Modelling Lasgit experiment

Document available in extended abstract form only. A large scale gas injection test (Lasgit) is carried out at the Aespoe Hard Rock Laboratory in Sweden. Lasgit is the first demonstration project designed to study gas migration in bentonite under full-scale repository conditions. The objective of this experimental programme is to provide data to improve process understanding and test/validate modelling approaches, which might be used in performance assessment. Specific objectives are: (1) perform and interpret a large-scale gas injection test based on the KBS-3 repository design concept, (2) examine issues relating to up-scaling and its effect on gas movement and buffer performance, (3) provide additional information on the process of gas migration, and (4) provide high-quality test data to test/validate modelling approaches. The test has a number of injection possibilities, some of them essentially intended for hydration and other intended for gas injection. The modelling work is significantly complex as the water injection phases are extremely fast and thus produce high velocities of water through the interfaces. In a real repository, the interfaces are expected to close when the bentonite expands. However, in a fast test if water penetrates before the saturation of the rings, the opposite effect may be produced, i.e. interfaces opening by pore pressure development. The modelling of this type of in situ experiments requires the incorporation of a constitutive model that is able to reproduce swelling induced by hydration of the buffer. The installation of the bentonite rings is done in such a way that the contact between the clay blocks and the canister remains as a discontinuity or interface. A similar situation happens on the contact surface between bentonite blocks. It has been observed that under natural hydration, these interfaces tend to close as hydration and swelling of the bentonite progresses. Normally, the interface sealing would happen and therefore special treatment of interfaces would perhaps not be necessary. Later, when gas generation by corrosion was generated, these interfaces would play a role and hence they cannot be neglected in a model. The special case of an experiment forces to include the interfaces. The necessity of including the interfaces explicitly appears both for the short term during water injection and for the long term as gas is generated. Even with forced hydration the interfaces manage to close, especially if the bentonite has a high swelling capacity. In case of mixtures of bentonite and sand, the interface closure during an experiment may be compromised as swelling capacity is lower. In any case, gas injection and migration is likely to be controlled by the presence of these interfaces. Therefore modelling of this kind of tests, especially if the water hydration is made very fast, requires the incorporation of interfaces and will improve the capabilities of the model. The present approach considers interfaces as a zone of a given thickness which contains a so called 'embedded' discontinuity model. This is necessary to model the penetration of water during the phase of hydration and later, the pressure responses induced by gas generation. The pressure response at sensors indicates that preferential paths through the discontinuities develop. Gas injection at a point in the canister wall generates a gas propagation front. The gas fluxes are higher at the interfaces than in the bentonite body. The interfaces play a major role as the penetration of the gas takes place in a preferential way. It can be seen that gas pressure development at points situated at some distance increases very suddenly, and at some distance it remains constant. Gas outflow is permitted on the model boundary in a zone representing the fractures of the host rock. According to the model, permeability may increase up to two orders of magnitude when the peak of water or gas pressure occurs. Hydration and the induced bentonite swelling counteract the opening effect and therefore, permeability decreases as the interfaces close, going back to low permeability values. One of the main drawbacks of the axisymmetric model is that the in situ gas injection is performed in reality as point source but in the 2D axisymmetry the point source becomes a ring source. This reduces the gas pressure increments for the same properties. This also happens for the flows. So developing a 3D model is a necessary but challenging task in view of the complex protocols that are considered in the test