Advances in image processing, high-performance computing and numerical methods for digital petrophysics using micro-ct imaging

In this thesis, I investigate modelling and analyzing the properties of oil-bearing rocks through the use of micro-CT technology, image processing and numerical methods. There are three main interlinked themes to this work. (i) Processing of three-dimensional image data from micro-CT to obtain accurate 3D digital models of the grain and pore arrangement of the rock: a 'digital rock ' model. Quantitative descriptors of the rock structure can then be investigated for correlation with physical properties of the rock. (ii) Use of forward modeling techniques to calculate for such 'digital rocks ' phys- ical properties such as acoustic/elastic and permeability behaviour, and ultimately to model three-phase flow, something that is difficult to do experimentally in labora- tories. The understanding and predictive capability gained from this modelling can then be applied to real-world problems in oil fields. (iii) Modelling these properties and processes requires iterative computations on very large datasets. This is not practicable without careful choices of the appropriate numerical methods, algorithmic implementations and hardware architecture. All these choices are interdependent. Chapters 1 and 2 explain the motivation for doing this work, as well as the basic concepts of 'digital rock ' technology. These include image acquisition and image processing techniques. Chapter 3 investigates the correlation between bulk-rock structural features and calculated and experimental elastic/acoustic properties. It is shown that for the reser- voir sandstone samples studied, acoustic anisotropy is largely determined by the dis- tribution of grain contact surfaces. Chapter 4 gives an account of elimination of an artefact that was discovered during early image processing work. Chapters 5-7 focus on the use of digital rocks in the more complex computations required for modeling transport processes in the pore space, i.e solving the Pois- son equation for conductivity and the Navier-Stokes equation for fluid flow. The very large micro-CT data sets require parallel processing for this to be practicable. Chapter 5 explores the interlinked issues of appropriate numerical methods, paral- lel algorithms and hardware architecture. Chapter 6 examines multigrid methods for solving large linear systems with complex boundary conditions, such as digital rocks, while Chapter 7 introduces a new approach using wavelets as an alternative