Research on spectral kurtosis for the rolling bearing fault diagnosis

As usually done in rolling bearing fault diagnosis, a band-pass filter is used to improve the signal- To-noise ratio (SNR) before carrying out envelope demodulation of the fault signal. Spectral kurtosis (SK) can indicate not only the transients in non-stationary signal, but also their locations in the frequency domain. Thus, SK is widely used to identify the resonance frequency band of the fault signal. However, when using SK to analyze the practical fault signal of the rolling bearing, it is found that the estimated SK of the signal with sparse pulses is abnormal which leads to a wrong determination of the resonance frequency band. In this paper a method is proposed to avoid this problem. Moreover, based on SK, Time-Spectral Kurtosis (TSK) is proposed in this paper. Just as its name implies, TSK indicates the locations of the transient components in both frequency domain and time domain. A noise cancelling filter is built using TSK, with which the out-band and in-band noise in the fault signal can both be filtered out. The envelope demodulation results demonstrate the validity of the proposed method. Copyright