Search

International audience; In this paper, we present simulations of turbulent thermal convection driven by free-surface evaporation above and by a heated wall below. A novel algorithm is proposed for predicting evaporation rates at a free surface, which we then validate against experimental data. At the top of a cuboidal domain, a shear-free boundary condition acts as an approximation of the free surface. We first focus on a domain of aspect ratio Γ = 1, where a fully resolved direct numerical simulation is carried out at a moderate Rayleigh number, Ra = 1.2 × 107, and we compare flow statistics with a Large-Eddy Simulation (LES) on a coarse grid. Both the fully resolved simulation and the LES predict well the time- and area-averaged evaporation rate and free surface temperature when compared with the experimental data. Next, we carry out a series of LES with an increasing lower wall temperature and, consequently, Ra. We then validate the evaporation model by comparing LES predictions of the time- and area-averaged mass flux and temperature at the upper boundary against the experimental measurements. The aspect ratio of the domain is then reduced, and we show, for the first time, the transition to a dual-roll state of the large-scale circulation (LSC) at the aspect ratio of Γ = 1/4 in a cuboidal domain. The temperature and velocity distributions at the free surface are impacted by the state of the LSC. However, we find that the water-side turbulence and aspect ratio play a negligible role on the evaporation rate above, in accordance with experimental observations.

This study deals with the fundamental problem of combustion dynamics in gas turbine combustors operating with liquid fuel. In this framework the present work proposes the study of an academic liquid fueled combustor sensitive to thermoacoustic instabilities, simulated via high-fidelity Large Eddy Simulations. The experimental setup addressed is SICCA-spray from EM2C laboratory featuring both stable and unstable flames depending on the combustion chamber length. The proposed analysis, based on the Euler-Lagrange modeling approach, studies the impact of the spray injection angle θ on both the stable flame and the triggering of the longitudinal combustor acoustic mode when using a longer quartz tube. For the liquid injection modeling, the FIM-UR semi-empirical model is adopted with three different θ values: θ = 35 ∘ , 45 ° and 60 ° . In stable conditions, the spray angle is proven to have a negligible impact on the flame anchoring point, however, the mean flame length and fuel distribution are found to be slightly modified by the velocity at which droplets enter the combustion chamber. For the thermoacoustically unstable conditions, two well-established stable limit cycles with the same frequency and similar amplitudes are obtained when fuel is injected at θ = 45 ∘ and 60 ° . Contrarily, the system stabilizes when θ = 35 ∘ pointing to the importance of the dynamics of the liquid film layer formed inside the injector for this setup. Likewise, this liquid film layer dynamics and its modeling appear critical as already suggested by previous studies on the same configuration. The detailed analysis of the thermoacoustically unstable different predictions is then performed through the investigation of the spatial fields of the local Rayleigh index obtained following the novel extension in the frequency domain of the Rayleigh criterion complemented by the application of Dynamic Mode Decomposition. It confirms that the injection angle of the liquid spray has a significant effect on the thermoacoustic response of the system. Indeed the influence of θ on the dynamics of the liquid fuel when entering the combustion chamber is proven to have an impact on the synchronization mechanism governing the liquid phase with respect to acoustics sustaining the observed limit cycles. More specifically, couplings at the liquid phase level are evidenced by introducing two novel indices correlating the fluctuations of liquid fuel volume fraction and evaporation rate with pressure.

International audience; In particle-laden flows, particles rebound from rough wall can influence significantly properties of particle and fluid phase which can be important in number of industry processes. 2D stochastic procedures for such a rebound have been developed for representing the effect of wall roughness although such mechanism is generally 3D due to the particle incident velocity and roughness geometry. Therefore we have developed a general approach allowing to represent the 3D particle bouncing on isotropic or non-isotropic rough wall. Thus effect of the isotropic and non-isotropic roughness is evaluated using DPS/LES in horizontal channel flow. Inter-particle collisions are calculated using soft sphere model. Different LES subgrid models of pure fluid flow are examined and influence of subgrid turbulent fluid velocity on particle statistics is evaluated. Detailed information regarding particle velocity statistics are discussed with respect to wall roughness characteristics.

International audience; Large Eddy Simulations (LES) are performed on a Vertical Axis Hydrokinetic Turbine (VAHT) at two different solidities, so as to enable a more complete physical description of the flow than for classic statistical calculations. To analyze the turbine performances, local quantities are defined to evaluate the contribution of the different turbine elements to the global VAHT power coefficient. For a deeper analysis of the major losses, the real turbine is also compared with an ideal turbine composed of only three infinite blades. It is observed that the ideal turbine with the lower solidity provides the best performance, but the losses due to the blade tips and the arms strongly increase for the real turbine at the same solidity. Consequently, for the considered real turbine, there is no clear gain to decrease the solidity. Simulations of the ideal turbine are performed for various solidities at their optimal Tip Speed Ratio (TSR) to study the evolution of optimal power coefficient as a function of solidity. A maximum power coefficient is obtained for a small value of the solidity. This is explained because this optimal solidity is obtained when the angles of incidences experienced by the blades at optimal TSR avoid a penalizing dynamic stall phenomenon but are high enough to produced an important positive torque. The design of an efficient turbine has then to limit losses, to be able to use small solidity and then to avoid dynamic stall phenomenon by having a high optimal TSR.

