The physical processes in lakes remain only partially understood despite successful data collection from a variety of sources spanning several decades. Although numerical models are already frequently employed to simulate the physics of lakes, especially in the context of water quality management, improved methods are necessary to better capture the wide array of dynamically important physical processes, spanning length scales from ~ 10 km (basin-scale oscillations) - 1 m (short internal waves). In this thesis, high-order numerical methods are explored for specialized model equations of lakes, so that their use can be taken into consideration in the next generation of more sophisticated models that will better capture important small scale features than their present day counterparts. The full three-dimensional incompressible density-stratified Navier-Stokes equations remain too computationally expensive to be solved for situations that involve both complicated geometries and require resolution of features at length-scales spanning four orders of magnitude. The main source of computational expense lay with the requirement of having to solve a three-dimensional Poisson equation for pressure at every time-step. Simplified model equations are thus the only way that numerical lake modelling can be carried out at present time, and progress can be made by seeking intelligent parameterizations as a means of capturing more physics within the framework of such simplified equation sets. In this thesis, we employ the long-accepted practice of sub-dividing the lake into vertical layers of different constant densities as an approximation to continuous vertical stratification. We build on this approach by including weakly non-hydrostatic dispersive correction terms in the model equations in order to parameterize the effects of small vertical accelerations that are often disregarded by operational models. Favouring the inclusion of weakly non-hydrostatic effects over the more popular hydrostatic approximation allows these models to capture the emergence of small-scale internal wave phenomena, such as internal solitary waves and undular bores, that are missed by purely hydrostatic models. The Fourier and Chebyshev pseudospectral methods are employed for these weakly non-hydrostatic layered models in simple idealized lake geometries, e.g., doubly periodic domains, periodic channels, and annular domains, for a set of test problems relevant to lake dynamics since they offer excellent resolution characteristics at minimal memory costs. This feature makes them an excellent benchmark to compare other methods against. The Discontinuous Galerkin Finite Element Method (DG-FEM) is then explored as a mid- to high-order method that can be used in arbitrary lake geometries. The DG-FEM can be interpreted as a domain-decomposition extension of a polynomial pseudospectral method and shares many of the same attractive features, such as fast convergence rates and the ability to resolve small-scale features with a relatively low number of grid points when compared to a low-order method. The DG-FEM is further complemented by certain desirable attributes it shares with the finite volume method, such as the freedom to specify upwind-biased numerical flux functions for advection-dominated flows, the flexibility to deal with complicated geometries, and the notion that each element (or cell) can be regarded as a control volume for conserved fluid quantities. Practical implementation details of the numerical methods used in this thesis are discussed, and the various modelling and methodology choices that have been made in the course of this work are justified as the difficulties that these choices address are revealed to the reader. Theoretical calculations are intermittently carried out throughout the thesis to help improve intuition in situations where numerical methods alone fall short of giving complete explanations of the physical processes under consideration. The utility of the DG-FEM method beyond purely hyperbolic systems is also a recurring theme in this thesis. The DG-FEM method is applied to dispersive shallow water type systems as well as incompressible flow situations. Furthermore, it is employed for eigenvalue problems where orthogonal bases must be constructed from the eigenspaces of elliptic operators. The technique is applied to the problem calculating the free modes of oscillation in rotating basins with irregular geometries where the corresponding linear operator is not self-adjoint.

