Advanced Space-Mapping-Based Approach for Accelerated Optimization of Metasurface-Combined Circularly Polarized Antenna

Integrating metasurfaces with circularly polarized (CP) antennas holds promise for applications requiring compact form factors and better performance. Traditional optimization methods for such complex structures often involve exhaustive electromagnetic (EM) simulations, which are computationally intensive and time-consuming. In this letter, we propose a space-mapping based neural network (SMNN) and trust region algorithm to expedite the optimization process. To further enhance the speed of optimization, we implement a parallel data generation strategy to generate the coarse-mesh and fine-mesh data simultaneously. Additionally, a trust-region algorithm is employed to refine the model prediction range, ensuring improved convergence at each step. The proposed approach achieves optimal antenna performance in fewer iterations, significantly reducing computational time and enhancing design efficiency