This thesis consists of two parts. We start with some background theory that will be used throughout the thesis. Then, in the first part, we investigate the existence and uniqueness of the solution of the stochastic partial differential equation with two reflecting walls. Then we establish the existence and uniqueness of invariant measure of this equation under some reasonable conditions. In the second part, we study the splitting-up method for approximating the solu- tions of stochastic Stokes equations using resolvent method.

In this thesis our primary interest is in developing adaptive solution methods for parabolic and elliptic partial differential equations. The convection-diffusion equation is used as a representative test problem. Investigations are made into adaptive temporal solvers implementing only a few changes to existing software. This includes a comparison of commercial code against some more academic releases. A novel way to select step sizes for an adaptive BDF2 code is introduced. A chapter is included introducing some functional analysis that is required to understand aspects of the finite element method and error estimation. Two error estimators are derived and proofs of their error bounds are covered. A new finite element package is written, implementing a rather interesting error estimator in one dimension to drive a rather standard refinement/coarsening type of adaptivity. This is compared to a commercially available partial differential equation solver and an investigation into the properties of the two inspires the development of a new method designed to very quickly and directly equidistribute the errors between elements. This new method is not really a refinement technique but doesn't quite fit the traditional description of a moving mesh either. We show that this method is far more effective at equidistribution of errors than a simple moving mesh method and the original simple adaptive method. A simple extension of the new method is proposed that would be a mesh reconstruction method. Finally the new code is extended to solve steady-state problems in two dimensions. The mesh refinement method from one dimension does not offer a simple extension, so the error estimator is used to supply an impression of the local topology of the error on each element. This in turn allows us to develop a new anisotropic refinement algorithm, which is more in tune with the nature of the error on the parent element. Whilst the benefits observed in one dimension are not directly transferred into the two-dimensional case, the obtained meshes seem to better capture the topology of the solution.

We introduce a new framework of generalized operators to handle vector valued distributions, intertwined evolution operators of B-evolution equations and Fokker Planck type evolution equations. Generalized operators capture these operators. The framework is a marriage between vector valued distribution theory and abstract harmonic analysis: a new convolution algebra is the offspring. The new algebra shows that convolution is more fundamental than operator composition. The framework is complete with a Hille-Yosida theorem for implicit evolution equations for generalized operators. Feller semigroups and processes fit perfectly into the framework of generalized operators. Feller semigroups are intertwined by the Chapman Kolmogorov equation. Our framework handles more complex intertwinements which naturally arise from a dynamic boundary approach to an absorbing barrier of a fly trap model: we construct an entwined pseudo Poisson process which is a pair of stochastic processes entwined by the extended Chapman Kolmogorov equation. Similarly, we introduce the idea of an entwined Brownian motion. We show that the diffusion equation of an entwined Brownian motion involves an implicit evolution equation on a suitable scalar test space. We end off by constructing a new convolution of operator valued measures which generalizes the convolution of Feller convolution semigroups. / Thesis (PhD)--University of Pretoria, 2015. / Mathematics and Applied Mathematics / Unrestricted

A presença de onda de choque e vórtices de pequena escala exigem métodos numéricos mais sofisticados para simular escoamentos compressíveis em velocidades altas. Alguns desses métodos produzem resultados adequados para regiões com função suave, embora os mesmos não possam ser utilizados diretamente em regiões com função descontínua, resultando em oscilações espúrias. Dessa forma, métodos foram desenvolvidos para solucionar esse problema, apresentando um bom desempenho para regiões com função descontínua; entretanto, estes possuem termos de alta dissipação. Para evitar os problemas encontrados, foram desenvolvidos os métodos híbridos, onde dois métodos com características ideais para cada região são combinados através de uma função detectora que analisa numericamente a variação de uma quantidade em uma região através de fórmulas que envolvem derivadas. Um detector de descontinuidades foi desenvolvido a partir da revisão bibliográfica de diversos métodos numéricos híbridos existentes, sendo avaliadas as principais desvantagens e limitações de cada um. Diversas comparações entre o novo detector e os detectores de descontinuidades já desenvolvidos foram realizadas através da aplicação em funções unidimensionais e bidimensionais. Finalmente, o método híbrido foi aplicado para a solução das equações de Euler unidimensionais e bidimensionais. / The presence of shock and small-scale vortices require more sophisticated numerical methods to simulate compressible flows at high speeds. Some of these methods produce good results for regions with smooth function, altough they cannot be used directly in regions with discontinuous functions, resulting in spurious oscillations. Thus, methods have been developed to solve this problem, showing a good performance for regions with discontinuous functions; however, these methods contain high dissipation terms. To avoid the problems encountered, hybrid methods have been developed, where two methods with ideal characteristics for each region are combined through a function that analyze numerically the variation of a quantity in the region using formulas involving derivatives. A discontinuity detector was developed from the literature review of several existing hybrid methods, evaluating the main disadvantages and limitations of each. The new detector and other developed discontinuity detectors were compared by applying on one and two-dimensional functions. Finally, the hybrid method was applied fo the solution of one and twodimensional Euler equations.

