›› 2019, Vol. 39 ›› Issue (1): 192-196.DOI: 10.3969/j.issn.1006-1355.2019.01.036

Previous Articles     Next Articles

Fault Diagnosis of Gearbox Bearings based on KICA-GDA and LSSVM

  

  • Received:2018-01-11 Revised:2018-05-13 Online:2019-02-18 Published:2019-02-18

基于KICA-GDA和LSSVM的齿轮箱轴承故障诊断

杨伟新1王平12, 李舜酩2   

  1. ( 1. 中国航发湖南动力机械研究所,湖南株洲412002; 2. 南京航空航天大学,南京21000 )
  • 通讯作者: 杨伟新

Abstract:

In order to achieve higher fault recognition rate of gearbox rolling bearing, a fault recognition method was proposed based on KICA (kemel independent component analysis)-GDA (generalized discriminant analysis) and LSSVM (least squares support vector machine). Firstly, fault features of rolling bearing vibration signal such as kurtosis and information entropy were computed and were recognized as initial feature vectors ,which were mapped into a kernel feature space with KICA to omit the redundancy and correlation among the fault features. Then the GDA was used to implement feature fusion and new features were inputted to LSSVM classifier for fault classification. The experimental results show the KICA-GDA and LSSVM method improves LSSVM’s classification performance by KICA-GDA obtaining additional discriminative information, and has more accurately rolling bearing fault classification than direct LSSVM method.

摘要:

摘 要:为了提高齿轮箱轴承故障识别率,提出基于核独立分量分析(KICA)、广义辨别分析(GDA)与最小二乘支持向量机(LSSVM)的轴承故障识别方法。首先将轴承故障振动信号的谱峭度、信息熵等故障特征作为原始特征向量,通过KICA方法将原始特征向量映射到核特征空间,从而去掉不同故障特征间的冗余并消除原始特征向量间的相关性。然后利用GDA方法对故障特征进行非线性融合,并构造新的特征向量。最后,将新的特征向量作为LSSVM分类器的输入,并实现轴承的故障分类。齿轮箱滚动轴承故障诊断试验结果表明:KICA-GDA和LSSVM的故障诊断方法可以识别出更多的轴承故障信息,且提高了LSSVM的分类性能,该方法相对于直接采用LSSVM进行分类的轴承故障方法具有更优秀的分类性能。

关键词: 振动与波, 滚动轴承, 齿轮箱, 故障诊断, KICA, GDA, LSSVM