Noise Removal from the Vibration Signals of the Rotating Machinery Using the Empirical Wavelet Transform and the Conventional Thresholding Methods

In this paper, a new method is presented for removing the noise from the vibration signals of the rotating machinery based on the empirical wavelet transform (EWT) and the soft thresholding function. The EWT is a new signal processing method that decomposes each signal into its constituent components based on its frequency information. After decomposing each signal, the soft thresholding method is performed to empirical modes and the denoised signal is reconstructed. For evaluating the proposed denoising approach, this technique is used for detecting the bearing fault. For this purpose, the kurtosis factor and the envelope spectrum of each denoised signal are calculated for detecting the presence of fault and diagnosing the fault type, respectively. The results illustrate that the proposed technique increases the quality of the vibration signals so that the obtained kurtosis value is more sensitive to the presence of fault in the inner ring and the outer ring. On the other hand, the type of fault is diagnosed by observing the appeared frequencies in the envelope spectrum of signals denoised with EWT. The results show that the EWT-based denoising approach is superior to the empirical mode decomposition-based denoising method in the rotating machinery fault diagnosis procedure.