Search

International audience; Multiple days assessments are frequent for the evaluation of candidates to living kidney donation, combined with an early GFR estimation (eGFR). Living kidney donation is questionable when eGFR is 55 years) in a population of 1825 French living kidney donors with a median follow-up of 5.9 years. In donors younger than 45, postdonation eGFR, absolute- and relative-eGFR variation were not different between the three groups. For older donors, postdonation eGFR was higher in S90 than in S80 or Sage but other comparators were identical. Postdonation eGFR slope was comparable between all groups. Our results are in favour of in-depth screening for all candidates to donation with a normal eGFR for age.

International audience; This paper proposes the application of the Finite Element Tearing and Interconnecting method (FETI-2LM) a Domain Decompositon Method (DDM) [1], [2] for the full-wave simulation of large antenna array (few thousands of radiating elements). The proposed methodology makes it possible to efficiently calculate in a post-processing step the antenna patterns for steering directions within the desired field of view while taking into account the active reflection coefficients. The methodology is based on an efficient extraction of the embedded radiating elements far field radiation patterns as well as the Generalized Scattering Matrix (GSM) [3] of the array by implementing massively parallel computers. The beauty of the GSM is its capability to handle the active S-parameters independently of the steering direction. This permits to generate the angular-dependent active reflection coefficients by post-processing. This implementation is a step forward to provide us with a tool that permits to evaluate by simulation large antenna array while taking into account all the phenomena and interactions that may degrade the key performances that are required at a higher level, to specify a global system before fabrication.

In this paper, a new framework to design high-order approximations in the context of node-centered FiniteVolumes on simplicial meshes is proposed. The major novelty of this method is that it relies on very simple andcompact differential operators, which is a critical point to achieve good performances in the High-PerformanceComputing (HPC) context. This method is based on deconvolution between nodal and volume-average valueswhich can be conducted to any order. The interest of the new method is illustrated through three differentapplications : mesh-to-mesh interpolation, levelset curvature computation and numerical scheme for convection.Higher-order can also be achieved within the present framework by introducing high-rank tensors. Althoughthese tensors feature many symmetries, their manipulation can quickly become an overwhelming task. For thisreason and without loss of generality, the present papers is limited to third-order expansion. This method,although tightly connected to thek-exact schemestheory, does not rely on successive corrections: the high-orderproperty is obtained in a single operation, which makes them more attractive in terms of performances.

International audience; View Video Presentation: https://doi.org/10.2514/6.2022-1073.vid Predicting the laminar to turbulent transition is an important aspect of computational fluid dynamics because of its impact on skin friction. The traditional methods of transition prediction do not make it possible to consider configurations where the boundary layer develops in the presence of surface defects (bumps, steps, gaps, etc.). A neural network approach is used in this paper, based on an extensive database of boundary layer stability studies in the presence of gap-like surface defects. These studies consists on Linearized Navier-Stokes (LNS) calculations and provide information on the effect of surface irregularity geometry and aerodynamic conditions on the transition to turbulence. The physical and geometrical parameters characteristic of the defect and the flow are then provided to a neural network whose outputs inform about the effect of a given gap on the transition through the ΔN method.

Tensor network states are used extensively as a mathematically convenient description of physically relevant states of many-body quantum systems. Those built on regular lattices, i.e. matrix product states (MPS) in dimension 1 and projected entangled pair states (PEPS) in dimension 2 or higher, are of particular interest in condensed matter physics. The general goal of this work is to characterize which features of MPS and PEPS are generic and which are, on the contrary, exceptional. This problem can be rephrased as follows: given an MPS or PEPS sampled at random, what are the features that it displays with either high or low probability? One property which we are particularly interested in is that of having either rapidly decaying or long-range correlations. In a nutshell, our main result is that translation-invariant MPS and PEPS typically exhibit exponential decay of correlations at a high rate. We have two distinct ways of getting to this conclusion, depending on the dimensional regime under consideration. Both yield intermediate results which are of independent interest, namely: the parent Hamiltonian and the transfer operator of such MPS and PEPS typically have a large spectral gap. In all these statements, our aim is to get a quantitative estimate of the considered quantity (generic correlation length or spectral gap), which has the best possible dependency on the physical and bond dimensions of the random MPS or PEPS., 53 pages, 7 figures. v2: a bug in Section 4 has been fixed and a few changes have been made to the presentation. v3: published version

International audience; The "extrapolated range" of electrons is a very practical physical parameter widely used in many applications. The commonly used analytical expressions are given classically for energies down to ~1 keV. Low energy Monte Carlo simulations have been performed to extract this parameter in various materials down to several tens of eV. These simulations have been used to parameter both, transmission probability and projected range analytical expressions. These empirical equations have been extended to 11 monoatomic elements (C, Be, Al, Si, Ti, Ni, Cu, Ge, Ag, Fe and W) and for energies ranging from ~10 eV up to some tens of MeV.; Le « parcours extrapolé » des électrons est un paramètre physique très pratique largement utilisé dans de nombreuses applications. Les expressions analytiques couramment utilisées sont données classiquement pour des énergies jusqu'à ~1 keV. Des simulations Monte Carlo basse énergie ont été réalisées pour extraire ce paramètre dans divers matériaux jusqu'à plusieurs dizaines d'eV. Ces simulations ont été utilisées pour paramétrer à la fois la probabilité de transmission et les expressions analytiques de la portée projetée. Ces équations empiriques ont été étendues à 11 éléments monoatomiques (C, Be, Al, Si, Ti, Ni, Cu, Ge, Ag, Fe et W) et pour des énergies allant de ~10 eV jusqu'à quelques dizaines de MeV.

International audience; Corona discharges have been designated as aeronautical VHF communications disturber. In order to better understand the discharges, the coupling phenomena involved and to propose an adequate protection for VHF systems, both laboratory testing and experiments on a 25 m high pylon are carried out here.Main characteristics of Corona pulses have been retrieved and analyzed wrt the experimental parameters. Leader propagation phenomena have also been identified.

