WorldCat Identities

CentraleSupélec (2015-....).

Overview
Works: 473 works in 474 publications in 2 languages and 474 library holdings
Roles: Other, Degree grantor, Editor
Publication Timeline
.
Most widely held works by CentraleSupélec (2015-....).
Optimisation des opérations dans les services d'urgence by Karim Ghanes( )

2 editions published in 2016 in English and held by 2 WorldCat member libraries worldwide

Emergency Department (ED) is the service within hospitals responsible for providing unscheduled care to a wide variety of patients over 24 hours a day, 7 days a week. As a result to the existing mismatch between available caring capacity and patients demand, EDs are currently facing a recurrent worldwide problem, namely overcrowding. The objective of this thesis is to develop internal and cost-effective solutions to alleviate overcrowding in EDs and improve their performance, using Operations Research/Operations Management (OR/OM) methods. Such solutions are of great interest for managers. In order to achieve this objective, we address a series of research questions.The first category of research questions include prospective questions about ED Key Performance Indicators (KPIs) and about the diverse factors contributing to overcrowding. We first conduct a detailed literature review on the commonly used KPIs from an OR/OM perspective. The review summarizes the advantages and drawbacks of each KPI and provides several useful insights. In addition, a series of statistical analysis are performed in the purpose of identifying the main influencing factors of performance.The second category consists in resource-related questions that are associated to the dimensioning of ED resource capacity. A realistic ED discrete-event simulation model is thus proposed. The model accounts for the most essential structural and functional characteristics of EDs thanks to a close collaboration with practitioners. Using simulation optimization, we minimize the patient average length of stay (LOS), by integrating a staffing budget constraint and a constraint securing that the most severe incidents will see a doctor within a specified time limit. The obtained results allow us to provide useful insights to managers about the impact of the budget on performance and how investments priorities should be allocated among resources, as well as the effect of combining two different major KPIs. Furthermore, we propose a heuristic for the optimization of the shifts of human resources. The method combines simulation optimization and linear programming.The third category of questions deals with process-related issues. We investigate potential alternative and innovative ED patient flow designs (with fixed budget). Typically in current ED practices, each patient is assigned to a single physician who will be exclusively responsible of him/her during all stages of the process (“Same Patient Same Physician”, SPSP rule). We conduct a survey which confirms that SPSP stands as the standard practice in most EDs worldwide, and that removing SPSP rule is very controversial among practitioners. We next use a complexity-augmented Erlang-R queueing network modeling. We show and quantify the potential benefits of removing the SPSP restriction as a function of the system parameters. For a second process-related issue, we compare the Triage Nurse ordering (TNO) diagnostic tests against the standard procedure, i.e., waiting for the physician to examine the patient and order tests. We demonstrate the efficiency of TNO on ED performance as a function of key parameters, such as triage nurse ability, system load and triage time extension.In summary, this thesis addresses innovative OM research questions for EDs. It provides decision makers with recommendations and tools in order to improve ED performance. It also highlights various avenues for future research related to the optimization of ED operations
Wetting on heterogeneous metal-oxides regular patterned surfaces by a non-reactive liquid metal by Moustapha Diallo( )

1 edition published in 2019 in English and held by 1 WorldCat member library worldwide

In hot-dip galvanizing, steel sheets are protected against corrosion by a thin layer of zinc obtained by immersion in a zinc alloy bath. Before this process, the steel sheets undergo recrystallization annealing to eliminate stresses after cold-rolling. The annealing conditions used reduce the native iron oxide film, which promotes the wettability of the steel surface with liquid zinc. However, new high-strength steels contain significant quantities of addition elements, such as silicon and manganese. These elements diffuse on the surface of the steel sheets during recrystallization annealing and form oxide particles or films by selective external oxidation. If pure iron is well wet with liquid zinc, these oxides are not and their presence on the surface can lead to defects in the final coating.To study the influence of oxide size and their distribution on liquid metal wetting, we studied a non-reactive wetting of liquid lead on a heterogeneous Fe / silica textured surface using the dispensed technique.These surfaces were designed by plasma-assisted chemical vapour deposition followed by a photolithographic process.After impact, the drop extends to its maximum spreading diameter. This is followed by a phase of drop receding. During this, the drop is more or less retained, depending on the silica coverage rate, on the pure iron: stick-slip motion. On surfaces with low silica content, this phenomenon causes a deformation of the drop shape which is more elongated in one direction and sometimes at the division of the drop.We showed that wetting is mainly affected by the surface fraction of silica.Finally, we modelled the different phases of drop spreading on these heterogeneous surfaces. Literature models were reviewed and adapted and macroscopic models of the oscillation of the drop during its spreading were proposed
Model-Based Testing of Timed Distributed Systems : A Constraint-Based Approach for Solving the Oracle Problem by Nassim Benharrat( )

