Cette thèse est dédiée à la description et à l'analyse mathématique de phénomènes d'hétérogénéités et de congestion dans les modèles de la mécanique des fluides.On montre un lien rigoureux entre des modèles de congestion douce de type Navier-Stokes compressible qui intègrent des forces de répulsion à très courte portée entre composants élémentaires; et des modèles de congestion dure de type compressible/incompressible décrivant les transitions entre zones libres et zones congestionnées.On s'intéresse ensuite à la modélisation macroscopique de mélanges formés par des particules solides immergées dans un fluide.On apporte dans ce cadre une première réponse mathématique à la question de la transition entre les régimes de suspensions dictés par les interactions hydrodynamiques et les régimes granulaires dictés par les contacts entre les particules solides.On met par cette démarche en évidence le rôle crucial joué par les effets de mémoire dans le régime granulaire.Cette approche permet également un nouveau point de vue pour l'étude mathématique des fluides avec viscosité dépendant de la pression.On s'intéresse enfin à la modélisation microscopique et macroscopique du trafic routier.Des schémas numériques originaux sont proposés afin de reproduire des phénomènes de persistance d'embouteillages. / This thesis is dedicated to the description and the mathematical analysis of heterogeneities and congestion phenomena in fluid mechanics models.A rigorous link between soft congestion models, based on the compressible Navier--Stokes equations which take into account short--range repulsive forces between elementary components; and hard congestion models which describe the transitions between free/compressible zones and congested/incompressible zones.We are interested then in the macroscopic modelling of mixtures composed solid particles immersed in a fluid.We provide a first mathematical answer to the question of the transition between the suspension regime dictated by hydrodynamical interactions and the granular regime dictated by the contacts between the solid particles.The method highlights the crucial role played by the memory effects in the granular regime.This approach enables also a new point of view concerning fluids with pressure-dependent viscosities.We finally deal with the microscopic and the macroscopic modelling of vehicular traffic.Original numerical schemes are proposed to robustly reproduce persistent traffic jams.

Yes / The Three Gorges Dam (TGD) constructed at the Yangtze River, China represents a revolutionary project to battle against the mage-scale flooding problems while improving the local economy at the same time. However, the large-scale fine-size sediment and pollutant material transport caused by the TGD operation are found to be inevitable and long-lasting. In this paper, a multi-fluid Incompressible Smoothed Particle Hydrodynamics (ISPH) model is used to simulate the multi-fluid flows similar to the fine sediment materials transport (in muddy flows) and water flow mixing process. The SPH method is a mesh-free particle modeling approach that can treat the free surfaces and multi-interfaces in a straightforward manner. The proposed model is based on the universal multi-fluid flow equations and a unified pressure equation is used to account for the interaction arising from the different fluid components. A Sub-Particle-Scale (SPS) turbulence model is included to address the turbulence effect generated during the flow process. The proposed model is used to investigate two cases of multi-fluid flows generated from the polluted flow intrusions into another fluid. The computations are found in good agreement with the practical situations. Sensitivity studies have also been carried out to evaluate the particle spatial resolution and turbulence modeling on the flow simulations. The proposed ISPH model could provide a promising tool to study the practical multi-fluid flows in the TGD operation environment. / The Major State Basic Research Development Program (973 program) of China (No. 2013CB036402) and the National Natural Science Foundation of China (No. 51479087).

This thesis examines relativistic fluids. We have used the variational approach to develop tools for studying the dynamics of relativistic fluids to apply this to cosmological modelling. Studies like these go beyond the standard model in cosmology. Researchers believe that such extensions to the standard cosmological model are pivotal to resolving some of the long-standing cosmological problems. An example of such problems is the origin, growth (from quantum electromagnetic fluctuations to large-scale magnetic fields during inflation) and evolution of cosmological magnetic fields that exhibit as large-scale (cosmological) magnetic fields in late time. One other example is the coincidence problem. The standard approach in such studies is to use modelling in the form of the single-fluid formalism. As an alternative one can consider the single-fluid and multi-fluid formalisms that incorporate aspects of electrodynamics and thermodynamics, respectively in the context of the variational approach. This might help us make progress in trying to either resolve some of these problems or at least open up new ways of addressing them. In this regard, we have extended the well-known M¨ueller-Israel-Stewart (hereafter MIS) formalism to allow us to examine the effect on fluid flow in which the components of the multi-species fluids interact thermodynamically. We use the extension to the MIS theory in the context of interacting species to study the growth of dark matter and dark energy, and find that either interaction or entrainment involving dark energy and dark matter suggests a mutual relative modulation of the growth behaviour of the two densities. This may aid in resolving the coincidence problem. Our examination of inflation-generated, large-scale magnetic fields reveals a super-adiabatically evolving mode from the beginning of the radiation-dominated epoch to either much later during the epoch or probably extending far into the era of matter domination which may account for late time, large-scale magnetic fields.