International audience; The total ionizing dose (TID) is a well-known reliability issue for integrated circuits (ICs). It consists in changes of MOSFETs dc characteristics following the irradiation of devices in a given radiative environment, such as space environment. TID effects have been investigated from an experimental and theoretical standpoint, and the compact modeling of TID effects has been included into BULK transistor compact models using PSP formalism in the case of uniform and nonuniform energetic distributions of interface traps. However, TID effects are still not included in fully depleted SOI (FDSOI) standard models. In this paper, TID effects modeling is proposed and included in Leti-UTSOI compact model for FDSOI MOSFETs. This model takes both uniform and nonuniform energetic distributions of both gate oxide/silicon and silicon/buried oxide (BOX) interface traps into account, as well as the contribution of oxide/BOX-trapped charges. Validation of the model has been done through TCAD simulations, and we used this model to extract TID parameters from the experimental data.

Light Detection And Ranging sensors (lidar) are key to autonomous driving, but their data is severely impacted by weather events (rain, fog, snow). To increase the safety and availability of self-driving vehicles, the analysis of the phenomena consequences at stake is necessary. This paper presents experiments performed in a climatic chamber with lidars of different technologies (spinning, Risley prisms, micro-motion and MEMS) that are compared in various artificial rain and fog conditions. A specific target with calibrated reflectance is used to make a first quantitative analysis. We observe different results depending on the sensors, valuable multi-echo information, and unexpected behaviors in the analysis with artificial rain are seen where higher rain rates do not necessarily mean higher degradations on lidar data.

Immersed boundary methods (IBMs) are an interesting alternative to the usual body-fitted mesh approach when dealing with complex geometries as they allow simpler mesh generation. The volume penalization method (an IBM) is commonly used for incompressible and compressible viscous flows but only one application to compressible inviscid flows can be found, which uses the characteristic-based volume penalization (CBVP) method. This approach penalizes the Euler equations to enforce a no-penetration velocity and an adiabatic wall while accounting for wall curvature. A new penalization method based on the CBVP is proposed to impose the conservation of entropy and total enthalpy in the normal direction to the wall instead of the classical adiabatic condition. The two approaches are compared and numerically tested on several cases: weakly compressible flow around a circular cylinder, subsonic flow around a NACA0012 airfoil and flow around a challenging high curvature ice horn. The new method is found to be more accurate than the CBVP on coarser meshes and better at retrieving attached flows for curved geometries. The paper concludes that the proposed method is suitable for general aerospace applications and beneficial for icing simulations which can exhibit highly curved geometries.

The Secondary Electron Emission (SEE) process plays an important role in the performance of various devices. Mitigating the multipactor phenomenon that may occur in radio-frequency components is a concern in many fields such as space technologies or electron microscopy. SEE is also a concern in the accelerator physics community, where the beam lines stability can strongly be affected by this phenomenon*,**. In that scope, the escaped depth and thus the range of emitted electrons is of great interest. Our goal, by means of simulations is to provide a better knowledge of SEE. We have developed a Monte Carlo electron transport code for low energy electrons [~eV, ~10keV], that is part of the Dec. 2020 release of GEANT4***. It has been used to study the practical range of low energy electrons. Our goal is to formulate, below ~10 keV, an analytic range vs. energy expression, and to relate it to fundamental physcial parameters such as the mean free paths of electrons in matter. The goal is to provide simple practical extrapolated range formula that can help to understand SEE phenomenon., Proceedings of the 12th International Particle Accelerator Conference, IPAC2021, Campinas, SP, Brazil

International audience; Monitoring plant metal uptake is essential for assessing the ecological risks of contaminated sites. While traditional techniques used to achieve this are destructive, Visible Near-Infrared (VNIR) reflectance spectroscopy represents a good alternative to monitor pollution remotely. Based on previous work, this study proposes a methodology for mapping the content of several metals in leaves (Cr, Cu, Ni and Zn) under realistic field conditions and from airborne imaging. For this purpose, the reflectance of Rubus fruticosus L., a pioneer species of industrial brownfields, was linked to leaf metal contents using optimized normalized vegetation indices. High correlations were found between the vegetation indices exploiting pigment-related wavelengths and leaf metal contents (r ≤ − 0.76 for Cr, Cu and Ni, and r ≥ 0.87 for Zn). This allowed predicting the metal contents with good accuracy in the field and on the image, especially Cu and Zn (r ≥ 0.84 and RPD ≥ 2.06). The same indices were applied over the entire study site to map the metal contents at very high spatial resolution. This study demonstrates the potential of remote sensing for assessing metal uptake by plants, opening perspectives of application in risk assessment and phytoextraction monitoring in the context of trace metal pollution.; La surveillance de l'absorption des métaux par les plantes est essentielle pour évaluer les risques écologiques des sites contaminés. Alors que les techniques traditionnelles utilisées pour y parvenir sont destructrices, la spectroscopie en réflectance visible - proche infrarouge représente une bonne alternative pour surveiller la pollution à distance. Sur la base de travaux antérieurs, cette étude propose une méthodologie pour cartographier la teneur de plusieurs métaux dans les feuilles (Cr, Cu, Ni et Zn) dans des conditions de terrain réalistes et à partir d'une imagerie aérienne. À cette fin, la réflectance de Rubus fruticosus L., une espèce pionnière des friches industrielles, a été liée aux teneurs en métaux des feuilles en utilisant des indices de végétation normalisés et optimisés. Des corrélations élevées ont été trouvées entre les indices de végétation exploitant les longueurs d'onde liées aux pigments et les teneurs en métaux des feuilles (r ≤ - 0,76 pour Cr, Cu et Ni, et r ≥ 0,87 pour Zn). Cela a permis de prédire les teneurs en métaux avec une bonne précision sur le terrain et sur l'image, en particulier Cu et Zn (r ≥ 0,84 et RPD ≥ 2,06). Les mêmes indices ont été appliqués sur l'ensemble du site d'étude pour cartographier les teneurs en métaux à une très haute résolution spatiale. Cette étude démontre le potentiel de la télédétection pour évaluer l'absorption de métaux par les plantes, ouvrant des perspectives d'application dans l'évaluation des risques et la surveillance de la phytoextraction dans le contexte de la pollution par les métaux traces.

