Skip to content

Latest commit

 

History

History
53 lines (43 loc) · 1.76 KB

README.md

File metadata and controls

53 lines (43 loc) · 1.76 KB

VBL-V001

Baseline methods for the paper Lab-scale Vibration Analysis Dataset and Baseline Methods for Machinery Fault Diagnosis with Machine Learning.

Dataset

Download from here: https://zenodo.org/record/7006575#.Y3W9lzPP2og.
Locate the dataset to path like '/data/VBL-VA001`.
Structure of dataset:

bagus@m049:VBL-VA001$ tree -L 2 . --filelimit 100
.
├── bearing [1000 entries exceeds filelimit, not opening dir]
├── misalignment [1000 entries exceeds filelimit, not opening dir]
├── normal [1000 entries exceeds filelimit, not opening dir]
└── unbalance [1000 entries exceeds filelimit, not opening dir]

4 directories, 4000 files

You can also try the extracted feature under data directory and run the following codes.

Running the program

# First, extract the feature
$ python3 extract_feature.py
# Then you can run any train_* program, i.e.,:
$ python3 train_svm.py
Shape of Train Data : (3200, 27)
Shape of Test Data : (800, 27)
Optimal C: 69
Max test accuracy: 1.0

Note on BPFO/BPFI

The BPFO and BPFI values are obtained from the pump bearing type datasheet, namely type NTN Bearing 6201 which has BPFO coefficient of 2.62 and BPFI coefficient of 4.38.

Citation (Bibtex)

@ARTICLE{Atmaja2023,  
  author = {Atmaja, Bagus Tris and Ihsannur, Haris and Suyanto and Arifianto, Dhany},  
  title = {Lab-Scale Vibration Analysis Dataset and Baseline Methods for Machinery Fault Diagnosis with Machine Learning},  
  year = {2023},  
  journal = {Journal of Vibration Engineering and Technologies},  
  doi = {10.1007/s42417-023-00959-9},  
  type = {Article},  
  publication_stage = {Article in press},  
  source = {Scopus},  
}