Application of slice bispectrum analysis to fault diagnosis of centrifugal pump

Bispectrum and slice bispectrum analysis techniques, which are excellent in non-Gaussian and nonlinear signal processing, were employed to analyze the vibration signals of a centrifugal pump under different running conditions. The results show that the bispectrum of the vibration signals have remarkably different features under the conditions of normal running, ground bolt loosening and cavitation. So, it can be used to classify the fault signals. By using the anti-diagonal slice attributes of bispectrum, the noise is suppressed markedly and the characteristic frequencies is well extracted. The examples of fault diagnosis indicate that the bispectrum and the slice bispectrum are effective in fault diagnosis of centrifugal pumps.