How to interpret AQL sampling tables

[v9991]

need guidance on interpretation of the sampling plan tables.

Background / Exclusions :-
* we operating Sample inspection level -II criteria ( thumb rule~ no questions asked on this! at this point of time)
* so viz., for 'lot size' of 3000, 'sample size code' is K
* so, we determine that the no. of samples is 125.
* Critical (0.065) - Major ( 0.25) - Other (1.5)
example for AQL table..https://elsmar.com/elsmarqualityforum/attachments/aqlchart-pdf.24029/

Q1. No. of samples to be selected.
option-1 :- as the AQL table indicates that for critical defects @ 0.065%, we require atleast 200 samples, we proceed to sample 200 and inspect for approving/rejecting
option-2:- as aql table is selected ( even though its an thumb rule at this point of time) we stick to 125 sample inspection for approving/rejecting

counter argument option-1 :- we can explain in certain instances, where more samples as worst case scenario, but same is not true for other scenarios, where samples required although higher, could lead to selection of first sample plan identified under the arrow.
counter argument option-2 :- this will lead to overkill of number of samples to be inspected for higher level of AQL. ( for major and others)..

Q2 :- so, is it acceptable to inspect different sample-sizes for different category of defects. ( viz., critical - major - others)

my opinion.,
different number of samples inspected for different category of defects ( or AQL levels) and we stick to the sample size determined by AQL ( subject to the option-1 only where it is extending the sample size)

 

[Miner]

Answer to Q1: The sample size is still 125. The accept on 0, Reject on 1 plan applies to all code letters from A through L.
Answer to Q2: No. The sample size is based on the lot size. AQL is based on the severity of the defect.

 

[v9991]

Answer to Q1: The sample size is still 125. The accept on 0, Reject on 1 plan applies to all code letters from A through L.
Answer to Q2: No. The sample size is based on the lot size. AQL is based on the severity of the defect.

one follow through question, agree that lot size is variable, but when the aql varies among defect categories,
isn't it obvious that the OC curve to gain recommended confidence would require lesser samples. ( as the sample size is from oc curve, for a given lot size, rate of defects, aql, etc)

 

[Toby M]

I've had a few discussions in the past about the interpretation of this table. It has always been my understanding that the whole sampling plan shifts up or down, otherwise what is the point of the arrows? The sampling plan includes the 3 elements (Sample Size, Ac and Re) and if we say that you don't change sample size, that means the table can be fully populated without needing arrows. This is not the case.

Also, initial sample size is based on lot size and if this doesn't change upwards with an arrow (for tighter AQLs), there will never be the situation where "If sample size equals, or exceeds, lot or batch size, do 100% inspection"....

 

[David-D]

I'd like to support Toby's comment; the arrow should force a change to the sample size so it should now be 200, not 125. Sample sizes less than 200 are inadequate to demonstrate an 0.065 AQL. I dont have a copy of Z1.4 readily available but to quote MIL-STD-105E, paragraph 4.9.3:

"... When no sampling plan is available for a given combination of AQL and code letter, the tables direct the user to a different letter. The sample size to be used is given by the new code letter, not the original letter. ..."

MIL-STD-105 (though obsolete and canceled) is available free from the USG at:

https://quicksearch.dla.mil/qaDocDetails.aspx?ident_number=35496

David

 

[David-D]

As an answer to question 2: Yes, you can have two different sample sizes, but you may shift the smaller sample sizes down to match the largest sample size (and get larger associated acc/rej quantities for those plans). To quote later in paragraph 4.9.3:

"... if this procedure leads to different sample sizes for different classes of defects, the code letter corresponding to the largest sample size derived may be used for all classes of defects. ..."

Going to larger sample sizes (farther down the table) results in more discriminating OC curves and reduced supplier and consumer risks (which is why going up is not acceptable).

David