A parallelization method of genetic algorithms for optimal design of microwave filter

The evolutional computing including generic algorithms (GAs) is applied to automatic design optimization in many practical problems. Using the method, we can find an optimum or a quasi optimum solution automatically, and may reduce the computational complexity about several to ten percent compared with the full searching in the parameter space. In general, an evaluation of fitness takes much time to be calculated. So it is necessary to speed up calculations by a parallel processing technique. In this paper, we discuss to employ two kinds of genetic algorithm (GA), parameter-free GA and micro-GA, for an automatic design optimization of dichroic band pass filter operating in the millimeter to terahertz frequency band. And their parallelization methods are proposed and compared with each other on the degree of acceleration. Electromagnetic simulations of dichroic filter are carried out using the FDTD technique.