Macro-permeability distribution and anisotropy in a 3D fissured and fractured clay rock: 'Excavation Damaged Zone' around a cylindrical drift in Callovo-Oxfordian Argilite (Bure)

摘要

The Underground Research Laboratory at Bure (CMHM), operated by ANDRA, the French National Radioactive Waste Management Agency, was developed for studying the disposal of radioactive waste in a deep clayey geologic repository. It comprises a network of underground galleries in a 130. m thick layer of Callovo-Oxfordian clay rock (depths 400-600. m). This work focuses on hydraulic homogenization (permeability upscaling) of the Excavation Damaged Zone (EDZ) around a cylindrical drift, taking into account: (1) the permeability of the intact porous rock matrix; (2) the geometric structure of micro-fissures and small fractures synthesized as a statistical set of planar discs; (3). the curved shapes of large 'chevron' fractures induced by excavation (periodically distributed).The method used for hydraulic homogenization (upscaling) of the 3D porous and fractured rock is based on a 'frozen gradient' superposition of individual fluxes pertaining to each fracture/matrix block, or 'unit block'. Each unit block comprises a prismatic block of permeable matrix (intact rock) obeying Darcy's law, crossed by a single piece of planar fracture obeying either Darcy or Poiseuille law. Polygonal as well as disc shaped fractures are accommodated. The result of upscaling is a tensorial Darcy law, with macro-permeability K_(ij)(x) distributed over a grid of upscaling sub-domains, or 'voxels'. Alternatively, K_(ij)(x) can be calculated point-wise using a moving window, e.g., for obtaining permeability profiles along 'numerical' boreholes. Because the permeable matrix is taken into account, the upscaling procedure can be implemented sequentially, as we do here: first, we embed the statistical fissures in the matrix, and secondly, we embed the large curved chevron fractures. The results of hydraulic upscaling are expressed first in terms of 'equivalent' macro-permeability tensors, K_(ij)(x, y, z) distributed around the drift. The statistically isotropic fissures are considered, first, without chevron fractures. There are 10,000 randomly isotropic fissures distributed over a 20m stretch of drift. The resulting spatially distributed K_(ij) tensor is nearly isotropic (as expected). At the scale of the whole EDZ, the global K_(FISSURES) is roughly 5000 times larger than permeability K_M. The detailed distribution of the equivalent K_(FISSURES)(x, y, z) defined on a grid of voxels is radially inhomogeneous, like the statistics of the disc fissures. In addition, a moving window procedure is used to compute detailed radial profiles of K_(FISSURES) versus distance (r) to drift wall, and the results compare favorably with insitu permeability profiles (numerical vs. experimental boreholes at Bure's GMR drift).Finally, including the large curved chevron fractures in addition to the random fissures, the resulting K_(ij)(x, y, z) appears strongly anisotropic locally. Its principal directions are spatially variable, and they tend to be aligned with the tangent planes of the chevron fracture surfaces. The global equivalent K_(ij) of the whole EDZ is also obtained: it is only weakly anisotropic, much less so than the local K_(ij)'s. However, because of the radially divergent structure of the 'chevrons' (although not quite cylindrical in geometry), it is recognized that the global K_(ij) due to chevrons lacks physical meaning as a tensor. Considering only the magnitude, it is found that the permeability due to 'chevrons' (K_(CHEVRONS)) is about 4 orders of magnitude larger than that due to statistical fissures (K_(FISSURES)), assuming a hydraulic aperture a_(CHEVRON)=100μm.