Fusing Vision-Dynamics Identification Method for the Wheel-Rail Maximum Adhesion Coefficient of Railway Vehicle Under Variable Rail Surface Condition
IEEE Sensors Journal · 15.04.2026 · 26 (8) · 11633-11645
DOI: 10.1109/jsen.2026.3669540 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 15.04.2026. Verfasst von Jinhai Wang und 5 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 43 Literaturangaben. Gefördert von: Beijing Natural Science Foundation.
Abstract (Originalsprache)
The wheel–rail adhesion coefficient plays a crucial role in the traction and braking control of railway vehicles. However, under dry or slippery rail surface conditions, the maximum adhesion coefficient exhibits significant fluctuations, posing challenges to reliable traction and braking control. To address this issue, this study proposes a fusion method that integrates vision and dynamics information for the identification of the maximum adhesion coefficient. First, an image classification network enhanced with a diffusion Fourier neural operator (DFNO) module is employed for real-time identification of rail surface conditions. Subsequently, a lattice Kalman filter (LKF) combined with a single-wheel vehicle model is employed to estimate the railway vehicle running speed. Considering that the sequence of rail surface images corresponds spatially and temporally to variations in the vehicle dynamics data, a spatiotemporal alignment mechanism is presented to synchronize the vision identification results with the vehicle running speed data. Finally, the maximum wheel–rail adhesion coefficient is calculated based on the aligned image recognition results, the estimated running speed, and empirical adhesion formulas. Simulation results demonstrate that the proposed method achieves high identification accuracy, with root mean square error (RMSE) = 0.0282 and mean absolute error (MAE) = 0.0175 across varying rail conditions. The method also exhibits strong adaptability to abrupt wet–dry transitions, showing only a short estimation delay before rapidly converging to the true adhesion curve. Furthermore, when image recognition remains stable, the estimated adhesion coefficient closely follows the ground truth, whereas consecutive recognition errors lead to noticeable degradation in estimation performance. These findings confirm that the proposed method effectively fuses vision and dynamics information, significantly improving the accuracy and robustness of maximum adhesion coefficient estimation under various operating scenarios, thereby providing strong support for traction and braking control of railway vehicles.
Förderung
- Beijing Natural Science Foundation
Zitieren
APA
Wang, J., Fan, M., Yang, J., Yao, D., Wen, Y., & Zhang, L. (2026). Fusing Vision-Dynamics Identification Method for the Wheel-Rail Maximum Adhesion Coefficient of Railway Vehicle Under Variable Rail Surface Condition. IEEE Sensors Journal, 26(8), 11633-11645. https://doi.org/10.1109/jsen.2026.3669540
Vancouver
Wang J, Fan M, Yang J, Yao D, Wen Y, Zhang L. Fusing Vision-Dynamics Identification Method for the Wheel-Rail Maximum Adhesion Coefficient of Railway Vehicle Under Variable Rail Surface Condition. IEEE Sensors Journal. 2026;26(8):11633-11645. doi:10.1109/jsen.2026.3669540
BibTeX
@article{Wang2026,
author = {Wang, Jinhai and Fan, Mengmeng and Yang, Jianwei and Yao, Dechen and Wen, Yilong and Zhang, Lei},
title = {Fusing Vision-Dynamics Identification Method for the Wheel-Rail Maximum Adhesion Coefficient of Railway Vehicle Under Variable Rail Surface Condition},
journal = {IEEE Sensors Journal},
year = {2026},
volume = {26},
number = {8},
pages = {11633--11645},
doi = {10.1109/jsen.2026.3669540}
}
EUVIMED