摘要
针对滚动轴承振动信号的非平稳以及非线性特点,提出了一种基于非线性流形的滚动轴承复合故障诊断方法。该方法首先提取振动信号的时域指标和小波包频带分解能量所构成的频域指标,组成原始特征空间,采用基于判别准则的邻域因子优化选择算法,运用基于局部切空间排列算法的非线性降维算法对原始特征空间进行学习,极大地保留了信号中内在的整体几何结构信息,从而提取出振动信号最优的敏感故障特征。试验结果表明,与经典的线性降维方法相比,该方法的聚类效果更好。
A combine fault diagnosis approach for roller bearing based on nonlinear manifold is proposed accord- ing to the fact that vibration signal of roller bearing is non - stationary and time - variation. After constructing the orig- inal feature space with the performance index of the vibration signal in time domain and frequency band energy decom- position using wavelet packets, according to optimal selection algorithm of neighborhood factor based on discriminate criterion, adopting a nonlinear dimensionality reduction algorithms based on local tangent space alignment algorithm to the original feature space, the whole geometry structure information embeded into the signal is hugely reserved so that the optimal sensitive fault feature of the vibration signal can be acquired. The experimental results show that the pro- posed method has better classification performance than the traditional linear dimensionality reduction method.
出处
《机械传动》
CSCD
北大核心
2012年第7期89-91,110,共4页
Journal of Mechanical Transmission
关键词
非线性流形
滚动轴承
复合故障
局部切空间排列算法
Nonlinear manifold Roller bearing Combine fault Local tangent space alignment algorithm
