Study On Fault Feature Extraction Of Rotating Machine Method Based On Demd Time-Frequency Analysis

Rotating machinery is the key equipment of electric power, metallurgy, machinery, aviation and some military industry. Whether can ensure the normal operation of the key equipment is directly related to all aspects of the development of an enterprise. With the development of modern industrial production, the mechanical fault diagnosis technology has been widely concerned. How to extract fault features from mechanical vibration signal and how to identify fault is the key of mechanical fault diagnosis, and signal processing and analysis is the most commonly method used in feature extraction method. In recent years, Empirical mode decomposition method as a major breakthrough in the analysis of non-stationary signal, is applied to analyze and process the rotating machinery vibration signal. However, EMD algorithm still exist deficiencies, which need further research to perfect. Differential-based empirical mode decomposition(DEMD) is an improvement method based on the theory of EMD. The proportion of different frequency components of the original signal is changed by differential, which is beneficial to extract the similar frequency or relatively weak high-frequency components. DEMD can effectively improve the mode aliasing problem of EMD.The article focuses on the time-frequency analysis method based on DEMD and its application in mechanical fault feature extraction.First, the instantaneous frequency, the single and multi-component signal, the basic concept of FM modulation signal is studied. The basic principle and algorithm of DEMD is analyzed. Then it is compared with empirical mode decomposition. The research results of the simulation show that the method of DEMD can extract the similar frequency or relatively weak high-frequency components and improve the mode aliasing problem of EMD effectively.Secondly, Aiming at the end effect of DEMD, a method combing support vector regression machine continuation with window function is proposed to inhibit DEMD endpoint effect. Using the constructed support vector regression model estimate the part of original signal of both ends and forecast the outboard, then the signal after continuation is conduct inner product operation with window function, which will suppress the end effect outside the original signal. Simulation and test show that the method can effectively restrain the end effect and improve the precision of DEMD. Aiming at the non-stationary and cyclical impact characteristics of vibration signals of rolling bearing, a method based on DEMD combining with symmetric difference energy operator demodulation is put forward. The Simulation and experimental results show that the method can accurately extract fault feature frequency of vibration signal and realize bearing fault diagnosis effectively.Lastly, based on the concept of fuzzy entropy, a comprehensive rotating machinery fault diagnosis method combining DEMD with fuzzy entropy and support vector machine(SVM) is proposed. Fault signal is decomposed by DEMD to obtain a certain number of intrinsic mode functions(IMFs) that have physical meaning. The fuzzy entropy of these IMFs are calculated and used as eigenvectors of fault signals, then the eigenvectors are put into SVM to identify the state and realize the identification and classification of the rotating machinery. By contrast, the method can be more accurately for the mechanical failure signal recognition and classification than the method based on EMD method combining with fuzzy entropy and the SVM.