This thesis explores the application of Radial Basis Functions (RBFs) to fluid dynamical problems. In particular, stationary Stokes and Navier-Stokes equations are solved using RBF collocation method. An existing approach from the literature, is enchanced by an additional polynomial basis and a new preconditioner. A faster method based on the partition of unity is introduced for stationary Stokes equations. Finally, a global method based on Picard linearization is introduced for stationary Navier-Stokes equations. / Denna avhandling utforskar tillämpningen av Radial Basis Functions (RBF) på dynamiska problem med vätskor. I synnerhet löses stationära Stokes och Navier-Stokes ekvationer lösas med hjälp av RBF-samlokaliseringsmetoden. En befintlig metod från litteraturen, förbättras genom en ytterligare polynombas och en ny förkonditionering. En snabbare metod baserad på enhetens partition introduceras för stationära Stokes-ekvationer. Slutligen introduceras en global metod baserad på Picard linjärisering för stationära Navier-Stokes ekvationer.

Meshless methods provide a highly continuous approximation field, convenient for thin structures like shells. Nevertheless, the lack of Kronecker Delta property makes the formulation of essential boundary conditions not straightforward, as the trial and test fields cannot be tailored to boundary values. Similar problem arise when different approximation regions must be joined, in a multi-region problem, such as kinks, folds or joints. This work presents three approaches to impose both kinematic conditions: the well known Lagrange Multiplier method, used since the beginning of the Element Free Galerkin method; a pure penalty approach; and the recently rediscovered alternative of Nitsche\'s Method. We use the EFG discretization technique for thick Reissner-Mindlin shells and adapt the weak form as to separate displacement and rotational degrees of freedom and obtain suitable and separate stabilization parameters. This approach enables the modeling of discontinuous shells and local refinement on multi-region problems. / Métodos sem malha geram campos de aproximação com alta continuidade, convenientes para estruturas finas como cascas. No entanto, a ausência da propriedade de Delta de Kronecker dificulta a formulação de condições essenciais de contorno, já que os campos de aproximação e teste não podem ser moldados aos valores de contorno. Um problema similar aparece quando diferentes regiões de aproximação precisam ser juntadas em um problema multi-regiões como dobras, vincos ou junções. Este trabalho apresenta três métodos de imposição ambas condições cinemáticas: o já conhecido método dos multiplicadores de Lagrange, usado desde o começo do método de Galekin sem elementos (EFG); uma abordagem de penalidade pura; e o recentemente redescoberto método de Nitsche. Nós usamos a técnica de discretização com EFG para cascas espessas de Reissner-Mindlin e adaptamos a forma fraca de forma a separar graus de liberdade de deslocamento e rotação e obter coeficientes de estabilização diferentes e apropriados. Essa abordagem permite a modelagem de cascas discontínuas e o refinamento local em problemas multi-regiões.

O Método dos Elementos Finitos Generalizados, MEFG, compartilha importantes características dos métodos sem malha. As funções de aproximação do MEFG, atreladas aos pontos nodais, são enriquecidas de modo análogo ao refinamento p realizado no Método das Nuvens hp. Por outro lado, por empregar uma malha de elementos para construir as funções partição da unidade, ele também pode ser entendido como uma forma não convencional do Método dos Elementos Finitos. Neste trabalho, ambas as interpretações são consideradas. Os métodos sem malha, particularmente o Método de Galerkin Livre de Elementos e o Método das Nuvens hp, são introduzidos com o propósito de estabelecer os conceitos fundamentais para a descrição do MEFG. Na seqüência, apresentam-se aplicações numéricas em análise linear e evidenciam-se características que tornam o MEFG interessante para a simulação da propagação de descontinuidades. Após discutir os modelos de dano adotados para representar o comportamento não-linear do material, são introduzidos exemplos de aplicação, inicialmente do Método das Nuvens hp e depois do MEFG, na análise de estruturas de