基于小波神经网络的模拟电路故障诊断

本文对模拟电路提出了一种基于小波神经网络的故障诊断方法。该法利用小波空间中函数的多分辨率分解思想，构造了一种激励函数为具有紧支撑集的尺度函数和小波函数的小波神经网络。这种小波神经网络隐层节点数的选取有理论根据，解决了传统神经网络隐层节点数难以确定的问题。分别用本文提出的小波神经网络和传统BP网络对实例电路进行故障诊断，结果发现，小波网络比传统BP网络方法不仅学习收敛速度快，而且有效地避免了局部最小值问题。

Fault diagnosis of analog circuit based on wavelet neural network

A method of analog circuit fault diagnosis based on wavelet neural network is presented in this paper. According to the multi-resolution analysis thought in the wavelet space, a wavelet network model for learning is obtained. In the model compactly supported scaling function and wavelets are adopted as the active function. The number of the nodes in the hidden layer of this wavelet neural network can be obtained based on theory criteria, which solves the difficulty of determining the number of nodes in traditional network. This wavelet neural network and traditional BP network are used respectively in the fault diagnosis of an example circuit. Simulation results indicate that the proposed method can advance convergence rate and avoid converging to local minimum effectively.