In this article, numerical results from a highly resolved large eddy simulations (LES) of an air-helium buoyant jet developing in a two vented cavity are presented. The simulated configuration mimics the helium-release experiment carried out at CEA Saclay in the framework of security assessment of indoor used hydrogen-based systems. The height of the enclosure was chosen so that a laminar-turbulent transition occurs approximately at the middle of the upstream direction. An exterior region, of different spatial dimensions, has been modeled in the computational domain in order to move the boundary conditions away from the vent surface and to approach the natural inlet/outlet conditions. A sensitivity analysis regarding the size of the exterior region is presented to define the minimum horizontal extension so that the flow inside the cavity is not furthermore influenced by bigger computational domains. We observe mainly that applying an ambient equilibrium-hydrostatic pressure outlet condition directly on the surface of the vent reduces the volume of the air inflow, and thus predicts larger helium mass inside the cavity, in contrary with the cases where an exterior region is considered. A qualification analysis shows that the sub-grid scale model plays a small role in the calculations and thus implies that the LES predictions approach the direct numerical simulation (DNS) solution. Analysis carried out on the time-averaged helium field depicts a concentration regime that is not classical in such configurations and thus the theoretical model used in safety pre-calculations can not be served.

The prediction of wall fluxes is a significant aspect in the life cycle of combustors, since it allows to prevent eventual wall damages. Heat transfer from flame to the walls is driven, apart from convection, also by radiation of burnt gases inside the chamber. In order to correctly account for both convective and radiative contributions to wall fluxes, the simultaneous solution of the radiative transfer equation (RTE) and the governing equations for reactive flows is required. However, multi-physics simulations where MC methods are coupled to Large Eddy Simulation (LES), remain very costly. The purpose of this study is then to investigate improvements of MC methods, by using an alternative sampling mechanism for numerical integration usually referred to as Quasi-Monte Carlo (QMC) integration. In this study, QMC method is coupled to Large Eddy Simulation (LES) of a configuration where the radiation plays an important role: the methane-air flame investigated during the experimental campaign Oxytec. Coupled and non-coupled simulations are compared and their comparison with experimental data shows that thermal radiation has an impact on both flow and flame topology. Finally a good agreement is found between numerical wall fluxes and experimental conductive fluxes., La prédiction des flux aux parois joue un rôle déterminant dans le cycle de vie des chambres de combustion. Le transfert de chaleur de la flamme aux parois est entraîné, outre la convection, également par le rayonnement des gaz chauds au sein de la chambre. Afin d’intégrer les contributions convectives et radiatives au flux pariétal il est nécessaire de résoudre simultanément l’équation de transfert radiatif et les équations régissant l’écoulement réactif. Quand les méthodes de Monte Carlo sont couplées aux simulations aux grandes échelles (LES), de telles simulations deviennent très coûteuses. L’objectif de cette thèse est donc d’investiguer une technique pour améliorer l’efficacité de la méthode MC, basée sur un mécanisme alternatif d’échantillonnage appelée intégration Quasi-Monte Carlo (QMC). Au cours de cette thèse, la méthode QMC a été couplée à une simulation LES dans une configuration où le rayonnement joue un rôle très important : la flamme méthane-air de la chambre Oxytec. La comparaison entre les simulations couplées et non couplées avec les données expérimentales montre que le rayonnement thermique a un impact sur la topologie de l’écoulement et de la flamme. Enfin, un bon accord est trouvé entre le flux de chaleur pariétal prédit par la simulation et les données expérimentales.

We present numerical results from large eddy simulations (LES) and coarse direct numerical simulations (DNS) of an air-helium buoyant jet rising in a two vented cavity. The geometrical configuration mimics the helium release experimental set-up studied at CEA Saclay in the framework of security assessment of hydrogen-based systems with an indoor usage. The dimension of the enclosure was chosen to ensure a laminar-turbulent transition occurring at about the middle height of the cavity. This study focuses mainly on three key points : the influence of the boundary conditions on the jet development and its interaction with the exterior environment, the validity of the numerical model which is analyzed by comparing the numerical velocity distribution versus the measured particle image velocimetry (PIV) ones, and finally understanding the distribution of air-helium and the stratification phenomenon that takes place inside the cavity. We observe at first that applying constant pressure outlet boundary conditions directly at the vent surfaces underestimates the volumetric flow rate of air entering the enclosure and thus overestimate the helium mass inside the cavity. On the contrary, modelling an exterior region in the computational domain better predicts the air flow-rate entrance and numerical results matches better with the experimental PIV data. It has been figured out that the coarse DNS predictions match better with the velocities PIV measurements, compared to the LES. Numerical prediction of the helium field depicts a homogeneous layer formed at the top of the cavity, with a concentration in good agreement with the theoretical model of Linden et al. 1990. However, the position and the thickness of the layer do not correspond to the theory. This is mainly due to the direct interactions between the buoyant jet and both the solid boundaries of the cavity and the exterior environment. Statistical analysis regarding the buoyancy production of the turbulent kinetic energy (TKE) served to identify the limits of the buoyant jet.; Nous cherchons à modéliser numériquement un jet flottant air-hélium dans une cavité avec deux ouvertures à partir de simulations aux grandes échelles (LES) et de simulations numériques directes (DNS). La configuration considérée est basée sur une étude expérimentale menée au CEA de Saclay reproduisant une fuite d'hydrogène en environnement confiné. La dimension de la cavité a été choisie pour permettre une transition laminaire-turbulent intervenant environ à la mi-hauteur de la cavité. Cette étude porte principalement sur trois points majeurs : l'influence des conditions aux limites sur le développement du jet et son interaction avec l'environnement extérieur, la validité du modèle numérique qui est analysée en comparant la distribution de vitesse obtenue numériquement aux mesures expérimentales (PIV) et, enfin, la compréhension de la distribution air-hélium et le phénomène de stratification qui s'établit à l'intérieur de la cavité. Nous observons dans un premier temps que des conditions limites de pression constante appliquées directement au ras des évents conduisent à une sous-estimation du débit volumique d'air entrant dans la cavité et donc à une surestimation de la masse de l'hélium à l'intérieur de la cavité, ce qui n'est pas acceptable dans un contexte d'évaluation du risque hydrogène. En revanche, la prise en compte, dans le domaine de calcul, d'une région extérieure à la cavité prédit correctement le flux d'air entrant. Les résultats numériques sont alors en bon accord avec les données PIV. Il a été montré que les prédictions de la DNS, par rapport à la LES, concordent mieux avec les mesures de vitesse par PIV. Le champ de concentration prédit numériquement présente une couche homogène en haut de la cavité, dont la concentration est en accord avec le modèle théorique de Linden et al. 1990. Cependant, sa position et son épaisseur ne correspondent pas au modèle. Ceci est principalement dû aux interactions directes entre le jet flottant et, d'une part, avec les limites solides de la cavité et d'autre part, avec l'environnement extérieur. L'analyse statistique concernant la production de la flottabilité de l'énergie cinétique turbulente (TKE) a permis d'identifier les limites du jet flottant.

