High-level waste, stemming from nuclear electricity generation poses significant environmental and safety concerns. Currently, high-level wastes are stored in interim facilities needing constant monitoring and waiting for a definitive solution. Deep geological repositories represent a globally recognized viable solution for the permanent disposal of those nuclear wastes and various concepts are currently being evaluated at the international level. In Switzerland, a multi-barrier approach is foreseen integrating engineered and natural barriers to prevent the migration of radionuclides into the environment.
This multi-barrier system encompasses the spent fuel assemblies safely emplaced in a stainless steel canister, which is embedded in a bentonite buffer and confined by the host rock.
Throughout its operation, the different barriers will be subjected to so-called repository-induced effects such as heat emitted by the decaying radioactive waste or gas production by anaerobic corrosion of the canisters. The thermo-hydro-mechanical (THM) evolution of the repository nearfield in the early post-closure period is of high significance for long-term performance of the entire multibarrier system. Confidence in the long-term safety of repositories relies on the model-supported design optimization of the multi-barrier system and on the model-based assessment of the long-term repository performance. The conceptual THM framework and implementation must be verified and validated against controlled experiments to be able to accurately simulate the repository-induced effects and forecast long-term interactions. This thesis thus aims to develop a comprehensive and rigorous methodology to verify and validate a fully coupled THM model of
the nearfield around the high-level waste emplacement tunnels of a geological repository.
This includes quantifying uncertainties during model abstraction, addressing parametric and conceptual uncertainties through sensitivity and uncertainty analyses. The goal is to showcase the repository's expected long-term performance and evaluate potential deviations due to uncertainties in parameters, concepts, and scenarios. The proposed methodology relies on a sensitivity analysis to assess parametric uncertainties and rank influential parameters and couplings using statistical metrics.
In this thesis, two methods are utilized, a screening method to rank parameters by importance and a variance-based method to assess parameter uncertainty and interactions. Used together, they offer a comprehensive understanding of parameter effects, aiding in parameter prioritization for calibration. Notably, the analysis reveals that heat transfer in the porous medium primarily occurs through conduction, enabling the separation of thermal and hydromechanical responses. Additionally, it highlights the strong coupling of hydro-mechanical parameters with permeabilities, air entry pressures, and stiffness, impacting model response significantly.
Following validation, modeling uncertainties are quantified using deterministic and stochastic tools to calibrate the model against in-situ experiments. Bayesian interpretation is employed to derive explicit modeling uncertainty, establishing 95% error bounds around the optimal solution. Calibration substantially enhances model response while maintaining low uncertainties, providing a robust foundation for future model-driven designs and safety assessments.