一种基于Transformer模型的轴承故障诊断方法
By combining multi-scale convolutional embedding based on the Transformer model, frequency-mixed attention, and multi-path convolutional feedforward modules, the shortcomings of existing bearing fault diagnosis methods in terms of accuracy and generalization ability are solved, and efficient fault detection in strong noise environment is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CAPITAL NORMAL UNIVERSITY
- Filing Date
- 2023-09-27
- Publication Date
- 2026-07-17
AI Technical Summary
Existing bearing fault diagnosis methods are insufficient in terms of accuracy and generalization ability. In particular, signal processing-based methods are affected by noise and the complexity of parameter settings, while knowledge-based methods are limited by expert knowledge and subjectivity, resulting in low detection accuracy.
A bearing fault diagnosis method based on the Transformer model is adopted. Multi-scale features are extracted by a multi-scale convolutional embedding module, global context information is captured by a frequency-mixed attention module, and feature map weights at different scales are learned by a multi-path convolutional feedforward module. Finally, a fully connected classification model is used for diagnosis.
It improves the accuracy and adaptability of bearing fault detection, can effectively perceive more contextual information in strong noise environments, reduces the number of model parameters and computational overhead, and enhances the modeling ability of non-stationary time series.
Smart Images
Figure CN117387948B_ABST