I know this is an old question, but I wanted to add a few comments.
My question is how do I generate a correct sampling plan? Customer goal is 0.08dphu
I can think of at least three ways to interpet this goal, and this interpretation has a big affect on the sampling you do:
1) The customer will accept no more than 0.08 dphu and will reject most lots that are any worse than 0.08 dphu.
2) The customer will accept an average of 0.08 dphu and will lots as long as the average doesn't get above about 0.08 dphu.
3) The customer will accept 0.08 dphu and will accept most lots that are no worse than 0.08 dphu.
The sampling plan used by a customer would vary greatly depending on which situations was the case. The required quality required from the producer also varies widely.
In case 1, you might need to aim for 0.04 dphu in order to avoid getting too many rejected lots back from the customer.
In case 2, you would aim for 0.08 (or a little less) dphu in order to avoid getting too many rejected lots back from the customer.
In case 3, you might get away with 0.12 dphu and still have most of the lots accepted.
The ASQ Z1.4 (MIL-STD-105E) tables are generally set up for situation 3. I.e, when you sample, you will accept most (say 90% to 99%) of lots with the specified AQL.
As a producer doing sampling, you need to decide your goal. Do you want to
ensure that nearly all lots
meet or accede the customer requirement? Or do you simple want to ensure that the customer's sampling plan won't reject too many shipments?
Do I need to know all the 4 information (AQL, alpha, LTPD, beta) to determine the sampling plan? I came across some example which only requires AQL and alpha. Pls. advise.
Yes and no. With the ASQ Z1.4 tables, you only need to specify AQL, because the four values are assumed to have reasonable values if you follow the plan. This upside is that you don't need to worry about the details, because the authors of the plan did that for you.
The downside is that you can't easily know what the actual numbers are. For example, you might think that the tables would be set up so that the alpha values would be something like tightened = alpha = 0.1, normal = 0.05, reduced = 0.01, but that isn't the case. There is a considerable variation in alpha depending on the AQL and the lot size. There is even less uniformity in the LTPD & beta. (Personally, I'd like to see these tables scrapped and replaced by a "standard" piece of software. You would input the AQL, LTPD, alpha, & beta, and it would output the sampling plan.
If you are motivated, you can look at the tables & OC curves at the back of the book to determine what the specific values are for each plan.
If you are really motivated (or buy the right software) you can specify all four values and generate your own plan to exactly meet your specs.
Tim F