Probabilistic structural analysis of a real-life corroding concrete bridge girder incorporating stochastic material and damage variables in a finite element approach

Abstract

A combined probabilistic approach based on response surface sampling (RSS) and importance sampling (IS) is developed to efficient assess the structural reliability of corroding concrete structures. In this approach, the failure probability and the corresponding reliability index can be calculated for any stochastic system, independent of the material and model complexity or the limit state function(s). The combined approach is applied to a real-life example where a corroding post-tensioned concrete bridge girder is considered. This bridge girder is modelled in a dedicated finite element analysis to assess the structural reliability under corrosion damage. Serviceability and ultimate limit state requirements are taken into account and the material and damage properties are the stochastic variables. A converged solution is found in 8 and 128 limit state function evaluations for the RSS and IS, respectively. The importance factors for the different variables are calculated. The full analysis highlights non-important parameters (strength of the traditional reinforcement) and indicates that the ultimate limit state of the girder is the limiting factor in all analyses. Further computational optimizations are proposed and implemented, highlighting the capabilities of the reduced model and the robustness of the proposed approach.