Circuit based artificial neural network model of microstrip line discontinuities

In this work, it is aimed to form the Equivalent Circuit Artificial Neural Network (EC-ANN)Black Box models of microstrip transmission lines.First, the most commonly known five microstrip line discontinuity types were investigated.These discontinuity types are open circuit, gap, step, T-Junction and Bend discontinuitieswhich were modeled with equivalent circuit parameters. Equivalent circuit parameter valuesthat were calculated by using analitical formulations were compared with the equivalentcircuit parameter values that were obtained by the education of the artificial neural network.By using the learning capability of artificial neural networks, it is aimed to obtain the EC-ANN Black Box Model whose inputs are the electrical properties of the substrate material, thedimensions and types of the discontinuities and the outputs are equivalent circuit parametersand microstrip line parameters. Two anisotropic materials PTFE/ microfiber glass, RT/Duroid 6006 and two isotropic materials Alumina, Gallium arsenide are used as substratematerials which are widely used in microwave technology.As a result, artificial neural network is educated and tested by the outputs of automatic dataprogram which is formed by the input and output formulations. By this way, EC-ANN BlackBox models of open circuit, gap, step, T-Junction and bend discontinuities are formed and itsperformance and the discontinuity effects are investigated on some graphics.Keywords: Microstrip Line, Discontinuity, Artificial Neural Network.JURY:1. Prof. Dr. Filiz GÜNEŞ (Supervisor) Date: 28.09.20062. Doç. Dr. Sedef KENT Page: 1053. Y. Doç. Dr. Hamid TORPİ