Pore-scale information has drawn great attention and has been widely used in estimating physical properties and their interrelations. A detailed 3D structure of pore geometry is essential to accurately determine rock properties. The most common method to obtain 3D pore structures is X-ray microtomography. We typically convert this tomogram (intensity values of attenuation) into pore-grain structure (binary values), by choosing a single threshold to classify pore and grain nodes. However, smoothing artifacts in 3D tomogram is the major problem when obtaining accurate geometry of pore structure. Especially, these artifacts are enhanced at the contacts between grains, which result in overestimating contact areas and thus the elastic stiffness of the rock. We tested two methods to improve the 3D reconstruction from X-ray tomogram: applying 3D sharpening filter and neural network classification. These two methods show observable improvements in grain contacts of 3D pore structures when compared to the thin section images of the same sandstone sample. In addition, numerical pore-scale simulations on the pore structures from our two methods also give better estimates of physical properties; however, the improvement in elastic stiffness is incremental. Although the new techniques improve the reconstruction of 3D pore geometry over the conventional single threshold method, the grain contacts are not resolved properly. For better estimation of elastic properties, we need to have tomograms that have much higher resolution than one we used in this study.


Physical Properties of rocks strongly depends on the pore microstructures. However, their complexity and prohibitive computational cost have been prevented us from using them in numerical simulations. With advent of highperformance and cheap computing power, we now can obtain realistic pore microstructure from high-resolution imaging technique, such as X-ray microtomography (Ketcham and Carlson, 2001). This technique currently can give us up to 2-micron resolution. While it is enough to accurately simulate transport properties of clastic rocks such as sandstones, it is sometimes not enough to resolve the contact area of grains thus gives overestimation of elastic properties. Moreover, one of drawbacks of tomographic inversion, the smoothing effect enhances this problem to cause much higher inaccuracy in elastic property estimation. In this paper, we are investigating the effect of the smoothing effect on the estimation of physical properties and trying to find a way to mitigate the smoothing effect for more accurate estimation of physical properties through pore-scale simulations.

Problem Description

X-ray Microtomogram to Pore Geometry The X-ray microtomography is one of widely-used technique to obtain a detailed microstructure of porous media, such as sandstones. Figure 1 shows the schematic procedures. The rock sample is Aztec sandstone from Valley of Fire National Park, NV, which is clean eolian sandstone with 23% porosity (Mollema and Antonellini, 1996; Keehm et al., 2006). An example of 2D tomogram has 1024x1024 pixels with the resolution of 11 micron and was obtained from the High-Resolution X-ray Computed Tomography Facility at the University of Texas at Austin (UTCT). After stacking and cropping the 2D tomogram, we can obtain 3D tomogram (right), which is an intensity map.

You can access this article if you purchase or spend a download.