This thesis presents the author's contribution to the design and implementation of COOLFluiD, an object oriented software platform for the high performance simulation of multi-physics phenomena on unstructured grids. In this context, the final goal has been to provide a reliable tool for handling high speed aerothermodynamic applications. To this end, we introduce a number of design techniques that have been developed in order to provide the framework with flexibility and reusability, allowing developers to easily integrate new functionalities such as arbitrary mesh-based data structures, numerical algorithms (space discretizations, time stepping schemes, linear system solvers, ...),and physical models. Furthermore, we describe the parallel algorithms that we have implemented in order to efficiently read/write generic computational meshes involving millions of degrees of freedom and partition them in a scalable way: benchmarks on HPC clusters with up to 512 processors show their effective suitability for large scale computing. Several systems of partial differential equations, characterizing flows in conditions of thermal and chemical equilibrium (with fixed and variable elemental fractions)and, particularly, nonequilibrium (multi-temperature models) have been integrated in the framework. In order to simulate such flows, we have developed two state-of-the-art flow solvers: 1- a parallel implicit 2D/3D steady and unsteady cell-centered Finite Volume (FV) solver for arbitrary systems of PDE's on hybrid unstructured meshes; 2- a parallel implicit 2D/3D steady vertex-centered Residual Distribution (RD) solver for arbitrary systems of PDE's on meshes with simplex elements (triangles and tetrahedra). The FV~code has been extended to handle all the available physical models, in regimes ranging from incompressible to hypersonic. As far as the RD code is concerned, the strictly conservative variant of the RD method, denominated CRD, has been applied for the first time in literature to solve high speed viscous flows in thermochemical nonequilibrium, yielding some preliminary outstanding results on a challenging double cone flow simulation. All the developments have been validated on real-life testcases of current interest in the aerospace community. A quantitative comparison with experimental measurements and/or literature has been performed whenever possible.

Au cours de ce travail, nous nous sommes attaché à la résolution numérique des équations de la Magnétohydrodynamique (MHD) auxquelles s'ajoute une loi hyperbolique de transport des erreurs de divergence.La première étape consista à symétriser le nouveau système de la MHD idéale afin d'en étudier le système propre, ce qui fut l'occasion de rappeler le rôle de l'entropie au niveau de ce calcul comme à celui de l'inégalité de Clausius-Duhem. La suite de cette thèse eut pour objectif la résolution de ces équations idéales à l'aide de schémas distribuant le résidu (notés RD). Les quatre principaux schémas connus furent testés, et nous avons montré entre autres que le schéma N, qui a fait ses preuves sur les équations d'Euler en mécanique des fluides, n'était pas adapté aux équations de la MHD. Les stratégies classiques de limitation et de stabilisation purent être revisitées à ce moment. Les équations étant instationnaires, il fallut intégrer une discrétisation en temps et une distribution spatiale des termes d'évolution (et d'éventuelles sources). Nous avons d'emblée opté pour une approche implicite permettant d'être performant sur les simulations longues des expériences de tokamaks, et de traiter la correction de la divergence d'une manière originale et efficace. Les problèmes de convergence de la méthode de Newton-Raphson n'ayant pas été pleinement résolus, nous nous sommes tournés vers une alternative explicite de type Runge-Kutta. Enfin, nous avons réétabli les principes de la montée en ordre (en théorie, jusqu'à des ordres arbitraires, en prenant en compte le phénomène de Gibbs) à l'aide de tout type d'élément fini (bien construit) 2D ou 3D, sans avoir pu valider tous ces aspects. Nous avons également pris en compte les équations complètes de la MHD réelle classique (i.e. sans effet Hall) à l'aide d'un couplage RD/Galerkin. / During this thesis, we worked on the numerical resolution of the Magnetohydrodynamic (MHD) equations, to which we added a hyperbolic transport equation for the divergence errors of the magnetic field.The first step consisted in symmetrizing the new ideal MHD system in order to study its eigensystem, which was the opportunity to remind the role of the entropy in this calculation as well as in the Clausius-Duhem inequality. Next, we aimed at solving these ideal equations by the mean of Residual Distribution (RD) schemes.The four main schemes were tested, and we showed among other things that the N scheme (although it has been proven very efficient with Euler equations in Fluid Mechanics) could not give satisfying results with the MHD equations. Classical strategies for the limitation and the stabilization were revisited then. Moreover,since we dealt with unsteady equations, we had to formulate atime discretization and a spatial distribution of the unsteady terms (as well as possible sources). We first choosed an implicit approach allowing us to be powerful on the long simulations needed for tokamak experiments, and to treat the divergence cleaning part in an original and efficient way. The convergence problems of our Newton-Raphson algorithm having not been fully resolved, we turned to an explicit alternative (Runge-Kutta type).Finally, we discussed about the principles of higher order schemes (theoretically, up to arbitrary orders, taking into account the Gibbs phenomenon) thanks to any type of 2D or 3D finite element (properly defined), without having been able to to validate all these aspects. We also implemented the dissipative part of the full MHD equations (in the classical sense, i.e. omitting the Hall effect) by the use of a RD/Galerkin coupling.

