La simulation numérique est l’un des outils les plus puissants pour concevoir etoptimiser les systèmes industriels. Dans le domaine de la Dynamique des FluidesNumériques (CFD, "Computational Fluid Dynamics"), la simulation auxgrandes échelles (LES, "Large Eddy Simulation") est aujourd’hui largementutilisée pour calculer les écoulements turbulents réactifs, où les tourbillons degrande taille sont calculés explicitement, tandis que l’effet de ceux de petitetaille est modelisé. Des modèles de sous-mailles sont requis pour fermer leséquations de transport en LES, et dans le contexte de la simulation de la combustionturbulente, le plissement de la surface de flamme de sous-maille doitêtre modélisé.En général, augmenter le plissement de la surface de flamme de sous-maille favorisela combustion. L’amplitude de la promotion est donnée par une fonctiond’efficacité, qui est dérivée d’une hypothèse d’équilibre entre la production etla destruction de la surface de flamme. Dans les méthodes conventionnelles,le calcul de la fonction d’efficacité nécessite une constante qui dépend de lagéométrie de la chambre de combustion, de l’intensité de turbulence, de larichesse du mélange de air-carburant etc, et cette constante doit être fixée audébut de la simulation. Autrement dit, elle doit être déterminé empiriquement.Cette thèse développe un modèle de sous-maille pour la LES en combustionturbulente, qui est appelé le modèle dynamique de flammelette épaissie (DTF,"dynamic thickened flamelet model"), qui détermine la valeur de la constanteen fonction des conditions de l’écoulement sans utiliser des données empiriques.Ce modèle est tout d’abord testé sur une flamme laminaire unidimensionnellepour vérifier la convergence de la fonction d’efficacité vers l’unité (aucun plissementde la surface de flamme de sous-maille). Puis il est appliqué en combinaisonavec le modèle dynamique de Smagorinsky (Dynamic Smagorinskymodel) aux simulations multidimensionnelles d’une flamme en V, stabilisée enaval d’un dièdre. Les résultats de la simulation en trois dimensions sont alorscomparés avec les données expérimentales obtenues sur une expérience de mêmegéométrie. La comparaison montre la faisabilité de la formulation dynamique. / Numerical simulation is one of the most powerful tools to design and optimizeindustrial facilities. In the field of Computational Fluid Dynamics (CFD),Large Eddy Simulation (LES) is widely used to compute turbulent reactingflows, where larger turbulent motions are explicitly computed, while only theeffect of smaller ones is modeled. Subgrid models are required to close thetransport equations in LES, and in the context of the simulation of turbulentcombustion, the subgrid-scale wrinkling of the flame front must be modeled.In general, subgrid-scale flame wrinkling promotes the chemical reaction. Themagnitude of the promotion is given through an efficiency function derivedfrom an equilibrium assumption between production and destruction of flamesurface. In conventional methods, the calculation of the efficiency functionrequires a constant which depends on the geometry of the combustion chamber,turbulence intensity, the equivalence ratio of the fuel-air mixture, and so on;this constant must be prescribed at the beginning of the simulation. In otherwords, empirical knowledge is required.This thesis develops a subgrid-scale model for LES of turbulent combustion,called the dynamic thickened flamelet (DTF) model, which determines the valueof the constant from the flow conditions without any empirical input.The model is first tested in a one-dimensional laminar flame to verify the convergenceof the efficiency function to unity (no subgrid-scale flame front wrinkling).Then it is applied to multi-dimensional simulations of V-shape flamestabilized downstream of a triangular flame holder in combination with the dynamicSmagorinsky model. The results of the three-dimensional simulation arethen compared with the experimental data obtained through the experimentof the same geometry. The comparison proves the feasibility of the dynamicformulation.

