Microscale flow modelling in geologic materials

Three-dimensional imaging techniques, numerical methods for simulating flow and transport, and emergent computational architectures are combined to enable fundamental studies of fluid flow at the pore scale. High resolution reconstructions of porous media obtained using laser scanning confocal microscopy reduce sampling artifacts to sub-micron features, and simultaneously capture multiple grain length scales. However, the volumetric image data sets are extremely large, and there are significant computational challenges in utilizing this information effectively. The principal problem lies in the complexity of the geometry and the retention of this structure in numerical analyses. Lattice Boltzmann (LB) methods provide a direct means to simulate transport processes in complex geometric domains due to the unique ability to treat accurately and efficiently the multitude of discrete boundary conditions. LB methods are numerically explicit as formulated, and this characteristic is exploited through a mapping of the numerical domain to distributed computing architectures. These techniques are applied to perform single phase flow simulations in 3D data sets obtained from cores of Berea sandstone using confocal microscopy. Simulations are performed using both a purpose-built distributed processor computer and a massively parallel processer (MPP) platform.