System Reliability of Bridge Structure Subjected to Chloride Ingress

Abstract

The parameters relevant for prediction of service lifetime with respect to chloride ingress are associated with large uncertainties. Full-scale measurements are in demand for conditions which are as homogeneous as possible. The present paper first summarizes statistical distributions which are obtained based on measurements from the Gimsøystraumen bridge in Norway. A large number of chloride profiles are available, and for each of these the diffusion coefficient and surface concentration (due to sea-spray) are estimated. Extensive measurements of concrete cover are also performed. These probability distributions are subsequently employed as input to a prediction model for chloride concentration at the steel reinforcement for a single but arbitrary position along the reinforment. Since the input parameters are represented in probabilistic terms, the chloride concentration is also a stochastic quantity. Furthermore, introducing the critical chloride concentration on a similar form, the probability of exceeding the critical threshold is determined as a function of time. In order to address chloride attack on the entire bridge, a system model with 90 components is next introduced. This model is employed in order to perform reliability updating based on observations at a number of sites along the bridge. First-order (FORM) reliability methods typically become inaccurate for large systems of this type. Crude Monte Carlo Simulation (which can be more accurate) will easily demand impractical efforts in terms of CPU-time, and a more efficient Monte Carlo simulation method is accordingly applied. It is shown that this typically reduces computation times by a factor of around 10.