An Improved Sparse Mode Decomposition Method for Pulse Signals

In signal processing field, mode decomposition is one of the most important branches. Many existing decomposition methods are mainly used to deal with narrow-band signals. If the analyzed signal is composed of bandwidth components, such as periodic pulse signals, traditional mode decomposition algorithms may have a performance deterioration. In order to overcome this problem, this paper proposes an improved sparse decomposition based on the group-sparse mode decomposition (GSMD) algorithm. The idea of the algorithm can be divided into two parts. First, the least square curve fitting technique is used to replace the average energy in GSMD algorithm with the fitted signal energy curve. Second, the bandwidth of each mode is adaptively reconstructed by the 3dB bandwidth criterion. The feasibility and superiority of the proposed method are verified by processing a set of actual bearing fault signal data and comparing with some existing methods.