1 edition published in 2018 in English and held by 1 WorldCat member library worldwide

Model-based testing of reactive systems is the process of checking if a System Under Test (SUT) conforms to its model. It consists of handling both test data generation and verdict computation by using models. We specify the behaviour of reactive systems using Timed Input Output Symbolic Transition Systems (TIOSTS) that are timed automata enriched with symbolic mechanisms to handle data. When TIOSTSs are used to test systems with a centralized interface, the user may completely order events occurring at this interface (i.e., inputs sent to the system and outputs produced from it). Interactions between the tester and the SUT are sequences of inputs and outputs named traces, separated by delays in the timed framework, to form so-called timed traces. Distributed systems are collections of communicating local components which interact with their environment at physically distributed interfaces. Interacting with such a distributed system requires exchanging values with it by means of several interfaces in the same testing process. Different events occurring at different interfaces cannot be ordered any more. This thesis focuses on conformance testing for distributed systems where a separate tester is placed at each localized interface and may only observe what happens at this interface. We assume that there is no global clock but only local clocks for each localized interface. The semantics of such systems can be seen as tuples of timed traces. We consider a framework for distributed testing from TIOSTS along with corresponding test hypotheses and a distributed conformance relation called dtioco. Global conformance can be tested in a distributed testing architecture using only local testers without any communication between them. We propose an algorithm to check communication policy for a tuple of timed traces by formulating the verification of message passing in terms of Constraint Satisfaction Problem (CSP). Hence, we were able to implement the computation of test verdicts by orchestrating both localised off-line testing algorithms and the verification of constraints defined by message passing that can be supported by a constraint solver. Lastly, we validated our approach on a real case study of a telecommunications distributed system
Diffuse interface models and adapted numerical schemes for the simulation of subcritical to supercritical flows by Milan Pelletier( )

1 edition published in 2019 in English and held by 1 WorldCat member library worldwide

Au cours de l'utilisation de certains systèmes propulsifs, tels que les moteurs fusées cryotechniques ou les moteurs Diesel, le point de fonctionnement peut varier sur une large plage de pressions. Ces variations de pression peuvent conduire à un changement de régime thermodynamique si la pression critique du fluide est franchie, l'injection initialement diphasique devenant alors transcritique. Ce changement modifie la topologie de l'écoulement, ainsi que la dynamique du mélange, ce qui impacte le comportement de la flamme. L'objectif de cette thèse est de développer une méthodologie originale capable de traiter au sein du même solveur des écoulements sous-critiques ainsi que supercritiques. Pour cela, une extension du solveur AVBP-RG aux écoulements diphasiques sous-critiques est proposée, basée sur des modèles d'interface diffuse. Les développements nécessaires à l'intégration de ces modèles dans le cadre du solveur aux éléments finis sont effectués. Des simulations numériques multidimensionnelles sont ensuite proposées de manière à confronter le modèle à des données exprérimentales, vis-à-vis desquelles un bon accord est observé. Cesrésultats offrent des perspectives encourageantes vers de futures améliorations du modèle et des applications à des configurations industrielles complexes
Unsupervised and weakly supervised deep learning methods for computer vision and medical imaging by Mihir Sahasrabudhe( )

1 edition published in 2020 in English and held by 1 WorldCat member library worldwide

The first two contributions of this thesis (Chapter 2 and 3) are models for unsupervised 2D alignment and learning 3D object surfaces, called Deforming Autoencoders (DAE) and Lifting Autoencoders (LAE). These models are capable of identifying canonical space in order to represent different object properties, for example, appearance in a canonical space, deformation associated with this appearance that maps it to the image space, and for human faces, a 3D model for a face, its facial expression, and the angle of the camera. We further illustrate applications of models to other domains_ alignment of lung MRI images in medical image analysis, and alignment of satellite images for remote sensing imagery. In Chapter 4, we concentrate on a problem in medical image analysis_ diagnosis of lymphocytosis. We propose a convolutional network to encode images of blood smears obtained from a patient, followed by an aggregation operation to gather information from all images in order to represent them in one feature vector which is used to determine the diagnosis. Our results show that the performance of the proposed models is at-par with biologists and can therefore augment their diagnosis
Traitement du Signal d'un LIDAR Doppler scannant dédié à la surveillance aéroportuaire by Alexandre Hallermeyer( )

