/PCCNN

This is a repository of code and experiment data for paper <A probability confidence CNN model and its application in mechanical fault diagnosis>

Primary LanguagePython

PCCNN

This is a repository of code and validation data for paper [A novel probability confidence CNN model and its application in mechanical fault diagnosis].

Publication: IEEE Transactions on Instrumentation & Measurement

DOI: 10.1109/TIM.2021.3077965

The description of folder as follow.

  1. code: running BATCH_ALL.py will automatically complete the data pre-processing, model training, comparison method calculations and result statistics, and save the results of the ten-fold cross validation in the result folder.
  2. data: raw validation data for 20 Datasets A~Q.
  3. model: trained PCCNN models that can be used for testing.
  4. result: the result of ten-fold cross-validation, including the proposed and comparison methods.

Required python and python libraries as follow.

  1. python==3.6 or 3.7.
  2. pytorch==1.4.0+cu101, CUDA==10.1.
  3. xlrd==1.2.0
  4. scikit-learn==0.23.2
  5. pandas==1.1.1
  6. numpy==1.19.2
  7. nptdms==0.28.0
  8. pywavelets==1.1.1
  9. matplotlib==2.2.5
  10. progressbar==2.5
  11. paramiko==2.7.2
  12. openpyxl==3.0.5
  13. cvxopt==1.2.5

The raw data file, trained model files and log files has been uploaded to network drive as follows. Network drive url: https://pan.baidu.com/s/1EF3pzDWTmjUKRD4tgh_drQ Extraction code: wz8e

Please do not hesitate to contact me if you have any queries, my e-mail address is caiweidon@qq.com.