The study of the long time behavior of nonlinear systems is not effortless, but it is very rewarding. The computation of invariant objects, in particular manifolds provide the scientist with the ability to make predictions at the frontiers of science. However, due to the presence of strong nonlinearities in many important applications, understanding the propagation of errors becomes necessary in order to quantify the reliability of these predictions, and to build sound foundations for future discoveries. This dissertation develops methods for the accurate computation of high-order polynomial approximations of stable/unstable manifolds attached to long periodic orbits in discrete time dynamical systems. For this purpose a multiple shooting scheme is applied to invariance equations for the manifolds obtained using the Parameterization Method developed by Xavier Cabre, Ernest Fontich and Rafael De La Llave in [CFdlL03a, CFdlL03b, CFdlL05]. / Includes bibliography. / Dissertation (Ph.D.)--Florida Atlantic University, 2020. / FAU Electronic Theses and Dissertations Collection

In this thesis a method for studying the asymptotic behavior of solutions to dissipative partial differential equations is developed, motivated by the study of the compressible Navier-Stokes equations in the past works of Hoff and Zumbrun,1995, Hoff and Zumbrun, 1997. In its most basic form, this method allows one to compute n^th order approximations in terms of Hermite functions of solutions of the heat equation having n^th order moments. The main advantage is that these approximations can be efficiently computed, and are often given explicitly in terms of elementary functions. It is shown how this method can be extended to increasingly complicated systems, leading the way toward the asymptotic analysis of the compressible Navier-Stokes equations. A number of challenges must be overcome to apply this method to the compressible Navier-Stokes system. For technical reasons, the analysis is carried out on the divergence and curl of the velocity field, and hence a means of recovering the velocity field from these quantities is established first. The linear part of the evolution is then studied, and an extended version of the artificial viscosity decomposition previously developed (Kawashima, Hoff and Zumbrun1995) is introduced. This decomposition is in terms of the heat and combined heat-wave operators, and hence general estimates on their evolution in weighted L^p spaces are obtained. A modified compressible Navier-Stokes system is then introduced which captures the dominant behavior of the linear evolution and possesses similar nonlinear terms. Solutions to this modified system are proven to exist in weighted spaces, showing that solutions initially having a certain number of moments possess this same number of moments for all time. An analysis of the asymptotic behavior of the modified compressible Navier-Stokes system is then carried out, and it is shown that the method developed herein extends and unifies the approach of Hoff and Zumbrun with that of Gallay and Wayne, 2002a, Gallay and Wayne, 2002b, where it was originally developed to study the behavior of the incompressible Navier-Stokes equations. The thesis is concluded with a discussion of how the results obtained for the modified compressible Navier-Stokes system pave the way for an analysis of the true compressible Navier-Stokes system, the generalization of this asymptotic analysis to arbitrary order, and with a comparison of this asymptotic analysis to that found in the recent work of Kagei and Okita, 2017.

The objective of this project is to get familiar with the numerical solution of partial differential equations. This solution will be implemented by using a grid refinement based on the aposteriory error estimation.

In this thesis we study a parameter identification problem for a stationary diffusion equation posed in heterogeneous media. This problem is closely related to the Calderón problem with anisotropic conductivities. The anisotropic case is particularly difficult and is ill-posed both in regards to uniqueness of solution and stability on the data. Since the present problem is posed in heterogeneous media, we can take advantage of multiscale modelling and the tools of homogenization theory in the study of the inverse problem, unlike the original Calderón problem. We investigate the possibilities of combining the theory of the Calderón problem with homogenization theory in order to obtain a well-posed parameter identification. We find that homogenization theory indeed can be used to make progress towards a well-posed identification of the diffusion coefficient. The success of the method is, however, dependent both on the precise structure of the heterogeneous media and on the modelling of the measurements in the invese problem framework. We have in mind a particular problem formulation which is motivated by an experiment to determine effective coefficients of materials used in food packaging. This experiment comes with a set of requirements on both the heterogeneous media and on the method for making measurements that, unfortunately, are in conflict with the currently available results for well-posedness. We study also an optimization approach to solving the inverse problem under these application specific requirements. Some progress towards well-posedness of the optimization problem is made by proving existence of minimizer, again with homogenization theory playing a key role in obtaining the result. In a proof-of-concept computational study this optimization approach is implemented and compared to two other optimization problems. For the two tested heterogeneous media, the only optimization method that manages to identify reasonably well the diffusion coefficient is the one which makes use of homogenization theory.

Abstract We are interested in studying Gaussian Markov random fields as correlation priors for Bayesian inversion. We construct the correlation priors to be discretisation-invariant, which means, loosely speaking, that the discrete priors converge to continuous priors at the discretisation limit. We construct the priors with stochastic partial differential equations, which guarantees computational efficiency via sparse matrix approximations. The stationary correlation priors have a clear statistical interpretation through the autocorrelation function. We also consider how to make structural model of an unknown object with anisotropic and inhomogeneous Gaussian Markov random fields. Finally we consider these fields on unstructured meshes, which are needed on complex domains. The publications in this thesis contain fundamental mathematical and computational results of correlation priors. We have considered one application in this thesis, the electrical impedance tomography. These fundamental results and application provide a platform for engineers and researchers to use correlation priors in other inverse problem applications.

<div>This thesis compiles my work on three projects.</div><div>The first project studies the cancellation of singularities in the inversion of two X-ray type transforms in the presence of conjugate points. The second project studies the recovery of singularities for the weighted cone transform. The third project studies the phenomenon of Rayleigh waves and Stoneley waves in the isotropic elastic wave equation of variable coefficients with a curved boundary.</div>