<p>The present dissertation focuses on the numerical solution of compressible flows with an emphasis on simulations of transitional hypersonic boundary layers. Initially, general concepts such as the governing equations, numerical approximations and theoretical modeling strategies are addressed. These are used as a basis to introduce two innovative techniques, the Quasi-Spectral Viscosity (QSV) method, applied to high-order finite difference settings and the Legendre Spectral Viscosity (LSV) approach, used in high-order flux reconstruction schemes. Such techniques are derived based on the mathematical formalism of the filtered compressible Navier-Stokes equations. While the latter perspective is only typically used for turbulence modeling in the context of Large-Eddy Simulations (LES), both the QSV and LSV subfilter scale (SFS) closure models are capable of performing simulations in the presence of shock-discontinuities. On top of that, the QSV approach is also shown to support dynamic subfilter turbulence modeling capabilities.</p> <p>QSV’s innovation lies in the introduction of a physical-space implementation of a spectral-like subfilter scale (SFS) dissipation term by leveraging residuals of filter operations, achiev- ing two goals: (1) estimating the energy of the resolved solution near the grid cutoff; (2) imposing a plateau-cusp shape to the spectral distribution of the added dissipation. The QSV approach was tested in a variety of flows to showcase its capability to act interchangeably as a shock capturing method or as a SFS turbulence closure. QSV performs well compared to previous eddy-viscosity closures and shock capturing methods. In a supersonic TGV flow, a case which exhibits shock/turbulence interactions, QSV alone outperforms the simple super- position of separate numerical treatments for SFS turbulence and shocks. QSV’s combined capability of simulating shocks and turbulence independently, as well as simultaneously, effectively achieves the unification of shock capturing and Large-Eddy Simulation.</p> <p>The LSV method extends the QSV idea to discontinuous numerical schemes making it suitable for unstructured solvers. LSV exploits the set of hierarchical basis functions formed by the Legendre polynomials to extract the information on the energy content near the resolution limit and estimate the overall magnitude of the required SFS dissipative terms, resulting in a scheme that dynamically activates only in cells where nonlinear behavior is important. Additionally, the modulation of such terms in the Legendre spectral space allows for the concentration of the dissipative action at small scales. The proposed method is tested in canonical shock-dominated flow setups in both one and two dimensions. These include the 1D Burgers’ problem, a 1D shock tube, a 1D shock-entropy wave interaction, a 2D inviscid shock-vortex interaction and a 2D double Mach reflection. Results showcase a high-degree of resolution power, achieving accurate results with a small number of degrees of freedom, and robustness, being able to capture shocks associated with the Burgers’ equation and the 1D shock tube within a single cell with discretization orders 120 and higher.</p> <p>After the introduction of these methods, the QSV-LES approach is leveraged to perform numerical simulations of hypersonic boundary layer transition delay on a 7^◦-half-angle cone for both sharp and 2.5 mm-nose tip radii due to porosity representative of carbon-fibre-reinforced carbon-matrix ceramics (C/C) in the Reynolds number range Re_m = 2.43 · 106 – 6.40 · 10⁶ m⁻¹ at the freestream Mach number of M_∞ = 7.4. A low-order impedance model was fitted through experimental measurements of acoustic absorption taken at discrete frequencies yielding a continuous representation in the frequency domain that was imposed in the simulations via a broadband time domain impedance boundary condition (TDIBC). The stability of the base flow is studied over impermeable and porous walls via pulse-perturbed axisymmetric simulations with second-mode spatial growth rates matching linear predictions. This shows that the QSV-LES approach is able to dynamically deactivate its dissipative action in laminar portions of the flow making it possible to accurately capture the boundary layer’s instability dynamics. Three-dimensional transitional LES were then performed with the introduction of grid independent pseudorandom pressure perturbations. Comparison against previous experiments were made regarding the frequency content of the disturbances in the transitional region with fairly good agreement capturing the shift to lower frequencies. Such shift is caused by the formation of near-wall low-temperature streaks that concentrate the pressure disturbances at locations with locally thicker boundary layers forming trapped wavetrains that can persist into the turbulent region. Additionally, it is shown that the presence of a porous wall representative of a C/C material does not affect turbulence significantly and simply shifts its onset downstream.</p>

Cette thèse s'intéresse à la conception et à l'analyse de méthodes de discrétisation hybrides pour les inégalités variationnelles non linéaires apparaissant en mécanique des fluides et des solides. Les principaux avantages de ces méthodes sont la conservation locale au niveau des mailles, la robustesse par rapport à différents régimes de paramètres et la possibilité d’utiliser des maillages polygonaux / polyédriques avec des nœuds non coïncidants, ce qui est très intéressant dans le contexte de l’adaptation de maillage. Les méthodes de discrétisation hybrides sont basées sur des inconnues discrètes attachées aux faces du maillage. Des inconnues discrètes attachées aux mailles sont également utilisées, mais elles peuvent être éliminées localement par condensation statique. Deux applications principales des discrétisations hybrides sont abordées dans cette thèse. La première est le traitement par la méthode de Nitsche du problème de contact de Signorini (dans le cas scalaire) avec une non-linéarité dans les conditions aux limites. Nous prouvons des estimations d'erreur optimales conduisant à des taux de convergence d'erreur d'énergie d'ordre (k + 1), si des polynômes de face de degré k >= 0 sont utilisés. La deuxième application principale concerne les fluides à seuil viscoplastiques. Nous concevons une méthode de Lagrangien augmenté discrète appliquée à la discrétisation hybride. Nous exploitons la capacité des méthodes hybrides d’utiliser des maillages polygonaux avec des nœuds non coïncidants afin d'effectuer l’adaptation de maillage local et mieux capturer la surface limite. La précision et la performance des schémas sont évaluées sur des cas tests bidimensionnels, y compris par des comparaisons avec la littérature / This thesis is concerned with the devising and the analysis of hybrid discretization methods for nonlinear variational inequalities arising in computational mechanics. Salient advantages of such methods are local conservation at the cell level, robustness in different regimes and the possibility to use polygonal/polyhedral meshes with hanging nodes, which is very attractive in the context of mesh adaptation. Hybrid discretizations methods are based on discrete unknowns attached to the mesh faces. Discrete unknowns attached to the mesh cells are also used, but they can be eliminated locally by static condensation. Two main applications of hybrid discretizations methods are addressed in this thesis. The first one is the treatment using Nitsche's method of Signorini's contact problem (in the scalar-valued case) with a nonlinearity in the boundary conditions. We prove optimal error estimates leading to energy-error convergence rates of order (k+1) if face polynomials of degree k >= 0 are used. The second main application is on viscoplastic yield flows. We devise a discrete augmented Lagrangian method applied to the present hybrid discretization. We exploit the capability of hybrid methods to use polygonal meshes with hanging nodes to perform local mesh adaptation and better capture the yield surface. The accuracy and performance of the present schemes is assessed on bi-dimensional test cases including comparisons with the literature

