Abstract Summary
High-speed railway networks in Europe include thousands of bridges, which constitute potential points of failure. Presently, the assessment of running safety is reflected in the Eurocode in the form of deck acceleration limits. However, this indirect metric does not relate unambiguously to an evaluation of reliability. In order to contribute to this discussion, it is necessary to perform probabilistic analysis on different scenarios of high-speed trains crossing bridges. Since whether dealing with track instability or derailment the target probabilities of failure are very low (in the order of 10^-5 to 10^-4), typical Monte Carlo trials demand a great deal of computation cost. This work presents a practical application of Subset Simulation to mitigate this issue. Uncertainty in geometrical, mechanical, and dynamic parameters of the bridges are represented by several random variables and different sources of lateral instability are considered, such as wind and rail irregularities. The optimization process includes the sensitivity analysis of different intermediate probability density functions and sample sizes. With this application it is possible to reduce the computation cost by 3 orders of magnitude, therefore facilitating future discussions of running safety.
