A new implicit and compact optimization-based method is presented for high order derivative calculation for finite-volume numerical method on unstructured meshes. Highorder approaches to gradient calculation are often based on variants of the Least-Squares (L-S) method, an explicit method that requires a stencil large enough to accommodate the necessary variable information to calculate the derivatives. The new scheme proposed here is applicable for an arbitrary order of accuracy (demonstrated here up to 3rd order), and uses just the first level of face neighbors to compute all derivatives, thus reducing stencil size and avoiding stiffness in the calculation matrix. Preliminary results for a static variable field example and solution of a simple scalar transport (advection) equation show that the proposed method is able to deliver numerical accuracy equivalent to (or better than) the nominal order of accuracy for both 2nd and 3rd order schemes in the presence of a smoothly distributed variable field (i.e., in the absence of discontinuities). This new Optimization-based Gradient REconstruction (herein denoted OGRE) scheme produces, for the simple scalar transport test case, lower error and demands less computational time (for a given level of required precision) for a 3rd order scheme when compared to an equivalent L-S approach on a two-dimensional framework. For three-dimensional simulations, where the L-S scheme fails to obtain convergence without the help of limiters, the new scheme obtains stable convergence and also produces lower error solution when compared to a third order MUSCL scheme. Furthermore, spectral analysis of results from the advection equation shows that the new scheme is better able to accurately resolve high wave number modes, which demonstrates its potential to better solve problems presenting a wide spectrum of wavelengths, for example unsteady turbulent flow simulations.

No description available.

No contexto de Dinâmica de Fluidos Computacional, métodos de simulação de objetos imersos em Malhas Cartesianas têm se mostrado vantajosos tanto em termos de Custo Computacional quanto em termos de precisão numérica. Entretanto, a representação física de objetos imersos nesses domínios computacionais impõe a perda de validade dos esquemas de Diferenças Finitas empregados, na região das superfícies introduzidas. Este trabalho analisa um Método de Interfaces Imersas quanto ao desempenho em aplicações a esquemas de solução numérica de Alta Ordem de precisão. Através de Testes de Refinamento de Malha, é feita a apreciação da ordem de decaimento dos erros das soluções numéricas em comparação com as soluções analíticas para 2 problemas unidimensionais. O primeiro envolve a solução da Equação de Calor unidimensional sujeita a uma Condição Inicial Unitária, e o segundo relaciona-se ao cálculo das duas primeiras derivadas espaciais das funções analíticas Seno e Tangente Hiperbólica. Também é promovida uma análise de forma fragmentária do método, a fim de individualizar a contribuição dos elementos envolvidos no comportamento das soluções geradas. Os resultados obtidos indicam eventuais alterações na ordem de precisão dos esquemas de Diferenças Finitas originalmente aplicados. Esse comportamento e visto como uma dependência que o método escolhido apresenta em relação a função discretizada. Por fim, são elaboradas considerações sobre restrições de aplicabilidade do método escolhido. / In the Computational Fluid Dynamics context, methods for simulating immersed objects in Cartesian Grids have shown advantages regarding both Computational Cost and numerical precision. Nevertheless, the physical representation of immersed objects within these computational domains leads to the loss of validity of the emplyed Finite Dierence Schemes near the immersed surfaces. This work analizes a Immersed Interface Method regarding its performance in High Order Schemes applications. The error decay order for numerical solutions of two 1D problems is observed. The rst problem relates to the solution of the Heat Equation subjected to the unitary initial condition. The second relates to the computation of the rst two derivatives of analytical functions Sin and Hyperbolic Tangent. It\'s also conducted a fragmentary analysis, which is intended to identify the contribution of each element of this method to the character of the generated solution. The results indicate some eventual changes in the Order of the Finite Dierences Schemes employed. This behaviour is regarded as a dependency of this method to the nature of the discretized function. Finaly, some remarks regarding restrictions to this method\'s applicability are made.

