Evaluation of Modeling Assumptions for Predicting Structural Damage and Train Derailment Under Earthquake Loading

ABSTRACT The rapid development of new railway networks and the aging of existing infrastructure in seismic‐prone regions continue to motivate the need for efficient methods to simulate the dynamic behavior of coupled train track structure systems. While detailed train–structure interaction (TSI) models can capture complex mechanisms, they are often too computationally demanding for parametric studies across large networks and multiple earthquake scenarios, and they may obscure which parameters most strongly govern response. This paper presents a simplified TSI model with three components: a multibody dynamics train, a track model with an explicit wheel rail contact interface, and a bridge model represented as a cantilever column with a concentrated plastic hinge that captures stiffness and strength degradation and pinching, typical of reinforced concrete response. The framework is used to run suites of earthquake simulations to evaluate key modeling assumptions and their influence on predicted demands in both the train and the structure. Results show that derailment predictions correlate better with structural response measures at the track level than with conventional ground motion intensity measures, highlighting the relevance of TSI in running safety during earthquakes. In contrast, structural demands were largely insensitive to the presence of the train; for the bridge typology considered, decoupled analysis provides an adequate approximation for evaluating structural response. Based on the dynamic analyses, derailment fragility curves are developed, and the underlying workflow for structure‐specific derailment fragility estimation is described. The approach is demonstrated through a case study, including a corridor‐level application to illustrate the end‐to‐end process and the impact of modeling assumptions on predicted derailment likelihood. The model's simplicity enables rapid evaluation across a range of geometries, supporting future network scale risk assessment beyond single bridge simulations.