Stochastic finite elements analysis of large concrete structures' serviceability under thermo-hydro-mechanical loads - Case of nuclear containment buildings

摘要

This work proposes a global Stochastic Finite Element Method (SFEM) to model the effects of concrete ageing uncertainties on the serviceability and durability of large reinforced and prestressed structures with a containment role. As their modelling requires strongly non-linear, coupled and expensive calculations with a large number of parameters, adapted and efficient probabilistic strategies need to be defined aiming at a stochastic analysis within a reasonable cost and a physically admissible representativeness. In this contribution, this is achieved through four steps: (a) the definition of a well-established physical framework based on a staggered Thermo-Hydro-Mechanical + Leakage (THM-L) model; (b) the limitation of random inputs for uncertainty propagation to the most influential ones using a variance-based Hierarchized and Local Sensitivity Analysis (HLSA); (c) the construction of a THM-L response metamodel using Polynomial Chaos Expansion (PCE); (d) the reliability analysis of serviceability criteria using Crude Monte Carlo Method (CMCM) applied to the developed metamodel. For validation purposes and demonstration of achievability within a complex industrial framework, this global methodology is applied to an experimental 1:3 scaled Containment Building of a nuclear reactor. Eventually, it is shown that a complete probabilistic analysis of a physically admissible total dry air leakage rate (indicative of a nuclear containment structure's performance) and its evolution in time are obtained within a computational time of tens of days only. Such result can provide insights and help during the decision-making process for the design, maintenance and risk assessment of large structures. For Nuclear Containment Buildings (NCB), a direct application would be the evaluation of lifespan extension based on a leakage-rate-defined criterion under operational loads.