A system for quality improvement via computer experiments

Abstract

Many products are now routinely designed with the aid of computer models. Given the inputs-designable engineering parameters and parameters representing manufacturing-process conditions-the model generates the product's quality characteristics. The quality improvement problem is to choose the designable engineering parameters such that the quality characteristics are uniformly good in the presence of variability in processing conditions. This article summarizes recent work to develop an efficient, systematic approach to quality improvement via such computer models. Using relatively few runs of the computationally expensive computer model, our approach builds approximating functions to be used during product-design optimization. We contrast several approximation strategies. We also discuss how to choose a loss to optimize when there are multiple, conflicting quality characteristics. Applications in the design of electronic circuits are given.

Keywords:

• circuit design
• computer code
• off-line quality control
• parameter design
• response surface
• stochastic process
• Taguchi method