Les écoulements chargés en particules sont présents dans de nombreuses applications industrielles telles que le transport de boues ou l’industrie chimique en général. Dans des mélanges constitués de particules solides immergées dans un fluide visqueux, les interactions entre particules jouent un rôle essentiel dans la viscosité globale du mélange.Le phénomène de suspension est causé par des interactions hydrodynamiques à courte distance, connues sous le nom de lubrification. Les forces de lubrification sont généralement sous-estimées en raison de leur nature et de la discrétisation spatiale du problème.Dans cette thèse, nous proposons un modèle de lubrification qui estime les forces et couples hydrodynamiques non résolues par un solveur couplant la résolution des équations de Navier-Stokes incompressible par une méthode de volumes pénalisés, à la résolution de la dynamique des particules par une méthode aux éléments discrets. Les corrections des contraintes hydrodynamiques sont faites localement sur la surface des particules en interaction sans aucune hypothèse sur la forme générale des particules. La version finale du modèle de lubrification proposée peut être utilisée pour des suspensions de particules convexes sans aucune tabulation. La méthode numérique a été validée avec des particules sphériques et des ellipsoïdes, en comparant des simulations à des données expérimentales.Dans le cas de particules sphériques, le modèle de lubrification est aussi précis que les modèles de lubrification existants qui sont limités à ce type de géométrie. La compatibilité du modèle avec des particules convexes a été validée en comparant des simulations,utilisant des ellipsoïdes, à des mesures expérimentales que nous avons réalisées. / Particle-laden flows can be found in many industrial applications such as slurry transport or the chemical industry in general. In mixtures made of solid particles emerged in a viscous fluid, particle interactions play an essential role in the overall mixture viscosity. The suspension phenomenon is caused by short-range hydrodynamic interactions, known as lubrication. Lubrication forces are usually underestimated due to their singularities and the spatial discretization of the numerical schemes. In this thesis, we propose a lubrication model for a coupled volume penalization method and discrete element method solver that estimates the unresolved hydrodynamic forces and torques in incompressible Navier-Stokes flows. Corrections are made locally on the surfaces of the interacting particles without any assumption on the global particle shapes. The final version of the local lubrication model can be used for suspension of convex particles without any tabulations. The numerical method has been validated against experimental data with spherical and ellipsoidal particles. With spherical particles, the lubrication model performs as well as existing numerical models that are limited to this specific particle shape. The model compatibility with convex particles has been validated by comparing simulations using ellipsoids to experimental measurements we made.

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.

L'objectif de ce travail est de contribuer à une meilleure prédiction de la rupture des Anévrismes de l'Aorte Abdominale. Pour ce faire, des simulations par éléments finis ont été mises en oeuvre sur des anévrismes modèles dans des conditions proches de la réalité physiopathologique, i.e. en tenant compte de l'anisotropie de la paroi anévrismale, du caractère poreux du thrombus et des Interactions Fluide-Structure. Dans la première partie, une étude statique en l'absence du thrombus a permis de mettre en évidence l'influence de la géométrie de l'anévrisme et du comportement mécanique (isotrope ou anisotrope) de la paroi sur la distribution des contraintes, i.e. la rupture potentielle de l'anévrisme, ainsi que sur l'évolution du module de Peterson. Dans la seconde partie, une modélisation poro-hyperélastique du thrombus est proposée, en s'appuyant sur des données expérimentales de la littérature. La prise en compte de ce comportement et des Interactions Fluide-Structure montre que la pression intra-thrombus demeure du même ordre de grandeur que la pression intra-luminale, conformément à des mesures in vivo réalisées par ailleurs. Enfin, nous montrons que ceci n'est pas en contradiction avec une réduction du risque de rupture potentielle de l'anévrisme. / The aim of this work is to contribute to a better prediction of the Abdominal Aortic Aneurysm rupture (AAA). For that purpose, finite elements simulations have been performed on idealized AAA models under physiopathological like conditions, by taking into account the aneurysmal wall anisotropy, the intra-luminal thrombus porosity and the Fluid-Structure Interactions. In the first part, the influence of the aneurysm geometry and its wall properties (isotropic or anisotropic hyperelasticity) on the wall stress distribution and the Peterson's modulus has been studied in a static analysis and without taking into account the thrombus. In the second part, based on the experimental results from the litterature, a porohyperelastic model has been proposed for the thrombus. By considering such behavior for the thrombus and the Fluid-Structure Interactions, we observe that the intra-thrombus pressure is the same order as the intra-luminal pressure, which is consistent with in vivo measurements. Our results show that despite this unchanged pressure, the maximum wall stress decreases leading to a decrease of the aneurysm potential rupture.

