ABSTRACT: From 1996 to 2021, uniaxial creep tests were performed on salt samples in dead-end drifts of the Varangéville (France) and Altaussee (Austria) mines to take advantage of constant temperature and hygrometry. The applied loads were from 0.05 MPa (relative) to 4.5 MPa, i.e., much smaller than the loads currently applied during standard creep tests performed at the laboratory. Main conclusions are: (1) Steady state is reached after a long period (longer than 8 months). (2) Cumulated transient creep is relatively large (3) Strain rates are faster (than extrapolated from high stresses) by 4-5 orders of magnitude (4) Steady state strain rate is a linear function of the applied stress (approximately) in the σ < 3 MPa domain (5) Strain rate is a decreasing function of grain size (6) The transition between the linear (n = 1) and the non-linear (n = 3 to 5) behavior seems to range between 3 MPa and 4.5 MPa (7) No creep is observed in a very dry environment (8) No threshold for salt creep (or smaller than 0.05 MPa) is observed (9) In the small stress domain, reverse creep is observed. 1. INTRODUCTION It has been suspected for long (Spiers et al., 1990; Urai and Spiers, 2007) that, in the small deviatoric stress domain (σ < 3 MPa), the governing mechanism for salt creep was pressure solution — rather than dislocation creep. A consequence should be that creep rate in this domain is much faster - by several orders of magnitude - than extrapolated from tests performed in the high stress domain. In addition, creep rate should be a decreasing function of grain size; it should be a linear function of the applied stress, and the presence of a small amount of brine at the grains interface should be a necessary condition for active creep. These statements were based on theoretical arguments, geological evidence and the results of tests performed on artificial salt.

ABSTRACT: Salt has long been considered as a potential host rock for the geological disposal of radioactive waste because of its favorable properties, including self-sealing, low permeability and high thermal conductivity. The feasibility of nuclear waste disposal within salt formations has been investigated mainly for small-sized canisters, widely considered in many national nuclear waste disposal programs. Direct disposal of larger-sized canisters originally designed for spent nuclear fuel storage and transportation has lately been examined as a promising and cost-effective alternative. However, the amount of decay heat released by large canisters may cause high temperature in the backfill surrounding the canisters and in the host rock. This results in pressure and stress changes that may affect the long-term repository performance. To analyze it, the TOUGH-FLAC simulator is used to conduct fully coupled Thermal-Hydro-Mechanical (THM) simulations of a generic salt repository for large-sized canisters. In this code, large deformation has been considered allowing to accommodate viscous compaction of the crushed salt backfill as well as the long-term stress changes in salt host rock. The simulations show that the peak temperature in the backfill and in the host rock can be reduced by adjusting the distance between the canisters along emplacement tunnels or the spacing between the tunnels themselves. Nevertheless, the thermal pressurization in the salt host rock and in the crushed salt backfill, following the reconsolidation of the backfill, seems to be less sensitive to these spacing adjustments. Consequently, the pore-pressure could potentially exceed the lithostatic stress causing an increase of permeability and a fluid infiltration into the rock mass that may last more than 1000 years after the waste canister emplacement. Thus, it is critically important to analyze and manage these coupled THM processes for a safe and effective disposal of nuclear waste in salt formations. 1 INTRODUCTION Geologic disposal of large-sized canisters, originally designed for spent nuclear fuel storage and transportation, has lately been examined as a promising and cost-effective alternative to conventional smaller-scale canisters commonly considered in many national nuclear waste disposal programs (Hardin et al., 2015).

In the energy transition context, salt caverns are probably the most promising storage solution that promotes the development of intermittent renewable energies, due to their flexible and high deliverability. However, their design is still challenging since it should account for their entire lifetime, from rapid cycling exploitation to centuries of abandonment. The key to an optimal design is a constitutive model for rock salt that ensures pertinent short and long-term predictions. In this paper, we confront the results of five experimental campaigns conducted on different salts with existing rheological models. This confrontation proved that the studied models are capable of describing laboratory tests, however their predictions for the long term are either too conservative or overly optimistic. In practice, conservative or optimistic approaches do not ensure the optimal design of the facility. For this reason, we propose a new constitutive model that provides pertinent long-term predictions while interpreting satisfactorily short-term and long-term laboratory tests.

We present results of a benchmark exercise between TU Clausthal and LBNL to validate two simulators for coupled thermal-hydraulic-mechanical processes, FLAC-TOUGH and TOUGH-FLAC. This benchmark exercise concerns the TSDE experiment, conducted in the Asse salt mine between 1990 and 1999. This experiment provides excellent data at field-scale for natural salt creep and crushed salt compaction. The main objective of the benchmark is to evaluate the capabilities of the two simulators to predict relevant processes under repository conditions for heat-generating nuclear waste in saliniferous media. Moreover, the three-dimensional modelling of the experiment enables the calibration of some parameters needed to describe the time-dependent response of salt rock mass, whose determination is difficult at laboratory-scale due to very low deviatoric stresses needed, resulting in extremely long tests. The results of the benchmark are satisfactory so far, and further modelling will be conducted to reduce remaining discrepancies between experimental data and numerical predictions.