Sequential Design Method for Geometric Optimization of an Electrothermal Microactuator Based on Dynamic Kriging Models

This paper proposes a sequential optimization methodology for designing an electrothermal polysilicon actuator in the presence of a fabrication tolerance. In the proposed method, a deterministic optimum is first sought from an initial design, and then a reliability-based robust design is obtained launching at the deterministic point. This serial design strategy can enhance numerical efficiency through minimizing the use of computationally expensive reliability-based design optimization. To effectively perform the robust design of very complex multiphysics problems, elaborate surrogate models based on the local window concept are exploited comprehensively. The proposed method is applied to an electrothermal polysilicon actuator with seven design random variables, and then three different nominal designs are examined in terms of maximum deflection, consumed power, and confidence level.