Research on Broadband W-Band Reflectarray Element Based on Machine Learning Optimization Method

Reflectarray antennas (RAs) have become a focal point of high-gain antenna research, renowned for their attributes including cost-effectiveness, lightweight, large communication capacity, and no complex feed network in millimeter-wave frequency. The reflective electromagnetic characteristics of element form the basis of broadband RAs. The conventional design methodology for element relies on brute-force parameter scanning, necessitating substantial computational resources and time for comprehensive full-wave simulations. Additionally, it heavily relies on electromagnetic theory and empirical knowledge, imposing significant constraints on the broader application of RA. To address these challenges, this study introduces a machine learning-based optimization method. The optimized element, after optimization, consistently maintains a reflective amplitude of -0.06 dB or higher. It achieves a linear reflective phase curve with a phase shift range exceeding 360 degrees. This result validates the effectiveness of the proposed method. Moreover, these results underscore the potential of the proposed method to establish an automated design and optimization framework for RAs with enhanced performance.