Journal of Xidian University

Previous Articles     Next Articles

Sparse feature extraction technique and its applications to machinery fault diagnosis

HE Wangpeng;SUN Wei;SU Bo;YANG Yunyi;GUO Baolong   

  1. (School of Aerospace Science & Technology, Xidian Univ., Xi'an 710071, China)
  • Received:2017-06-12 Online:2018-04-20 Published:2018-06-06


To address the problem of extracting periodic-group-sparse features for the purpose of detecting machinery faults, the periodic overlapping group sparsity (POGS) method is thoroughly investigated. The POGS method formulates a convex optimization problem to extract periodic sparse features based on the prior knowledge of machinery fault diagnosis. The non-convex penalty functions are employed to further enhance the sparsity of useful fault features. Moreover, the convexity condition of the POGS optimization problem is provided. A fast iterative algorithm is given for its optimal solution based on the majorization-minimization approach. A simulated signal is formulated to verify the performance of the POGS method for periodic feature extraction. Finally, the POGS method is applied to process experimental data for detecting bearing faults. The estimated results demonstrate that the POGS method can effectively extract the periodic-group-sparse fault features.

Key words: rotating machines, fault diagnosis, convex optimization, sparse feature extraction