1 edition published in 2017 in French and held by 1 WorldCat member library worldwide

Un algorithme permettant d'estimer précisément les paramètres des tourbillons de sillage (positions et circulations) en utilisant les données spectrales fournies par un LIDAR a été développé. Il s'articule en 3 grandes étapes : La première permet de détecter la présence de tourbillon et d'en faire une localisation grossière grâce à la méthode des enveloppes de vitesses. La seconde étape a pour but d'affiner l'estimation des positions des tourbillons en utilisant une optimisation du critère des moindres carrés. Cette étape permet également de faire une première estimation de la circulation des tourbillons. La troisième et dernière étape se concentre sur l'estimation des circulations des tourbillons en maximisant le critère de vraisemblance. Les estimations sont de plus en plus fines et se concentrent au fur et à mesure sur les paramètres les plus critiques. La mise au point de cet algorithme a nécessité d'utiliser plusieurs modèles (LIDAR, tourbillons de sillage, atmosphère) et de formuler un certain nombre d'hypothèses et approximations simplificatrices afin d'atteindre un coût calculatoire raisonnable. L'algorithme proposé a ensuite fait l'objet d'une évaluation de performances, l'intérêt étant porté sur la robustesse par rapport aux différents bruits altérant la mesure, en particulier celui lié à la turbulence atmosphérique et par rapport aux erreurs de modèle. Cette évaluation a été menée à la fois sur des données simulées à l'aide de modèles paramétriques simplifiés, et sur des données de simulations aux grandes échelles.Les paramètres instrumentaux du LIDAR constituent de potentiels degrés de liberté pour améliorer les performances de l'estimateur, en particulier pour les grandeurs les plus critiques, c'est-à-dire les valeurs de circulation. Le calcul des performances de l'estimateur nécessitant un coût de calcul non négligeable, il se prête mal à des fins d'optimisation. C'est pourquoi une étude de l'influence des paramètres du LIDAR sur la Borne de Cramér-Rao (BCR) a été menée. Cette étude a permis de mieux comprendre l'influence des paramètres instrumentaux et d'aboutir à une configuration optimale pour la BCR
Etude d'interface entre matrice polymère et renforts à base de carbone, à l'aide d'observations multiéchelles et multimodales en microscopie électronique by Yu Liu( )

1 edition published in 2017 in French and held by 1 WorldCat member library worldwide

This thesis aims to investigate the multiscale (nano-, micro-, and macro-scopic) behavior of the composites based on a fine investigation using the most modern techniques, to understand the interfaces and to quantify them. Two series of reinforcements on a micrometer scale, carbon fibers (CFs) and graphene-based materials, were studied here. To improve the interactions between these nanofillers and the surrounding polymer matrix, two major routes were used in this thesis: the oxidation of the fillers and the grafting of carbon nanotubes on their surface.The study itself was conducted on a microscopic scale on the interfacial strength between CFs and the epoxy matrix, with tensile tests carried out in-situ in the chamber of a double-column FIB-SEM microscope (scanning electron microscope coupled to a focused ion beam). The ion beam was used to mill a thin bond-shaped tensile specimen of composite containing both an epoxy and a CF part. Thetensile stress field was applied using the nanomanipulator and the test was observed both via the ionic and the electronic columns (with two different angles of view) to estimate the strain field, hence the interfacial strength when the failure is observed. A similar experiment was led on a composite with GNPs.Finally, the transmission electron microscopy (TEM) study of the interface region between the epoxy and the graphene-based nanofillers revealed the existence of an interphase and allowed to measure its thickness and give an indication of its nature. For this purpose, an EELS (electron energy-loss spectroscopy) analysis was carried out, making it possible to measure the density of the sample very locally (probe size of the order of a tenth of a nanometer) across or parallelly to an interface. A scenario on the chemical bonding modes between the two media as a function of the surface treatment used makes it possible to explain the nature of the observed interphases
Contribution à l'étude de l'échantillonnage non uniforme dans le domaine de la radio intelligente. by Samba Traore( )

1 edition published in 2015 in French and held by 1 WorldCat member library worldwide

