Development of an upscalable HM model for representing advective gas migration through saturated bentonite

Highlights

• Gas transport in water-saturated bentonite/clays.

• Novel upscaled model developed for Task A of the DECOVALEX-2019 project.

• Upscaled hydro-mechanical model representing gas flow through dilating capillaries.

Abstract

Understanding gas transport through engineered and natural clays is important for their use in buffers and seals in a range of Geological Disposal Facility (GDF) concepts. Gas migration via processes of multi-phase flow of water and bulk gas, and transport of dissolved gases, have been widely studied and are well understood for traditional porous media. However, models using only these processes are unable to represent the complex behaviour of gas flow in water-saturated clays, where gas migrates through the creation of additional flow paths (via dilatant pathways/tensile fractures). A bespoke fully-coupled hydro-mechanical model representing gas flow through dilating capillaries is presented, where gas in the model is considered to have a separate permeability to water within the saturated bentonite, dependent on the in situ stress conditions. The model permits a coarse spatial discretisation, compared to standard multiphase flow approaches, while still providing a good match to experimental datasets. This should allow for future upscaling (in space and/or time), which may be required if such modelling is necessary in a performance assessment (PA) or GDF safety case context. The model has been tested against small-scale laboratory experiments investigating gas flow through saturated MX-80 bentonite samples as part of participation in Task A of the DECOVALEX-2019 project, and has provided useful experience in the interpretation of the experimental configuration and its representation in models. Good agreement of the model results to the experimental data can be obtained after calibration of a few key parameters, which also provide a good match to an additional historical dataset. The significance of these gas flows is considered in a PA context and potential future improvements are discussed.

Introduction

Many Geological Disposal Facility (GDF) concepts for radioactive waste expect gas to be produced through mechanisms including corrosion of metals (e.g. H₂), radioactive decay (e.g. Rn) and radiolysis of water (H₂), as discussed by Rodwell et al.¹ It is therefore important to understand the migration of gas (by molecular diffusion and bulk advection) that has been generated within the repository context, both in terms of bulk gas (potential GDF and host rock pressurisation) and trace gas transport. For disposal concepts involving bentonite as a barrier within the GDF and/or clay host rocks, a discrete gas phase could form if the rate of gas generation exceeds the rate of diffusion of the gas into the clay and/or the rate of dissolution of gas into the pore water. Transport of water, bulk gas and dissolved gases through multi-phase flow processes is well understood for traditional porous media. However, models using only these processes are unable to represent the complex behaviour of gas flow in water-saturated clays, where gas migrates through the creation of additional flow paths.^2,3 These emergent gas pathways are observed and conceptualised by Marschall et al.⁴ via the processes of pathway dilation and development of tensile fractures. Such processes can change the hydraulic and mechanical properties of the medium suggesting strong hydro-mechanical (HM) coupling in the system. It is therefore important in terms of a GDF safety case and performance assessment (PA) context to develop a sufficient understanding of these gas migration processes, since they could affect the ability of clay buffers and seals to perform as required by the safety case.

Two gas flow experiments carried out by the British Geological Survey (BGS) have been considered as part of Task A of the DECOVALEX-2019 (D-2019) project, whereby gas is injected into a small fully-saturated compacted MX-80 bentonite cylindrical sample under constant volume conditions. The first is a one-dimensional gas injection test⁵ in which gas is injected from one end of the sample and outflow measured at the opposite end. The second experiment is a point injection gas test⁶ where gas is injected through a steel rod into the centre of the sample and outflow measured at the outer radial surface. These experiments are analysed to build a conceptual understanding of the key processes involved in such gas tests and then used to inform the numerical model presented herein. One of the important findings for modelling these experiments has been to include an accurate representation of the gas injection pressure boundary condition, which is tightly coupled to the rate at which gas is able to enter the clay. It was therefore found to be necessary to explicitly model the gas injector mass balance, to allow proper comparison to the experimental data.

A bespoke fully-coupled hydro-mechanical model has been developed7, 8, 9 to help conceptualise and understand the D-2019 Task A experiments, which represents gas flow through dilating capillaries with a separate permeability to water. This approach follows conceptually from previous work^10,11 whereby gas migration is modelled through a predefined network of capillaries within MX-80 bentonite. Modelling the gas using a single-phase flow approach was found to be sufficient to capture the key gas migration features from these laboratory-scale experiments, but does not preclude using the model in conjunction with conventional multi-phase flow approaches. To enable the model to be applied in future to upscaled geometries it has been developed using a continuum modelling approach. This was one of the main objectives of the model since upscaling (in space and/or time) may be necessary if the model is to be applied in a PA or GDF safety case context. For this reason, development of the model has also focussed on representing the key deterministic features from the experiments rather than reproducing fluctuations (e.g. gas “burst-type” behaviour) relating to pathway instability as observed in the experimental data.

The setup and results of the D-2019 Task A gas flow experiments are described in Section 2 and the formulation of the model is given in Section 3. Results are compared with these experiments in Section 4, where the MX80-8 dataset from Harrington and Horseman¹² is used as an additional dataset outside of D-2019 Task A to build confidence in the modelling approach. This is followed by some discussion in Section 5 of conceptual and modelling uncertainties, and consideration for possible future application to radioactive waste disposal systems.