concreto. Os resultados obtidos servem de argumento para a implementação de um procedimento p-adaptativo, particularmente com o MEFG. Propõe-se, então a adaptação do Método dos Resíduos em Elementos Equilibrados à formulação do MEFG. Com vistas ao seu emprego em problemas não-lineares, algumas modificações são introduzidas à formulação do estimador. Mostra-se que a medida obtida para representar o erro, apesar de fundamentada em diversas hipóteses nem sempre possíveis de serem satisfeitas, ainda assim viabiliza a análise não-linear p-adaptativa. Ao final, são enumeradas propostas para a aplicação do MEFG em problemas caracterizados pela propagação de defeitos / The Generalized Finite Element Method, GFEM, shares several features with the so called meshless methods. The approximation functions used in the GFEM are associated with nodal points like in meshless methods. In addition, the enrichment of the approximation spaces can be done in the same fashion as in the meshless hp-Cloud method. On the other hand, the partition of unity used in the GFEM is provided by Lagrangian finite element shape functions. Therefore, this method can also be understood as a variation of the Finite Element Method. Indeed, both interpretations of the GFEM are valid and give unique insights into the method. The meshless character of the GFEM justified the investigation of meshless methods in this work. Among them, the Element Free Galerkin Method and the hp-Cloud Method are described aiming to introduce key concepts of the GFEM formulation. Following that, several linear problems are solved using these three methods. Such linear analysis demonstrates several features of the GFEM and its suitability to simulate propagating discontinuities. Next, damage models employed to model the nonlinear behavior of concrete structures are discussed and numerical analysis using the hp-Cloud Method and the GFEM are presented. The results motivate the implementation of a p-adaptive procedure tailored to the GFEM. The technique adopted is the Equilibrated Element Residual Method. The estimator is modified to take into account nonlinear peculiarities of the problems considered. The hypotheses assumed in the definition of the error measure are sometimes violated. Nonetheless, it is shown that the proposed error indicator is effective for the class of p-adaptive nonlinear analysis investigated. Finally, several suggestions are enumerated considering future applications of the GFEM, specially for the simulation of damage and crack propagation

Cette thèse porte sur des méthode de résolution d'équations matricielles appliquées à la résolution numérique d'équations aux dérivées partielles ou des problèmes de contrôle linéaire. On s'intéressen en premier lieu à des équations matricielles linéaires. Après avoir donné un aperçu des méthodes classiques employées pour les équations de Sylvester et de Lyapunov, on s'intéresse au cas d'équations linéaires générales de la forme M(X)=C, où M est un opérateur linéaire matriciel. On expose la méthode de GMRES globale qui s'avère particulièrement utile dans le cas où M(X) ne peut s'exprimer comme un polynôme du premier degré en X à coefficients matriciels, ce qui est le cas dans certains problèmes de résolution numérique d'équations aux dérivées partielles. Nous proposons une approche, noté LR-BA-ADI consistant à utiliser un préconditionnement de type ADI qui transforme l'équation de Sylvester en une équation de Stein que nous résolvons par une méthode de Krylox par blocs. Enfin, nous proposons une méthode de type Newton-Krylov par blocs avec préconditionnement ADI pour les équations de Riccati issues de problèmes de contrôle linéaire quadratique. Cette méthode est dérivée de la méthode LR-BA-ADI. Des résultats de convergence et de majoration de l'erreur sont donnés. Dans la seconde partie de ce travail, nous appliquons les méthodes exposées dans la première partie de ce travail à des problèmes d'équations aux dérivées partielles. Nous nous intéressons d'abord à la résolution numérique d'équations couplées de type Burgers évolutives en dimension 2. Ensuite, nous nous intéressons au cas où le domaine borné est choisi quelconque. Nous établissons des résultats théoriques de l'existence de tels interpolants faisant appel à des techniques d'algèbre linéaire.