Cette thèse s’inscrit dans le contexte de la simulation et du contrôle des vibrations de structures montées sur ressort qui peuvent apparaître sous l’effet de l’interaction avec l’écoulement de sillage instationnaire. Le contrôle de ce phénomène, appelé vibrations induites par vortex (VIV), est un enjeu critique dans l’optimisation de nombreux systèmes. Une méthode de frontières immergées (IBM) a été intégrée dans l’algorithme PISO du code OpenFOAM, dédié à la simulation d’écoulements fluides incompressibles. La méthode IBM permet une représentation précise de corps fixes ou en mouvement, tout en conservant des maillages structurés conduisant à des algorithmes plus précis et efficaces en termes de performances numériques. Pour calculer la divergence de l’équation de quantité de mouvement dans une boucle PISO et l’interpolation des flux, un calcul hybride orignal a été proposé avec une résolution analytique utilisant l’équation de la fonction noyau des quantités impliquant le terme force de l’IBM (quantités singulières). La méthode a été étendu au formalisme d’écoulements en régimes turbulents. Une loi de paroi a été intégrée permettant de modéliser la couche limite à grand nombre de Reynolds. Le travail de validation a été réalisé au regard des données expérimentales et numériques disponibles dans la littérature pour l’étude d’écoulements autour de cylindres et de sphères, sur une large gamme de nombres de Reynolds. Avec l’objectif de développer des lois de contrôle optimal pour le VIV, basées sur les mécanismes d’instabilité linéaire du système couplé dans le cadre de la théorie du contrôle, un solveur adjoint a été développé et validé. / This thesis is related to the simulation and the control of the vortex induced vibrations phenomenon (VIV), which can result from the fluid structure interactions between an unsteady wake and the body, when the shedding frequency in the wake is close to the natural frequency of the body. The control of VIV is a critical issue when optimizing many systems. An Immersed Boundaries Method (IBM) was implemented into the PISO algorithm as a new library of OpenFOAM, in order to perform reliable simulations of incompressible flows around bluff bodies.To compute the divergence of the momentum equation and the interpolation of the fluxes, an hybrid calculation with an analytical resolution of the quantities involving the force term (singular quantities) has been proposed. The mesh convergence of several errors was shown by means of a manufactured solution, allowing to analyze both the errors irelated to the discretization and to the IBM. The new algorithm was subsequently extended to the RANS and DDES formalism proposed in OpenFOAM for the simulation of turbulent flows. A wall law was integrated into theIBM method to model the boundary layers that develop around the bodies at large Reynolds numbers. Various 2D and 3D well-documented test cases of academic flows around fixed or moving solid bodies (cylinderand sphere) have been simulated and carefully validated against existing data from the literature in a large range of Reynolds numbers. With the objective of developing optimal control laws for VIV, based on the linear instability mechanisms of the coupled system within the framework of the control theory, a new adjoint solver was also developed and validated in OpenFOAM.

