基于GLM的双响应曲面法及其稳健设计

Dual response surface methodology based on generalized linear models and its application on robust design

针对非正态响应的稳健设计,首先在均值与散度的联合广义线性模型基础上构建了基于广义线性模型(generalized linear model,GLM)的双响应曲面模型。然后,鉴于所构建的双响应曲面模型为高度复杂的非线性函数,运用遗传算法与模式搜索的混合算法对其进行参数优化,获得可控因子的最佳参数设计值。最后,运用所提出方法对某测试晶片电阻率的参数设计进行了分析。研究结果表明,该方法能有效地减少测试晶片电阻率的质量波动,提高了产品质量的稳健性。

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.