Automatic instantaneous frequency order (IFO) extraction via integration strategy and multi-demodulation for bearing fault diagnosis under variable speed operation

• Published in: Journal of intelligent & fuzzy systems Vol. 34; no. 6; pp. 3547 - 3563
• Main Authors: Shi, Juanjuan, Jiang, Xingxing, Liang, Ming, Ding, Rongmei, Zhu, Zhongkui
• Format: Journal Article
• Language: English
• Published: London, England SAGE Publications 01.01.2018; Sage Publications Ltd
• Subjects: Bearing; Computer simulation; Demodulation; Expert systems; Fault detection; Fault diagnosis; Resampling; Search algorithms; integration strategy; Time-varying speed operation; bearing fault diagnosis; generalized demodulation; resampling-free method
• ISSN: 1064-1246; 1875-8967
• DOI: 10.3233/JIFS-169533

Bearing fault diagnosis under variable speed often faces two obstacles: a) blurry time frequency representation (TFR) and thus ambiguous and even unattainable instantaneous frequency (IF) for resampling, and b) complicated and error-prone resampling processes. To address such problems, this paper proposes a new tacholess and resampling-free method for bearing fault diagnosis under variable speed conditions. This method consists of two main steps: a) extract an accurate IF from the vibration data following a dual pre-IF integration strategy and a regional peak search algorithm to search the frequency bins point by point at local frequency regions, and b) with the accurate IF estimator (either shaft IF, instantaneous fault characteristic frequency (IFCF) or their harmonics), multi-demodulate the signal and superpose the resulting frequency spectra of all demodulated signal components using an order peak highlighting method. Then, the instantaneous frequency order (IFO) of signal components of interest contained in the original signal can be highlighted and the IFO spectra can be obtained for bearing fault diagnosis under variable speed conditions. In this manner, the bearing fault can be diagnosed without tachometer devices and resampling procedure. Therefore, the proposed method can substantially reduce human involvement and facilitate its implementation in a fault detection expert system. The effectiveness of the proposed method is validated using both simulated and experimental data.