In this study, numerical simulations of a low emission lab-scale non-premixed combustor are conducted and analyzed. The objectives are to provide new insight into the physical phenomena in the SPRF (Stagnation Point Reverse Flow) combustor built in the Georgia Tech Combustion Lab, and to compare three Large Eddy Simulation (LES) combustion models (Eddy Break-Up [EBU], Steady Flamelet [SF] and Linear Eddy Model [LEM]) for non-premixed combustion. The nominal operating condition of the SPRF combustor achieves very low NOx and CO emissions by combining turbulent mixing of exhaust gases with preheated reactants and chemical kinetics. The SPRF numerical simulation focuses on capturing the complex interaction between turbulent mixing and heat release. LES simulations have been carried out for a non-reactive case in order to analyze the turbulent mixing inside the combustor. The LES results have been compared to PIV experimental data and the code has been validated. The dominating features of the operational mode of the SPRF combustor (dilution of hot products into reactants, pre-heating and pre-mixing) have been analyzed, and results from the EBU-LES, SF-LES and LEM-LES simulations have been compared. Analysis shows that the LEM-LES simulation achieves the best agreement with the observed flame structure and is the only model that captures the stabilization processes observed in the experiments. EBU-LES and SF-LES do not predict the correct flow pattern because of the inaccurate modeling of sub-grid scale mixing and turbulence-combustion interaction. Limitations of these two models for this type of combustor are discussed.

Le contexte de cette étude est l’optimisation de la combustion turbulente prémélangée de syngaz pour la production propre d’énergie. Un brûleur CH4/air de type bec Bunsen avec forçage turbulent multi-échelles produit par un système de trois grilles, est simulé numériquement par différentes techniques de simulation des grandes échelles (SGE), et les résultats sont comparés à l’expérience. On a développé et appliqué une formulation bas-Mach du solveur Navier-Stokes basé sur différents schémas numériques, allant des différences finies centrées d’ordre 4 à des versions avancées des schémas WENO d’ordre 5. La méthodologie est évaluée sur une série de cas-tests classiques (flamme laminaire 1D prémélangée, turbulence homogène et isotrope en auto-amortissement), et sur des simulations 2D de la flamme turbulente prémélangée expérimentale. Les SGE implicites (ILES), i.e. sans aucune modélisation sous-maille, et explicites avec le modèle de flamme épaissie et un modèle de plissement sous-maille nouvellement élaboré (TFLES), sont appliquées à la simulation 3D du brûleur expérimental. Les résultats montrent que l’approche TFLES avec un schéma d’ordre élevé à faible dissipation numérique prédit correctement la longueur de la flamme et la densité de surface de flamme. La SGE implicite avec un schéma WENO avancé produit une flamme trop courte mais réaliste à condition que la taille de la maille soit de l’ordre de l’épaisseur de flamme laminaire. La représentation des interactions flamme/turbulence est néanmoins très différente entre TFLES et ILES. / The context of this study is the optimization of premixed turbulent combustion of syngas for clean energy production. A Bunsen-type CH4/air turbulent premixed burner with a multi-scale grid generator is simulated with different Large Eddy Simulation (LES) strategies and compared to experimental results. A low-Mach formulation of a compressible Navier-Stokes solver based on different numerical methods, ranging from 4th order central finite difference to 5th order advanced WENO schemes, is developed and applied. Classical test cases (1D laminar premixed flame, decaying HIT), and 2D simulations of the turbulent premixed flame are performed to assess the numerical methodology. Implicit LES (ILES), i.e. LES without any explicit subgrid modeling, and explicit LES with the Thickened Flame model and subgrid scale flame wrinkling modelling (TFLES) are applied to simulate numerically the 3D experimental burner. Results show that TFLES with a high-order low dissipation scheme predicts quite well the experimental flame length and flame surface density. ILES with advanced WENO schemes produces a slightly shorter although realistic flame provided the grid spacing is of order of the laminar flame thickness. The representation of flame/turbulence interactions in TFLES and ILES are however quite different.