In France, Article 7 of the "Charte nationale de déontologie des métiers de la recherche (2015)", currently signed by more than 60 institutions, states that "Ethical rules must be integrated into the training programs, in particular within the master's and doctoral programs, and their learning must be considered as participating in the mastery of the specific research field." In addition, the ministerial decree of May 2016 setting the National framework for training and the modalities leading to the delivery of the National doctoral degree, stipulates that doctoral schools "ensure that each doctoral student receives training in research ethics and research integrity."In this context, it seemed relevant a few years later to draw up a sufficiently exhaustive map of training and specific programs in research integrity and research ethics, in order to better identify them. To this end, a working group involving members of the French network of research integrity officers and of the French doctoral schools network, together with the French office for research integrity has designed an online survey that was sent to all the French higher education and research institutions, research integrity officers network, and doctoral schools network in late 2019. One hundred and twenty responses corresponding to 109 separate institutions were collected.Out of these 109 respondent institutions, 72% have implemented training in research integrity and ethics for doctoral students; this figure rises to 94% if the scope is limited to the 51 respondent universities. These results confirm that these initiatives are now widely deployed. Moreover, half of the 109 responding institutions offer training for staff other than doctoral students. The characteristics of the doctoral training courses offered by 48 of the responding universities in terms of format, language, disciplinary mix, subjects covered, validation system and evaluation method are briefly presented in this document. A more detailed report will soon be available on HAL.Beyond the numerical results confirming the actual implementation of training in research integrity and ethics at the national level since 2016, this survey allows us to draw up a more qualitative representation, taking into account both common modalities and local specificities, and allowing us to suggest avenues for further thoughts in this area., En France, l'article 7 de la Charte nationale de déontologie des métiers de la recherche (2015), actuellement signée par plus de soixante établissements, stipule que « Les règles déontologiques doivent être intégrées aux cursus de formation, en particulier au sein des cursus de master et de doctorat, et leur apprentissage doit être considéré comme participant à la maîtrise du domaine spécifique de recherche. » De plus, l’arrêté ministériel de mai 2016 fixant le cadre national de la formation et les modalités conduisant à la délivrance du diplôme national de doctorat, stipule que les écoles doctorales ou les collèges doctoraux « veillent à ce que chaque doctorant reçoive une formation à l'éthique de la recherche et à l'intégrité scientifique ».Dans ce cadre, il a semblé pertinent quelques années plus tard de dresser une carte suffisamment exhaustive des formations et des programmes spécifiques en matière d'intégrité scientifique et d’éthique de la recherche, afin de mieux les identifier. C'est ainsi que nous avons constitué un groupe de travail associant des membres du RESINT (RESeau des référents INTégrité scientifique), du RNCD (Réseau National des Collèges Doctoraux) et de l'OFIS (Office Français de l'Intégrité Scientifique), au sein duquel nous avons conçu une enquête en ligne qui a été envoyée à tous les établissements d'enseignement supérieur et de recherche, ainsi qu'à la liste de diffusion du RESINT et aux membres du RNCD. Cent vingt réponses correspondant à 109 établissements distincts ont été recueillies.Sur ces 109 établissements répondants, 72% ont mis en place une formation à l’intégrité scientifique et à l’éthique de la recherche pour les doctorants ; et ce chiffre atteint 94% si le périmètre est limité aux 51 universités répondantes. Ceci confirme que ces initiatives sont désormais largement déployées. Pour les formations destinées aux publics autres que les doctorants, il en existe dans la moitié des 109 établissements répondants. Les caractéristiques des formations doctorales proposées par 48 des universités répondantes, concernant leur format, la langue utilisée, la pluridisciplinarité, les sujets abordés, le système de validation ou la méthode d'évaluation, sont brièvement présentées dans le présent document. Un rapport plus détaillé sera prochainement disponible sur HAL.Au-delà des résultats chiffrés confirmant la mise en place effective d’une formation à l’intégrité scientifique et à l’éthique de la recherche en France depuis 2016, cette enquête permet de dresser une représentation qualitative plus fine, rendant compte à la fois de modalités communes et de spécificités locales, et permettant de suggérer des pistes de réflexion.

International audience; The current work presents a study on spray dynamics in aeronautical injectors. More particularly, the experimental characterization of a liquid jet in an oscillating gaseous cross flow (LJIOGCF) atomization is presented and compared to numerical simulations. The experimental setup consists of a water jet transversally injected into an oscillating subsonic air flow at ambient conditions. In a first step, a cross-comparison is performed in order to validate the numerical model and the tools used to quantitatively analyze the experimental data in the case of steady air flows. In a second step, the air flow is submitted to longitudinal acoustic waves using a pneumatic loudspeaker. A detailed database, representative of the actual dynamics of LJIOGCF configurations, is obtained using both experiments and numerical simulations. Visualizations with a back lighting approach are employed to characterize the liquid jet close to the injection location while phase doppler anemometry is used to determine the characteristics of the crossflow and the spray in terms of droplet size, velocity and concentration. Phase-averaging is performed to characterize the response of the liquid jet, the air velocity field and the spray oscillations during the excitation cycle. The numerical simulation relies on a multi-scale large eddy simulation approach. This method couples a multi-fluid solver for the liquid jet main body with a dispersed phase solver for the atomized spray. The acoustic perturbation is imposed as a fluctuating air inflow condition. The experimental results show that the acoustic forcing induces a flapping motion of the liquid jet. As a consequence, velocity and concentration waves travelling downstream the liquid jet appear. Coupling phenomena between the crossflow, the atomization of the liquid jet and the transport of droplets are observed, revealing different wave transport velocities. It further appears that spray dynamics are both driven by liquid column and crossflow oscillations. The numerical simulation is able to capture the global flapping dynamics of the liquid jet’s main body. More quantitative comparisons show a very good agreement between simulation and experiments regarding the jet trajectories over the entire excitation cycle. Numerical droplet average velocities are also in good agreement with experiments. Finally, the numerical simulation partly reproduces the coupling between acoustics and spray dynamics observed in the experiments.

