Reseach On The Application Of Spectral Kurtosis Algorithm Based On Lmd In The Rolling Bearing Fault Diagnosis

Rolling bearing plays an important rule in mechanical equipments. It is one of the most common sources of the equipment failture. Therefore, it is worthy to monitor and diagnosis the condition of rolling bearings. The complete rolling bearing fault diagnosis process is composed of signal acquisition, feature extraction, fault diagnosis(analysis), and decision-making intervention. The rolling bearing fault mechanism is studied in this paper. Rolling bearing fault vibration signals have the characteristics like non-stationary, modulation. So the local mean decomposition, a new time-frequency analysis method, is adopted into the rolling bearing fault diagnosis further combining with the spectral kurtosis. The main contents are as follows.Firstly, the local mean decomposition method is studied. The non-stationary nonlinear multi-component signals are decomposed into a series of PF component. The combination of the instantaneous amplitude and instantaneous frequency, which belong to the PF component, is the original signal time-frequency distribution. A new Matlab program was compiled in order to simulation and verification. The verification includes two different types of signals and some data from Case Western Reserve University.The filtering parameters could not be determined accurately only rely on historical data and human experience, when filtering analysis the rolling bearing fault vibration signal. So the spectral kurtosis algorithm based on local mean decomposition was put forward. The local mean decomposition method was carried out on the original signal time-frequency analysis. Based on the time-frequency distribution of signal, the signal is decomposed into several different frequencies. Then calculate the spectral kurtosis and draw the corresponding kurtosis figurewith Matlab. Based on kurtosis maximum principle, the best filter band was selected for FIR filting. Finally, the modulation information of the filted signal would be extracted after the e Envelope spectrum analysis. A new Matlab program was compiled in order to simulation and verification. The verification includes a multi-component signal and some data from Case Western Reserve University.In the precision machinery research institute rotating machinery fault detection laboratory, using the spectral kurtosis algorithm based on local mean decomposition, the rolling bearing inner ring and outer ring, rolling body fault and normal bearing are analyzed. The four experiments, succeeded in extracting the failure frequency of the roller bearing inner ring, outer ring and the roller, and the natural vibration frequency of normal runtime. The feasibility of using the spectral kurtosis algorithm based on local mean decomposition in the rolling bearing fault diagnosis was verified by those experiments.