Torsional vibration provides critical information for assessing the health of rotating shafts, whereas contact sensors can be difficult to deploy and may suffer from signal contamination in industrial environments. This study proposes a non-contact torsional vibration monitoring framework targeting High-Voltage Direct Current (HVDC) converter-station auxiliary drives, using a dual-beam differential laser Doppler vibrometer (LDV) and entropy-based lightweight learning. Two spatially separated points on the shaft are measured simultaneously and differentially processed to suppress common-mode rigid-body motion and enhance torsional sensitivity. The beat-frequency signal is demodulated to obtain an instantaneous frequency (IF) sequence. Variational mode decomposition (VMD) is then applied to isolate torsion-related intrinsic mode functions, from which bubble entropy features are extracted to form compact feature vectors. A particle swarm optimization–optimized extreme learning machine (PSO-ELM) is finally employed for supervised condition identification. Experiments on a rotating-shaft test rig representing typical auxiliary-drive operating states demonstrate an overall classification accuracy of 99.52%, outperforming a baseline ELM classifier (92.38%) under the same data partition. The results suggest that the proposed dual-beam differential LDV combined with entropy-based learning is feasible for non-contact torsional vibration monitoring and condition identification of rotating shafts in HVDC converter-station auxiliary-drive applications.
Ding et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: