RADIOENGINEERING, cilt.19, sa.4, ss.645-652, 2010 (SCI-Expanded)
Artificial Neural Networks (ANNs) gained importance on the RF microwave (MW) design area and behavioral modeling of MW components in the past few decades. This paper presents a cost effective neural network (NN) approach to overcome design, modeling and optimization problems of an 180 degrees ring hybrid coupler operating in C-Band. The proposed NN model is trained by data sets obtained from electromagnetic (EM) simulators and neural test results are compared with simulator findings to determine the network accuracy. Moreover, necessary trade-offs are applied to improve the networks' performance. Finally correlation factors, which are defined as comparison criteria between EM-simulator and proposed neural models, are calculated for each trade-off case.