基于LLE和LSSVM的滚动轴承故障诊断  

Research on Rolling Bearings Fault Diagnosis Based on LLE and LSSVM

作　　者：李力[1] 李冕[1] 陈法法[1] 

机构地区：[1]三峡大学水电机械设备设计与维护湖北省重点实验室,湖北宜昌443002

出　　处：《煤矿机械》2015年第7期308-310,共3页Coal Mine Machinery

基　　金：国家自然科学基金资助(51405264)

摘　　要：针对滚动轴承故障特征混叠难以有效区分的问题,提出基于局部线性嵌入(LLE)与最小二乘支持向量机(LSSVM)结合的故障诊断方法。在由振动信号时域和频域统计指标构造的多维特征空间中,通过LLE算法对多维特征空间进行非线性降维处理,得到初始低维流形结构。将低维流形结构导入LSSVM中进行学习训练与故障辨识。应用于滚动轴承故障分析表明,该方法不仅对高维复杂的非线性故障特征具有良好的降维性能,而且故障识别率较之传统方法有明显提高,能够有效识别出高维特征空间的非线性故障特征。In consideration of the overlapping of rolling bearings fault features and the difficulty to distinguish these features, a rolling bearing fault diagnosis method based on local linear embedding（LLE） and least squares support vector machine（LSSVM） was proposed. In the multi-dimensionality feature space constructed by time and frequency domain statistic indices, the nonlinear multi-dimension reduction based on LLE was introduced to get the initial low-dimensional manifold features value, and then the low-dimensional feature vector was regarded as the input features vector of the LSSVM for rolling bearings fault classification. The analysis of the experiment data for rolling bearings show that this method not only has good dimensionality reduction performance for high-dimensional complex nonlinear fault features, but also significantly improves the fault diagnosis accuracy compare with traditional methods. This method can effectively identify the nonlinear fault features existing in high dimension space.

关 键 词：局部线性嵌入(LLE) 最小二乘支持向量机(LSSVM) 滚动轴承故障 诊断 

分 类 号：TH133.3[机械工程—机械制造及自动化]