The Elsmar Cove Wiki More Free Files The Elsmar Cove Forums Discussion Thread Index Post Attachments Listing Failure Modes Services and Solutions to Problems Elsmar cove Forums Main Page Elsmar Cove Home Page
Google
  Web Elsmar.com
*Please be aware that SOME RECENT forum threads may not yet be indexed by Google.

View Full Version : Need to set up sampling plans in Receiving Inspection


Reggie
16th April 2007, 02:17 PM
Hello all,
I need help setting up sampling plans, not sure how to set up the plan or where to start, I tried explaining that we need to find the AQL but they just want to start.
Any help in setting up the AQL or the OC Curves would be helpful.
Regards,
Liz Piotrowski

ScottK
16th April 2007, 02:34 PM
(I'm assuming attribute sampling)

the AQL is kind of key. Can't use the charts without it.

What percent defective is management willing to accept?

Alex Kobzar
16th April 2007, 02:48 PM
I'd recommend to obtain a precision guidance on this task from Quality Manager of your Company.
Also I'd suggest you to bear in mind that your goal is to keep the Incoming QC office comfortably busy.

ScottK
16th April 2007, 03:34 PM
Also I'd suggest you to bear in mind that your goal is to keep the Incoming QC office comfortably busy.

Isn't that just creating work for the sake of looking busy?

Alex Kobzar
16th April 2007, 03:54 PM
Isn't that just creating work for the sake of looking busy?

I guess that at some places creating work for the sake of looking busy may be the main purpose of an Incoming QC.
For example, in a situation where there are ppl willing to start doing incoming QC and there is no plan on why and how, then the goal of keeping the office busy may prevail.

Jim Wynne
16th April 2007, 04:23 PM
Hello all,
I need help setting up sampling plans, not sure how to set up the plan or where to start, I tried explaining that we need to find the AQL but they just want to start.
Any help in setting up the AQL or the OC Curves would be helpful.
Regards,
Liz Piotrowski

Do a search of the Cove, starting with the related threads down towards the end of this page. The first one includes a spreadsheet from Tim Folkerts that you might find useful. You should also remember that random sampling is important; all bets are off with regard to sampling efficacy if the sampling isn't random (meaning that each member of the sampled population has an equal chance of being selected).

Reggie
16th April 2007, 04:25 PM
I need help in how to do the AQL, it is not just to keep QC Busy. My manager is not sure on the AQL. Any suggestions

Alex Kobzar
16th April 2007, 04:38 PM
Keeping the office busy is not such a bad goal as it may sound.
See what you can do using your resources, then determine AQL using this knowledge.
Then you will be able to adjust the AQL on specific items (e.g. by going to increased or reduced inspection when problems or no problems will be found).

ScottK
16th April 2007, 04:59 PM
I need help in how to do the AQL, it is not just to keep QC Busy. My manager is not sure on the AQL. Any suggestions


AQL is the wost quality level that is still considered satisfactory.
to choose an AQL you need to know how many defectives are allowed per 100 parts.

ScottK
16th April 2007, 05:02 PM
Keeping the office busy is not such a bad goal as it may sound.


I'd rather keep my staff busy being productive.
Inspection after the fact, while required in some cases, is just not productive.
You can't inspect quality into a part, right?

Tim Folkerts
16th April 2007, 05:13 PM
AQL is basically the level of quality that will usually be accepted. The level that will usually be rejected is typically 3-10 times higher, depending on the particular sampling plan. And depending on what you mean by "usually"! All of this can be found in the OC curves and the tables with them in th Z1.4 standards.

Setting the AQL should be a business decision based on some sort of cost-benefit analysis based on factors like:

how much does inspection cost?
how much does it cost when you let bad pieces through?
how many bad pieces do you typically get?
how likely are you to get an especially bad lot?From such info, you can start to determine the benefit of testing.

Also, when possible, it is much better to use a variable data (ie a continuous variable), rather than attribute data (i.e. pass/fail type data). You can get much better results from much smaller sample sizes.

Better yet, use SPC so that inspection isn't needed.


Tim F

reynald
16th April 2007, 11:32 PM
Just for reference:
http://www.variation.com/anonftp/pub/milstd105e.pdf

Al Rosen
17th April 2007, 11:42 AM
Look at Selecting Statistically Valid Sampling Plans. (http://www.variation.com/techlib/as-7.html)