Nonstationary Vibration Signal Analysis of a Hydroturbine Based on Adaptive Chirplet Decomposition

Abstract

During transient processes, such as start-up and shut-down, vibration responses of hydroturbine rotors are typically frequency modulated signals. In this study, Gaussian chirplet is shown to be complete enough to match the characteristics of transient rotor vibration signals during speed varying processes. In order to avoid the amplitude distortion found in adaptive chirplet spectrograms, the Wigner—Ville distribution of normalized optimal chirplets is employed as a time—frequency window to filter the time—frequency distribution. The improved adaptive chirplet spectrogram possesses the best time—frequency resolution, and is free of cross term interference and amplitude distortion. Using this decomposition, the characteristic frequency components and their time evolution of main shaft vibration signals during the shut-down process of a hydroturbine are identified successfully. The vibration response from the upper part of the hydroturbine, which is the generator, is found to be closely related to rotor rotation, as expected. The signal is dominated by the rotational frequency of the rotor, and the frequency modulation phenomenon is the characteristic of a speed varying process. On the other hand, the vibration response from the lower part of the hydroturbine, which is the water turbine, is found to be closely related to hydraulic pressure fluctuation, and is directly influenced by the guide vane opening.