Neural network based CAD models for analysis and design of fin-lines for mm-wave applications

Abstract

Neural network (NN) based CAD models have been developed for analysis and design of fin-lines. In the analysis phase, the network takes the dimensional parameters of the finline structure as its input and gives the characteristic impedance (Z0) and normalized guided wavelength (b/lambdacf) as its output. Another network, trained for design, takes Z0 along with other dimensions as input and produces the normalized gap width between fins (w/b) as output. A multilayer perceptron trained in the back-propagation mode is used to develop the networks. Results for unilateral fin-line is presented in this paper.