基于拉普拉斯特征映射的滚动轴承故障识别

用传统的线性方法对非平稳和非线性运行状态的滚动轴承进行故障诊断时,效果欠佳。为了及时、准确地监测轴承的运行状态,提出了将拉普拉斯特征映射算法(Laplacian Eigenmap LE)应用到滚动轴承的故障识别中。在振动信号构建的时域和频域高维特征空间矩阵中,充分利用LE算法在非线性特征提取和降维的优点,进行学习,提取表征轴承状态的特征量,并以可视化的聚类结果进行表示。实验模拟了轴承的4种不同类型故障以及滚动体的4种不同受损程度,采用模式识别中聚类性的类内距和类间距两个参数作为衡量指标。与PCA和KPCA两种方法对比,LE不仅明显识别出四种故障类型和有效的区分出滚动体的不同受损程度,而且识别率大大提高。并通过测试样本组验证了LE方法的有效性。

A Method for Fault Diagnosis of Rolling Bearings Based on Laplacian Eigenmap

The traditional linear diagnosis methods for rolling bearing fault with non-stationary and nonlinear running status was not effective. In order to monitor the rolling status accurately and timely, a new diagnosis method was put forward by applying the algorithm of Laplacian Eigenmap (LE) to the diagnosis of rolling. It fully used the advantage of LE algorithm for extracting nonlinear features and reducing dimension for characteristic space matrix in the time domain and frequency domain constructed by vibration signal，extracted the features of running status of external rolling and visualized the clustering results. The experiments using two parameters (between-class scatter and within-distance in pattern recognition) as the measurable indicators simulated four different faults of the bearings and the four different extent of the damage of balls in bearing. Compared with PCA & KPCA, LE clearly identifies the four different faults and the different extent of the damage of balls, and its recognition rate rises greatly.  The effectiveness of LE has been Verified by testing samples.