Neural Network Modeling for the Solution of the Inverse Loop Antenna Radiation Problem

Abstract

Soft computing techniques are used, in this paper, to model and solve the inverse problem of a thin, circular, loop antenna that radiates in free space. The electromagnetic field intensity serves as the input to the inverse model, whereas the antenna radius is the output. Three different architectures, based on artificial neural networks (ANNs), are implemented and various training algorithms are tested in order to obtain the optimum performance. The effect of the size of the training data set and the number of the observers on the accuracy of the results are investigated. Specific information for the selection of the appropriate ANN architecture is provided, depending on the constraints imposed by various parameters of the problem. Extensive numerical tests indicate that the results predicted by the proposed models are in excellent agreement with the theoretical data obtained from the existing analytical solutions of the forward problem.