基于LLTSA的轴承故障数据降维方法研究

Dimension Reduction of Bearing Fault Data Based on LLTSA

针对工作中轴承内外圈及滚动体难以分别监测以及轴承故障数据特征指标多、维数高的特点,考虑对轴承状态数据进行降维。考虑到常用降维方法自身存在的各种缺陷,采用线性局部切空间排列算法(LLTSA)对轴承故障数据进行降维并投影到三维空间,并与PCA以及KPCA方法进行比较。结果表明LLTSA算法对于滚动轴承内外圈以及滚动体不同故障具有较好的分类性能。

Aiming at the difficulty in separate monitor of bearing inner race, outter race and ball,as well as the high-dimensional property of fault data, dimension reduction of fault data is considered. To avoid the defects of normal ways, LLTSA method is adoptted to extract the eigenvectors of high dimensional matrixes and eigenvectors are projected to visual space. LLTSA method shows a better performance in bearing fault pattern recognition in comparison to the PCA and KPCA methods.