Understanding the gas movement phenomenon within the deep geological repository is essential for assessing the disposal system’s long-term stability. The primary gas transport mechanism through the bentonite is dilatancy-controlled flow, which differs from gas flow in general porous media. This flow is characterized by gas movement through microcracks created under relatively high gas pressure conditions, and the intrinsic permeability, air-entry pressure, and mechanical strength of the medium change due to the generation and propagation of these microcracks. Therefore, dilatancy-controlled flow cannot be simulated using the classical two-phase flow modelling technique. This study constructed the H2MD (two-phase hydraulic-mechanical-damage) numerical model by combining a damage model to simulate material degradation and the resulting change in intrinsic permeability with a classical two-phase flow model. In addition, the numerical model was tested against a 1D laboratory gas injection test investing gas flow mechanisms in the buffer, and a sensitivity analysis was performed on tensile strength, a key factor in the damage model for gas movement phenomenon. In the validation study, the proposed model successfully simulated the key features observed in the test: rapid stress and pressure increase trends, changes in intrinsic permeability due to damage, and the resulting flow rate. In addition, the effect of heterogeneity on the strength characteristics of each material and interfaces between materials was analyzed through field-scale test simulations, and the applicability of the model to upscaling analysis was examined. The study of heterogeneity effects confirmed that incorporating the strength characteristics of interfaces accurately simulates the gas flow path observed in actual tests. However, the model overestimated the gas flow before the gas breakthrough and underestimated the evolution of the damaged area within the buffer. Therefore, additional research on relative permeability and mechanical constitutive models is needed to improve the reliability of the current model.
Read full abstract