Keywords: vibration fingerprinting, condition monitoring, railway vehicle, multibody simulation, physics-informed, suspension stiffness, track irregularities.

Sparse vibration fingerprints, constructed from physics-informed spectral peaks of short-time Fourier transform spectrograms, provide an interpretable representation for railway vehicle condition monitoring. The method eliminates amplitude dependence and emphasise vehicle eigenmodes through frequency-band weighting. Using multibody simulations with suspension stiffness and track quality variations, vibration fingerprints from carbody, bogie and axlebox vertical accelerations are evaluated across three experiments. A simple matching score achieves clear separation between baseline, vehicle-related and track-related conditions, with axlebox fingerprints showing the strongest discrimination capabilities and minimal cross-confusion between vehicle and track changes. Gradual stiffness variations are detected with sensitivity to the direction of change, particularly at wheelset level where approximate severity levels are also distinguished. Noise robustness analysis reveals sensor-dependent degradation: axlebox fingerprints develop track-change bias, bogie fingerprints remain unchanged, and carbody fingerprints gradually lose their ability to detect track changes and develop a bias towards vehicle change. The proposed fingerprinting algorithm thus proves to be a promising method for physics-informed condition monitoring that balances data sparsity, interpretability and accuracy.

download the full-text of this paper (PDF, 14 pages, 944 Kb)

go to the previous paper
go to the next paper
return to the table of contents
return to the volume description