<abstract xmlns="http://www.w3.org/1999/xhtml"><p>Detection of incipient degradation demands extracting sensitive features accurately when signal-to-noise ratio (SNR) is very poor, which appears in most industrial environments. Vibration signals of rolling bearings are widely used for bearing fault diagnosis. In this paper, we propose a feature extraction method that combines Blind Source Separation (BSS) and Spectral Kurtosis (SK) to separate independent noise sources. Normal, and incipient fault signals from vibration tests of rolling bearings are processed. We studied 16 groups of vibration signals (which all display an increase in kurtosis) of incipient degradation after they are processed by a BSS filter. Compared with conventional kurtosis, theoretical studies of SK trends show that the SK levels vary with frequencies and some experimental studies show that SK trends of measured vibration signals of bearings vary with the amount and level of impulses in both vibration and noise signals due to bearing faults. It is found that the peak values of SK increase when vibration signals of incipient faults are processed by a BSS filter. This pre-processing by a BSS filter makes SK more sensitive to impulses caused by performance degradation of bearings.</p></abstract>

Detection of incipient degradation demands extracting sensitive features accurately when signal-to-noise ratio (SNR) is very poor, which appears in most industrial environments. Vibration signals of rolling bearings are widely used for bearing fault diagnosis. In this paper, we propose a feature extraction method that combines Blind Source Separation (BSS) and Spectral Kurtosis (SK) to separate independent noise sources. Normal, and incipient fault signals from vibration tests of rolling bearings are processed. We studied 16 groups of vibration signals (which all display an increase in kurtosis) of incipient degradation after they are processed by a BSS filter. Compared with conventional kurtosis, theoretical studies of SK trends show that the SK levels vary with frequencies and some experimental studies show that SK trends of measured vibration signals of bearings vary with the amount and level of impulses in both vibration and noise signals due to bearing faults. It is found that the peak values of SK increase when vibration signals of incipient faults are processed by a BSS filter. This pre-processing by a BSS filter makes SK more sensitive to impulses caused by performance degradation of bearings.

A Feature Extraction Method for Vibration Signal of Bearing Incipient Degradation

School of Mechanical Engineering, Southwest Jiaotong University, 111 Section One, North Second Ring Road, 610031, Chengdu, China

School of Transportation and Logistics, Southwest Jiaotong University, 111 Section One, North Second Ring Road, 610031, Chengdu, China

School of Engineering, University of Liverpool, the Quadrangle, L69 3GH, Liverpool, U.K.

Measurement Science Review

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Detection of incipient degradation demands extracting sensitive features accurately when signal-to-noise ratio (SNR) is very poor, which appears in most...

A Feature Extraction Method for Vibration Signal of Bearing Incipient Degradation

Online veröffentlicht: 11. Juni 2016

Seitenbereich: 149 - 159

Eingereicht: 05. Jan. 2016

Akzeptiert: 27. Mai 2016

DOI: https://doi.org/10.1515/msr-2016-0018

Schlüsselwörter
Feature extraction, incipient degradation, bearing, Blind Separation Source, Spectral Kurtosis

© 2016 Haifeng Huang et al., published by De Gruyter Open

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

A Feature Extraction Method for Vibration Signal of Bearing Incipient Degradation

Online veröffentlicht: 11. Juni 2016

Seitenbereich: 149 - 159

Eingereicht: 05. Jan. 2016

Akzeptiert: 27. Mai 2016

DOI: https://doi.org/10.1515/msr-2016-0018

SchlüsselwörterFeature extraction, incipient degradation, bearing, Blind Separation Source, Spectral Kurtosis

© 2016 Haifeng Huang et al., published by De Gruyter Open

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Schlüsselwörter
Feature extraction, incipient degradation, bearing, Blind Separation Source, Spectral Kurtosis