E-Thesis 193 views
A novel DEM method to investigate the deformation of porous media and fast prediction of relative permeability and stress-sensitive permeability / SHAN ZHONG
Swansea University Author: SHAN ZHONG
DOI (Published version): 10.23889/SUThesis.67186
Abstract
Porous materials play a pivotal role in various natural and industrial processes, with their properties being significantly influenced by stress conditions. Understanding and accurately predicting these properties are crucial for optimizing processes such as oil and gas recovery.Considerable efforts h...
| Published: |
Swansea University, Wales, UK
2024
|
|---|---|
| Institution: | Swansea University |
| Degree level: | Doctoral |
| Degree name: | Ph.D |
| Supervisor: | Li, C. |
| URI: | https://cronfa.swan.ac.uk/Record/cronfa67186 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Abstract: |
Porous materials play a pivotal role in various natural and industrial processes, with their properties being significantly influenced by stress conditions. Understanding and accurately predicting these properties are crucial for optimizing processes such as oil and gas recovery.Considerable efforts have been made in researching the absolute permeability of porous media, encompassing experimental, numerical, empirical, and deep learning methods. How-ever, most research relies on rock samples analyzed in laboratory conditions, which do not accurately reflect their operational environment. Consequently, the stress-sensitivity of permeability has emerged as a focal point of research.Additionally, numerous researchers have conducted laboratory experiments on this topic and have achieved promising results. Permeability typically follows a power law with respect to compressive stress, although the parameters may vary among different samples. These experiments are invariably time-consuming, and many interference factors cannot be entirely eliminated. With the advancement of Micro-CT technology, simulations based on digital rock samples have become an indispensable tool. There exists an extensive body of literature on the deformation of porous media through the creation of numerical models from digital rock samples. However, the research focused on simulating fluid flow within porous media under varying stress conditions remains an area that requires substantial effort.This thesis endeavors to establish a comprehensive framework for investigating the behavior of porous materials under varying stress conditions, with the ultimate goal of achieving precise real-world property approximations. The thesis adopts a multi-faceted approach to tackle the challenges associated with the analysis of porous materials under stress. Initially, it reviews various methods for obtaining permeability in a succinct yet comprehensive manner. Subsequently, it introduces a Discrete Element Method (DEM) for modeling the shapes of real digital rock samples. The control parameters influencing the quality of the final discrete element model are quantitatively analyzed, demonstrating the method’s efficacy in constructing discrete element models of porous media based on 3D digital images. Furthermore, the thesis calculates and presents the trends of several descriptors along with the deformation of porous media. Leveraging the permeability calculated under different deformations, and taking descriptors and 3D images as inputs, a neural network is shown to proficiently predict permeability across different deformation levels (with strain ranging from 0 to 0.03). Additionally, the polynomial interpolation method is employed to provide an explicit equation for calculating permeability from descriptors. The thesis then proceeds to validate an open-source lattice Boltzmann method based on Palabos for simulating multiphase flow in porous media, including a quantitative analysis of key parameters to offer insights into their selection. Following this parameter selection, the relative permeability of the sphere-packing model is calculated, and the neural network is applied to forecast relative permeability in two-phase flow within porous media.To further enhance this research, future endeavors are recommended. These include incorporating a more diverse range of porous media in the neural network’s training dataset and developing empirical formulas and feature selection methods for greater reusability. Addi-tionally, the analysis of different calculation methods for descriptors lacking clear definitions,such as to rtuosity and fractal dimension, holds promise.In summary, this thesis establishes a foundational framework for investigating porous materials’ properties under varying stress, providing valuable insights and a solid base for future research in this critical field. |
|---|---|
| Item Description: |
A selection of content is redacted or is partially redacted from this thesis to protect sensitive and personal information. |
| Keywords: |
Porous media, permeability, DEM, deep learning |
| College: |
Faculty of Science and Engineering |
| Funders: |
China Scholarship Council, Zienkiewicz Center for Computational Engineering |