VMDPgram: An Effective Signal Decomposition Method for Rolling Bearing Fault Diagnosis

Due to the powerful signal decomposition performance, variational mode decomposition (VMD) is popularly used in numerous fields. Multiparameter optimization has always been the focus of VMD and has also plagued the development of its theory. In order to overcome this obstacle, a new concept of VMD packet (VMDP) is creatively proposed based on making full use of the excellent properties of VMD and combining the binary decomposition structure of wavelet packet transform (WPT). The idea of VMDP is to realize the signal decomposition mode of WPT based on VMD, whereby the original signal can be decomposed into a series of subintrinsic mode functions (sub-IMFs) of different layers, in which each VMD generates two intrinsic mode functions (IMFs), thus simplifying multiparameter optimization. In order to intuitively display each sub-IMF of different layers, the tiled structure of VMDP, namely, VMDPgram, is constructed. Moreover, a scheme based on envelope kurtosis as the sparsity measure to locate sub-IMF containing abundant fault information in VMDPgram is explored. The simulation and actual cases of bearing fault are used to verify the robustness of VMDPgram’s signal decomposition and fault detection, indicating the outstanding application potential of VMDPgram.