Estimation and Design of U-slot Physical Patch Parameters with Artificial Neural Networks

In this study, physical U slot parameters of rectangular microstrip patch antennas as vertical and horizontal slot lengths and widths with the patch lengths and widths are determined by the help of Artificial Neural Networks. The aim of the study is calculation of physical U slot patch parameters without any mathematical expressions or long and complex numeric calculations with a neural network model. Experimental results in the literature are used as the training data for the network by using Gradient Descent with Adaptive Learning Rate Back Propagation learning algorithm. The resonant frequency, dielectric constant of the substrate and the dielectric substrate thickness values are the inputs of the neural network and the patch length, patch width, the lengths and widths of the vertical and horizontal slots are the network outputs. The test output data of the network are used for simulations and the results are confirmed by these simulations. S-11 responses, simulation frequency, impedance bandwidth, directivity, gain and radiation efficiency values of the antennas are investigated by HFSS. Simulation results are compatible with test outputs. The high training success of the network and R-2 values very close to 1 show that physical patch parameters of U-slot rectangular microstrip antennas can be calculated with this Artificial Neural Network model with high accuracy.