Statistical modelling of cracking in large concrete structures under Thermo-Hydro-Mechanical loads: Application to Nuclear Containment Buildings. Part 2: Sensitivity analysis

Highlights: • 1st order sensitivity analysis of the suggested THM model's response using the OFAT method. • Analysis of the influence of statistical size effects and the THM inputs' intrinsic variation on concrete cracking. • Definition of the most influential parameters with regards to concrete cracking and its ageing in time. • Approximation of the variation domains of various THM inputs based on in situ observations of a Nuclear Containment Building. - Abstract: As demonstrated in Part I of this contribution, the precise and full prediction of the cracking patterns and concrete's global behaviour in ageing structures is a complex task. Though the suggested modelling strategy allows the prediction of the main cracking patterns, its drawbacks are mainly related to (a) the use of the so-called Statistical Size Effect Law requiring various random field realizations (more than 30) and (b) the identification of a consequent number of Thermo-Hydro-Mechanical parameters (more than 50) which variations may also affect the computed cracking patterns and ageing behaviour. In part II of this contribution, the aim is to evaluate the effect of such uncertainty-related variations on concrete's early age and long term behaviours (in comparison with the effect of the intrinsic spatial variation of mechanical properties in part I). With that regard, a 1st order sensitivity analysis to the various input's variation is performed using the OFAT method. Throughout the study, the robustness (model's convergence) and predictiveness (physical representativeness of numerical results) of the suggested model in Part I are evaluated within the identified inputs' variation domains. The obtained results, in terms of the 1st order global sensitivity indexes, provide a subjectively quantitative and objectively qualitative ordering of the most influential parameters within the model's associated physical hypotheses. In particular, the obtained results show (a) the relevance of the Gaussian function to describe the spatial correlation of the Young's modulus property (b) the dependence of early age behaviour on, both, the spatial scattering of the mechanical properties and the maturity process; but mostly, on the structural size effect assessment (c) the main dependence of long term behaviour on the drying history and applied prestressing loads and (d) the importance of uncertainties propagation through the Thermo-Hydro-Mechanical calculations and through the operational lifespan of ageing concrete structures.