ABSTRACT
Predicting the performance and remaining service life of deteriorated reinforced concrete (RC) structures based on observational information obtained by means such as visual inspection and monitoring remains a very challenging task in the structural engineering community and the subject of this keynote paper. After experimentally studying the process of steel corrosion and the subsequential impact that corrosion has on the structural performance and integrity of deteriorated RC structures, it was determined that existing numerical simulation techniques need to be improved to reproduce the phenomena observed. With the application of machine learning techniques, the mechanical properties of concrete and rebar and their bonding over the entire RC structure should be determined quantitatively to replicate the observational information. In addition, new inspection technologies, such as those that use drones, must be utilized to obtain required observational information more efficiently. To address this challenging task, it is of vital importance to further integrate various advanced technologies and properly consider the uncertainties involved in the estimation and evaluation of each stage of deterioration and its propagation. In this study, a probabilistic framework for estimating the structural performance of corroded RC structures using observational information is presented using the associated core technologies in the literature, which include: (a) X-ray and digital image processing techniques to understand spatial steel corrosion, the product generated from corrosion and the associated crack width distributions, (b) the finite element method to simulate the structural performance and cracking behavior of corroded RC members, (c) the stochastic simulation technique to reproduce the characteristics of spatial steel corrosion variation, and (d) Bayesian updating and machine learning methods to estimate the degradation state from observational information as an inverse problem. For illustrative purposes, the effect of the observed corrosion-induced crack width distribution on the probability density function of the structural capacity of corroded RC beams is investigated based on the proposed framework.
