A Comprehensive Mobile Phone Antenna Performance Evaluation Model Based on Deep Learning

Machine learning methods such as deep learning are beneficial to solving complex multidimensional modeling problems, so have great potential in electromagnetic(EM) structure research. In this work, we introduce a general method for predicting antenna performance using deep neural networks. Specifically, we use near-field data from artificial neural network (ANN) to estimate the far-field radiation and S 11 of antennas such as rectangular patch antennas in real time. This method is versatile because it can learn from a range of design parameters for various substrate materials and operating frequencies and can accurately predict all electromagnetic characteristics of the desired antenna (e.g., near and far field, S-parameters). Compared to numerical methods, we achieve acceptable antenna performance estimates.