International audience; Degradation of InGaAsN pinsubcell under 1MeV electron irradiation was studied by characterizing solar cells and dilute nitride bulklayers before and after irradiation. The measurements show that these cells retain more than 94% of their original photocurrent after an 1015cm-2 electron-irradiation. Moreover, no significant degradation of the optoelectronic properties is observed after irradiation

International audience; A bi-static short-range elastic backscatter micro-lidar, named Colibri, has been developed for quantitative aerosol profiling with high range and temporal resolution within the first hundred meters. The geometric (i.e., overlap) and radiometric (i.e., lidar constant) calibrations were performed along with dark current and background noise characterizations. Results of a measurement campaign have demonstrated the capability of our system to characterize aerosol plumes with high range-resolution (

Models used for studies of atmospheric reentry vehicles are mainly developed with data from wind tunnel test campaigns. Despite such data are highly instructive, they do not reflect all the conditions encountered during reentry. Measurements carried out during atmospheric flights can be used to deal with this lack of data. Within this context, the present work consisted in an aerothermodynamic analysis of the atmospheric reentry of a reusable lifting body from a low Earth orbit. This study used thermocouple and infrared thermographic temperature measurements carried out during the atmospheric reentry of the IXV (Intermediate eXperimental Vehicle) spaceplane. This vehicle was an automatic lifting body designed and launched by the European Space Agency. The aerothermodynamic analysis involved two approaches. The first one is based on inverse heat conduction problems: the purpose was to estimate the unknown convecto-diffusive heat flux at the windward surface of IXV using temperature measurements on the leeward surface of a flap or inside the thermal protection systems. However, estimates from inverse analyses do not bring information about physical phenomena in the shock layer which could generate convecto-diffusive heat flux at the surface. This point justifies the work from the second approach, based on 1D and 3D Navier-Stokes simulations of the hypersonic flow around IXV. The purposes of these simulations were to analyse and identify the physical phenomena which allowed the convecto-diffusive heat flux from inverse problems to be reproduced. This association between inverse methods and Navier-Stokes simulations was previously tested on data from the orbiter Columbia during the STS-2 atmospheric reentry. The long-term objective is to improve aerothermodynamic databases and tools for the preparation of Europe’s future automated reentry systems.; Les modèles développés pour l'étude des rentrées atmosphériques sont essentiellement basés sur des mesures en soufflerie. Bien que ces dernières soient riches en enseignements, elles ne sont pas représentatives de l'ensemble des conditions rencontrées lors d'une rentrée atmosphérique. Pour combler ce manque de données il est nécessaire de traiter des mesures réalisées en vol. Les présents travaux de thèse s'inscrivent dans ce contexte et proposent une analyse aérothermodynamique de la rentrée atmosphérique d'un véhicule réutilisable, de type corps portant, depuis une orbite basse terrestre. Plus précisément cette analyse traite du flux de chaleur convecto-diffusif à l'intrados du véhicule. Pour remplir cet objectif l'étude s'appuie principalement sur les mesures de température effectuées au cours de la rentrée atmosphérique de l’IXV (Intermediate eXperimental Vehicle), le premier véhicule autonome de l'Agence Spatiale Européenne piloté sur une trajectoire planante. Les mesures de température ont été réalisées dans les protections thermiques de l’intrados et à la surface extrados des gouvernes à l’aide de thermocouples et d'une caméra infrarouge. L'analyse aérothermodynamique s'articule autour de deux approches complémentaires. La première est une estimation par méthodes inverses 1D et 3D du flux de chaleur convecto-diffusif à la surface intrados de l'IXV à partir des mesures de température. Cependant ces estimations ne permettent pas à elles seules de remonter jusqu'à l'historique des phénomènes physico-chimiques dans la couche de choc. Ce point justifie la présence de la deuxième approche basée sur des simulations Navier-Stokes 1D et 3D de l'écoulement. Ces simulations ont été menées afin d'identifier et d'analyser de possibles scénarios des phénomènes physico-chimiques permettant de reproduire les niveaux du flux de chaleur estimés par méthode inverse. Cette méthode, basée sur les résultats d'une approche par méthode inverse et d'une approche par simulations de l'écoulement, a préalablement été testée sur les données des premiers vols de la navette spatiale américaine Columbia. Finalement, ces travaux de thèse présentent une lecture approfondie des mesures réalisées au cours du vol de l'IXV. Sur le long terme ces données permettront de compléter les bases de données aérothermodynamiques et d'améliorer les outils de dimensionnement des véhicules de rentrée atmosphérique.

International audience; This paper investigates the time-local discretization, using Gaussian quadrature, of a class of diffusive operators that includes fractional operators, for application in fractional differential equations and related eigenvalue problems. A discretization based on the Gauss-Legendre quadrature rule is analyzed both theoretically and numerically. Numerical comparisons with two existing methods, one optimization-based and one quadrature-based, highlight its applicability. In addition, it is shown, on the example of a fractional delay differential equation, that quadrature-based discretization methods are spectrally correct, i.e. that they yield an unpolluted and convergent approximation of the essential spectrum linked to the fractional derivative, by contrast with optimization based methods that can yield polluted spectra whose convergence is difficult to assess.

International audience; Electron radiation belts are regions surrounding Earth, filled with highly energetic electrons and overlapping the majority of satellite orbits. Their multi-scale and rapidly evolving dynamics are modelled by the mean of a diffusion equation involving a highly anisotropic and inhomogeneous diffusion tensor. Finite difference based methods have been the preferred method of discretization in physical codes, starting from ONERA’s Salammbô-Electron model, the pioneering 3-dimensional code in the radiation belts community. This choice however does not prevent several numerically induced constraints impacting the reliability of the code as well as its computational cost. Thus in this paper we present the outcome of our investigation to improve Salammbô’s numerical core. In particular, we present our special diffusion frame and its numerically induced limitations on our finite difference based scheme. Then we test potential alternative finite volume schemes with physically relevant properties and we finally highlight with a real case simulation the contribution of the positivity preserving scheme to evaluate the impact of cross diffusion.

