Stochastic service life prediction of reinforced concrete structures subjected to chloride-induced corrosion -

Abstract

A main reason behind reinforced concrete structural deterioration is chloride-induced corrosion. Once the critical chloride concentration is exceeded at the rebar level, the structure becomes susceptible to corrosion initiation. Corrosion propagates progressively, degrades the resistance capacity of the structure and decreases the design safety margin. To mitigate this risk, a continuous monitoring scheme, in conjunction with state estimation and parametric calibration, is suggested, thus yielding an accurate prediction of corrosion initiation and consequently efficient implementation of maintenance schedules. Chloride concentration measurements coupled with sequential data filtering are employed for health monitoring of the structure and for achieving more accurate reliability analysis. Several sources of uncertainties are identified once developing an accurate mathematical tool to simulate chloride ingress, like process noise of the physical model, parametric errors, boundary condition errors, and time independent sensors errors. These uncertainties limit the capability of traditional deterministic and probabilistic models to predict the corrosion initiation time in RC structures. Therefore, to accurately predict corrosion initiation time, this study presents a data assimilation framework that incorporates real time monitored chloride concentration and system dynamics in a probabilistic setting. As a first attempt, an Ensemble Kalman filter (EnKF) based framework is developed to try to predict the corrosion-free service life of reinforced concrete structures. The application of the EnKF scheme to 1-D and 2-D problems, using simulated measurements, demonstrate its ability to accurately estimate variability of the model parameters and consequently predict the temporal changes in the chloride concentration profile. However, once the EnKF is applied in real settings, where there is typically significant uncertainties and spatial variability, sampling errors become a concern that creates an unfavorable tradeof