基于神经网络的微波滤波器设计综述

摘要：

随着微波器件结构复杂度和性能要求的提高,建模和设计过程的时间成本也逐渐增大.在设计过程中引入优化算法可以有效地提高微波器件的设计效率.对神经网络在微波滤波器设计领域的方法进行了系统的综述,提出了基于训练和精调神经网络的LC电路参数提取方法、基于数据自生成神经网络的结构参数估计方法和带宽可重构的滤波器设计方法.实验结果表明,神经网络可以利用滤波器的S参数信息,在较短时间内预测出满足设计要求的LC电路参数/器件结构参数,有效减少设计成本,提高设计效率.

With the increase of the structure complexity and the performance requirements of the microwave devices,the time cost of the modeling and the design process increases gradually.The design efficiency of microwave devices can be improved by introducing optimization algorithm in the design process.In this paper,the methods of the neural network in the field of microwave filter design are systematically reviewed,and the LC circuit parameter extraction method based on training and fine-tuning neural network is proposed.A method for estimating the structural parameters of the bandpass filters and designing the bandpass filters with bandwidth reconfigurable is proposed based on the self-generated neural networks.The experimental results show that the neural network can use the S parameter information of the filter to predict the LC circuit parameters/structure parameters that meet the design requirements in a relatively short time,reduce the design cost effectively and improve the design efficiency.

作者：

张安学 杜浩 戴新月 杨倩 郭诚 廖学文

Zhang Anxue;Du Hao;Dai Xinyue;Yang Qian;Guo Cheng;Liao Xuewen(Department of Electronic and Information Engineering,Xi'an Jiaotong University,Xi'an 710049,China)

机构地区：

西安交通大学电子与信息学部

出处：

《betway官方app 学报:自然科学版》 CAS 北大核心 2023年第6期1-14,F0002,共15页

Journal of Henan Normal University(Natural Science Edition)

基金：

国家自然科学基金(61801367) 陕西省深空探测智能信息技术重点实验室资助(2021SYS-04).

关键词：

神经网络 精调 结构参数估计 带通滤波器 可重构

neural network fine-tuning structural parameter estimation bandpass filter reconfigurable

分类号：

TP391 [自动化与计算机技术—计算机应用技术]