9 replies to this topic

[Joncer Naibaho]

hello, my name is Joncer.

help me to create sampling plan, how to make a decision the percent of aql which we can use.

 

thank you

[Snookie]

 

We will need a lot more information in order to help.  What type of product you make, what the purpose of the sampling plan is for example do you mean environmental, raw product, final product. 

[RG3]

 Joncer Naibaho. We will definitely need more information. Especially for AQL.

[mdean1124]

There are numerous publications on the topic of "acceptance sampling".  Look for a nice paper by Nicholas Squeglia called, "Zero Acceptance Number Sampling Plans" from ASQ.  It's a good place to start on the industrial side.  Also see...

 

Ryan, T.P. (2000). Statistical Methods for Quality Improvement, 2nd ed., Wiley, New York, NY.

 

Army Chemical Corps (1953). Master Sampling Plans for Single, Duplicate, Double and Multiple Sampling, Manual No. 2. (This is the basis for NUMEROUS texts on the subject)

 

ISO 7966 Acceptance Control Charts

 

And tons of free "stuff" on the internet! 

 

You might also join a web site called the "C*#!! C*#!".  There are lots of FREE quality management documents and references there... including some spreadsheets and databases with everything you need to set up and run sampling plans.  There's a small fee... but it's worth every penny!

 

Good luck!

 

 

[Joncer Naibaho]

thanks for all,

snookie and RG3, i have try to create and almost finish. but i have a problem about to make a decision in AQL. there are some choices about what the percent we will use. it is like 0,010; 0,015; 0,025; 0,040, .... how do i choose that, after make condition analysis.

 

mdean112, thanks your advise, i will..

[Charles.C]

thanks for all,

snookie and RG3, i have try to create and almost finish. but i have a problem about to make a decision in AQL. there are some choices about what the percent we will use. it is like 0,010; 0,015; 0,025; 0,040, .... how do i choose that, after make condition analysis.

 

mdean112, thanks your advise, i will..

 

It depends on yr quality specification. And the specific sampling situation/level. For example bananas might well be different to steel bars.

 

AQL means what it says, eg, for example,  AQL = 4.0 means the number of defective cans which can be tolerated per 100 cans in a batch is 4. 

 

One can predict that, following the tabulated procedure (ie select sample size/AQL=X), lots with true values of X% defective will, in the long run, have a, typically, 95% probability of acceptance. The details are derived from OC curves.

 

Rgds / Charles.C

[Joncer Naibaho]

great!

thankyou charles, i will text soon what i have created. i need more your advise. i can not now because i need to translate in to english.

 

kind regards

[Charles.C]

Dear Joncer Naibaho,

 

The relevance will depend on yr specific objective but here is a litlle more info. on sampling. TBH it's an enormous field.

 

Just FYI, as a general comment, there are 2 common, simple, approaches to sampling lots, ie (1) if the primary objective is to accept/reject the lot, (2) if the primary objective is to estimate the percentage of some quality -related characteristic of the lot.

 

The theoretical bases for 1,2 are inter-related but the equations tend to have a different emphasis.

 

The usual practical problem with the normal AQL tables is that the sample sizes are impossibly large, especially if destructive sampling is involved. if so one zigzag is to make use of the "special" sampling tables ("S" from memory) also included and which allow much smaller samples.

 

Some treatments convert AQL tables (category 2) into category (1) by re-stating that a specific lot has, with for example 95% confidence, a defect% <= X if the measured total defect number for a tabled sample size matched to AQL=X complies with the maximum defect number in the AQL=X column. This is theoretically questionable but often an acceptable approximation for simple random sampling. Similar results can be obtained via basic sampling theory which ignores lot size for adequately small samples.

 

And then comes ANOVA etc.

 

Rgds / Charles.C

[Joncer Naibaho]

Hi Mr.Charles,,Please see attached.

 

Need your advice to develop it.

 

Kind regards

 

Joncer

Attached Files

•  SAMPLING PLAN English Version.docx   15.89KB   220 downloads

[Charles.C]

Hi Mr.Charles,,Please see attached.

 

Need your advice to develop it.

 

Kind regards

 

Joncer

Dear Joncer Naibaho,

 

Thks for the attached "sampling plan".

 

The file appears to be a description of the process and target limits for various process variables.

 

I don't see any proposed sampling plan unless I have lost part of the document (I had to convert .docx to .doc so maybe some error at my end) ?

 

There appear to be 7 stages in the process including the final QC activity.

 

Which variables / stages are you trying to design a sampling plan for ?

 

Rgds / Charles.C

 

PS (added later)  –

 

Offhand, yr attachment seems  more oriented to an analysis of the type noted in category (2) of my  previous post rather than an AQL-based implementation. The point is that one method estimates the value for a given lot of a variable like moisture so that you can then decide what to do with the lot. The other, AQL, approach attempts to ensure acceptance of lots within  a certain defect level (ie the AQL)  while also acknowledging that some incorrect decisions will also occur, ie good lots be rejected, bad lots be accepted. The choice depends on your objective.

 

Just as an example, here is a quote from a  Quality Control  text –

As stated earlier, AQL, by definition, represents that quality level which, for purposes of sampling inspection, can be considered satisfactory as a process average. The probability of acceptance of lots with a quality level equal to AQL is usually 95%.

AQL levels are normally specified by the procurer. For example, the government may specify a 1.0% AQL inspection plan for a contracted item. If  the manufacturer, based on previous production experience, or from intimate knowledge of his process capability, has established that his production indeed runs at 1% defective, he may anticipate acceptance of 95 of his 100 production lots. Knowledge of how many lots will, in all likelihood, be rejected, enables the manufacturer to compute his anticipated cost of manufacture under this plan.

AQL plans are basically producer-oriented; they keep the latter’s risk of having lots rejected, even though they meet the specification sampling criteria to 5% (or to 1 lot in 20). The consumer protection afforded by AQL plans in MIL-STD-105D varies considerably from plan to plan. Unlike the producer’s protection, it is not maintained at any specific value.

 

(Note that the term "quality level" refers to percent defectives)