Fossil fuel reserves are projected to be decreasing, and emission regulations are becoming more stringent due to increasing atmospheric pollution. Alternative fuels for power generation in industrial gas turbines are thus required able to meet the above demands. Examples of such fuels are synthetic gas, blast furnace gas and coke oven gas. A common characteristic of these fuels is that they are multi-component fuels, whose composition varies greatly depending on their production process. This implies that their combustion characteristics will also vary significantly. Thus, accurate and yet flexible enough combustion sub-models are required for such fuels, which are used during the design stage, to ensure optimum performance during practical operating conditions. Most combustion sub-model development and validation is based on Direct Numerical Simulation (DNS) studies. DNS however is computationally expensive. This, has so far limited DNS to single-component fuels such as methane and hydrogen. Furthermore, the majority of DNS conducted to date used one-step chemistry in 3D, and skeletal chemistry in 2D only. The need for 3D DNS using skeletal chemistry is thus apparent. In this study, an accurate reduced chemical mechanism suitable for multi-component fuel-air combustion is developed from a skeletal mechanism. Three-dimensional DNS of a freely propagating turbulent premixed flame is then conducted using both mechanisms to shed some light into the flame structure and turbulence-scalar interaction of such multi-component fuel flames. It is found that for the multi-component fuel flame heat is released over a wider temperature range contrary to a methane flame. This, results from the presence of individual species reactions zones which do not all overlap. The performance of the reduced mechanism is also validated using the DNS data. Results suggest it to be a good substitute of the skeletal mechanism, resulting in significant time and memory savings. The flame markers commonly used to visualize heat release rate in laser diagnostics are found to be inadequate for the multi-component fuel flame, and alternative markers are proposed. Finally, some popular mean reaction rate closures are tested for the multi-component fuel flame. Significant differences are observed between the models’ performance at the highest turbulence level considered in this study. These arise from the chemical complexity of the fuel, and further parametric studies using skeletal chemistry DNS would be useful for the refinement of the models.

The present thesis deals with the fundamental aspects of turbulent mixingand non-premixed combustion in wall-jet flows. Direct numerical simulations(DNS) of compressible turbulent flows are performed in a wall-jet configura-tion, which has a close resemblance to many industrial combustion applica-tions. The triple ”turbulence-chemistry-wall” interactions are also present inthis flow set-up. These interactions have been addressed by first focusing onturbulent flow effects on the isothermal reaction, including the near-wall issues.Then, by adding heat-release to the simulations, it has been concentrated onheat-release effects on various phenomena that occur in the reacting turbulentwall-jet flow. In the computational domain, fuel and oxidizer enter separatelyin a non-premixed manner and the flow is fully turbulent and subsonic in allsimulations. In the first phase of this study, the case of a turbulent wall-jetincluding an isothermal reaction without heat release is addressed in order toisolate the near-wall effects and the mixing characteristics of the flow and thekey statistics for combustion are studied in the absence of thermal effects. Adeeper insight into three-dimensional mixing and reaction characteristics in aturbulent wall-jet has been gained through investigation of the probability den-sity functions, higher order moments of velocities and reacting scalars and thescalar dissipation rates of different species. In the second phase, DNS of turbu-lent reacting wall-jets including heat release is performed, where a single-stepglobal exothermic reaction with an Arrhenius-type reaction rate is considered.The main target was to identify the heat-release effects on different mixingscales of turbulent wall-jet flow. The scalar dissipation rates, time