Search

This project has received funding from the European Union’s Horizon 2020 research and innovation program under the Marie Sklodowska-Curie grant agreement No 643134., info:eu-repo/semantics/published

Modern society will have to meet its energy demands while ensuring low or virtually zero emissions in order to meet future challenges associated to air pollution, climate change and energy storage. Very often, renewable sources cannot be directly employed because of their intermittent nature and because many applications such as transport and other in- dustrial processes require high energy densities. Therefore, novel storage solutions for the energy that renewable sources contribute to produce is necessary and the transformation of this energy into chemical compounds represents the best choice in order to meet the aforementioned demands, which requires novel combustion technologies, such as Moder- ate and Intense Low-oxygen Dilution (MILD) combustion, to be efficient and fuel-flexible. In order to develop such technologies, several studies are being proposed and terabytes of data collected as more and more experiments and high-fidelity simulations are carried out. However, there are two main challenges to this: the huge amount of data available makes it hard for the researcher to distinguish useful from redundant data, with the risk that useful information might stay hidden; the production process of these data-sets re- quires substantial resources as combustion process are multi-physics, multi-scale and thus require high-fidelity computationally-intensive simulations and experiments over a wide range for their operating conditions or input parameters. Digital twins and Artificial Intel- ligence (AI) are shaping the fourth industrial revolution by building data-driven models that make use of machine learning. It makes sense then to extend this approach to combus- tion applications in order to alleviate the two aforementioned issues: the use of machine learning techniques can help automate the process of data interpretation as well as pro- vide a low-dimensional representation of the high-dimensional data produced by either experiments or simulations; they can speed up the data production process by building reduced-order models that can foresee the outcome of a certain simulation with reduced or negligible computational cost. Besides, such reduced-order models are the foundations for the development of virtual counterparts of real physical systems, which can be employed for system control, non-destructive testing and visualization.With the final objective being to develop reduced-order models for combustion appli- cations, unsupervised and supervised machine learning techniques were tested and com- bined in the work of the present Thesis for feature extraction and the construction of reduced-order models. Thus, the application of data-driven techniques for the detection of features from turbulent combustion data sets (direct numerical simulation) was inves- tigated on two H2/CO flames: a spatially-evolving (DNS1) and a temporally-evolving jet (DNS2). Methods such as Principal Component Analysis (PCA), Local Principal Compo- nent Analysis (LPCA), Non-negative Matrix Factorization (NMF) and Autoencoders were explored for this purpose. It was shown that various factors could affect the performance of these methods, such as the criteria employed for the centering and the scaling of the original data or the choice of the number of dimensions in the low-rank approximations. A set of guidelines was presented that can aid the process of identifying meaningful physical features from turbulent reactive flows data. Data compression methods such as Principal Component Analysis (PCA) and variations were combined with interpolation methods such as Kriging, for the construction of computationally affordable reduced-order models for the prediction of the state of a combustion system for unseen operating conditions or combinations of model input parameter values. The methodology was first tested for the prediction of 1D flames with an increasing number of input parameters (equivalence ra- tio, fuel composition and inlet temperature), with variations of the classic PCA approach, namely constrained PCA and local PCA, being applied to combustion cases for the first time in combination with an interpolation technique. The positive outcome of the study led to the application of the proposed methodology to 2D flames with two input parameters, namely fuel composition and inlet velocity, which produced satisfactory results. Alterna- tives to the chosen unsupervised and supervised methods were also tested on the same 2D data. The use of non-negative matrix factorization (NMF) for low-rank approximation was investigated because of the ability of the method to represent positive-valued data, which helps the non-violation of important physical laws such as positivity of chemical species mass fractions, and compared to PCA. As alternative supervised methods, the combination of polynomial chaos expansion (PCE) and Kriging and the use of artificial neural networks (ANNs) were tested. Results from the mentioned work paved the way for the development of a digital twin of a combustion furnace from a set of 3D simulations. The combination of PCA and Kriging was also employed in the context of uncertainty quantification (UQ), specifically in the bound-to-bound data collaboration framework (B2B-DC), which led to the introduction of the reduced-order B2B-DC procedure as for the first time the B2B-DC was developed in terms of latent variables and not in terms of original physical variables., Doctorat en Sciences de l'ingénieur et technologie, info:eu-repo/semantics/nonPublished

