一种基于经验模态分解和流形学习的滚动轴承故障诊断方法

A ROLLER BEARING FAULT DIAGNOSIS METHOD BASED ON THE EMPIRICAL MODE DECOMPOSITION (EMD) AND MANIFOLD LEARNING (LTSA)

提出一种基于经验模态分解(EMD)和流形学习(LTSA)的滚动轴承故障诊断方法。首先,利用EMD对滚动轴承振动信号进行自适应分解,计算IMF分量的协方差矩阵特征值,组成滚动轴承状态原始特征集;然后利用LTSA对原始特征集进一步的融合提取;将所得新特征输入到K-means分类器中进行故障识别与聚类。实验分析结果表明:该方法可以有效地对滚动轴承的工作状态和故障类型进行识别。

A roller bearing fault diagnosis method based on the empirical mode decomposition （EMD） and manifold learning （LTSA） was presented. After the adaptive decomposition of the roller beating vibration signal by the EMD technique, its original state feature set of the rolling bearing was acquired by calculating eigenvalues of IMF＇s covariance matrix. The extraction performance of the original feature set was further fusion implemented by using the LTSA. The new features obtained were input into a K-means classifier, and the output of the K-means classifier was clustering results. Finally, the experiment results show that the proposed method can effectively identify work status and fault type of roller bearing.