Research On The Improvement Of Empirical Mode Decomposition

Empirical mode decomposition(EMD) is an effective method to deal with nonlinear and non-stationary signals. The method decomposes the signal into a number of intrinsic mode function(IMF) based on the characteristics of its own signal.Therefore, it can make up the shortage of the short time Fourier, the wavelet decomposition and the Wigner-Ville distribution. At present, EMD is widely used in the field of mechanical fault diagnosis, biomedical signal analysis and communication signal analysis. In this paper, the theoretical system of empirical mode decomposition is studied, with the inherent shortcomings of the algorithm as a breakthrough.And research work is carried out on the end effect problem and the mode mixing problem in EMD, and the corresponding solutions are given. The research contents of this paper are as follows:(1)The commonly used time-frequency analysis methods of machine fault signal are introduced, such as short time Fourier transform, Wigner-Ville distribution and wavelet transform. The characteristics and disadvantages of these methods are summarized. Then, Hilbert-Huang transform(HHT) as well as the essence of the empirical mode decomposition(EMD) are introduced, and several problems of EMD algorithm which need to be solved, such as the mode mixing, end effect.(2) Using signal waveform monotonic consistency to dea l with the extreme value at the end point of the signal. Empirical mode decomposition is required to describe the upper and lower envelope of the signal through the extreme points, but it is difficult to estimate the maximum and minimum value of the signal at both ends which causes the envelope to be variable. Thus empirical mode decomposition process will produce the boundary error, with the decomposition of the boundary, the error will be propagated inward, so as to pollute the internal data and cause the decomposition result not reasonable. Through the analysis of several typical methods of suppression of end effect, the monotone method is introduced into EMD to obtain the end point of the signal, this method is simple and can effectively suppress the end effect.(3) A singular value decomposition denoising method is proposed to suppress the mode mixing problem in the EMD process and a EMD method based on energy separation(SEMD) is proposed to avoid the non event mode mixing phenomenon. The abnormal signal interval is selected by the method of adaptive mode mixing and overlapping first, and then using singular value decomposition to remove abnormal events which made that the abnormal events no longer so obvious and the signal envelope becomes more natural, which can effectively suppress the mode mixing phenomenon and improve the overall efficiency of EMD. Compared with the traditional EMD method, the improved method can effectively suppress problem of the mode mixing.Non event mode mixing are aggregated into high energy signal and low energy signal by singular value decomposition, and then decomposed by EMD method to get the IMF components. The simulation results show that the SEMD method can effectively separate the signal components, and compare d with the traditional EMD method, the SEMD method has obvious advantages.(4) Based on the above study, a fault identification method using the improved EMD method based on the Monotonic consistency and singular value decomposition(MSEMD)is proposed, and the MSEMD mthod is applied to analysis the Case Western Reserve University bearing data center data. And based on the comparison of the fault frequency identification result of the MSEMD and empirical mode decomposition, the conclusion is made that the MSEMD decomposition effect is superior to the empirical mode decomposition(EMD), and the bearing fault can be identified by the MSEMD method.