An artificial neural model of the microstrip lines

Abstract

In this work,the black-box model of a microstrip transmission line is worked out. Thickness of the substrate H width of the conductor strip W the conductor thickness T as the geometric dimensions of the system the normalized frequency fH (GHz mm), and the dielectric constants epsilon(x), epsilon(y) are taken as the input free variables, so the functions of the characteristic empedance Z(0) and the effective dielectric constant epsilon(eff) are resulted at the output of the black-box. A simple neural network with a single hidden layer is employed for the evaluation inside the black-box, which is activated by the sigmoid function and trained by the Levenberg-Marquard algorithm. The approximate analytic solutions with the empirical adjustment of their numerical constants to achieve the desired accuracy is utilized to obtain the training and test data. The commonly used materials such as Alumina, PTFE /Microfiber Glass, Gallium-Arsenide, and RT/Duroid 6006 are applied to the neural network and their characteristic empedance Z(0)s and effective dielectric constant epsilon(eff)s are obtained as the functions of the geometric dimensions H,W/H,T the normalized frequency fH and the dielectric constants epsilon(x), epsilon(y). So this neural network model can be used for the analysis and the synthesis of the microstrip circuits including monolithic microwave integrated circuits.