Higher order spectral analysis for vibration signals of the large steam turbine in slow-down process

Shaft vibration of steam turbine usually has characteristics of strong harmonics and typical sub-Gaussian distribution. In the process of startup and shutdown of units, there are always some changes of harmonic components with nonlinear coupling reciprocity each other. That makes the technology of power spectral analysis based on second-order statistics restricted in application. This paper made a deep research on the measured vibration signals of a large steam turbine in its slow-down process. Used traditional methods and higher order spectral analysis (HOSA) methods separately to analyze the shaft vibration signals and bearing vibration signals. And it clarified the characteristic of nonlinear harmonic coupling between the fundamental frequency of shaft rotation and its harmonics. The result shows that bispectrum and 1(1/2) dimensional spectrum based on third-order cumulant can clearly reveal the large changes of higher order spectral characteristics when the unit is under the varying operating conditions. And they can also reveal the changes of non-Gaussian and nonlinear characteristics for the vibration signals. HOSA has a high sensitivity and can effectively restrain the Gaussian noise of the measured vibration signals. So it builds a good new foundation for feature extraction and fault diagnosis.