A Sparse Mode Decomposition Method Based on Gaussian Window for Pulse Signals

A decomposition method called the improved sparse mode decomposition (ISMD) method is based on the sparse properties of frequency domain pulse signals. This method has a good performance of identifying the number of modes, and has a good application prospect for bearing fault diagnosis. However, the vast size of the real-world data set and the pulse signal's spectral distribution characteristics are not taken into account by the ISMD method. As a result, processing actual data by ISMD is incredibly sluggish and the decomposed signal has partially distortion. To overcome these limitations, a sparse mode decomposition method based on Gaussian window is proposed. Firstly, the least squares curve fitting in ISMD is replaced by the piecewise quantile regression method. Simultaneously, the Gaussian window function is used to replace the original filter, and on this basis, the parameters of the Gaussian function are determined by combining the energy entropy as the optimization index. Finally, an experimental project is studied and compared with the existing methods to confirm the feasibility and superiority of the proposed approach.