Uncertainty Quantification of the Short‐ and Long‐Term Behavior of COx Claystone and Reliability Analysis of the IL‐LLW Repository's Concrete Liner Based on In Situ Convergence Data

Abstract

The excavation‐induced fractured zone (EFZ) and the anisotropic evolution in time of drifts’ convergence, as observed in the Mesue/Haute‐Marne Underground Researche Laboratory (MHM URL), reveal the complex behavior of Callovo‐Oxfordian (COx) claystone, the host formation for geological radioactive waste disposal project (Cigéo) in France. Especially, the dispersion of the in situ convergence monitoring exhibits the non‐negligible uncertainty of the host rock properties that must be quantified and considered in the stability analysis of the repositories’ support systems. In this work, the well‐known Bayesian inference with the transitional Markov chain Monte Carlo sampling technique is chosen to quantify the uncertainty of the short‐ and long‐term behavior of COx claystone and EFZ using the convergence data of a drift excavated in the major horizontal stress direction. An engineering approach is adopted to simulate the anisotropy of drift convergence. To reduce the computational cost of the numerical model prediction during the probabilistic inversion, the artificial neural network–based surrogate is chosen. The uncertainty of the visco‐elastoplastic behavior of COx claystone, as well as the uncertainty of the EFZ shape, is then considered in the reliability analysis of the concrete liner of an intermediate‐level long‐lived radioactive (IL‐LLW) repository. The numerical applications allow verifying the robustness of the current design for the repository support system.