38 Geoscience and Hydrogeology in the AI Era: From Predictive Models to Real-Time Applications

This session offers a unique platform to showcase advancements in Artificial Intelligence (AI), Machine Learning (ML), and physics-informed methodologies, focusing on topics in hydrogeology, subsurface flow and transport, real-time geophysical inversion, critical minerals, and energy applications.

Sunday, 19 October 2025
1:30 PM - 5:30 PM
HBGCC, 210AB

Physics-Informed Neural Network Framework for Modeling Flow in Dual-Pore Porous Media

Paper number: 38-11
Program: Technical Sessions
Time: 4:20 PM – 4:35 PM

Authors

Abstract

Porous media span a remarkably diverse range of systems—including fractured carbonate reservoirs, ultra-tight shale formations, architected metallic foams, tissue-engineering scaffolds, and even plant xylem conduits. In many of these environments, flow and transport processes are governed by two hierarchically distinct but hydraulically coupled pore domains: a larger, highly permeable network that serves as the main conduit for flow, and a finer, more tortuous micro-pore network that controls storage, diffusion, and exchange dynamics. To accurately capture the resulting multiscale mass and solute transfer, the double-porosity/permeability (DPP) framework offers a powerful modeling approach. However, classical discretization methods—such as finite volume, finite difference, or finite element techniques—are often inadequate for workflows requiring rapid scenario screening, real-time data assimilation, or high-dimensional inverse analysis under sparse and noisy observations. In this talk, we introduce a physics-informed neural network (PINN) framework tailored to address these limitations. The method employs a shared neural backbone through which residual-balancing dynamically reconciles competing physical constraints. An error-guided collocation strategy further enhances accuracy by adaptively concentrating sampling points near steep pressure gradients and evolving flow fronts. Representative numerical results demonstrate that this mesh-free formulation offers multiple performance advantages: it resolves discontinuities without artificial diffusion, maintains numerical stability under extreme permeability contrasts, and avoids the spurious oscillations commonly observed in traditional mixed-element schemes.A particularly attractive feature of the proposed framework is its adaptability: the network can be incrementally retrained as new laboratory measurements, wireline logs, or in-situ sensor data become available. This enables continuously updated forecasting capabilities for critical applications such as critical-mineral extraction, tight-shale production, CO₂ and hydrogen storage integrity, and geothermal heat-exchange optimization.

Geological Society of America Abstracts with Programs. Vol. 57, No. 6

doi: 10.1130/abs/2025AM-413842

© Copyright 2025 The Geological Society of America (GSA), all rights reserved. Permission is hereby granted to the author(s) of this abstract to reproduce and distribute it freely, for noncommercial purposes. Permission is hereby granted to any individual scientist to download a single copy of this electronic file and reproduce up to 20 paper copies for noncommercial purposes advancing science and education, including classroom use, providing all reproductions include the complete content shown here, including the author information. All other forms of reproduction and/or transmittal are prohibited without written permission from GSA Copyright Permissions.