An improved Autogram based on multi-indicator weighted fusion for rolling bearing fault diagnosis method

Aiming at the issue that the kurtosis index in Autogram is insufficient for the strong impact interference and the index is relatively single, which cannot accurately select the fault frequency band of the bearing, an improved Autogram based on multi-indicator weighted fusion is proposed in this article. Firstly, the reweighted kurtosis (RK) and spectral negative entropy (SNE) of each frequency band are computed, and a novel index is derived from weighted fusion of the two metrics by information entropy (IE), which can effectively overcome the inadequacy of individual index failing to integrate the impact and cyclic stability of the signal. Next, the novel indicator is utilized to choose the optimal demodulation frequency band, in which signal reconstruction and envelope demodulation analysis are conducted to extract the bearing fault information. Finally, through analyzing the data of single and compound faults of bearing, it is confirmed that the improved Autogram can exactly diagnose the fault types of the bearing, with better diagnostic effect than the comparative methods.