An Improved Fast Spectral Kurtosis Algorithm in Bearing Fault Diagnosis for a Motor Drive System

Bearing is one of the most easily damaged parts in the electric drive system of electric vehicles. With the silicon carbide motor controller put into use, higher switching frequency produces greater shaft current, which makes the bearing more prone to failure. The accurate diagnosis of initial bearing faults is of great significance for prolonging the service life of electric drive system and improving the operation reliability of equipment. The fast spectral kurtosis algorithm can screen out the center frequency and bandwidth to band-pass filter the bearing vibration signal, and then the envelope analysis technology can be used to identify the bearing fault characteristic frequency, so as to determine whether the bearing fault exists and the fault type. However, due to the presence of transient noise in the system, the fast spectral kurtosis algorithm can sometimes be affected by the transient noise and cannot identify the appropriate center frequency and bandwidth. An improved fast spectral kurtosis algorithm is proposed in this paper, that is, a pre-whitening process is added before the traditional fast spectral kurtosis algorithm, and this method is applied to the bearing public data set. The experimental results show that the improved fast spectral kurtosis algorithm is superior to the conventional spectral kurtosis algorithm and manual empirical method, and its diagnosis results are more robust.