神经网络加速PSO算法的超材料吸波体设计  

Design of metamaterial absorber based on neural network accelerated PSO algorithm

作　　者：戴书浩 孙俊[1] 彭艺[1] 罗会龙 张莉 DAI Shuhao;SUN Jun;PENG Yi;LUO Huilong;ZHANG Li(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650500

出　　处：《传感器与微系统》2025年第2期90-94,共5页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(52166001)。

摘　　要：在超材料吸波体的设计过程中,研究人员常采用耗时长的全波仿真方法,设计思路主要以耗时长的参数扫描和经验设计为主。为了减少设计耗时,本文提出了一种基于神经网络加速粒子群优化(PSO)算法的快速设计方法。该方法利用神经网络对超材料吸波体的电磁参数进行准确地预测,其预测结果与仿真结果均方误差(MSE)不超过0.0011。在PSO算法对结构参数空间进行搜索的过程中,预测结果被用于算法优化过程中的适应度计算,PSO算法能够根据不同的适应度值自动调节结构参数以到达电磁波宽频带吸收的目的。该方法将设计耗时缩短为全波仿真设计耗时的0.3%。通过该方法设计的超材料吸波体在8.5~17.9 GHz频段内的吸波率大于90%,吸波带宽为9.4 GHz。此外该方法优化过程避免了人工干扰,能够移植到超材料的其他应用设计中。In the design process of metamaterial absorbers,researchers often use time-consuming full wave simulation methods,and the design ideas mainly focus on time-consuming parameter scanning and empirical design.In order to reduce time-consuming of design,a fast design method based on neural network-accelerated particle swarm optimization(PSO)algorithm is proposed.This method uses neural network to accurately predict the electromagnetic parameters of metamaterial absorber,and the mean square error(MSE)between the prediction results and the simulation results is not more than 0.0011.In the process of searching the structure parameter space by the PSO algorithm,the prediction results are used to calculate the fitness in the algorithm optimization process.The PSO algorithm can automatically adjust the structure parameters according to different fitness values to achieve the purpose of electromagnetic wave broadband absorption.This method reduces the design time to 0.3%of the full wave simulation design time-consuming.The absorption rate of the metamaterial absorber designed by this method is more than 90%in the frequency band of 8.5~17.9GHz,and the absorption bandwidth is 9.4 GHz.In addition,the optimization process of this method avoids artificial interference and can be transplanted to other application designs of metamaterial.

关 键 词：超材料吸波体 神经网络 粒子群优化算法 

分 类 号：TB391[一般工业技术—材料科学与工程] TB34