In signal processing, existing conventional indicators such as kurtosis, peak, margin, spectral kurtosis and other factors are usually very sensitive to accidental singularity on the signal data, which results in false judgments in bearing condition monitoring. To solve this problem, a spectrum band entropy (SBE) methods based on time-frequency is proposed. First, have a time-frequency transform to the signal, and then calculate the spectrum entropy of each frequency along the time axis, which is the SBE of the signal, and it can be used as a feature for bearing fault identification. SBE characterizes the complexity of frequency components change with time. The complexity is different for bearings’ normal state and fault state, and the performance of SBE is different too, so it can be used for bearings fault identification. At the same time, the data singularity caused by causal factors has little effect to the complexity of the frequency components change, SBE can automatically shield these effects, thereby reducing the false judgments. It’s proved to be effective for SBE applied in actual bearing failure identification. Compared with kurtosis, peak, spectral kurtosis and other indicators, SBE can effectively exclude interference from data, and identify the bearing state accurately.
WANG Xiao-ling;CHEN Jin;CONG Fei-yun.
Application of spectrum band entropy method in rolling bearings fault diagnosisbased on time-frequency analysis[J]. Journal of Vibration and Shock, 2012, 31(18): 29-33
PDF(1227 KB)
