In this dissertation we discuss bilinear immersed finite elements (IFE) for solving interface problems. The related research works can be categorized into three aspects: (1) the construction of the bilinear immersed finite element spaces; (2) numerical methods based on these IFE spaces for solving interface problems; and (3) the corresponding error analysis. All of these together form a solid foundation for the bilinear IFEs. The research on immersed finite elements is motivated by many real world applications, in which a simulation domain is often formed by several materials separated from each other by curves or surfaces while a mesh independent of interface instead of a body-fitting mesh is preferred. The bilinear IFE spaces are nonconforming finite element spaces and the mesh can be independent of interface. The error estimates for the interpolation of a Sobolev function in a bilinear IFE space indicate that this space has the usual approximation capability expected from bilinear polynomials, which is <i>O</i>(<i>h</i>²) in <i>L</i>² norm and <i>O</i>(<i>h</i>) in <i>H</i>¹ norm. Then the immersed spaces are applied in Galerkin, finite volume element (FVE) and discontinuous Galerkin (DG) methods for solving interface problems. Numerical examples show that these methods based on the bilinear IFE spaces have the same optimal convergence rates as those based on the standard bilinear finite element for solutions with certain smoothness. For the symmetric selective immersed discontinuous Galerkin method based on bilinear IFE, we have established its optimal convergence rate. For the Galerkin method based on bilinear IFE, we have also established its convergence. One of the important advantages of the discontinuous Galerkin method is its flexibility for both <i>p</i> and <i>h</i> mesh refinement. Because IFEs can use a mesh independent of interface, such as a structured mesh, the combination of a DG method and IFEs allows a flexible adaptive mesh independent of interface to be used for solving interface problems. That is, a mesh independent of interface can be refined wherever needed, such as around the interface and the singular source. We also develop an efficient selective immersed discontinuous Galerkin method. It uses the sophisticated discontinuous Galerkin formulation only around the locations needed, but uses the simpler Galerkin formulation everywhere else. This selective formulation leads to an algebraic system with far less unknowns than the immersed DG method without scarifying the accuracy; hence it is far more efficient than the conventional discontinuous Galerkin formulations. / Ph. D.

published_or_final_version / Mechanical Engineering / Doctoral / Doctor of Philosophy

Near shore hydrodynamics has been an important research area dealing with coastal processes. The nearshore coastal region is the region between the shoreline and a fictive offshore limit which usually is defined as the limit where the depth becomes so large that it no longer influences the waves. This spatially limited but highly energetic zone is where water waves shoal, break and transmit energy to the shoreline and are governed by highly dispersive and non-linear effects. An accurate understanding of this phenomena is extremely useful, especially in emergency situations during hurricanes and storms. While the shallow water assumption is valid in regions where the characteristic wavelength exceeds a typical depth by orders of magnitude, Boussinesq-type equations have been used to model near-shore wave motion. Unfortunately these equations are complex system of coupled non-linear and dispersive differential equations that have made the developement of numerical approximations extremely challenging. In this dissertation, a local discontinuous Galerkin method for Boussinesq-Green Naghdi Equations is presented and validated against experimental results. Currently Green-Naghdi equations have many variants. We develop a numerical method in one horizontal dimension for the Green-Naghdi equations based on rotational characteristics in the velocity field. Stability criterion is also established for the linearized Green-Naghdi equations and a careful proof of linear stability of the numerical method is carried out. Verification is done against a linearized standing wave problem in flat bathymetry and h,p (denoted by K in this thesis) error rates are plotted. The numerical method is validated with experimental data from dispersive and non-linear test cases. / text

