Un metodo para el diseño de circuitos de microondas apantallados basado en redes neuronales y algoritmos genéticos

Abstract

The present work develops a method to design shielded printed microwave filters based on Genetic Algorithms and Neural Networks. Genetic Algorithms techniques have been applied in the microwave optimization problems, but they are usually very time consuming. Thanks to the Neural Networks, the Genetic Algorithms operation time in the filter design task is reduced. The Neural Networks approximates the Green’s functions employed in the Integral Equation approach. This neural method allows fast analysis of any printed circuit inside the shielded structure. In order to improve the Genetic Algorithm performance a new specific fitness function suitable to microwave problems has also been developed. Results show the ability of the new method in the design of practical printed shielded microwave filters.