基于小波分形分析和脊波网络的模拟电路故障诊断方法

摘要

A novel ridgelet network approach for diagnosing faults of analog circuits is proposed.This systematic approach is performed in such way that utilizing wavelet fractal analysis,PCA and data normalization preprocess the fault response signals and using ridgelet network classifies the faulty components.Meanwhile,the PCA method is used for selecting the reasonable numbers of ridgelet units of hidden layer of related networks.The simulation results show that the proposed diagnostic system can perform analog fault diagnosis effectively,and can diagnose not only the single faults and but also the multiple faults effectively.%提出了一种新型的脊波网络方法来进行模拟电路的故障诊断,这种系统化的方法采用小波分形分析、主元分析和数据归一化作为数据预处理器来进行故障响应信号的预处理,采用脊波网络进行故障元件的分类,并采用主元分析方法选择脊波网络隐层脊波元的数目。仿真结果表明,提出的诊断系统能有效地实现模拟电路的故障诊断,不但能有效地诊断模拟电路的单故障情况,还能诊断多故障情况。