No description available.

A novel approach to modelling highly compressible and reactive flows is formulated to provide high resolution closure of turbulent-scale reaction rates in the presence of very rapid transients in pressure and energy. For such flows, treatment of turbulent-micro scales are generally unattainable through traditional modelling techniques. To address this, the modelling strategy developed here is based on the Linear Eddy Model for Large Eddy Simulation (LEM-LES); a technique which has only previously been applied to weakly compressible flows. In the current formulation of the Compressible LEM-LES (CLEM-LES), special treatment of the energy balance on the model subgrid is accounted for in order for the model reaction rates to respond accordingly to strong shocks and rapid expansions, both of which may be present in reactive and supersonic flow fields. In the current study, the model implemented is verified and validated for various 1D and 2D flow configurations in a compressible Adaptive Mesh Refinement (AMR) framework. In 1D test cases, laminar and turbulent flame speeds and structure have been reproduced. Also, detonation speeds and initiation events are also captured with the model. For 2D model validation, unsteady and turbulent detonation propagation and initiation events, in a narrow channel, are simulated. Both test cases involve premixed methane-oxygen mixture at low pressures. The model is found to capture well the two-dimensional detonation cellular structure, behaviour, and initiation events that are observed in corresponding shock tube experiments. Furthermore, the effect of turbulent mixing rates is investigated though a single tuning constant. It was found that by increasing the intensity of turbulent fluctuations present, detonations exhibit larger and more irregular cell structures. Furthermore, the intensity of turbulent fluctuations is found to also have an effect on initiation events.

Subfilter scalar variance is a key quantity for scalar mixing at the small scales of a turbulent flow and thus plays a crucial role in large eddy simulation (LES) of combustion. While prior studies have mainly focused on the physical aspects of modeling subfilter variance, the current work discusses variance models in conjunction with numerical errors due to their implementation using finite difference methods. Because of the prevalence of grid-based filtering in practical LES, the smallest filtered scales are generally under-resolved. These scales, however, are often important in determining the values of subfilter models. A priori tests on data from direct numerical simulation (DNS) of homogenous isotropic turbulence are performed to evaluate the numerical implications of specific model forms in the context of practical LES evaluated with finite differences. As with other subfilter quantities, such as kinetic energy, subfilter variance can be modeled according to one of two general methodologies. In the first of these, an algebraic equation relating the variance to gradients of the filtered scalar field is coupled with a dynamic procedure for coefficient estimation. Although finite difference methods substantially underpredict the gradient of the filtered scalar field, the dynamic method is shown to mitigate this error through overestimation of the model coefficient. The second group of models utilizes a transport equation for the subfilter variance itself or for the second moment of the scalar. Here, it is shown that the model formulation based on the variance transport equation is consistently biased toward underprediction of the subfilter variance. The numerical issues stem from making discrete approximations to the chain rule manipulations used to derive convective and diffusive terms in the variance transport equation associated with the square of the filtered scalar. This set of approximations can be avoided by solving the equation for the second moment of the scalar, suggesting that model's numerical superiority. / text

This thesis describes the implementation of the conditional moment closure (CMC) combustion model in a numerical scheme and its application to the modelling of enclosure fires. Prediction of carbon monoxide (CO) in the upper smoke layer of enclosure fires is of primary interest because it is a common cause of death. The CO concentration cannot be easily predicted by empirical means, so a method is needed which models the chemistry of a quenched, turbulent fire plume and subsequent mixing within an enclosed space. CMC is a turbulent combustion model which has been researched for over a decade. It has provided predictions of major and minor species in jet diffusion flames. The extension to enclosure fires is a new application for which the flow is complex and temperatures are well below adiabatic conditions. Advances are made in the numerical implementation of CMC. The governing combustion equations are cast in a conserved, finite volume formulation for which boundary conditions are uniquely defined. Computational efficiency is improved through two criteria which allow the reduction in the size of the computational domain without any loss of accuracy. Modelling results are compared to experimental data for natural gas fires burning under a hood. Comparison is made in the recirculating, post-flame region of the flow where temperatures are low and reactions are quenched. Due to the spatial flux terms contained in the governing equations, CMC is able to model the situation where chemical species are produced in the high temperature fire-plume and then transported to non-reacting regions. Predictions of CO and other species are in reasonable agreement with the experimental data over a range of lean and rich hood-fire conditions. Sensitivity of results to chemistry, temperature and modelling closures is inves- tigated. Species predictions are shown to be quite different for the two detailed chemical mechanisms used. Temperature conditions within the hood effect the for- mation of species in the plume prior to quenching and subsequently species predic- tions in the post-flame region are also effected. Clipped Gaussian and ß-function probability density functions (PDFs) are used for the stochastic mixture fraction. Species predictions in the plume are sensitive to the form of the PDF but in the post-flame region, where the ß-function approaches a Gaussian form, predictions are relatively insensitive. Two models are used for the conditional scalar dissipation: a uniform model, where the conditional quantity is set equal to the unconditional scalar dissipation across all mixture fraction space; and a model which is consistent with the PDF transport equation. In the plume, predictions of minor species are sensitive to the modelling used, but in the recirculating, post-flame region species are not significantly effected.

