ARTIFICIAL NEURAL NETWORK TO PREDICT RESONANT FREQUENCY OF DIPOLE FREQUENCY SELECTIVE SURFACE

摘要

This paper deals with a frequency selective structure comprising of two dimensional arrays of dipole patches. The characteristics of Frequency Selective Surface can be calculated by Finite Element Method, Method of Moment etc. But these methods are very complicated and time consuming. Efforts have been given to develop a method which may relate the resonant frequency with the periodicities and dielectric constant of an FSS having dipole patches. Firstly resonant frequencies for each of the combinations have been calculated using Method of Moment after selecting proper basis function. Some of the resonant frequencies obtained have also been compared with the measured results for validation. These results have been used to train an artificial neural network. From the trained network resonant frequencies for given periodicities and dielectric constant may be readily available. It is observed that the results obtained by the trained artificial neural network are very fast and accurate.