Comparison of simulation methods applied to steel bridge reliability evaluations

Abstract

Steel bridges are in general subjected to fatigue deterioration and the structural reliability of bridges will thus reduce over time. There are multiple simulation-based procedures available to perform structural probabilistic studies with several classes of uncertainty taken into account. Since the crack propagation is highly nonlinear and the limit state function (LSF) is multi-dimensional, it imposes specific demands on the simulation methods. Monte Carlo simulation (MCS) has been widely applied in various of fields, however, it requires a great amount of samples and long computation time to reach a high level of accuracy. A more advanced method, Subset Simulation (SS), compensates this shortage. It calculates the product of conditional probabilities of several chosen intermediate failure events. In this paper, the performance of each method was evaluated and compared against fatigue deterioration for a selected bridge detail. A probabilistic model was defined and both prior and updated reliability estimation were performed. The results showed that SS is a good option to deal with fatigue problem with high nonlinearity and multi-dimensional LSF, and shows outstanding time efficiency compared to MCS to reach a comparable accuracy.