A GENERAL ALGORITHM FOR DESIGN FOR QUALITY

摘要

The goal of engineering design is to design engineering systems which operate as the designers intend them to. The design process is complicated by the fact that no component value is purely deterministic―any manufacturing process and even nature imposes variability in component values. How do we ensure that these specifications are met with highest possible probability and smallest variability given that the component values have some arbitrary distributions? A well-known approach is Design for Quality, which comprises the complementary phases of parameter design and tolerance design. In this paper we focus on the parameter design phase for systems with implicit multiple responses with double- or single-sided specifications while including probabilistic information on both the control and internal and external noise factors. The salient feature of our work is a computer program which automates the Design for Quality process. The program allows the user to formulate a Design for Quality problem in symbolic form, which is then converted to numeric form using the Matlab Symbolic Toolbox. Lastly, we illustrate the method with the aid of a practical example.