No description available.

The propagation of premixed flames in turbulent flows is a problem of wide physical and technological interest, with a significant literature on their propagation speed and front topology. While certain scalings and parametric dependencies are well understood, a variety of problems remain. One major challenge, and focus of this thesis, is to model the influence of fuel/oxidizer composition on turbulent burning rates. Classical explanations for augmentation of turbulent burning rates by turbulent velocity fluctuations rely on global arguments - i.e., the turbulent burning velocity increase is directly proportional to the increase in flame surface area and mean local burning rate along the flame. However, the development of such global approaches is complicated by the abundance of phenomena influencing the propagation of turbulent premixed flames. Emphasizing key governing processes and cutting-off interesting but marginal phenomena appears to be necessary to make further progress in understanding the subject. An alternative approach to understand turbulent augmentation of burning rates is based upon so-called "leading points", which are intrinsically local properties of the turbulent flame. Leading points concepts suggest that the key physical mechanism controlling turbulent burning velocities of premixed flames is the velocity of the points on the flame that propagate farthest out into the reactants. It is postulated that modifications in the overall turbulent combustion speed depend solely on modifications of the burning rate at the leading points since an increase (decrease) in the average propagation speed of these points causes more (less) flame area to be produced behind them. In this framework, modeling of turbulent burning rates can be thought as consisting of two sub-problems: the modeling of (1) burning rates at the leading points and of (2) the dynamics/statistics of the leading points in the turbulent flame. The main objective of this thesis is to critically address both aspects, providing validation and development of the physical description put forward by leading point concepts. To address the first sub-problem, a comparison between numerical simulations of one-dimensional laminar flames in different geometrical configurations and statistics from a database of direct numerical simulations (DNS) is detailed. In this thesis, it is shown that the leading portions of the turbulent flame front display a structure that on average can be reproduced reasonably well by results obtained from model geometries with the same curvature. However, the comparison between model laminar flame computations and highly curved flamelets is complicated by the presence of negative (i.e., compressive) strain rates, due to gas expansion. For the highest turbulent intensity investigated, local consumption speeds, curvatures, strain rates and flame thicknesses approach the maximum values obtained by the laminar model geometries, while other cases display substantially lower values. To address the second sub-problem, the dynamics of flame propagation in simplified flow geometries is studied theoretically. Utilizing results for Hamilton-Jacobi equations from the Aubry-Mather theory, it is shown how the overall flame front progation under certain conditions is controlled only by discrete points on the flame. Based on these results, definitions of leading points are proposed and their dynamics is studied. These results validate some basic ideas from leading points arguments, but also modify them appreciably. For the simple case of a front propagating in a one-dimensional shear flow, these results clearly show that the front displacement speed is controlled by velocity field characteristics at discrete points on the flame only when the amplitude of the shear flow is sufficiently large and does not vary too rapidly in time. However, these points do not generally lie on the farthest forward point of the front. On the contrary, for sufficiently weak or unsteady flow perturbations, the front displacement speed is not controlled by discrete points, but rather by the entire spatial distribution of the velocity field. For these conditions, the leading points do not have any dynamical significance in controlling the front displacement speed. Finally, these results clearly show that the effects of flame curvature sensitivity in modifying the front displacement speed can be successfully interpreted in term of leading point concepts.

No description available.

