A novel reactive transport framework for fluid-rock interaction analysis
Hax Damiani, Leonardo · BORIS Theses (Bern Open Research Information System) (Bern University Library, Hochschulstrasse 6, 3012 Bern, Switzerland) · 2021
Not cached yet. This record was fetched live for this request and is stored after the lookup. Reopening the page should serve the cached copy.
BORIS Theses (Bern Open Research Information System) (Bern University Library, Hochschulstrasse 6, 3012 Bern, Switzerland) · Nakuru County Council · dissertation · 2021
Abstract
Reactive transport modeling is a fundamental tool in modern computational geochemistry. However, its predictive accuracy is often constrained by the limitations of existing numerical tools, which struggle to capture the complexity of natural phenomena and process couplings. This thesis presents an open-source simulation framework designed to address these challenges. The framework, presented in Chapter 2, offers a versatile, goal-oriented, and easy-to-learn environment for reactive transport modeling.
It provides users with full control over the numerical methods and integrates various computing engines optimized for state-of-the-art hardware platforms to model both physical and chemical processes. Basic reactive transport equations predefined for typical geochemical problems can be easily scaled and extended to include additional physical and chemical phenomena through a high-level, intuitive programming interface. Additionally, the framework’s capabilities were demonstrated through multiple benchmark analyses, focusing on electrochemically coupled multicomponent diffusion across charged and uncharged membranes, combined with complex geochemical reactions.
Building on this, Chapter 3 demonstrated its application to the Hydrogen Transfer (HT) experiment conducted at the Mont Terri underground rock laboratory in Switzerland. This experiment required a coupled geochemical model to describe the transport of dissolved gases in the surrounding Opalinus Clay formation. The model successfully reproduced the observed gas composition evolution in the borehole and the composition of extracted water. Importantly, it highlighted that gas fluxes across the borehole wall were primarily governed by the diffusive transport of dissolved gases, with the estimated diffusion coefficients showing strong agreement with prior studies and experiments.
A key innovation of this framework, presented in Chapter 4, lies in the integration of machine learning (ML) techniques to accelerate chemical equilibria calculations, overcoming common bottlenecks in reactive transport modeling. The ML-assisted geochemical modeling framework generates training datasets from geochemical solvers, while neural networks are trained, tested, and validated against traditional geochemical solvers. This novel approach overcomes the inherent limitations of chemical solvers, such as their non-parallelizable nature and non-transferability across different computer architectures.
In summary, this research clearly demonstrates the power of combining modern numerical methods with machine learning techniques to significantly enhance the efficiency and scalability of reactive transport models. The modularity of the framework allows for flexible combinations of solvers and approaches, providing a significant leap in efficiency, accuracy, and scalability for simulating complex geochemical processes.
Open the resolved locator · 10.48549/5486
Metadata licence: OpenAlex metadata (CC0)