摘要
对齿轮箱故障诊断问题进行研究,由于齿轮的振动信号是非平稳信号,常规的齿轮特征提取方法难以从振动信号中提取有效故障特征信息。笔者采用小波包理论对齿轮振动信号应用db12小波进行多层分解后,从而对信号进行消噪,并对消噪后的信号进行小波包3层分解及系数重构,再次对各频段能量进行处理分析从而得到特征向量。最终应用归一化方法对特征向量处理后再结合RBF神经网络进行故障诊断,并且取得了良好的诊断效果。
In this article, gearbox fault diagnosis is mainly researched. Because the vibration signal of gear is non - stationary, the fault feature information can not be obtained availably by regular methods to extract the feature by fault gears. The wavelet package theory and multi - wavelet db12 are used to analyze the signal for de - noising the vibration signal,and the three - layer wavelet packet is used to decom pose the de - noising signal. Then the energy of each frequency band is analyzed in order to get the characteristic vector. Finally normalization method and RBF neural network are used to diagnose the faults, and the result of the fault diagnosis is very well.
出处
《机械研究与应用》
2010年第1期21-24,共4页
Mechanical Research & Application
关键词
齿轮
小波包
RBF神经网络
故障诊断
gears
wavelet package
RBF neural network
fault diagnosis