Residual-distribution methods offer several potential benefits over classical methods, such as a means of applying upwinding in a multi-dimensional manner and a multi-dimensional positivity property. While it is apparent that residual-distribution methods also offer higher accuracy than finite-volume methods on similar meshes, few studies have directly compared the performance of the two approaches in a systematic and quantitative manner. In this study, comparisons between residual distribution and finite volume are made for steady-state smooth and discontinuous flows of gas dynamics, governed by hyperbolic conservation laws, to illustrate the strengths and deficiencies of the residual-distribution method. Deficiencies which reduce the accuracy are analyzed and a new nonlinear scheme is proposed that closely reproduces or surpasses the accuracy of the best linear residual-distribution scheme. The accuracy is further improved by extending the scheme to fourth order using established finite-element techniques. Finally, the compact stencil, arithmetic workload, and data parallelism of the fourth-order residual-distribution scheme are exploited to accelerate parallel computations on an architecture consisting of both CPU cores and a graphics processing unit. Numerical experiments are used to assess the gains to efficiency and possible monetary savings that may be provided by accelerated architectures.

Residual-distribution methods offer several potential benefits over classical methods, such as a means of applying upwinding in a multi-dimensional manner and a multi-dimensional positivity property. While it is apparent that residual-distribution methods also offer higher accuracy than finite-volume methods on similar meshes, few studies have directly compared the performance of the two approaches in a systematic and quantitative manner. In this study, comparisons between residual distribution and finite volume are made for steady-state smooth and discontinuous flows of gas dynamics, governed by hyperbolic conservation laws, to illustrate the strengths and deficiencies of the residual-distribution method. Deficiencies which reduce the accuracy are analyzed and a new nonlinear scheme is proposed that closely reproduces or surpasses the accuracy of the best linear residual-distribution scheme. The accuracy is further improved by extending the scheme to fourth order using established finite-element techniques. Finally, the compact stencil, arithmetic workload, and data parallelism of the fourth-order residual-distribution scheme are exploited to accelerate parallel computations on an architecture consisting of both CPU cores and a graphics processing unit. Numerical experiments are used to assess the gains to efficiency and possible monetary savings that may be provided by accelerated architectures.

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.