La simulation numérique représente un outil important pour la conception des systèmes de combustion. La prédiction des flux aux parois joue un rôle déterminant dans le cycle de vie des chambres de combustion, car elle permet de prédire la fatigue thermique des parois. 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. Pour évaluer les transferts thermiques aux parois il faut donc tenir compte des flux radiatifs. Les pertes thermiques aux parois dépendent de la répartition de la température des gaz qui est, à son tour, fortement modifiée par le rayonnement des gaz brûlés. Afin d’intégrer les contributions convectives et radiatives au flux pariétal dans des simulations numériques, il est nécessaire de résoudre simultanément l'équation de transfert radiatif et les équations régissant l’écoulement réactif. De nos jours, les simulations couplées impliquant combustion et transfert de chaleur radiatif sont de plus en plus utilisées et ciblées. Grâce à l'augmentation de la puissance de calcul, l'utilisation des méthodes de Monte Carlo (MC) dans des simulations 3D instationnaires, telles que les simulations numériques directes (DNS) et les simulations aux grandes échelles (LES), est devenue. Cependant, de telles simulations restent 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 et appelé intégration Quasi-Monte Carlo (QMC). Cette méthode a rarement été utilisée pour la résolution numérique du rayonnement thermique. Dans cette étude, la méthode QMC est appliquée sur plusieurs configurations 3D et comparée à celle de MC. L’amélioration de l’efficacité obtenue par QMC fait de cette méthode un candidat idéal pour des simulations haute-fidélité couplées avec des simulations LES ou DNS. Au cours de cette thèse, la méthode QMC a pu être appliquée à une configuration où le rayonnement joue un rôle important :la chambre Oxytec, étudiée expérimentalement au laboratoire EM2C. Deux flammes prémélangées swirlées à pression atmosphérique ont été étudiées expérimentalement :une flamme méthane-air (Flamme A) et une oxy-flamme de méthane diluée en CO2 (Flamme B). Malgré leur composition différente, ces flammes partagent de nombreuses caractéristiques communes. Les premières simulations numériques de la chambre Oxytec sont réalisées dans ce travail :une approche QMC, permettant de résoudre l’équation de transfert radiatif avec des propriétés radiatives détaillées des gaz et des parties solides de la chambre, est couplée au solveur LES pour la simulation de la Flamme A. Des simulations couplées LES-QMC sont effectuées en imposant la température mesurée aux parties solides de la chambre de combustion. 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. De plus, les pertes radiatives représentent le 20% de la puissance thermique de la flamme et environ 35% de la puissance radiative émise et absorbée par la flamme est transmise à l'extérieur à travers les fenêtres en quartz. Enfin, un bon accord est trouvé entre le flux de chaleur pariétal numerique et les données expérimentales. Dans la dernière partie de cette thèse, l’étude se concentre sur la flamme B, où l’on s’attend que la concentration élevée de CO2 dans les gaz brûlés augmente le transfert de chaleur radiatif. Il est d’abord montré que la présence d’une espèce absorbante telle que le CO2 dans les gaz frais augmente la vitesse laminaire de flamme d’un facteur qui dépend de la taille de la configuration étudiée. Ensuite, les premiers résultats issus des calculs LES de la Flamme B sont présentés :les résultats préliminaires sur le transfert radiatif sont discutés et comparés à ceux obtenus à partir des simulations couplées de Flamme A., Doctorat en Sciences de l'ingénieur et technologie, info:eu-repo/semantics/nonPublished

Energy generation through combustion of hydrocarbons continues to dominate, as the most common method for energy generation. In the U.S. nearly 84% of the energy consump- tion comes from the combustion of fossil fuels. Because of this demand there is a continued need for improvement, enhancement and understanding of the combustion process. As computational power increases, and our methods for modelling these complex combustion systems improve, combustion modelling has become an important tool in gaining deeper insight and understanding for these complex systems. The constant state of change in computational ability lead to a continual need for new combustion models that can take full advantage of the latest computational resources. To this end, the research presented here encompasses the development of new models, which can be tailored to the available resources, allowing one to increase or decrease the amount of modelling error based on the available computational resources, and desired accuracy. Principal component analysis (PCA) is used to identify the low-dimensional manifolds which exist in turbulent combustion systems. These manifolds are unique in there ability to represent a larger dimensional space with fewer components resulting in a minimal addition of error. PCA is well suited for the problem at hand because of its ability to allow the user to define the amount of error in approximation, depending on the resources at hand. The research presented here looks into various methods which exploit the benefits of PCA in modelling combustion systems, demonstrating several models, and providing new and interesting perspectives for the PCA based approaches to modelling turbulent combustion., Doctorat en Sciences de l'ingénieur, info:eu-repo/semantics/published