<p>Domain decomposition method provides a solution for the very large electromagnetic</p><p>system which are impossible for single domain methods. Discontinuous Galerkin</p><p>(DG) method can be viewed as an extreme version of the domain decomposition,</p><p>i.e., each element is regarded as one subdomain. The whole system is solved element</p><p>by element, thus the inversion of the large global system matrix is no longer necessary,</p><p>and much larger system can be solved with the DG method compared to the</p><p>continuous Galerkin (CG) method.</p><p>In this work, the DG method is implemented on a subdomain level, that is, each subdomain contains multiple elements. The numerical flux only applies on the</p><p>interfaces between adjacent subdomains. The subodmain level DG method divides</p><p>the original large global system into a few smaller ones, which are easier to solve,</p><p>and it also provides the possibility of parallelization. Compared to the conventional</p><p>element level DG method, the subdomain level DG has the advantage of less total</p><p>DoFs and fexibility in interface choice. In addition, the implicit time stepping is </p><p>relatively much easier for the subdomain level DG, and the total CPU time can be</p><p>much less for the electrically small or multiscale problems.</p><p>The hybrid of elements are employed to reduce the total DoF of the system.</p><p>Low-order tetrahedrons are used to catch the geometry ne parts and high-order</p><p>hexahedrons are used to discretize the homogeneous and/or geometry coarse parts.</p><p>In addition, the non-conformal mesh not only allow dierent kinds of elements but</p><p>also sharp change of the element size, therefore the DoF can be further decreased.</p><p>The DGTD method in this research is based on the EB scheme to replace the</p><p>previous EH scheme. Dierent from the requirement of mixed order basis functions</p><p>for the led variables E and H in the EH scheme, the EB scheme can suppress the</p><p>spurious modes with same order of basis functions for E and B. One order lower in</p><p>the basis functions in B brings great benets because the DoFs can be signicantly</p><p>reduced, especially for the tetrahedrons parts.</p><p>With the basis functions for both E and B, the EB scheme upwind </p><p>ux and</p><p>EB scheme Maxwellian PML, the eigen-analysis and numerical results shows the</p><p>eectiveness of the proposed DGTD method, and multiscale problems are solved</p><p>eciently combined with the implicit-explicit hybrid time stepping scheme and multiple</p><p>kinds of elements.</p><p>The EB scheme DGTD method is further developed to allow arbitrary anisotropic</p><p>media via new anisotropic EB scheme upwind </p><p>ux and anisotropic EB scheme</p><p>Maxwellian PML. The anisotropic M-PML is long time stable and absorb the outgoing</p><p>wave eectively. A new TF/SF boundary condition is brought forward to</p><p>simulate the half space case. The negative refraction in YVO4 bicrystal is simulated</p><p>with the anisotropic DGTD and half space TF/SF condition for the rst time with</p><p>numerical methods.</p> / Dissertation

This work deals with numerical modeling of electromagnetic induction in 3D environment with heterogeneous conductivity. We develop a program to solve Maxwell's equations in quasistatic approximation by using Continuous and Discontinuous Finite Elements. Their implementation in the numerical library deal.ii is discussed. The obtained numerical results are compared with each other and also with a quasianalytic solution for an environment with 1D heterogeneous conductivity. We discuss different numerical methods, limits of our code for practical use and possible future enhancements.