A series of benchmarks based on the physical situation of "phase inversion" between two immiscible liquids is presented. These benchmarks aim at progressing toward the direct numerical simulation of two-phase flows. Several CFD codes developed in French laboratories and using either Volume-of-Fluid or Level-Set interface tracking methods are used to provide physical solutions of the benchmarks, convergence studies and code comparisons. Two typical configurations are retained, with integral scale Reynolds numbers of 13.700 and 433.000, respectively. The physics of the problem are probed through macroscopic quantities such as potential and kinetic energies, or enstrophy. In addition, scaling laws for the temporal decay of the kinetic energy are derived to check the physical relevance of the simulations. Finally the droplet size distribution is probed. Additional test problems are also reported to estimate the influence of viscous effects in the vicinity of the interface., 35 pages, 35 figures, third revision submitted to J. Comput. Phys

International audience; For the last 30 years, experimental and modeling studies have been carried out on fluidized bed reactors from laboratory up to industrial scales. The application of developed models for predictive simulations has however been strongly limited by the available computational power and the capability of computational fluid dynamics software to handle large enough simulations. In recent years, both aspects have made significant advances and we thus now demonstrate the feasibility of a massively parallel simulation, on whole supercomputers using NEPTUNE_CFD, of an industrial-scale polydispersed fluidized-bed reactor. This simulation of an olefin polymerization reactor makes use of an unsteady Eulerianmulti-fluid approach and relies on a billion cellsmeshing. This is a worldwide premiere as the obtained accuracy is yet unmatched for such a large-scale system. The interest of this work is two-fold. In terms of High Performance Computation (HPC), all steps of setting-up the simulation, running it with NEPTUNE_CFD, and post-processing results induce multiple challenges due to the scale of the simulation. The simulation ran using 1260 up to 36,000 cores on supercomputers, used 15 millions of CPU hours and generated 200 TB of rawdata for a simulated physical time of 25s. This article details the methodology applied to handle this simulation, and also focuses on computation performances in terms of profiling, code efficiency and partitioning method suitability. Though being by itself interesting, the HPC challenge is not the only goal of this work as performing this highly-resolved simulation will benefit chemical engineering and CFD communities.Indeed, this computation enables the possibility to account, in a realistic way, for complex flows in an industrial-scale reactor. The predicted behavior is described, and results are post-processed to develop sub-grid models. These will allow for lower-cost simulations with coarser meshes while still encompassing local phenomenon.

An inversion algorithm to estimate the discharge of rivers observed by the forthcoming SWOT mission (wide swath altimetry) is developed and assessed in detail. The algorithm relies on an advanced variational data assimilation formulation applied to the Saint-Venant equations (1D shallow-water) combined with a lower complexity model (low Froude assumption and locally stationary). This modeling approach enables to estimate from the altimetry measurements the three flow features: the discharge Q(t) associated with an effective bathymetry b(x) and a (non constant) roughness coefficient K. The river geometry is built up with effective cross sections defined from the altimetry measurements. The numerical results are analyzed for three rivers (\sim100 km long each) presenting rapid flow variations compared to the observation frequency. Two scenarios of observation are considered: frequent satellite overpasses corresponding to the SWOT Cal-Val orbit (\sim1 day period) and SWOT like data corresponding to a \sim21 days period with 1 to 4 passes at mid-latitudes. Given prior mean values (of Q or/and b), the numerical experiments demonstrate that the inference of the river discharge Q(t) and the bathymetry profile b(x) may be accurate. Various prior values sources are investigated in view of worldwide applications. Once the assimilation of a long period of measurements is done (e.g. one year long time series), the low complexity flow model is correctly calibrated and provides accurate discharge estimations in real-time from newly acquired measurements.

In the framework of the supervised learning of a real function defined on a space X , the so called Kriging method stands on a real Gaussian field defined on X. The Euclidean case is well known and has been widely studied. In this paper, we explore the less classical case where X is the non commutative finite group of permutations. In this setting, we propose and study an harmonic analysis of the covariance operators that enables to consider Gaussian processes models and forecasting issues. Our theory is motivated by statistical ranking problems.

International audience; Transistor arrangement influences the performance of logic cells. Complex logic cells can be used to minimize area, delay, and power. However, with the increasing relevance of nanometer challenges such as process variability and radiation effects, it is also necessary to consider these factors at logic level design. The main contribution of this work is to evaluate the advantages of adopting a multi-level logic design instead of using complex gates to mitigate process variability and radiation effects. A justification for choosing the parameter that will characterize process variability is presented. Also, the analysis with different transistor arrangements and sizing is carried out to provide a basis for a better understanding of the obtained results. The experimental evaluation considers two topologies: complex gates and multi-level using basic gates, for a set of logic gates using the 7nm FinFET technology at the layout level. As expected, at nominal conditions, the best choice is to adopt complex gate topology to optimize area, power consumption, and performance. However, the logic multi-level arrangements became the functions up to 50% less sensitive to the transient faults and at least 30% more robust to the process variability effects. A trade-off needs to be done, considering the area and power constraints.

International audience; In this paper we study the convergence properties of a Nesterov’s family of inertial schemes which is a specific case of inertial Gradient Descent algorithm in the context of a smooth convex minimization problem, under some additional hypotheses on the local geometry of the objective function F, such as the growth (or Łojasiewicz) condition. In particular we study the different convergence rates for the objective function and the local variation, depending on these geometric conditions. In this setting we can give optimal convergence rates for this Nesterov scheme. Our analysis shows that there are some situations when Nesterov’s family of inertial schemes is asymptotically less efficient than the gradient descent (e.g. in the case when the objective function is quadratic).

International audience; In this study, we present a method that combines a Long Short-Term Memory (LSTM) recurrent neural network with a Gaussian Process (GP) model to provide up to 6-hour ahead probabilistic forecasts of the Dst geomagnetic index. The proposed approach brings together the sequence modelling capabilities of a recurrent neural network with the error bars and confidence bounds provided by a Gaussian process. Our model is trained using the hourly OMNI and GPS databases, both of which are publicly available.We first develop a Long Short-Term Memory network to get a single point prediction of Dst, . This model yields great accuracy in forecasting the Dst index from 1h ahead to 6h ahead, with a correlation coefficient always higher than 0.873 and a root mean square error lower than 9.86. However, even if global metrics show excellent performance, it remains poor in predicting intense storms (Dst

International audience; This paper evaluates a set of complex cells with different transistor arrangements that implement the same logic function. These cells were evaluated under nominal conditions and with gate variability at layout level. The purpose is to verify what topology is more appropriate to increase the robustness of cells regarding the process variability issues. Results emphasize the importance of investigating the effects caused by process variability in FinFET technologies, as the electrical characteristics of circuits suffer significant changes. In general, the best choice is to use the network that the transistor in series is as far as possible to the output node. However, a trade-off needs to be done due to performance and power consumption penalties.

