L’allumage constitue une phase critique dans de nombreuses applications de combustion et plus particulièrement dans celles qui sont liées à la propulsion aéronautique et spatiale. Un des défis actuels a été de développer des simulations aux grandes échelles de ce phénomène transitoire dans des configurations réalistes comme celles trouvées dans les moteurs aéronautiques. A cet égard, le travail pionnier de Boileau et al. (2008)a indiqué que des calculs complets de ce processus pouvaient être réalisés dans des géométries complètes de chambre de combustion annulaire et que ces calculs pouvaient fournir des informations de première main sur le processus d’allumage circulaire. Il était toutefois important devoir si la simulation pouvait reproduiredes données expérimentales bien contrôlées.Ceci est accompli dans le présent travail qui utilise un dispositif expérimental nouveau désigné sous le nom de MICCA.La thèse décrit l’ensemble des données recueillies au cours d’essais systématiques sur MICCA, la méthode de calcul aux grandes échelles et sa validation dans une configuration de brûleur simplifiée, les résultats de simulations aux grandes échelles du processus d’allumage circulaire, une analyse détaillée des résultats numériques et enfin une modélisation simplifiée du processus d’allumage fondée sur des équations de bilan macroscopiques. / Ignition constitutes a critical phase in many combustion applications and specifically those related to aerospace propulsion. One of the current challenges has been to develop large eddy simulations of this transient phenomenon in realistic configurations like those found in aeroengines.In this respect, the pioneering work of Boileau et a. (2008) indicated that complete calculations of this process in a full annular combustor geometry could be carried out and that they provided first hand information on the light-round process.It was however important to see if the simulation can match well controlled experimental data. This is accomplished in the present work which uses a novel experimental device named MICCA. The thesis describes the experimental set of data,the calculation methodology and its validation in a single burner configuration,results of large eddy simulation of the full light round process, a detailed analysis of the numerical results and an attempt to build a simplified model of the process based on macroscopic balance equations.

Les préoccupations environnementales actuelles ont conduit les constructeurs automobiles à proposer de nouvelles technologies dans le but de réduire les émissions de CO2. Parmi ces technologies, le downsizing appliqué aux moteurs turbocompressés à allumage commandé est une des solutions privilégiées, car permettant d'atteindre des points de fonctionnement fortement chargés, avec un meilleur rendement thermique. Cependant, les fortes charges favorisent l'apparition de cliquetis, un phénomène potentiellement dommageable pour le moteur et qui l'empêche d'exploiter tout son potentiel. Du fait des variabilités cycliques de combustion dans le moteur, le cliquetis, qui dépend des conditions locales dans la chambre de combustion, peut apparaître uniquement sur quelques cycles, à différents endroits et instants. Dans cette thèse, une approche par Simulation aux Grandes Échelles (SGE) a été choisi, dans le but d'étudier et d'améliorer notre compréhension du cliquetis. L'étude se base sur la SGE d'un moteur industriel, le RENAULT 1.2 TCe 115. Un premier ensemble de 30 cycles a été simulé sur un point de fonctionnement de référence, correspondant à un point cliquetant dans la base de données banc d'essais fournie par RENAULT. Les résultats de simulation ont été comparés aux résultats expérimentaux, aussi bien en termes de variabilités cycliques de combustion que de cliquetis. A la suite, un balayage d'avance à l'allumage a été simulé pour étendre la base de données LES à des points plus faiblement et plus fortement cliquetants. La base de données résultante se compose de 150 cycles de combustion, utilisés pour développer des méthodologies et outils, dans le but de mieux caractériser le cliquetis et d'approfondir sa compréhension. L'accès numérique à toute grandeur dans la chambre de combustion, combiné à la description séparée dans cette simulation entre la flamme de pré-mélange initiée par la bougie et l'auto-inflammation dans les gaz frais, ont permis de caractériser le cliquetis en se focalisant sur son origine : l'auto-inflammation. A la suite, les méthodologies et outils développés ont soutenu une analyse détaillée des mécanismes qui contrôlent l'apparition du cliquetis. En particulier, le lien entre le cliquetis et les variabilités cycliques de combustion a été exploré. Les résultats mettent notamment en évidence l'impact des variabilités cycliques, aussi bien de la vitesse de propagation que de la forme de la flamme de pré-mélange, sur le cliquetis. / The rising concerns about the environment have led car manufacturers to come up with new engine technologies, in order to reduce the impact of internal combustion engines on CO2 emissions. In this context, downsizing of turbocharged spark-ignition engines has become a commonly used technology, the advantage of which is to operate the engine under thermally more efficient high loads. However, these high loads favour the appearance of potentially damaging knock phenomena, which prevent the engine to fully exploit its potential. Because of cyclic combustion variability (CCV) in the engine, knock, which depends on the local conditions inside the combustion chamber, can appear at different locations and timings and not in all engine cycles. In this thesis, a Large-Eddy Simulation (LES) approach was selected to investigate and further improve our understanding of the appearance of knock. The study is based on the LES of a production engine, the RENAULT 1.2 TCe 115. For this engine, a set of 30 cycles was initially simulated at a single operating point, corresponding to a knocking point in the test bench database from RENAULT. The results were compared to experimental findings, both in terms of CCV and knock. Subsequently, a spark-timing sweep was simulated in order to enlarge the LES database to also include weaker and stronger knock levels. The resulting LES, which consists of 150 combustion cycles, was used to develop methodologies and tools with the objective to better characterize and understand knock. The computational access to any quantity inside the combustion chamber, together with the separate description with the present LES approach between the spark-triggered premixed flame propagation and auto-ignition, were exploited to characterize knock focusing on its source: autoignition in the fresh gases. Then, the developed methodologies and tools supported a detailed analysis of the mechanisms that control the knock onset. In particular, its link with CCV was explored. The results point out the impact of the cyclic variability in the premixed flame propagation speed and shape on knock.