This thesis describes the implementation of the conditional moment closure (CMC) combustion model in a numerical scheme and its application to the modelling of enclosure fires. Prediction of carbon monoxide (CO) in the upper smoke layer of enclosure fires is of primary interest because it is a common cause of death. The CO concentration cannot be easily predicted by empirical means, so a method is needed which models the chemistry of a quenched, turbulent fire plume and subsequent mixing within an enclosed space. CMC is a turbulent combustion model which has been researched for over a decade. It has provided predictions of major and minor species in jet diffusion flames. The extension to enclosure fires is a new application for which the flow is complex and temperatures are well below adiabatic conditions. Advances are made in the numerical implementation of CMC. The governing combustion equations are cast in a conserved, finite volume formulation for which boundary conditions are uniquely defined. Computational efficiency is improved through two criteria which allow the reduction in the size of the computational domain without any loss of accuracy. Modelling results are compared to experimental data for natural gas fires burning under a hood. Comparison is made in the recirculating, post-flame region of the flow where temperatures are low and reactions are quenched. Due to the spatial flux terms contained in the governing equations, CMC is able to model the situation where chemical species are produced in the high temperature fire-plume and then transported to non-reacting regions. Predictions of CO and other species are in reasonable agreement with the experimental data over a range of lean and rich hood-fire conditions. Sensitivity of results to chemistry, temperature and modelling closures is inves- tigated. Species predictions are shown to be quite different for the two detailed chemical mechanisms used. Temperature conditions within the hood effect the for- mation of species in the plume prior to quenching and subsequently species predic- tions in the post-flame region are also effected. Clipped Gaussian and ß-function probability density functions (PDFs) are used for the stochastic mixture fraction. Species predictions in the plume are sensitive to the form of the PDF but in the post-flame region, where the ß-function approaches a Gaussian form, predictions are relatively insensitive. Two models are used for the conditional scalar dissipation: a uniform model, where the conditional quantity is set equal to the unconditional scalar dissipation across all mixture fraction space; and a model which is consistent with the PDF transport equation. In the plume, predictions of minor species are sensitive to the modelling used, but in the recirculating, post-flame region species are not significantly effected.

Pour satisfaire les exigences environnementales et d'agrément de conduite, le moteur automobile a évolué en une vingtaine d'années en un système très complexe combinant de nombreux composants de haute technologie avec des stratégies de contrôle très élaborées. L’optimisation et le contrôle de ce système sont alors devenus de véritables challenges pour les constructeurs automobiles. Ces derniers points sont aujourd'hui d'autant plus complexes que le contexte actuel de raréfaction des ressources impose de plus en plus le couplage ou le remplacement des carburants conventionnels par des carburants alternatifs tels que l’éthanol, le gaz naturel ou encore l’hydrogène. Ces nouveaux carburants présentent, en plus de leur intérêt économique, un certain nombre de propriétés physico-chimiques favorisant un meilleur rendement du moteur ainsi que la réduction des gaz à effet de serre. L’élaboration de ces nouveaux moteurs est finalement rendue possible par l'utilisation de dispositifs physiques et numériques de plus en plus sophistiqués. Dans ce contexte, les outils de simulation système destinés aux groupes motopropulseurs se sont démocratisés et peuvent aujourd'hui être utilisés à toutes les étapes de développement des moteurs, du choix de l’architecture au développement des stratégies de contrôle et à la calibration. Cependant, l'efficacité de tels outils demande encore à être améliorée afin de fournir un haut niveau de prédictivité couplé à un temps de calcul proche du temps réel. Les travaux réalisés lors de cette thèse ont visé à contribuer au développement du modèle de combustion 0-dimensionnel CFM1D (Coherent Flame Model) afin d’améliorer la prédiction du dégagement d'énergie, des polluants et des phénomènes d'auto-inammation (AI) dans les moteurs à allumage commandé lorsque des variations de la composition du carburant sont considérées. Le formalisme CFM distingue deux zones : les gaz frais et les gaz brûlés qui sont séparés par un front de flamme et qui sont entièrement décrits par leur masse, température et composition. Dans ce formalisme, le taux de consommation des espèces est directement lié aux processus de combustion et de post-oxydation assujettis aux mécanismes de chimie et de turbulence. Dans la version initiale du CFM1D, ces mécanismes sont représentés par des approches simples pouvant souffrir d'un manque de prédictivité. Ainsi, la prédiction de la formation de polluants peut être limitée par les chimies simples ou réduites la décrivant. Ces dernières sont en effet généralement définies dans des domaines de validité restreints en température, pression et composition. De la même manière, le calcul de la vitesse de flamme laminaire, de l'étirement de la flamme ou encore des éventuels délais d'auto-inammation intervenant dans l'évaluation du dégagement d'énergie met en jeux des corrélations phénoménologiques initialement développées sur un nombre limités de points de validation. Toutes ces limitations peuvent finalement entraîner une mauvaise réaction du modèle de combustion à des variations thermodynamiques ou de compositions et ont donc nécessite un certain nombre d'améliorations présentées dans ce manuscrit. L'originalité des développements réside dans l'intégration de chimie complexe dans le modèle CFM1D en utilisant des méthodes inspirées de récents travaux de CFD (Computational Fluid Dynamics) 3D. / A promising way to reduce green house gases emissions of spark ignition (SI) engines is to burn alternative fuels like bio-mass-derived products, hydrogen or compressed natural gas. However, their use strongly impacts combustion processes in terms of burning velocity and emissions. Specific engine architectures as well as dedicated control strategies should then be optimized to take advantage of these fuels. Such developments are today increasingly performed using complete engine simulators running in times close to the real time and thus requiring very CPU efficient models. For this purpose, 0-dimensional models are commonly used to describe combustion processes in the cylinders. These models are expected to reproduce the engine response for all possible fuels, which is not an obvious task regarding the mentioned CPU constraints. Works performed in this thesis aimed at developing the 0-dimensional combustion model CFM1D (Coherent Flame Model) to improve the prediction of heat release, pollutants emissions and auto-ignition phenomena in SI engines when fuel composition variations are considered. The CFM formalism distinguishes two zones: the fresh and the burnt gases, which are separated by a flame front and are both described by their temperature, mass and composition. In this formalism, the rate of consumption of species is directly linked to the combustion and post-oxidation processes highly dependent on chemistry and turbulence mechanisms. In the original version of CFM1D, these mechanisms are represented by simple approaches which can suffer from a lack of predictivity. The prediction of pollutant formation can therefore be limited by the simple or reduced chemistries used to describe kinetics in the chamber. These latter are indeed defined in very restrictive validity domains in terms of temperature, pressure and composition. In the same way, the flame velocity, wrinkling or potential auto-ignition delays stepping in the heat release computation are defined by phenomenological correlations initially developed under a limited number of validation points. All these limitations can finally lead to a wrong behavior of the combustion model to thermodynamic and compositions variations and therefore required a number of improvements presented in this manuscript. The originality of the model derives from the fact it is based on the integration of complex chemistry in CFM1D using methods inspired from recent 3D (Computational Fluid Dynamics) CFD works.

