DIMENSIONING OF FIXED FREQUENCY PATCH ANTENNAS BASED ON NEURAL NETWORKS

Rectangular patch antenna, method of moments, resonance, neural networks, networks based on electromagnetic knowledge.

Authors

Vol. 10 No. 04 (2022)
Engineering and Computer Science
April 30, 2022

Downloads

During the last decades the race for innovation in communication systems has not ceased to evolve, which has led to important studies in the field of antennas, which today have very different forms depending on the applications such as: mobile telecommunications, television, radio, satellites, communicating systems, radar, remote sensing, radio astronomy [1]. Among the most used antenna families are the microstrip antennas (also called printed antennas or patch antennas) [2]. These antennas are characterized by: low manufacturing cost, mass production possible, linear and circular polarization, feed and matching networks manufactured simultaneously with the antenna.

The limitations of conventional neural modeling have been overcome by the introduction of neural networks based on electromagnetic knowledge. A neural network using effective parameters in conjunction with the GALERKIN function has been developed for modeling the resonant frequency of a rectangular antenna printed on a substrate [3]. Thus, we set up a rectangular PATCH antenna model on HFSS for a resonant frequency close to 6GHZ and extract the parameters S11, Z11, and VSWR, and also check the training errors on MATLAB by generating the error types on a 4-input model (Thickness, Length, Width and substrate type) and one (1) output (Resonant Frequency) that will allow to extract the final dimensions of the antennas thanks to this neural model.