Help interpreting MIL-STD-105E Single Sampling Plans Tables

[Mark Meer]

Hi All,

I'm needing some help interpreting some of the tables in the MIL-STD-105 (E) standard. In particular the Single Sampling Plans tables (Table II).

My confusion is this: how/why is it that the AQLs are specified up to 1000?

Aren't AQLs expressed as percentages? In practice, is there any reason to even consider the columns of AQL >= 100?

Also...
I know that MIL-STD-105E has been superceded by ANSI/ASQC Z1.4, but as I understand, they are still quite similar. So for anyone (like me) just wanting to explore acceptance sampling without having to buy the ANSI/ASQC standard, the MIL-STD-105E can still be a good starting point.
Can anyone familiar with the differences between the two please confirm?

Thanks in advance!

 

[WCHorn]

Re: Help interpreting MIL-STD-105E tables

There is the potential for defects per piece inspected, so the higher AQLs are for that purpose.

Z1.4 and 105E yield the same sample sizes and accept/reject numbers. There may be some subtle differences but I don't think those differences affect the accept/reject decision.

 

[Mark Meer]

Re: Help interpreting MIL-STD-105E tables

Thanks for the prompt reply WCHorn.
...but I'm afraid my statistically-challenged brain is still confused.

I've seen some industry guidances, and SOPs from other companies, specifying a default AQL of 1.0. This makes sense to me: we want confidence that the percent defects in any given accepted lot are never above 1%.
But an AQL of 65-100, let alone 101-1000? Doesn't make sense...

Also, the acceptance and rejection numbers for AQL >= 65 are larger than the total sample sizes. This too, I find very confusing...

In the definition note for AQL, the standard states:
"...the AQL is a designated value of the percent defective...for which lots will be accepted most of the time..."

So, doesn't this mean that selecting an AQL of 100 would allow lot acceptance even if 100% of the samples were found defective? (which clearly makes no sense... ).

Can you possibly give (or reference) a simple example in which these higher AQLs might be useful?

 

[WCHorn]

Re: Help interpreting MIL-STD-105E tables

This makes sense to me: we want confidence that the percent defects in any given accepted lot are never above 1%.

Starting here, you are actually assuring that the average percent lot defective of a series of lots is 1 percent or less (assuming you're using the normal/tightened/reduced switching procedures). If you want to limit the percent lot defective in a single lot, the standard gives limiting quality (LQ) plans for that purpose.

Can you possibly give (or reference) a simple example in which these higher AQLs might be useful?

If you're making something that has the possibility of several defects per unit (an automotive fender with paint defects, for example, where there is a limit of, say, 5 per unit), then you could use an AQL of 250 (250 defects per hundred units). That would mean the average defects per unit would be 2.5 or less, again using the switching procedures.

If a single defect on a unit makes the unit a defect, then you're perception is correct; it doesn't make sense. If there are some number of defects per unit allowed before the unit is declared defective, then the larger numbers make sense. I'm sure other's on the Cove may be able to give you better examples. I made this one up.

 

[David-D]

It goes back to the earlier poster's comments about counting defects, rather than defective units. For example you'd count how many total scratches you had on your sample of 3 doors.

For the example of a sample size of 3 (code B) with an AQL of 1000%, you would inspect three items and if there were 44 or less defects (scratches, gouges, etc) on all 3 of the sample doors in total. If there are 45 or more defects, the lot would be rejected.

As it is a 1000% AQL, if the supplier was running at 10 defects per item, them they would consistently have the product accepted.

Hope that helps.

David

 

[Mark Meer]

Ah-ha! Thanks! The numbers make sense now.

Obviously I was (mistakenly) assuming that each unit was either declared defective or not (pass/fail type inspection).
...I had not considered the possibility of multiple defects per unit...

That being said, it's hard to imagine ever considering 30 defects in 2 samples (Sample code A, AQL=1000) to be acceptable. I guess if said defects were really minor...

In anycase, at least I understand the numbers now... thanks again!

...now, on with the rest of the standard....