Numerical design technique for waveguide T-junction in H-plane

A microwave device is usually designed by repeated trial and error until its performance satisfies its design goal. Numerical analysis has been applied to predict the performance of the microwave device. Numerical analysis gives the accurate performance of the device even though the shape of the device is complicated. However, the numerical analysis gives merely the performance not the designed shape of the device. Thus, designing using numerical analysis is usually performed as follows: many models are analyzed by changing the shapes. Among them, the best one is adopted. This method is not efficient when the design variables are multiple. Therefore, it is desirable to develop a more systematic design algorithm. The optimization algorithm proposed in the paper is the genetic algorithm (GA) coupled with the design sensitivity analysis, which the authors call GACS. The proposed algorithm can overcome the demerits of GA and design sensitivity analysis. The drawback is that it needs a huge computing time, especially when it is applied with finite element analysis. The demerit of sensitivity analysis is that it is not suitable when the problem has multiple local minima. The usefulness of the proposed optimization algorithm, GACS, is verified through the simulations and experiments.