La conception d'un statoréacteur bénificie aujourd'hui des progrès divers des outils numériques permettant par la même occasion d'alléger les différentes étapes préliminaires de tests en géométrie réelle nécessaires au développement de telle configuration. L'objectif de cette thèse est de développer une méthodologie s'appuyant sur la Simulation aux Grandes Echelles (SGE) afin de contribuer à la validation de ce nouvel outil numérique pour la simulation de statoréacteur et ainsi de contribuer à la compréhension des phénomènes mis en jeu dans ces chambres de combustion. L'outil numérique est tout d'abord adapté pour la simulation des écoulements réactifs fortement turbulents avec un accent mis sur la gestion des chocs avec des schémas centrés et la discrétisation de la convection pour la simulation Eulérienne de la phase dispersée. La configuration cible est le "Statoréacteur de Recherche'' étudié expérimentalement par l'ONERA. Sa simulation est réalisée de manière graduelle. Tout d'abord, il est montré que la simulation de la totalité de la configuration, y compris les diffuseurs d'entrée où se positionne un réseau de choc, est essentielle afin de considérer une géométrie acoustiquement close pour reproduire correctement les modes d'oscillation du statoréacteur. La pertinence du schéma cinétique est aussi étudiée et il est montré l'importance de bien reproduire l'évolution de la vitesse de flamme adiabatique pour une plage de richesse grande, en raison du régime de combustion partiellement prémélangé. Finalement, trois cas à richesse différente sont simulés et un excellent accord est trouvé avec l'expérience. La phénoménologie et les mécanismes pilotant la combustion sont alors étudiés pour ces trois cas. / Design of ramjets benefits today from the progress of numerical tools which relieve the various test stages of real engines that remain necessary for the development of such a kind of configuration. The objective of this dissertation is to develop a methodology based on the Large Eddy Simulation (LES) to contribute to the validation of this new type of advanced numerical tool for the simulation of ramjets and improve the understanding of combustion in these devices. The numerical tool is first adapted for the simulation of highly turbulent reacting flows with emphases on the management of shocks with centered schemes and the discretization of convection for the Eulerian simulation of the dispersed phase. The target configuration is the “Research ramjet” experimentally studied by ONERA. Simulation is carried out gradually. First, it is shown that the simulation of the entire configuration, including diffusers at the inlets where shocks appear is essential to consider an acoustically close geometry to properly reproduce the oscillation modes of combustion. The relevance of the kinetic scheme is also studied. It is shown that reproducing the evolution of the adiabatic flame speed for a wide range of equivalence ratio is critical because of the partially premixed combustion regime involved in this configuration. Finally, three different cases are simulated and excellent agreement is found with experimental data. The phenomenology and the different mechanisms governing combustion are studied for these three cases.

