Notched planar square monopole antenna design by using artificial neural network

In this paper, artificial neural network based on multilayer perceptron (MLP) model is applied to the metal-plate notches are carved into the square monopole antenna. Backpropagation algorithms that are Levenberg-Marquardt (LM), scaled conjugate gradient (SCG), Bayesian regularization (BR) used to train the neural network. The outputs of the algorithms were transferred to the electro-magnetic (EM) analysis program and the simulation results are compared. The accuracy is calculated by correlation between EM simulations and neural network output. In order to train the neural network, dataset is created by parametric analysis of the antenna design parameters in computer-aided design (CAD). Although the dataset is generated to capture critical design parameters of frequency spectrum of L-band and S-band, the out of band performance has a good agreement with the EM simulation result. The results presented indicate that the neural network can predict antenna design parameters by using S-parameters and effects of the design parameters of the antenna on the S-parameters.