In this work we consider the problem of designing an effective sampling scheme for sparse multi-band signals. Based on previous results on periodic non-uniform sampling (Multi-Coset) and using the well known Non-Uniform Fourier Transform through Bartlett's method for Power Spectral Density estimation, we propose a new sampling scheme named the Dynamic Single Branch Non-uniform Sampler (DSB-NUS). The idea of the proposed scheme is to reduce the average sampling frequency, the number of samples collected, and consequently the power consumption of the Analog to Digital Converter (ADC). In addition to that our proposed method detects the location of the bands in order to adapt the sampling rate. In this thesis, we show through simulation results that compared to existing multi-coset based samplers, our proposed sampler provides superior performance, both in terms of sampling rate and energy consumption. It is notconstrained by the inflexibility of hardware circuitry and is easily reconfigurable. We also show the effect of the false detection of active bands on the average sampling rate of our new adaptive non-uniform sub-Nyquist sampler scheme
Modélisation du transfert thermique par marcheurs browniens dans des milieux hétérogènes by Vincent Gonneau( )

1 edition published in 2020 in French and held by 1 WorldCat member library worldwide

This work deals with the modeling of transient conduction heat transfer by brownian walkers within a heterogenous medium. The medium is described by a 3D voxelised structure. Each walker carries an elementary enthalpy during its movement. This enthalpy motion represents the thermal flux and allows to solve quantitatively the transient thermal conduction. A study demonstrates the importance of the time step chosen in the simulations. An empirical value of this parameter, based on the spatial resolution and the thermal diffusivities of the constituents, has been established to model accurately the conduction at each step of the heat transfer. Several problems related to the behavior of the walkers had to be solved to model two well-known experimental thermal characterization techniques: the hot guarded plate technique and the flash method. An imposed temperature condition is modeled by a reservoir where the number of walkers is regulated. An adiabatic wall condition imposes a specular reflection to the walkers. A stochastic transmission criterion, based on the thermal effusivities, was established to treat the behavior of a walker at voxel-voxel interface between two different constituents. Convective losses at a boundary of the numerical structure are represented by an analogous stochastic transmission criterion involving the effusivity, the convective exchange coefficient and the time step. An imposed thermal flux condition, of any temporal profile, is modeled by an injection of walkers through the boundary. The complex combination of these computing blocks allowed to model transient heat transfer within voxelised heterogeneous structures. The comparison between our results and those issued from more classical approaches allowed to validate the ability of our model to characterize the phonic thermal properties of virtual and real structures. We conclude this PhD work with preliminary activities aiming at the simulation of conduction-radiation coupling at the voxel scale of the structure: the radiative contribution is modeled by an internal power source term within the voxels
Méthodologie pour génération de modèles réduits dynamiques multiphysiques : application aux open rotors by Hadrien Tournaire( )

1 edition published in 2017 in French and held by 1 WorldCat member library worldwide

The mechanical design of a system involves many investigations, notably the validation of its structural behaviour over its operating frequency range. This kind of analysis can be numerically performed using the finite element method, however in such a context, the required accuracy and detail level imply models whose significant sizes lead to time consuming simulations. Moreover, the optimization process of such a system may request numerous validation computations that turn out extremely slow the design process. In the framework of this PhD we target a reduction methodology whose main features are: being compact, dealing with non-linear displacement and recovering the damping effects of the model joint due to the contact-friction phenomenon
Exploitation des mesures électriques en vue de la surveillance et du diagnostic en temps réel des piles à combustible pour application transport automobile by Miassa Taleb( )

1 edition published in 2015 in French and held by 1 WorldCat member library worldwide