scale ratios,two-point correlations, one and two-dimensional premultiplied spectra are usedto illustrate the heat release induced modifications. It is observed that heatrelease effects delay the transition process in the chemically reacting cases andenlarge the fluctuation intensities of density and pressure, but have a dampingeffect on all velocity fluctuation intensities. Finer small mixing scales were ob-served in the isothermal simulations and larger vortical structures formed afteradding significant amounts of heat-release. Simulations with different Damk ̈h-  oler numbers, but comparable temperature-rise are performed and the expectedbehavior, a thinner flame with increasing Damk ̈hler number, is observed. Finally, some heat transfer related quantities are examined. The wall heat fluxand the corresponding Nusselt numbers are addressed. The near-wall reactioneffects on the skin friction coefficient are studied and further the reaction char-acteristics are investigated throughout the domain. / QC 20110908

[pt] Chamas pré-misturadas em escoamentos turbulentos com rotação são encontradas em diversos sistemas de engenharia, como turbinas a gás e motores a jato. Neste trabalho, regimes de chamas característicos de tais sistemas são estudados numericamente num queimador de escala laboratorial. O estado da arte dos estudos numéricos de tais chamas é revisado, com respeito a simulações de grandes escalas, bem como o de modelos computacionais baseados em médias de Reynolds. Um estudo isotérmico é feito no escoamento turbulento, num domínio computacional que consiste de um swirler radial e uma câmara de combustão. O impacto de diferentes modelos de turbulência, níveis de refinamentos de malha e condições de contorno no número de swirl e na estrutura do escoamento é investigado. Os resultados revelam que os três modelos de turbulência propostos resultam em campos de escoamento e número de swirl similares, enquanto o nível de refinamento de malha e a condição de contorno de parede deslizante alteram o número de swirl significativamente. Utilizando as equações de média de Reynolds, com o fechamento do modelo k − E realizável, acoplado a um modelo de duas equações para chamas pré-misturadas de metano e ar, dois regimes de chamas são analisados. Estes regimes correspondem à chama de recirculação externa (chama tipo M) e um regime de instabilidade, que ocorre na transição entre a chama tipo V e chama tornado. A estrutura do escoamento é caracterizada em termos de velocidade e propriedades de turbulência e combustão. Uma comparação entre variáveis de progresso também é feita, utilizando resultados experimentais prévios, levando a boa concordância qualitativa para os dois regimes estudados. / [en] Lean premixed turbulent swirling flames are found in many engineering systems, such as gas turbines and jet engines. This work aims to numerically study flame regimes, representative of such systems, stabilized in a laboratory scale burner. The state of the art of the numerical studies concerning these types of flames is reviewed, with respect to Reynolds-Averaged NavierStokes and Large Eddy Simulations. A turbulent, isothermal flow study is performed within the radial swirler and the combustion chamber. The impact of different turbulence models (realizable k − E, RNG k − E and SST k−W), mesh refinement levels and boundary conditions on the swirl number and overall flow structure is investigated. The results show that the three tested turbulence models yield similar results, with respect to the obtained flow field, whereas the mesh refinement level and slip wall boundary condition alter the swirl number significantly. Using Reynolds-Averaged NavierStokes transport equations, closed by the realizable k − E model, coupled with a two-equation premixed combustion model for methane/air mixtures, two combustion regimes are analyzed. These regimes correspond to the outer recirculation zone flame (M-shaped flame) and an unstable regime, which occurs at the transition between the V-shaped flame and tornado-flame. The flow structure is characterized in terms of velocity fields, turbulence and combustion properties. A reaction progress variable comparison is also performed, using existing experimental results, yielding qualitatively similar results for both studied regimes.