A coupled solution approach is presented for numerically simulating a near-field underwater explosion (UNDEX). An UNDEX consists of a complicated sequence of events over a wide range of time scales. Due to the complex physics, separate simulations for near/far-field and early/late-time are common in practice. This work focuses on near-field early-time UNDEX simulations. Using the assumption of compressible, inviscid and adiabatic flow, the fluid flow is governed by a set of Euler fluid equations. In practical simulations, we often encounter computational difficulties that include large displacements, shocks, multi-fluid flows with cavitation, spurious waves reflecting from boundaries and fluid-structure coupling. Existing methods and codes are not able to simultaneously consider all of these characteristics. A robust numerical method that is capable of treating large displacements, capturing shocks, handling two-fluid flows with cavitation, imposing non-reflecting boundary conditions (NRBC) and allowing the movement of fluid grids is required. This method is developed by combining numerical techniques that include a high-order accurate numerical method with a shock capturing scheme, a multi-fluid method to handle explosive gas-water flows and cavitating flows, and an Arbitrary Lagrangian Eulerian (ALE) deformable fluid mesh. These combined approaches are unique for numerically simulating various near-field UNDEX phenomena within a robust single framework. A review of the literature indicates that a fully coupled methodology with all of these characteristics for near-field UNDEX phenomena has not yet been developed. A set of governing equations in the ALE description is discretized by a Runge Kutta Discontinuous Galerkin (RKDG) method. For multi-fluid flows, a Direct Ghost Fluid (DGF) Method coupled with the Level Set (LS) interface method is incorporated in the RKDG framework. The combination of RKDG and DGF methods (RKDG-DGF) is the main contribution of this work which improves the quality and stability of near-field UNDEX flow simulations. Unlike other methods, this method is simpler to apply for various UNDEX applications and easier to extend to multi-dimensions. / Ph. D.

This thesis is concerned with the development of numerical techniques to simulate compressible multi-phase flows, in particular a high-accuracy numerical approach with mesh adaptivity. The Discontinuous Galerkin (DG) method was chosen as the framework for this work for being characterized for its high-order of accuracy -thus low numerical diffusion- and being compatible with mesh adaptivity due to its locality. A DG solver named DiGGIT (Discontinuous Galerkin at the Georgia Institute of Technology) has been developed and several aspects of the method have been studied. The Local Discontinuous Galerkin (LDG) method -an extension of DG for equations with high-order derivatives- was extended to solve multiphase flows using Diffused Interface Methods (DIM). This multi-phase model includes the convection of the volume fraction, which is treated as a Hamilton-Jacobi equation. This is the first study, to the author's knowledge, in which the volume fraction of a DIM is solved using the DG and the LDG methods. The formulation is independent of the Equation of State (EOS) and it can differ for each phase. This allows for a more accurate representation of the different fluids by using cubic EOSs, like the Peng-Robinson and the van der Waals models. Surface tension is modeled with a new numerical technique appropriate for LDG. Spurious oscillations due to surface tension are common to all the capturing schemes, and this new approach presents oscillations comparable in magnitude to the most common schemes. The moment limiter (ML) was generalized for non-uniform grids with hanging nodes that result from adaptive mesh refinement (AMR). The effect of characteristic, primitive, or conservative decomposition in the limiting stage was studied. The characteristic option cannot be used with the ML in multi-dimensions. In general, primitive variable decomposition is a better option than with conservative variables, particularly for multiphase flows, since the former type of decomposition reduces the numerical oscillations at material discontinuities. An additional limiting technique was introduced for DIM to preserve positivity while minimizing the numerical diffusion, which is especially important at the interface. The accuracy-preserving total variation diminishing (AP-TVD) marker for ``troubled-cell' detection, which uses an averaged-derivative basis, was modified to use the Legendre polynomial basis. Given that the latest basis is generally used for DG, the new approach avoids transforming to the averaged-derivative basis, what results in a more efficient technique. Furthermore, a new error estimator was proposed to determine where to refine or coarsen the grid. This estimator was compared against other estimator used in the literature and it showed an improved performance. In order to provide equal order of accuracy in time as in space, the commonly used 3rd-order TVD Runge-Kutta (RK) scheme in the DG method was replaced in some cases by the Spectral Deferred Correction (SDC) technique. High orders in time were shown to only be required when the error in time is significant. For instance, convection-dominated compressible flows require for stability a time step much smaller than is required for accuracy, so in such cases 3rd-order TVD RK resulted to be more efficient than SDC with higher orders. All these new capabilities were included in DiGGIT and have provided a generalized approach capable of solving sub- and super-critical flows at sub- and super-sonic speeds, using a high-order scheme in space and time, and with AMR. Canonical test cases are presented to verify and validate the formulation in one, two, and three dimensions. Finally, the solver is applied to practical applications. Shock-bubble interaction is studied and the effect of the different thermodynamic closures is assessed. Interaction between single-drops and a wall is simulated. Sticking and the onset of splashing are observed. In addition, the solver is used to simulate turbulent flows, where the high-order of accuracy clearly shows its benefits. Finally, the methodology is challenged with the simulation of a liquid jet in cross flow.