In the current global energy context, proton exchange membrane fuel cells represent a promising solution to the future development of a new generation of electrified vehicles, allowing greater autonomy than electrified vehicles using batteries.Nevertheless, the large-scale development of fuel cells remains limited due to some technological locks, such as water management. To enable mass production of fuel cells, such problems must be solved. Several working axes may be envisaged both on the hardware aspects of the fuel cell structure, and from the point of view of control, by developing algorithmic tools for monitoring the operating state of the system to detect any failures, or degradations that may occur.The work of this thesis falls within this second approach and focuses specifically on the identification of drying and drowning phenomena which can appear in a fuel cell, to diagnose any moisture problems leading to yield reduction.The methods developed in this work are based on the monitoring of relevant parameters of the fuel cell model which changes, compared to reference values, are characteristic of the state of the fuel cell hydration.The real-time monitoring of these parameters can highlight the drying and drowning phenomena.Adopted models for this work are based on a representation of the electrical impedance of the fuel cell.Thus, following this approach, the adopted strategy is then based on the development of two electrical models: an integer order model and a fractional order model. It appears that the second model formulation is closer to the physical reality of transport phenomena occurring in the fuel cell. It allows a better representation of the fuel cell behavior in time and frequency domain. Indeed, the analyzes based on experimental results performed using a single fuel cell (100 cm2 active area designed by UBzM company) have validated that the fractional order model, in return for an increase of complexity, allows better reproduce, in the one hand of the fuel cell time-series voltage response (voltage monitoring for a given current profile), on the other hand a better approximation of the measured impedance. Conventional and of fractional order parametric identification methods are then used to extract the model's parameters from time-series experimental data (voltage / current from the battery) or frequency data (impedance spectroscopy).A sensitivity analysis allows then the defining of the most indicative parameters of the drowning and drying phenomena. The evolution of these parameters associated with the voltage and impedance spectrum of the fuel cell are then combined to build a diagnosis strategy of the fuel cell water management
Compositional and Efficient Controller Synthesis for Cyber-Physical Systems by Adnane Saoud( )

1 edition published in 2019 in English and held by 1 WorldCat member library worldwide

This thesis focus on the development of compositional and efficient controller synthesis approaches for cyber-physical systems (CPS). Indeed, while model-based techniques for CPS design have been the subject of a large amount of research in the last decade, scalability of these techniques remains an issue. In his thesis, we contribute to make such approaches more scalable.The focus of the first part is on compositional approaches. A general framework for compositional reasoning using assume-guarantee contracts is proposed. This framework is then combined with symbolic control techniques and applied to a controller synthesis problem for multiperiodic distributed sampled-data systems, where symbolic approaches have been used to synthesize controllers enforcing a given assume-guarantee contract. Then, a new approach to the compositional computation of symbolic abstractions is proposed based on the notion of approximate composition, allowing to deal with heterogeneous abstractions and arbitrary interconnections.The second part is about efficient abstraction and controller synthesis techniques. Two new abstractions schemes have been developed for incrementally stable switched systems. The first approach is based on multirate sampling where we established the existence of an optimal multirate sampling parameter that results in a symbolic model with a minimal number of transitions. The second approach is based on event-based sampling, where the duration of transitions in the symbolic model is determined by some triggering mechanism, which makes it possible to reduce the conservatism with respect to the periodic case. Combination with lazy controller synthesis techniques has been proposed allowing the synthesis at a reduced computational cost. Finally, a new lazy approach has been develop for monotone transition systems and directed safety specifications. Several case studies have been considered in this thesis such as temperature regulation in buildings, control of power converters, vehicle platooning and voltage control in DC micro-grids
Generation and Dynamic Update of Attack Graphs in Cloud Providers Infrastructures by Pernelle Mensah( )

1 edition published in 2019 in English and held by 1 WorldCat member library worldwide

Dans les infrastructures traditionnelles, les graphes d'attaque permettent de brosser un tableau de la sécurité, car ils sont un modèle décrivant les différentes étapes suivies par un attaquant dans le but de compromettre un actif du réseau. Ces graphes peuvent ainsi servir de base à l'évaluation automatisée des risques, en s'appuyant sur l'identification et l'évaluation des actifs essentiels. Cela permet de concevoir des contre-mesures proactives et réactives pour la réduction des risques et peut être utilisé pour la surveillance et le renforcement de la sécurité du réseau.Cette thèse vise à appliquer une approche similaire dans les environnements Cloud, ce qui implique de prendre en compte les nouveaux défis posés par ces infrastructures modernes, la majorité des graphes d'attaque étant conçue pour une application dans des environnements traditionnels. Les nouveaux scénarios d'attaque liés à la virtualisation, ainsi que les propriétés inhérentes du Cloud, à savoir l'élasticité et le caractère dynamique, sont quelques-uns des obstacles à franchir à cette fin.Ainsi, pour atteindre cet objectif, un inventaire complet des vulnérabilités liées à la virtualisation a été effectué, permettant d'inclure cette nouvelle dimension dans les graphes d'attaque existants. Par l'utilisation d'un modèle adapté à l'échelle du Cloud, nous avons pu tirer parti des technologies Cloud et SDN, dans le but de construire des graphes d'attaque et de les maintenir à jour. Des algorithmes capables de faire face aux modifications fréquentes survenant dans les environnements virtualisés ont été conçus et testés à grande échelle sur une plateforme Cloud réelle afin d'évaluer les performances et confirmer la validité des méthodes proposées dans cette thèse pour permettre à l'administrateur de Cloud de disposer d'un graphe d'attaque à jour dans cet environnent
Contribution à l'étude du transport d'énergie dans la matière condensée : phonons, électrons et photons by Benoit Latour( )

