Accelerator architectures are used to accelerate the simulation of nonlinear hyperbolic PDEs. Three different architectures, a multicore CPU using threading, IBM’s Cell Processor, and Nvidia’s Tesla GPUs are investigated. Speed-ups of between 40-75× relative to a single CPU core in single precision are obtained using the Cell processor and the GPU. The three implementations are extended to parallel computing clusters by making use of the Message Passing Interface (MPI). The resulting hybrid-parallel code is investigated for performance and scalability on both a GPU and Cell computing cluster.

Accelerator architectures are used to accelerate the simulation of nonlinear hyperbolic PDEs. Three different architectures, a multicore CPU using threading, IBM’s Cell Processor, and Nvidia’s Tesla GPUs are investigated. Speed-ups of between 40-75× relative to a single CPU core in single precision are obtained using the Cell processor and the GPU. The three implementations are extended to parallel computing clusters by making use of the Message Passing Interface (MPI). The resulting hybrid-parallel code is investigated for performance and scalability on both a GPU and Cell computing cluster.

We present immersed discontinuous Galerkin finite element methods for one and two dimensional acoustic wave propagation problems in inhomogeneous media where elements are allowed to be cut by the material interface. The proposed methods use the standard discontinuous Galerkin finite element formulation with polynomial approximation on elements that contain one fluid while on interface elements containing more than one fluid they use specially-built piecewise polynomial shape functions that satisfy appropriate interface jump conditions. The finite element spaces on interface elements satisfy physical interface conditions from the acoustic problem in addition to extended conditions derived from the system of partial differential equations. Additional curl-free and consistency conditions are added to generate bilinear and biquadratic piecewise shape functions for two dimensional problems. We established the existence and uniqueness of one dimensional immersed finite element shape functions and existence of two dimensional bilinear immersed finite element shape functions for the velocity. The proposed methods are tested on one dimensional problems and are extended to two dimensional problems where the problem is defined on a domain split by an interface into two different media. Our methods exhibit optimal $O(h^{p+1})$ convergence rates for one and two dimensional problems. However it is observed that one of the proposed methods is not stable for two dimensional interface problems with high contrast media such as water/air. We performed an analysis to prove that our immersed Petrov-Galerkin method is stable for interface problems with high jumps across the interface. Local time-stepping and parallel algorithms are used to speed up computation. Several realistic interface problems such as ether/glycerol, water/methyl-alcohol and water/air with a circular interface are solved to show the stability and robustness of our methods. / Ph. D.

L’analyse de sensibilité (AS) concerne la quantification des changements dans la solution d’un système d’équations aux dérivées partielles (EDP) dus aux varia- tions des paramètres d’entrée du modèle. Les techniques standard d’AS pour les EDP, comme la méthode d’équation de sensibilité continue, requirent de dériver la variable d’état. Cependant, dans le cas d’équations hyperboliques l’état peut présenter des dis- continuités, qui donc génèrent des Dirac dans la sensibilité. Le but de ce travail est de modifier les équations de sensibilité pour obtenir un syst‘eme valable même dans le cas discontinu et obtenir des sensibilités qui ne présentent pas de Dirac. Ceci est motivé par plusieurs raisons : d’abord, un Dirac ne peut pas être saisi numériquement, ce qui pourvoit une solution incorrecte de la sensibilité au voisinage de la discontinuité ; deuxièmement, les pics dans la solution numérique des équations de sensibilité non cor- rigées rendent ces sensibilités inutilisables pour certaines applications. Par conséquent, nous ajoutons un terme de correction aux équations de sensibilité. Nous faisons cela pour une hiérarchie de modèles de complexité croissante : de l’équation de Burgers non visqueuse au système d’Euler quasi-1D. Nous montrons l’influence de ce terme de correction sur un problème d’optimisation et sur un de quantification d’incertitude. / Sensitivity analysis (SA) concerns the quantification of changes in Partial Differential Equations (PDEs) solution due to perturbations in the model input. Stan- dard SA techniques for PDEs, such as the continuous sensitivity equation method, rely on the differentiation of the state variable. However, if the governing equations are hyperbolic PDEs, the state can exhibit discontinuities yielding Dirac delta functions in the sensitivity. We aim at modifying the sensitivity equations to obtain a solution without delta functions. This is motivated by several reasons: firstly, a Dirac delta function cannot be seized numerically, leading to an incorrect solution for the sensi- tivity in the neighbourhood of the state discontinuity; secondly, the spikes appearing in the numerical solution of the original sensitivity equations make such sensitivities unusable for some applications. Therefore, we add a correction term to the sensitivity equations. We do this for a hierarchy of models of increasing complexity: starting from the inviscid Burgers’ equation, to the quasi 1D Euler system. We show the influence of such correction term on an optimization algorithm and on an uncertainty quantification problem.