Afin de pouvoir simuler la formation des principaux polluants au sein de foyers aéronautiques réalistes, un modèle de réduction de la chimie détaillée (FPI), basé sur la construction de tables à partir de calculs de flammes de prémélange laminaires élémentaires, est adapté et couplé au code d’aérothermochimie CEDRE de l’ONERA. Après une brève validation de ce modèle via la simulation de flammes laminaires canoniques, les interactions chimie/turbulence sont modélisées sous l’hypothèse des flammelettes, en approchant les PDF des paramètres d’entrée des tables par des fonctions beta. Cette approche complète est appliquée à la simulation numérique de l’écoulement au sein d’une configuration plus appliquée : la chambre PRECCINSTA. Ce cas bien connu a permis notamment l’évaluation des capacités du modèle dans un contexte plus industriel par comparaison des résultats de calcul aux données expérimentales disponibles. Il a en particulier permis de tester l’approche FPI étendue à la modélisation de la combustion partiellement prémélangée. Par ailleurs, l’utilisation d’un modèle de chimie réduite s’avère particulièrement appropriée pour prédire l’émission de substances polluantes, par exemple CO. Cependant, lorsque l’on considère la formation de NO, FPI ne peut pas être utilisé directement du fait de la lente dynamique chimique de cette espèce.Pour pallier à cette limitation, deux approches permettant de modéliser la production de NO au sein d’écoulements complexes sont proposées, fondées sur l’utilisation des tables chimiques FPI. Les capacités de ces modèles sont finalement analysées à l’aide de calculs effectués sur la configuration PRECCINSTA. / In order to simulate major pollutant formation inside realistic aeronautical combustion chambers, a detailed chemistry reduction technique (FPI), based on the construction of databases from elementary laminar premixed flame calculations, is adapted and coupled to the ONERA household CFD code : CEDRE. After a short validation of this model based on the numerical simulation of simplified laminar flames, the chemistry turbulence interactions are modeled under the laminar flamelet hypothesis, by assuming the shape of the FPI progress variable PDFs using beta functions. This comprehensive approach is then applied to the numerical simulation of the flow inside a realistic geometry :the PRECCINSTA combustion chamber. This well-known configuration has enabled the evaluation of the model’s abilities within an industrial framework using numerical/experimental results comparisons. It has especially allowed to test an extension of the model to partially premixed combution. Furthermore, the use of a tabulated chemistry model turns out to be particularly appropriate to predict pollutant species formation such as CO. However, when considering the formation of nitrogen oxides,FPI cannot be applied directly because of the slow dynamics of the chemical processes involved. Toovercome these limitations, two approaches allowing NO production modeling within complexe flowsare proposed, derived from the use of the tabulated data. The capacities of these models are finally analysed using computations performed on the PRECCINSTA chamber.

<div>Studies of turbulent combustion in the past mainly focus on problems with single-regime combustion. In practical combustion systems, however, combustion rarely occurs in a single regime, and different regimes of combustion can be observed in the same system. This creates a significant gap between our existing knowledge of combustion in single regime and the practical need in multi-regime combustion. In this work, we aim to extend the traditional single-regime combustion models to problems involving different regimes of combustion. Among the existing modeling methods, Transported Probability Density Function (PDF) method is attractive for its intrinsic closure of treating detailed chemical kinetics and has been demonstrated to be promising in predicting low-probability but practically important combustion events like local extinction and re-ignition. In this work, we focus on the model assessment and advancement of the Large Eddy Simulation (LES)/ PDF method in predicting turbulent multi-regime combustion.</div><div><br></div><div><div>Two combustion benchmark problems are considered for the model assessment. One is a recently designed turbulent piloted jet flame that features statistically transient processes, the Sydney turbulent pulsed piloted jet flame. A direct comparison of the predicted and measured time series of the axial velocity demonstrates a satisfactory prediction of the flow and turbulence fields of the pulsed jet flame by the employed LES/PDF modeling method. A comparison of the PLIF-OH images and the predicted OH mass fraction contours at a few selected times shows that the method captures the different combustion stages including healthy burning, significant extinction, and the re-establishment of healthy burning, in the statistically transient process. The temporal history of the conditional PDF of OH mass fraction/temperature at around stoichiometric conditions at different axial locations suggests that the method predicts the extinction and re-establishment timings accurately at upstream locations but less accurately at downstream locations with a delay of burning reestablishment. The other test case is a unified series of existing turbulent piloted flames. To facilitate model assessment across different combustion regimes, we develop a model validation framework by unifying several existing pilot stabilized turbulent jet flames in different combustion regimes. The characteristic similarity and difference of the employed piloted flames are examined, including the Sydney piloted flames L, B, and M, the Sandia piloted flames D, E, and F, a series of piloted premixed Bunsen flames, and the Sydney/Sandia inhomogeneous inlet piloted jet flames. Proper parameterization and a regime diagram are introduced to characterize the pilot stabilized flames covering non-premixed, partially premixed, and premixed flames. A preliminary model assessment is carried out to examine the simultaneous model performance of the LES/PDF method for the piloted jet flames across different combustion regimes.