1 edition published in 2015 in French and held by 1 WorldCat member library worldwide

Nous avons étudié durant cette thèse les transferts de chaleur mettant en jeu différents types de porteur d'énergie - phonons, électrons et photons - dans des matériaux nanostructurés. A ces échelles, les lois régissant les phénomènes physiques sont différentes des lois macroscopiques. Il est donc nécessaire de développer de nouveaux outils pour étudier ces nouveaux mécanismes. Dans une première partie, nous nous sommes intéressés aux propriétés ondulatoires des phonons thermiques. Nous avons ainsi développé une théorie pour quantifier leur cohérence temporelle et spatiale. Dans une seconde partie, nous nous sommes tournés vers la thermo-plasmonique, c'est-à-dire vers le chauffage par absorption de lumière d'un métal et la redistribution de l'énergie au réseau cristallin par interactions électron/phonon. Dans une dernière partie, nous avons porté notre étude sur la possibilité d'incluer les effets quantiques dans la Dynamique Moléculaire, ouvrant ainsi l'accès aux propriétés thermiques de nanomatériaux aux basses températures
Simulation aux grandes échelles de l'allumage par bougie turbulent et de la propagation de la flamme dans les Moteurs à allumage commandé by Sophie Mouriaux( )

1 edition published in 2016 in English and held by 1 WorldCat member library worldwide

The use of lean equivalence ratios or high EGR rates in spark ignition engines (SIE) enables to optimize CO2 and NOx emissions; however too important dilution rates leads to increased cycle-to-cycle variability. These latter are mostly due to the ignition phase, which becomes critical when dilution rates are important and requires high ignition energy. The ECFM-LES model currently used in IFPEN, which is based on the flame surface density concept, is not sufficient to describe ignition in these critical conditions. The TF-LES approach was chosen in this study, principally because it directly resolved chemistry and can thus model ignition via a local raise of the temperature. The present work defines and evaluates a simulation strategy for TF-LES in SIE configurations, that enables a fine prediction of critical ignitions and of the turbulent flame propagation.In the first part, DNS of turbulent ignition were performed. The ignition phase was modeled using a thermal energy deposit (ED model, Lacaze et al.). Simulations reproduced the ignition experiments of Cardin et al. who determined the minimum ignition energy (MIE) of lean premixed methane/air mixtures, for different turbulence characteristics. The main purpose of the study was to determine the numerical and physical model parameters, which enable to reproduce Cardin et al. experiments. Two types of kinetic schemes were evaluated: a simplified kinetic scheme and an analytical kinetic scheme (ARC), that can predict both the auto-ignition delays and the laminar flame speed, while keeping affordable CPU times. Results analysis enabled to define ignition criteria and to highlight the differences in terms of ignition prediction using the two kinetic schemes. Results also demonstrated that the chosen approach could recover correct levels of ignition energy for laminar and low Karlovitz number cases (Ka<10). For higher Karlovitz number cases, the ED model was found to be insufficient to predict the ignition and a finer description of the energy deposit is required.In the second part, a dynamic wrinkling model (Wang et al., 2012) was studied to describe the out-of-equilibrium behavior of the flame during the propagation phase. Studies on laminar spherical flames were first performed, to assess the laminar degeneration of the model. Then, as first tests in an engine configuration have revealed incompatibilities of the model, modifications were proposed. The modified dynamic model was finally tested in the ICAMDAC engine configuration. Results of the simulations were compared against previous results of Robert et al. obtained with the ECFM-LES model using a transport equation for the flame surface density that can describe the out-of-equilibrium wrinkling of the flame. Results obtained with the dynamic model are in very good agreement with the ones of Robert et al., thus demonstrating the ability of the dynamic model to predict out-of-equilibrium values in the engine configuration. Besides, the dynamic model self-adapts to the turbulence conditions, hence does not require any model parameter adjustment, as is it the case for models based on the flame surface density transport equation
Injection and combustion dynamics in swirled spray flames and azimuthal coupling in annular combustors by Guillaume Vignat( )

1 edition published in 2020 in English and held by 1 WorldCat member library worldwide

