Theses and Dissertations

ORCID

https://orcid.org/0000-0001-9250-0990

Issuing Body

Mississippi State University

Advisor

Ramirez-Avila, John J.

Committee Member

Vahedifard, Farshid

Committee Member

Peters, John F.

Committee Member

Amirlatifi, Amin

Committee Member

Freyne, Seamus; Stache, Jeremiah M.

Date of Degree

12-13-2024

Original embargo terms

Worldwide

Document Type

Dissertation - Open Access

Major

Engineering (Civil Engineering)

Degree Name

Doctor of Philosophy (Ph.D.)

College

James Worth Bagley College of Engineering

Department

Richard A. Rula School of Civil and Environmental Engineering

Abstract

Understanding the behavior of unsaturated soils under multi-physics loading is crucial for addressing several challenges related to emerging geo-energy technologies, climate change, and geohazard mitigation. The main objective of this research is to enhance the state of the art for modeling unsaturated soils under multi-physics loading conditions through physics-based and data-driven approaches. Toward this objective, this dissertation is divided into two distinct yet interrelated parts. In the first part (Chapters 2, 3, and 4), thermodynamic principles are employed to develop new constitutive functions to describe soil suction and stress-dilatancy of unsaturated soils under temperature-dependent conditions. Building upon the insights gained from the first part, the second part of the dissertation (Chapter 5) presents a novel physics-informed data-driven constitutive model for unsaturated soils under thermo-hydro-mechanical (THM) conditions via a generative machine learning method. In Chapter 2, a novel model for temperature-dependent suction is developed by expanding the Gibbs free energy to address both internal chemical exchanges and the thermodynamic system as a whole. The suction model is then utilized in Chapter 3 to establish a stress-dilatancy relationship for unsaturated soils, considering the difference between net and effective stresses to derive equations for dissipative and internal energies. The proposed dilatancy model is extended in Chapter 4 to temperature-dependent conditions considering various sources of energy dissipation, including entropy, water flow, friction, and energies associated with volume change and soil grain rearrangement. Based on the insights gained from the first section, Chapter 5 develops a new physics-informed data-driven constitutive model by utilizing a hybrid Variational Autoencoder (VAE) and Generative Adversarial Network (GAN) framework to predict the THM response of unsaturated soils. The hybrid VAE-GAN model integrates physical constraints such as physical consistency and conservation laws. Experimental data from triaxial tests under varying THM conditions were used to train, validate, and test the model. The findings of this research demonstrate the great yet unexplored potential of physics-informed data-driven models for modeling THM problems in unsaturated soils, offering a powerful tool for future research and practical applications.

Download

Share

COinS