Recent technological advances are based on effectively using complex multiphysics concepts. Therefore, there is an ever increasing need for accurate numerical al-gorithms of reduced complexity for solving multiphysics problems. Traditional mesh-based simulation methods depend on a neighbor connectivity information for formulation of operators like derivatives. In large deformation problems, de-pendence on a mesh could prove a limitation in terms of accuracy and cost of preprocessing. Meshless methods obviate the need to construct meshes thus al-lowing simulations involving severe geometric deformations such as breakup of a contiguous domain into multiple fragments. Smoothed Particle Hydrodynamics (SPH) is a meshless particle based Lagrangian numerical method that has the longest continuous history of development ever since it was introduced in 1977. Commensurate with the significant growth in computational power, SPH has been increasingly applied to solve problems of greater complexity in fluid mechanics, solid mechanics, interfacial flows and astrophysics to name a few. The SPH approximation of the continuity and momentum equations govern-ing fluid flow traditionally involves a stiff equation of state relating pressure and density, when applied to incompressible flow problems. Incompressible Smoothed Particle Hydrodynamics (ISPH) is a variant of SPH that replaces this weak com-pressibility approach with a pressure equation that gives a hydrostatic pressure field which ensures a divergence-free (or density invariant) velocity field. The present study explains the development of an ISPH algorithm and its implementa-tion with focus on application to free surface flows, interaction of fluid with rigid bodies and coupling of incompressible fluids with a compressible second phase. Several improvements to the exiting ISPH algorithm are proposed in this study. A semi-analytic free surface model which is more accurate and robust compared to existing algorithms used in ISPH methods is introduced, validated against experi-ments and grid based CFD results. A surface tension model with specific applica-bility to free surfaces is presented and tested using 2D and 3D simulations. Using theoretical arguments, a volume conservation error in existing particle methods in general is demonstrated. A deformation gradient based approach is used to derive a new pressure equation which reduces these errors. The method is ap-plied to both free surface and internal flow problems and is shown to have better volume conservation and therefore reduced density fluctuations. Also, comments on instabilities arising from particle distributions are made and the role of the smoothing functions in such instabilities is discussed. The challenges in imple-menting the ISPH algorithm in a computer code are discussed and the experience of developing an in-house ISPH code is described. A parametric study on water entry of cylinders of different shapes, angular velocity and density is performed and aspects such as surface profiles, impact pressures and penetration velocities are compared. An analysis on the energy transfer between the solid and the fluid is also performed. Low Froude number water entry of a sphere is studied and the impact pressure is compared with the theoretical estimates. The Incompressible SPH formulation, employing the proposed improvements from the study is then coupled with a compressible SPH formulation to perform two phase flow simulations interacting compressible and incompressible fluids. To gain confidence in its applicability, the simulations are compared against the theoretical predication given by the Rayleigh-Plesset equation for the problem of compressible drop in an incompressible fluid.

Cette thèse consiste à explorer une nouvelle approche pour la simulation d'objets flexibles par la mécanique des milieux continus, dans le cadre d'applications graphiques interactives telles que le jeu vidéo ou l'entraînement aux gestes chirurgicaux. Elle s'inscrit en continuité d'un stage de M2-R sur ce même sujet. Il est important de pouvoir régler simplement un compromis entre précision et temps de calcul suivant la nature de l'application. Les approches actuelles de simulation utilisent principalement la méthode des éléments finis. Celle-ci repose sur un maillage volumique des objets qu'il est souvent difficile d'adapter dynamiquement aux besoins de l'application. La nouveauté introduite par cette thèse est d'utiliser des repères déformables comme primitives cinématiques, avec des champs de déplacements inspirés des méthodes de 'skinning' utilisées en informatique graphique. Le but est d'éviter ainsi les difficultés liées au maillage volumique, ainsi que de faciliter le raffinement et la simplification adaptatives par simple ajout ou suppression de repère déformable là où c'est souhaitable. Ce travail est financé par le projet européen 'Passport for Virtual Surgery', dont le but est de créer automatiquement des modèles physiques pour l'entraînement aux gestes de chirurgie hépatique, à partir de données médicales et anatomiques personnalisées. Dans ce contexte, Guillaume, en collaboration avec d'autres membres du projet, mettra en place les outils nécessaires pour construire la scène physique à partir d'images médicales segmentées et de connaissances anatomiques génériques. Le foie sera dans un premier temps représenté par des modèles physiques précédemment développés à EVASION et étendus aux opérations de découpe. Par la suite, il y appliquera son nouveau modèle mécanique basé sur des repères déformables. The aim of this thesis is to develop a new approach for the simulation of flexible objects based on the continous middle method, related with