We present numerical results from large eddy simulations (LES) and coarse direct numerical simulations (DNS) of an air-helium buoyant jet rising in a two vented cavity. The geometrical configuration mimics the helium release experimental set-up studied at CEA Saclay in the framework of security assessment of hydrogen-based systems with an indoor usage. The dimension of the enclosure was chosen to ensure a laminar-turbulent transition occurring at about the middle height of the cavity. This study focuses mainly on three key points : the influence of the boundary conditions on the jet development and its interaction with the exterior environment, the validity of the numerical model which is analyzed by comparing the numerical velocity distribution versus the measured particle image velocimetry (PIV) ones, and finally understanding the distribution of air-helium and the stratification phenomenon that takes place inside the cavity. We observe at first that applying constant pressure outlet boundary conditions directly at the vent surfaces underestimates the volumetric flow rate of air entering the enclosure and thus overestimate the helium mass inside the cavity. On the contrary, modelling an exterior region in the computational domain better predicts the air flow-rate entrance and numerical results matches better with the experimental PIV data. It has been figured out that the coarse DNS predictions match better with the velocities PIV measurements, compared to the LES. Numerical prediction of the helium field depicts a homogeneous layer formed at the top of the cavity, with a concentration in good agreement with the theoretical model of Linden et al. 1990. However, the position and the thickness of the layer do not correspond to the theory. This is mainly due to the direct interactions between the buoyant jet and both the solid boundaries of the cavity and the exterior environment. Statistical analysis regarding the buoyancy production of the turbulent kinetic energy (TKE) served to identify the limits of the buoyant jet.; Nous cherchons à modéliser numériquement un jet flottant air-hélium dans une cavité avec deux ouvertures à partir de simulations aux grandes échelles (LES) et de simulations numériques directes (DNS). La configuration considérée est basée sur une étude expérimentale menée au CEA de Saclay reproduisant une fuite d'hydrogène en environnement confiné. La dimension de la cavité a été choisie pour permettre une transition laminaire-turbulent intervenant environ à la mi-hauteur de la cavité. Cette étude porte principalement sur trois points majeurs : l'influence des conditions aux limites sur le développement du jet et son interaction avec l'environnement extérieur, la validité du modèle numérique qui est analysée en comparant la distribution de vitesse obtenue numériquement aux mesures expérimentales (PIV) et, enfin, la compréhension de la distribution air-hélium et le phénomène de stratification qui s'établit à l'intérieur de la cavité. Nous observons dans un premier temps que des conditions limites de pression constante appliquées directement au ras des évents conduisent à une sous-estimation du débit volumique d'air entrant dans la cavité et donc à une surestimation de la masse de l'hélium à l'intérieur de la cavité, ce qui n'est pas acceptable dans un contexte d'évaluation du risque hydrogène. En revanche, la prise en compte, dans le domaine de calcul, d'une région extérieure à la cavité prédit correctement le flux d'air entrant. Les résultats numériques sont alors en bon accord avec les données PIV. Il a été montré que les prédictions de la DNS, par rapport à la LES, concordent mieux avec les mesures de vitesse par PIV. Le champ de concentration prédit numériquement présente une couche homogène en haut de la cavité, dont la concentration est en accord avec le modèle théorique de Linden et al. 1990. Cependant, sa position et son épaisseur ne correspondent pas au modèle. Ceci est principalement dû aux interactions directes entre le jet flottant et, d'une part, avec les limites solides de la cavité et d'autre part, avec l'environnement extérieur. L'analyse statistique concernant la production de la flottabilité de l'énergie cinétique turbulente (TKE) a permis d'identifier les limites du jet flottant.

