Upscaling porosity and permeability in core plug sample from core plugs simulations using machine learning methods

Characterizing accurately rock properties at core scale is critical for reservoir scale modeling. This step is especially complex when dealing with carbonate rocks because of their inherent heterogeneities at several length scales. The deposition and diagenesis processes result in diverse pore-shaped geometries, significantly influencing petrophysical properties. Notably, two core plug samples taken from adjacent locations within a carbonate core may exhibit substantial differences in rock properties. While standard core analysis methods offer precise experimental measurements in the laboratory, they fail to consider the pore-scale variability within core plug samples. Digital rock physics (DRP) emerges as an approach aiming to characterize rock properties at the pore scale through the utilization of X-ray micro-tomography and numerical simulation methods. DRP has been widely employed to estimate various rock properties, including porosity, permeability, and elastic moduli, in both siliciclastics and carbonate rocks. However, a well-defined workflow for numerically characterizing rock properties in carbonates is currently absent. This study introduces a novel multiscale method for simulating permeability and porosity in heterogeneous carbonate samples using 3D X-ray computed tomography images. The unique aspect of our approach lies in incorporating a quantitative description of heterogeneity through texture classification using machine learning. The results of rock texture classification are then utilized to scale up simulations of rock properties from a fine to a coarse scale. The fine-scale properties are examined using the lattice Boltzmann method, while a Darcy-scale flow simulator is applied to estimate coarse-scale properties. In addition, due to the critical role played by petrophysical properties at fine scale, we employ a 3D printing technique to experimentally validate the numerical simulations. Lastly, we demonstrate the application of our proposed approach on two carbonate samples from a Middle East carbonate oilfield reservoir.