To achieve a high performance level during ground operations, the lateral dynamics of an aircraft must be controlled using all available actuators (rudder, nose-wheel steering system, engines and brakes) and under various constraints, which gives rise to a challenging allocation problem. To address this issue, a simple yet accurate design-oriented on-ground aircraft model is first developed. It takes into account the effects of aerodynamics, thrust and tire-ground interactions, both laterally and longitudinally, and for several runway states. It is validated on a high-fidelity Airbus simulator and a complete set of numerical values representative of a commercial aircraft is given, as well as design objectives, so as to provide the control community with a challenging benchmark. After an extensive literature review and an evaluation of the pros and cons of many existing control allocation techniques, a novel and easily implementable algorithm is then developed, which meets actuator and implementation constraints. It automatically manages the trade-off between two antagonistic objectives, namely minimizing the control effort and attaining the maximum virtual control. Its validation on both the design-oriented model and the high-fidelity simulator shows promising results, and flight tests on a real aircraft are scheduled in the short term., Pour atteindre un niveau de performance élevé lors des opérations au sol, la dynamique latérale d’un aéronef doit être contrôlée à l’aide de tous les actionneurs disponibles (gouverne de direction, roulette de nez, moteurs et freins) et ce sous diverses contraintes, ce qui donne lieu à un problème d’allocation. Dans cette thèse, un modèle d’avion au sol est développé spécifiquement pour cette problématique. Il inclut les différents actionneurs, les efforts aérodynamiques et de poussée, et considère le couplage latéral-longitudinal de l’interaction pneu-sol. Plus particulièrement, cette dernière prend en compte le glissement combiné, l’état de la piste et la variation de la charge normale. Le modèle proposé est validé par comparaison avec un simulateur haute-fidélité d’Airbus. Un ensemble complet de valeurs numériques représentatives d’un aéronef commercial, ainsi que les objectifs de commande associés, sont également donnés afin de fournir à la communauté un problème stimulant. Après un état de l’art exhaustif des techniques d’allocation et une évaluation approfondie de ces dernières, un nouvel algorithme est ensuite développé afin de permettre une utilisation raisonnée du freinage différentiel en plus de la roulette et de la gouverne de direction. L’allocateur proposé, en sus du respect de diverses contraintes (temps de calcul, certification. . . ), gère intelligemment le compromis entre la minimisation de l’utilisation des actionneurs et la réalisation de la commande virtuelle maximale. Sa validation sur le modèle d’avion développé dans la thèse et le simulateur haute fidélité montre des résultats prometteurs, et des essais sur un avion réel sont prévus à court terme.

International audience; The first cases of Severe Acute Respiratory Syndrome coronavirus 2 (SARS‐CoV‐2) occurred in France in early February and in our region (Southwest France) in early March 2020.

This article proposes to study the impact of ground-level enhancement (GLE) induced by extreme solar flares on the soft error rate (SER) for flight representatives to the world-air traffic. A GLE physical model was confronted to cosmic ray variations measured with neutron monitors during the GLEs 5 and 69. Based on said GLE physical model and single-event upset occurrence models, investigations were performed to assess the SER during these GLEs for a bulk 90-nm commercial off-the-shelf SRAM. The main conclusion is that the GLE impact must be considered in the risk assessment, which requires a physical description of the additional solar cosmic rays and a realistic consideration of the air traffic.

In this article, we perform an asymptotic analysis of a nonlocal reaction-diffusion equation, with a fractional laplacian as the diffusion term and with a nonlocal reaction term. Such equation models the evolutionary dynamics of a phenotypically structured population. We perform a rescaling considering large time and small effect of mutations, but still with algebraic law. With such rescaling, we expect that the phenotypic density will concentrate as a Dirac mass which evolves in time. To study such concentration phenomenon, we extend an approach based on Hamilton-Jacobi equations with constraint, that has been developed to study models from evolutionary biology, to the case of fat-tailed mutation kernels. However, unlike previous works within this approach, the WKB transformation of the solution does not converge to a viscosity solution of a Hamilton-Jacobi equation but to a viscosity supersolution of such equation which is minimal in a certain class of supersolutions. Such property allows to derive the concentration of the population density as an evolving Dirac mass, under monotony conditions on the growth rate, similarly to the case with thin-tailed mutation kernels.

International audience; The multipactor phenomenon is characterized by a very fast growth of the electronic population in Radio-Frequency (RF) devices under vacuum. As this effect limits the transmissible RF power and can harm RF systems, it has been widely studied during the last decades. Due to the high cost of experimental tests, simulation tools are heavily used to predict the threshold of multipactor growth. However, their reliability is limited for complex configurations, e.g. when dielectrics or magnetic fields are present. A crucial element of these multipactor simulations is the secondary-emission model. Dionne model is able to model both metals and dielectrics secondary emission but is one-dimension only. As the three-dimensional aspect is essential for complex configurations, the Dionne model is extended do three-dimensions. Measurements of the total electron emission yield have been carried out at the ONERA and shows a good agreement for low-impact energy and low-impact angle electrons, which is relevant in multipactor simulations.

International audience; We study the asymptotic behavior of solutions to the Vlasov equation in the presence of a strong external magnetic field. In particular we provide a mathematically rigorous derivation of the guiding-center approximation in the general three dimensional setting under the action of large inhomogeneous magnetic fields. First order corrections are computed and justified as well, including electric cross field, magnetic gradient and magnetic curvature drifts. We also treat long time behaviors on two specific examples, the two dimensional case in carte-sian coordinates and a poloidal axi-symmetric geometry, the former for expository purposes. Algebraic manipulations that underlie concrete computations make the most of the linearity of the stiffest part of the system of characteristics instead of relying on any particular variational structure.

