We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
A rolling bearing fault evolution state indicator based on deep learning and its application.
- Authors
Liu, Xiyang; Chen, Guo; Wei, Xunkai; Liu, Yaobin; Wang, Hao
- Abstract
Aiming at the limitation of early fault warning and the diagnosis of aero-engine main bearing when there are only normal operation data, a rolling bearing fault evolution state indicator based on deep convolutional neural network (CNN) and wavelet analysis was proposed. To be specific, firstly, the wavelet band envelope method was adopted to identify the early fault evolution process, and the feature distance between the degraded data and the normal ones was extracted by using deep CNN to develop the evolution state indicator. Then, the evolution stages were divided by using unsupervised clustering method. Finally, the remaining useful life (RUL) was predicted based on particle filter (PF). Three different groups of whole life cycle data of rolling bearings under various working conditions were used to prove the feasibility of the indicator. The results show that the wavelet-CNN features of completely different fault data show similar evolution trends, and the normalization of warning threshold can be realized based on the train labels. In conclusion, the results are of great significance for the early fault evolution monitoring, condition evaluation and remaining useful life prediction of rolling bearings without the absence of fault samples under actual aeroengine operation.
- Subjects
DEEP learning; ROLLER bearings; CONVOLUTIONAL neural networks; REMAINING useful life; LIFE cycles (Biology); WAVELETS (Mathematics)
- Publication
Journal of Mechanical Science & Technology, 2023, Vol 37, Issue 6, p2755
- ISSN
1738-494X
- Publication type
Article
- DOI
10.1007/s12206-023-0504-8