摘要
目前,设计和优化微带滤波器的方法是通过电磁仿真软件EM进行的。该方法存在两个不足之处:一是拟优化变量的初始值需要人工猜测,但受制于人工经验,致使猜测值与最优值差别较大,很容易使得优化结果陷入局部最优;二是单次仿真时间长导致优化过程耗时很长~([1])。文中利用并行神经网络并结合矢量拟合建立了替代模型进而优化滤波器。为了验证方法的可行性,文中设计了微带发卡滤波器并进行仿真和实验测试,仿真和实验测试结果基本一致。文中方法克服了目前滤波器设计和优化方法的两个不足之处,为快速设计高性能滤波器提供了方法。
Currently,the method for designing and optimizing microstrip filters is based on electromagnetic simulation EM software.However,this method has two major drawbacks.First,the initial value of the optimization variable needs to be guessed manually,but due to human experience limitations,the guessed value may deviate greatly from the optimal value,leading to optimization results trapped in local optima.Second,the long simulation time per iteration results in a lengthy optimization process[1].This paper proposes a proxy model using a parallel neural network and vector fitting to optimize the filter.To verify the credibility of this method,a microstrip filter is designed and tested via simulation and experiments,which showed consistent results.The proposed method overcomes the two major drawbacks of the current filter design and optimization methods,which provides a fast and efficient approach for designing high-performance filters.
作者
于新华
李志豪
赵洲
冯林平
连丽梅
YU Xinhua;LI Zhihao;ZHAO Zhou;FENG Linping;LIAN Limei(School of Information and Communication,Guilin University of Electronic Technology,Guilin 541000,China;School of Marine Engineering,Guilin University of Electronic Technology,Guilin 541000,China;School of Computer Science,Central China Normal University,Wuhan 430079,China;School of Microelectronics,Xi'an Jiaotong University,Xi′an 710000,China;State Key Laboratory of Millimeter Waves,Nanjing 210096,China)
出处
《微波学报》
CSCD
北大核心
2024年第6期54-59,共6页
Journal of Microwaves
基金
国家自然科学基金(62001170)
广东省基础与应用基础研究基金(区域联合基金-青年项目)(2019A1515110417)
中国博士后科学基金资助项目(2022M712513)
毫米波全国重点实验室基金(K202332)
广西科技规划项目(AD22080042)
广西研究生教育创新项目(YCSW2023492)
广西研究生教育创新项目(YCSW2023477)
桂林电子科技大学研究生教育创新计划项目(2023YCXS028)。
关键词
神经网络
耦合矩阵
微波滤波器
快速优化
neural network
coupling matrix
microwave filter
fast optimization
