Rolling Bearing Fault Diagnosis Based On WPD-HHT

The roller bearing is not only one of the most ordinary in the rotating machinery, but also is liable to be damaged. It is of great realistic significance to study fault diagnosis technical for it. In the process of fault diagnosis, the key point is extracting fault feature from the fault vibration signals. And the modern signal processing technology used in the extracting process plays a central role.The traditional time-frequency analysis methods have some limitations for the non-stationary and non-linear signals. This paper briefly describes concepts and theories of the most commonly used time-frequency analysis method, then it discusses the scope of application of various methods and their advantages and disadvantages.Followed by the introduction of the instantaneous frequency, the intrinsic mode function, the Hilbert spectrum, the Hilbert Marginal spectrum and other basic concepts, it is described the empirical mode decomposition method, and verified that EMD has adaptability completeness and orthogonality. Hilbert-Huang Transform is a new non-linear, non-stationary time-frequency analysis method, and widely used in various fields. But in complex conditions, failure impulse noise is easily drowned by noise, and it is difficult to get accurate results by using the original HHT.According to the problems such as mode mixing and undesirable intrinsic mode functions at low frequency region EMD generated, it improves Hilbert-Huang Transform method by combining wavelet analysis and empirical mode decomposition. First, the signal is decomposed by WPD, and it is carried out the first layer filtering. Second, the signal is reconstructed after screening, and the reconstructed signal is then decomposed using EMD method. Third, it is carried out the second layer filtering, and finally the failure can be diagnosed according to Hilbert spectrum and Hilbert marginal spectrum. This method is used for simulation and experimental signals, and the results show that the method can extract a periodic pulse and suppress noise components.