</div><div><br></div><div>With the assessment work in the above two test cases, it is found that the LES/PDF method can predict the statistically transient combustion and multi-regime combustion reasonably well but some modeling limitations are also identified. Thus, further model advancement is needed for the LES/PDF method. In this work, we focus on two model advancement studies related to the molecular diffusion and sub-filter scale mixing processes in turbulent combustion. The first study is to deal with differential molecular diffusion (DMD) among different species. The importance of theDMD effects on combustion has been found in many applications. However, in most previous combustion models equal molecular diffusivity is assumed. To incorporate the DMD effects accurately, we develop a model called Variance Consistent Mean Shift (VCMS) model. The second model advancement focuses on the sub-filter scale mixing in high-Karlovitz (Ka) number turbulent combustion. We analyze the DNS data of a Sandia high-Ka premixed jet flame to gain insights into the modeling of sub-filter scale mixing. A sub-filter scale mixing time scale is analyzed with respect to the filter size to examine the validity of a power-law scaling model for the mixing time scale.</div></div>

Le moteur essence à allumage par compression (GCI), reposant sur la combustion partiellement prémélangée de l'essence (GPPC), peut potentiellement assurer des opérations efficaces et propres. Le moteur GCI s'est avéré efficace à forte charge, mais l'indice d'octane élevé de l'essence limite considérablement les opérations à faible charge. Le présent travail étudie le potentiel de l'utilisation de l'ozone, fort agent oxydant, pour améliorer la réactivité de l'essence et permettre le fonctionnement à faible charge de GCI. L'ozone peut être produit on board en équipant le moteur d'un générateur d'ozone, sans impact dramatique sur le coût du moteur et sur la complexité du contrôle du moteur. Les essais effectués avec un moteur monocylindre ont montré que l'ozone favorise la combustion HCCI de l'essence, permettant d'étendre la limite d’auto-inflammation et de réduire la température minimale nécessaire de celle-ci. Les diagnostics optiques ont montré que ces propriétés sont liées à une prolifération radicale accrue, amenées par des réactions à basse température induites par l'ozone. En parallèle, le processus de combustion GCI a été étudié dans des conditions de faible charge. Sans ozone, la température d'admission doit être considérablement augmentée pour permettre l'auto-inflammationdes mélanges essence-air pauvres. De plus, les résultats indiquent que le monoxyde d’azote (NO) contenu dans les gaz brûlés résiduels peut, dans certaines conditions, favoriser fortement la combustion GCI. Ensuite,l'effet de l'ozone a été étudié dans des conditions d'injection directe GCI. Les résultats démontrent qu’une stratégie avec double injection est nécessaire pour maximiser l’effet promoteur de l’ozone et pour contrôler le processus de combustion GCI. Enfin, l'utilisation d’une forte concentration d’ozone a permis d’atteindre des opérations à faible charge en mode GCI, avec des faibles émissions de NOx et de suie, et cela, sans avoir besoin d'augmenter la température ou la pression d'admission. / Gasoline Compression Ignition (GCI) engine, relying on Gasoline Partially Premixed Combustion (GPPC) has potential for efficient and clean operations. GCI engine showed to be effective at high load, however, the highoctane number of gasoline dramatically limits low load operations. The present work investigates the potential of using ozone, a strong oxidizing agent, to improve gasoline reactivity and enabling low load GCI operation.Ozone can be produced in-situ and on-demand by equipping the engine with an ozone generator, without a dramatic impact on the engine cost and the engine control complexity. Experiments in a single cylinder engine showed that ozone promotes gasoline HCCI combustion, making possible to extend the lean limit and reducing the minimum temperature needed for autoignition. Optical diagnostics showed that these properties are related to an increased radical proliferation related to ozone-induced low temperature reactions. In parallel, GCI combustion process was investigated under low load conditions. Without ozone, the intake temperature should be considerable increased to enable auto ignition of lean gasoline-air mixtures. Moreover, results indicated that the NO contained into residual burnt gases can strongly promote GCI low load combustion. Finally, the effect of ozone was investigated under GCI direct-injection conditions, demonstrating that low load GCI operation with low NOx and Soot emission can be achieved by seeding the intake of the engine with ozone without needing of increasing the intake charge temperature or boosting the intake pressure.

