Machine Learning Application for Permeability Estimation of Three-Dimensional Rock Images

Abstract:

Estimation of permeability in porous media is fundamental to understanding coupled multi-physics processes critical to various geoscience and environmental applications such as geologic carbon storage, subsurface energy recovery, and environmental fate and transport. Although a pore-scale model provides fundamental mechanistic explanations of flow and reactive transport processes, pore scale modeling often faces challenges in computational expense (in terms of both time and memory) and explicit knowledge of pore structures. Recent emerging machine learning methods with physics-based constraints and/or physical properties can provide a new means to improve computational efficiency while improving machine learning-based prediction by accounting for physical information during training. Here we first used 3D real rock images to estimate permeability of fractured and porous media using 3D convolutional neural networks (CNNs) coupled with physics-informed pore topology characteristics (e.g., porosity, surface area, connectivity) during the training stage. Training data including permeability were generated using both lattice Boltzmann simulations of segmented real rock images and pore network modeling of simplified networks. Both methods were successfully compared to each other. In particular, we explore the opportunity to estimate permeability at smaller volumes of rock samples and use the ML-based permeability values for upscaling to perform fine-scale continuum transport processes. Comparison of pore-scale reactive transport and an ML-based continuum approach will be presented to discuss upscaling strategy of reactive transport problems.