International audience; is paper presents numerical analysis of an airfoil geometry inspired by the down coat of the night owl. e objective is to understand the mechanisms of airfoil trailing edge noise reduction that has been observed with such designs in previous experiments. e bioinspired geometry consists of an array of "fences" that are applied near the trailing edge of the NACA-baseline airfoil. Wall-resolved large eddy simulations are performed over the baseline and the bioinspired airfoil geometries and the aeroacoustic performance of the two geometries are contrasted. Both models are simulated at chord-based Reynolds number Re c =5× 10 5 , flow Mach number, M ∞ =0.2, and angle of a ack, α =0 •. Farfield noise spectra comparisons between the baseline and the bioinspired airfoil near the airfoil trailing edge show reductions with the fences of up to dB. e simulations reveal that the fences li turbulence eddies away from the airfoil trailing (sca ering) edge hence reducing sca ering efficiency. ese findings suggest that one of the mechanisms of noise reduction is the increased source-sca ering edge separation distance.

Colloque avec actes et comité de lecture. Internationale.; International audience; This work is devoted to the security assessment of systems using hydrogen as energy carriers, where the problematic is mainly with the hydrogen leakage which is not only dangerous, but can be destructive when mixed with air. The french Atomic Energy Commission (CEA) is showing a great interest to the presented problem and have been participating in several concerned projects that focus on reducing the risks in such systems. We present a numerical Large Eddy Simulation (LES) study to mimic the experiment held at CEA-Saclay, where the above problem is modelled by injecting Helium gas (replacing the Hydrogen in the real problem) through a cylindrical pipe that enters a two-vented parallelepiped cavity filled initially with air. Understanding the mixing/dispersion of the binary fluids at the transient and quasi-steady state is one of the goals that we target on. In spite of the fact that the experiments had been held over a wide range of varied injection flow-rates, we restrict our study to the case of a 5 Nl/min injection resulting in a formation of a buoyant jet that is mainly influenced by momentum in the near source and buoyancy dominant upstream. The dimensions of the cavity had been chosen so that the interesting physical phenomena of the problem is captured; basically the transition from a laminar regime to a turbulent one and the self-similarity of the jet horizontal profiles. In this paper, we present some of our LES results obtained by making use of the CEA code TRUST and the Smagoranski model. We point out basically to the influence of the exterior domain, if it is/isn't considered in the computational domain, on the flow inside the cavity. We demonstrate that imposing classical fixed pressure boundary conditions directly at the vents positions, when compared with the experimental Particle Image Velocitimetry (PIV) measurements, tends to under-estimate the air flow in the cavity and thus leading to over-estimate the maximum Helium concentration. Different simulations had been achieved by varying the size of the exterior domain which is opened from all directions not in contact with the plexi-glass of the cavity. Following a space and a temporal convergence study, our statistical results seems to be satisfactory when compared to the experimental data. This work is a first step as we look on developing new outlet boundary conditions which can be directly applied at the vents positions for a perfect Direct Numerical Simulation (DNS), at the Kolmogorov scales, with a better computational cost.

The Smoothed Particle Hydrodynamics (SPH) method, often used for the modelling of the Navier- Stokes equations by a meshless Lagrangian approach, is revisited from the point of view of Large Eddy Simulation (LES). To this aim, the LES filtering procedure is recast in a Lagrangian framework by defining a filter that moves with the positions of the fluid particles at the filtered velocity. It is shown that the SPH smoothing procedure can be reinterpreted as a sort of LES Lagrangian filter- ing, and that, besides the terms coming from the LES convolution, additional contributions (never accounted for in the SPH literature) appear in the equations when formulated in a filtered fashion. Appropriate closure formulas are derived for the additional terms and a preliminary numerical test is provided to show the main features of the proposed LES-SPH model.