La prédiction de mouvement de solide évoluant dans un fluide présente un réel intérêt pour des applications industrielles telle que l’accrétion de glace sur des surfaces aérodynamiques. Dans ce contexte, en considérant des systèmes de dégivrage, la prévision des trajectoire de glace est nécessaire pour éviter des risques de collision/ingestion de glace sur/dans des zones sensibles de l’avion. Ce type d’application soulève de nombreux challenges d’un point de vue numérique, en particulier concernant la génération/l’adaptation de maillage au cours du mouvement du solide dans le domaine. Pour gérer ces difficultés, dans cette étude, les solides sont définis de manière implicite via une fonction level set. Une méthode de type frontière immergée, appelée Pénalization, est utilisée pour imposer les conditions de bords. Pour améliorer la précision de l’interface, les équations sont résolues sur des maillages non structurés adaptatifs. Cela permet d’obtenir un raffinement proche des bords du solide et ainsi d’améliorer sa définition, permettant un meilleure impositions des conditions de bord. Pour économiser du temps de calcul, et éviter de coûteuses étapes de remaillage/interpolation, la stratégie adoptée pour les simulations instationnaires est d’utiliser une adaptation de maillage à connectivité constante, aussi appelée r-adaptation. / The prediction of solid motion evolving in a fluid presents a real interest for engineering application such as ice accretion on aerodynamics bodies.In this context, considering de-icing systems, the ice shedding trajectory is needed to prevent the risk of collision/ingestion of the ice in/with some sensitive part of the aircraft. This application raises many challenges from a numerical point of view, especially concerning mesh generation/adaptation as the solid moves in the computational domain. To handle this issue, in this work the solids are known implicitly on the mesh via a level set function. An immersed boundary method, called penalization, is employed to impose the wall boundary conditions. To improve the resolution of these boundaries, the equations are solved on adaptive unstructured grids. This allows to have are finement close to the solid boundary and thus increases the solid definition,leading to a more accurate imposition of the wall conditions. To save computational time, and avoid costly remeshing/interpolation steps, the strategy chosen for unsteady simulations is to use a constant connectivity mesh adaptation,also known as r-adaptation

Cette thèse a pour objectif la construction de schémas d’ordre élevé et peu diffusifs pour le transport d’un scalaire dans les écoulements à surface libre, en deux ou trois dimensions. On souhaite en particulier obtenir des schémas robustes, qui gardent au niveau discret les propriétés mathématiques de l’équation de transport avec une faible diffusion numérique, et les utiliser sur des cas industriels. Dans ce travail deux méthodes numériques sont envisagées : une méthode aux volumes finis (VF) et une méthode aux résidus distribués (RD). Dans les deux cas, l’équation de transport est résolue avec une approche découplée, qui est la solution la plus avantageuse en termes de précision et de coûts de calcul. Pour ce qui concerne la méthode aux volumes finis, les équations de Saint-Venant couplées à l’équation du transport sont d’abord résolues avec un schéma dit vertex-centred où le flux numérique est approximé avec un solveur de Riemann appelé Harten-Lax-Van Leer-Contact [135]. A partir de cette approche, une formulation découplée est proposée. Cette dernière permet de résoudre l’équation du transport avec un pas de temps plus grand que celui de la formulation couplée. Cette idée a été d’abord proposée pour d’autres schémas dans [13]. Pour augmenter l’ordre de précision en espace, la technique MUSCL [89] est utilisée en combinaison avec l’approche découplée. Finalement, la problématique des zones sèches est abordée. Dans le cas de la méthode aux résidus distribués, les équations de Saint-Venant sont résolues avec une méthode éléments finis, et la méthode RD est utilisée seulement pour discrétiser l’équation du transport, en focalisant l’attention sur les problèmes non stationnaires. L’équation de continuité du fluide discrétisée est employée pour garantir la conservation de la masse et le principe du maximum. Pour obtenir des schémas d’ordre deux dans les problèmes non stationnaires, un schéma prédicteur-correcteur [112] est utilisé, en l’adaptant au cas de concentration moyennée sur la verticale. Une version d’ordre 1 mais peu diffusive, est aussi présentée dans ce travail. De plus, un schéma localement implicite, complètement nouveau, est aussi formulé pour pouvoir traiter le problème des bancs découvrant. Les deux techniques sont validées d’abord sur des cas simples, pour évaluer l’ordre de précision des schémas et ensuite sur des cas plus complexes pour vérifier aussi les autres propriétés numériques. Les résultats montrent que les nouveaux schémas sont à la fois précis et conservatifs, tout en gardant la monotonie comme le prévoient les démonstrations. Un cas d’application industriel est aussi présenté en conclusion. Le schéma prédicteur-correcteur RD est adapté aussi au cas 3D, sans aucun problème théorique nouveau, par rapport au cas 2D. Les propriétés de base des schémas sont validées sur des cas test préliminaires / The purpose of this thesis is to build higher order and less diffusive schemes for pollutant transport in shallow water flows or 3D free surface flows. We want robust schemes which respect the main mathematical properties of the advection equation with relatively low numerical diffusion and apply them to environmental industrial applications. Two techniques are tested in this work: a classical finite volume method and a residual distribution technique combined with a finite element method. For both methods we propose a decoupled approach since it is the most advantageous in terms of accuracy and CPU time. Concerning the first technique, a vertex-centred finite volume method is used to solve the augmented shallow water system where the numerical flux is computed through an Harten-Lax-Van Leer-Contact Riemannsolver [135]. Starting from this solution, a decoupled approach is formulated and is preferred since it allows to compute with a larger time step the advection of a tracer. This idea was inspired by [13]. The Monotonic Upwind Scheme for Conservation Law [89], combined with the decoupled approach, is then used for the second order extension in space. The wetting and drying problem is also analysed and a possible solution is presented. In the second case, the shallow water system is entirely solved using the finite element technique and the residual distribution method is applied to the solution of the tracer equation, focusing on the case of time-dependent problems. However, for consistency reasons the resolution of the continuity equation must be considered in the numerical discretization of the tracer. In order to get second order schemes for unsteady cases a predictor-corrector scheme [112] is used in this work. A first order but less diffusive version of the predictor-corrector scheme is also introduced. Moreover, we also present a new locally semi-implicit version of the residual distribution method which, in addition to good properties in terms of accuracy and stability, has the advantage to cope with dry zones. The two methods are first validated on academical test cases with analytical solution in order to assess the order of the schemes. Then more complex cases are addressed to test the robustness of the schemes and their performance under different flow conditions. Finally a real test case for which real data are available is carried out. An extension of the predictor-corrector residual distribution schemes to the 3D case is presented as final contribution. Even in this case the RD technique is completely compatible with the finite element framework used for the Navier-Stokes equations, thus its extension to the 3D case does not present any extra theoretical problem. The method is tested on preliminary cases

