Multifidelity methods for simulation of wax deposition in single-phase and two-phase flow
File(s)
Author(s)
Farah Noroes Goncalves, Gabriel
Type
Thesis
Abstract
The description of wax deposition in the presence of (often turbulent) flow is challenging since it requires the understanding of the fluid dynamics, heat and mass transfer, the thermodynamic effects, and the deposit formation and removal. The length and timescales involved vary over many orders of magnitude making a complete simulation of the full phenomena prohibitively expensive for any industrial situations. Different simplifications may be used, depending on the desired compromise between generality, robustness, speed, and accuracy.
In this work, a set of different modelling techniques for wax deposition based on fundamental methods or data-driven approaches is proposed, implemented, and validated. These included more complete, local descriptions of the flow, namely interface-capturing, graph networks, and thin film models, as well as global models using mechanistic considerations and/or regressions based on data.
An interface-capturing model from the literature is implemented and tested, and, due to its high computational cost, a graph network based data-driven approach is tested for acceleration of the computations. Following that, a simplification based on a thin film assumption is also examined, which can result in significant savings in computational cost. Focusing on a more industrial application, a mechanistic model for pipe flows is proposed and extensively validated with data from single and multiphase flows. Finally, a framework for generation of surrogate models based on an active-learning technique, for complex models, is analyzed.
The models presented here have been validated whenever possible, and as far as possible, against experimental data from the literature and could be used in the design and/or operation of equipment subject to wax deposition (and, more generally, any deposit resulting from a phase change-type transformation), or used in the development of other novel techniques.
In this work, a set of different modelling techniques for wax deposition based on fundamental methods or data-driven approaches is proposed, implemented, and validated. These included more complete, local descriptions of the flow, namely interface-capturing, graph networks, and thin film models, as well as global models using mechanistic considerations and/or regressions based on data.
An interface-capturing model from the literature is implemented and tested, and, due to its high computational cost, a graph network based data-driven approach is tested for acceleration of the computations. Following that, a simplification based on a thin film assumption is also examined, which can result in significant savings in computational cost. Focusing on a more industrial application, a mechanistic model for pipe flows is proposed and extensively validated with data from single and multiphase flows. Finally, a framework for generation of surrogate models based on an active-learning technique, for complex models, is analyzed.
The models presented here have been validated whenever possible, and as far as possible, against experimental data from the literature and could be used in the design and/or operation of equipment subject to wax deposition (and, more generally, any deposit resulting from a phase change-type transformation), or used in the development of other novel techniques.
Version
Open Access
Date Issued
2022-08
Date Awarded
2023-02
Copyright Statement
Creative Commons Attribution NonCommercial Licence
License URL
Advisor
Matar, Omar
Sponsor
Conselho Nacional de Desenvolvimento Científico e Tecnológico
Publisher Department
Chemical Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)