Design Of Product Specifications For Multi-Characteristic I

Abstract

Inspecting a product for each characteristic independently ignores interactions among characteristics in deciding the disposition of an item. Two multicharacteristic screening models are proposed with different information processing requirements. In both, screening specifications are jointly determined by considering all the economic and stochastic factors linked with the characteristics of interest. Model 1 uses separate screening specifications, and Model 2 uses a joint screening rule based on an aggregation of characteristics. A numerical study is used to compare the cost performance and other plan characteristics of the independently determined single characteristic models and the 2 multicharacteristic models. It is noted that the acceptance region of Model 1 actually is not much different from that of Model 2. Thus, the cost difference is not substantial, especially when the probability density in the nonoverlapping area is small. This suggests that a correct use of the loss function is very important to a multicharacteristic problem.