<p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="mso-ansi-language: EN-US;" lang="EN-US"><span style="font-size: small;"><span style="font-family: Times New Roman;">This thesis consists of two different parts, related to two different fields in mathematical physics: wave propagation and kinetic theory of gases. Various mathematical and computational problems for equations from these areas are treated.</span></span></span></p><p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="mso-ansi-language: EN-US;" lang="EN-US"><span style="font-size: small; font-family: Times New Roman;"> </span></span></p><p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="mso-ansi-language: EN-US;" lang="EN-US"><span style="font-size: small;"><span style="font-family: Times New Roman;">The first part is devoted to high order finite difference methods for the Helmholtz equation and the wave equation. Compact schemes with high order accuracy are obtained from an investigation of the function derivatives in the truncation error. With the help of the equation itself, it is possible to transfer high order derivatives to lower order or to transfer time derivatives to space derivatives. For the Helmholtz equation, a compact scheme based on this principle is compared to standard schemes and to deferred correction schemes, and the characteristics of the errors for the different methods are demonstrated and discussed. For the wave equation, a finite difference scheme with fourth order accuracy in both space and time is constructed and applied to a problem in discontinuous media.</span></span></span></p><p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="mso-ansi-language: EN-US;" lang="EN-US"><span style="font-size: small; font-family: Times New Roman;"> </span></span></p><p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="mso-ansi-language: EN-US;" lang="EN-US"><span style="font-size: small;"><span style="font-family: Times New Roman;">The second part addresses some problems related to kinetic equations. A direct simulation Monte-Carlo method is constructed for the Landau-Fokker-Planck equation, and numerical tests are performed to verify the accuracy of the algorithm. A formal derivation of the method from the Boltzmann equation with grazing collisions is performed. The linear and linearized Boltzmann collision operators for the hard sphere molecular model are studied using exact reduction of integral equations to ordinary differential equations. It is demonstrated how the eigenvalues of the operators are found from these equations, and numerical values are computed. A proof of existence of non-zero discrete eigenvalues is given. The ordinary diffential equations are also used for investigation of the Chapman-Enskog distribution function with respect to its asymptotic behavior.</span></span></span></p><p class="MsoNormal" style="margin: 0cm 0cm 0pt;"><span style="mso-ansi-language: EN-US;" lang="EN-US"><span style="font-size: small; font-family: Times New Roman;"> </span></span></p>

The construction of compact high-order Residual Distribution schemes for the discretizationof steady multidimensional advection-diffusion problems on unstructuredgrids is presented. Linear and non-linear scheme are considered. A piecewise continuouspolynomial approximation of the solution is adopted and a gradient reconstructionprocedure is used in order to have a continuous representation of both thenumerical solution and its gradient. It is shown that the gradient must be reconstructedwith the same accuracy of the solution, otherwise the formal accuracy ofthe numerical scheme is lost in applications in which diffusive effects prevail overthe advective ones, and when advection and diffusion are equally important. Thenthe method is extended to systems of equations, with particular emphasis on theNavier-Stokes and RANS equations. The accuracy, efficiency, and robustness of theimplicit RD solver is demonstrated using a variety of challenging aerodynamic testproblems.

