Surrogate Modeling with Hybrid CNN-RNN and GAN Architectures for Antenna Designs

Accurate modeling of antennas presents significant challenges due to their complex geometries and performance requirements. This paper is devoted to presenting an automated surrogate modeling for sizing the antenna and also predicting the radiation patterns (RPs). First, a hybrid convolutional neural network (CNN)- recurrent neural network (RNN) is employed for optimizing the antenna in terms of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$S_{11}$</tex> and gain performances, and afterward, the generative adversarial network (GAN) is used for modelling the RPs in both E-plane and H-plane, respectively. For both of the networks, the hyperparameters are achieved by using ‘Bayesian optimization’. All the optimization processes are executed automatically by a combination of an electronic design automation (EDA) tool (CST Studio Suite) and a numerical analyzer (MATLAB <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${ }^{\text {TM }}$</tex>). The practical use of the proposed methodology is validated by designing and optimizing an array antenna operating from 12.9 GHz to 13.7 GHz, which results in much faster simulations compared to the only-use shallow neural network and the traditional methods existing in the commercial computational environment.