Doctor en Ciencias de la Ingeniería, Mención Modelación Matemática / En cotutela con la Universidad de Borgoña Franco-Condado / Esta tesis está dedicada al estudio de inclusiones diferenciales con conos normales de conjuntos no regulares en espacios de Hilbert. En particular, nos interesa el proceso de arrastre y sus variantes. El proceso de arrastre es una inclusión diferencial restringida con conos normales que aparece naturalmente en varias aplicaciones tales como elastoplasticidad, histéresis, circuitos eléctricos, movimiento de multitudes, etc. Este trabajo está dividido conceptualmente en tres partes: Estudio de los conjuntos "alpha-far'', existencia de soluciones para las inclusiones diferenciales con conos normales y caracterizaciones de los pares de Lyapunov para el proceso de arrastre en espacios de Hilbert separable. En la primera parte (Capítulo 2), investigamos la clase de conjuntos positivamente "alpha-far''. Esta clase de conjuntos no regulares es muy general e incluye los conjuntos convexos, uniformemente prox-regulares y uniformemente sub-lisos, entre otros. Esta clase de conjuntos es la mejor adaptada al estudio de inclusiones diferenciales con conos normales. En la segunda parte (Capítulo 3 hasta la primera parte del Capítulo 8), se entregan varios resultados de existencia para el proceso de arrastre y sus variantes. Para ello, consideramos tres enfoques: el algoritmo de rectificación (Catching-up algorithm), el método de tipo Galerkin y la regularización de Moreau-Yosida. El primer método es el más clásico en el estudio de inclusiones diferenciales gobernadas por conos normales. Aquí es utilizado en el caso donde el conjunto considerado es fijo. El segundo método (de tipo Galerkin) consiste en aproximar el problema original proyectando el estado sobre un espacio de Hilbert de dimensión finita, pero no la velocidad. Los problemas aproximados siempre tienen una solución y, bajo ciertas condiciones de compacidad, se demuestra que ellos convergen fuertemente (salvo subsucesión) a una solución de la inclusión diferencial original. Más aún, se muestra que este método está bien adaptado para tratar inclusiones diferenciales con conos normales, proporcionando resultados generales de existencia para el proceso de arrastre generalizado. En consecuencia, se obtiene la existencia de soluciones para el proceso de arrastre de primer y segundo orden. Adicionalmente, este método es utilizado para mostrar la existencia de soluciones del proceso de arrastre con condiciones iniciales no locales. El tercer método es la técnica de regularización de Moreau-Yosida que consiste en aproximar una inclusión diferencial por una penalizada, en función de un parámetro positivo, para luego pasar al límite cuando el parámetro tiende a cero. Este método es utilizado para tratar el proceso de arrastre dependiente del estado gobernado por conjuntos uniformemente sub-lisos. Finalmente, en la tercera parte (segunda parte del Capítulo 8 y Capítulo 9), se proporcionan algunas caracterizaciones de los pares de Lyapunov débiles y la invariancia débil para el proceso de arrastre perturbado con conjuntos uniformemente sub-lisos. / Este trabajo ha sido parcialmente financiado por CONICYT-Beca Doctorado Nacional 2013.

Tesis para optar al grado de Magíster en Ciencias de la Ingeniería, Mención Mecánica / En la integración numérica de los métodos de Galerkin sin malla, debido a la complejidad de las funciones de forma, es necesario utilizar una gran cantidad de puntos de integración para lograr que el método sea estable, lo que aumenta los tiempos de cómputo. Por otro lado, la integración directa en el nodo puede ser deseable porque se basa en menos evaluaciones de puntos de integración, pero conduce a inestabilidad numérica debido a un mecanismo similar a la subintegración y al desvanecimiento de las derivadas de las funciones de base en los nodos. En este trabajo se propone un esquema de integración nodal consistente y estable para el método de Galerkin sin malla para problemas de mecánica de sólidos lineal. Para el desarrollo de este esquema se utiliza la descomposición del elemento virtual, la cual fue previamente desarrollada para afrontar problemas de integración numérica en elementos poligonales. El método propuesto en esta tesis se ha denominado NIVED (Nodal Integration using the Virtual Element Descomposition). La integración nodal se evalúa sobre las celdas representativas para cada nodo, basadas en diagramas de Voronoi o en polígonos construidos a partir de mallas de triángulos, donde el centroide de los triángulos representan los vértices de los polígonos. En esta tesis, el esquema se implementa utilizando las funciones de base de la máxima entropía. Para estudiar y demostrar la precisión y la robustez del método de integración nodal se implementan varios problemas de referencia en elastostática y elastodinámica lineal bidimensional. Adicionalmente los problemas estáticos se comparan con el desempeño de un método de Galerkin sin malla utilizando integración de Gauss y el problema dinámico con el desempeño del método del punto material. Se demuestra que el esquema propuesto satisface el test de la parcela lineal entregando error de máquina. Finalmente NIVED demostró ser un esquema consistente y estable.

Este trabalho consiste de uma análise particularizada sobre diferenças finitas. Nele procurou-se desenvolver uma metodologia específica , trabalhando-se com splines definidos em um espaço de Hilbert e e associados a distribuições n pertencentes ao seu dual e'; do tipo ~n = I Cp 0p que : de tal modo metodologia foi desenvolvida, adotando-se em particular e como sendo um espaço de Sobolev, associado a um operador linear L, tendo como ênfase o desenvolvimento de métodos para a obtenção de splines os quais permitiram a obtenção de esquemas projecionais de diferenças finitas, que levam a valores nodais exatos da solução do problema de contorno. Como um enfoque complementar, foi analisada a aplicação de métodos projecionais análogos ao método de Galerkin, com a preocupação da geração de esquemas em diferenças, os quais podem em alguns casos levar a valores exatos para a solução do problema de contorno em estudo XI

Mak, Tsz Fan. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2013. / Includes bibliographical references (leaves 33-36). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstracts also in Chinese.

