metadata

dataset: Wind-Turbine-5MW-Drivetrain-Bearing-Damage-Dataset
language:
  - en
tags:
  - bearing-fault
  - drivetrain
  - condition-monitoring
  - vibration-signal
  - wind-turbine
  - fault-diagnosis
license: cc-by-4.0
dataset_info:
  features:
    - name: signal_a1a
      sequence:
        sequence: float32
    - name: signal_a3a
      sequence:
        sequence: float32
    - name: signal_length
      dtype: int32
    - name: label
      dtype:
        class_label:
          names:
            '0': healthy
            '1': main_bearing
            '2': high_speed_bearing
            '3': low_speed_planet_bearing
    - name: fault_condition
      dtype: string
    - name: environmental_condition
      dtype: string
    - name: sampling_frequency
      dtype: float32
    - name: wind_speed
      dtype: float32
  splits:
    - name: train
      num_bytes: 69120747
      num_examples: 12
  download_size: 70324111
  dataset_size: 69120747
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
pretty_name: 5MW offshore wind turbine drivetrain bearing damage dataset

5MW Wind Turbine Drivetrain Bearing Damage Dataset

Dataset Description

Dataset containing vibration signals from the NREL 5MW reference drivetrain model with different bearing conditions. The reference paper for this dataset is available at: DOI: 10.1016/j.renene.2022.12.049 This is a subset of the original dataset available at Zenodo. For the full dataset and additional information, please visit: DOI: 10.5281/zenodo.7674842

Features:

signal_a1a: Vibration time waveform from sensor A1A
signal_a3a: Vibration time waveform from sensor A3A
signal_length: Length of the signals
label: Fault condition (0: healthy, 1: main bearing, 2: high-speed bearing, 3: low-speed planet bearing)
environmental_condition: Operating conditions (EC1, EC2, EC3)
sampling_frequency: 200.0 Hz
wind_speed: Operating wind speeds (7.0, 12.0, 14.0 m/s)

Data Split

The dataset includes samples from:

4 different conditions (healthy + 3 fault locations)
3 different operating conditions (EC1, EC2, EC3)
Total of 12 different scenarios

Implementation

https://github.com/alidi24/deep-learning-fault-diagnosis

License

This dataset is licensed under CC BY 4.0.
This means:

✅ You are free to share and adapt the data
✅ You must give appropriate credit
✅ You can use it for any purpose, including commercial use

Citation

In case using this dataset for your research, you are welcome to cite the paper:

@article{DIBAJ2023161,
title = {Fault detection of offshore wind turbine drivetrains in different environmental conditions through optimal selection of vibration measurements},
journal = {Renewable Energy},
volume = {203},
pages = {161-176},
year = {2023},
issn = {0960-1481},
doi = {https://doi.org/10.1016/j.renene.2022.12.049},
url = {https://www.sciencedirect.com/science/article/pii/S0960148122018341},
author = {Ali Dibaj and Zhen Gao and Amir R. Nejad},
keywords = {Offshore wind turbine, Drivetrain system, Fault detection, Optimal vibration measurements},
}