/FRAN

This repo provides source code for cross-domain machine fault diagnosis using an unsupervised domain adaptation approach (Feature Representation Alignment Networks).

Primary LanguagePythonApache License 2.0Apache-2.0

FRAN: Unsupervised Cross-domain Fault Diagnosis Using Feature Representation Alignment Networks for Rotating Machinery

This is the Python+PyTorch code to reproduce the results of Fault Severity Diagnosis in paper 'Unsupervised Cross-domain Fault Diagnosis Using Feature Representation Alignment Networks for Rotating Machinery'.

Requirements

  • Platform : Linux
  • Computing Environment:
    • CUDA 10.1
    • TensorFlow 1.6.0
  • Packages: pandas, numpy, scipy, argparse, tqdm.
  • Hardware (optional) : Nvidia GPU (requires around 7GB of GPU memory)

Getting Started

  1. Computing environment set up can be refered to this repo.
  2. Extract preprocessed CWRU data files in './CWRU_dataset'.
  3. Run the code:
  • For training:
bash batchrun.sh
  • For visualization:
python correlationMatrix.py

Citation

Please cite our paper and the dataset if you found them usefull.

@ARTICLE{chen2020unsupervised,
  author={J. {Chen} and J. {Wang} and J. {Zhu} and T. H. {Lee} and C. {De Silva}},
  journal={IEEE/ASME Transactions on Mechatronics}, 
  title={Unsupervised Cross-domain Fault Diagnosis Using Feature Representation Alignment Networks for Rotating Machinery}, 
  year={2020},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/TMECH.2020.3046277}}