Study On Mechanical Fault Diagnosis Methods Based On Local Mean Decompostion

This dissertation is supported by the National Natural Science Foundation of China (No.:50775208) and Natural Science Foundation of Education Department of Henan Province (No.:2006460005,2008C460003). This paper introduces Local Mean Decomposition (LMD) into the field of mechanical fault diagnosis,and has made thorough research in the mechanical fault diagnosis method based on LMD,and achieved some innovations.The main contents of this paper are as follows:In chapter one, the subject is proposed, and its significance is discussed. Then the development and application status of time-frequency analysis are recited, the research and application status of LMD are described. At last the general structure and innovations of this thesis are proposed.In chapter two, the basic theories and algorithm of LMD are discussed in detail,and compared with EMD in theory.LMD can decompose original signal into a series of production function(PF),which has obvious physical meaning and can reveal the system's nonlinearity and nonstationarity better. Because of the good adaptability of LMD,it is introduced into mechanical fault diagnosis.A mechanical fault diagnosis method based on LMD is proposed,and its major steps are provided.The simulation results show that the proposed method can handle nonstationary signal adaptivly,and effectively extract the instantaneous frequency components which has obvious physical meaning, will not make information lost.Finally,the proposed method is applied to the rolling bearing fault diagnosis,the results further show that this method is effective.In chapter three, the definition and algorithms of WHOS are described. To suppress the cross-term effect of WHOS,we propose a mechanical fault diagnosis method based on LMD and WHOS.In this method, firstly, original signal is decomposed into a series of PF which is a single component. Then the Wigner Bispectrums of the real components are obtained and superposed, as a result, the cross-terms of WHOS can be inhibited effectively.Through simulation study, the proposed method is compared with direct Wigner Bispectrums,the result shows that the cross-term effect is effectively inhibited,and the proposed method keeps the nature of the original signal exactly. Compared with Choi-Williams kernel filter Wigner Bispectrums, the proposed method.can show the time-frequency characteristics more truly. At last, this method is successfully applied to the rolling bearing fault diagnosis. The experiment result further verifies the effectivity of the proposed method, which offers a new way for the cross-term suppression of WHOS.In chapter four, the basic theory and algorithm of BSS are introduced. Combining the features of LMD with BSS, an underdetermined blind source separation method of mechanical fault diagnosis based on LMD is proposed. In this method, the original signal is decomposed into a series of PF firstly, then all the PFs components and the original signal are combined into a new observation signal, further through whitening and joint approximate diagonalization (JADE), estimated source signals are acquired. The simulation result shows that this method is superior to the traditional blind source separation method which is based on time-frequency distribution,can process the underdetermined blind source separation of non-stationary signal effectively.Finally, the proposed method is applied to faults separation of rolling bearing, and the experiment result further verifies the effectivity of this method.In chapter five,combining LMD, Envelope analysis and Support vector machine, we propose a mechanical fault diagnosis method based on LMD and SVM. In this method, the original signal is decomposed into a series of PF firstly, then envelope analysis is processed on the previous few PFs, further from envelope spectrum, characteristic amplitude ratios are extracted as feature vectors to input into SVM for recognition. The experiment result shows that the proposed method is effective for the faults recognition of rolling bearing.In the last chapter, the conclusions of the dissertation are summarized. Future research of LMD is prospected.