interactive graphics applications such as video games or training in surgery. It is a continuity of the M2 research internship on the same topic. It is important to simply settle a compromise between accuracy and time computing according to the application. Current simulation approaches mainly use the finite element method, which is based on a volumetric mesh of the simulated objects. It is often difficult to dynamically adapt the needs to the application. The novelty of this thesis is to use deformable reference frames as kinematic primitives, with displacement fields based on 'skinning' methods used in computer graphics. The aim is to avoid the difficulties associated with volumetric mesh, and make the refinement and the adaptive simplification easier by adding or deleting deformable reference frames if necessary. This work is funded by the European project 'Passport for Virtual Surgery', which aims to automatically create models for physical training in gestures of liver surgery, from medical and anatomical custom data. In this context, Guillaume, in collaboration with other members of the project, will develop the tools necessary to build the physical scene from segmented medical images and generic anatomical knowledge. The liver will initially be represented by physical models previously developed in the EVASION team and then extended to cutting operations. Thereafter, Guillaume will apply his new mechanical model based on deformable reference frames. / Physically based deformable models have become ubiquitous in computer graphics. It allow to synthetize real behaviors, based on the physical laws from continuum mechanics. In this thesis, we focus on interactive simulations such as to video games or surgical simulators. The majority of the existing works focused up to here on the animation of objects made of homogeneous materials. Nevertheless, plenty of real objects, for instance like the biological structures, consist of multiple imbricated materials. Their decomposition in homogeneous zones requires a high-resolution spatial discretization to solve the variations of the material properties, which requires prohibitive computation time. In this context, we present new real time simulation techniques for deformable objects which can be cut. First of all, we present a real time method for cutting deformable objects in which, contrary to the previous methods, the object deforms on the cutting tool contact and cuts occur only when the pressure reaches a certain level. The independence of the physical, collision and visual models makes the topological changes easier. The GPU computing and local modifications enable fast execution. Then, a dynamic meshless method is described, which uses reference frames as control nodes instead of using points, with a displacement field formulation similar to skinning. It allows to easily tune the weights and benefits from the rigor of physical methods as the finite elements. The introduction of integration points, reducing the samples number by a least squares approximation, speeds up the spatial integrations. Other pre-computations are proposed in order to speed up the simulation time. Finally, new anisotropic shape functions are defined to encode the variations of material properties thanks to the introduction of the compliance distance. These complex shape functions uncouple the material resolution of the displacement functions ones. It allow an extremely sparse nodes sampling. The use of the compliance distance allows an automatic nodes distribution with regard to the material properties.

Meshless methods provide a highly continuous approximation field, convenient for thin structures like shells. Nevertheless, the lack of Kronecker Delta property makes the formulation of essential boundary conditions not straightforward, as the trial and test fields cannot be tailored to boundary values. Similar problem arise when different approximation regions must be joined, in a multi-region problem, such as kinks, folds or joints. This work presents three approaches to impose both kinematic conditions: the well known Lagrange Multiplier method, used since the beginning of the Element Free Galerkin method; a pure penalty approach; and the recently rediscovered alternative of Nitsche\'s Method. We use the EFG discretization technique for thick Reissner-Mindlin shells and adapt the weak form as to separate displacement and rotational degrees of freedom and obtain suitable and separate stabilization parameters. This approach enables the modeling of discontinuous shells and local refinement on multi-region problems. / Métodos sem malha geram campos de aproximação com alta continuidade, convenientes para estruturas finas como cascas. No entanto, a ausência da propriedade de Delta de Kronecker dificulta a formulação de condições essenciais de contorno, já que os campos de aproximação e teste não podem ser moldados aos valores de contorno. Um problema similar aparece quando diferentes regiões de aproximação precisam ser juntadas em um problema multi-regiões como dobras, vincos ou junções. Este trabalho apresenta três métodos de imposição ambas condições cinemáticas: o já conhecido método dos multiplicadores de Lagrange, usado desde o começo do método de Galekin sem elementos (EFG); uma abordagem de penalidade pura; e o recentemente redescoberto método de Nitsche. Nós usamos a técnica de discretização com EFG para cascas espessas de Reissner-Mindlin e adaptamos a forma fraca de forma a separar graus de liberdade de deslocamento e rotação e obter coeficientes de estabilização diferentes e apropriados. Essa abordagem permite a modelagem de cascas discontínuas e o refinamento local em problemas multi-regiões.

