Back to search

Georgia Institute of Technology

Performance Prediction and Optimization of Graphene Aerogel for Membrane Distillation Using Experimental, Molecular Dynamics, and Machine Learning Approaches

Abstract

dc:description.abstract

The growing global demand for freshwater has spurred the development of advanced desalination technologies, with membrane distillation (MemD) recognized as a promising approach due to its capacity to harness low-grade thermal energy (LoT). This thesis aims to optimize the mass and heat transfer performance of graphene aerogel (GA)-based membranes for the MemD process to enable more efficient water purification. However, the relationship between GA's morphology and its performance during MemD is still lacking in studies and remains an unveiled riddle. Traditional methods for MemD performance characterization are typically sample-destructive, inaccurate, and time-consuming. These limitations pose substantial barriers to the practical application of GA-based membranes and constrain further innovation in MemD development. To address these challenges, this thesis develops a systematic, scalable, and generalizable framework for predicting and optimizing the heat and mass transfer properties of GA-based membranes by integrating molecular dynamics (MDyn) simulations with machine learning (ML) techniques. Additionally, a fully automated, modular testing platform for MemD is designed and constructed, providing a standardized protocol for future experimental studies. Together, these efforts offer a rapid and accurate methodology for characterizing and predicting MemD performance at both experimental and computational levels. The thesis comprises five key components: 1. In the first part, the mechanism, consequences, and mitigation strategies of inorganic scaling during the MemD process are systematically analyzed in the form of a literature review. 2. In the second part, a custom-built direct contact membrane distillation (DCMD) testing platform was developed using Python scripting and 3D-printing techniques to evaluate the effects of inorganic scaling and operational parameters (e.g., flow rate and feed temperature) on the permeate flux of commercial PTFE membranes. 3. In the third part, MDyn simulations were employed to model the vacuum membrane distillation (VMemD) process using GA-based membranes. We investigated how various input parameters influence the resulting GA morphology, and how structural features such as pore diameter, porosity, and tortuosity affect water vapor diffusivity and thereby the permeate flux. 4. In the fourth part, MDyn simulations were further utilized to study the thermal transport properties of GA. Specifically, the relationship between the geometric features of local graphene sheets and their phonon transport capabilities was analyzed in detail. 5. In the fifth part, a transformer-enhanced 3D convolutional neural network (CNN) model (TRM-CNN) was developed to predict the morphological, thermal, and mass transfer properties of GA during the MemD process, leveraging the dataset generated from the previous two MDyn simulations. Key Findings are listed below: 1. The custom-built DCMD testing platform achieved long-term, fully automated, and stable operation, offering an accurate, reliable, and open-source protocol for future researchers. Meanwhile, it was observed that the permeate flux is time and temperature-dependent and highly influenced by multiple operational parameters simultaneously. The characterization process remains inherently sample-destructive and time-consuming, which further underscores the importance and necessity of incorporating simulation and ML techniques to accelerate research in the MemD field. 2. Among all morphological and structural properties studied, the MDyn simulations demonstrate that GA's pore diameter and porosity strongly influence water diffusivity and flux in VMemD, with thermal conductivity inversely related to porosity and tortuosity. GA pore channels featuring smoother surfaces and composed of high-curvature graphene sheets are more favorable for efficient water vapor transport and enhanced phonon scattering. When utilizing GA morphological data to compute water vapor diffusivity, permeate flux, and thermal conductivity, our constructed equations demonstrated comparable or superior predictive accuracy relative to traditional empirical formulas, while maintaining better physical interpretability. 3. The self-developed transformer-enhanced 3D CNN model outperforms conventional CNN architectures, achieving both high accuracy and computational efficiency in predicting the mass and heat transfer properties of GA. This model offers a reliable, scalable, and generalizable platform for future predictive studies, while also providing actionable insights for the structural optimization of membrane materials.

Degree

thesis:*
Level thesis:degree_level
Doctoral
Department dc:contributor.department
Materials Science and Engineering
Grantor dc:publisher
Georgia Institute of Technology
Year dc:date.issued
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Liu, Honglin
Advisors dc:contributor.advisor
  • Wang, Youjiang
  • Jacob, Karl I.
Committee members dc:contributor.committeemember
  • Yao, Donggang
  • Garmestani, Hamid
  • Ghiaasiaan, Seyed Mostafa

Subjects

dc:subject × 9

Rights

Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/1853/82036

Chain of custody

source
Harvested from
Georgia Tech
Base URL
repository.gatech.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Liu, Honglin. Performance Prediction and Optimization of Graphene Aerogel for Membrane Distillation Using Experimental, Molecular Dynamics, and Machine Learning Approaches. Doctoral thesis, Georgia Institute of Technology, 2025. https://hdl.handle.net/1853/82036