| Conventional (primary) oil recovery leaves as much as 80% of the original oil in place (Lake, 1989), and therefore secondary and tertiary recovery schemes, such as the injection of water, steam, surfactant, or carbon dioxide, become necessary to recover additional oil. Since these displacement processes depend largely on the topology of the porous medium, it is very important to accurately characterize the pore structure in order to develop effective recovery strategies (Dullien, 1992).; The objective of this work is to develop a fundamental understanding of the relationship between pore structure and its characterizing information and properties, as well as to propose and test reconstruction algorithms designed to simulate real porous media based on limited input information. A novel and powerful application of reconstructing three-dimensional media from two-dimensional thin section data is also investigated, as well as two-phase experimental studies into non-Darcy flow behavior. Extensive characterization is performed using a host of original algorithms, including an important skeletonization technique that rivals a commercial software package. Three stochastic reconstruction techniques are investigated in this work, including a linear filter method, a simulated annealing method, and a hybrid of the two. Reconstructions are performed on a 3D digital image of a sample of Brent Triassic sandstone provided by Mobil and show very good agreement to the Mobil image over a wide range of properties if the chosen input information is well matched by the methods. The simulated annealing method is shown to be the most simple and versatile though having a computation time limitation on larger systems. |