[pt] As interações fluido-estrutura são muito comuns na engenharia mecânica e civil porque muitas estruturas, como pontes, plataformas de petróleo, linhas de transmissão ou turbinas eólicas, estão diretamente em contato com um fluido, que pode ser o ar, no caso de vento, ou água, que irá perturbar a estrutura através de ondas. Um papel importante do engenheiro é prevenir a falha da estrutura devido às instabilidades criadas pelas interações fluidoestrutura. Este trabalho apresentará em primeiro lugar todos os conceitos básicos necessários para o estudo de problemas de interação fluido-estrutura. Assim, é realizada uma análise dimensional visando classificar os problemas de fluido-estrutura. A classificação é baseada na velocidade reduzida, e algumas conclusões sobre as conseqüências das interações fluido-estrutura podem ser feitas em termos de estabilidade ou, o que é mais interessante, de instabilidade. De fato, usando modelos simplificados, pode-se mostrar instabilidades estáticas e dinâmicas, induzidas por fluxo, que podem ser críticas para a estrutura. As partes finais do trabalho apresentarão uma estrutura não-linear específica, uma ponte suspensa. Primeiro, a formulação de um modelo simplificado unidimensional é explicada e, em seguida, através de uma discretização por elementos finitos, é realizado um estudo dinâmico. Além disso, algumas conclusões são apresentadas sobre a dinâmica das pontes suspensas. A última parte deste trabalho apresenta um método que foi uma importante fonte de publicação para nós, o método de decomposição regular. / [en] Fluid-structure interactions are very common in mechanical and civil engineering because many structures, as bridges, offshore risers, transmission lines or wind turbines are directly in contact with a fluid, which can be air, which will be source of wind, or water, which will perturb the structure through waves. An important role of the engineer is to prevent structure failure due to instabilities created by the fluid-structure interactions. This work will first present all the basic concepts needed for the study of fluid-structure interaction problems. Thus, a dimensional analysis of those problems is performed and also all the equations governing such cases are presented. Then, thanks to the dimensional analysis made, a classification of problems, namely based on the reduced velocity, can be done and some conclusions concerning the consequences of the fluid-structure interactions can be drawn in terms of stability or, which is more interesting, instability. Indeed, using simplified models one can show static and dynamic flow-induced instabilities that may be critical for the structure. The final parts of the work will present a specific non-linear structure, a suspension bridge. First the formulation of a simplified one-dimensional model is explained and then, through a finite element discretization, a dynamical study is performed. Also, some conclusions are made concerning the dynamic of suspension bridges. The last part of this work presents a method that was an important source of publication for us, the Smooth Decomposition method.

One of the primary objectives of voice research is to better understand the biomechanics of voice production and how changes in properties of the vocal folds (VFs) affect voice ability and quality. Synthetic VF models provide a way to observe how changes in geometry and material property affect voice biomechanics. This thesis seeks to evaluate an approach of using a genetic algorithm to design synthetic VF models in three ways: first, through the development of a computationally cost-effective 3D vocal fold model; second, by creating and optimizing a variation of this model; and third, by validating the approach. To reduce computation times, a user-defined function (UDF) was implemented in low-fidelity 2D and 3D computational VF models. The UDF replaced the conventional meshed fluid domain with the mechanical energy equation. The UDF was implemented in the commercial finite element code ADINA and verified to produce results that were similar to those of 2D and 3D VF models with meshed fluid domains. Computation times were reduced by 86% for 2D VF models and 74% for 3D VF models while core vibratory characteristic changes were less than 5%. The results from using the UDF demonstrate that computation times could be reduced while still producing acceptable results. A genetic algorithm optimizer was developed to study the effects of altering geometry and material elasticity on frequency, closed quotient (CQ), and maximum flow declination rate (MFDR). The objective was to achieve frequency and CQ values within the normal human physiological range while maximizing MFDR. The resulting models enabled an exploration of trends between objective and design variables. Significant trends and aspects of model variability are discussed. The results demonstrate the benefit of using a structured model exploration method to create models with desirable characteristics. Two synthetic VF models were fabricated to validate predictions made by models produced by the genetic algorithm. Fabricated models were subjected to tests where frequency, CQ, and sound pressure level were measured. Trends between computational and synthetic VF model responses are discussed. The results show that predicted frequency trends between computational and synthetic models were similar, trends for closed quotient were inconclusive, and relationships between MFDR and sound pressure level remained consistent. Overall, while discrepancies between computational and synthetic VF model results were observed and areas in need of further study are noted, the study results provide evidence of potential for using the present optimization method to design synthetic VF models.