A generalized model for bubble coalescence and breakup has been developed, which is based on a comprehensive survey of existing theories and models. One important feature of the model is that all important mechanisms leading to bubble coalescence and breakup in a turbulent gas-liquid flow are considered. The new model is tested extensively in a 1D Test Solver and a 3D CFD code ANSYS CFX for the case of vertical gas-liquid pipe flow under adiabatic conditions, respectively. Two kinds of extensions of the standard multi-fluid model, i.e. the discrete population model and the inhomogeneous MUSIG (multiple-size group) model, are available in the two solvers, respectively. These extensions with suitable closure models such as those for coalescence and breakup are able to predict the evolution of bubble size distribution in dispersed flows and to overcome the mono-dispersed flow limitation of the standard multi-fluid model. For the validation of the model the high quality database of the TOPFLOW L12 experiments for air-water flow in a vertical pipe was employed. A wide range of test points, which cover the bubbly flow, turbulent-churn flow as well as the transition regime, is involved in the simulations. The comparison between the simulated results such as bubble size distribution, gas velocity and volume fraction and the measured ones indicates a generally good agreement for all selected test points. As the superficial gas velocity increases, bubble size distribution evolves via coalescence dominant regimes first, then breakup-dominant regimes and finally turns into a bimodal distribution. The tendency of the evolution is well reproduced by the model. However, the tendency is almost always overestimated, i.e. too much coalescence in the coalescence dominant case while too much breakup in breakup dominant ones. The reason of this problem is discussed by studying the contribution of each coalescence and breakup mechanism at different test points. The redistribution of the gaseous phase from the injection position at the pipe wall to the whole cross section is overpredicted by the Test Solver especially for the test points with high superficial gas velocity. Besides the models for bubble forces, the simplification of the Test Solver to a 1D model has an influence on the redistribution process. Simulations performed using CFX show that a considerable improvement is achieved with comparison to the results delivered by the standard closure models. For the breakup-dominant cases, the breakup rate is again overestimated and the contribution of wake entrainment of large bubbles is underestimated. Furthermore, inlet conditions for the liquid phase, bubble forces as well as turbulence modeling are shown to have a noticeable influence, especially on the redistribution of the gaseous phase.