O Método dos Elementos Finitos Generalizados, MEFG, compartilha importantes características dos métodos sem malha. As funções de aproximação do MEFG, atreladas aos pontos nodais, são enriquecidas de modo análogo ao refinamento p realizado no Método das Nuvens hp. Por outro lado, por empregar uma malha de elementos para construir as funções partição da unidade, ele também pode ser entendido como uma forma não convencional do Método dos Elementos Finitos. Neste trabalho, ambas as interpretações são consideradas. Os métodos sem malha, particularmente o Método de Galerkin Livre de Elementos e o Método das Nuvens hp, são introduzidos com o propósito de estabelecer os conceitos fundamentais para a descrição do MEFG. Na seqüência, apresentam-se aplicações numéricas em análise linear e evidenciam-se características que tornam o MEFG interessante para a simulação da propagação de descontinuidades. Após discutir os modelos de dano adotados para representar o comportamento não-linear do material, são introduzidos exemplos de aplicação, inicialmente do Método das Nuvens hp e depois do MEFG, na análise de estruturas de concreto. Os resultados obtidos servem de argumento para a implementação de um procedimento p-adaptativo, particularmente com o MEFG. Propõe-se, então a adaptação do Método dos Resíduos em Elementos Equilibrados à formulação do MEFG. Com vistas ao seu emprego em problemas não-lineares, algumas modificações são introduzidas à formulação do estimador. Mostra-se que a medida obtida para representar o erro, apesar de fundamentada em diversas hipóteses nem sempre possíveis de serem satisfeitas, ainda assim viabiliza a análise não-linear p-adaptativa. Ao final, são enumeradas propostas para a aplicação do MEFG em problemas caracterizados pela propagação de defeitos / The Generalized Finite Element Method, GFEM, shares several features with the so called meshless methods. The approximation functions used in the GFEM are associated with nodal points like in meshless methods. In addition, the enrichment of the approximation spaces can be done in the same fashion as in the meshless hp-Cloud method. On the other hand, the partition of unity used in the GFEM is provided by Lagrangian finite element shape functions. Therefore, this method can also be understood as a variation of the Finite Element Method. Indeed, both interpretations of the GFEM are valid and give unique insights into the method. The meshless character of the GFEM justified the investigation of meshless methods in this work. Among them, the Element Free Galerkin Method and the hp-Cloud Method are described aiming to introduce key concepts of the GFEM formulation. Following that, several linear problems are solved using these three methods. Such linear analysis demonstrates several features of the GFEM and its suitability to simulate propagating discontinuities. Next, damage models employed to model the nonlinear behavior of concrete structures are discussed and numerical analysis using the hp-Cloud Method and the GFEM are presented. The results motivate the implementation of a p-adaptive procedure tailored to the GFEM. The technique adopted is the Equilibrated Element Residual Method. The estimator is modified to take into account nonlinear peculiarities of the problems considered. The hypotheses assumed in the definition of the error measure are sometimes violated. Nonetheless, it is shown that the proposed error indicator is effective for the class of p-adaptive nonlinear analysis investigated. Finally, several suggestions are enumerated considering future applications of the GFEM, specially for the simulation of damage and crack propagation

Avec l'avancé des méthodes numériques, de nouvelles méthodes dites « sans maillage » sont apparues pour remédier à certaines limitations de la méthode des éléments finis. Ces méthodes ont la particularité de n’employer aucun maillage prédéfini : elles utilisent un ensemble de nœuds dispersés dans le domaine considéré et sur ses frontières. L’objectif de cette étude est de montrer l’intérêt de l’application des méthodes sans maillage basées sur les fonctions de base radiale pour la simulation des procédés de mise en forme en général et de poinçonnage rapide en particulier. Une attention particulière sera portée sur la méthode RPIM qui offre l’avantage de proposer une interpolation nodale. La démarche proposée dans ce document consiste: à rappeler succinctement les principes essentiels des méthodes sans maillage en précisant leurs avantages par rapport aux méthodes classiques, à présenter et à mettre en œuvre la technique numérique de la méthode sans maillage RPIM avec un calibrage des paramètres caractéristiques, et enfin, à traiter des exemples numériques de procédés de mise en forme avec amorçage et propagation de fissure qui confirmeront ces avantages. / In recent years, new methods named Meshfree methods have been developed to surmount limitations of the finite element method. The main characteristic of these methods is to not employ any pre-defined mesh: they use a set of nodes scattered within the problem domain as well as sets of nodes scattered on the boundaries of the domain. A particular attention is paid to the RPIM method, which proposes a nodal interpolation. The followed steps are: a presentation of « Meshfree methods » and their advantages compared to the traditional methods, an introduction to the meshfree RPIM method with a calibration of its associated parameters, and finally, a discussion on results obtained with the RPIM in forming processes exhibiting initiation and propagation of a crack showing the interest of the approach.

