C=0 vs. AQL MIL Standard 105 E - Sampling Plans

[carryn1]

I am trying to understand C=0 sampling plan. My customer is requesting me to refer to MIl stand 105E . I kind of confuse, since C=0 mean zero defective . assume i carry out level 2 inspection. Why i need refer to AQL still ? can't I just reject the lot whenver defect detected regardless the lot size & AQL ?

[rexch]

It is insane just give "C=0" as a sample plan. It means no critical issue is allowed. But normal issues must be determined. Cause if normal issue is not allowed in any sample plan, the rework cost is going to be insane. Just IMO.

[Jim Wynne]

In Reply to Parent Post by carryn1

I am trying to understand C=0 sampling plan. My customer is requesting me to refer to MIl stand 105E . I kind of confuse, since C=0 mean zero defective . assume i carry out level 2 inspection. Why i need refer to AQL still ? can't I just reject the lot whenver defect detected regardless the lot size & AQL ?

What your customer undoubtedly wants is for you to use 105E to determine sample quantities, but to disregard the associated accept/reject numbers and reject a lot whenever a nonconforming thing is encountered.

[Tim Folkerts]

One thing you could use the Z1.4 (ie MIL-STD-105) for is to "work backwards" to find a sampling plan that corresponds to what you are doing anyway; or corresponds to what you need.


For example, if sampling 32 units sounds good, then you could look in Table II-A (Single sampling plans for normal inspection) and see that a c=0 plan for 32 units is AQL = 0.4 and the code letter is "G".

Looking back at Table I, you could find that "G" corresponds to General inspection level II for any batch up to 280 items. So now you know that for any batch up to 280 items, the C=0, n = 32 sampling plan is indeed AQL 0.4, General Level II, Single Sampling Plan for normal inspection of lots up to 280 units (or it is Level I for lots up to 500; or Level S-2 for any size lots).

Looking ahead to Table X-G, you can find the OC curve for this plan. For example, this plan has a 95% chance of accepting lots with 0.16% nonconforming, and a 5% chance of accepting lots with 8.94% nonconforming. This table lets you determine the actual alpha and beta risks for the plan you are considering.


If the plan you were considering doesn't match your expectation for alpha and beta risk, then you can go back and redo this all for a different row (ie different sample size) in Table II-A.

[theva]

In Reply to Parent Post by Jim Wynne

What your customer undoubtedly wants is for you to use 105E to determine sample quantities, but to disregard the associated accept/reject numbers and reject a lot whenever a nonconforming thing is encountered.

I agree with Jim, some customer much prefer C=0 and sample size aligned with ANSI ASQC Z1.4. Please take note on the sampling switching rule too.

[Statistical Steven]

In Reply to Parent Post by theva

I agree with Jim, some customer much prefer C=0 and sample size aligned with ANSI ASQC Z1.4. Please take note on the sampling switching rule too.

Why not just use the square root of N+1 with c=0? Or just 10% of the lot size. Using the sample size from the 105E without the accompanying acc/rej is a misapplication and misuse of the standard.

[Geoff Withnell]

For the best layout of C=0 vs Z1.4 Mil105 plans, see "Zero Acceptance Number Sampling Plans" by Nicholas L. Squeglia. There are advantages to statistically correct c=0 plans, such as low sample size. But very few people are qualified to "just find a plan that matches what they are doing" and know what their results actually mean in terms of AOQL.

Geoff Withnell