Combustion science must necessarily go through a deep process of innovation, as only improving the energy efficiency and the fuel flexibility it will be possible to mitigate the impact of the anthropogenic activities on the climate and the environment. Because of the strong relation that is observed in chemically reacting flows between the fluid-dynamic conditions and the chemical kinetics, the use of Computational Fluid Dynamics (CFD) simulations with detailed kinetic mechanisms represents the best tool to optimize and develop novel combustion systems. In fact, while the CFD provides for the possibility to retrieve information that cannot be extracted by using experimental means (such as the turbulence-chemistry interaction and the local straining rates) and it avoids the costs associated to the scale-up process from laboratory scale experiments, the use of detailed kinetic mechanisms offers the possibility to correctly describe process conditions which are relevant from an industrial point of view (i.e. in which the chemical and mixing time scales are comparable), as well as to predict the formation of complex chemical species, such as the pollutants. Nevertheless, the use of detailed kinetic mechanisms in numerical simulations adds a considerable number of differential equations to be solved (because of the large number of species which are taken into account), and therefore increases the computational complexity of the CFD model. Thus, Machine Learning (ML) algorithms and Reduced-Order Models (ROMs) can be effectively included in the numerical description of chemically reacting flows. In fact, they can be used either to reduce the computational cost associated to the large number of equations in CFD simulations carried out with detailed chemistry, or to leverage the detailed information which can be found in massive, high-fidelity, data obtained from Direct Numerical Simulations (DNS), for model development and validation. In this Thesis, unsupervised and supervised learning algorithms were employed to design a novel adaptive-chemistry approach: the Sample-Partitioning Adaptive Reduced Chemistry (SPARC). This framework can be used to reduce the computational effort required by detailed CFD simulations thanks to a kinetic reduction accomplished in light of the local conditions of the thermochemical field. Several machine-learning algorithms, such as the Principal Component Analysis (PCA), the Local Principal Component Analysis (LPCA), and Artificial Neural Networks (ANNs) were coupled with the Direct Relation Graph with Error Propagation (DRGEP), a graph-based tool for the automatic reduction of kinetic mechanisms. The aforementioned algorithms were compared to achieve the optimal formulation of the adaptive approach, such that the best performances, in terms of accuracy and computational speed-up with respect to the CFD simulation carried out with detailed kinetics, could be obtained. Finally, PCA-based algorithms were proposed and tested to perform feature extraction and local feature selection from high-fidelity data, which were obtained by means of a DNS of a n-heptane jet reacting in air. The PCA, as well as two formulations of LPCA, and the Procrustes analysis were employed and compared with the aim to extract the main features of the turbulent reacting jet in an unsupervised fashion (i.e. to perform data mining tasks), as well as to aid the formulation of local optimized ROMs. All the codes employed to perform the unsupervised and supervised machine learning tasks in the current work were also included in an open-source Python framework, called OpenMORe, designed to perform reduction, clustering and data analysis, and specifically conceived for reacting flows. In fact, although many open-source Python software are already available, they often cannot be adapted to the user’s specific needs, unlike OpenMORe. In addition, many features such as the PCA-based clustering algorithm, or the local feature selection via PCA, are not yet available on any commercial or open-source software, to the best of the author’s knowledge. / Doctorat en Sciences de l'ingénieur et technologie / This thesis is submitted to the Université Libre de Bruxelles (ULB) and to the Politecnico di Milano for the degree of philosophy doctor. This doctoral work has been performed at the Université Libre de Bruxelles, École polytechnique de Bruxelles, Aero-Thermo-Mechanics Laboratory, Bruxelles, Belgium with Professor Alessandro Parente and at the Politecnico di Milano, CRECK Modelling Lab, Department of Chemistry, Materials and Chemical Engineering, Milan, Italy with Professor Alberto Cuoci. / info:eu-repo/semantics/nonPublished