Ce travail a pour thème la simulation numérique d’écoulements diphasiques dans un contexte industriel. En effet, la simulation d’écoulements diphasiques est un domaine qui présente de nombreux défis, en raison de phénomènes complexes qui surviennent, comme la cavitation et autres transferts entre les phases. En outre, ces écoulements se déroulent généralement dans des géométries complexes rendant difficile une résolution efficiente. Les modèles que nous considérons font partie de la catégorie des modèles à interfaces diffuses et permettent de prendre en compte aisément les différents transferts entre les phases. Cette classe de modèles inclut une hiérarchie de sous-modèles pouvant simuler plus ou moins d’interactions entre les phases. Pour mener à bien cette étude nous avons en premier lieu comparé les modèles diphasiques dits à quatre équations et six équations, en incluant les effets de transfert de masse. Nous avons ensuite choisi de nous concentrer sur le modèle à quatre équations. L’objectif majeur de notre travail a alors été d’étendre les méthodes aux résidus distribués à ce modèle. Dans le contexte des méthodes de résolution numérique, il est courant d’utiliser la forme conservative des équations de bilan. En effet, la résolution sous forme non-conservative conduit à une mauvaise résolution du problème. Cependant, résoudre les équations sous forme non-conservative peut s’avérer plus intéressant d’un point de vue industriel. Dans ce but, nous utilisons une approche développée récemment permettant d’assurer la conservation en résolvant un système sous forme non-conservative, à condition que la forme conservative soit connue. Nous validons ensuite notre méthode et l’appliquons à des problèmes en géométries complexes. Finalement, la dernière partie de notre travail est dédiée à étudier la validité des modèles à interfaces diffuses pour des applications à des problèmes industriels réels. On cherche alors, en utilisant des méthodes de quantification d’incertitude, à obtenir les paramètres rendant nos simulations les plus vraisemblables et cibler les éventuels développements pouvant rendre nos simulations plus réalistes. / The topic of this thesis is the numerical simulation of two-phase flows in an industrial framework. Two-phase flows modelling is a challenging domain to explore, mainly because of the complex phenomena involved, such as cavitation and other transfer processes between phases. Furthermore, these flows occur generally in complex geometries, which makes difficult the development of efficient resolution methods. The models that we consider belong to the class of diffuse interface models, and they allow an easy modelling of transfers between phases. The considered class of models includes a hierarchy of sub-models, which take into account different levels of interactions between phases. To pursue our studies, first we have compared the so-called four-equation and six-equation two-phase flow models, including the effects of mass transfer processes. We have then chosen to focus on the four-equation model. One of the main objective of our work has been to extend residual distribution schemes to this model. In the context of numerical solution methods, it is common to use the conservative form of the balance law. In fact, the solution of the equations under a non-conservative form may lead to a wrong solution to the problem. Nonetheless, solving the equations in non-conservative form may be more interesting from an industrial point of view. To this aim, we employ a recent approach, which allows us to ensure conservation while solving a non-conservative system, at the condition of knowing its conservative form. We then validate our method and apply it to problems with complex geometry. Finally, the last part of our work is dedicated to the evaluation of the validity of the considered diffuse interface model for applications to real industrial problems. By using uncertainty quantification methods, the objective is to get parameters that make our simulations the most plausible, and to target the possible extensions that can make our simulations more realistic.