Control of pollutants and emissions has become a major factor in the design of modern combustion systems. The “Liquid Hydrogen Fueled Aircraft - System Analysis” project funded in 2000 by the European Commission can be seen as such an initiative. Within the framework of this project, the Aachen University of Applied Sciences developed experimentally the “Micromix” hydrogen combustion principle and implemented it successfully in the Honeywell APU GTCP 36-300 gas turbine engine. Lowering the reaction temperature, eliminating hot spots from the reaction zone and keeping the time available for the formation of NOx to a minimum are the prime drivers towards NOx reduction. The “Micromix” hydrogen combustion principle meets those requirements by minimizing the flame temperature working at small equivalence ratios, improving the mixing by means of Jets In Cross-Flow and reducing the residence time in adopting a combustor geometry that provides a very large number of very small diffusion flames. In terms of pollutant emissions, compared to the unconverted APU, an essential reduction in emitted NOx was observed, stressing the potential of this innovative burning principle.The objective of this thesis is to investigate the “Micromix” hydrogen combustion principle with the ultimate goal of an improved prediction during the design process. Due to the complex interrelation of chemical kinetics and flow dynamics, the “Micromixing” was analyzed first. Stereoscopic Particle Image Velocimetry was used to provide insight into the mixing process. A “simplified” set-up, that allowed to investigate the flow characteristics in great detail while retaining the same local characteristics of its “real” counterparts, was considered. The driving vortical structures were identified. To further investigate the physics involved and to extend the experimental results, numerical computations were carried out on the same “simplified” set-up as on a literature test case. In general, a number of physical issues were clarified. In particular, the interaction between the different vortical structures was looked into, and a kinematically consistent vortex model is proposed. After demonstrating the development of the mixing, the “cold flow” study was extended to a single injector. The double backward-facing step injector geometry was addressed experimentally and numerically. At design geometry, the flow appeared to behave single backward-facing like, with respect to the first gradation. In terms of varying step configurations, the flow was seen to be dependent on the periodic perturbation arising from the graded series of backward-facing steps. During the second part of the investigation, the “hot flow” was analyzed. Considering combustor similar operating conditions, a test burner was experimented on an atmospheric test rig. NOx emissions were traced by exhaust gas analysis for different working conditions. Particular flame patterns, such as a regular attached flame as well as lifted flames were observed. In parallel with the experimental work, numerical computations on a pair of opposite injectors, permitted to classify the combustion regime and the main factors involved in the NOx formation. Accordingly, NOx emission enhancing design changes are proposed. Finally, the demanding computational effort, worthy of acceptance for academic purposes, is found not agreeable as future design tool and improvements to speed up the design process are projected., Doctorat en Sciences de l'ingénieur, info:eu-repo/semantics/nonPublished

La première partie de cette thèse traite du développement du code de simulation numérique directe YWC, principalement du développement des conditions aux limites. En effet, une forte contribution scientifique a été apportée aux conditions aux limites appelées "Three dimensional Navier-Stokes characteristic boundary condtions" (3D-NSCBC). Premièrement, la formulation de ces conditions aux arêtes et coins a été complétée, ensuite une extension de la formulation a été proposée pour supprimer les déformations observées en sortie dans le cas d'écoulements non-perpendiculaires à la frontière. De plus, ces conditions ont été étendues au cas des gaz réels et une nouvelle définition du facteur de relaxation pour la pression a été proposée. Ce nouveau facteur de relaxation permet de supprimer les déformations observées en sortie pour des écoulements transcritiques. Les résultats obtenus avec le code YWC ont ensuite été utilisés dans la seconde partie de la thèse pour développer une nouvelle méthode de tabulation basée sur l'analyse en composantes principales. Par rapport aux méthodes existante telles que FPI ou SLFM, la technique proposée, permet une identification automatique des variables à transporter et n'est, de plus, pas lié à un régime de combustion spécifique. Cette technique a permis d'effectuer des calculs d'interaction flamme-vortex en ne transportant que 5 espèces à la place des 9 requises pour le calcul en chimie détaillée complète, sans pour autant perdre en précision. Finalement, dans le but de réduire encore le nombre d'espèces transportées, les techniques T-BAKED et HT-BAKED PCA ont été introduites. En utilisant une pondération des points sous-représentés, ces deux techniques permettent d'augmenter la précision de l'analyse par composantes principales dans le cadre des phénomènes de combustion., Doctorat en Sciences de l'ingénieur, info:eu-repo/semantics/nonPublished