Skip to content

Fraunhofer151

Fraunhofer_151 dataset.

Vibration Measurements on a Rotating Shaft at Different Unbalance Strengths

Description

This dataset contains vibration data recorded on a rotating drive train. This drive train consists of an electronically commutated DC motor and a shaft driven by it, which passes through a roller bearing. With the help of a 3D-printed holder, unbalances with different weights and different radii were attached to the shaft. Besides the strength of the unbalances, the rotation speed of the motor was also varied.

This dataset can be used to develop and test algorithms for the automatic detection of unbalances on drive trains. Datasets for 4 differently sized unbalances and for the unbalance-free case were recorded. The vibration data was recorded at a sampling rate of 4096 values per second. Datasets for development (ID "D[0-4]") as well as for evaluation (ID "E[0-4]") are available for each unbalance strength. The rotation speed was varied between approx. 630 and 2330 RPM in the development datasets and between approx. 1060 and 1900 RPM in the evaluation datasets. For each measurement of the development dataset there are approx. 107min of continuous measurement data available, for each measurement of the evaluation dataset 28min.

Details of the recorded measurements and the used unbalance strengths are documented in the README.md file.

Overview of the dataset components:

ID Radius [mm] Mass [g]
0D/ 0E - -
1D/ 1E 14 3.281
2D/ 2E 18.5 3.281
3D/ 3E 23 3.281
4D/ 4E 23 6.614

Homepage

https://fordatis.fraunhofer.de/handle/fordatis/151.2

Original Dataset

  • Type of experiments: labelled data.
  • Format: CSV files
  • Date of acquisition: 2020
  • Channels: voltage(?), measured rpm and 3 vibrations
  • Split: development and evaluation
  • Operating conditions: 630~2330 rpm for the development set and 1060~1900 rpm for the evaluation set, continuous record.
  • Sampling rate: 4096 Hz
  • Recording duration: 107 minutes for development set and 28 minutes for evaluation set.
  • Faults: 4 differently sized unbalances and unbalance-free case
  • Size: ~ 11 Gb, unzipped

Download

https://fordatis.fraunhofer.de/bitstream/fordatis/151.2/1/fraunhofer_eas_dataset_for_unbalance_detection_v1.zip

https://fordatis.fraunhofer.de/bitstream/fordatis/151.2/3/README.md

Built Dataset

Split: ['train', 'test'].

Features

  • 'signal': {'V_in', 'Measured_RPM', 'Vibration'},
  • 'sampling_rate: 4096,
  • 'metadata':
    • 'LoadRadius': radius of the load,
    • 'Label': ['Normal', 'Unbalanced']
    • 'LoadMass': mass of the load,
    • 'TrunkIndex': time index of the signal in the original file.
    • 'FileName': original file name,

Notes

The original record consists of two periods where the rotation speed increases linearly. Their durations are almost the same but depend on the split set. In the processed data these two periods are separated. Moreover, we truncate the few seconds (~10s) near the beginning and the end of each period which seem to correspond to initialization. The field TrunkIndex was added in metadata for record.

Tutorial