Neural networks modeling and parameterization applied to coplanar waveguide components

Abstract

The present work describes the use of neural networks (NN) for multi-parametric design and parameterization of coplanar waveguide (CPW) components. This technique allows one to reduce the CPU time required for intensive electromagnetic (EM) simulations in a classical optimization procedure. Neural networks are used for modeling the high complex relationship between the physical parameters of a CPW circuit and its various frequency responses. In this paper, the multi-layer perceptron neural network is used with one or two hidden layers due to its great capability for modeling complex structure behavior using data obtained from electromagnetic (EM) simulations. The validity of the neural modeling is demonstrated by studying a CPW T-junction. Our proposed technique is applied on the modeling of a CPW low pass filter and a slot antenna fed by a CPW line. © 2000 John Wiley & Sons, Inc. Int J RF and Microwave CAE 10: 296–307, 2000.