This thesis consists of two different parts, related to two different fields in mathematical physics: wave propagation and kinetic theory of gases. Various mathematical and computational problems for equations from these areas are treated.   The first part is devoted to high order finite difference methods for the Helmholtz equation and the wave equation. Compact schemes with high order accuracy are obtained from an investigation of the function derivatives in the truncation error. With the help of the equation itself, it is possible to transfer high order derivatives to lower order or to transfer time derivatives to space derivatives. For the Helmholtz equation, a compact scheme based on this principle is compared to standard schemes and to deferred correction schemes, and the characteristics of the errors for the different methods are demonstrated and discussed. For the wave equation, a finite difference scheme with fourth order accuracy in both space and time is constructed and applied to a problem in discontinuous media.   The second part addresses some problems related to kinetic equations. A direct simulation Monte-Carlo method is constructed for the Landau-Fokker-Planck equation, and numerical tests are performed to verify the accuracy of the algorithm. A formal derivation of the method from the Boltzmann equation with grazing collisions is performed. The linear and linearized Boltzmann collision operators for the hard sphere molecular model are studied using exact reduction of integral equations to ordinary differential equations. It is demonstrated how the eigenvalues of the operators are found from these equations, and numerical values are computed. A proof of existence of non-zero discrete eigenvalues is given. The ordinary diffential equations are also used for investigation of the Chapman-Enskog distribution function with respect to its asymptotic behavior.

Le but de cette thèse est de développer et d’analyser les méthodes Hybrides d’Ordre Élevé (HHO: Hybrid High-Order, en anglais) pour des problèmes d’interfaces. Nous nous intéressons à deux types d’interfaces (i) les interfaces diffuses, et (ii) les interfaces traitées comme frontières internes du domaine computationnel. La première moitié de ce manuscrit est consacrée aux interfaces diffuses, et plus précisément aux célèbres équations de Cahn–Hilliard qui modélisent le processus de séparation de phase par lequel les deux composants d’un fluide binaire se séparent pour former des domaines purs en chaque composant. Dans la deuxième moitié, nous considérons des modèles à dimension hybride pour la simulation d’écoulements de Darcy et de transports passifs en milieu poreux fracturé, dans lequel la fracture est considérée comme un hyperplan (d’où le terme hybride) qui traverse le domaine computationnel. / The purpose of this Ph.D. thesis is to design and analyse Hybrid High-Order (HHO) methods on some interface problems. By interface, we mean (i) diffuse interface, and (ii) interface as an immersed boundary. The first half of this manuscrit is dedicated to diffuse interface, more precisely we consider the so called Cahn–Hilliard problem that models the process of phase separation, by which the two components of a binary fluid spontaneously separate and form domains pure in each component. In the second half, we deal with the interface as an immersed boundary and consider a hybrid dimensional model for the simulation of Darcy flows and passive transport in fractured porous media, in which the fracture is considered as an hyperplane that crosses our domain of interest.

Orientador: Marco Lucio Bittencourt / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Mecanica / Made available in DSpace on 2018-08-10T12:57:32Z (GMT). No. of bitstreams: 1 Vazquez_ThaisGodoy_D.pdf: 11719751 bytes, checksum: c6d385d6a6414705c9f468358b8d3bea (MD5) Previous issue date: 2008 / Resumo: Este trabalho tem por objetivo principal o desenvolvimento de funções de interpolaçao e regras de integraçao tensorizaveis para o Metodo dos Elementos Finitos (MEF) de alta ordem hp, considerando os sistemas de referencias locais dos elementos. Para isso, primeiramente, determinam-se ponderaçoes especficas para as bases de funçoes de triangulos e tetraedros, formada pelo produto tensorial de polinomios de Jacobi, de forma a se obter melhor esparsidade e condicionamento das matrizes de massa e rigidez dos elementos. Alem disso, procuram-se novas funçoes de base para tornar as matrizes de massa e rigidez mais esparsas possiveis. Em seguida, escolhe-se os pontos de integraçao que otimizam o custo do calculo dos coeficientes das matrizes de massa e rigidez usando as regras de quadratura de Gauss-Jacobi, Gauss-Radau-Jacobi e Gauss-Lobatto-Jacobi. Por fim, mostra-se a construçao de uma base unidimensional nodal que permite obter uma matriz de rigidez praticamente diagonal para problemas de Poisson unidimensionais. Discute-se ainda extensoes para elementos bi e tridimensionais / Abstract: The main purpose of this work is the development of tensor-based interpolation functions and integration rules for the hp High-order Finite Element Method (FEM), considering the local reference systems of the elements. We first determine specific weights for the shape functions of triangles and tetrahedra, constructed by the tensorial product of Jacobi polynomials, aiming to obtain better sparsity and numerical conditioning for the mass and stiffness matrices of the elements. Moreover, new shape functions are proposed to obtain more sparse mass and stiffness matrices. After that, integration points are chosen that optimize the cost for the calculation of the coefficients of the mass and stiffness matrices using the rules of quadrature of Gauss-Jacobi, Gauss-Radau-Jacobi and Gauss-Lobatto-Jacobi. Finally, we construct an one-dimensional nodal shape function that obtains an almost diagonal stiffness matrix for the 1D Poisson problem. Extensions to two and three-dimensional elements are discussed. / Doutorado / Mecanica dos Sólidos e Projeto Mecanico / Doutor em Engenharia Mecânica

