Abstract This paper thus proposes a new method combining Empirical Mode Decomposition (EMD) and Singular Value Decomposition (SVD) for bearing fault diagnosis. The method includes three steps. First, the signal is decomposed using EMD. Second, the instantaneous amplitudes are computed for each component using the Hilbert Transform (HT). Lastly, the Singular Value Vector is applied to the matrix of Cross-Power Spectral Density (CPSD) of the instantaneous amplitude matrix and the SVD versus frequency is analysed.  The proposed method is first validated by using various noisy simulated signals. The results show that the proposed method is robust versus the noise to detect the bearing frequencies that are representative of the defect even in a very noisy environment and that the amplitude of the first SVD at each bearing frequency is very sensitive to the defect severity. The proposed method is also applied to two different experimental cases with very low degradation. The results show that the proposed method is able to detect bearing defects at an early stage of degradation for both experimental cases.

Keywords: Bearing fault, Empirical Mode Decomposition (EMD), Hilbert transform (HT), Cross-Power Spectral Density (CPSD), Singular Value Decomposition (SVD).

References

Lei, Y. G., Lin, J., He, Z. J. and Zuo, M. J. “A review on empirical mode decomposition in fault diagnosis of rotating machinery”. Mechanical Systems and Signal Processing, vol. 35, no. 1-2, (2012): 108-126.

Cong, F., Chen, J., Dong, G. and Zhao, F. “Short-time matrix series based singular value decomposition for rolling bearing fault diagnosis”. Mechanical Systems and Signal Processing, vol. 34, no. 1 (2013): 218-230.