No contexto de Dinâmica de Fluidos Computacional, métodos de simulação de objetos imersos em Malhas Cartesianas têm se mostrado vantajosos tanto em termos de Custo Computacional quanto em termos de precisão numérica. Entretanto, a representação física de objetos imersos nesses domínios computacionais impõe a perda de validade dos esquemas de Diferenças Finitas empregados, na região das superfícies introduzidas. Este trabalho analisa um Método de Interfaces Imersas quanto ao desempenho em aplicações a esquemas de solução numérica de Alta Ordem de precisão. Através de Testes de Refinamento de Malha, é feita a apreciação da ordem de decaimento dos erros das soluções numéricas em comparação com as soluções analíticas para 2 problemas unidimensionais. O primeiro envolve a solução da Equação de Calor unidimensional sujeita a uma Condição Inicial Unitária, e o segundo relaciona-se ao cálculo das duas primeiras derivadas espaciais das funções analíticas Seno e Tangente Hiperbólica. Também é promovida uma análise de forma fragmentária do método, a fim de individualizar a contribuição dos elementos envolvidos no comportamento das soluções geradas. Os resultados obtidos indicam eventuais alterações na ordem de precisão dos esquemas de Diferenças Finitas originalmente aplicados. Esse comportamento e visto como uma dependência que o método escolhido apresenta em relação a função discretizada. Por fim, são elaboradas considerações sobre restrições de aplicabilidade do método escolhido. / In the Computational Fluid Dynamics context, methods for simulating immersed objects in Cartesian Grids have shown advantages regarding both Computational Cost and numerical precision. Nevertheless, the physical representation of immersed objects within these computational domains leads to the loss of validity of the emplyed Finite Dierence Schemes near the immersed surfaces. This work analizes a Immersed Interface Method regarding its performance in High Order Schemes applications. The error decay order for numerical solutions of two 1D problems is observed. The rst problem relates to the solution of the Heat Equation subjected to the unitary initial condition. The second relates to the computation of the rst two derivatives of analytical functions Sin and Hyperbolic Tangent. It\'s also conducted a fragmentary analysis, which is intended to identify the contribution of each element of this method to the character of the generated solution. The results indicate some eventual changes in the Order of the Finite Dierences Schemes employed. This behaviour is regarded as a dependency of this method to the nature of the discretized function. Finaly, some remarks regarding restrictions to this method\'s applicability are made.

As aplicações aeroespaciais estão frequentemente associadas a escoamentos compressíveis com altíssimos números de Reynolds. No entanto, existem no contexto aeroespacial importantes aplicações que envolvem escoamentos compressíveis a Reynolds relativamente baixos. Entre eles se destacam o escoamento em pás de turbina a gás e ao redor de dispositivos de alta sustentação como eslates e flapes em grandes ângulos de ataque. Pode-se destacar também o processo de combustão supersônica que está intimamente ligado e é fortemente beneficiado pelo presente estudo. Nas aplicações aerodinâmicas em baixos números de Reynolds frequentemente uma parcela significativa do escoamento se apresenta no regime de transição para turbulência, ou nos estágios iniciais do escoamento turbulento. O objetivo do presente projeto é a simulação numérica direta de escoamentos compressíveis transicionais com desenvolvimento de um código para simulação em três dimensões de escoamentos alto subsônicos. O escoamento a ser estudado no projeto é a evolução linear e não linear de trens de onda e pacotes de onda em uma camada de mistura compressível. A solução das equações de Navier-Stokes é obtida através do método das diferenças finitas. As derivadas espaciais são resolvidas através de um método compacto de sexta ordem, enquanto que as derivadas temporais são resolvidas através do método de Runge-Kutta de quarta ordem. Os métodos de aproximação foram modificados para trabalhar com malhas não uniformes visando refinar a malha em pontos em que o fenômeno ocorre e, consequentemente, reduzir o custo computacional. A investigação numérica inicia-se com a análise da taxa de amplificação dos trens de ondas fortemente modulados em regime linear. Os resultados obtidos foram comparados favoravelmente com a teoria linear. Os testes foram estendidos para a análise não linear, e consequentemente, foi possível reproduzir os fenômenos clássicos de instabilidade hidrodinâmica através da evolução dos trens de ondas oblíquos. / Aerospace applications are frequently associated with compressible flows at relatively high Reynolds number. Nevertheless important applications involve compressible flows at relatively low Reynolds number in the aerospace context. Among them, the flow on gas turbine blades and high lift devices such as slats and flaps at high angle of attack are particulary important. Besides, progress in aeroespace research is dependent on developing more efficient propulsion systems. In aerodynamic applications at low Reynolds number, often a substancial portion of the flow is in the transition regime, or in the initial stages of a turbulent flow. The objective of the present study is the Direct Numerical Simulation of three-dimensional transition of compressible flows in a mixing layer. Inspired on the worked devoted to modulated waves, the current work investigates the linear and nonlinear temporal evolution of wavetrains in this phenomenon. The Navier-Stokes equations were solved with a sixth-order compact finite-difference schemes. The time integration was performed by a fourth-order Runge-Kutta scheme. Moreover, the methods to solve the spatial derivatives were modified to work with non-uniform grids. This technique was implemented with the objective to improve the resolution of the grid where the phenomenon occurs and to reduce the computational cost. The numerical investigation starts with an analysis of the growth rate of the wavetrains in linear regime to verify the numerical code. The results compared favourably with linear theory. Tests were also performed in the nonlinear regime to simulate the oblique wavetrains and it was possible to reproduce the classical hydrodynamic instability phenomena.