Heat exchangers are widely used in the power generation industries. The cross-flow type of heat exchangers are more common. The rate of heat transfer is enhanced by operating the heat exchangers at higher flow rates by means of the increased flow turbulence. Although, the high flow rate operations are favoured, there are side effects in terms of the flow-induced vibrations. In the last few decades, the topic (flow induced vibrations in heat exchanger tube bundles) is studied extensively, especially in order to understand the fluidelastic vibration. The preventive measures can be taken in other types of vibration mechanisms, such as the vortex induced vibrations and the acoustic resonance in tube bundles. The turbulence induced vibrations generally take long term to deteriorate the performance of heat exchangers, hence it involves a lesser risk of immediate damage to the heat exchangers. The failure due to the fluidelastic instability occurs suddenly and it can pose a serious risk in the plant operations. Besides the devastating nature of the fluidelastic instability, it is not well understood yet. In the first part of this thesis, the fluidelastic instability is explored by means of performing numerical simulations. The flow induced vibrations in the heat exchanger tube bundles are reviewed historically. The other mechanisms of vibrations, namely, vortex induced vibrations, turbulent buffeting and acoustic resonance in the tube arrays are briefed. In addition, thetheoretical models of the fluidelastic instability are revised in order to understand the different approaches used to model the instability. Computational Fluid Dynamics (CFD) simulations are performed, first by using the Unsteady Reynolds Averaged Navier-Stokes (URANS) approach of modeling the flow turbulence, in order to verify the capability of URANS modelsto predict the instability thresholds dynamically. Secondly, the transient nature of fluidelastic instability is investigated by means of the Large Eddy Simulations (LES) approach of the turbulence modeling. Although the LES approach is computationally expensive in comparison with the URANS approach, the dynamic interactions between the interstitial fluid flow and a single tube from an in-line tube bundle are well captured by the LES. The post-processing of the LES results is comprised of the dynamics of fluid forces acting on a single cylinder from an array, transient surface pressure profiles on the cylinder and the interstitial velocity flow fields as a consequence of the increasing flow velocity until the onset of fluidelastic instability. A mathematical model for the fluidelastic instability is developed based on the transient interaction between the interstitial flow through an in-line cylinders array and a single cylinder form the array. Although there are significant advances in the computers today, the Direct Numerical Simulations (DNS) of large dynamic systems are infeasible. Model reduction also known as Reduced-Order Modeling (ROM) has gained an importance in almost all fields of computational sciences. In the second part of the thesis, firstly, a short introduction to the model orderreduction is provided. The Proper Orthogonal Decomposition (POD) and Galerkin projections are commonly used in model reduction of the fluid systems. Almost all reduced-order models derived from the traditional POD-Galerkin ROM require the stability enablers. A novel Galerkin-free approach for model reduction of the Navier-Stokes equations is proposed in thisthesis. The method uses the periodicity of the POD time coefficients and a linear interpolation technique in order to construct the off-reference reduced solutions. A test case of the flow past a cylinder at low Reynolds numbers (Re ∼ 125) is used for the demonstration of the proposed ROM. In the formulation of the proposed Galerkin-free ROM, the variables of a dynamicalsystem are treated independently. Therefore, the method can be conveniently extended for the multi-physics dynamical systems. Lastly, the method of Galerkin-free ROM is applied to a fluid-structure interaction problem, where the moving mesh is a part of the solution state vector. A test case of the vortex induced vibration in a cylinder at Reynolds number Re = 100and the mass ratio as the controlling parameter is considered for the demonstration., Les échangeurs de chaleur sont largement utilisés dans les industries de production d’énergie. Le type d’échangeurs de chaleur à flux croisés sont les plus fréquents. Le taux de transfert de chaleur est amélioré en faisant fonctionner les échangeurs de chaleur à des débits plus élevés au moyen de la turbulence d’écoulement accrue. Bien que, les opérations de débit élevé sont favorisés, il y des effets secondaires en termes de vibrations induites par l’écoulement. Dans les dernières décennies, le sujet (débit vibrations induites dans l’échangeur de chaleur des faisceaux de tubes) a été étudié en profondeur, en particulier afin de comprendre les vibrations fluide-élastique. Des mesures préventives peuvent être prises dans d’autres types de mécanismes de vibration, comme les vibrations induites par vortex et la résonance acoustiquedans les faisceaux de tubes. Les vibrations induites par turbulence prennent généralement longtemps à détériorer les structures des échangeurs de chaleur, d’où elle implique un risque moindre de dommages immédiats aux échangeurs de chaleur. L’échec dû à l’instabilité fluide-élastique se produit dans une question d’heures et il peut poser un risque grave dans les opérations de l’usine; alors même que la nature dévastatrice. De l’instabilité fluide-élastique n’est pas encore bien comprise. Dans la première partie de cette thèse, l’instabilité fluide-élastique est explorée en utilisant des simulations numérique. Les vibrations induites par l’écoulement dans les faisceaux de tubes sont examinés historiquement. Les autres mécanismes de vibrations, à savoir, des vibrations induites par vortex, et par la turbulente et la résonance acoustique dans les fais-ceaux de tubes sont bien documentés. De plus, les modèles théoriques de l’instabilité fluide-élastique sont examinés. Les simulations de Mécanique des Fluides Numérique (MFN) ou en anglais Computational Fluid Dynamics (CFD) sont réalisées, d’abord en utilisant l’approache Unsteady Reynolds Averaged Navier-Stokes (URANS) pour la modélisation de la turbulence,afin de vérifier la capacité des modèles de URANS à prédire les seuils d’instabilité dynamique. Deuxièmement, la nature transitoire de l’instabilité fluide-élastique est étudiée au moyen de la Simulation Grandes Échelles (Large Eddy Simulation LES) des structure turbulent. Bien que l’approche LES est chère en comparaison avec l’approche URANS, les interactions dy-namiques entre le flux de liquide interstitiel et un seul tube à partir d’un faisceau de tubes en ligne sont bien capturés par la LES. Le post-traitement des résultats LES composée de la dynamique des forces du fluide agissant sur un seul cylindre dans un groupe, les profils de pression sur le cylindre et les champs d’écoulement de vitesse interstitiels en augmentant lavitesse d’écoulement jusqu’à l’apparition de l’instabilité fluide-élastique. Un modèle mathématique pour l’instabilité fluide-élastique est développé sur la base de l’interaction transitoire entre le débit interstitiel à travers une gamme de cylindres en ligne et une seule forme de cylindre du faisceau. Bien qu’il y ait des progrès significatifs dans les capacités de calcul, la simulation numérique direct (Direct Numerical Simulation DNS) des grands systémes dynamique sont touhours infaisable. La Réduction de modèles aussi connue comme Reduced Order Modeling (ROM) a acquis une importance dans presque tous les domaines des sciences informatiques. Dans ladeuxième partie de la thèse, d’une part, une introduction à la réduction de l’ordre de modèle est fournie. La décomposition en modes propres orthogonaux (Proper Orthogonal Decomposition POD) et les projections de type Galerkin sont couramment utilisées dans la réduction de modèle des écoulement de fluides. Presque tous les modèles d’ordre réduit provenant de la technique POD-Galerkin exigent des l’introduction de stabilisations. Une nouvelle approche de type “Galerkin-free" pour la réduction de modèle des équations de Navier-Stokes est proposée dans cette thèse. La méthode utilise la périodicité des coefficients temporelle de POD et une technique d’interpolation linéaire afin de construiredes solutions d’ordre réduit. Le cas test de l’écoulement autour d’un cylindre à faible nombre de Reynolds (Re ∼ 125) est utilisé pour la démonstration de modéle d’ordre réduit proposé. Dans la formulation de modéle d’ordre réduit “Galerkin-free" POD, les variables d’un système dynamique sont traitées indépendamment. Par conséquent, la méthode peut être facile-ment étendue aux systèmes dynamiques multi-physiques. Enfin, la technique de Galerkin-free modéle d’ordre réduit est appliquée à un problème d’interaction fluide-structure, où la maille mobile fait partie du vecteur d’état de la solution. Un cas test de vibrations d’induite par vortex d’un cylindre pour un nombre de Reynolds de 100 avec le rapport de masse comme paramètre de contrôle est considéré pour la démonstration.

