Audits in Volume Manufacturing vs. Confidence/AQL Levels

S

Steve1152

Hi,

In a situation where 2000 units are being manufactured per day and the current audit plan is to sample 10% of the lot (200 pcs.), what confidence level or equivant AQL is this?

If it changes to a 5% audit audit per day (100 pcs.), what confidence level or AQL does this go to?

Thanks, Steve
 

Jim Wynne

Leader
Admin
Re: Question on Audits in Volume Manufacturing vs. confidence/AQL levels

Hi,

In a situation where 2000 units are being manufactured per day and the current audit plan is to sample 10% of the lot (200 pcs.), what confidence level or equivant AQL is this?

If it changes to a 5% audit audit per day (100 pcs.), what confidence level or AQL does this go to?

Thanks, Steve

If there's a lot size of 2000 that has a single defective, it's a different situation than if there are 200 defectives. Your chances of finding defective units increases as the sample size increases, but you need to start by determining the level of risk you're willing to assume. The type of inspection and its associated costs are also relevant. If you're looking for one black thing in a 2000-piece lot of white things, it's a lot different than if there's time-consuming testing involved.

In short, there's no way to make a backwards calculation of the sort you're looking for. If you can tell us more about what you're doing, we can help with advice on sampling.
 

AndyN

Moved On
Re: Question on Audits in Volume Manufacturing vs. Confidence/AQL Levels

Sampling like this is normally based on the tables found in Mil-I-105E (or whatever the equivalent is these days). The sampling tables are derived from equating produces/users risks - a bit like a capability study. The result is that the closer you get to accepting zero risk, the higher the sampling becomes, given a lot size.

Generally - and as Jim is pointing out - it's not a simple case of taking a percentage based sample and halving it. There's more to it than that. I've used an AQL in Receiving Inspection, at 0.4% and inspected 32 of just about everything that came through the door - until, one day it hit me, that we were doing the wrong thing! The issue is, unless you do a (capability) study on process/part variation, your sampling is only going to be like a wet finger held into the air!
 
S

Steve1152

Re: Question on Audits in Volume Manufacturing vs. confidence/AQL levels

If there's a lot size of 2000 that has a single defective, it's a different situation than if there are 200 defectives. Your chances of finding defective units increases as the sample size increases, but you need to start by determining the level of risk you're willing to assume. The type of inspection and its associated costs are also relevant. If you're looking for one black thing in a 2000-piece lot of white things, it's a lot different than if there's time-consuming testing involved.

In short, there's no way to make a backwards calculation of the sort you're looking for. If you can tell us more about what you're doing, we can help with advice on sampling.
Good point. This audit is taking units from production into a 5 hour Burn-In at Room Temperature and afterwards a functional test is performed. Typically the lot is accepted when there are zero failures. The current sample has been 10% for several years but I wanted to know if it goes to a lower sample percentage (of total lot size) of say 5% what confidence does that still give us. - Steve
 
S

Steve1152

Re: Question on Audits in Volume Manufacturing vs. Confidence/AQL Levels

Sampling like this is normally based on the tables found in Mil-I-105E (or whatever the equivalent is these days). The sampling tables are derived from equating produces/users risks - a bit like a capability study. The result is that the closer you get to accepting zero risk, the higher the sampling becomes, given a lot size.

Generally - and as Jim is pointing out - it's not a simple case of taking a percentage based sample and halving it. There's more to it than that. I've used an AQL in Receiving Inspection, at 0.4% and inspected 32 of just about everything that came through the door - until, one day it hit me, that we were doing the wrong thing! The issue is, unless you do a (capability) study on process/part variation, your sampling is only going to be like a wet finger held into the air!
Thanks. I did study the standard ANSI ASQ Z1.4-2008 Sampling Plan procedure to get a feel for the statistics today. When you say capability study, are you referring to Cpk or something similar? Let me know. Thanks, Steve
 

AndyN

Moved On
Re: Question on Audits in Volume Manufacturing vs. Confidence/AQL Levels

Thanks. I did study the standard ANSI ASQ Z1.4-2008 Sampling Plan procedure to get a feel for the statistics today. When you say capability study, are you referring to Cpk or something similar? Let me know. Thanks, Steve

Not really, Steve. From memory (and I was using BS 6001 in those days) you need to study the process variation over time (somewhat like the traditional 50 piece study) to get a feel for the natural variation. Don't forget that much of these AQL sampling plans are directed at sampling by attributes - it's there or not - rather than by variables. In my situation, someone had set it up as a sample plan by variables, then treated it as an attribute - which is when I decided that we could do anything we wanted, it wouldn't make a difference!

Reading through the variables section, it seems that by the time you're done, you've pretty much led yourself to a fully fledged SPC technique.
 
Last edited:

Jim Wynne

Leader
Admin
Re: Question on Audits in Volume Manufacturing vs. confidence/AQL levels

Good point. This audit is taking units from production into a 5 hour Burn-In at Room Temperature and afterwards a functional test is performed. Typically the lot is accepted when there are zero failures. The current sample has been 10% for several years but I wanted to know if it goes to a lower sample percentage (of total lot size) of say 5% what confidence does that still give us. - Steve

If you've been doing this sampling/testing for several years, you should know something about the efficacy of it. If your urge is to reduce the number of units tested because of past results, you should already know what you need to know, no?

Remember that what acceptance sampling should do for is help you to predict that if there are x defectives in a lot of y population size you have a z% of discovering at least one, with a given confidence level. If your current scheme is working--i.e., the number of escapes is acceptably low--I wouldn't be so quick to try and do a lot of statistics. You don't need a weatherman to know which way the wind blows.
 

Tim Folkerts

Trusted Information Resource
Re: Question on Audits in Volume Manufacturing vs. Confidence/AQL Levels

There was a thread a while back on confidence and reliability for sampling. Check here: elsmar.com/Forums/showthread.php?t=33830

It includes a few tables, some formulas, and then a spreadsheet I created. For a sample of 200 with no defects, you can be ...

  • 86% confident that at least 99% are good
  • 95% confident that at least 98.5% are good
  • 99.99% confident that at least 95% are good

If you reduce that to a sample of 100, then you can be

  • 63% confident that at least 99% are good
  • 95% confident that at least 97% are good
  • 99.4% confident that at least 95% are good

There are any number of other combinations for either sample size that would also work.


Tim
 
J

JJC1979

Re: Question on Audits in Volume Manufacturing vs. Confidence/AQL Levels

Reviewing the ANSI Z1.4 is a good start to determine what you are looking for. My thought is to look back at the history and see what you have been producing and what defects you have found, if any. If you produce 2000 pcs and sample 10%, this is a standard place to be in the realm of AQL levels. It is like an AQL of 1.0 to 1.5 with normal inspection. Have you noticed any returns back from customers? This will clue you in on if you can decrease the inspection amount. If your returns are high, your inspection is doing nothing for you...especially if you are inspecting 200 pcs and find no defects.

I would say do not decrease unless the history and data shows you can you possibly can, without a doubt. Always follow your data. And if you have no data yet, start at 10% first until you obtain enough data.
 
S

Steve1152

Re: Question on Audits in Volume Manufacturing vs. Confidence/AQL Levels

In your tables and analysis, does a lot size of 2000 fall into the range of your confidence calculations? I say this since we are taking this sample from a daily lot size in production of 2000 units. Thanks, Steve
 
Top Bottom