Rolling Element Bearing Fault Investigation Based on Translation Invariant Wavelet Means Denoising and Empirical Mode Decomposition (EMD)

Abstract

The defected rolling element bearing produces a non-stationary vibration signal, and these vibration signals are generally immersed with heavy noise. When non-stationary and background noise vibration signals are observed, the feature frequency of defected bearings is challenging to extract. This research paper introduces a hybrid rolling element-bearing fault investigation approach. This hybrid method of fault investigation uses transient invariant (TI) denoising with a circular shift to diminish the background noise data from the original signal. Then purified vibration signal is obtained by eliminating the background noise from the original signal. After removing background noise, the adaptive method called the empirical mode decomposition (EMD) is applied to decompose a non-stationary signal. EMD decomposes the non-stationary signal into an intrinsic mode function (IMF), a stationary component. However, the selection of maximum energies IMFs is a challenging task. Correlation coefficient investigation is then applied to choose the IMFs having maximum energies IMFs and reject the other pseudo-low-frequency component of IMFs. Implement envelope spectrum investigation to extract fault feature frequency from selected IMFs. Numerical simulation and test data from the defected bearing validate the suggested hybrid method with inner race bearing, outer race bearing, and rolling element defects. The outcome shows that the TI denoising with cycle spin, the EMD, and the envelope spectrum is feasible and effective in detecting rolling element-bearing faults.