Artificial neural networks for planar waveguide junctions

In recent years, artificial neural networks (ANNs) have been introduced as fast and accurate tools for modeling simulating and optimizing a great variety of microwave structures. In this paper we propose a multilayer perceptron neural network (MLPNN) for modelling planar waveguide junctions involving waveguides with arbitrary shapes and with variable geometrical features. The artificial neural network has been used to represent the modal spectrum of the arbitrary waveguides, and the coupling-integrals between such waveguides and standard rectangular waveguides, as a function of the variable geometrical feature(s). The training data set has been obtained,using an accurate but slow electromagnetic simulation algorithm. The result obtained is shown to be a very accurate and extremely, efficient representation of the waveguide junction.