O estudo do transporte de sedimentos, partindo da estimativa de estruturas turbulentas relevantes, constituiu o principal foco deste trabalho. Assim, a equação de transporte de massa (advecção-difusão) foi aplicada em conjunto com as equações de Navier-Stokes e da continuidade filtradas. Neste contexto, houve a necessidade de uma descrição conveniente da turbulência, o que ocorreu mediante a aplicação da simulação de grandes escalas acoplada a modelos de viscosidade turbulenta sub-malha. O método de fronteira imersa foi utilizado na modelação da interface sólido/fluido, representada pela geometria de fundo dos canais. As equações de Navier-Stokes filtradas e da continuidade foram resolvidas numericamente pelo método de passos fracionados, o qual estabeleceu o almejado acoplamento entre ambas. Na discretização das equações governantes foi utilizado o método de diferenças finitas, aplicado sobre malhas deslocadas. Os esquemas explícitos de Adams-Bashforth (de segunda e quarta ordens) foram utilizados no avanço temporal das velocidades do escoamento e das concentrações de sedimentos. Uma nova formulação para a velocidade de sedimentação foi desenvolvida analiticamente, enquanto que eventuais fluxos de ressuspensão foram impostos como condição de contorno no fundo do canal. Todos os códigos computacionais, que estabeleceram as diretrizes e a lógica de cálculo, foram criados no contexto deste trabalho. Os resultados obtidos indicam que a simulação de grandes escalas, associada ao método de fronteira imersa, considerando velocidade de sedimentação conforme aqui modelada, e ainda utilizando a equação de advecção-difusão para o transporte de massa, constituem ferramentas altamente adequadas à estimativa do transporte de sedimentos pela água. / The goal of this work is the research of sediment transport phenomena, deriving from outstanding turbulent eddies estimative. Thus, the mass transport equation (advection-diffusion) was connected with the filtered Navier-Stokes and continuity equations. In this context, the large-eddy simulation and sub-grid viscosity modeling established a convenient description of turbulence effects. The immersed boundary method was applied to model solid/fluid interface, represented here by the shapes of channel bottom. The filtered Navier-Stokes and continuity equations were solved by the fractional step method. The equations were discretized with the finite difference method, applied over staggered grids, whereas explicit Adams-Bashforth schemes (second and forth orders) were used in temporal advancement of velocities and sediment concentration fields. A new analytical formulation for settling velocity was obtained, while fortuitous re-suspension flux was applied like a boundary condition in the channel bottom. The computational code was totally developed in this work. The results of present simulations show that large-eddy simulation coupled to the immersed boundary method, considering, yet, the settling velocity of particles and the advection-diffusion equation for mass transport, constitute potential tools for sediment transport evaluation in water flows.

Large-eddy simulation (LES) of wind and wave forced oceanic turbulent boundary layers is performed using the residual-based variational multiscale method (RBVMS) and near-wall modeling. Wind and surface gravity wave forcing generates Langmuir turbulence characterized by Langmuir circulation (LC) with largest scales consisting of streamwise vortices aligned in the direction of the wind, acting as a secondary flow structure to the primary wind-driven component of the flow. The LES here is representative of a shallow water continental shelf flow (10 to 30 meters in depth) far from lateral boundaries in which LC engulfs the full depth of the water column and disrupts the bottom log layer. Field measurements indicate that occurrence of full-depth LC is typical during the passage of storms. The RBVMS method with quadratic NURBS (Non-Uniform Rational B-splines) with near-wall resolution is shown to possess good convergence characteristics for this flow. The use of near-wall modeling facilitates simulations with expanded domains over horizontal directions. Thus, these simulations are able to resolve multiple Langmuir cells permitting analysis of the interaction between the cells. Results in terms of velocity statistics are presented from simulations performed with various domain sizes and distinct near-wall treatments: (1) the classical treatment based on prescription of the wall shear stress assuming a law of the wall and (2) a recent treatment based on weak imposition of the no-slip bottom boundary condition.