The deployment of low-emission technologies in aero-engines does not only require that new designs produce reduced amounts of pollutants, but also that their dynamical behavior (ignition, blow-off, and combustion instabilities) be compatible with the high safety standards prevailing in aeronautics. This research is focused on the latter aspect of combustor design. A transparent annular combustor equipped with 16 swirled spray injectors is used to represent at the laboratory scale the combustion chamber of a jet engine. This system is used in conjunction with a single sector rig to investigate dynamical issues.Theory, experimentation and large eddy simulation are combined to examine a range of items pertaining to the injector dynamics, flow structure, swirl number determination, spray characteristics, and coupling between injector flow and acoustic field. A database of injectors is introduced to investigate the impact of injection parameters on combustion dynamics. These injectors are examined under steady and unsteady conditions by combining laser diagnostics and high-fidelity simulations which allows the characterization of spray-specific behaviors of relevance to the dynamics of injection systems. One important result is that the presence of a liquid film formed on the wall of the injection unit gives rise to a multi-modal distribution of droplet velocities. A novel method is introduced to examine the space-time behavior of the flow and flame of a swirling injector submitted to axial modulations. A detailed investigation of the process by which convective perturbations couple with the acoustic field allows to examine the time lags that control combustion instability and sort out the respective roles of convection and droplet spray evaporation. High speed tomography relying on SnO2 particles provides major results on the Precessing Vortex Core structure and show its behavior under acoustic forcing. The impact of injector head loss on combustion instabilities is examined using several injection systems with similar levels of swirl. The head loss is shown to play a major role in the coupling between the flame and the upstream plenum. The previous results obtained in a single sector rig are used to guide experiments on the annular combustor. The focus is placed on high amplitude combustion oscillations coupled by a standing azimuthal mode inducing flame blow-off near the pressure nodal line. The deformation of the acoustic distribution is tracked using a novel expansion on azimuthal harmonics allowing the determination of the critical conditions leading to this phenomenon. New results are also presented about the transient dynamics of an injector during ignition and about lean blow out with a proof of concept extension of the LBO limit by nanosecond plasma discharges
Apprentissage de graphes structuré et parcimonieux dans des données de haute dimension avec applications à l'imagerie cérébrale by Eugene Belilovsky( )

1 edition published in 2018 in French and held by 1 WorldCat member library worldwide

This dissertation presents novel structured sparse learning methods on graphs that address commonly found problems in the analysis of neuroimaging data as well as other high dimensional data with few samples. The first part of the thesis proposes convex relaxations of discrete and combinatorial penalties involving sparsity and bounded total variation on a graph as well as bounded `2 norm. These are developed with the aim of learning an interpretable predictive linear model and we demonstrate their effectiveness on neuroimaging data as well as a sparse image recovery problem.The subsequent parts of the thesis considers structure discovery of undirected graphical models from few observational data. In particular we focus on invoking sparsity and other structured assumptions in Gaussian Graphical Models (GGMs). To this end we make two contributions. We show an approach to identify differences in Gaussian Graphical Models (GGMs) known to have similar structure. We derive the distribution of parameter differences under a joint penalty when parameters are known to be sparse in the difference. We then show how this approach can be used to obtain confidence intervals on edge differences in GGMs. We then introduce a novel learning based approach to the problem structure discovery of undirected graphical models from observational data. We demonstrate how neural networks can be used to learn effective estimators for this problem. This is empirically shown to be flexible and efficient alternatives to existing techniques
Développement d'une méthodologie pour l'évaluation de l'exposition réelle des personnes aux champs électromagnétiques by Ashish Rojatkar( )

1 edition published in 2016 in English and held by 1 WorldCat member library worldwide

