DCA-BiGRU

The pytorch implementation of the paper Fault diagnosis for small samples based on attention mechanism

However, in fact, the title Fault diagnosis for small samples based on interpretable improved space-channel attention mechanism and improved regularization algorithms fits the research content of the paper better.

The dataset comes from 12khz, 1hp

Contributions:

1. 1D-signal attention mechanism [code]
2. AMSGradP [code]
3. 1D-Meta-ACON [code]
4. At the beginning, I found that many model designs did not connect GAP operation after BiGRU/BiLSTM, which is the basically routine operation. I found that GAP works very well. [code]
5. 1D-Grad-CAM++ [code]
6. AdaBN [code]

Attention Block（SCA）

How does it work?

If it is helpful for your research, please kindly cite this work:

@article{he2025pifcapsule,
  title={Human prior knowledge-embedded first-layer interpretable paradigm for rail transit vehicle human-computer collaboration monitoring},
  author={He, Chao and Shi, Hongmei and Liao, Jing-Xiao and Liu, Qiuhai and Li, Jianbo and Yu, Zujun},
  journal={Journal of Industrial Information Integration},
  volume={51},
  pages={101068},
  year={2025},
  doi={10.1016/j.jii.2026.101068},
  publisher={Elsevier}
}

Our other works

@article{HE2023110846,
title = {IDSN: A one-stage interpretable and differentiable STFT domain adaptation network for traction motor of high-speed trains cross-machine diagnosis},
journal = {Mechanical Systems and Signal Processing},
volume = {205},
pages = {110846},
year = {2023},
doi = {https://doi.org/10.1016/j.ymssp.2023.110846},
author = {Chao He and Hongmei Shi and Jianbo Li},
} 

Environment

pytorch == 1.10.0
python == 3.8
cuda == 10.2

Contact

• Chao He
• chaohe#bjtu.edu.cn (please replace # by @)

Views