Date of Award
May 2020
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Mathematical Sciences
Committee Member
Vincent Ervin
Committee Member
Eleanor Jenkins
Committee Member
Leo Rebholz
Committee Member
Fei Xue
Committee Member
Chris Kees
Abstract
This dissertation explores ways to improve the computational efficiency of linear elasticity and the variable density/viscosity Navier--Stokes equations. While the approaches explored for these two problems are much different in nature, the end goal is the same - to reduce the computational effort required to form reliable numerical approximations.\\
The first topic considered is the axisymmetric linear elasticity problem. While the linear elasticity problem has been studied extensively in the finite-element literature, to the author's knowledge, this is the first study of the elasticity problem in an axisymmetric setting. Indeed, the axisymmetric nature of the problem means that a change of variables to cylindrical coordinates reduces a three-dimensional problem into a decoupled one-dimensional and two-dimensional problem. The change of variables to cylindrical coordinates, however, affects the functional form of the divergence operator and the definition of the inner products. To develop a computational framework for the linear elasticity problem in this context, a new projection operator is defined that is tailored to the cylindrical form of the divergence and inner products. Using this framework, a stable finite-element quadruple is derived for $k=1,2$. These computational rates are then validated with a few computational examples.\\
The second topic addressed in this work is the development of a new Schur complement approach for preconditioning the two-phase Navier--Stokes equations. Considerable research effort has been invested in the development of Schur complement preconditioning techniques for the Navier--Stokes equations, with the pressure-convection diffusion (PCD) operator and the least-squares commutator being among the most popular. Furthermore, more recently researchers have begun examining preconditioning strategies for variable density / viscosity Stokes and Navier--Stokes equations. This work contributes to recent work that has extended the PCD Schur complement approach for single phase flow to the variable phase case. Specifically, this work studies the effectiveness of a new two-phase PCD operator when applied to dynamic two-phase simulations that use the two-phase Navier--Stokes equations. To demonstrate the new two-phase PCD operators effectiveness, results are presented for standard benchmark problems, as well as parallel scaling results are presented for large-scale dynamic simulations for three-dimensional problems.
Recommended Citation
Bentley, Alistair R., "A Computational Framework for Axisymmetric Linear Elasticity and Parallel Iterative Solvers for Two-Phase Navier–Stokes" (2020). All Dissertations. 2638.
https://tigerprints.clemson.edu/all_dissertations/2638