Non-destructive data assimilation as a tool to diagnose corrosion rate in reinforced concrete structures

Reinforcement corrosion is a major problem in the long-term management of reinforced concrete structures. With sustainability in perspective, knowledge of the corrosion rate (V-cor) makes it possible to estimate the kinetics of the corrosion phenomenon and helps in refining the maintenance strategy of such structures. Although in situ V-cor measurements are possible, data acquisition is time-consuming because of the protocol intrinsic to its measurement (reinforcement polarization made point by point). Therefore, in the context of site diagnostics, these methods cannot reasonably be used systematically on site and must be combined with high performance non-destructive testing (NDT) surface methods (GPR, capacimetry, half-cell corrosion). In addition, depending on the case, V-cor (point measurements) and NDT (surface) data are statistically related. However, there is a lack of efficient data assimilation tools permitting accurate translation of NDT data into pseudo V-cor data. In this paper, we present a numerical tool allowing prediction of V-cor values from NDT measurements. The tool permits application of different data assimilation techniques, i.e., cokriging, Bayesian sequential simulation, and a decision tree-driven learning depending on statistical behavior and available data. The efficiency of our numerical tool has been tested on a dataset acquired on a structure located in the French Alps. Results show that, for the case study, our data assimilation tool allows prediction of V-cor with accuracy compared to in situ measurements and also permits one to infer the uncertainty of the prediction. This opens the door for quantitative use of multiple NDT in the management of reinforced concrete structures.