| description abstract | Variational mode decomposition (VMD) is a typical signal-processing method for fault identification in rolling bearing. In extracting fault characteristics from vibration signals of bearing with VMD algorithm, an inaccurate modal number will probably result in incorrect decomposition of signal and difficulty in identifying a fault. For that, first, the paper adaptively chose from component signals obtained by VMD according to the mean value of kurtosis. Second, according to chosen component signals, a new Weighted-kurtosis was built to adaptively determine the weight coefficient of chosen component signals. Third, rebuilding was implemented with weight coefficient and component signals to enhance fault features of the signal; meanwhile, concerning about the sensitivity of margin factor to impact features in the early stage of fault, margin factor of reconstructed signals was used to adaptively determine optimal modal number of VMD. Finally, compound faults of bearings were recognized by the spectrum of autocorrelation function (AF) of reconstructed signals corresponding to optimal modal number. The effectiveness of proposed method was validated by analyzing the vibration data of different compound fault types and sensor positions. The result has indicated that the proposed method is more effective than classical method to suppress noise interference, enhance fault features, and precisely identify the combined fault types of rolling bearings. | |
