DGAT-LPS: A new semi-supervised fault diagnosis method called dynamic graph attention network with label propagation strategy
- Core codes for the paper "Semi-supervised fault diagnosis of machinery using LPS-DGAT under speed fluctuation and extremely low labeled rates"
- Created by Shen Yan, Haidong Shao, Yiming Xiao, Jian Zhou, Yuandong Xu and Jiafu Wan.
- Journal: Advanced Engineering Informatics
- Python 3.8
- torch-geometric 2.2.0
- pytorch 1.10.1
- pandas 1.5.3
- numpy 1.23.5
- and other necessary libs
- This repository provides a concise framework for semi-supervised fault diagnosis. It includes a demo dataset; the pre-processing and graph composition process for the data and the model proposed in the paper.
- You just need to run
train_test_graph.py
. You can also adjust the structure and parameters of the model to suit your needs.
data
contians a demo datasetdatasets
contians the pre-processing and graph composition process for the datamodels
contians the model proposed in the paper
- The DGAT is derived from the paper: arXiv:2105.14491
- Special thanks to Li et al. for the GNN base framework provided by PHMGNNBenchmark
If you use our work as a comparison model, please cite:
@paper{DGAT-LPS,
title = {Semi-supervised fault diagnosis of machinery using LPS-DGAT under speed fluctuation and extremely low labeled rates},
author = {Shen Yan, Haidong Shao, Yiming Xiao, Jian Zhou, Yuandong Xu and Jiafu Wan},
journal = {Advanced Engineering Informatics},
volume = {53},
pages = {101648},
year = {2022},
doi = {https://doi.org/10.1016/j.aei.2022.101648},
url = {https://www.sciencedirect.com/science/article/abs/pii/S1474034622001124},
}
If our work is useful to you, please star it, it is the greatest encouragement to our open source work, thank you very much!