The early steps of activation are crucial in deciding the fate of T-cells leading to the proliferation. These steps strongly depend on the initial conditions, especially the avidity of the T-cell receptor for the specific ligand and the concentration of this ligand. The recognition induces a rapid decrease of membrane TCR-CD3 complexes inside the T-cell, then the up-regulation of CD25 and then CD25-IL2 binding which down-regulates into the T-cell. This process can be monitored by flow cytometry technique. We propose several models based on the level of complexity by using population balance modeling technique to study the dynamics of T-cells population density during the activation process. These models provide us a relation between the population of T-cells with their intracellular and extracellular components. Moreover, the hypotheses are proposed for the activation process of daughter T-cells after proliferation. The corresponding population balance equations (PBEs) include reaction term (i.e. assimilated as growth term) and activation term (i.e. assimilated as nucleation term). Further the PBEs are solved by newly developed method that is validated against analytical method wherever possible and various approximate techniques available in the literature.

The manuscript presents my research on hyperbolic Partial Differential Equations (PDE), especially on conservation laws. My works began with this thought in my mind: ''Existence and uniqueness of solutions is not the end but merely the beginning of a theory of differential equations. The really interesting questions concern the behavior of solutions.'' (P.D. Lax, The formation and decay of shock waves 1974). To study or highlight some behaviors, I started by working on geometric optics expansions (WKB) for hyperbolic PDEs. For conservation laws, existence of solutions is still a problem (for large data, $L^\infty$ data), so I early learned method of characteristics, Riemann problem, $BV$ spaces, Glimm and Godunov schemes, \ldots In this report I emphasize my last works since 2006 when I became assistant professor. I use geometric optics method to investigate a conjecture of Lions-Perthame-Tadmor on the maximal smoothing effect for scalar multidimensional conservation laws. With Christian Bourdarias and Marguerite Gisclon from the LAMA (Laboratoire de \\ Mathématiques de l'Université de Savoie), we have obtained the first mathematical results on a $2\times2$ system of conservation laws arising in gas chromatography. Of course, I tried to put high oscillations in this system. We have obtained a propagation result exhibiting a stratified structure of the velocity, and we have shown that a blow up occurs when there are too high oscillations on the hyperbolic boundary. I finish this subject with some works on kinetic équations. In particular, a kinetic formulation of the gas chromatography system, some averaging lemmas for Vlasov equation, and a recent model of a continuous rating system with large interactions are discussed. Bernard Rousselet (Laboratoire JAD Université de Nice Sophia-Antipolis) introduced me to some periodic solutions related to crak problems and the so called nonlinear normal modes (NNM). Then I became a member of the European GDR: ''Wave Propagation in Complex Media for Quantitative and non Destructive Evaluation.'' In 2008, I started a collaboration with Bruno Lombard, LMA (Laboratoire de Mécanique et d'Acoustique, Marseille). We details mathematical results and challenges we have identified for a linear elasticity model with nonlinear interfaces. It leads to consider original neutral delay differential systems.

