One dimensional profile reconstruction of dielectric cylinder by using neural network

Abstract

In this paper, Neural Network (NN) is used for profile reconstruction of dielectric constant of thecylinder. The region is assumed to be filled with dielectric medium of which dielectric constantchanges along the radial direction. Radial variation of dielectric constant is represented by Fourierseries. Scattered fields, which are obtained by numerically Method of Moments (MoM) technique, andFourier coefficients are used for training the NN as inputs and outputs respectively. In numericalexamples, Scattered fields are applied trained NN, outputs of NN are compared the exact profiles andgood reconstructions are observed.