Techniques in Global Optimal Design for MEMS & Their Applications

Abstract

This chapter presents an overview on techniques in global optimal design for MEMS & their applications. We address single-objective and multi-objective functions optimizations using a Simulated Annealing (SA) method, which has been used by us to handle some constraints as well. This optimization method is essentially an iterative random search procedure with adaptive moves along the coordinate directions. It permits downhill or uphill moves under the control of a probabilistic criterion using Metropolis criterion. Thus, it tends to avoid the first local maxima or minima encountered. The SA method is usually used for a single objective optimization. However, the use of the SA for multi-objective functions is also described in this chapter. Besides, the evolutionary algorithms, which are widely used for multi-objective problems, are briefly presented. For design applications, our device example is a multimorph. The SA exhibits a promising superiority over other algorithms using the gradient methods, have a greater search flexibility and efficiency in exploring the neighborhood of the solutions on the constraints boundaries to find the global optimum solutions. In fact, the SA has potential solutions in the multi-objective optimization field.