小波包和BP神经网络在齿轮箱故障诊断中的应用

研究如何采用处理非平稳性的实用方法以提高监测诊断效率及水平是国内外专家一直研究的课题之一。小波包技术将信号中不同的分量无冗余、无疏漏、正交地分解到独立的频带内,这些频带里的信号能量守衡,每个频带里信号的能量对于状态监测和故障诊断都是十分有用的信息。本文对齿轮箱振动信号应用小波包分解提取故障特征向量,进一步用特征向量训练前向传播BP人工神经网络,建立齿轮运行状态分类器,对齿轮故障进行识别。实验结果表明,本文方法对齿轮箱故障诊断十分有效。

Fault Diagnosis of Gearbox Using Wavelet Package and BP Neural Network

The study of how to use practical methods for non-stationary signals to improve the level and efficiency of monitoring and diagnosis is one of the important research subjects studied by many researchers all over the world. Wavelet package non-redundantly, non-omissively and orthogonally divide the different parts of signal into unattached frequency band. It maintains the conservation of energy in these bands so that the energy of each band is very useful information for condition monition and fault diagnosis. The wavelet package is applied to decompose vibration signal of gearbox to get fault eigenvector. Then, the eigenvector is employed to train the forward transmitting BP neural network so as to set up the classifier of operational state to identity gear fault .The experimental results testify that the proposed method is a useful tool to diagnose gearbox fault.