考虑质量损失与后验概率的多响应稳健参数设计

Multi-response robust parameter design generally involves a series of questions including robustness measurement of multivariate process, correlation of multiple responses, confliction among multiple goals, prediction performance of process model and the reliability assessment for optimization results, etc. In this paper we present a new approach for multi-response robust parameter design to deal with the above questions. The proposed approach integrates an improved multivariate quality loss function with Bayesian posterior predictive approach based on the framework of Bayesian statistical modeling. The case study shows that the proposed approach not only focuses on the robustness of multivariate process by the multivariate quality loss function, but also evaluates the reliability of an acceptable optimization result for any set of operating conditions via the Bayesian posterior probability method in a united framework of Bayesian model. Therefore, the proposed approach can obtain more robust and reliable value of design parameters than previous research methods.