摘要
分析了模拟电路故障诊断中故障类重叠.针对在该情况下神经网络训练困难与故障诊断正确率低的问题,提出了一种适于模拟电路的、基于神经网络故障诊断的故障重分类方法,给出了该方法的数学模型.通过诊断示例表明,该方法在故障类存在重叠时,降低了神经网络的训练难度,故障诊断的正确率达到99%以上.
The faults overlap circumstances in analog fault diagnosis are analyzed, In order to solve the problems of high difficulty of neural network training and low accuracy of fault diagnosis under above circumstances, a fault re-classification method based on neural network fault diagnosis in analog circuit is proposed and the mathematics model of the method is given. The fault diagnosis example shows that the difficulty of neural network training is diminished and the fault diagnosis accuracy can reach more then 99 % when faults overlaps exist.
出处
《北京交通大学学报》
EI
CAS
CSCD
北大核心
2006年第2期53-56,64,共5页
JOURNAL OF BEIJING JIAOTONG UNIVERSITY
关键词
重分类
故障诊断
模拟电路
神经网络
re-classification
fault diagnosis
analog circuit
neural network
作者简介
张屹（1979-）．男．辽宁义县人．博士生．email：zhycare@126．com
魏学业（1963-）．男，山东临朐人，教授，博士．博士生导师