The class of hyperbolic conservation laws model the phenomena of non-linear wave propagation, including the presence and propagation of discontinuities and expansion waves. Such nonlinear systems can generate discontinuities in the so-lution even for smooth initial conditions. Presence of discontinuities results in break down of a solution in the classical sense and to show existence, weak for-mulation of a problem is required. Moreover, closed form solutions are di cult to obtain and in some cases such solutions are even unavailable. Thus, numerical algorithms play an important role in solving such systems. There are several dis-cretization techniques to solve hyperbolic systems numerically and Finite Volume Method (FVM) is one of such important frameworks. Numerical algorithms based on FVM are broadly classi ed into two categories, central discretization methods and upwind discretization methods. Various upwind and central discretization methods developed so far di er widely in terms of robustness, accuracy and ef-ciency and an ideal scheme with all these characteristics is yet to emerge. In this thesis, novel upwind and central schemes are formulated for various hyper-bolic systems, with the aim of maintaining right balance between accuracy and robustness. This thesis is divided into two parts. First part consists of the formulation of upwind methods to simulate genuine weakly hyperbolic (GWH) systems. Such systems do not possess full set of linearly independent (LI) eigenvectors and some of the examples include pressureless gas dynamics system, modi ed Burgers' sys-tem and further modi ed Burgers' system. The main challenge while formulating an upwind solver for GWH systems, using the concept of Flux Di erence Splitting (FDS), is to recover full set of LI eigenvectors, which is done through addition of generalized eigenvectors using the theory of Jordan Canonical Forms. Once the defective set of LI eigenvectors are completed, a novel (FDS-J) solver is for-mulated in such a manner that it is independent of generalized eigenvectors, as they are not unique. FDS-J solver is capable of capturing various shocks such as -shocks, 0-shocks and 00-shocks accurately. In this thesis, the FDS-J schemes are proposed for those GWH systems each of which have one particular repeated eigenvalue with arithmetic multiplicity (AM) greater than one. Moreover, each ux Jacobian matrix corresponding to such systems is similar to a unique Jordan matrix. After the successful treatment of genuine weakly hyperbolic systems, this strategy is further applied to those weakly hyperbolic subsystems which result on employ-ing various convection-pressure splittings to the Euler ux function. For example, Toro-Vazquez (TV) splitting and Zha-Bilgen (ZB) type splitting approaches to split the Euler ux function yield genuine weakly hyperbolic convective parts and strict hyperbolic pressure parts. Moreover, the ux Jacobian of each convective part is similar to a Jordan matrix with at least two lower order Jordan blocks. Based on the lines of FDS-J scheme, we develop two numerical schemes for Eu-ler equations using TV splitting and ZB type splitting. Both the new ZBS-FDS and TVS-FDS schemes are tested on various 1-D shock tube problems and out of two, contact capturing ZBS-FDS scheme is extended to 2-dimensional Euler system where it is tested successfully on various test cases including many shock instability problems. Second part of the thesis is associated with the development of simple, robust and accurate central solvers for systems of hyperbolic conservation laws. The idea of splitting schemes together with the notion of FDS is not easily extendable to systems such as shallow water equations. Thus, a novel central solver Convection Isolated Discontinuity Recognizing Algorithm (CIDRA) is formulated for shallow water equations. As the name suggests, the convective ux is isolated from the total ux in such a way that other ux, in present case other ux represents celerity part, must possess non-zero eigenvalue contribution. FVM framework is applied to each part separately and ux equivalence principle is used to x the coe cient of numerical di usion. CIDRA for SWE is computed on various 1-D and 2-D benchmark problems and extended to Euler systems e ortlessly. As a further improvement, a scalar di usion based algorithm CIDRA-1 is designed for v Euler systems. The scalar di usion coe cient depends on that particular part of the Rankine-Hugoniot (R-H) condition which involves total energy of the system as a direct contribution. This algorithm is applied to a variety of shock tube test cases including a class of low density ow problems and also to various 2-D test problems successfully. vi

