Reliability assessment of railway bridges designed for high-speed traffic: Modeling strategies and stochastic simulation

Due to the development and construction of new high-speed lines there is a high demand for reliable design and assessment of railway bridge constructions crossed by trains with high speed. Excessive vibrations of railway bridges lead to amplified deformations, stresses and accelerations of the structure and must be limited to prevent derailment and structural failure. Additional complexity of this reliability problem is caused by uncertainties that influence the resistance and the load level of the structure, such as material dispersion, irregular rail surfaces or local ballast settlements, undetermined and unpredictable energy dissipation mechanisms, and the environmental impact on the dynamic behavior of the bridge.

In contrast to the standard engineering design assessment based on partial safety factors, in this dissertation a probabilistic approach is used to quantify structural reliability of railway bridges, which are dynamically excited by crossing trains. Random variables describe the inherent uncertainties of the problem, and the compliance of performance criteria is checked by probabilistic numerical simulations, yielding estimates of the probability of failure of the train-bridge interaction system for assessing reliability.

A sophisticated numerical model is developed for the dynamic assessment of railway bridge structures, capable of taking into account vehicle-bridge interaction and track irregularities. By applying a substructure approach, which couples bridge and train models, different degrees of idealization can be implemented for the individual parts of the problem. For the mechanical description of the bridge structure simple beam models, and plane and three-dimensional finite element models are utilized. The crossing train is either idealized by concentrated forces at the train axles or by a multi-body system. The influence and validity of different degrees of sophistication of the numerical model on the resonance prediction is evaluated for deterministic example structures. The environmental impact on the bridge structure from seasonal changing temperatures and frost in subsoil and ballast is taken into account within a stochastic description. Based on these models in a probabilistic analysis the probability of failure due to acceleration thresholds is evaluated with simulation methods. Additionally to Monte Carlo simulations with crude direct and Latin Hypercube samples also more elaborated methods such as line sampling and subset simulation are applied to the reliability problem.

On a case study object the outcomes of the probabilistic assessment are compared to the standardized engineering design concept for dynamically loaded railway bridges based on European and Austrian building codes. Monte Carlo simulations provide a view on the scatter of the dynamic response and its variability. It is found that track irregularities significantly affect the variance and magnitude of the peak acceleration response. Although being conservative, a good agreement of the results of code based and stochastic analysis is found, leading to the conclusion that the applicable codes provide reasonable design rules.