Using a multilayer perceptrons for accurate modeling of quasi-fractal patch antennas

A multilayer perceptrons (MLP) artificial neural network (ANN) with one hidden layer and trained through the efficient resilient backpropagation (RPROP) algorithm is used for modeling quasi-fractal patch antennas. The design of the proposed antenna is based on the application of rectangular Koch fractal curve to the edges of a conventional microstrip inset-fed patch antenna. The electromagnetic (EM) characterization of the patch antennas was performed using the Ansoft DesignerTM software that uses the method of moments. A parametric analysis was developed as function of the dielectric substrate thickness and size of the quasi-fractal patch antennas. Considering the region of interest of the design parameters a representative EM-dataset was obtained to develop the MLP network model using the conventional EM-ANN neuromodeling technique. The MLP model is able to estimates the behavior of the antennas with very good accuracy and low computational cost. Good agreement is observed between simulated and measured results.