This thesis presents the construction, the analysis and the verication of compact residual discretizations for the solution of conservation laws on unstructured meshes. The schemes considered belong to the class of residual distribution (RD) or fluctuation splitting (FS) schemes. The methodology presented relies on three main elements: design of compact linear first-order stable schemes for linear hyperbolic PDEs, a positivity preserving procedure mapping stable first-order linear schemes onto nonlinear second-order schemes with non-oscillatory shock capturing capabilities, and a conservative formulation enabling to extend the schemes to nonlinear CLs. These three design steps, and the underlying theoretical tools, are discussed in depth. The nonlinear RD schemes resulting from this construction are tested on a large set of problems involving the solution of scalar models, and systems of CLs. This extensive verification fills the gaps left open, where no theoretical analysis is possible. Numerical results are presented on the Euler equations of a perfect gas, on a two-phase flow model with highly nonlinear thermodynamics, and on the shallow-water equations. On irregular grids, the schemes proposed yield quite accurate and stable solutions even on very difficult computations. Direct comparisone show that these results are more accurate than the ones given by FV and WENO schemes. Moreover, our schemes have a compact nearest-neighbor stencil. This encourages to further develop our approach, toward the design of very high-order schemes, which would represent a very appealing alternative, both in terms of accuracy and efficiency, to now classical FV and ENO/WENO discretizations. These schemes might also be very competitive with respect to very high-order DG schemes.

<p align="justify">This work deals with the development of numerical methods for compressible flow simulation with application to the interaction of fluid flows and structural bodies.</p> <p align="justify">First, we develop numerical methods based on multidimensional upwind residual distribution (RD) schemes. Theoretical results for the stability and accuracy of the methods are given. Then, the RD schemes for unsteady problems are extended for computations on moving meshes. As a second approach, cell centered and vertex centered finite volume (FV) schemes are considered. The RD schemes are compared to FV schemes by means of the 1D modified equation and by the comparison of the numerical results for scalar problems and system of Euler equations. We present a number of two and three dimensional steady and unsteady test cases, illustrating properties of the numerical methods. The results are compared with the theoretical solution and experimental data.</p> <p align="justify">In the second part, a numerical method for fluid-structure interaction problems is developed. The problem is divided into three distinct sub-problems: Computational Fluid Dynamics, Computational Solid Mechanics and the problem of fluid mesh movement. The problem of Computational Solid Mechanics is formulated as a system of partial differential equations for an anisotropic elastic continuum and solved by the finite element method. The mesh movement is determined using the pseudo-elastic continuum approach and solved again by the finite element method. The coupling of the problems is achieved by a simple sub-iterative approach. Capabilities of the methods are demonstrated on computations of 2D supersonic panel flutter and 3D transonic flutter of the AGARD 445.6 wing. In the first case, the results are compared with the theoretical solution and the numerical computations given in the references. In the second case the comparison with experimental data is presented.</p>