Research On Lens Antennas Based On Phase Gradient Metasurfaces

With the rapid development of microwave millimeter wave circuit systems and information and communication fields,the phase-gradient metasurface lens antenna has become one of the current hot spots of research in academia and engineering fields due to its ability to control the wavefront phase and amplitude of electromagnetic waves,thus realizing the regulation of electromagnetic wave propagation direction and form.Aiming at the problems of narrow bandwidth of traditional lens antennas and difficulty in realizing anisotropic correlation functions,this paper focuses on broadband metasurface lens antennas and a machine learning design method applied to anisotropic metasurface lens antennas.This thesis focuses on the phase gradient metasurface theory,by comparing the difference before and after adjusting the phase with the traditional compensation phase calculation method,leading to the broadband compensation phase calculation and scheduling method,and verifying that the method reduces the compensation phase difference of the metasurface lens in the broadband range,thus enhancing the gain and bandwidth of the metasurface lens antenna.The broadband compensated phase calculation method is used to design three highperformance broadband antennas and simulate and analyze the antenna-related performance.Among them,the metasurface lens antenna has 6GHz 1dB gain bandwidth;the broadband circularly polarized metasurface lens antenna achieves 12GHz 1dB gain bandwidth with 3dB axial ratio bandwidth;the broadband vortex wave metasurface lens antenna generates firstorder vortex wave in the broadband range of 22-30GHz with an average energy weight of 64.6%and a vortex wave bandwidth of 38%.A machine learning design method is proposed for anisotropic metasurface lens antennas,which consists of linear function-based DNN,phase-classification-based DNN and gray wolf optimization algorithm.The machine learning design method can be used to achieve accurate prediction of the transmission phase of the dielectric base unit in a large frequency range,thus realizing the fast design of anisotropic metasurface lens antennas.The proposed machine learning design method achieves a fast mapping between the dielectric base unit size parameters and the TE and TM transmission phases with more than 95%accuracy in the range of 10 to 25 GHz.The design of a vortex wave multiplexing metasurface lens antenna is completed using the designed machine learning method,which can achieve vortex wave multiplexing of orbital angular momentum order 1 and-1 at 24 GHz,confirming the effectiveness of the designed machine learning method.