De nombreux procédés industriels nécessitent la manipulation de matériaux sous forme pulvérulente. L’émission de poussières générée par leur manipulation peut s’avérer dangereuse pour la santé des travailleurs ou bien causer un risque d’explosion. Afin de mieux comprendre les mécanismes de dispersion des poussières, le cas de la décharge d’un silo est étudié par simulation numérique avec une approche Euler-Euler. Deux configurations ont été étudiées au cours de cette thèse. La première, sans silo, a permis d’étudier l’influence de perturbations de vitesses imposées à l’entrée de la chambre de dispersion en lieu et place du silo. Cette étude a révélé que ces perturbations peuvent influencer l’élargissement du panache de poudre. Seules les perturbations avec une corrélation temporelle ont généré une ouverture importante du jet tombant semblable à celle relevée expérimentalement. Dans la deuxième configuration, le silo et la chambre de dispersion sont représentés afin d’étudier le couplage entre la dispersion du jet et l’écoulement dans le silo. L’une des difficultés de ces simulations est de prédire les différents régimes d’écoulements granulaires, allant de l’état quasi-statique dans le silo au régime très dilué lors de la dispersion du jet tombant, en passant par le régime collisionnel à la sortie du silo. La théorie cinétique permet de modéliser le régime dilué et collisionnel. En revanche pour la partie quasi-statique un modèle semi-empirique a été utilisé, implémenté et validé sur différentes configurations. La seconde étude a montré l’importance du rapport entre le diamètre de l’orifice et le diamètre des particules sur la structure du jet. En effet, lorsque ce paramètre est faible, le coeur du jet se contracte immédiatement après la sortie du silo puis s’ouvre en aval. Pour des valeurs grandes, l’ouverture du jet est négligeable. Cependant, il semblerait que l’angle du silo modifie le comportement de l’écoulement, ce qui nécessitera des études supplémentaires. / A wide range of industrial processes requires the handling of granular material in a pulverulent form. The subsequent dust emissions due to these processes can be harmful to the health of workers or hazardous explosion risks. In order to understand dust dispersion mechanisms, a case of a free falling granular jet discharged from a silo is studied by numerical simulations using an Euler-Euler approach. Two types of numerical simulation are conducted. First, the influence of velocity fluctuations at the inlet chamber is studied on the plume behavior, instead of the silo. This study reveals that fluctuations are enable to reproduce the jet expansion. It is established that only fluctuations with temporal correlation generate a large jet opening similar to the experiment. The second type of setup shows the coupling between the silo and the chamber. One of the major challenges is the ability to predict the different flow regimes going from quasi-static regime inside the silo, to the very dilute regime in the dust spread and include the collisional regime occurs through the silo. Kinetic theory allows modeling of the dilute and collisional regime. By contrast, frictional models have been used, implemented and validated in different cases. The second study highlights the key role of the ratio defined by the orifice diameter on the particle diameter. Indeed, when this parameter is small, the jet powder core contracts immediately after the exit of the silo dump plane and expands downstream. For high values, the granular jet does not exhibit dispersion anymore. This study suggests that the silo half-angle has an impact on the flow field which justifies the need for further investigations.

A generalized model for bubble coalescence and breakup has been developed, which is based on a comprehensive survey of existing theories and models. One important feature of the model is that all important mechanisms leading to bubble coalescence and breakup in a turbulent gas-liquid flow are considered. The new model is tested extensively in a 1D Test Solver and a 3D CFD code ANSYS CFX for the case of vertical gas-liquid pipe flow under adiabatic conditions, respectively. Two kinds of extensions of the standard multi-fluid model, i.e. the discrete population model and the inhomogeneous MUSIG (multiple-size group) model, are available in the two solvers, respectively. These extensions with suitable closure models such as those for coalescence and breakup are able to predict the evolution of bubble size distribution in dispersed flows and to overcome the mono-dispersed flow limitation of the standard multi-fluid model. For the validation of the model the high quality database of the TOPFLOW L12 experiments for air-water flow in a vertical pipe was employed. A wide range of test points, which cover the bubbly flow, turbulent-churn flow as well as the transition regime, is involved in the simulations. The comparison between the simulated results such as bubble size distribution, gas velocity and volume fraction and the measured ones indicates a generally good agreement for all selected test points. As the superficial gas velocity increases, bubble size distribution evolves via coalescence dominant regimes first, then breakup-dominant regimes and finally turns into a bimodal distribution. The tendency of the evolution is well reproduced by the model. However, the tendency is almost always overestimated, i.e. too much coalescence in the coalescence dominant case while too much breakup in breakup dominant ones. The reason of this problem is discussed by studying the contribution of each coalescence and breakup mechanism at different test points. The redistribution of the gaseous phase from the injection position at the pipe wall to the whole cross section is overpredicted by the Test Solver especially for the test points with high superficial gas velocity. Besides the models for bubble forces, the simplification of the Test Solver to a 1D model has an influence on the redistribution process. Simulations performed using CFX show that a considerable improvement is achieved with comparison to the results delivered by the standard closure models. For the breakup-dominant cases, the breakup rate is again overestimated and the contribution of wake entrainment of large bubbles is underestimated. Furthermore, inlet conditions for the liquid phase, bubble forces as well as turbulence modeling are shown to have a noticeable influence, especially on the redistribution of the gaseous phase.

