NEURAL-NETWORK SIMULATION OF A DIELECTRIC RING-RESONATOR ANTENNA

This paper discusses the application of neural network (NN) techniques in the modelling of an electromagnetic (EM) field problem consisting of a dielectric-ring resonator antenna. The NN takes into consideratio n the geometrical and material parameters of the antenna, and the inpu t reflection coefficient values for different frequencies in the bandw idth of interest. The NN is trained off-line using results obtained wi th the finite element method (FEM). The NN is considered as a nonlinea r function approximator and the Levenberg-Marquardt algorithm is used to solve the corresponding least-squares optimization problem. The NN approach is found to provide a tremendous computational speed improvem ent as well as computer memory reduction compared to the traditional n umerical methods of analysis.