论文部分内容阅读
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。
This paper presents a fault diagnosis method based on wavelet neural network for analog circuits. This method constructs a wavelet neural network whose excitation function is a scale function and a wavelet function with a compact support set, using the idea of multi-resolution decomposition of functions in the wavelet space. The selection of hidden layer nodes in this wavelet neural network has a theoretical basis and solves the problem that the number of hidden layer nodes in a traditional neural network is difficult to determine. The wavelet neural network and traditional BP network proposed in this paper are respectively used to diagnose the fault of the circuit. It is found that the wavelet network not only has faster learning convergence than the traditional BP neural network, but also effectively avoids the local minimum problem.