Le récent développement des menaces terroristes renforce l'effort de recherche du CEA et d'EDF pour la protection des citoyens et des installations. De nombreux scénarios doivent être envisagés comme, par exemple, la chute d'un avion de ligne sur une structure de génie civil. La dispersion du carburant dans la structure, son embrasement sous forme de boule de feu et les effets thermiques associés sont des éléments essentiels du problème. L'utilisation de modèles numériques est indispensable car des expériences seraient difficiles à mettre en œuvre, coûteuses et dangereuses. Le problème type que l'on cherche à modéliser est donc l'impact d'un réservoir rempli de fluide, sa déchirure et la dispersion de son contenu. C'est un problème complexe qui fait intervenir une structure mince avec un comportement fortement non-linéaire allant jusqu'à rupture, un fluide dont la surface libre peut varier drastiquement et des interactions fluide-structure non permanentes. L'utilisation des méthodes numériques traditionnelles pour résoudre ce problème semble difficile, essentiellement parce qu'elles reposent sur un maillage. Cela complique la gestion des grandes déformations, la modélisation des interfaces variables et l'introduction de discontinuités telles que les fissures. Afin de s'affranchir de ces problèmes, la méthode sans maillage SPH (\og Smoothed Particle Hydrodynamics \fg) a été utilisée pour modéliser le fluide et la structure. Ce travail, inscrit dans la continuité de recherches précédentes, a permis d'étendre un modèle de coque SPH à la modélisation des ruptures. Un algorithme de gestion des interactions fluide-structure a également été adapté à la topologie particulière des coques. Afin de réduire les coûts de calcul importants liés à ce modèle, un couplage avec la méthode des éléments finis a également été élaboré. Il permet de n'utiliser les SPH que dans les zones d'intérêt où la rupture est attendue. Finalement, des essais réalisés par l'ONERA sont étudiés pour valider la méthode. Ces travaux ont permis de doter le logiciel de dynamique rapide Europlexus d'un outil original et efficace pour la simulation des impacts de structures minces en interaction avec un fluide. Un calcul démonstratif montre enfin la pertinence de l'approche et sa mise en œuvre dans un cadre industriel. / The recent development of terrorist threats increases the research effort of the french Atomic Energy Commission (CEA) and the French Electricity company (EDF) for the protection of citizens and facilities. Many scenarios should be considered as, for example, the fall of an airliner on a civil engineering structure. The dispersion of fuel in the structure, the formation of a fireball and associated thermal effects are essential elements of the problem. The use of numerical models is essential because experiences would be difficult to organize, costly and dangerous. The typical problem that we want to model is the impact of a tank filled with fluid, its rupture and the dispersion of its contents. It is a complex problem which involves a thin structure with a highly non-linear behavior up to rupture, a fluid with a free surface that can vary drastically and non permanent fluid-structure interactions. The use of traditional numerical methods to solve this kind of problems is difficult, mainly because they rely on a mesh. This complicates the management of large deformations, the modeling of moving interfaces and the introduction of discontinuities such as cracks. To overcome these problems, the meshfree method SPH (Smoothed Particle Hydrodynamics) was used to model both the fluid and the structure. This work, which is a continuation of previous research, has extended a model of SPH shell to the modeling of ruptures. An algorithm for managing fluid-structure interactions has also been adapted to the particular topology of shells. To reduce the important computational costs associated with this model, a coupling with the finite element method was also developed. It allows the use of SPH in areas of interest where the rupture is expected. Finally, tests performed by the french Aerospace Lab (ONERA) are studied to validate the method. This work helped to provide fast dynamic software Europlexus an original and effective tool for the simulation of the impact of thin structures interacting with fluid. A demonstrative calculation finally shows the relevance of the approach and its use within an industrial framework.