In this thesis study, Detached Eddy Simulation turbulence model is studied in two dimension mainly for flow over single element airfoils in high Reynolds numbers to gain experience with model before applying it to a three dimensional simulations. For this aim, Spalart-Allmaras and standard DES ,DES97, turbulence models are implemented to parallel, viscous, hybrid grid flow solver. The flow solver ,Set2d, is written in FORTRAN language. The Navier-Stokes equations are discretized by first order accurately cell centered finite volume method and solved explicitly by using Runge-Kutta dual time integration technique. Inviscid fluxes are computed using Roe flux difference splitting method. The numerical simulations are performed in parallel environment using domain decomposition and PVM library routines for inter-process communications. To take into account the effect of unsteadyness after the convergence is ensured by local time stepping technique for four order magnitude drop in density residual, global time stepping is applied for 20000 iterations. The solution algorithm is validated aganist the numerical and experimental studies for single element airfoils in subsonic and transonic flows. It is seen that Spalart-Allmaras and DES97 turbulence models give the same results in the non-seperated flows. Grey area is investigated by changing $C_{DES}$ coefficient. Modeled Stress Depletion which cause reduction of eddy viscosity is observed.

This thesis deals with Large Eddy Simulation (LES) of impinging air jets. The impinging jet configuration features heated circular jets impinging onto a flat plate. The problem addressed here is of generic nature, with applications in many engineering devices, such as cooling of components in gas turbines, in cars and electronic devices. The flow is inherently unsteady and contains relatively slowly varying coherent structures. Therefore, LES is the method of choice when the Reynolds number is large enough to exclude Direct Numerical Simulations (DNS). The present LES model is a basic model without explicit Sub-Grid-Scale (SGS) modeling and without explicit filtering. Instead, the numerical scheme is used to account for the necessary amount of dissipation. By using the computational grid as a filter the cutoff wavenumber depends explicitly on the grid spacing. The underlying computational grid is staggered and constructed in a Cartesian coordinate system. Heat transfer is modeled by the transport equation for a passive scalar. This is possible due to the negligible influence of buoyancy which implies constant density throughout the flow field. The present method provides accurate results for simple geometries in an efficient manner. A great variety of inlet conditions have been considered in order to elucidate how the dynamics of the flow and heat transfer are affected. The considered studies include top-hat and mollified mean velocity profiles subjected to random and sinusoidal perturbations and top-hat profiles superimposed with solid body rotation. It has been found that the shape of the mean inlet velocity profile has a decisive influence on the development of the flow and scalar fields, whereas the characteristics of the imposed artificial disturbances (under consideration) have somewhat weaker effect. In order to obtain results unequivocally comparable to experimental data on turbulent impinging jets both space and time correlations of the inflow data must be considered, so also the spectral content. This is particularly important if the region of interest is close to the velocity inlet, i.e. for small nozzle-to-plate spacings. Within this work mainly small nozzle-toplate spacings are considered (within the range of 0.25 and 4 nozzle diameters), which emphasizes the importance of the inflow conditions. Thus, additional to the basic methods also turbulent inflow conditions, acquired from a precursor pipe simulation, have been examined. Both for swirling and non-swirling flows. This method emulates fully developed turbulent pipe flow conditions and is the best in the sense of being well defined, but it demands a great deal of computing power and is also rather inflexibility. In case of the basic randomly perturbed methods the top-hat approach has been found to produce results in closest agreement with those originating from turbulent inlet conditions. In the present simulations the growth of individual instability modes is clearly detected. The character of the instability is strongly influenced by the imposed boundary conditions. Due to the lack of correlation random superimposed fluctuations have only a weak influence on the developing flow field. The shape of the mean profile, on the other hand, influences both the growth rate and the frequency of the dominant modes. The top-hat profile yields a higher natural frequency than the mollified. Furthermore, for the top-hat profile coalescence of pairs of vortices takes place within the shear-layer of the axial jet, whereas for the mollified profile (for the considered degree of mollification) it takes place within the wall jet. This indicates that the transition process is delayed for smoother profiles. The amount of wall heat transfer is directly influenced by the character of the convective vortical structures. For the mollified cases wall heat transfer originates predominantly from the dynamics of discrete coherent structures. The influence from eddy structures is low and hence Reynolds analogy is applicable, at least in regions of attached flow. The top-hat and the turbulent inflow conditions yield a higher rate of incoherent small scale structures. This strongly affects the character of wall heat transfer. Also the applied level of swirl at the velocity inlet has significant influence on the rate of heat transfer. The turbulence level increases with swirl, which is positive for heat transfer, and so also the spreading of the jet. The latter effect has a negative influence on wall heat transfer, particularly in the center most regions. This however depends also on the details of the inflow data. / QC 20100831

