AQL - How do I find Acceptable Quality Level?

Thank You Gordan,

For your help on this procedure and the wording. Sorry to everyone if I sounded a little (ok a lot) grumpy yesterday. I have revised it again and feel much more confident this time.

Please take a look and tell me what you think. I have changed a lot and am still apprehensive about some of the wording and "redunantcy". I have put the chart back in, but if you read 2.0 and 4.0 you'll see why. It may be that I can do without it still.

But as I entered just below it, the wording is from DoD, not mine. (Could probably leave that out too) I don't know...

This forum has been very helpful to me. Not just this thread, I have also been reading in other areas as well. And I have made a few friends...I hope

Thanks to all of you who helped, (even you Bev) for making me think.
QMSLADY

Bev D said:

I think the troubling part that remains unanswered is the random recording. We understand that you are inmeasuring all dimensions on all parts. Good without SPC and proven capability most arespace Customers require 100% final inspection.

However, you still need to justify why you are only recording some of the measurements. You also need to specify how you will "randomly" determine which parts will be recorded. Are you going to use a random number generator each time?

Click to expand...

I am also troubled...for two reasons.

1) If 100% inspection is being done because a customer wants it, wouldn't that customer also expect there to be evidence of the 100% inspection? That is, inspect 100% and record 100%, or inspect 10% and record that same 10%.

2) The sampling plan's entire purpose is to inspect less with some defined degree of confidence (that degree of confidence is what gets decided on as AQL) because costs of inspecting 100% are high. Inspecting 100% and writing down 10% is not sampling.

Am I missing something here?

Have you received permission to reprint that table? ASQ sells those sampling plans, and to my knowledge they are copyrighted material.

Jennifer Kirley said:

I am also troubled...for two reasons.

1) If 100% inspection is being done because a customer wants it, wouldn't that customer also expect there to be evidence of the 100% inspection? That is, inspect 100% and record 100%, or inspect 10% and record that same 10%.

2) The sampling plan's entire purpose is to inspect less with some defined degree of confidence (that degree of confidence is what gets decided on as AQL) because costs of inspecting 100% are high. Inspecting 100% and writing down 10% is not sampling.

Am I missing something here?

Have you received permission to reprint that table? ASQ sells those sampling plans, and to my knowledge they are copyrighted material.

Click to expand...

Thanks Jen,

Yes...to all the above, I do hear your point, but read further in the threads or just skip to this mornings post by me to Gordan and view the Procedure again...just forget the "word" random...it no longer exists. ")
qmslady

Sorry. I've read the revised procedure and I don't see that the initial confusion has been cleared up regarding recording of only a partial set of results. You seem to be changing words but not the meaning of your sentences. You point to the AQL Plan but provide no guidance on how to determine how many values will be recorded nor how you will determine which of values will be selected. Which "AQL level" (for recording purposes) do they choose? It is also not sufficient statistical justification for recording only a portion of the results to simply point to a statistical sampling plan table.

More importantly, WHY are you determined to record some smaller portion of the values? you seem to be very adamant about it, so if we can get to the heart of that question, perhaps we can provide better guidance as to the real justification for it. What is it about a recording a portion of the results that is more valuable - or less wasteful - to you than recording all of them, or just the min/max values?

additionally the wording "first part, during and after machining" doesn't clear up the instruction. Are you trying to say that each part will be measured after it has completed the machining operation, even if other parts are not yet complete? In other words you are saying "don't wait until they are all done before measuring them"?

Jennifer Kirley said:

I am also troubled...for two reasons.

Have you received permission to reprint that table? ASQ sells those sampling plans, and to my knowledge they are copyrighted material.

Click to expand...

I'm not "troubled" (puzzled?) but I must say the two points are very relevant.

I honestly still don't see the need for a sampling plan - at least certainly not the one shown. If it is a "must" (for some obscure reason) then it'd be easier to make a simple one relevant to the number quantity your company produces.

I also read the procedure, and it clearly enough states that parts will be 100% inspected but recorded as per the sampling plan. That is not acceptance sampling. Period. Those tables are not intended for the purpose of minimizing recording, they are intended for the purpose of minimizing inspection where it's allowable.

Just a thought or the beginning of an idea. Why don't you mark all dimensions on the customer drawing with 3 colours?
Red could mean 100% inspection and measurement documentation.
Yellow could be 100% inspection by gauge.
Green could be a sample size - i.e. 1:3 or 1:5 (or whatever) after each dimension.
The colours would/could also signify criticality.
Your customer would then know you'd been over the drawing carefully and given it some thought.
It's just an idea, so any remarks/suggestions would be regarded as a plus rather than critique

I'd definately drop the "official" standard sampling plan.

By the way - it's GORDON with two Os
My name is common in English speaking countries, but I'm the only one in Denmark with that name

Suggestion for a revision to your procedure

Hi,

I've ben reading yours and Bev's stuff ;

(Bev)"It is also not sufficient statistical justification for recording only a portion of the results to simply point to a statistical sampling plan table."

O.K. (I think)

More importantly, WHY are you determined to record some smaller portion of the values? you seem to be very adamant about it, so if we can get to the heart of that question, perhaps we can provide better guidance as to the real justification for it. What is it about a recording a portion of the results that is more valuable - or less wasteful - to you than recording all of them, or just the min/max values?

"I" am not the one making these rules. There are procedures that B.T. had in place before I came here. If they are spinning "their" wheels by recording useless information, then I need to tell them to stop. But...

According to AS9100-2009, 7.1 (c & d) and Procedure 7.1.2-RM, Risk Management and the "requirements of the organization"???

I went out on the floor and questioned some of the guys about how they determine these things since you all put the question to me. They are writing down "critical dims" for parts with less than .002 tolerance and using the 10 % method. I asked one guy, after picking up a part, "which one is this?" He smiled and shrugged??? I asked "How do I know you aren't writing down the same piece over and over?" He says "I guess I'll have to write it on the part". O.K.

Now...How does he justify the sequence? We came up with dividing the AQL number under 1.0% into the lot size. i.e. 90 parts '/. 13=6.92 or every seven parts. O.K. But now, how do we justify using the AQL 1.0% ??? How do we determine that number?

(Bev)"additionally the wording "first part, during and after machining" doesn't clear up the instruction. Are you trying to say that each part will be measured after it has completed the machining operation, even if other parts are not yet complete? In other words you are saying "don't wait until they are all done before measuring them"? No

I fixed that part to be more specific,

Gotta go...will check in tomorrow...Thx..qms

AQL is roughly the fraction defective that would usually be accepted. If AQL = 1, then most lots with 1% defective would be accepted. The defect rate that will usually be rejected is several times higher. This doesn't really seem to be what you want.

If you want to be (fairly) sure that (most of) the parts are good, then this thread on reliability and sample size gives some good advice. https://elsmar.com/Forums/showthread.php?t=33830 (Note that the sample sizes are nearly nearly independent of sample size. So picking a fixed samlpe size - like "60" - is statistically more valid than picking a fixed % -- like "10%".)

If you can record & analyze the actual numbers, then you might be able to justify smaller sample sizes. For example, 30 good parts all near the center of the specs would be much better than 30 good parts all hugging one spec limit.




It almost sounds like you are saying

• "we test all the parts because it is valuable and important that all the parts are good."
• "we record some of the results just because it was written into the procedures."

This might be a good place to try some root cause analysis to find out why the procedures are set up this way - like "5 Whys". "Why do we record just some of the data?" ... "Then why ...?"


Tim F