基于小波变换的容差电路故障诊断

Fault Diagnosis of Tolerance Circuits Based on Wavelet Method

针对目前模拟电路中电子元器件存在的容差与非线性导致电路故障难以检测的现状,设计了适用于诊断由器件超出容差所引起的模拟电路故障的小波分析诊断方法。通过设定故障进行蒙特卡罗容差实验,采用小波神经网络,对故障输出信号进行小波分析提取其小波高频系数参量,经PCA分析和归一化后形成训练特征向量,并经过BP神经网络训练后,故障信号通过小波神经网络后能够快速精准的对故障器件进行定位。通过大量样本进行仿真计算表明所设计的小波特征参量故障诊断法对于模拟电路具有很好的故障分辨率。

A systematic method of fault diagnosis for analogue circuits based on the wavelet transform with neural network is presented in the paper .This method is designed for diagnosis faults led by tolerance and nonlinear electronic components ,which is detected difficultly with current methods .With the Monte Carlo tolerance fault experiment ,the feature information of fault circuit output signal is extracted by wavelet , then analyzed by PCA and normalized to form the training feature vectors ,w hich is used to train the BP neural network .The simulation shows the fault can be positioned to component with the training wavelet neural network .Large number of simulation experiments show that the designed wavelet neural network method for analog circuit fault diagnosis has a good fault resolution .