Jupiter has the most complex and energetic radiation belts in our Solar System and one of the most challenging space environments to measure and characterize in-depth. Their hazardous environment is also a reason why so many spacecraft avoid flying directly through their most intense regions, thus explaining how Jupiter’s radiation belts have kept many of their secrets so well hidden, despite having been studied for decades. In this paper we argue why these secrets are worth unveiling. Jupiter’s radiation belts and the vast magnetosphere that encloses them constitute an unprecedented physical laboratory, suitable for interdisciplinary and novel scientific investigations: from studying fundamental high energy plasma physics processes which operate throughout the Universe, such as adiabatic charged particle acceleration and nonlinear wave-particle interactions, to exploiting the astrobiological consequences of energetic particle radiation. The in-situ exploration of the uninviting environment of Jupiter’s radiation belts presents us with many challenges in mission design, science planning, instrumentation, and technology. We address these challenges by reviewing the different options that exist for direct and indirect observations of this unique system. We stress the need for new instruments, the value of synergistic Earth and Jupiter-based remote sensing and in-situ investigations, and the vital importance of multi-spacecraft in-situ measurements. While simultaneous, multi-point in-situ observations have long become the standard for exploring electromagnetic interactions in the inner Solar System, they have never taken place at Jupiter or any strongly magnetized planet besides Earth. We conclude that a dedicated multi-spacecraft mission to Jupiter is an essential and obvious way forward for exploring the planet’s radiation belts. Besides guaranteeing numerous discoveries and huge leaps in our understanding of radiation belt systems, such a mission would also enable us to view Jupiter, its extended magnetosphere, moons, and rings under new light, with great benefits for space, planetary, and astrophysical sciences. For all these reasons, in-situ investigations of Jupiter’s radiation belts deserve to be given a high priority in the future exploration of our Solar System. This article is based on a White Paper submitted in response to the European Space Agency’s call for science themes for its Voyage 2050 programme.

An unanticipated and unexpected problem occurred when astronauts first walked on the Moon : the strong adhesion of lunar regolith (the dust covering the whole surface of the Moon) to every man-made equipment. Next space missions will be guided thanks to robots. It therefore becomes crucial to understand and bring under control the problem of lunar dust adhesion. This Ph.D thesis aims at studying and characterizing the physical phenomena at stake.The first work done had been the characterization of the Van der Waals adhesion force. This force acting between a grain and a substrate is theoretically defined but remains poorly experimentally characterized. An external motor has been implemented within the DROP (Dust RegOlith and Particles) vacuum chamber in order to determine the adhesion force using the centrifugal force, which is hypothesized to be equal to the former when the particle is removed. Results from this first work showed the importance of several parameters: composition and substrate roughness in association with the particle shape.Electrostatic forces are the second forces responsible for dust contamination. They result from the interaction of the dust grain and space environment and are all electrostatic charge dependent. Two main phenomenon act on charge settlement: charging current collection (from solar wind and photoemission) and conduction current through neighboring surfaces. Second work has been charge characterization, as this parameter is also not well known. Study of the charging currents has been made thanks to numerical simulations done on the SPIS software and showed complex charge pattern. Conduction current through a dust layer was made using two methods: broadband dielectric spectroscopy (BDS) measurements done at CIRIMAT and DC-currents at ONERA. Study of different structures such as layers and monolayers of soda-lime glass spheres gave a better understanding of charge leakage.These conjoint studies (collection and leakage current) help determine the charge balance within a grain, and by expansion electrostatic forces. The outcome, combined to previous results obtained using centrifugal method allow to better understand forces acting on a grain at the surface of the Moon. It also enables us to justify to manufactures of parameters and solutions limiting dust contamination.; Les missions Apollo ont rapporté le problème d’adhésion du régolithe lunaire, où l’épaisse couche de grains présente à la surface de la Lune adhérait aux combinaisons spatiales et à la surface de nombreux équipements techniques et scientifiques. Cette contamination reste un frein pour la suite de l’exploration spatiale, qu’il est nécessaire de caractériser. Sous environnement spatial, les forces responsables de l’adhésion sont les forces de Van der Waals ainsi que les forces électrostatiques. Ces dernières résultent de l’interaction entre la poussière et l’environnement spatial et dépendent de la charge électrostatique. L’objectif du travail présenté lors de cette soutenance est centré sur la compréhension de la dynamique de la charge et de l’adhésion du régolithe lunaire sous environnement spatial. Plus particulièrement, c’est la caractérisation de l’adhésion en termes de force et de la modélisation de la charge électrostatique à l’échelle microscopique qui sont visées. Pour cela, plusieurs outils ont été utilisés : le caisson sous vide DROP qui permet de faire les mesures expérimentales, des simulants de poussières lunaires ainsi que le logiciel de simulations SPIS.La première force à avoir été évaluée est la force de Van der Waals à l’aide d’un dispositif centrifuge dimensionné et implanté dans le caisson DROP. Les résultats de cette première campagne d’essais ont montré que plusieurs facteurs jouaient un rôle important dans l’adhésion : la rugosité et la composition du substrat ainsi que la forme de la particule. Des mesures XPS réalisées parallèlement ont montré une forte oxydation de surface des échantillons suite à leur mise en contact avec l’atmosphère. Cette couche diélectrique pourrait être à l’origine de l’apparition de forces électrostatiques supplémentaires. Les forces électrostatiques ont ensuite été étudiées au travers de la charge électrostatique. Deux phénomènes participent à l’établissement de cette dernière sur un grain de poussière : la collection de courant chargeant (vent solaire, photo-émission) et le courant de fuite (conduction) à travers les surfaces environnantes. L’étude du courant de conduction à travers une couche de poussière a été effectuée grâce à deux techniques de mesures de la conductivité : la spectroscopie diélectrique dynamique (SDD) au CIRIMAT et une mesure de courant DC réalisée à l’ONERA. Les données expérimentales obtenues ont été comparées à un modèle développé durant cette thèse et qui prédit bien les conductivités. Ce même modèle a aussi permis de caractériser des simulants de poussières lunaires en déterminant entre autre leur conductivité bulk. La collection des courants chargeants a ensuite été étudiée via des simulations réalisées avec le logiciel SPIS. Ces modélisations ont permis de mettre en évidence un phénomène de charge complexe à l’échelle d’un grain microscopique, avec des zones positives et négatives. Un deuxième travail couplant des expérimentations et des simulations a permis de mieux comprendre les mécanismes de charge et relaxation de charges au travers d’une couche de poussières, sur des échelles autant macroscopique que microscopique. L’utilisation de SPIS en tant que modèle prédictif de la charge est satisfaisante et l’on préconise maintenant de lui adjoindre un module de calcul des forces électrostatiques.

