GPU-based pore-scale modelling of two-phase flow and transport in porous media

  • Senyou An

Student thesis: Phd


Fluid flow and solute/heat transport in porous materials are ubiquitous in nature, as well as in many scientific problems and engineering applications. Despite the importance, there is yet a significant room for in-depth understanding and enhancing the predictive capabilities for processes, such as multi-phase flow, fluid-solid interaction, reactive transport, and non-Newtonian fluid flow in porous media. In this thesis, advanced computational methods and algorithms are developed to simulate and analyze the aforementioned processes, which are directly linked to many subsurface processes such as enhanced oil recovery, geological CO2 storage, and contaminant hydrogeology. All of the algorithms are accelerated using the MPI-Cuda program with developed GPU-based solvers for the linear and non-linear systems with millions of equations. The developed algorithms have the features of higher accuracy, lower computational cost, and larger simulation domain compared to competitive pore-scale methods. In this thesis, the following topics have been thoroughly investigated: the effect of spatial heterogeneity on constitutive relations (e.g., capillary pressure, relative permeability, and dispersivity) for two-phase flow and solute transport; heat/mass transport for Newtonian and non-Newtonian fluid flow through porous media; two-phase flow regimes and residual saturation for viscous and capillary fingering; fluid-solid interaction during CO2 storage; and wettability alteration during controlled salinity waterflooding. To study these complex phenomena, I have developed the following computational methods: a) Flow in porous media. Proposed models include quasi-static and fully implicit dynamic pore network models for two-phase flow, single- and two-phase volumetric lattice Boltzmann methods to simulate flow in image-based structure, and single-phase pore network model for non-Newtonian fluid flow in the porous material. In addition, an upscaling method has been applied to the volumetric lattice Boltzmann model to simulate flow in larger domains. b) Solute/heat transport for both Newtonian and non-Newtonian fluids flow through porous media. Algorithms for conservative advection-dispersion transport are proposed based on the finite volume method. Also, reactive transport with local equilibrium and kinetic reactions are realized by coupling volumetric lattice Boltzmann model with PhreeqcRM. c) Pore-network generation and extraction for the analysis of pore morphology. The pore-network generation code is able to create pore networks with specific geostatistical parameters, e.g., spatial correlation length. d) Image segmentation. For complex X-ray images and low-contrast multi-phase images, algorithms for segmentation are proposed based on the active contour method, local threshold method, and local refinement method. Additionally, some X-ray tomography and microfluidic laboratory experiments are included in this thesis to investigate the fluid dynamics and solute transport in porous media, and/or validate the proposed computational models. The novel insights into the physics of the porous media obtained using the developed computational methods have been fully discussed in the chapters.
Date of Award31 Dec 2021
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorKonstantinos Theodoropoulos (Supervisor) & Vahid Joekar-Niasar (Supervisor)


  • CO2 geo-storage
  • Contaminant hydrogeology
  • Enhanced oil recovery
  • Porous materials
  • GPU-based simulation
  • Heat transfer
  • Fluid rheology
  • Solute transport
  • Reactive flow
  • Multiphase flow

Cite this