Computational Technology Resources - CCC

Keywords: railway bridge dynamics, dynamics, modal parameters, resonance, transient analyses, statistical analysis, predictive model, random forest algorithm, machine learning.

Abstract

This paper presents a comprehensive study on the dynamic response of railway bridges, based on a data collection of over 500 structures and 2000 train passages developed within the InBridge4EU project, funded by Europe's Rail Joint Undertaking under the Horizon Europe programme 2020-2027. Initially, the vertical accelerations due to traffic loads of 509 bridges are computed using numerical models with different levels of complexity. Then, through a statistical procedure the fundamental frequency and structural configuration are identified as the primary parameters responsible for the maximum deck vertical acceleration, highlighting simply-supported bridges as the most susceptible to resonance. While traditional Stepwise Linear Regression (SLR) provides a baseline, it fails to capture the inherent non-linearity of the dynamic problem and its dependence on its governing parameters and on their combinations, achieving coefficient of determination R^2 approximately 0.50. To overcome these limitations, an optimized Random Forest (RF) machine learning model is proposed, enhanced by physical feature engineering through the introduction of design parameters. The RF model significantly outperforms linear methods, reaching an R^2 = 0.88 and RMSE of 0.9312 m/s^2 on unseen data. Interpretability analysis through Partial Dependence Plots reveals that the model successfully reproduces the physical trigger mechanisms of resonance due to railway traffic. The predictive engine is further integrated into a design application that incorporates typology-specific RMSE offsets to provide conservative acceleration estimations to be used in the design stage. These results demonstrate that the proposed framework acts as an accurate surrogate for the preliminary and fast dynamic screening of railway bridges, optimizing resources by limiting complex simulations only to specific cases.

download the full-text of this paper (PDF, 11653 Kb)

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

	Computational & Technology Resources an online resource for computational, engineering & technology publications
	not logged in - login
Front Page Browse CCC CCP CSETS CTR IJRT Other Authors Search Purchase Guide FAQ Contact us	Civil-Comp Conferences ISSN 2753-3239 CCC: 15 PROCEEDINGS OF THE SEVENTH INTERNATIONAL CONFERENCE ON RAILWAY TECHNOLOGY: RESEARCH, DEVELOPMENT AND MAINTENANCE Edited by: J. Pombo Paper 2.4 Machine Learning-Based Prediction of Railway Bridge Vertical Accelerations P. Galvín¹, E. Moliner², M.D. Martínez-Rodrigo² and A. Romero¹ ¹Escuela Técnica Superior de Ingeniería, Universidad de Sevilla, Spain ²Department of Mechanical Engineering and Construction, Universitat Jaume I, Castellón, Spain doi:10.4203/ccc.15.2.4 Full Bibliographic Reference for this paper P. Galvín, E. Moliner, M.D. Martínez-Rodrigo, A. Romero, "Machine Learning-Based Prediction of Railway Bridge Vertical Accelerations", in J. Pombo, (Editor), "Proceedings of the Seventh International Conference on Railway Technology: Research, Development and Maintenance ", Civil-Comp Press, Edinburgh, UK, Online volume: CCC 15, Paper 2.4, 2026, doi:10.4203/ccc.15.2.4 Keywords: railway bridge dynamics, dynamics, modal parameters, resonance, transient analyses, statistical analysis, predictive model, random forest algorithm, machine learning. Abstract This paper presents a comprehensive study on the dynamic response of railway bridges, based on a data collection of over 500 structures and 2000 train passages developed within the InBridge4EU project, funded by Europe's Rail Joint Undertaking under the Horizon Europe programme 2020-2027. Initially, the vertical accelerations due to traffic loads of 509 bridges are computed using numerical models with different levels of complexity. Then, through a statistical procedure the fundamental frequency and structural configuration are identified as the primary parameters responsible for the maximum deck vertical acceleration, highlighting simply-supported bridges as the most susceptible to resonance. While traditional Stepwise Linear Regression (SLR) provides a baseline, it fails to capture the inherent non-linearity of the dynamic problem and its dependence on its governing parameters and on their combinations, achieving coefficient of determination R^2 approximately 0.50. To overcome these limitations, an optimized Random Forest (RF) machine learning model is proposed, enhanced by physical feature engineering through the introduction of design parameters. The RF model significantly outperforms linear methods, reaching an R^2 = 0.88 and RMSE of 0.9312 m/s^2 on unseen data. Interpretability analysis through Partial Dependence Plots reveals that the model successfully reproduces the physical trigger mechanisms of resonance due to railway traffic. The predictive engine is further integrated into a design application that incorporates typology-specific RMSE offsets to provide conservative acceleration estimations to be used in the design stage. These results demonstrate that the proposed framework acts as an accurate surrogate for the preliminary and fast dynamic screening of railway bridges, optimizing resources by limiting complex simulations only to specific cases. download the full-text of this paper (PDF, 11653 Kb) go to the previous paper go to the next paper return to the table of contents return to the volume description
Back to top	©Civil-Comp Limited 2023 - terms & conditions