Les premières étapes d’activation des lymphocytes T sont cruciales pour déterminer leur comportement, ainsi que leur prolifération. Ces étapes dépendent fortement des conditions initiales, particulièrement de l’avidité du récepteur du lymphocyte (TCR) pour le ligand spécifique provenant de l’antigène. La reconnaissance du virus entraine une séquence des réactions biochimiques mettant en œuvre de protéines membranaires et cellulaires. Le processus peut être mesuré par cytométrie en flux. On propose ici plusieurs modèles de différents niveaux de complexité. Ces modèles décrivent une relation entre la population de lymphocytes T et leurs composants intracellulaires et extracellulaires. Ils conduisent à des systèmes d’EDO et d’EDP dont la résolution permet d’étudier la dynamique de la densité de population des lymphocytes au cours du processus d'activation. En outre, différentes hypothèses sont proposées pour le processus d'activation des cellules filles après prolifération. Les équations de bilan de population (EBPs) sont résolues par une nouvelle méthode validée par une solution analytique quand elle existe, ou par comparaison à différentes méthodes numériques disponibles dans la littérature. L’avantage de cette nouvelle méthode est d’être utilisable dans certains cas où les méthodes classiques ne le sont pas. / The early steps of activation are crucial in deciding the fate of T-cells leading to the proliferation. These steps strongly depend on the initial conditions, especially the avidity of the T-cell receptor for the specific ligand and the concentration of this ligand. The recognition induces a rapid decrease of membrane TCR-CD3 complexes inside the T-cell, then the up-regulation of CD25 and then CD25–IL2 binding which down-regulates into the T-cell. This process can be monitored by flow cytometry technique. We propose several models based on the level of complexity by using population balance modeling technique to study the dynamics of T-cells population density during the activation process. These models provide us a relation between the population of T-cells with their intracellular and extracellular components. Moreover, the hypotheses are proposed for the activation process of daughter T-cells after proliferation. The corresponding population balance equations (PBEs) include reaction term (i.e. assimilated as growth term) and activation term (i.e. assimilated as nucleation term). Further the PBEs are solved by newly developed method that is validated against analytical method wherever possible and various approximate techniques available in the literature.

Les systèmes d'Equations aux Dérivées Partielles Hyperboliques Linéaires du Premier Ordre (EDPs HLPO) permettent de modéliser de nombreux systèmes de lois de conservation. Ils apparaissent, par exemple, lors de la modélisation de problèmes de trafic routier, d'échangeurs de chaleurs, ou de problèmes multiphasiques. Différentes approches ont été proposées pour stabiliser ou observer de tels systèmes. Parmi elles, la méthode de backstepping consiste à transformer le système originel en un système découplé pour lequel la synthèse de la loi de commande est plus simple. Les contrôleur obtenus par cette méthode sont explicites.Dans la première partie de cette thèse, nous présentons des résultats généraux de théorie des systèmes. Plus précisément, nous résolvons les problèmes de stabilisation en temps fini pour une classe générale d'EDPs HLPO. Le temps de convergence minimal atteignable dépend du nombre d'actionneurs disponibles. Les observateurs associés à ces contrôleurs (nécessaires pour envisager une utilisation industrielle de tels contrôleurs) sont obtenus via une approche duale. Un des avantages importants de l'approche considérée dans cette thèse est de montrer que l'espace généré par les solutions de l'EDPs HLPO considérée est isomorphe à l'espace généré par les solutions d'une système neutre à retards distribués.Dans la seconde partie de cette thèse, nous montrons la nécessité d'un changement de stratégie pour résoudre les problèmes de contrôle robuste. Ces questions surviennent nécessairement lorsque sont considérées des applications industrielles pour lesquelles les différents paramètres du système peuvent être mal connus, pour lesquelles des dynamiques peuvent avoir été négligées, de même que des retards agissant sur la commande ou sur la mesure, ou encore pour lesquels les mesures sont bruitées. Nous proposons ainsi quelques modifications sur les lois de commande précédemment développées en y incorporant plusieurs degrés de liberté permettant d'effectuer un compromis entre performance et robustesse. L'analyse de stabilité et de robustesse sous-jacente est rendue possible en utilisant l'isomorphisme précédemment introduit. / Linear First-Order Hyperbolic Partial Differential Equations (LFOH PDEs) represent systems of conservation and balance law and are predominant in modeling of traffic flow, heat exchanger, open channel flow or multiphase flow. Different control