Digital rock analysis for accurate prediction of fractured media permeability

Abstract

Determining the permeability of fractured rocks depends significantly on fracture geometry and topology. To date, no practical approach has been proposed to produce realistic binary images of fractured rocks from micro-computed tomography (micro-CT) images for fluid flow simulation. In this paper, a novel method is developed for generating a binary image that contains the true characteristics of fractured rocks for accurate computation of permeability. The method employs high-resolution scanning electron microscope (SEM) data for calibration of micro-CT images. Micro-CT is used to obtain 3D images of a highly fractured rock sample with a resolution of 16.5μm and SEM is applied to obtain images with nanometer resolution from polished surfaces of the sample. The SEM images are then registered to the micro-CT images to facilitate image segmentation and generate a calibration curve. The calibration curve correlates the grayscale values at the midpoint of each fracture to the true aperture size measured from SEM data. Thinned fractures of two subsets are extracted and used to obtain the gray-scale values at the midpoint of fractures. These are converted to the true aperture value using the calibration curve and subsequently grown by an adjustment algorithm to produce 3D calibrated binary images that are representative of the true fracture system. The connectivity and aperture size distribution of the subsets before and after calibration are quantified. The permeability of the subsets before and after calibration are computed using a direct numerical simulator and compared with experimental measurements. The computed permeabilities demonstrate that using non-calibrated images generates massive permeability errors whereas calibrated images produce accurate permeability results that nearly coincide with experimental measurments. The method can be applied to fractured rocks for better prediction of permeability and other petrophysical properties.