Comparison and Selection of Wavelets for Vibration Signals Denoising and Fault Detection of Rotating Machines using Neighborhood Correlation of SWT Coefficients

The vabartion signals carries the dynamic informations of the machine, which is useful for fault identification of the machine. The practical raw signals suffer from heavy noise, it is essential to retrieve the original signal from noisy raw signals. The signal processing techniques are very effective in such situations. The presence of noise in vibration signal is so large that its elimination results in information loss. In the proposed work noisy vibration signals corrupted by AWGN noise of faulty gear box faulty gear box are denoised using many wavelets and the results are compared. Stationary Wavelet Transform (SWT) is time invariant transform due to this property it is free from the aliasing problem. In this work SWT has been employed for vibration signal denoising and the results have been compared to the discrete wavelet Transfrom(DWT). The fault detection becomes easy for denoised signal and the analysis of its spectrum predicts the actual condition of the machine. The denoised results are compared on the basis of kurtosis.