Sparsity guided empirical wavelet transform for fault diagnosis of rolling element bearings. (15th February 2018)

• Record Type: Journal Article
• Authors: Wang, Dong; Zhao, Yang; Yi, Cai; Tsui, Kwok-Leung; Lin, Jianhui
• Is Part Of: Mechanical systems and signal processing. Volume 101(2018)
• Page Start: 292
• Page End: 308
• Publication Date: 2018-02-15
• Subjects: Empirical wavelet transform -- Sparsity -- Bearing fault diagnosis -- Fourier segments -- Multiple bearing defects -- Railway axle bearings; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2017.08.038
• Languages: English
• ISSNs: 0888-3270

Abstract

Graphical abstract: The proposed sparsity guided empirical wavelet transform for bearing fault diagnosis: (a) a temporal bearing fault signal; (b) Fourier spectrum of the bearing fault signal; (c) the temporal signal reprocessed by autoregressive model; (d) Fourier spectrum of figure (c); (e) identification of Fourier segments by using Steps 1–3 of the sparsity guided wavelet transform; (f) wavelet with Fourier support [ f l 1, f p ] ; (g) Fourier spectrum of the signal filtered by the wavelet in figure (f); (h) squared envelope spectrum of the signal filtered by the wavelet in figure (f); (i) wavelet with Fourier support [ f p, f h 1 ] ; (j) Fourier spectrum of the signal filtered by the wavelet in figure (i); (k) squared envelope spectrum of the signal filtered by the wavelet in figure (i); (l) wavelet with Fourier support [ f h 1, f h 2 ] ; (m) Fourier spectrum of the signal filtered by the wavelet in figure (l); (n) squared envelope spectrum of the signal filtered by the wavelet in figure (l). Highlights: Sparsity is introduced to guide empirical wavelet transform. Fourier segments required in empirical wavelet transform are automatically determined. Several resonant frequency bands are identified for bearing fault diagnosis. Single and multiple railway axle bearing defects are detected. Different resonant frequency bands are sensitive to different bearing defects. Abstract: Rolling element bearings are widely used in various industrial machines, such as electric motors, generators, pumps, gearboxes, railway axles, turbines, and helicopter transmissions. Fault diagnosis of rolling element bearings is beneficial to preventing any unexpected accident and reducing economic loss. In the past years, many bearing fault detection methods have been developed. Recently, a new adaptive signal processing method called empirical wavelet transform attracts much attention from readers and engineers and its applications to bearing fault diagnosis have been reported. The main problem of empirical wavelet transform is that Fourier segments required in empirical wavelet transform are strongly dependent on the local maxima of the amplitudes of the Fourier spectrum of a signal, which connotes that Fourier segments are not always reliable and effective if the Fourier spectrum of the signal is complicated and overwhelmed by heavy noises and other strong vibration components. In this paper, sparsity guided empirical wavelet transform is proposed to automatically establish Fourier segments required in empirical wavelet transform for fault diagnosis of rolling element bearings. Industrial bearing fault signals caused by single and multiple railway axle bearing defects are used to verify the effectiveness of the proposed sparsity guided empirical wavelet transform. Results show that the proposed method can automatically discover Fourier segments required in empirical wavelet transform and reveal single and multiple railway axle bearing defects. Besides, some comparisons with three popular signal processing methods including ensemble empirical mode decomposition, the fast kurtogram and the fast spectral correlation are conducted to highlight the superiority of the proposed method.