Analysis and synthesis of coplanar waveguide-feed using Multilayer Perceptron Feed Forward Neural Networks

Analysis and synthesis of coplanar waveguide feed used in planar antennas have been proposed. Neural network-I has been proposed to analyze the coplanar waveguide feed. It dispenses the use of very long and complex formulae. Neural model-II has been proposed to remove the problems in conventional approach. It synthesizes, even if the dielectric constant is less than 6, unlike the previous approach. It gives the accurate result because thickness of the conductor also is taken into account unlike the conventional approach. The proposed neural networks are trained with three training algorithms like Levenberg-Marquart (LM), Conjugate Gradient of Fletcher-Powell (CGF) and Quasi-Neuron (QN). The LM algorithm is found to be the best algorithms among all based on Mean Squared Error (MSE) values of them.