基于小波能量熵和LLE的滚动轴承故障特征提取方法

Rolling bearing fault feature extraction method based on wavelet energy entropy and LLE

为解决滚动轴承中故障振动信号成分复杂和信息冗余引起的故障诊断困难,提出一种基于小波能量熵和局部线性嵌入(LLE)相结合的滚动轴承故障特征提取方法。首先,对原始信号进行小波分解并提取信号的能量熵,同时提取时域特征变量,组成特征向量集;其次,采用LLE对高维特征向量进行特征融合,实现特征信号的降维;最后,训练支持向量机对滚动轴承故障进行识别。实验结果表明,该方法能够有效识别滚动轴承故障类型,准确率达到96.25%。

In order to solve the difficult problem of fault diagnosis caused by complex vibration signal components and redundant information in rolling bearings, a method based on combination of wavelet energy entropy and the locally linear embedding(LLE) for rolling bearing fault feature extraction was proposed. Firstly, the original signal was decomposed and reconstructed by wavelet, and the energy entropy of the signal was extracted. Meanwhile, the characteristic variables in time domain were extracted to form the feature vector set. Secondly, LLE was used to carry out feature fusion of high-dimensional feature vectors to achieve dimensionality reduction of feature signals. Lastly, support vector machine(SVM) was built to identify rolling bearing faults. The results show that this method can effectively identify the types of rolling bearing faults, and the accuracy rate reaches 96.25%.