These thesis review some recent results on the construction of very high order multidimensional upwind schemes for the solution of steady and unsteady conservation laws on unstructured triangular grids. We also consider the extension to the approximation of solutions to conservation laws containing second order dissipative terms. To build this high order schemes we use a sub-triangulation of the triangular Pk elements where we apply the distribution used for a P1 element. This manuscript is divided in two parts. The first part is dedicated to the design of the high order schemes for scalar equations and focus more on the theoretical design of the schemes. The second part deals with the extension to system of equations, in particular we will compare the performances of 2nd, 3rd and 4th order schemes. The first part is subdivided in four chapters: The aim of the second chapter is to present the multidimensional upwind residual distributive schmes and to explain what was the status of their development at the beginning of this work. The third chapter is dedicated to the first contribution: the design of 3rd and 4th order quasi non-oscillatory schemes. The fourth chapter is composed of two parts: We start by understanding the non-uniformity of the accuracy of the 2nd order schemes for advection-diffusion problem. To solve this issue we use a Finite Element hybridisation. This deep study of the 2nd order scheme is used as a basis to design a 3rd order scheme for advection-diffusion. Finally, in the fifth chapter we extend the high order quasi non-oscillatory schemes to unsteady problems. In the second part, we extend the schemes of the first part to systems of equations as follows: The sixth chapter deals with the extension to steady systems of hyperbolic equations. In particular, we discuss how to solve some issues such as boundary conditions and the discretisation of curved geometries. Then, we look at the performance of 2nd and 3rd order schemes on viscous flow. Finally, we test the space-time schemes on several test cases. In particular, we will test the monotonicity of the space-time non-oscillatory schemes and we apply residual distributive schemes to acoustic problems.

Dans cette thèse, nous étudions et développons différentes familles de schémas d’intégration en temps pour les EDO linéaires. Dans la première partie, après avoir introduit les définitions et propriétés utilisées pour construire les schémas en temps, nous présentons deux méthodes de discrétisation en espace et une revue des schémas de Runge-Kutta (RK) qui sont couramment utilisés dans la littérature. Dans la seconde partie on présente une méthodologie pour construire deux familles de schémas A-stable pour un ordre quelcomque. Puis on fournit des schémas explicites, construits en maximisant leur nombre CFL pour un profil de spectre donné. Ces schémas explicites sont ensuite combinés aux schémas implicites A-stable, pour construire des schémas localement implicites que nous décrivons. En plus des tests de validations des schémas pour des problèmes en dimension un et deux de l’espace, nous présentons des résultats numériques obtenus en résolvant des problèmes de propagation d’ondes acoustiques et électromagnétiques en dimensions trois dans la troisième partie. / In this thesis, we study and develop different families of time integration schemes for linear ODEs. After presenting the space discretisation methods and a review of classical Runge-Kutta schemes in the first part, we construct high-order A-stable time integration schemes for an arbitrary order with low-dissipation and low-dispersion effects in the second part. Then we develop explicit schemes with an optimal CFL number for a typical profile of spectrum. The obtained CFL number and the efficiency on the typical profile for each explicit scheme are given. Pursuing our aim, we propose a methodology to construct locally implicit methods of arbitrary order. We present the locally implicit methods obtained from the combination of the A-stable implicit schemes we have developed and explicit schemes with optimal CFL number. We use them to solve the acoustic wave equation and provide convergence curves demonstrating the performance of the obtained schemes. In addition of the different 1D and 2D validation tests performed while solving the acoustic wave equation, we present numerical simulation results for 3D acoustic wave and the Maxwell’s equations in the last part.

