Image-Based Multi-scale Reconstruction of Unresolved Microporosity in 3D Heterogeneous Rock Digital Twins Using Cross-Correlation Simulation and Watershed Algorithm

Multi-scale modelling techniques are commonly used to characterize heterogeneous rock samples. However, open challenges limit the efficiency of these models. A significant issue is the tradeoff between resolution and field of view (FoV) during imaging. Capturing an image of a heterogeneous rock sample that includes pores of different scales with a large FoV is impossible. Various novel approaches have attempted to solve this problem, but they have inherent limitations such as unrealistic results and high computational costs. In this study, we propose a novel method to generate 3D multiscale images of two heterogeneous rock samples: Berea sandstone and Edward Brown carbonate. We scanned both samples at low and high (HR) resolutions using X-ray microtomography. Our approach involves distinct reconstruction of resolved and unresolved porosity in rock images at lower resolutions. We divide the unresolved porosity into smaller sections, called unresolved templates, using the watershed algorithm to reduce memory allocation. The cross-correlation based simulation approach then finds a suitable replacement template from the HR images, which contain a significant number of micro-pores, using a modified overlap region selection procedure in 3D. We compare the geometrical and petrophysical properties of the reconstructed multi-scale images with those of the HR rock images. The results show good agreement with the HR image properties computed from the direct numerical simulation approach. Additionally, our thus validated method is two to four times faster in constructing multi-scale images. Article Highlights An efficient multi-scale image reconstruction approach for heterogeneous rock samples Utilize watershed algorithm to reduce memory consumption and increase run time Improve accuracy of image reconstruction by considering different overlap area