Journal
JOURNAL OF MECHANICAL SCIENCE AND TECHNOLOGY
Volume 37, Issue 4, Pages 1637-1651Publisher
KOREAN SOC MECHANICAL ENGINEERS
DOI: 10.1007/s12206-023-0306-z
Keywords
Adversarial learning; Domain adaptation; Fault diagnosis; Rolling bearing; Wind turbines
Categories
Ask authors/readers for more resources
Due to the lack of labeled data on wind turbine bearings, a new wind turbine bearing fault diagnosis method based on a dynamic multi-adversarial adaptive network (DMAAN) was proposed. The method utilized laboratory data to obtain fault diagnosis models and involved evaluating distribution differences and reducing them through adversarial training. The validity of DMAAN was verified through transfer experiments, showing higher diagnostic accuracy and better transmission capability compared to existing methods.
Owing to the shortage of available labeled data on wind turbine bearings, a new wind turbine bearing fault diagnosis method based on a dynamic multi-adversarial adaptive network (DMAAN) was proposed. In this new method, a laboratory data were used to obtain fault diagnosis models for wind turbine bearings. The first step was evaluating the interdomain distribution difference and intraclass distribution differences between domains. The second step was setting a dynamic adversarial factor to dynamically measure the relative contribution of the two different distributions. The last step was, reducing the distribution difference through multiple adversarial training, to obtain the diagnosis results. The validity of DMAAN was verified via the transfer experiments of laboratory datasets and wind turbine generator measured datasets. The results showed that DMAAN has a higher diagnostic accuracy and better transmission capability in cross-machine transfer fault diagnosis in compare with the existing methods.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available