基于Elman反馈神经网络的导线串扰问题预测  

WIRE CROSSTALK PREDICTION BASED ON ELMAN FEEDBACK NEURAL NETWORK

作　　者：王锦锦 聂鑫[1] 王伟[1] 石跃武 相辉[1] 杨静[1] 朱志臻[1] Wang Jinjin;Nie Xin;Wang Wei;Shi Yuewu;Xiang Hui;Yang Jing;Zhu Zhizhen(State Key Laboratory of Intense Pulsed Radiation Simulation and Effect , Northwest Institute of Nuclear Technology, Xi＇ an 710024, Shaanxi , China)

机构地区：[1]西北核技术研究所强脉冲辐射环境模拟与效应国家重点实验室,陕西西安710024

出　　处：《计算机应用与软件》2018年第5期33-36,66,共5页Computer Applications and Software

摘　　要：为了更好地预测导线串扰问题,提出使用Elman反馈神经网络方法。该反馈神经网络方法具有很强的联想记忆和优化计算功能,使用与导线串扰响应有关的参数作为网络的输入,将预测到的导线的耦合电压值作为输出。Elman反馈神经网络采用的训练数据为多导体传输线方法计算得到的。采用训练得到的该反馈神经网络预测模型对未知样本进行预测,比较预测结果和真实测试结果。实验表明建立的Elman神经网络对于导线串扰问题的预测误差较小,结果比较准确。In order to better predict the wire crosstalk problem,an Elman feedback neural network method was proposed. The feedback neural network method had a strong associative memory and optimization calculation function,used the parameters related to the wire crosstalk response as the input of the network,and took the predicted coupling voltage value of the wire as an output. The training data used by the Elman feedback neural network was calculated for the multi-conductor transmission line method. Finally,the feedback neural network prediction model obtained by training was used to predict unknown samples,and the prediction results and real test results were compared. The Experimental results show that the established Elman neural network had less prediction error for wire crosstalk problems and the result was more accurate.

关 键 词：Elman反馈神经网络 预测 导线串扰 电磁兼容 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]