International audience; Three LES models devoted to the NO prediction in under-adiabatic furnaces are evaluated in this paper: the NORA (NO relaxation Approach) model, based on the NO relaxation towards equilibrium, the linear model (LM) which employs a linear relation to rescale the NO consumption rate, and a new model, DF-NORA, in which the linear approximation of the LM is replaced by a tabulation of the reaction rate as a function of a NO progress variable. To generate this table, NO relaxation complex chemistry calculations are used like in NORA, but the homogeneous reactor is replaced by a steady laminar diffusion flame. These models are validated on Sandia Flame D and on the flameless case of Verissimo et al. (Ener. Fuel. 25, 2469–2480 ([32])). For both cases, NORA underpredicts the NO production due to its insensitivity to strain, while LM overpredicts NO by a factor 2 on Flame D and a factor 13 on the flameless case. DF-NORA presents the best prediction with a maximal underprediction of 30% on Flame D and an over-prediction of 30% on the final NO yield of the flameless case. The impact of a radiative source term is also assessed on Flame D, showing a local decrease of NO by less than 7% compared to the adiabatic calculation for the DF-NORA model.

International audience; This study analyses the mixing and thermal fluctuations induced in a mixing tee junction with circularcross-sections when cold water flowing in a pipe is joined by hot water from a branch pipe. This config-uration is representative of industrial piping systems in which temperature fluctuations in the fluid maycause thermal fatigue damage on the walls. Implicit large-eddy simulations (LES) are performed for equalinflow rates corresponding to a bulk Reynolds number Re= 39,080. Two different thermal boundary conditions are studied for the pipe walls; an insulating adiabatic boundary and a conducting steel wall boundary. The predicted flow structures show a satisfactory agreement with the literature. The velocityand thermal fields (including high-order statistics) are not affected by the heat transfer with the steelwalls. However, predicted thermal fluctuations at the boundary are not the same between the flowand the solid, showing that solid thermal fluctuations cannot be predicted by the knowledge of the fluidthermal fluctuations alone. The analysis of high-order turbulent statistics provides a better understand-ing of the turbulence features. In particular, the budgets of the turbulent kinetic energy and temperaturevariance allows a comparative analysis of dissipation, production and transport terms. It is found that theturbulent transport term is an important term that acts to balance the production. We therefore usea priori tests to evaluate three different models for the triple correlation

The interactions surrounding the coupling between surface energy balance and a boundary layer with shallow cumuli are investigated using the National Center for Atmospheric Research’s large-eddy simulation code coupled to the Noah land surface model. The simulated cloudy boundary layer is based on the already well-documented and previously simulated 21 June 1997 case at the Atmospheric Radiation Measurement Southern Great Plains central facility. The surface energy balance response to cloud shading is highly nonlinear, leading to different partitioning between sensible and latent heat flux compared to the surface not impacted by cloud. The evaporative fraction increases by about 2%–3% in the presence of shallow cumuli at the regional scale but can increase by up to 30% at any individual location. As expected, the cloud’s reduction of solar irradiance largely controls the surface’s response. However, the turbulence and secondary circulations associated with the cloud dynamics increases the surface flux variability. Even though they are less than 1 km in horizontal scale, the cloud-induced surface heterogeneities impact the vertical flux of heat and moisture up to approximately 20% of the height of the subcloud layer zsl, higher than the surface layer’s typical extent. Above 0.2zsl, the cloud root tends to amplify the drying and the cooling of the subcloud layer. Near the entrainment zone, the cloud-induced latent heat flux increase and sensible heat flux decrease compensate each other with respect to total buoyancy and therefore do not significantly modify the subcloud-layer entrainment rate over large time scales.

