As hydroelectric generating units take on increasing responsibilities in peak load shaving and frequency regulation, they require more frequent start-up, shutdown, and load variation operations. Therefore, selecting an appropriate start-up strategy is essential for ensuring the safe and stable operation of the units. In this context, strain signals can effectively capture the vibration characteristics during the start-up process. The strain signal generated during the start-up of a hydroelectric generating unit exhibits nonlinear and non-stationary behavior, making the selection of an appropriate time-frequency analysis method critically important for accurate signal interpretation. Studies have shown that combining ridge line extraction with wavelet transform can effectively extract spectral features from such nonlinear and non-stationary signals. Based on fieldmeasured strain signals from key components of the turbine unit, the vibration characteristics under different start-up modes were evaluated. After denoising the acquired strain signals, the CWT-energy optimal ridge line algorithm was applied to extract the spectral features of the signals. The results demonstrate that this algorithm effectively captures the spectral characteristics of the strain signals, accurately reflecting the vibration variations of the unit under different start-up conditions.