Abstract:

As for robust design with non-normal responses, dual response surface models are established based on the jointed generalized linear models of the mean and dispersion. Then, in view of the very complicated nonlinear functions of the dual response surface models, the hybrid algorithm based on genetic algorithm and pattern search is used to optimize the dual response surface models and obtain the optimum parameter values of the controllable factors. Finally, a robust parameter design of the resistivity for some test wafer is illustrated by the proposed approach. The results reveal that the proposed approach can effectively reduce the variation of the resistivity of test wafers and improve the robustness of product quality.