On-line bearing fault diagnosis based on signal analysis and rough set

Bearing defects are categorized as localized and distributed. For on-line bearing fault diagnosis, in this paper, the time-domain kurtosis calculation and the frequency domain wavelet analysis are used to extract the transitory features of non-stationary vibration signal produced by the bearing localized defects. To distributed defects, bearing fault diagnosis is built on the reducing decision based on rough set. This algorithm, making use of conditional entropy and the importance of it, without calculating the attribute core, get the optimization and minimum reduction set, and improves the on-line diagnosis speed and increases the fault diagnosis reliability. The feasibility and the robustness of this algorithm is demonstrated in a real-world application.