Development of Information-Driven Robust Design:A Bayesian View

摘要

The Bayesian approach is a proven method for formally incorporating scientific knowledge, expertise, and informedrnjudgment into statistical analysis; however, there is room for improvement. The conventional approach considers arnrandom sample to determine the mean and variance of the current system performance, yet there are other ways torndetermine this type of information. One such method is robust design, which attempts to determine the optimumrnoperating conditions for a system. The techniques used in robust design that could improve the estimation of arnposterior distribution include design of experiments, consideration of uncontrollable (noise) factors, and experimentsrnin which restrictions and constraints are imposed. In this paper, we propose a robust design-based Bayesian model tornestimate the mean and variance of a posterior distribution more precisely. The proposed model is illustrated throughrnan example and is compared to the traditional Bayesian approach.