With the increase of hydropower units and installed capacity in China,the fault problem during the transient process occurs frequently.The signal analysis method is widely used to deal with the problem.The variational mode decomposition (VMD) method,a relatively new self-adaptive method for processing non-linear and non-stationary signals,was adopted to investigate the vibration signals during the transient process of a hydropower unit in the Gezhouba power plant.The first step was to use VMD to analyze the simulated vibration signals of the transient process,and to verify the effectiveness of the VMD method in dealing with non-linear and unstable signals.The Hilbert-Huang transform (HHT) was then used to calculate the instantaneous frequency of each intrinsic mode functions (IMF) decomposed by VMD to obtain the corresponding Hilbert spectrum.These results were compared with those from empirical mode decomposition (EMD) method.From the final vibration signal analysis,the IMF decomposed by VMD could be divided into three parts.The first part was the trend of signal development.The second part was the basic decomposition of the signal with a strong regularity of "spindle" signal reflecting the change law of the transient process.The third part was high-frequency characteristic signals mainly caused by interference noise or unit failure.The corresponding Hilbert spectrum had the characteristics of small fluctuation and stable amplitude.The results showed good consistency between the frequency change of each component and the time-dependent rule of the rotational speed.This suggested that VMD was capable of extracting characteristic frequencies to facilitate the analysis of vibration characteristics.Compared with EMD,the VMD outperformed in self-adaptivity.It provided more accurate and effective analysis results and a better representation of the vibration rule during the transient process of the hydropower unit.