Dans certaines simulations numériques exigeantes de mécanique des fluides, ilest nécessaire de simuler des écoulements multiphasiques impliquant de nombreuses contraintes simultanées : nombre de fluides important, évolutions compressibles à la fois isentropes et fortement choquées, équations d’états variables et contrastées, déformations importantes et transport surdes longues distances. Afin de remplir ces objectifs de manière robuste, il est nécessaire que la cohérence thermodynamique du schéma numérique soit vérifiée.Dans le premier chapitre, un schéma de type Lagrange plus projection est proposé pour la simulation d’écoulements diphasiques avec un modèle squelette à six équations et sans termes de dissipation. L’importance de la propriété de préservation des écoulements isentropiques est mise en évidence à l’aide d’une comparaison avec des résultats issus de la littérature pour le test deRansom. Ce chapitre souligne aussi certaines limitations de l’approche Lagrange plus projection pour simuler des modèles multiphasiques.Afin de pallier à ces limitations, une nouvelle procédure de dérivation est proposée afin de construire un schéma mimétique pour la simulation d’écoulements instationnaires compressibles dans un formalisme ALE direct (Arbitrary Lagrangian–Eulerian). La possibilité de choisir a prioriles degrés de liberté permet de s’inscrire dans une continuité avec les schémas historiques décalés, tout en imposant les conservations au niveau discret. L’équation de quantité de mouvement discrèteest obtenue par application d’un principe variationnel, assurant par construction la cohérence thermodynamique des efforts de pression. Cette approche est appliquée au cas d’écoulements monofluides comme preuve de concept au Chapitre 3, puis elle est étendue au cas d’écoulements à Nphasescompressibles au Chapitre 4. Des tests mono et multiphasiques montrent un comportement satisfaisant en terme de conservativité, versatilité aux mouvements de grilles et robustesse. / In some highly demanding fluid dynamics simulations, it appears necessary tosimulate multiphase flows involving numerous constraints at the same time : large numbers of fluids, both isentropic and strongly shocked compressible evolution, highly variable and contrasted equations of state, large deformations, and transport over large distances. Fulfilling such a challengein a robust and tractable way demands that thermodynamic consistency of the numerical scheme be carefully ensured.In the first chapter, a Lagrange plus remap scheme is proposed for the simulation of two-phase flows with a dissipation-free six-equation bakcbone model. The importance of the property of isentropic flow preservation is highlighted with a comparison with Ransom test results fromthe literature. This chapter also also point out certain limitations of the Lagrange plus remap approach for multiphase simulations.In order to overcome these limitations, a novel derivation procedure is proposed to construct a mimetic scheme for the simulation of unsteady and compressible flows in a direct ALE (ArbitraryLagrangian-Eulerian) formalism. The possibility to choose a priori the degrees of freedom allows to obtain a continuity with historical staggered scheme, while imposing conservativity at discretelevel. The discrete momentum evolution equation is obtained by application of a variational principle, thus natively ensuring the thermodynamic consistency of pressure efforts. This approach is applied to single-fluid flows as a proof of concept in Chapter 3, then it is extended to N-phasecompressible flows in Chapter 4. Single- and multi-phase tests show satisfactory behavior in terms on conservation, versatility to grid motions, and robustness.

Cette thèse s'intéresse à la simulation numérique de l'explosion vapeur. Ce phénomène correspond à une vaporisation instantanée d'un volume d'eau liquide entraînant un choc de pression. Nous nous y intéressons dans le cadre de la sûreté nucléaire. En effet, lors d'un accident entraînant la fusion du cœur du réacteur, du métal fondu pourrait interagir avec de l'eau liquide et entraîner un tel choc. On voudrait alors connaître l'ampleur de ce phénomène et les risques d'endommagements de la centrale qu'il implique. Pour y parvenir, nous utilisons pour modèle les équations d'Euler dans un cadre Lagrangien. Cette description a l'avantage de suivre les fluides au cours du temps et donc de parfaitement conserver les interfaces entre l'eau liquide et sa vapeur. Pour résoudre numériquement les équations obtenues, nous développons un nouveau schéma de type Godunov utilisant des flux nodaux. Le solveur nodal développé durant cette thèse ne dépend que de la répartition angulaire des variables physiques autour du nœud. De plus, nous nous intéressons aux changements de phase liquide-vapeur. Nous proposons une méthode pour les prendre en compte et mettons en avant les avantages qu'il y a à l'implémentation de ce phénomène dans un algorithme Lagrangien. / This thesis studies numerical simulation of steam explosion. This phenomenon correspond to a fast vaporization of a liquid leading to a pressure shock. It is of interest in the nuclear safety field. During a core-meltdown crisis, molten fuel rods interacting with water could lead to steam explosion. Consequently we want to evaluate the risks created by this phenomenon.In order to do it, we use Euler equations written in a Lagrangian form. This description has the advantage of following the fluid motion and consequently preserves interfaces between the liquid and its vapor. To solve these equations, we develop a new Godunov type scheme using nodal fluxes. The nodal solver developed here only depends on the angular repartition of the physical variables around the node.Moreover, we study liquid-vapor phase changes. We describe a method to take it into account and highlight the advantages of using this method into a Lagrangian framework.