Dans le cadre du développement récent des schémas numériques compacts d’ordre élevé, tels que la méthode de Galerkin discontinu (discontinuous Galerkin) ou la méthode des différences spectrales (spectral differences), nous nous intéressons aux difficultés liées à l’utilisation de ces méthodes lors de la simulation de solutions discontinues.L’utilisation par ces schémas numériques d’une représentation polynomiale des champs les prédisposent à fournir des solutions fortement oscillantes aux abords des discontinuités. Ces oscillations pouvant aller jusqu’à l’arrêt du processus de simulation, l’utilisation d’un dispositif numérique de détection et de contrôle de ces oscillations est alors un prérequis nécessaire au bon déroulement du calcul. Les processus de limitation les plus courants tels que les algorithmes WENO ou l’utilisation d’une viscosité artificielle ont d’ores et déjà été adaptés aux différentes méthodes compactes d’ordres élevés et ont permis d’appliquer ces méthodes à la classe des écoulements compressibles. Les différences entre les stencils utilisés par ces processus de limitation et les schémas numériques compacts peuvent néanmoins être une source importante de perte de performances. Dans cette thèse nous détaillons les concepts et le cheminement permettant d’aboutir à la définition d’un processus de limitation compact adapté à la description polynomiale des champs. Suite à une étude de configurations monodimensionnels, différentes projections polynomiales sont introduites et permettent la construction d’un processus de limitation préservant l’ordre élevé. Nous présentons ensuite l’extension de cette méthodologie à la simulation d’écoulements compressibles bidimensionnels et tridimensionnels. Nous avons en effet développé les schémas de discrétisation des différences spectrales dans un code CFD non structuré, massivement parallèle et basé historiquement sur une méthodologie volumes finis. Nous présentons en particulier différents résultats obtenus lors de la simulation de l’interaction entre une onde de choc et une couche limite turbulente. / Following the recent development of high order compact schemes such as the discontinuous Galerkin or the spectraldifferences, this thesis investigates the issues encountered with the simulation of discontinuous flows. High order compactschemes use polynomial representations which tends to introduce spurious oscillations around discontinuities that can lead to computational failure. To prevent the emergence of these numerical issues, it is necessary to improve the schemewith an additional procedure that can detect and control its behaviour in the neighbourhood of the discontinuities,usually referred to as a limiting procedure or a limiter. Most usual limiters include either the WENO procedure, TVB schemes or the use of an artificial viscosity. All of these solutions have already been adapted to high order compact schemes but none of these techniques takes a real advantage of the richness offered by the polynomial structure. What’s more, the original compactness of the scheme is generally deteriorated and losses of scalability can occur. This thesis investigates the concept of a compact limiter based on the polynomial structure of the solution. A monodimensional study allows us to define some algebraic projections that can be used as a high-order tool for the limiting procedure. The extension of this methodology is then evaluated thanks to the simulation of different 2D and 3D test cases. Those results have been obtained thanks to the development of a parallel solver which have been based on a existing unstructured finite volume CFD code. The different exposed studies detailed end up to the numerical simulation of the shock turbulent boundary layer.

As aplicações aeroespaciais estão frequentemente associadas a escoamentos compressíveis com altíssimos números de Reynolds. No entanto, existem no contexto aeroespacial importantes aplicações que envolvem escoamentos compressíveis a Reynolds relativamente baixos. Entre eles se destacam o escoamento em pás de turbina a gás e ao redor de dispositivos de alta sustentação como eslates e flapes em grandes ângulos de ataque. Pode-se destacar também o processo de combustão supersônica que está intimamente ligado e é fortemente beneficiado pelo presente estudo. Nas aplicações aerodinâmicas em baixos números de Reynolds frequentemente uma parcela significativa do escoamento se apresenta no regime de transição para turbulência, ou nos estágios iniciais do escoamento turbulento. O objetivo do presente projeto é a simulação numérica direta de escoamentos compressíveis transicionais com desenvolvimento de um código para simulação em três dimensões de escoamentos alto subsônicos. O escoamento a ser estudado no projeto é a evolução linear e não linear de trens de onda e pacotes de onda em uma camada de mistura compressível. A solução das equações de Navier-Stokes é obtida através do método das diferenças finitas. As derivadas espaciais são resolvidas através de um método compacto de sexta ordem, enquanto que as derivadas temporais são resolvidas através do método de Runge-Kutta de quarta ordem. Os métodos de aproximação foram modificados para trabalhar com malhas não uniformes visando refinar a malha em pontos em que o fenômeno ocorre e, consequentemente, reduzir o custo computacional. A investigação numérica inicia-se com a análise da taxa de amplificação dos trens de ondas fortemente modulados em regime linear. Os resultados obtidos foram comparados favoravelmente com a teoria linear. Os testes foram estendidos para a análise não linear, e consequentemente, foi possível reproduzir os fenômenos clássicos de instabilidade hidrodinâmica através da evolução dos trens de ondas oblíquos. / Aerospace applications are frequently associated with compressible flows at relatively high Reynolds number. Nevertheless important applications involve compressible flows at relatively low Reynolds number in the aerospace context. Among them, the flow on gas turbine blades and high lift devices such as slats and flaps at high angle of attack are particulary important. Besides, progress in aeroespace research is dependent on developing more efficient propulsion systems. In aerodynamic applications at low Reynolds number, often a substancial portion of the flow is in the transition regime, or in the initial stages of a turbulent flow. The objective of the present study is the Direct Numerical Simulation of three-dimensional transition of compressible flows in a mixing layer. Inspired on the worked devoted to modulated waves, the current work investigates the linear and nonlinear temporal evolution of wavetrains in this phenomenon. The Navier-Stokes equations were solved with a sixth-order compact finite-difference schemes. The time integration was performed by a fourth-order Runge-Kutta scheme. Moreover, the methods to solve the spatial derivatives were modified to work with non-uniform grids. This technique was implemented with the objective to improve the resolution of the grid where the phenomenon occurs and to reduce the computational cost. The numerical investigation starts with an analysis of the growth rate of the wavetrains in linear regime to verify the numerical code. The results compared favourably with linear theory. Tests were also performed in the nonlinear regime to simulate the oblique wavetrains and it was possible to reproduce the classical hydrodynamic instability phenomena.

