A new optimization model，integrating quality loss function and posterior probability approach in the framework of Bayesian statistical modeling，is proposed to solve the problem of multi-response robust parameter design． The proposed method not only assesses the expected probability of each response which falls within its respective specification limit ( i． e． ，the reliability of optimization results) using posterior probability approach，but also measures the robustness of multivariate process with quality loss function． In addition，this paper discusses，by illustrative examples， the relationship between joint posterior probability and marginal posterior probability， the influence of different expected probability on the optimization results of the proposed approach，and how to obtain the optimum balance between quality loss and posterior probability． The results show that the proposed method can simultaneously take into consideration the robustness of multivariate process and the reliability of optimization results，and provide a relatively satisfactory optimization result from several respects ( e． g． ，robustness of multivariate process，the reliability of optimization results) to achieve robust parameter design with multiple responses．