Upscaling Strategy to Simulate Permeability in a Carbonate Sample Using Machine Learning and 3D Printing

Abstract

Characterizing heterogeneity is crucial to assess the variability of rock properties in carbonate reservoir samples. This work introduces an original multiscale approach to simulate permeability and porosity in heterogeneous carbonate samples using 3D X-ray computed tomography images. The main novelty of our approach is to introduce a quantitative heterogeneity description in terms of texture classification using machine learning. The rock texture classification result is then used to upscale rock properties simulations from fine to coarse scale. The fine scale properties are investigated based lattice Boltzmann method, while a Darcy-scale flow simulator is adopted for estimating coarse scale properties. In addition, due to the critical role played by petrophysical properties at fine scale, a 3D printing technique is employed to validate experimentally the numerical simulations at this scale. Finally, we present an application of our proposed approach on a real carbonate sample from the Middle East carbonate oilfield reservoir.

Keywords

• Machine learning
• micro-computed tomography
• permeability
• upscaling
• 3D printing