Ce manuscrit de thèse est consacré à l'analyse mathématique et numérique des systèmes de l'hydrodynamique compressible et de la photonique. Plus particulièrement, on étudie la construction de méthodes numériques dans des systèmes de coordonnées 2D polaires (une coordonnée radiale et une coordonnée d'angle) et où les équations sont discrétisées sur des maillages polaires structurés. Ces méthodes sont adaptées à la simulation d'écoulements à symétrie polaire puisqu'elles préservent ces symétries par construction. En revanche, ces systèmes de coordonnées introduisent des singularités géométriques et des termes sources géométriques qui doivent être traités avec attention. Dans la première partie de ce document, consacrée à l'hydrodynamique, on propose une classe de schémas numériques d'ordre arbitrairement élevé pour la résolution des équations d'Euler. Ces schémas utilisent des méthodes de résolution à directions alternées où chaque sous-système est résolu par un solveur Lagrange+projection. On étudie l'influence de la singularité géométrique r=0 des systèmes de coordonnées cylindriques et sphériques sur la précision du solveur 2D développé. La deuxième partie de ce manuscrit est consacrée à l'étude des équations de la photonique. Ces équations font intervenir un grand nombre de dimensions mathématiques et un terme source pouvant être raide. La principale difficulté ici est de capturer le bon régime asymptotique sur maillage grossier. On construit d'abord une classe de modèles où l'intensité radiative est projetée sur une base d'harmoniques sphériques afin de réduire le nombre de dimensions. Puis on propose un schéma numérique en coordonnées polaires et on prouve que le schéma restitue la bonne limite de diffusion aussi bien dans la direction radiale que dans la direction angulaire. / This thesis deals with the mathematical and numerical analysis of the systems of compressible hydrodynamics and radiative transfer. More precisely, we study the derivation of numerical methods with 2D polar coordinates (one for the radius, one for the angle) where equations are discretized on regular polar grids. On one hand, these methods are well-suited for the simulation of flows with polar symetries since they preserve these symetries by construction. On the other hand, such coordinates systems introduce geometrical singularities as well as geometrical source terms which must be carefully treated. The first part of this document is devoted to the study of hydrodynamics equations, or Euler equations. We propose a new class of arbitrary high-order numerical schemes in both space and time and rely on directional splitting methods for the resolution of 2D equations. Each sub-system is solved using a Lagrange+Remap solver. We study the influence of the r=0 geometrical singularities of the cylindrical and spherical coordinates systems on the precision of the 2D numerical solutions. The second part of this document is devoted to the study of radiative transfer equations. In these equations, the unknowns depend on a large number of variables and a stiff source term is involved. The main difficulty consists in capturing the correct asymptotic behavior on coarse grids. We first construct a class of models where the radiative intensity is projected on a truncated spherical harmonics basis in order to lower the number of mathematical dimensions. Then we propose an Asymptotic Preserving scheme built in polar coordinates and we show that the scheme capture the correct diffusion limit in the radial direction as well as in the polar direction.