Low-Wind-Load Broadband Dual-Polarized Antenna and Array Designs Using Sequential Multiphysics Machine-Learning-Assisted Optimization

Abstract

Low-wind-load designs are increasingly crucial for ultralarge-scale base station arrays operating in the sub-6 GHz band. Here, a sequential multiphysics machine-learning-assisted optimization (MLAO) method is proposed for the rapid design of a compact antenna with aerodynamic favorability and excellent electromagnetic (EM) performance. The total project area is reduced by designing a compact radiator and replacing the traditional metal ground with a topologically innovative metal ground. A $4 \times 4$ array formed by the above antenna showcases a remarkable 73% reduction in the wind load, when each antenna element is independently packaged in a small radome. Moreover, the EM performance is greatly enhanced by optimizing the dipole arm topologies. The antenna prototype is fabricated and measured with a broad impedance bandwidth of 3.2–5.0 GHz, isolation higher than 20 dB, and realized gain of $6.5 \pm 1.2$ dBi. The $4 \times 4$ array shows a front-to-back ratio greater than 20 dB, cross-polarization discrimination greater than 15 dB, and realized gain of $18.6 \pm 1.5$ dBi. These results demonstrate that the proposed antenna is suitable for 5G new radio frequency bands n77/n78/n79.