基于小波神经网络的容差模拟电路故障诊断方法

Diagnosis Method for Tolerance Analog Circuit Fault Based on Wavelet Neural Network

以模拟电路的故障诊断为例,利用小波分析,将电路故障信号进行层次分解,获得不同频段的信号成分,取其能反映故障信号特征的成分作为电路故障特征,再输入给神经网络,大大减少了神经网络的输入数目、简化了神经网络的结构、减少了它的训练时间,并提高了辨识故障类别的能力.

Wavelet neural network （WNN）combines the time-frequency location and multiple-scale analysis of wavelet transform and the self-organization and learning ability of the neural network, which is very suitable for fault diagnosis. In this paper, the fault diagnosis method for analog circuit based on WNN is introduced. Wavelet Transform is used to decompose fault signal in levels and acquire signal component in various frequency segment, using the ingredient that could reflect the fault signal character as circuits fault character to input ANN to neural network. With this method, the input number is reduced, the structure is simplified and the training time is reduced. Therefore, the ability to identify fault sort is enhanced.