0
Research Papers

Fault Diagnosis of Rolling Bearings Using Data Mining Techniques and Boosting

[+] Author and Article Information
Muhammet Unal

Technology Faculty,
Mechatronic Engineering,
Marmara University,
G207, Kadikoy,
Istanbul 34722, Turkey
e-mail: munal@marmara.edu.tr

Yusuf Sahin

Engineering Faculty,
Electrical and Electronics Engineering,
Marmara University,
Kadikoy,
Istanbul 34722, Turkey
e-mail: ysahin@marmara.edu.tr

Mustafa Onat

Engineering Faculty,
Electrical & Electronics Engineering,
Marmara University,
Kadikoy,
Istanbul 34722, Turkey
e-mail: monat@marmara.edu.tr

Mustafa Demetgul

Technology Faculty,
Mechatronic Engineering,
Marmara University,
G204, Kadikoy,
Istanbul 34722, Turkey
e-mail: mdemetgul@marmara.edu.tr

Haluk Kucuk

Technology Faculty,
Mechatronic Engineering,
Marmara University,
Kadikoy,
Istanbul 34722, Turkey
e-mail: halukkucuk@marmara.edu.tr

1Corresponding author.

Contributed by the Dynamic Systems Division of ASME for publication in the JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CONTROL. Manuscript received November 7, 2014; final manuscript received August 24, 2016; published online November 8, 2016. Editor: Joseph Beaman.

J. Dyn. Sys., Meas., Control 139(2), 021003 (Nov 08, 2016) (12 pages) Paper No: DS-14-1459; doi: 10.1115/1.4034604 History: Received November 07, 2014; Revised August 24, 2016

Rolling bearings are key components in most mechanical facilities; hence, the diagnosis of their faults is very important in predictive maintenance. Up to date, vibration analysis has been widely used for fault diagnosis in practice. However, acoustic analysis is still a novel approach. In this study, acoustic analysis with classification is used for fault diagnosis of rolling bearings. First, Hilbert transform (HT) and power spectral density (PSD) are used to extract features from the original sound signal. Then, decision tree algorithm C5.0, support vector machines (SVMs) and the ensemble method boosting are used to build models to classify the instances for three different classification tasks. Performances of the classifiers are compared w.r.t. accuracy and receiver operating characteristic (ROC) curves. Although C5.0 and SVM show comparable performances, C5.0 with boosting classifier indicates the highest performance and perfectly discriminates normal instances from the faulty ones in each task. The defect sizes to create faults used in this study are notably small compared to previous studies. Moreover, fault diagnosis is done for rolling bearings operating at different loading conditions and speeds. Furthermore, one of the classification tasks incorporates diagnosis of five states including four different faults. Thus, these models, due to their high performance in classifying multiple defect scenarios having different loading conditions and speeds, can be readily implemented and applied to real-life situations to detect and classify even incipient faults of rolling bearings of any rotating machinery.

Copyright © 2017 by ASME
Your Session has timed out. Please sign back in to continue.

References

Figures

Grahic Jump Location
Fig. 1

Experimental setup and test rig

Grahic Jump Location
Fig. 2

Cylindrical test rolling bearings: (a) FAG N208-E-TVP2 and (b) FAG NU208-E-TVP2

Grahic Jump Location
Fig. 3

Inner and outer race defect: (a) 05x05 ORD and (b) 05x05 IRD

Grahic Jump Location
Fig. 4

Design of the fault diagnosis process

Grahic Jump Location
Fig. 5

Raw sound data under 3 kN loads: (a) Fault Free IR, (b) 03x03 IRD, (c) 05x05 IRD, (d) Fault Free OR, (e) 03x03 ORD, and (f) 05x05 ORD

Grahic Jump Location
Fig. 7

Peak-to-peak values of raw sound data at different loads: (a) 2 kN, (b) 2 kN, (c) 3 kN, (d) 3 kN, (e) 4 kN, and (f) 4 kN

Grahic Jump Location
Fig. 8

RMS values of raw sound data at different loads: (a) 2 kN, (b) 2 kN, (c) 3 kN, (d) 3 kN, (e) 4 kN, and (f) 4 kN

Grahic Jump Location
Fig. 9

Schematic of envelope detector and its spectrum: (a) sound signal, band-pass filtered signal and its envelope of 03x03 IRD (speed: 3000 rpm, load: 300 kg), (b) spectrum of sound signal, and (c) spectrum of envelope detector

Grahic Jump Location
Fig. 10

Feature extraction process

Grahic Jump Location
Fig. 11

Selected frequencies and amplitudes as features (speed: 3000 rpm, load: 300 kg): (a) IRD signal and (b) ORD signal

Grahic Jump Location
Fig. 12

Frequency components and corresponding harmonics of sound signal at 3 kN; (a) IRD frequencies after ED and FFT, (b) IRD frequencies after ED and PSD, (c) ORD frequencies after ED and FFT, and (d) ORD frequencies after ED and PSD

Grahic Jump Location
Fig. 13

Classification methods: (a) decision tree and (b) support vector machine

Grahic Jump Location
Fig. 14

Performance of C5.0 with boosting classifier w.r.t. ROC curves: (a) normal–faulty, (b) normal–IRD–ORD, and (c) all classes

Grahic Jump Location
Fig. 15

Performance of C5.0 classifier w.r.t. ROC curves: (a) normal–faulty, (b) normal–IRD–ORD, and (c) all classes

Grahic Jump Location
Fig. 16

Performance of SWM classifier w.r.t. ROC curves: (a) normal–faulty, (b) normal–IRD–ORD, and (c) all classes

Grahic Jump Location
Fig. 17

Confusion matrixes of classification of normal–faulty: (a) C5.0 with boosting, (b) C5.0, and (c) SVM (linear)

Grahic Jump Location
Fig. 18

Confusion matrixes of classification of normal–IRD–ORD: (a) C5.0 with boosting, (b) C5.0, and (c) SVM (linear)

Grahic Jump Location
Fig. 19

Confusion matrixes of classification of all classes: (a) C5.0 with boosting, (b) C5.0, and (c) SVM (linear)

Tables

Errata

Discussions

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In