基于卷积神经网络的微带滤波器的参数预测  

Parameter Prediction of Microstrip Filter Based on Convolutional Neural Network

作　　者：揭智航 杨维明[1] 李进[1] JIE Zhihang;YANG Weiming;LI Jin(College of Artificial Intelligence,Hubei University,Wuhan Hubei 430062,China)

机构地区：[1]湖北大学人工智能学院,湖北武汉430062

出　　处：《电子器件》2024年第2期490-495,共6页Chinese Journal of Electron Devices

基　　金：2018年国家“新工科”研究与实践项目(教高厅函[2018]17号)。

摘　　要：提出了一种基于卷积神经网络的微带带通滤波器的参数预测方法,针对一个三频带带通滤波器进行预测。首先利用HFSS电磁仿真软件训练数据得到S参数的数据集作为真实值数据集,将微带滤波器的S参数作为输入,物理结构参数作为输出,通过卷积神经网络进行训练,最后将目标S参数作为输入进行参数预测。相比于简单的全连接神经网络,卷积神经网络不仅能够大幅度地减少网络参数,还有效避免了过拟合情况的出现,解决了全连接神经网络耗时长的问题,并且由于卷积神经网络对于结构参数的预测是直接的,即使对于初学者也可以节省大量设计时间。仿真结果表明,目标S参数与卷积神经网络预测后得到的S参数拟合程度很高,证明了该方法对微带滤波器物理结构参数的预测有较高的准确率。A parameter prediction method for microstrip bandpass filter based on a convolutional neural network is proposed,and the corresponding prediction is performed for a three⁃band bandpass filter.Firstly,the HFSS electromagnetic simulation software training da⁃ta is used to obtain the data set of S parameters.The S parameters of the microstrip filter are used as the input,and the physical struc⁃ture parameters are used as the output.The convolutional neural network is used for training.Finally,the target S parameters are used as the input for parameter prediction.Compared with the simple fully connected neural network,the convolutional neural network can not only greatly reduce the network parameters,but also effectively avoid the occurrence of over fitting,and solve the problem of long time consumption of the fully connected neural network.Moreover,because the prediction of structural parameters is direct,even for begin⁃ners,the convolutional neural network can save a lot of design time.The simulation results show that the fitting degree between the tar⁃get S parameters and the S parameters predicted by the convolutional neural network is very high,which proves that the method has a high accuracy in predicting the physical structure parameters of the microstrip filter.

关 键 词：卷积神经网络 参数预测 三频带带通滤波器 HFSS电磁仿真 

分 类 号：TN713.5[电子电信—电路与系统] TP183[自动化与计算机技术—控制理论与控制工程]