ABSTRACT
Corrosion damages reinforced concrete (RC) bridges primarily through the loss of steel rebar volume. Theoretical corrosion growth models are usually employed to predict this loss of steel over the life of a RC structure. Typically, these models fail to provide realistic estimates of the actual corrosion observed in the field. Alternatively, non-destructive testing/evaluation (NDT/E) or monitoring is used to evaluate the current state of a structure. However, either of these two approaches, alone, fails to provide a comprehensive understanding of how a bridge is expected to deteriorate over its lifetime. In order to be able to manage a degrading bridge over its service-life, it will be beneficial to scientifically combine the predictive capability of the theoretical degradation model and the data obtained from NDT/E for assessing the residual life of a structure. In this paper, this is achieved through a sequential Bayesian updating to address the issue of uncertainty present in the (i) physical process of corrosion and in the (ii) corrosion growth model used for prediction. The parameters of the updated corrosion growth model are stochastic in nature. A robust estimation of the time-varying steel loss due to corrosion is obtained incorporating these parameter uncertainties. The updated model is further used for the reliability assessment of a corroding RC bridge slab. Based on the target reliability, service-life of a bridge is obtained considering flexural failure.