The work presented in the thesis is directed towards addressing the requirement for determining the radio frequency (RF) exposure due to mobile phones under typical usage/ real-life scenarios and also to develop a method to predict and compare mobile phones for their real-life RF exposure. The mobile phones are characterized for their specific absorption rate (SAR) and for transmit and receive performance given by the over-the-air (OTA) characterization. Using the SAR and the total radiated power (TRP) characterization, an exposure index referred to as the SAROTA index was previously proposed to predict the real-life exposure due to mobile phones which would also serve as a metric to compare individual phones. In order to experimentally determine the real-life RF exposure, various software modified phones (SMP) are utilized for the study. These phones contain an embedded software capable of recording the network parameters. The study is undertaken in the following order: (a) Characterization of the available tools and resources for performing targeted measurements/experiments, (b) identifying the important radio resource parameters and metrics to perform the targeted measurements, (c) investigation of the actual implementation of the power control mechanism in a live network for various received signal level and received quality environments, (d) investigating the correlation of the over-the-air performance of the mobile phones and the extent of actual power control realization, (e) comparing the actual exposure and the real-life exposure as predicted by the SAROTA index. Based on the logistical and technical challenges encountered, the experiments were restricted to indoor environments to enable repeatability. During the first phase of the study, the stability of the indoor environment was evaluated. During the second phase, the influence of hand phantom on the SAR and TRP of the mobile phones and the capability of the SAROTA index to predict the exposure was investigated. Further developing on the insights from the hand phantom experiments, in the third phase, a set of identical software modified phones were externally modified to alter the TRP performance and the methodology to determine the real-life exposure and also verify the capability of the SAROTA index to predict the exposure levels was investigated. The experiments demonstrate that the SAROTA index is capable of predicting the real-life exposure and comparing the mobile phones
Contribution au développement d'une méthode analytique pour l'identification des dépendances et la propagation des changements d'ingénierie en re-conception de produits by Mahmoud Masmoudi( )

1 edition published in 2017 in French and held by 1 WorldCat member library worldwide

La gestion des changements d'ingénierie, appelée aussi ingénierie du changement, est un domaine de recherche qui s'intéresse à l'étude des changements et des modifications apportés par les concepteurs industriels sur un produit durant l'une des phases de son cycle de vie. Ces changements peuvent exister dans le cadre de renouvellement d'un produit dans un but d'amélioration ou de reconception. Dans la littérature, plusieurs méthodes et outils ont été mis en place pour évaluer l'impact d'un changement en utilisant les dépendances existantes entre les éléments d'un produit, prédire de façon efficace la propagation du changement et évaluer ensuite ses conséquences possibles. Mais aucun d'entre eux n'a tenté de caractériser la relation de dépendance sous la forme d'une fonction polynomiale, en plus des caractérisations classiques par valeurs qualitatives et quantitatives. Nous proposons donc, une nouvelle méthode pour générer les variations possibles d'un changement, afin de modéliser ces variations par paire sous formule analytique, d'étendre les variations par paire aux variations de plusieurs variables et de vérifier formellement la cohérence des résultats obtenus. Visuellement, un graphe de dépendance est construit oú les noeuds représentent les éléments de produit et les arcs portent les fonctions qui les relient. Nos résultats montrent que cette caractérisation fonctionnelle de la dépendance est instructive et permet une propagation de changement et une évaluation efficace. Un modèle géométrique 2D d'un vélo est utilisé pour illustrer l'approche proposée
Leveraging User-Generated Content for Enhancing and Personalizing News Recommendation. by Youssef Meguebli( )

1 edition published in 2015 in English and held by 1 WorldCat member library worldwide

La motivation principale de cette thèse est de proposer un système de recommandation personnalisé pour les plateformes d'informations. Pour cela, nous avons démontré que les opinions peuvent constituer un descripteur efficace pour améliorer la qualité de la recommandation. Au cours de cette thèse, nous avons abordé ce problème en proposant trois contributions principales. Tout d'abord, nous avons proposé un modèle de profil qui décrit avec précision les intérêts des utilisateurs ainsi que le contenu des articles de presse. Le modèle de profil proposé repose sur trois éléments : les entités nommées, les aspects et les sentiments. Nous avons testé notre modèle de profil sur les trois applications différentes que sont l'identification des orientations politiques des utilisateurs, la recommandation personnalisée des articles de presse et enfin la diversification de la liste des articles recommandés. Deuxièmement, nous avons proposé une approche de classement des opinions permettant de filtrer et sélectionner seulement les opinions pertinentes. Pour cela, nous avons utilisé une variation de la technique de PageRank pour définir le score de chaque opinion. Les résultats montrent que notre approche surpasse deux approches récemment proposées pour le classement des opinions. Troisièmement, nous avons étudié différentes façons d'enrichir le contenu des articles de presse par les opinions : par toutes les opinions, par seulement le topk des opinions, et enfin par un ensemble d'opinions diversifiées. Les résultats montrent que l'enrichissement des contenus des articles de presse
 
moreShow More Titles
fewerShow Fewer Titles
Audience Level
0
Audience Level
1
  Kids General Special  
Audience level: 0.99 (from 0.94 for Optimisati ... to 0.99 for Optimisati ...)

Alternative Names
Ecole Centrale Paris

Supélec

Languages
English (12)

French (9)