Combustion research currently aims to reduce emissions, whilst improving the fuel economy. Burning fuel in excess of air, or lean-burn combustion, is a promising alternative to conventional combustion, and can achieve these requirements simultaneously. However, lean-burn combustion poses new challenges, especially for internal combustion (IC) engines. Therefore, models used to predict such combustion have to be reliable, accurate and robust. In this work, the flamelet approach in the Reynolds-Averaged Navier- Stokes framework, is used to simulate flames relevant to spark-ignition IC engines. A central quantity in the current modelling approach is the scalar dissipation rate, which represents coupling between reaction and diffusion, as well as the flame front dynamics. In the first part of this thesis, the predictive ability of two reaction rate closures, viz. strained and unstrained flamelet models, are assessed through a series of experimental test cases. These cases are: spherically propagating methane- and hydrogen-air flames and combustion in a closed vessel. In addition to these models, simpler algebraic closures are also used for comparison. It is shown that the strained flamelet model can predict unconfined, spherically propagating methane-air flames reasonably well. By comparing spherical flame results with planar flames, under identical thermochemical and turbulence conditions, it is shown that the turbulent flame speed of spherical flames are 10 to 20% higher than that of planar flames, whilst the mean reaction rates are less influenced by the flame geometry. Growth of the flame brush thickness in unsteady spherical flames have been attributed to turbulent diffusion in past studies. However, the present analyses revealed that the dominant cause for this increase is the heat-release induced convective effects, which is a novel observation. Unlike methane-air flames, hydrogen-air flames have non-unity Lewis numbers. Hence, a novel two degrees of freedom approach, using two progress variables, is used to describe the thermochemistry of hydrogen-air flames. Again, it is shown that the strained flamelet model is able to predict the experimental flame growth for stoichiometric hydrogen-air flames. However, none of the models used in this work were able to predict lean hydrogen-air flames. This is because these flames are thermo-diffusively unstable and the current approach is inadequate to represent them. When combustion takes place inside a closed vessel, the compression of the end gases by the propagating flame causes the pressure to rise. This is more representative of real IC engines, where intermittent combustion takes place. The combustion models are implemented in a commercial computational fluid dynamics (CFD) code, STAR-CD, and it is shown that both strained and unstrained flamelet models are able to predict the experimental pressure rise in a closed vessel. In the final part of this work, a spark-ignition engine is simulated in STAR-CD using the flamelet model verified for simpler geometries. It is shown that this model, together with a skeletal mechanism for iso-octane, compares reasonably well with experimental cylinder pressure rise. Results obtained from this model are compared with two models available in STAR- CD. These models require some level of tuning to match the experiments, whereas the modelling approach used in this work does not involve any tunable parameters.