Use of sampling plans like MIL-STD-105 E or ANSI Z1.4 for setting AQL's?

AndyN

Moved On
I have a question for my fellow mods. Does anyone here have a good grasp of the use of sampling plans like MIL 105 E or ANSI Z1.4? I'd like to discuss the setting of AQL's. It's been bugging me for over 20 years and I'd like to lay it to rest!!
 

Coury Ferguson

Moderator here to help
Trusted Information Resource
Re: Pick your brains - Use of sampling plans like MIL-STD-105 E or ANSI Z1.4?

I have a question for my fellow mods. Does anyone here have a good grasp of the use of sampling plans like MIL 105 E or ANSI Z1.4? I'd like to discuss the setting of AQL's. It's been bugging me for over 20 years and I'd like to lay it to rest!!

Do you have copies of the standards? I can only help a little since I haven't used sampling plans in over 20 years.

If you need a copy of the standards, I can send them to you.
 

AndyN

Moved On
Re: Pick your brains - Use of sampling plans like MIL-STD-105 E or ANSI Z1.4?

I have a copy of the ANSI, it's the 'logic' I need some guidance with, Coury. Thanks!
 

Tim Folkerts

Trusted Information Resource
Re: Pick your brains - Use of sampling plans like MIL-STD-105 E or ANSI Z1.4?

A variety of moderators have experience with these standards, I'm sure. I know the statistics pretty well, although setting the AQL often has little to do with the statistics and more to do with economics or simple "comfort level".

Pose your question and I'm sure you will get some responses.


Tim F
 

GStough

Leader
Super Moderator
Re: Pick your brains - Use of sampling plans like MIL-STD-105 E or ANSI Z1.4?

I have a copy of the ANSI, it's the 'logic' I need some guidance with, Coury. Thanks!

I'm not sure if I can help, Andy, but what is your question?
 

AndyN

Moved On
Re: Pick your brains - Use of sampling plans like MIL-STD-105 E or ANSI Z1.4?

O.K - here's some background, then my question(s)

I inherited, many years ago, the Receiving Function. A predecessor had set an AQL on all incoming parts at 0.4%, so on a (usual) lot size of 151 - 280 (letter 'G') we sampled 32, accepted 0 defectives and rejected on 1 defective found. I'm cool with the use of the tables!

I could never get a satisfactory answer on why a 0.4% AQL was chosen and it has bugged me ever since (well, maybe once a year when I indulge in some deja vu of my enjoyable worklife experiences!!)

If I read table X-H-1, on page 46 of the ANSI Z1.4 (1993) doc, I need to reconcile my expected percentage of lots to be accepted 'Pa', with the 'p' value (% non-conforming) of submitted product, to derive my AQL, I think.

So, if I expect that 99% of received product is conforming, then I choose an AQL of 4.0%, and follow the acceptance/rejection rules from table II A (page 12) - I'm actually (statistically) likely to 'see' 3.69% defectives (from estimating on the curve or reading off the table)

Did I get the rationale correct? If so, was it simply a guestimate if the originator of the 0.4% AQL if they couldn't explain what the 'Pa' value was or the impact (risk of the 'p' value), either?

Or am I hosed in my understanding?
 
Last edited:

Coury Ferguson

Moderator here to help
Trusted Information Resource
Re: Pick your brains - Use of sampling plans like MIL-STD-105 E or ANSI Z1.4?

O.K - here's some background, then my question(s)

I inherited, many years ago, the Receiving Function. A predecessor had set an AQL on all incoming parts at 0.4%, so on a (usual) lot size of 151 - 280 (letter 'G') we sampled 32, accepted 0 defectives and rejected on 1 defective found. I'm cool with the use of the tables!

I could never get a satisfactory answer on why a 0.4% AQL was chosen and it has bugged me ever since (well, maybe once a year when I indulge in some deja vu of my enjoyable worklife experiences!!)

If I read table X-H-1, on page 46 of the ANSI Z1.4 (1993) doc, I need to reconcile my expected percentage of lots to be accepted 'Pa', with the 'p' value (% non-conforming) of submitted product, to derive my AQL, I think.

So, if I expect that 99% of received product is conforming, then I choose an AQL of 4.0%, and follow the acceptance/rejection rules from table II A (page 12) - I'm actually (statistically) likely to 'see' 3.69% defectives (from estimating on the curve or reading off the table)

Did I get the rationale correct? If so, was simply a guestimate if the originator of the 0.4% AQL if they couldn't explain what the 'Pa' value was or the impact (risk of the 'p' value), either?

Or am I hosed in my understanding?

Andy,

The choice of the AQL has really nothing to do with Stats. It is based upon what they are willing to take the risk of a number of unacceptable product based upon certain factors.

When the AQL is specified, it is usually identified like this:

4.0 Minor characteristics (+/- .010)

2.5 Major characteristics (+/- .005)

.65 Critical characteristics (+/- .001)

You should download MIL-STD-105E it would give you a basic understanding of the logic. If you want to download it, which is free at this website: Standards

That would be your best bet.
 

AndyN

Moved On
Re: Pick your brains - Use of sampling plans like MIL-STD-105 E or ANSI Z1.4?

O.K so this is where I start getting frustrated - ANSI Z1.4 is for sampling by attributes - not variables, so I can't see how tolerances, as you've indicated Coury, have any bearing on the situation I described........

All I wanted to do was to have a good 'story' on the 'agreement' to use an AQL - when parts arrived, I merely wanted to accept them, (based on finding 0 defects) or reject the lot, (based on finding 1 or more) in the sample.

I didn't care (back in the day) to negotiate (with the supplier) on the amount of variation in their product characteristic(s), if they weren't to print, then I'd reject them - purely because I didn't want to waste money on sorting etc. If I found 1 defective, then statistically there's likely to be more in there, so to speak........

Thanks for the link - but it didn't work for me for some reason
 
Top Bottom