approaches have been tackled for the stabilization or observation of such systems. Among them, the backstepping method consists to map the original system to a simpler system for which the control design is easier. The resulting controllers are explicit.In the first part of this thesis, we develop some general results in control theory. More precisely, we solve the problem of finite-time stabilization of a general class of LFOH PDEs using the backstepping methodology. The minimum stabilization time reachable may change depending on the number of available actuators. The corresponding boundary observers (crucial to envision industrial applications) are obtained through a dual approach. An important by-product of the proposed approach is to derive an explicit mapping from the space generated by the solutions of the considered LFOH PDEs to the space generated by the solutions of a general class of neutral systems with distributed delays. This mapping opens new prospects in terms of stability analysis for LFOH PDEs, extending the stability analysis methods developed for neutral systems.In the second part of the thesis, we prove the necessity of a change of strategy for robust control while considering industrial applications, for which the major limitation is known to be the robustness of the resulting control law to uncertainties in the parameters, delays in the loop, neglected dynamics or disturbances and noise acting on the system. In some situations, one may have to renounce to finite-time stabilization to ensure the existence of robustness margins. We propose some adjustments in the previously designed control laws by means of several degrees of freedom enabling trade-offs between performance and robustness. The robustness analysis is fulfilled using the explicit mapping between LFOH PDEs and neutral systems previously introduced.

Ce manuscrit porte sur les transitions compressible-incompressible dans les équations aux dérivées partielles de la mécanique des fluides. On s'intéresse à deux problèmes : les structures flottantes et les fluides en rotation. Dans le premier problème, l'introduction d'un objet flottant dans les vagues induit une contrainte sur le fluide et les équations gouvernant le mouvement acquièrent une structure compressible-incompressible. Dans le deuxième problème, le mouvement de fluides géophysiques compressibles est influencé par la rotation de la Terre. L'étude de la limite à rotation rapide montre que le champ vectoriel de vitesse tend vers une configuration horizontale et incompressible.Les structures flottantes constituent un exemple particulier d'interaction fluide-structure, où un solide partiellement immergé flotte à la surface du fluide. Ce problème mathématique modélise le mouvement de convertisseurs d'énergie marine. En particulier, on s'intéresse aux bouées pilonnantes, installées proche de la côte où les modèles asymptotiques en eaux peu profondes sont valables. On étudie les équations de Saint-Venant axisymétriques en dimension deux avec un objet flottant à murs verticaux se déplaçant seulement verticalement. Les hypothèses sur le solide permettent de supprimer le problème à bord libre associé avec la ligne de contact entre l'air, le fluide et le solide. Les équations pour le fluide dans le domaine extérieur au solide sont donc écrites comme un problème au bord quasi-linéaire hyperbolique. Celui-ci est couplé avec une EDO non-linéaire du second ordre qui est dérivée de l'équation de Newton pour le mouvement libre du solide. On montre le caractère bien posé localement en temps du système couplé lorsque que les données initiales satisfont des conditions de compatibilité afin de générer des solutions régulières.Ensuite on considère une configuration particulière: le retour à l'équilibre. Il s'agit de considérer un solide partiellement immergé dans un fluide initialement au repos et de le laisser retourner à sa position d'équilibre. Pour cela, on utilise un modèle hydrodynamique différent, où les équations sont linearisées dans le domaine extérieur, tandis que les effets non-linéaires sont considérés en dessous du solide. Le mouvement du solide est décrit par une équation intégro-différentielle non-linéaire du second ordre qui justifie rigoureusement l'équation de Cummins, utilisée par les ingénieurs pour les mouvements des objets flottants. L'équation que l'on dérive améliore l'approche linéaire de Cummins en tenant compte des effets non-linéaires. On montre l'existence et l'unicité globale de la solution pour des données petites en utilisant la conservation de l'énergie du système fluide-structure.Dans la deuxième partie du manuscrit, on étudie les fluides en rotation rapide. Ce problème mathématique modélise le mouvement des flots géophysiques à grandes échelles influencés par la rotation de la Terre. Le mouvement est aussi affecté par la gravité, ce qui donne lieu