Machine Learning Assisted Optimization of an Irregular Slot Antenna

In this paper, to address the problem of long simulation time during antenna design, it is proposed to use Multilayer Perceptron (MLP) to learn the complex nonlinear functional relationship between antenna structural parameters and performance parameters. MLP combined with Genetic Algorithm (GA) for antenna optimization can quickly reach the optimization goal. The impedance bandwidth obtained by manual optimization of the irregular slot antenna is 45.9% (2.28-3.61 GHz). After the optimization, the impedance bandwidth of the antenna reaches 49.24% (2.22-3.67 GHz), the experimental results show that the set optimization goal is well accomplished. The comparison between the predicted results and the HFSS simulation results basically matches, proving the validity and reliability of the method.