Method Research Of Weak Fault Feature Extraction Of Rolling Bearing Based On EEMD

The structure of mechanical equipment is very complicated, and each part is closely related to each other. Any part appear damage, will affect the normal operation of the machine. Bearing is one of the most easily damaged parts in rotating machinery. Therefore, study the effective bearing fault diagnosis methods, especially the method for extracting the feature of weak fault. It is of great significance to find and lift fault of these key parts timely.Rolling bearing fault signals, especially early weak fault signals are often submerged in the strong background noise and not easy to extract. Noise seriously affects the accuracy of the fault feature extraction. These vibration signal has characteristics of nonlinear and non-stationary. So from two aspects of reducing noise and nonlinear signal processing methods, the main research on the vibration signal of rolling bearing is as follows:Through theoretical and simulation analysis, it was founded that the EEMD had better anti aliasing effect than the EMD in the nonlinear signal processing aspect. The autocorrelation function could reduce noise and retain the original frequency components of the signal. Delay correlation demodulation had a great advantage in the early fault feature extraction.Through theoretical analysis, applied the on the optimization of IMF component.The experiment showed that the relevant kurtosis method is much better than the kurtosis method and correlation coefficient method.To obvious frequency modulation signal, put forward a based on EEMD and kurtosis of time delay correlation demodulation method to extract fault feature. Firstly make EEMD, use correlation kurtosis to choose the effective IMF to make up kurtosis method or correlation coefficient method. Then the delay correlation demodulation is performed on the reconstructed signal, which effectively avoids the noise from EEMD. Through the simulation analysis, the method is better than the method ofdirect demodulation after EEMD, it can accurately extract the fault frequency and frequency doubling to improve the accuracy of fault diagnosis. And also shows this method has advantage in the early fault feature extraction.To the signal modulation frequency is not obvious, put forward a based on EEMD and autocorrelation of marginal spectrum method. In this method, make EEMD firstly, and apply autocorrelated on the IMF that contain low frequency fault,then make marginal spectrum of the disposed IMF component. Through the simulation experiment, it is found that the method can not only remove the frequency band of the failure frequency, but also can be more accurate and effective than the direct EEMD for marginal spectrum analysis.Establish a set of fault diagnosis system, and analyzed the experimental data. A based on EEMD and kurtosis of time delay correlation demodulation method is much better than the direct demodulation based on EEMD, witch is verified. As well as, a based on EEMD and autocorrelation of marginal spectrum method is much accurate than the direct EEMD for marginal spectrum analysis. Both of these two methods have better effect in early fault diagnosis.