Non-stationary Signal Analysis For Rolling Bearing Condition Monitoring And Fault Diagnosis

Rolling bearing is widely used in factory equipments, and is one of the mostimportant parts in machinery. The operating status of rolling bearing is the most importantfactor that if the machine can operate normally. Therefore it’s very valuable to monitor anddiagnose the operating status of rolling bearing.In The third chapter, from the perspective of time-frequency analysis aspact, weproposes the Harmonic Wavelet and Hilbert-Huang Transform (HHT) to diagnose thevibration signal of rolling bearing Harmonic wavelet has box-shaped spectrum feature thatit can decompose any signal to independent band orthogonally without leaks andredundancy, so harmonic wavelet can be used as band-pass filter. It can separate thespecific frequency band components from signal frequency components, in order to reducethe influence of other frequency band components to the specific frequency bandcomponents, so that the signal to noise ratio is raised, and the weak signal can be seen inthe noise. Hilbert-Huang transform method can do time-frequency analysis adaptively, anddecompose the local time-frequency characteristics of signals. The rolling bearing signal isseparated from high frequency interference noise after filtering. Then do HHTperformance on the signal, in order to get bearing fault frequency in Hilbert marginalspectrum. It is helpful to locate the bearing fault.In the fourth chapter, from the bearing vibration signal transient characterization andanalysis of non-stationary characteristics aspact, we build Fault Diagnosis Model oftransient signal feature extraction based on Smoothed Pseudo Wigner-Ville distribution(SPWVD) of spectral entropy and Support Vector Machine (SVM). SPWVD can showsignals intuitively and properly, and is a quadratic time-frequency distribution method. Itsuppressed cross-term function effectively by smoothing window, solved the disadvantageof Winger-Ville distribution. Us the support vector machine theory in bearing faultdiagnose field, we can realize the intelligent classification of rolling bearing operatingstatus. Combine Smoothed Pseudo Wigner-Ville distribution of spectral entropy theoryand support vector machine theory, and apply it to case study, research shows that supportvector machine intelligent classifier can distinguish the different operation status of rolling bearing, and the resolution is very high, and it is very suitable to be used in operationstatus diagnose of rolling bearing.The paper analysis the rolling bearing fault data by two differante method,above-mentioned method is verified by the analysis of bearing fault vibration signal ofCase Western Reserve University. The experiment shows that the above-mentionedmethod can diagnose the bearing fault effectively.