Doctorat en sciences appliquées / info:eu-repo/semantics/nonPublished

Doctorat en sciences appliquées / info:eu-repo/semantics/nonPublished

Les cristaux photoniques (CP) nous fournissent un moyen sans précédent de contrôler et de confiner les photons. En particulier, les cristaux photoniques membranaires (CP-2D) (confinement vertical par réflexion totale interne), et l'effet de bande interdite photonique (confinement dans le plan), jouent un rôle très important en nanophotonique. En introduisant des défauts dans le cristal (omission d'un ou de plusieurs motifs), il est possible de générer des modes optiques très localisés, possédant un faible volume modal et un grand facteur de qualité. Coupler un émetteur unique à ce type de mode peut être alors utile pour exalter ou inhiber sa dynamique. L'efficacité du couplage dépend à la fois de l'accord en longueur d'onde entre l'émetteur et le mode optique ainsi que de leur recouvrement spatial, c'est pourquoi le contrôle du couplage entre l'émetteur et le mode localisé doit être optimal. Il est donc crucial de connaître le profil du mode dans la cavité à une échelle suffisamment petite. Jusqu'à présent, le SNOM s'est révélé un outil de caractérisation particulièrement bien adapté à l'observation directe du champ dans des structures nanophotoniques. En effet, il permet d'accéder aux parties évanescentes des modes, livrant ainsi des informations locales, inaccessibles par des mesures classiques en champ lointain. Cependant, le potentiel des structures à base de CP-2D pour développer de nouveaux composants photoniques serait considérablement accru si leurs propriétés optiques pouvaient être modifiées après fabrication. En particulier, pour les structures actives, la capacité à accorder de manière réversible une cavité à l'emission ou à l'absorption d'une source est de grand intérêt. Dans cette étude, l'idée est donc d'utiliser la sonde d'un SNOM dans le but de perturber la résonnance (longueur d'onde, facteur de qualité) d'un mode. Dans ce travail, Nous concevons et optimisons une cavité linéaire dans laquelle sept trous ont été omis (CL7). Nous étudions théoriquement les propriétés de la cavité et particulièrement un mode présentant un bon facteur de qualité. Ensuite, l'interaction de pointes champ proche avec les modes de la CL7 et en particulier l'effet du matériau de la pointe sur l'émission en longueur d'onde et les pertes induites relatives sont étudiés. Dans ce but, une pointe est introduite dans des simulations FDTD-3D, soit avec l'indice de la silice (1.44), soit avec l'indice du silicium (3.4). Ensuite, l'effet de la forme de la pointe sur les cartographies est étudié, et particulièrement l'influence de la polarisabilité et de la section efficace. Les cavités CL7 sont fabriquées et caractérisées par un dispositif d'optique réfractive et par un microscope optique de champ proche. Le mode fondamental, présentant un effet laser, est utilisé pour étudier l'interaction ave la pointe champ proche. Des caractérisations SNOM avec une pointe en silice et une pointe hybride silice/silicium sont réalisées sur la cavité CL7 et nous démontrons que la pointe en silice recouverte de silicium provoque un décalage en longueur d'onde de l'ordre de quelques nanomètres, de 5 à 10 fois supérieur que la largeur intrinsèque du pic. La pointe en silice induitdes décalages en longueur d'onde de l'ordre du dixième de nanomètre, qui n'est pas détecté par notre montage expérimental. Nous démontrons également l'importance de la forme de la pointe lors de l'observation directe et locale de la distribution du champ avec le SNOM. Nous montrons qu'une pointe isotrope de bas indice peut être utilisée comme un outil de caractérisation passive car la cartographie champ proche donne une bonne approximation du mode tel qu'il existe dans la structure sans présence de la pointe. A l'inverse, une pointe anisotrope donne une information partielle car elle ne convertie le champ que dans la direction du petit axe.

Pas de résumé

Le bruit de rugosité, rayonné lors du frottement entre surfaces rugueuses sous faible chargement, est un phénomène complexe dépendant de plusieurs paramètres dont la rugosité de surface, la vitesse de glissement, le chargement normal appliqué et les matériaux frottés. Il se distingue des autres types du bruit de frottement par un spectre fréquentiel large bande, un niveau sonore faible, une faible pression de contact et par conséquent un couplage dynamique faible entre les solides en contact. Pour étudier ce type du bruit, deux approches ont été développées: expérimentale et numérique. L'approche expérimentale a permis de proposer des lois de variation des niveaux sonores et vibratoires en fonction des paramètres tribologiques (principalement la rugosité) et cinématique (principalement la vitesse de glissement). L'approche numérique est consacrée à l'étude de la dynamique locale des interfaces multicontact. Elle a permis ainsi d'accéder aux paramètres inaccessibles à la mesure (forces locales, durée de chocs inter-aspérités, fréquence de chocs) et d'établir une comparaison avec les mesures sonores et vibratoires. Les résultats aussi bien expérimentaux que numériques montrent que le bruit de rugosité est une fonction croissante de la rugosité de surface et de la vitesse de glissement. La phénoménologie du bruit de rugosité fait appel à trois disciplines différentes: la tribologie (génération des vibrations à l'interface multicontact dues aux variations des efforts de contact suite aux chocs inter-aspérités), la dynamique (propagation de ces vibrations dans les solides en contact) et l'acoustique (une partie de ces vibrations est dissipée sous forme de bruit de rugosité). Le mécanisme fondamental du bruit de rugosité est ainsi la mise en vibration des solides en contact générée par les chocs inter-aspérités. Ces chocs sont caractérisés par leur fréquence, leur percussion et leur durée. La fréquence diminue quand la rugosité ou la vitesse de glissement augmente, la percussion augmente avec ces deux paramètres et la durée de choc est indépendante de la rugosité de surface et la vitesse de glissement.

