Skip to content

Paderborn

Paderborn University Bearing DataCenter.

Experimental bearing data sets for condition monitoring (CM) based on vibration and motor current signals, provided by the Chair of Design and Drive Technology, Paderborn University, Germany.

Description

The main characteristic of the data set are:

  • Synchronously measured motor currents and vibration signals with high resolution and sampling rate of 26 damaged bearing states and 6 undamaged (healthy) states for reference.
  • Supportive measurement of speed, torque, radial load, and temperature.
  • Four different operating conditions (see operating conditions).
  • 20 measurements of 4 seconds each for each setting, saved as a MatLab file with a name consisting of the code of the operating condition and the four-digit bearing code (e.g. N15_M07_F10_KA01_1.mat).
  • Systematic description of the bearing damage by uniform fact sheets and a measuring log, which can be downloaded with the data.

In total, experiments with 32 different bearing damages in ball bearings of type 6203 were performed:

  • Undamaged (healthy) bearings (6x), see Table 6 in (pdf).
  • Artificially damaged bearings (12x), see Table 4 in (pdf).
  • Bearings with real damages caused by accelerated lifetime tests, (14x) see Table 5 in (pdf)

Homepage

https://mb.uni-paderborn.de/kat/forschung/datacenter/bearing-datacenter

Original Dataset

  • Type of experiments: initiated faults & run-to-failure
  • Format: Matlab
  • Size: ~ 20.8 Gb, unzipped
  • Year of acquisition: 2015
  • Channels & Sampling rate:
    • motor current sampling frequency: 64 kHz, 2 channels ['phase_current_1', 'phase_current_2']
    • vibration sampling frequency: 64 kHz, 1 channel ['vibration_1']
    • sampling frequency of mechanic parameters (load force, load torque, speed): 4 kHz, 3 channels ['force', 'torque', 'speed']
    • temperature sampling frequency: 1 Hz, 1 channel ['temp_2_bearing_module']
  • Recording duration: 4 seconds
  • Load and Rotational Speed:
    • For undamaged and artificially damaged bearings test: not used
    • For lifetime test: 3800 N and 2900 rpm
  • Split: Healthy + Artificially damaged, Accelerated lifetime test

Download

https://mb.uni-paderborn.de/en/kat/main-research/datacenter/bearing-datacenter/data-sets-and-download

Built Dataset

  • Split: ['artificial', 'lifetime'] for initiated faults (merged with 'healthy') & run-to-failure experiments respectively.

Features

  • 'signal':
    • 'vibration': 1 channel
    • 'current': 2 channels ['phase_current_1', 'phase_current_2']
    • 'mechanic': 3 channels ['force', 'speed', 'torque']
    • 'temperature': 1 channel
  • 'sampling_rate':
    • 'vibration': 64 kHz
    • 'current': 64 kHz
    • 'mechanic': 4 kHz
    • 'temperature': 1 Hz
  • 'metadata':
    • 'FaultComponent': 'None', 'Inner Ring', 'Outer Ring', 'Inner Ring+Outer Ring'
    • 'FaultExtend': 0,1,2,3,
    • 'DamageMethod': 'Healthy', 'Aritificial', 'Lifetime'
    • 'FaultType': how fault is introduced
    • 'FileName': original filename.

References

  • Lessmeier, Christian; Kimotho, J. Kimotho; Zimmer, Detmar; Sextro, Walter: A Benchmark Data Set for Data-Driven Classification, European Conference of the Prognostics and Health Management Society (PHM Society), Bilbao (Spain), 05.-08.07.2016. (pdf)
  • ISO 15243:2017 Rolling bearings — Damage and failures — Terms, characteristics and causes https://www.iso.org/standard/59619.html

Notes

  • The file KA08/N15_M01_F10_KA08_2.mat cannot be loaded by scipy.io.loadmat.

Paderborn

Bases: tfds.core.GeneratorBasedBuilder

DatasetBuilder for Paderborn dataset.

Tutorial