Research on bearing fault diagnosis method based on transformer neural network
Исследование метода диагностики неисправностей подшипников на основе трансформерной нейронной сети
2022-04-12
SCID: 54.1/sxt4uxx6
Discuss with AI
Case Western Reserve University datasetattention mechanismbearing fault diagnosissignal Transformer neural networkvibration time-series signals
Figures from the paper
Abstract (AI)
Abstract The Attention mechanism (AM) has been widely used for fault diagnosis and identifying the health of industrial equipment. Existing research has only used AM in combination with deep networks, or to replace certain components of these deep networks. This reliance on deep networks severely limits the feature extraction capability of AM. In this paper, a bearing fault diagnosis method is proposed based on a signal Transformer neural network (SiT) with pure AM. First, the raw one-dimensional vibration time-series signal is segmented and a new segmented learning strategy is introduced. Second, linear encoding and position encoding are performed on the segmented subsequences. Finally, the encoded subsequence is fed to the Transformer for feature extraction to achieve fault identification. The validity of the proposed method is verified using the Case Western Reserve University dataset and the self-priming centrifugal pump bearing dataset. Compared with other existing methods, the proposed method still achieves the highest average diagnostic accuracy without any data preprocessing. The results demonstrate that the proposed SiT based on pure AM can extract features and identify faults from the raw vibration signal, and has superior diagnostic performance.
Key Findings
1
A signal Transformer (SiT) using pure attention mechanisms is proposed for bearing fault diagnosis without relying on deep-network feature extractors.
2
Compared with existing methods, the proposed SiT achieves the highest average diagnostic accuracy without any vibration-signal preprocessing.
3
Experiments on the Case Western Reserve University and self-priming centrifugal pump bearing datasets validate the proposed diagnostic approach.
4
The method segments raw one-dimensional vibration signals and applies segmented learning, linear encoding, and positional encoding before Transformer-based feature extraction.
5
The results show that pure-attention Transformers can extract fault features directly from raw vibration signals and provide superior diagnostic performance.
Research Object
industrial equipment bearings represented by raw one-dimensional vibration time-series signals, including bearings from the Case Western Reserve University and self-priming centrifugal pump datasets
Research Subject
bearing fault identification and diagnostic accuracy based on feature extraction from raw vibration signals using a pure-attention Signal Transformer
Publication Details
Publication Date
2022-04-12
Journal
Publisher
ISSN
Cited by
90
Access Type
Author Information
Download PDF
Subscribe to digest