Upscaling THM modeling from small-scale to full-scale in-situ experiments in the Callovo-Oxfordian claystone

Abstract

The coupled Thermo-Hydro-Mechanical (THM) behavior of the Callovo-Oxfordian claystone (COx) is of great importance for the design and safety calculations of the high-level radioactive waste disposal project in this potential host rock in France. The heat emitted by the waste causes a pore pressure increase within the surrounding rock essentially due to the differential thermal expansion of the pore water and the solid skeleton. The low permeability of the COx and its relative rigidity inhibits the discharge of the induced pressure build-up. Moreover, thermal loading may provoke thermo-mechanical stresses within the formation due to mechanical confinement by the rigidity of the surrounding host rock. An important research program has been conducted by the French National Radioactive Waste Management Agency (Andra) since 2003 in order to investigate the THM response of the COx under thermal loading, through laboratory tests, in-situ experiments, model development and numerical modeling. Within Task E of the DECOVALEX-2019 project, five research teams investigated upscaling THM modeling from a small-scale in-situ experiment (TED) to a full-scale in-situ experiment (ALC). The upscaling modeling started with a verification test to validate the numerical codes. Then, an interpretative modeling of the TED experiment was performed to calibrate the THM parameters of the COx. Finally, the calibrated THM parameters were used for a blind prediction of the ALC experiment. The modeling teams each adopted a thermo-poro-elastic approach which yielded satisfactory results. The blind prediction of the temperature field showed an overestimation of less than 2 °C which was considered acceptable. On the other hand, pore pressure was well predicted only in the direction parallel to the bedding whereas the slow dissipation of the pore pressure in the direction perpendicular to the bedding was not captured by any of the modeling teams – which remains an open question of the present study.