La propagation de flammes turbulentes dans des milieux réactifs inhomogènes concerne un grand nombre d’applications pratiques, y compris celles qui reposent sur des cycles de combustion à volume constant. Les hétérogénéités de composition (richesse, température,dilution par des gaz brûlés, etc.) sont issues de plusieurs facteurs distincts tels que la dispersion du spray de gouttelettes de combustible et son évaporation, la topologie de l’écoulement ainsi que la présence éventuelle de gaz brûlés résiduels issus du cycle précédent. La structure des flammes partiellement prémélangées qui en résultent est significativement plus complexe que celles des flammes plus classiques de diffusion ou de prémélange. L’objectif de ce travail de thèse est donc de contribuer à l’amélioration de leur connaissance, en s’appuyant sur la génération et l’analyse de base de données de simulations numériques directes ou DNS (Direct Numerical Simulation). Celles-ci sont conduites avec le code de calcul Asphodele qui est basé sur l’approximation de faible nombre de Mach. Le combustible de référence retenu est l’iso-octane.La base de données est structurée suivant cinq paramètres qui permettent de caractériser l’écoulement turbulent ainsi que l’hétérogénéité de composition du milieu réactif. Dans un premier temps, des configurations bidimensionnelles ont été considérées en raison du coût élevé induit par la description détaillée de la cinétique chimique. L’étude des ces différents cas de calcul a permis de mettre en lumière plusieurs mécanismes fondamentaux de propagation dans les milieux hétérogènes en composition. Une réduction significative des coûts de calcula pu ensuite être obtenue grâce au développement d’un modèle chimique simplifié optimisé.Son utilisation a permis d’étendre les analyses à de / The propagation of turbulent flames in non-homogeneous reactive mixtures of reactants concerns a large number of practical applications, including those based on constant volume combustion cycles. The composition heterogeneities (equivalence ratio, temperature, dilution by burnt gases, etc.) result from several distinct factors such as the dispersion of the spray of fuel droplets and its evaporation, the flow field topology as well as the possible presence of residual burnt gases issued from the previous cycle. The resulting partially premixed flames structure is significantly more complex than the one of more conventional diffusion or premixed flames.The aim of this thesis work is therefore to contribute to the improvement of their understanding, by proceeding to the generation and analysis of a new set of direct numerical simulations (DNS) databases. The present computations are performed with the low-Mach number DNS solver Asphodele. The database is structured according to five parameters that characterize the turbulent flow as well as the composition heterogeneity of the reactive mixture. First, because of the high numerical costs induced by the detailed description of chemical kinetics, two-dimensional configurations were considered. The study of these various simulations highlights several fundamental mechanisms of flame propagation in heterogeneous mixtures. Then, a significant computational cost saving has been achieved through the development of an optimized simplified chemistry model. The use of the latter allowed to overcome the major bottleneck of high CPU costs related to chemical kinetics description and thus to extend the analysis to three-dimensional configurations. Some of the conclusions obtained previously were reinforced.