Co-encadrement de la thèse : Élie Hachem; The main objective of this thesis is to develop a complete simulation code for large pieces (up to several meters) metal casting. This kind of processes reveals several coupled physical phenomena and needs adapted numerical methods. Molten metal has usually a low viscosity (around 10-6 m2/s) hence in order to simulate the filling process the use of a turbulence model is required. At the same time, the liquid/air interface must be tracked along with the numerical simulation. In the first part of this work, a stabilized finite element method base on the Variational MultiScale approach (VMS) required for computing the incompressible flows is proposed and analyzed. The latter approach has proved to be efficient in simulating moderate Reynolds numbers, and combined with a convected level-set method, it showed an ability to determine accurately and at any moment the position of the liquid / air interface. Turbulence is taken into account by developing a dynamic Large Eddy Simulation (L.E.S.) model with no empirical data. For validation, each of these methods has been compared to experimental, numerical or analytical results. We also designed our own water flow experimental setting in order to validate our solvers coupling for a representative case. Another feature of these processes having a relatively long duration (over several ten minutes) is the influence of conjugate heat transfers. This can play a major role on the solidification of the metal that is being filled. It is then convenient to develop a stabilized finite element method for convection dominated heat transfer. This method should take into account previously introduced turbulent variables. Finally, we propose an innovative method to simulate the phase change based on a germination/growth approach, with a level-set function. The appliance of all these methods to the moving ice filling case allows the validation of the computational robustness of our code and the well-coupling of its various entities.; L'objectif de cette thèse est de développer un code de simulation complet pour le remplissage en fonderie de pièces de grandes dimensions (jusqu'à plusieurs mètres). Ce type de procédé fait entrer en jeu de nombreux phénomènes physiques couplés, nécessitant des méthodes numériques adaptées. La faible viscosité du métal liquide (de l'ordre de 10−6 m2/s) requiert l'emploi d'un modèle de turbulence basé sur un solveur Navier Stokes stabilisé et une méthode de suivi/capture d'interface. Nous avons pour cela choisi un approche stabilisée de type Variational Multi Scales (VMS), qui s'est révélée efficace pour simuler des nombres de Reynolds modérés, alliée à une méthode level-set permettant de déterminer de manière précise et à tout moment la position de l'interface liquide/air. La turbulence est quant à elle prise en compte grâce à un modèle dynamique de type Large Eddy Simulations (L.E.S.), ne faisant pas apparaître de paramètre empirique. Chacune de ces méthodes numériques a été confrontée à des résultats expérimentaux, numériques ou analytiques. Nous avons également conçu notre propre maquette expérimentale de remplissage d'eau, afin de valider le couplage des solveurs pour un cas représentatif. Une autre caractéristique de ces procédés à durée relativement longue (jusqu'à plusieurs dizaines de minutes) est l'importance des transferts thermiques, pouvant mener à la solidification du métal en cours de remplissage. Il convient donc de développer une méthode de résolution stabilisée de la thermique avec convection dominante. Cette méthode doit prendre en compte les variables turbulentes introduites précédemment. Enfin, nous proposons une méthode innovante pour simuler le changement de phase, basée sur une approche germination/ croissance avec fonction level-set. L'application de toutes ces méthodes au cas du remplissage avec glaçon mobile a enfin permis de valider la robustesse numérique de notre code et le bon couplage de ses différentes entités.

Available on: http://congress.cimne.com/CMWR2010/frontal/default.asp; International audience; We propose a stabilized finite element method to address complex high Reynolds flows with free surface. An implicit stabilized finite element method (SFEM) is used for solving the incompressible two-phase Navier-Stokes equations in three-dimensions. A novel approach to deal with turbulent two-phase flows is highlighted by coupling a local convected levelset method to a Large Eddy simulation turbulent modeling. A comparison between the static and dynamic eddy-viscosity models is analyzed. We assess the behaviour and accuracy of the proposed stabilized finite element method coupled to the two-phase turbulent approximation in the simulation of complex 3D flows, such as the flow in a partially filled two communicating tanks. Results are compared with the experimental data and show that the present implementation is able to exhibit good stability and accuracy properties for high Reynolds number flows using unstructured meshes.

Avalaible on: http://web.univ-ubs.fr/limatb/EG2M/Disc_Seminaire/ECCOMAS-CFD2010/papers/01019.pdf; International audience; We propose a stabilized finite element method to address complex high Reynolds flows with free surface. An implicit stabilized finite element method (SFEM) is used for solving the incompressible two-phase Navier-Stokes equations in three-dimensions. A novel approach to deal with turbulent two-phase flows is highlighted by coupling a local convected levelset method to a Large Eddy simulation turbulent modeling. A comparison between the static and dynamic eddy-viscosity models is analyzed. We assess the behaviour and accuracy of the proposed stabilized finite element method coupled to the two-phase turbulent approximation in the simulation of complex 3D flows, such as the flow in a partially filled two communicating tanks. Results are compared with the experimental data and show that the present implementation is able to exhibit good stability and accuracy properties for high Reynolds number flows using unstructured meshes.

