Comprehensive uncertainty quantification of damage-based modeling of cracking in reinforced concrete structures

This work addresses the uncertainties inherent in civil engineering, arising from various sources such as the spatiotemporal variations in material properties, the complexity of concrete behavior, variations in applied loads and the impreciseness of theoretical models. This research aims to explore the advantages of accounting for some uncertain parameters in the numerical damage- based modeling of cracking in reinforced concrete structures, specifically investigating their impact on the accuracy and reliability of simulated outputs. By accounting for these uncertainties, the study aims to improve the predictive capability of numerical models, resulting in simulated responses that align more closely with observed on-site behavior. To achieve this aim, a case study involving a Representative Structural Volume (RSV) of a part of a 1450MWe nuclear power plant containment building is considered. It is based on the PACE- 1450 experimental campaign [1], which aims to thoroughly characterize cracking and air flow through these cracks for various tensile loads and temperatures. Experimental results show strong asymmetric cracking even though the applied tensile loads are mostly unidirectional using hydraulic jacks. To accurately represent the experimental distribution of the cracking pattern, several input parameters including material properties and boundary conditions during testing are considered as uncertain. The spatial variability of concrete properties is described using discretized random fields associated with the tensile strength [2]. In addition, uncertainties are considered in the angle of application of tensile loads applied via hydraulic jacks, which are suspected to deviate slightly from the surface normal rather than being perfectly perpendicular. These deviations may significantly influence the observed cracking behavior. First, a sensitivity analysis is performed to identify the most influential input parameters on the computed cracking patterns. Second, based on the obtained results, an attempt is made to determine the set of input parameters that replicate the observed cracking patterns. The obtained results highlight the critical need for comprehensive uncertainties quantification to objectively assess the results of complex experimental campaigns. They also indicate that accounting for random uncertainty alone is insufficient to replicate the experimental results. Indeed, considering the epistemic uncertainty is needed, particularly regarding the boundary and loading for a comprehensive consideration of uncertainties and better understanding of the observed behavior. 1 H. AL ELANI, D. BOUHJITI, L. JASON and B.