摘要
为精确诊断转子故障,采用了基于小波包能量特征向量的弹性BP神经网络和最速下降BP算法神经网络的故障诊断方法,对采集到的信号进行3层小波包分解,构造小波包特征向量,对样本进行3层BP网络训练,实现智能化故障诊断。结果表明采用改进的BP算法优于最速下降BP算法,训练的网络可以很好地诊断转子故障。
In order to diagnose the rotor fault precisely, this paper applies the resilient back propagation neural network and steepest descent back propagation neural network which based on wavelet packet energy eigenvector. It adopts three-layer wavelet packet to decompose the signal which acquisited from the rotor experiment, and constructs the wavelet packet energy eigenvector, take the energy eigenvector as fault samples to train three-layer BP neural network, finally it realizes the intelligent fault diagnosis. The results shows that the ameliorated BP algorithm is better than the steepest descent algorithm, the trained network can exactly diagnose the rotor fault.
出处
《汽轮机技术》
北大核心
2008年第6期437-439,共3页
Turbine Technology
基金
国家自然科学基金资助项目(50676031)
关键词
神经网络
转子
故障诊断
弹性BP算法
小波包特征向量
neural network
rotor
fault diagnosis
resilient back-propagation
wavelet packet energy eigenvector