Elastic Moduli of Arenites From Microtomographic Images: A Practical Digital Rock Physics Workflow

Abstract

Numerical computation from high-resolution 3-D microtomographic images of rocks (known as digital rock physics) has the potential to predict elastic properties more accurately. However, successful examples are limited to samples with simple structure and mineralogy. The physical size of sample is often too small to present heterogeneities at a larger scale and the image resolution is insufficient to characterize the details of rocks. Also, the grayscale values of different minerals in microtomographic images are often similar, and previous attempts to segment them as separate phases are not very successful. Here, we propose a practical digital rock physics workflow for somewhat more complex and ubiquitous rocks, namely, sandstones that contain mostly quartz and a small fraction of dispersed clay (known as arenites). Based on a set of images, we obtain a suite of postcomputation corrections to compensate for the effects of sample size and resolution of the microtomographic images. Furthermore, we build a segmentation workflow that effectively detects feldspar and clay minerals, despite their grayscale similarity to quartz. A moduli-porosity trend is derived from the subsamples of the original digital images. Bulk moduli agree well with the ultrasonic measurements on the dry samples at 40 MPa. Shear moduli remain overestimated, which is likely caused by poor knowledge of the mineral stiffness. We compensate for this effect using a heuristic correction to the matrix moduli. The final version of the workflow provides accurate elastic moduli trends with porosity and clay content based on only two samples of Bentheimer sandstone.

High‐Precision Tracking of Sandstone Deformation From Micro‐CT Images

“…After locating the same physical part of the S60 sample on the micro‐CT images scanned under different stress conditions, we segment the greyscale images into pore and solid phase with the Otsu method (Liang, Gurevich, et al., 2020; Otsu, 1979). The porosity of the same part of S60 is 0.1753 at 0 MPa (without residual stress), 0.1673 at 0 MPa (with residual stress), 0.1621 at 20 MPa, and 0.1625 at 36 MPa.…”

“…Rock properties, especially elasticity, can be strongly pressure dependent (Eberhart‐Phillips et al., 1989; Han et al., 1986; Zimmerman, 1990). However, micro‐CT images, as the foundation of DRP, are mostly scanned under ambient pressure (Ahmed et al., 2017; Arns et al., 2002; Liang, Glubokovskikh, et al., 2020; Liang, Gurevich, et al., 2020; Shulakova et al., 2013). There are a few exceptional attempts to obtain medical CT and micro‐CT images of rock at higher pressure.…”

Hierarchical Homogenization With Deep‐Learning‐Based Surrogate Model for Rapid Estimation of Effective Permeability From Digital Rocks

“…In principle, DRP provides a non‐destructive way to study porous media in a repeatable and more efficient manner as a complement to conventional laboratory measurements. The typical DRP workflow includes three steps: (a) acquiring raw digital rock data by tomographic imaging (Mees et al., 2003); (b) segmenting raw digital rock images to different phases (Sezgin and Iassonov et al., 2009) and (c) calculating physical properties by numerical simulation on the segmented images (Bazaikin et al., 2017; Keehm et al., 2001; Knackstedt et al., 2009; Liang et al., 2020; Saxena et al., 2017). Specifically, common numerical methods for fluid flow simulation on digital rock images include the finite difference method (Smith, 1985), the finite element method (Brenner et al., 2008), the finite volume method (Moukalled et al., 2016) and the lattice Boltzmann method (Chen & Doolen, 1998).…”