Sampling is used for acceptance tests of products. It can be used during the production process, at the end of it, before sending to customer, and by customer while receiving ordered products. It should be noted, that quality inspection can allow reduce number of defective products send to customer, but it increases quality costs. Quality inspection doesn't create a value for customer (customer won't pay more if you perform the check 1 or 5 times), therefore in modern enterprise it is better to assure quality self-inspection.
Sampling is based on two factors:
- frequency of inspection,
- number of parts checked in each batch.
Those two factors should be determined in agreement or standard.
Methods of sampling
Sampling can be based on agreements or internal/external standards that define acceptable quality level (AQL). A list of standards and some calculators can be found on SQC Online website. The main types of sampling are:
- single-stage sampling plan,
- double-stage sampling plan,
- accept on zero,
- continuous sampling.
Single-stage sampling plan
In single-stage sampling plan quality inspector chooses certain number of products to check. If the number of defective products is lower than AQL the batch passes the check. If it is above AQL - it's rejected.
Double-stage sampling plan
In double-stage sampling plan a smaller number of products is chosen. If it is below AQL - whole batch passes the check. If it is above - another samples are taken. This approach should be used when searching for conformity, e.g. by auditors while auditing quality management systems.
Accept on zero
In production practice we know, that there is no 100% quality. However sometimes the product quality is so important, that we can't afford even one defective piece. The AQL is set to 0 defective parts. In that case quality inspector will reject every batch that has even one defective part. It may become very expensive, so be careful.
In case of continuous flow it's hard to talk about batch and sample size. Therefore we can make random sampling during the production. The most effective way of doing this is to make more intensive screening while production starts. If everything is OK, inspector can switch to random sampling. If he/she finds defective part, he/she should switch to intensive screening again.
Evaluation of the sample
Quality inspector should be instructed on how to distinct defective parts. There are two methods of distinction:
- discrete evaluation,
- evaluation using variables.
It works or it doesn't. It's as simple as that. Well, no. In practice it is hard to use discrete method, which uses only one criterion and two bins for non-defective and defective parts. The problem is, that quality inspector is human, and he/she also can make mistakes. Discrete method tends to increase number of false-positives or false-negatives.
Evaluation using variables
We can set tolerance limits for certain variables related to important characteristics of the product, e.g. length, weight. If product is within tolerance for each variable it passes, otherwise - it is rejected.
What to do with defective parts?
Quality policy can allow several ways of proceeding with defective parts:
- defective part can be repaired and be fully functional,
- classification as lower quality product and selling it cheaper to the customer that accepts lower quality and understands effects of using defective product (this can be prohibited by law in some cases),
- scrapping or recycling.
- Aparisi F (1997) Sampling plans for the multivariate T2 control chart, Quality Engineering, Volume 10, Issue 1
- Bousetta A, Cross AJ (2005) Adaptive sampling methodology for in-line defect inspection, Advanced Semiconductor Manufacturing Conference and Workshop
Author: Slawomir Wawak