Under long-term service conditions, vibration signals of water-lubricated stern bearings exhibit strong nonlinearity, nonstationarity, and multicomponent coupling, which makes accurate feature extraction challenging. To address this issue, this study proposes a progressive EEMD-TFMST-based analysis framework that combines spectral localization, adaptive signal decomposition, noise suppression, and high-resolution time–frequency characterization. Rotational-speed tests and long-duration wear tests were conducted using an SSB-100 test rig, and the lubrication regimes were identified based on friction coefficient variations. The results show that the dominant vibration features are strongly dependent on the lubrication regime and wear stage. With increasing rotational speed, the vibration response evolves from isolated peaks near 400 and 600 Hz under boundary lubrication to enhanced 300–400 Hz components under mixed lubrication, and further to broadband responses within 0–1000 Hz under hydrodynamic lubrication, with dominant peaks mainly concentrated in the 300–500 Hz range. With increasing rotational speed, the lubrication regime gradually changes from boundary lubrication to hydrodynamic lubrication, accompanied by a transition of vibration energy from single-IMF concentration to broadband distribution across multiple IMF components. Long-term operation induces stage-dependent changes in lubrication and vibration behavior: moderate wear improves vibration stability, whereas excessive wear deteriorates lubrication, increases the proportion of mixed lubrication, and promotes energy migration toward lower frequencies with additional high-frequency excitation. Under prolonged high-speed operation, lubrication degradation further induces broadband vibration. The proposed method enables accurate quantification of vibration features and provides a useful basis for service-performance evaluation and early fault warning of water-lubricated stern bearings.
Liu et al. (Wed,) studied this question.