Rapid and Accurate Prediction of Radiation Patterns of Reconfigurable Reflectarray Using Deep Learning

This paper presents a rapid and accurate radiation pattern prediction model for large-aperture reflectarray based on Convolutional Neural Network (CNN). The model takes into account complex effects in the array such as mutual couplings among antenna elements and truncation effect for edge elements, which are typically challenging for traditional methods in characterizing electromagnetic property of large-size reflectarrays. Compared to full-wave simulations, the proposed method significantly reduces computational time while maintaining high prediction accuracy. Numerical experimental results indicate that the CNN model provides reliable and robust predictions, making it an effective tool for array antenna design and real-time optimization.