A feature extraction method based on LMD and MSE and its application for fault diagnosis of reciprocating compressor

Hai-yang Zhao, Jindong Wang, Junjie Xing, Yi-qi GaoJournal of Vibroengineering 2015

Abstract

According to the nonlinearity, nonstationarity and multi-component coupling characteristics of reciprocating compressor vibration signal, a feature extraction method based on Local mean decomposition (LMD) and Multiscale entropy (MSE) is proposed for the diagnosis of reciprocating compressor oversized bearing clearance faults. Vibration signals in each state are decomposed into a series of PF components with LMD method, and the highlighted PF components which contain the main information of fault state were chosen according to the correlation coefficient. MSE of the selected PF components were calculated, and the optimized scale factor was selected based on the maximum of average distances between different states, so the eigenvectors which have the best divisibility were extracted. Taken SVM as pattern classifier, the eigenvectors of four bearing clearance faults were diagnosed, and superiority of this method is verified by comparing the recognition results of the eigenvectors extracted by three other methods.
BibTeX
@article{zhao2015feature,
  author = {Hai-yang Zhao and Jindong Wang and Junjie Xing and Yi-qi Gao},
  title = {{A feature extraction method based on LMD and MSE and its application for fault diagnosis of reciprocating compressor}},
  journal = {Journal of Vibroengineering},
  year = {2015},
}
Figure