Efficient Array Pattern Prediction Using a Deep Learning Algorithm

In this article, a deep-learning (DL) algorithm is proposed to efficiently predict the pattern of the antenna array. By establishing the Bayesian-optimized convolutional neural network (CNN) framework, the far-field radiation pattern of the array with arbitrary given array geometry is successfully predicted. Specifically, the CNN is utilized to predict the pattern from the input array geometry and element excitations. The Bayesian optimization algorithm is used to optimize the network architecture to enhance the fitting ability of the CNN. To validate the effectiveness, the proposed algorithm is employed to predict the shaped beam patterns for both linear and planar arrays with various array geometries. Numerical results demonstrate that the proposed algorithm can predict the accurate direction pattern in ms time, and its response speed is more than 6 orders of magnitude faster than that of conventional full-wave simulation.