基于S变换和小波神经网络的容差模拟电路故障诊断  

Fault Diagnosis of Tolerence Analog Circuits Based on S-transform and Wavelet Neural Network

作　　者：罗琨[1] 何怡刚[1] 方葛丰[1] 

机构地区：[1]湖南大学电气与信息工程学院,湖南长沙410082

出　　处：《计算机测量与控制》2011年第6期1304-1307,1310,共5页Computer Measurement &Control

基　　金：国家自然科学基金项目(60876022);国家863计划(2006AA04A104);国家杰出青年科学基金(50925727);湖南省科技计划项目(2008Gk2022);广东省教育部产学研项目(2009B090300196)资助

摘　　要：提出了一种结合S变换和小波神经网络的容差模拟电路故障诊断新方法;该方法通过对被测电路的冲激响应进行S变换,提取信号的时频信息做为特征量,并将所提取的特征量做为小波神经网络的输入进行训练并分类;仿真实验结果表明该方法诊断速度快且故障定位准确率高,在噪声影响、故障类型的特征向量重叠率高以及可测节点不足的情况下,具有良好的故障识别效果。A new fault diagnosis method of tolerence analog circuits based on the combination of S-transform and wavelet neural net- work is proposed. S--transform is used to analysis the impulse response of circuits under test （CUT）, whose results of time and frequency im^ormation are extracted as features. These features are applied to train wavelet neural network and classify fault classes, the simulation results show that the scheme reduce the diagnosis time and locate faults accurately, it can also identify faults effectively in the situation of high level noises, overlap in fault feature vectors and insufficient accessible nodes.

关 键 词：故障诊断 容差模拟电路 S变换 小波神经网络 

分 类 号：TP206.1[自动化与计算机技术—检测技术与自动化装置]