Double-stage inspection for screening multi-characteristic items

Final product inspection of a multi-attribute product, such as in the electronic assembly industry, involves expensive facilities. The correlation among attributes may be used for reducing the efforts needed for screening the products; however, engineers without appropriate statistical-economical analysis tools do not take risks, and they designate full inspection of each item. We propose a double stage inspection program for reducing inspection efforts. Assuming that the joint distribution is known, the conditional probability that a product is ‘good’ may be evaluated conditional upon the observation of the product's first-stage inspected quality attributes. Then, an expected cost minimization is implemented in order to decide whether a second inspection stage is required or a classification should be based solely on the first inspection stage. The cost factors include inspection and false classification. The method is illustrated on a real data set from a particular electronic product of Motorola-Arad Ltd.