La déflection statique d'une structure flexible exposée à un écoulement transverse permet classiquement de réduire la traînée à laquelle elle est soumise. Dans le domaine de la biomécanique, la déformation induite par l'écoulement d'éléments végétaux flexibles conduisant à une réduction du chargement est désignée par le terme `reconfiguration' pour souligner le caractère avantageux de ce processus adaptatif. Dans cette thèse, nous examinons les mécanismes qui sous-tendent le processus de reconfiguration, dans des systèmes fluide-structure présentant une variabilité spatiale, ou de la dynamique provenant au choix de l'instationnarité de l'écoulement de base, d'un couplage fluide-structure conduisant à une instabilité, ou de vibrations induites par vortex. Nous montrons que l'aptitude des structures flexibles à réduire l'intensité du chargement imposé par l'écoulement est preservée en présence de non-uniformités ou de dynamique, à condition que le design de la structure soit tel que la traînée résistive domine les forces inertiels. Nous montrons de plus que la capacité à se déformer présente l'avantage supplémentaire de permettre la réduction des vibrations induites par vortex. Notre travail indique également que des structures légères et élancées sont les mieux adaptées pour supporter les chargements induits par l'écoulement en se reconfigurant, et que l'efficacité de la réduction du chargement par reconfiguration élastique dépend faiblement de la distribution spatiale des propriétés du système. Finalement, la réduction des chargements résulte toujours, indépendamment du régime de reconfiguration, de la concentration de la déformation sur une longueur caractéristique inférieure à la longueur réelle de la structure. / The static deflection of a flexible structure exposed to a transverse flow is classically known to reduce the drag it has to withstand. In the field of biomechanics, the flow-induced deformation of flexible plant elements leading to a reduction of the loads is referred to as `reconfiguration', in order to highlight the alleged benefits of such adaptive process. In this thesis, we investigate the mechanisms underpinning the reconfiguration in flow-structure systems featuring some spatial variability, or some dynamics arising either from the unsteadiness of the free-stream, from a flow-structure coupling leading to an instability, or from vortex-induced vibrations. We show that the ability of flexible structures to reduce the magnitude of the flow-induced loads is preserved in the presence of non-uniformities or dynamics, provided that the design of the structure is such that resistive drag dominates over inertial forces. We also show that the ability to deform has the added benefit of reducing the magnitude of the vortex-induced vibrations. Our work further indicates that light, slender structures are better suited to accommodate the flow-induced loads by reconfiguring, and that the efficiency of the process of load reduction by elastic reconfiguration is weakly sensitive to the spatial distribution of the system properties. Finally, regardless of the regime of reconfiguration, the reduction of the load always results from the concentration of the deformation on a characteristic bending length smaller than the actual length of the structure.

Indiana University-Purdue University Indianapolis (IUPUI) / LBM-IB method is useful and popular simulation technique that is adopted ubiquitously to solve Fluid-Structure interaction problems in computational fluid dynamics. These problems are known for utilizing computing resources intensively while solving mathematical equations involved in simulations. Problems involving such interactions are omnipresent, therefore, it is eminent that a faster and accurate algorithm exists for solving these equations, to reproduce a real-life model of such complex analytical problems in a shorter time period. LBM-IB being inherently parallel, proves to be an ideal candidate for developing a parallel software. This research focuses on developing a parallel software library, LBM-IB based on the algorithm proposed by [1] which is first of its kind that utilizes the high performance computing abilities of supercomputers procurable today. An initial sequential version of LBM-IB is developed that is used as a benchmark for correctness and performance evaluation of shared memory parallel versions. Two shared memory parallel versions of LBM-IB have been developed using OpenMP and Pthread library respectively. The OpenMP version is able to scale well enough, as good as 83% speedup on multicore machines for <=8 cores. Based on the profiling and instrumentation done on this version, to improve the data-locality and increase the degree of parallelism, Pthread based data centric version is developed which is able to outperform the OpenMP version by 53% on manycore machines. A distributed version using the MPI interfaces on top of the cube based Pthread version has also been designed to be used by extreme scale distributed memory manycore systems.

No description available.

No description available.