<p>Large-scale coherent structures are systematically investigated in terms of their geometric attributes, importance toward describing turbulent exchange of energy, momentum and mass as well as their relationship to landscape features in the context of land-atmosphere interaction. In the first chapter, we present the motivation of this work as well as a background review of large-scale coherent structures in land-atmosphere interaction. In the second chapter, the methodology of large-eddy simulation (LES) and the proper orthogonal decomposition (POD) is introduced. LES was used to serve as a virtual laboratory to simulate typical scenarios in land-atmosphere interaction and the POD was used as the major technique to educe the coherent structures from turbulent flows in land-atmosphere interaction. In the third chapter, we justify the use of the LES to simulate the realistic coherent structures in the atmospheric boundary layer (ABL) by comparing results obtained from LES of the ABL and direct numerical simulation (DNS) of channel flow. In the fourth chapter, we investigate the effects of a wide range of vegetation density on the coherent structures within the air space within and just above the canopy (the so-called canopy sublayer, CSL). The fifth chapter presents an analysis of the coherent structures across a periodic forest-clearing-forest transition in the steamwise direction. The sixth chapter focuses on the role of coherent structures in explaining scalar dissimilarity in the CSL. The seventh chapter summarizes this dissertation and provides suggestions for future study.</p> / Dissertation

Two new models to calculate the species instantaneous and filtered reaction rates for multi-step, multi-species chemical kinetics mechanisms are developed based on the artificial neural networks (ANN) approach. The proposed methodologies depend on training the ANNs off-line on a thermo-chemical database representative of the actual composition and turbulence level of interest. The thermo-chemical database is constructed by stand-alone linear eddy mixing (LEM) model simulations under both premixed and non-premixed conditions, where the unsteady interaction of turbulence with chemical kinetics is included as a part of the training database. In this approach, the information regarding the actual geometry of interest is not needed within the LEM computations. The developed models are validated extensively on the large eddy simulations (LES) of (i) premixed laminar-flame-vortex-turbulence interaction, (ii) temporally mixing non-premixed flame with extinction-reignition characteristics, and (iii) stagnation point reverse flow combustor, which utilizes exhaust gas re-circulation technique. Results in general are satisfactory, and it is shown that the ANN provides considerable amount of memory saving and speed-up with reasonable and reliable accuracy. The speed-up is strongly affected by the stiffness of the reduced mechanism used for the computations, whereas the memory saving is considerable regardless.

A cause de sa température élevée, le jet propulsif d'un avion de combat est une source de rayonnement infrarouge qui rend l'appareil très vulnérable. Ce rayonnement peut toutefois être réduit en accélérant le mélange du jet avec l'atmosphère. <br />Cette thèse est consacrée à la simulation numérique d'un jet propulsif réaliste contrôlé par des injections radiales et à l'analyse physique des mécanismes d'augmentation de son mélange. <br />Deux types de simulations, RANS et ZDES, ont été réalisés sur la base du modèle de Spalart-Allmaras. Dans le modèle ZDES, une nouvelle longueur caractéristique de maille est formulée et améliore sensiblement la prévision de la région initiale du jet. Globalement, les simulations ZDES restituent fidèlement le champ moyen du jet supersonique sans et avec contrôle, aussi bien les cellules de détente/compression que la diffusion turbulente. <br />L'analyse physique est dédiée à la compréhension d'une part des mécanismes compressibles concentrées au coeur du jet et d'autre part des mécanismes tourbillonnaires périphériques ainsi qu'à l'évaluation de leurs rôles respectifs dans l'augmentation du mélange. Il en ressort que l'augmentation du mélange est exclusivement due aux mécanismes tourbillonnaires. Une étude paramétrique fournit des indications pour concevoir un mélangeur efficace. L'analyse des tourbillons focalisée sur le régime lointain quasi-bidimensionnel souligne leur dynamique moyenne et fait apparaître l'action des fluctuations turbulentes sur leur taux de dégénérescence. Enfin, deux régimes de contrôle sont identifiés et associés aux pénétrations respectivement quasi-stationnaire et intermittente des jets secondaires.