Optimal design technique for waveguide device

Abstract

A modified genetic algorithm, which we call the genetic algorithm coupled with sensitivity analysis (GACS), is proposed in comparison with the genetic algorithm (GA) and the optimization process using the design sensitivity analysis (DSA). The GACS reduces computing time by appending the DSA optimization process to the genetic algorithm. As an example, a waveguide T-junction is designed for the purpose of minimizing the reflected power. As an analysis tool the two dimensional vector finite element method (VFEM), which does not generate spurious solutions having a fatal effect on the optimization process, is used. The reflected power of the T-junction designed by the GACS shows a characteristic of less than -20 dB near 2.2 GHz and the T-junction designed by GA also shows similar characteristics. However the one designed by DSA optimization shows a considerably worse performance than those of the T-junction designed by the GACS or the GA process. The validity of the proposed algorithm is verified through experiment.