à une stratification de la densité dans les fluides compressibles. La rotation génère de l'anisotropie dans les flots visqueux et la viscosité turbulente verticale tend vers zéro dans la limite à rotation rapide. Notre interêt porte sur ce problème de limite singulière en tenant compte des effets gravitationnels et compressibles. On étudie les équations de Navier-Stokes-Coriolis anisotropes compressibles avec force gravitationnelle dans la bande infinie horizontale avec une condition au bord de non glissement. Celle-ci et la force de Coriolis donnent lieu à l'apparition des couches d'Ekman proche du bord. Dans ce travail on considère des données initiales bien préparées. On montre un résultat de stabilité des solutions faibles globales pour des lois de pression particulières. La dynamique limite est décrite par une équation quasi-géostrophique visqueuse en dimension deux avec un terme d'amortissement qui tient compte des couches limites. / This manuscript deals with compressible-incompressible transitions arising in partial differential equations of fluid mechanics. We investigate two problems: floating structures and rotating fluids. In the first problem, the introduction of a floating object into water waves enforces a constraint on the fluid and the governing equations turn out to have a compressible-incompressible structure. In the second problem, the motion of geophysical compressible fluids is affected by the Earth's rotation and the study of the high rotation limit shows that the velocity vector field tends to be horizontal and with an incompressibility constraint.Floating structures are a particular example of fluid-structure interaction, in which a partially immersed solid is floating at the fluid surface. This mathematical problem models the motion of wave energy converters in sea water. In particular, we focus on heaving buoys, usually implemented in the near-shore zone, where the shallow water asymptotic models describe accurately the motion of waves. We study the two-dimensional nonlinear shallow water equations in the axisymmetric configuration in the presence of a floating object with vertical side-walls moving only vertically. The assumptions on the solid permit to avoid the free boundary problem associated with the moving contact line between the air, the water and the solid. Hence, in the domain exterior to the solid the fluid equations can be written as an hyperbolic quasilinear initial boundary value problem. This couples with a nonlinear second order ODE derived from Newton's law for the free solid motion. Local in time well-posedness of the coupled system is shown provided some compatibility conditions are satisfied by the initial data in order to generate smooth solutions.Afterwards, we address a particular configuration of this fluid-structure interaction: the return to equilibrium. It consists in releasing a partially immersed solid body into a fluid initially at rest and letting it evolve towards its equilibrium position. A different hydrodynamical model is used. In the exterior domain the equations are linearized but the nonlinear effects are taken into account under the solid. The equation for the solid motion becomes a nonlinear second order integro-differential equation which rigorously justifies the Cummins equation, assumed by engineers to govern the motion of floating objects. Moreover, the equation derived improves the linear approach of Cummins by taking into account the nonlinear effects. The global existence and uniqueness of the solution is shown for small data using the conservation of the energy of the fluid-structure system.In the second part of the manuscript, highly rotating fluids are studied. This mathematical problem models the motion of geophysical flows at large scales affected by the Earth's rotation, such as massive oceanic and atmospheric currents. The motion is also influenced by the gravity, which causes a stratification of the density in compressible fluids. The rotation generates anisotropy in viscous flows and the vertical turbulent viscosity tends to zero in the high rotation limit. Our interest lies in this singular limit problem taking into account gravitational and compressible effects. We study the compressible anisotropic Navier-Stokes-Coriolis equations with gravitational force in the horizontal infinite slab with no-slip boundary condition. Both this condition and the Coriolis force cause the apparition of Ekman layers near the boundary. They are taken into account in the analysis by adding corrector terms which decay in the interior of the domain. In this work well-prepared initial data are considered. A stability result of global weak solutions is shown for power-type pressure laws. The limit dynamics is described by a two-dimensional viscous quasi-geostrophic equation with a damping term that accounts for the boundary layers.