International audience; Pilot flames, created by additional injectors of pure fuel, are often used in turbulent burners to enhance flame stabilisation and reduce combustion instabilities. The exact mechanisms through which these additional rich zones modify the flame anchoring location and the combustion dynamics are often difficult to identify, especially when they include unsteady hydrodynamic motion. This study presents Large Eddy Simulations (LES) of the reacting flow within a large-scale gas turbine burner for two different cases of piloting, where either 2 or 6 percent of the total methane used in the burner is injected through additional pilot flame lines. For each case, LES shows how the pilot fuel injection affects both flame stabilisation and flame stability. The 6 percent case leads to a stable flame and limited hydrodynamic perturbations in the initial flame zone. The 2 percent case is less stable, with a small-lift-off of the flame and a Precessing Vortex Core (PVC) in the cold stabilisation zone. This PVC traps some of the lean cold gases issuing from the pilot passage stream, changes the flame stabilisation point and induces instability.

International audience; The Rayleigh criterion (which measures the correlation between pressure and heat release) is the standard tool used to investigate and predict combustion instabilities in both experimental and numerical studies. However, the Rayleigh term is just one of the terms appearing in the acoustic energy equation. The recent development of large eddy simulations for combustion chambers allows complete closure of the budget and analysis of all terms in this equation. This task leads to unexpected difficulties and requires some basic work, as multiple definitions of the energy of fluctuations in a reacting compressible flow can be derived. The objective of this article is to revisit the theoretical derivations of the fluctuation energy equations. Two forms of energy are defined: The first is the classic acoustic energy (AE) introduced by various authors. The second is the fluctuation energy (FE) presented by B.T. Chu [Acta Mech. (1965) 215-234]. Both equations are rederived in a compact manner starting from full nonlinear forms. It is shown that the classic Rayleigh criterion naturally appears as the source term of the AE equation, while the FE form leads to a different criterion stating that temperature and heat release must be in phase for the instability to be fed by the flame/acoustics coupling. The FE form also integrates the fluctuations of three variables (pressure, velocity, entropy), while the AE form uses only pressure and velocity perturbations. It is shown that only the FE form should be used in flames, in contradiction to many current studies performed for combustion instabilities.

Président : Pierre Sagaut ; Rapporteurs : Michel Deville et Eric Lamballais ; Examinateurs : Frédéric Ducros et Franck Nicoud; LES of turbulent flows with heat transfer was used within the framework of conjugate heat transfer problems. The objective of this work lies not only in identifying the various elements likely to impair temperature fluctuations estimations at the fluid/solid interface but also to introduce adequate wall modelings. The choice of a proper convection scheme for the transport of passive scalars led to the adoption of a high order upwind scheme with slope limiter. The use of classical wall models having shown some weaknesses as for the estimation of parietal temperature fluctuations, two new approaches are proposed and tested. The first one relies on a complete resolution of the Navier-Stokes equations on a refined grid close to the wall making it possible to rebuild the temperature fluctuations near the wall. The second one relies on the simultaneous and one dimensional resolution of a turbulent boundary layer equation and a variance transport equation near the wall.; La Simulations des Grandes Echelles d'écoulements turbulents anisothermes a été utilisée dans le cadre de problèmes d'interaction thermique fluide/solide. Le but de ce travail réside non seulement dans l'identification des divers éléments susceptibles de mésestimer les fluctuations de température à la paroi mais aussi à l'introduction de modélisations de parois adaptées. Le choix d'un schéma de convection "optimisant" le transport de scalaire a conduit à l'adoption d'un schéma décentré avec limiteur de pente : le schéma QUICK. Deux nouvelles approches de reconstruction de fluctuations de température en proche paroi sont proposées et testées. La première repose sur une résolution complète des équations de Navier-Stokes sur une grille fine en proche paroi permettant de reconstruire le champ fluctuant. Une seconde méthode repose sur la résolution 1D et simultanée d'une équation de température moyenne et d'une équation de transport pour la variance de température en proche paroi.

Président : Pierre Sagaut ; Rapporteurs : Michel Deville et Eric Lamballais ; Examinateurs : Frédéric Ducros et Franck Nicoud; LES of turbulent flows with heat transfer was used within the framework of conjugate heat transfer problems. The objective of this work lies not only in identifying the various elements likely to impair temperature fluctuations estimations at the fluid/solid interface but also to introduce adequate wall modelings. The choice of a proper convection scheme for the transport of passive scalars led to the adoption of a high order upwind scheme with slope limiter. The use of classical wall models having shown some weaknesses as for the estimation of parietal temperature fluctuations, two new approaches are proposed and tested. The first one relies on a complete resolution of the Navier-Stokes equations on a refined grid close to the wall making it possible to rebuild the temperature fluctuations near the wall. The second one relies on the simultaneous and one dimensional resolution of a turbulent boundary layer equation and a variance transport equation near the wall.; La Simulations des Grandes Echelles d'écoulements turbulents anisothermes a été utilisée dans le cadre de problèmes d'interaction thermique fluide/solide. Le but de ce travail réside non seulement dans l'identification des divers éléments susceptibles de mésestimer les fluctuations de température à la paroi mais aussi à l'introduction de modélisations de parois adaptées. Le choix d'un schéma de convection "optimisant" le transport de scalaire a conduit à l'adoption d'un schéma décentré avec limiteur de pente : le schéma QUICK. Deux nouvelles approches de reconstruction de fluctuations de température en proche paroi sont proposées et testées. La première repose sur une résolution complète des équations de Navier-Stokes sur une grille fine en proche paroi permettant de reconstruire le champ fluctuant. Une seconde méthode repose sur la résolution 1D et simultanée d'une équation de température moyenne et d'une équation de transport pour la variance de température en proche paroi.