Assessing the Capability of Analytical Carbonation Models to Propagate Uncertainties and Spatial Variability of Reinforced Concrete Structures

Abstract

Most of the approaches for diagnosis or prognosis of deteriorated reinforced concrete (RC) structures are based on two stages: acquiring data (concrete properties, quantitative degradation information), and then predicting the evolution of degradation by using appropriate models. Spatial variability of both properties and degradation processes cannot be neglected in the lifecycle assessment and implies that (i) data should be acquired for a representative part of the concrete surface and (ii) models should be capable of dealing with this variability. However, the assessment and modeling of spatial variability is not a straightforward task particularly when uncertainties affect the measurements or when the number of measurements is limited. The present paper aims at studying the capability of analytical carbonation models to deal with the spatial variability of model inputs in terms of spatial correlation of model outputs. Analytical models are considered herein because they provide practical and usual tools in engineering. This paper focuses on the case of a RC wall exposed to atmospheric carbonation where concrete properties and carbonation depths were measured by destructive techniques at several points over a linear portion of a wall within the framework of the French ANR EVADEOS project. Uncertainties due to experimental devices and procedures are estimated and propagated throughout random field models to account for spatial variability of spatial observations. Correspondence indexes are proposed to rank carbonation models with respect to their ability of reflecting the observed correlation profiles of carbonation depth. It was found that for the available database the proposed correspondence index that incorporates uncertainties was useful to assess the capabilities of models to deal with the spatial variability.