Automatic synthesis of MEMS devices using self-adaptive hybrid metaheuristics

This paper introduces a multi-objective optimization ap- proach for layout synthesis of MEMS components. A case study of layout synthesis of a comb-driven micro-resonator shows that the approach proposed in this paper can lead to design results accommodating two design objectives, i.e. si- multaneous minimization of size and power input of a MEMS device, while investigating optimum geometrical conguration as the main concern.

The major contribution of this paper is the application of self-adaptive memetic computing in MEMS design. An evolutionary multi-objective optimization (EMO) technique, in particular non-dominated sorting genetic algorithm (NSGA-II), has been applied to- gether with a pattern recognition statistical tool, i.e.

Principal Component Analysis (PCA), to nd multiple trade-o solutions in an ecient manner. Following this, a gradient- based local search, i.e. sequential quadratic programming (SQP), is applied to improve and speed up the convergence of the obtained Pareto-optimal front. In order to reduce the number of function evaluations in the local search procedure, the obtained non-dominated solutions are clustered in the objective space and consequently, a post-optimality study is manually performed to nd out some common design principles among those solutions.

Finally, two reasonable design choices have been oered based on manufacturability issues.