模拟电路故障诊断的小波神经网络方法  

Fault Diagnosis of Analog Circuit Based on Wavelet and Neural Network

作　　者：阳辉[1] 罗琨[1] 何怡刚[1] 

机构地区：[1]湖南大学电气与信息工程学院,长沙410082

出　　处：《微电子学》2010年第1期50-53,57,共5页Microelectronics

基　　金：国家自然科学基金资助项目(50677014;60876022);高校博士点基金资助项目(20060532002);国家高技术研究发展(863)计划基金资助项目(2006AA04A104);湖南省科技计划项目资助(2008Gk2022)

摘　　要：利用小波和改进型BP神经网络相结合的方法,进行模拟电路故障诊断;通过分析被测电路的冲激响应,来识别电路中的故障元件;用小波对冲激响应信号进行多尺度分解,进行归一化后,提取特征信息作为神经网络的输入而进行分类。该方法减少了神经网络的输入、简化了其结构、并缩短了训练和处理时间。实验仿真结果表明:该方法能准确实现故障定位且准确率高。Fault diagnosis of analog circuit was accomplished based on improved BP Neural Network （BPNN） combined with wavelet transform. Faulty components were identified by analyzing impulse response of circuit under test. After multi-scale decomposition and normalization of impulse response signal, feature information was extracted and categorized as BPNN input. This method reduces the number of inputs to neural network, simplifies its architecture and minimizes training and processing time. Simulation results showed that the proposed method could locate fault with high accuracy.

关 键 词：模拟电路 故障诊断 小波变换 神经网络 

分 类 号：TN407[电子电信—微电子学与固体电子学]