Integrated parameter and tolerance design based on a multivariate Gaussian process model

Abstract

Parameter design and tolerance design are commonly used methods in continuous quality improvement. However, it is difficult to use univariate Gaussian process models to perform the parameter and tolerance design for multi-response physics experiments. In this article, an integrated parameter and tolerance design (IPTD) method is proposed based on a multivariate Gaussian process (MGP) model. MGP models are used to estimate the relationship models for physical experiments with input uncertainty, and a covariance structure with the input noise distribution is constructed. Then an optimization scheme integrating the quality loss function and tolerance cost function is built to find the compromise solutions. The optimal solution is selected from the Pareto set by using grey relation analysis (GRA) technique. Two examples are used to illustrate the effectiveness of the proposed approach. The results show that the proposed approach can obtain optimal settings that minimize quality loss and tolerance cost simultaneously.