基于神经网络训练样本优选算法的模拟电路软故障诊断  

Soft Fault Diagnosis of Analog Circuit Based on Optimized Algorithm of Neural Network Training Samples

作　　者：李晓建[1] 赵大炜[2] 陈祥初[1] 吴涛[1] 

机构地区：[1]第二炮兵装备研究院,100085 [2]机械制造系统工程国家重点实验室

出　　处：《电气应用》2005年第3期88-91,共4页Electrotechnical Application

摘　　要：为了解决非线性模拟电路软故障不易诊断的难题 ,根据模拟电路软故障的特点 ,提出了基于σ判据的训练样本优选算法 ,保证训练样本在数量较少的情况下 ,充分包含电路软故障特征。诊断实例证明 ,应用此优选算法对样本进行筛选 ,神经网络的训练速度快 ,对于模拟电路软故障的诊断正确率高。In order to solve the problem in soft fault diagnosis of nonlinear analog circuit,according to the characteristic of nonlinear analog circuit soft fault,an optimized algorithm of neural network training samples based on criterion is proposed,which can ensure the full coverage of fault information in the condition of few training samples It was proved that,the fast training velocity of neural network and the high correct rate of analog circuit soft fault diagnosis can be obtained by the samples selected through the optimized algorithm

关 键 词：非线性模拟电路 软故障 故障诊断 神经网络 训练样本优选算法 

分 类 号：TN710[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]