Fatigue assessment of railway noise barriers based on field monitoring, ML-driven stress prediction and numerical modeling

Abstract

Railway noise barriers are frequently subjected to train-induced aerodynamic loads during long-term service, which may lead to fatigue problems at the welded base of steel posts. Conventional fatigue assessment methods often fail to account for the complex operational and environmental conditions experienced by real structures. To address this issue, this study proposes a fatigue evaluation framework that integrates long-term field monitoring data with machine learning (ML) and numerical simulation. Based on a field test database, a data-driven calculation model was developed to predict the stress responses of barrier steel posts under actual operational and environmental conditions. A finite element (FE) model validated by field data was further used to establish a quantitative relationship between the monitored stress responses and the maximum stress at the steel post base. Fatigue life for weld toe and weld root failure modes was evaluated using Palmgren-Miner linear damage rule and the nominal stress method in the Eurocode. Results show that weld root failure governs the fatigue performance of the steel post base, with significantly shorter fatigue lives than weld toe failure. Nose waves dominate fatigue damage, while tail waves still contribute about 14–23%. The fatigue life is also strongly influenced by train operating speeds and train type distribution, highlighting the importance of considering realistic train operating conditions in fatigue assessment and design. The proposed framework provides a practical and scalable tool for fatigue assessment and lifecycle maintenance planning of railway noise barrier systems. © 2026