在本文中，我們為了三維空間的馬克士威方程組(Maxwell’s equation)制定和分析了一套新種類的交錯間斷伽遼金(discontinuous Galerkin)方法，同時考慮了時間依賴性和時間諧波的馬克士威方程組。我們用了空間離散上交錯笛卡兒網格，這種方法具有許多良好的性質。首先，我們的方法所得出的數值解保留了電磁能量，並自動符合了高斯定律的離散版本。第二，質量矩陣是對角矩陣，從而時間推進是顯式和非常有效的。第三，我們的方法是高階準確，最佳收斂性在這裏會被嚴格地證明。第四，基於笛卡兒網格，它也很容易被執行，並可視為是典型的Yee’s Scheme的以及四邊形的邊有限元的推廣。最後，超收斂結果也會在這裏被證明。 / 在本文中，我們還提供了幾個數值結果驗證了理論的陳述。我們計算了時間依賴性和時間諧波的馬克士威方程組數值收斂結果。此外，我們計算時間諧波馬克士威方程組特徵值問題的數值特徵值，並與理論特徵值比較結果。最後，完美匹配層(Perfect Matching Layer)吸收邊界的問題也有實行其數值結果。 / We develop and analyze a new type of staggered discontinuous Galerkin methods for the three dimensional Maxwell’s equations in this paper. Both time-dependent and time-harmonic Maxwell’s equations are considered. The spatial discretization is based on staggered Cartesian grids which possess many good properties. First of all, our method has the advantages that the numerical solution preserves the electromagnetic energy and automatically fulfills a discrete version of the Gauss law. Second, the mass matrices are diagonal, thus time marching is explicit and is very efficient. Third, our method is high order accurate and the optimal order of convergence is rigorously proved. Fourth, it is also very easy to implement due to its Cartesian structure and can be regarded as a generalization of the classical Yee’s scheme as well as the quadrilateral edge finite elements. Lastly, a superconvergence result, that is the convergence rate is one order higher at interpolation nodes, is proved. / In this paper, we also provide several numerical results to verify the theoretical statements. We compute the numerical convergence order using L2-norm and discrete-norm respectively for both the time-dependent and time-harmonic Maxwell’s equations. Also, we compute the numerical eigenvalues for the time-harmonic eigenvalue problem and compare the result with the theoretical eigenvalues. Lastly, applications to problems in unbounded domains with the use of PML are also presented. / Detailed summary in vernacular field only. / Detailed summary in vernacular field only. / Yu, Tang Fei. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2012. / Includes bibliographical references (leaves 46-49). / Abstracts also in Chinese. / Chapter 1 --- Introduction and Model Problems --- p.1 / Chapter 2 --- Staggered DG Spaces --- p.4 / Chapter 2.1 --- Review on Gauss-Radau and Gaussisan points --- p.5 / Chapter 2.2 --- Basis functions --- p.6 / Chapter 2.3 --- Finite Elements space --- p.7 / Chapter 3 --- Method derivation --- p.14 / Chapter 3.1 --- Method --- p.14 / Chapter 3.2 --- Time discretization --- p.17 / Chapter 4 --- Energy conservation and Discrete Gauss law --- p.19 / Chapter 4.1 --- Energy conservation --- p.19 / Chapter 4.2 --- Discrete Gauss law --- p.22 / Chapter 5 --- Error analysis --- p.24 / Chapter 6 --- Numerical examples --- p.29 / Chapter 6.1 --- Convergence tests --- p.30 / Chapter 6.2 --- Diffraction by a perfectly conducting object --- p.30 / Chapter 6.3 --- Perfectly matched layers --- p.37 / Chapter 7 --- Time Harmonic Maxwell’s equations --- p.40 / Chapter 7.1 --- Model Problems --- p.40 / Chapter 7.2 --- Numerical examples --- p.40 / Chapter 7.2.1 --- Convergence tests --- p.41 / Chapter 7.2.2 --- Eigenvalues tests --- p.41 / Chapter 8 --- Conclusion --- p.45 / Bibliography --- p.46