 minor defects. And then decide what was the appropriate AQL and sampling plan for this "blended" score. AQL works the way it does because the tables were set up to look at one AQL and sample at a time, not to "add" results of different samples and AQL levels.

Geoff Withnell

Thanks for your answers. It is more clear now.

Damien.

The sampling tables work well in ideal situations.They are pure theory, basesd on probability.We can neither ignore the sampling tables nor strictly adhere to them. The quality inspector has to use lot of his experiance and discretion in using the tables.
This may or may not be true. If the inspector has experience, we may allow the person to use that experience to influence the decision. If the inspector doesn't have the right experience, we shouldn't rely on their experience to make judgment on the lot. More often than not, I see that the rules are set by the organization and must be followed.
The Quality plan /control plan specification on AQL has to be fixed carefully, because Control Plan is mandatory.
V.J.Brahmaiah
Control plans are mandatory only for TS 16949. I'm not aware of any other standards or regulations that require them. Of course, TS 16949 only allows acceptance with zero defects (C=0) for attribute sampling.

Hey Brah!!

I was having some similar problems and questions a few days ago and these guys really helped. Since i had to do a lot of research on AQL, I know what you mean about getting confused with the way you look at the chart.

My problem being that we do not accept defects at all. Either in the parts we make or recieve from suppliers. So the Zero Based Acceptance Plan will probably help us because we inspect 100% but write down only "about" 10% of the results, AS9100 requires that I have a "statistically valid" plan that indicates how I have chosen to only write down a certain percentage instead of every one.

So..I have attached the procedure with the ZBA Plan I got. (from the book by Nicholas Squeglia), I found this on line and then more info on pdf from the DoD on MIL-STD-1916. Lots of stuff I downloaded and put together that helped me to understand some, how it works, hopefully enough to get by.

I also found a 30 day free trial on-line for AQL software that I played with all week. The more I used it and the more I read up, the more I knew about how to enter the correct info to get the right results from each plan and chart it makes. Search "Taylor Enterprises-Sampling Plan Analyzer" great product, but I don't think my boss will buy it for me....yet.

:agree1: Well enough from me...good Luck!!
QMSLADY

Good insights from people who are using the tools. Here's another spin on this that I would like to get some feedback on ... it is really about using AOQL.

We have a Chinese supplier who will be a making a product and they always want to drop to the bottome line ... "What is the warranty expense? How many extra do I need to make to ensure that if there are problems in the field I have made enough to cover it?"

Obviously, not a process-driven quality approach because they are planning for some level of defects and not designing to a level of process quality. We want to work in a conservative AOQL to an overall agreement so that if there are escapes, we have some leverage. Here's the rub. On similar product built in the US or at South of the Border Contract Manufacturers the Field Failure rate is 0.25% with most of the fall-out happening early in production / deployment with no defined common failure modes (really IPC workmanship issues and PCBA process control lapses resulting in a multiple distributions of failures).

What is a good AOQL (that we can measure from field returns within a 2-3 year warranty period) that gives us protection for start-up concerns in China?