AQL - How to count samples with defects for each defect class

[Achilleas]

Hi,

We are reviewing our incoming acceptance activities and we would like clarification on how to count the count samples with defects.
Should we count all samples with defects that belong to the same defect class, even if these defects are related to different specifications?

An example:

We use the following defect classification:
- Critical defect
- Major defect
- Minor defect

We have the following inspection specifications that are related to the major defect:
1. Visual inspection of material deformation
2. Visual inspection of material cracks
3. Dimensional inspection of dimension A
4. Dimensional inspection of dimension B

Let's assume that based on the selected AQL for the major defect, the acceptance number of defects is 2.

During an inspection of a specific lot:
one sample fails in inspection 1,
another sample in inspection 3,
another sample in inspection 4,
and all other samples passed.

In our example, should we say that 3 samples have at least 1 major defect, so the lot should be rejected as the limit is 2?
Or should we assess each individual inspection specification within the major class independently, e.g. no individual inspection test (1, 2, 3, or 4) had more than 2 defects, so the lot should be accepted?

Could you please help me to clarify which approach is suggested according to ISO2859 and followed by the industry?


Thanks a lot.

 

[Jim Wynne]

Unless you're going to sample the lot separately for each defect type, you reject the lot based on your example results. Of course, the assumption is that the samples have been actually taken randomly--each member of the entire lot has an equal chance of being selected--otherwise you can't rely on the results as being representative.

 

[Achilleas]

Unless you're going to sample the lot separately for each defect type, you reject the lot based on your example results. Of course, the assumption is that the samples have been actually taken randomly--each member of the entire lot has an equal chance of being selected--otherwise you can't rely on the results as being representative.

Thank you Jim. Just for clarity, you mean that we should add the samples found with at least one major defect, which gives us 3 defective samples and reject the lot, right?

 

[Steve Prevette]

I've written sampling plans before which take into account different severities of defects. For example, a criteria for rejection might be:

1. Any significant defect (reject on one)
2. Reject on two major defects.
3. Reject on one major defect and two minors.
4. Reject on four minor defects.

And of course, one needs to write the recovery plan as to how to recover from a rejected sample, and will that vary with which of the four criteria were tripped? Perhaps on significant defect is important enough to go to 100% sampling, while the others may only increase the sample size.

 

[Miner]

Aren't sampling plans based on the number of defective units, not the number of defects?

 

[Steve Prevette]

Not necessarily. If it is possible to have multiple defects per unit, I am overlooking considerable information if I only count the number of units with defects, regardless of severity or quantity. So I may desire to account for severity and counts per unit in the sampling plan.

Or in the example I gave, a "defective unit" is defined as 1 Critical defect, two major defects, etc. in the case where you want to only count the number of defective units.

https://saaq.gouv.qc.ca/fileadmin/documents/publications/table-new-minor-major-defects.pdf
Different AQL sampling size for critical, major and minor defects | AQF

 

[Chris64]

Hi,

I have the same question as the user Achilleas but I am not sure if I understand your answer correctly.

Would reject or accept the lot knowing that:

Acceptance Criteria: 2,3 (Acc,Rej) (=Given as result of LOT size and AQL)
Defect Class: Minor
Number of Characteristics to check: 3
Number of defect samples: 3

Sample 1 fails on Characteristic 1
Sample 2 fails on Characteristic 1
Sample 3 fails on Characteristic 2

Accordign to the Acceptance Creiteria (2,3) there are 2 possible options:

- Yes, I would accept the lot because the maximum number of defect samples per characteristic for this defect class is 2
- No, I would reject the lot because the total number of defect samples for this defect class is 3

Which option would you choose? Would you accept or reject the lot?

I would appreciate an answer with explanation.

Thank you!
Chris

 

[Steve Prevette]

My response is - you need to plan ahead and write in the sampling plan what the failure criteria are, and if there are multiple things being monitored in the same sampling, then you've got to write ahead - it takes 1 Apple and 3 Oranges and 2 Cherries to count the unit as defective.

 

[Chris64]

So you're saying that I can define my own criteria without using the acceptance criteria of the ISO2589?
In this case are for you both options of my example acceptable?

Assuming I wrote in the sampling plan that it takes 1 characteristic(1 Apple) to count the unit as defective. Based on the above results of inspection, how would you decide for the lot? Is it accepted or rejected?

 

[Steve Prevette]

Here's the thing. ISO 2589, and the predecessor MIL STD 105 (which I was trained in) gives you the basic sample sizes and accept/ reject numbers. And there are several structures out there, c=0 plans, double sampling plans. They give you the basics. But in your sampling plan, you must still define what is a "reject" and you still must define what do you do in response to a reject, and in some cases, how do you relax the sample sizes if you have gone a long time with no rejects., But that has to be determined AHEAD OF TIME. Not - gee I have two minor defects and a major on this lot- what do I do with that?