info:eu-repo/semantics/nonPublished

The motion of the free liquid surface inside a reservoir is called sloshing. Itis of large interest in different industrial fields such as satellite and spacecrafttrajectory control, automotive industry, nuclear engineering, buildingdesign, etc. The framework of propellant management on spacecraft is ofmain interest for this PhD thesis, even if its outcome can be applied to manyother fields concerned by sloshing.Being able to understand the behavior of the fluid in a reservoir subjectedto extreme environmental conditions means being able to predict its positionand topology inside the tank, for a given external and gravitationalacceleration and a determined thermodynamic condition. The predictionand control of this motion is far from being understood due to the differentparameters that play a role in the dynamic system such as the geometryof the container, the type of external excitation (shape, frequency contentand amplitude), the level of the liquid and finally the kind of liquid. In particular,the design of propulsion systems are affected by this phenomenon,still hampered by the unavailability of validated CFD models. Moreover theexisting experimental studies are mainly based on intrusive and local singlepoint measurement techniques, which give no information on the behaviorof the 3D liquid interface and on the velocity field inside the liquid phase.The main goal of this project has been to extend the experimental approachof liquid sloshing investigation in space propulsion, studying, developing andimproving non-intrusive measurement techniques for free surface behaviorand velocity characterization in the liquid phase. In particular, the free surfacebehavior have been studied by means of Laser Detection and Recordingtechnique (LeDaR), retrieving the profile of the interface over a line, andReference Image Topography technique (RIT), capturing the instantaneous3D interface shape. In addition, Particle Image Velocimetry (PIV) have beenused to measure the 2D velocity field in the main section of the reservoir.Tests performed with water were used as simpler test case to perform thetechniques while liquid nitrogen has been used as replacement uid havingphysical properties similar to real space propellants.The experimental problems of the selected measurement techniques relatedto the particular application have been addressed and a solution has beenproposed. Especially, the selection of tracers which could comply with theuse of a cryogenic fluid while for RIT the possibility to deal with circulardomains and to measure the absolute value of the liquid level. Finally, PIV in wavy ows needed to deal with dynamic curved interfaces for whicha widely-accepted processing algorithm was not available in literature andbesides, the choice of the particles and their seeding procedure in cryogenicsfluids had to be solved.Some applications are shown, which present the potentiality of the techniquesfor a new insight on sloshing flows with the future purpose of providingan accurate database for the verification and validation of numericalsimulations and a better understanding of the phenomena. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

