An application of PSO algorithm for multi-criteria geometry optimization of printed low pass filter based on periodic structures

In this paper we discussed the results of the multi-criteria optimization scheme as well as numerical results of the periodic conductive structures with selected geometry and electrical properties. Thin printed structures embedded on a flexible dielectric substrate may be applied as a simple, cheap, passive low-pass filters with an adjustable cutoff frequency in the low (up to 1 MHz) radio frequency range. The analysis of the electromagnetic phenomena in the presented structures was realized on the basis of the three-dimensional numerical model of three proposed geometries of periodic elements. The finite element method (FEM) was used to obtain the solution of the electromagnetic harmonic field. The equivalent, lumped electrical parameters of the printed cells obtained in such manner determine the shape of the amplitude transmission characteristic of the low-pass filter. The nonlinear influence of a printed cell geometry on the equivalent parameters of their electric model, makes it difficult to find the desired optimal solution. Therefore, the optimization problem of optimal cell geometry estimation with regard to approximation of determined amplitude transmission characteristic with adjusted cutoff frequency, was obtained by the particle swarm optimization (PSO) algorithm. The dynamically suitable inertia factor was also introduced to the algorithm to improve a convergence to the global extremity of the multimodal objective function. The numerical results as well as PSO simulation results were characterized in terms of approximation accuracy of the predefined amplitude characteristics in pass band, stop band and cutoff frequency. Three geometries of varying degrees of complexity were considered and their use in signal processing systems was evaluated.

Author: Adam Steckiewicz
Conference: Title