Orientador: Marco Lucio Bittencourt / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Mecanica / Made available in DSpace on 2018-08-13T18:20:59Z (GMT). No. of bitstreams: 1 Bargos_FabianoFernandes_M.pdf: 7838730 bytes, checksum: fc693b4a6996fada9f50dfaa2a0a102b (MD5) Previous issue date: 2009 / Resumo: Esse trabalho apresenta uma implementação, em ambiente MatLab, de códigos para o Método dos Elementos Finitos de Alta Ordem em malhas estruturadas e não estruturadas para aplicação em problemas 2D e 3D. Apresenta-se um resumo dos procedimentos para construção das bases de funções para quadrados, triângulos, hexaedros e tetraedros através do produto tensorial. Faz-se um estudo detalhado da continuidade C0 da aproximação para expansões modais em quadrados e mostra-se que com uma numeração adequada das funções de aresta a continuidade é automaticamente obtida. Por fim, através da imposição de uma solução analítica, analisam-se os problemas de projeção e Poisson, 2D e 3D, em malhas de quadrados, triângulos e hexaedros, para refinamentos h e p / Abstract: An implementation in MatLab environment of a code for the High Order Finite Element Method on structured and non-structured mesh for 2D and 3D application problems is showed. The construction of basis functions for squares, triangles, hexahedral and tetrahedral, based on tensorial product, is briefly presented. It is showed that the approximation continuity in modal expansions for squares can be reached with a suitable functions numbering. Finally, through a analytical solution, the 2D and 3D projection and Poisson problems are investigates in squares, triangles and hexahedrons meshes with h and p refinements / Mestrado / Mecanica dos Sólidos e Projeto Mecanico / Mestre em Engenharia Mecânica

La simulation numérique de phénomènes physiques complexes requiert généralement l’utilisation d’un maillage. En mécanique des fluides numérique, cela consisteà représenter un objet dans un gros volume de contrôle. Cet objet étant celui dont l’on souhaite simuler le comportement. Usuellement, l’objet et la boîte englobante sont représentés par des maillage de surface linéaires et la zone intermédiaire est remplie par un maillage volumique. L’objectif de cette thèse est de s’intéresser à deux manières différentes de représenter cet objet. La première approche dite immergée consiste à mailler intégralement le volume de contrôle et ensuite à simuler le comportement autour de l’objet sans avoir à mailler explicitement dans le volume ladite géometrie. L’objet étant implicitement pris en compte par le schéma numérique. Le couplage de cette méthode avec de l’adaptation de maillage linéaire est notamment étudié. La deuxième approche dite d’ordre élevé consiste quant à elle consiste à augmenter le degré polynomial du maillage de surface de l’objet. La première étape consiste donc à générer le maillage de surface de degré élevé et ensuite àpropager l’information de degré élevé dans les éléments volumiques environnants si nécessaire. Dans ce cadre-là, il s’agit de s’assurer de la validité de telles modifications et à considérer l’extension des méthodes classiques de modification de maillages linéaires. / The numerical simulation of complex physical phenomenons usually requires a mesh. In Computational Fluid Dynamics, it consists in representing an object inside a huge control volume. This object is then the subject of some physical study. In general, this object and its bounding box are represented by linear surface meshes and the intermediary zone is filled by a volume mesh. The aim of this thesis is to have a look on two different approaches for representing the object. The first approach called embedded method consist in integrally meshing the bounding box volume without explicitly meshing the object in it. In this case, the presence of the object is implicitly simulated by the CFD solver. The coupling of this method with linear mesh adaptation is in particular discussed.The second approach called high-order method consist on the contrary by increasing the polynomial order of the surface mesh of the object. The first step is therefore to generate a suitable high-order mesh and then to propagate the high-order information in the neighboring volume if necessary. In this context, it is mandatory to make sure that such modifications are valid and then the extension of classic mesh modification techniques has to be considered.