Human-swarm interaction studies how human beings can interact with a robotswarm---a large number of robots cooperating with each other without any form of centralizedcontrol. In today's human-swarm interaction literature, the large majority of the works investigatehow human beings can issue commands to and receive feedback from a robot swarm. However, only a few ofthese works study the effect of the interaction with a robot swarm on human psychology (e.g. on thehuman stress or on the human workload). Understanding human psychology in human-swarm interaction isimportant because the human psychological state can have significant impact on the way humansinteract with robot swarms (e.g. a high level of stress can cause a human operator to freeze in themiddle of a critical task, such as a search-and-rescue task). Most existing works that study human psychology in human-swarm interaction conduct their experimentsusing robot swarms simulated on a computer screen. The use of simulation is convenient becauseexperimental conditions can be repeated perfectly in different experimental runs and becauseexperimentation using real robots is expensive both in money and time. However, simulation suffersfrom the so-called reality gap: the inherent discrepancy between simulation and reality. Itis therefore important to study whether this inherent discrepancy can affect humanpsychology---human operators interacting with a simulated robot swarm can react differently thanwhen interacting with a real robot swarm.A large literature in human-robot interaction has studied the psychological impact of theinteraction between human beings and single robots. This literature could in principle be highlyrelevant to human-swarm interaction. However, an inherent difference between human-robot interactionand human-swarm interaction is that in the latter, human operators interact with a large number ofrobots. This large number of robots can affect human psychology---human operators interacting with alarge number of robots can react differently than when interacting with a single robot or with asmall number of robots. It is therefore important to understand whether the large number of robotsthat composes a robot swarm affects human psychology. In fact, if this is the case, it would not bepossible to directly apply the results of human-robot interaction research to human-swarminteraction.We conducted several experiments in order to understand the effect of the reality gap and the effectof the group size (i.e. the number of robots that composes a robot swarm) on the humanpsychological state. In these experiments our participants are exposed to swarms of robots and arepurely passive---they do not issue commands nor receive feedback from the robots. Making theinteraction passive allowed us to study the effects of the reality gap and of the group size on thehuman psychological state without the risk that an interaction interface (such as a joystick)influences the psychological responses of the participants (and thus limiting the visibility of both thereality gap and group size effects). In the reality gap experiments, participants are exposed tosimulated robot swarms displayed either on a computer screen or in a virtual reality environment, and toreal robot swarms. In the group size experiments, participants are exposed to an increasing numberof real robots.In this thesis, we show that the reality gap and the group size affect the human psychological stateby collecting psychophysiological measures (heart rate and skin conductance), self-reported (viaquestionnaires) affective state measures (arousal and valence), self-reported workload (the amountof mental resource needed to carry out a task) and reaction time (the time needed to respond to astimulus). Firstly, we show with our results that our participants' psychophysiological measures,affective state measures, workload and reaction time are significantly higher when they interactwith a real robot swarm compared to when they interact with a robot swarm simulated on a computerscreen, confirming that the reality gap significantly affects the human psychological state.Moreover, we show that it is possible to mitigate the effect of the reality gap using virtualreality---our participants' arousal, workload and reaction time are significantly higher when theyinteract with a simulated robot swarm displayed in a virtual reality environment as opposed to whenit is displayed on a computer screen. Secondly, we show that our participants' psychophysiologicalmeasures and affective state measures increase when the number of robots they are exposed toincreases. Our results have important implications for research in human-swarm interaction. Firstly, for thefirst time, we show that experiments in simulation change the human psychological state compared toexperiments with real robots. Secondly, we show that a characteristic that is inherent to thedefinition of swarm robotics---the large number of robots that composes a robotswarm---significantly affects the human psychological state. Finally, our results show thatpsychophysiological measures, such as heart rate and skin conductance, provide researchers with moreinformation on human psychology than the information provided by using traditional self-reportedmeasures (collected via psychological questionnaires). / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished

Chemical strengthening via ion exchange is commonly used in the industry to produce ultra-thin and resistant glasses used for high-tech applications. This process results in the formation of an ion-exchanged region near the glass surface that is under significant compressive stress, which tremendously improves the mechanical strength and the scratch resistance of the glass. The aim of this thesis is to deepen our fundamental understanding of the ion-exchange mechanisms. Since the mechanical strength against brittle fracture of ion-exchanged glasses depends on the compressive stress at the surface and the depth of interdiffusion, special attention was paid to the two main parameters affecting those properties: the parent glass composition and stress relaxation processes.First, the effect of the parent glass composition on interdiffusion properties was analyzed. We focused our study on the effect of the alkali oxide:alumina ratio and the ionic potential (charge:radius) of divalent cations on the interdiffusional properties. An increase of the Al2O3 content in Na+/K+ ion-exchanged silicate glasses leads to a reduction of the depth of interdiffusion layer and an enhancement of the surface compressive stress. These trends are shown to be consistent with the compositional variation of the glass connectivity and hence, the stress relaxation occurring during the ion-exchange. It is also found that the depth of interdiffusion is higher for glasses containing divalent cations with a higher ionic potential such as Mg and Zn. This is attributed to the larger free volume available and the weaker Na-O bonds in those glasses. Second, the structural origin of stress relaxation is investigated. In the Na+/K+ ion exchanged glasses, the K+-foreign cations are introduced in the cages of the Na+-host cations. Due to stress relaxation, the size of the K-O coordination shell in ion-exchanged glasses increases. This increase is achieved by two structural adaptation mechanisms: a contraction of the Na-O, Ca-O and Mg-O coordination shell and both a shortening of the Si-NBO distances and an opening of the Q4 Si-O-Si angles. The greater knowledge developed in this thesis on the effect of parent glass composition on interdiffusion mechanisms and stress relaxation processes provides further insight into practical ways to optimize the mechanical strength of ion-exchanged glasses. / Doctorat en Sciences de l'ingénieur et technologie / info:eu-repo/semantics/nonPublished