Research On Automatic Design And Optimization Technology Of Reconfigurable Metasurfaces

Metamaterials are artificial electromagnetic materials that break through the limitations of the electromagnetic properties of natural materials.Metasurfaces are composed of subwavelength-sized metamaterial elements arranged in a 2-dimensional(2D)plane,which can control the amplitude,phase,polarization and other characteristics of electromagnetic waves,thereby owing huge applications in the field of electromagnetic stealth,radar detection and wireless communication systems.In this thesis,considering the application of reconfigurable metasurfaces in the field of electronic countermeasure,the research of automatic design and optimization technology is carried out to simplify the design process,and save the time consumption and the labor costs.The main contents are summarized as follows:1.The concept of machine learning is introduced into the design process of reconfigurable metasurfaces,and a fast design method of reconfigurable metasurfaces based on the support vector machine(SVM)is proposed.Four stages are included: a data collection stage,a data preprocessing stage,a training stage and a reverse design stage.In the data collection stage,a pixel modeling method is proposed to generate random topologies of reconfigurable metasurfaces;In the data preprocessing stage,an undercomplete autoencoder is used to extract features from electromagnetic characteristic data;In the training stage,the SVM model is generated based on the training set to describe relationships between structures and corresponding electromagnetic characteristic data,and the Particle Swarm Optimization(PSO)is used to improve the generalization ability of the SVM model;In the reverse design stage,two design objectives are proposed and inputted into the SVM model,and then the corresponding topologies of reconfigurable metasurfaces will be obtained.After verification,the simulation results match well with the design objectives proposed.2.A model named PNMD(PSO-Nurbs Metasurface Design)is proposed to realize the automatic design of reconfigurable metasurfaces,in which a pixel modeling method based on the Nurbs curve is proposed for the topology design,and the PSO is used to realize the automatic optimization of the topology.Using this model,a broadband wave-absorbing reconfigurable metasurface is designed,fabricated and tested.It has been proved that PNMD can automatically design and optimize wave-absorbing reconfigurable metasurfaces,which has potential applications in the field of electromagnetic stealth.3.A reconfigurable metasurfaces automatic design model named NNMD(NSGA-Nurbs Metasurface Design)is proposed for the realization of multi-dimensional(amplitude,phase,etc.)modulations of electromagnetic waves,in which a pixel modeling method based on the Nurbs curve is used to design the topology of the reconfigurable metasurfaces,and the multi-objective optimization algorithm NSGA-II is introduced to optimize the topology under multiple design objectives.Using NNMD,a reconfigurable metasurface with independent manipulations of amplitude and phase is designed,fabricated and tested.It has been proved that NNMD can design reconfigurable metasurfaces with multiple electromagnetic wave manipulations,which has potential applications in the field of radar detection and wireless communication systems.