For nearly two decades, patrolling has received significant attention from the multiagent community. Multiagent patrolling (MAP) consists in modelling a patrol task to optimise as a multiagent system. The problem of optimising a patrol task is to distribute the most efficiently agents over the area to patrol in space and time, which constitutes a decision-making problem. A range of algorithms based on reactive, cognitive, reinforcement learning, centralised and decentralised strategies, amongst others, have been developed to make such a task ever more efficient. However, the existing patrolling-specific approaches based on supervised learning were still at preliminary stages, although a few works addressed this issue. Central to supervised learning, which is a set of methods and tools that allow inferring new knowledge, is the idea of learning a function mapping any input to an output from a sample of data composed of input-output pairs; learning, in this case, enables the system to generalise to new data never observed before.Until now, the best online MAP strategy, namely without precalculation, has turned out to be a centralised strategy with a coordinator. However, as for any centralised decision process in general, such a strategy is hardly scalable. The purpose of this work is then to develop and implement a new methodology aiming at turning any high-performance centralised strategy into a distributed strategy. Indeed, distributed strategies are by design resilient, more adaptive to changes in the environment, and scalable. In doing so, the centralised decision process, generally represented in MAP by a coordinator, is distributed into patrolling agents by means of supervised learning methods, so that each agent of the resultant distributed strategy tends to capture a part of the algorithm executed by the centralised decision process. The outcome is a new distributed decision-making algorithm based on machine learning. In this dissertation therefore, such a procedure of distribution of centralised strategy is established, then concretely implemented using some artificial neural networks architectures. By doing so, after having exposed the context and motivations of this work, we pose the problematic that led our study. The main multiagent strategies devised until now as part of MAP are then described, particularly a high-performance coordinated strategy, which is the centralised strategy studied in this work, as well as a simple decentralised strategy used as reference for decentralised strategies. Among others, some existing strategies based on supervised learning are also described. Thereafter, the model as well as certain of key concepts of MAP are defined. We also define the methodology laid down to address and study this problematic. This methodology comes in the form of a procedure that allows decentralising any centralised strategy by means of supervised learning. Then, the software ecosystem we developed for the needs of this work is also described, particularly PyTrol a discrete-time simulator dedicated to MAP developed with the aim of performing MAP simulation, to assess strategies and generate data, and MAPTrainer, a framework hinging on the PyTorch machine learning library, dedicated to research in machine learning in the context of MAP. Two MAP strategies relying on Long Short-Term Memory (LSTM) networks are then defined: RLPM and RAMPAGER. In those strategies, the LSTM network is used as a predictor that agents use to select the next place to visit in the area to patrol. It is trained over data generated by the centralised strategy. We also show that the stochastic selection of the next place to visit leads to better performance. RAMPAGER, which relies on analytical initialisation of the LSTM network guided by the structure of the area to patrol, turns out to be the best decentralised strategy based on LSTM networks. We then present a new generic type of strategy, called Idleness Estimator, relying on value estimation. In the strategies of this type, each agent embeds an estimator to estimate the time elapsed since the latest visit of any agent on each place to monitor. This estimator is trained over data generated by a high-performance centralised strategy as previously. Different strategies can then be derived according to the used estimator. In this dissertation we study three types of estimators: artificial neural networks, and particularly MultiLayer Perceptrons (MLPs), a linear model, and the mean. An interaction scheme is lastly set up to make agents communicate and improve their individual estimate through interaction. Finally, the Idleness Estimator strategies, either with or without interaction, turn out to be the best decentralised strategies studied in this dissertation.; Depuis presque deux décennies, la tâche de la patrouille a fait l'objet d'une attention toute particulière de la part de la communauté multiagent. La patrouille multiagent consiste à modéliser comme un système multiagent une tâche de patrouille à optimiser. Cette optimisation revient à répartir dans l'espace et le temps les agents patrouilleurs sur la zone à surveiller, cela le plus efficacement possible; un tel problème constitue par là même un problème de décision. Un large éventail d'algorithmes basés sur des stratégies d’agent réactives, cognitives, d’apprentissage par renforcement, centralisées et décentralisées, entre autres, ont été développés pour rendre les stratégies de patrouille toujours plus performantes. Cependant, les approches existantes basées sur de l'apprentissage supervisé avaient peu été étudiées jusqu’à présent, bien que quelques travaux aient abordé cette question. L’idée principale et sous-jacente à l'apprentissage supervisé, qui n’est rien de plus qu’un ensemble de méthodes et d'outils permettant d’inférer de nouvelles connaissances, est d’apprendre une fonction associant à tout élément en entrée un élément en sortie, à partir d'un ensemble de données composé de paires d'éléments entrées-sorties; l'apprentissage, dans ce cas, permet au système de faire de la généralisation à de nouvelles données jamais observées auparavant.Jusqu'à présent, la meilleure stratégie de patrouille multiagent en ligne, à savoir sans calcul préalable, s'est avérée être une stratégie centralisée à coordinateur. Cependant, comme pour tout processus de décision centralisé généralement, une telle stratégie est difficilement échelonnable.L'objectif de ce travail est alors de développer et de mettre en œuvre une nouvelle méthodologie visant à transformer toute stratégie centralisée performante en stratégie distribuée, c'est-à-dire par nature résiliente, plus adaptative aux changements de l'environnement et échelonnable. Ce faisant, le processus de décision centralisé, généralement représenté par un coordinateur dans la patrouille multiagent, est distribué sur les agents patrouilleurs au moyen de méthodes d’apprentissage supervisé, de sorte que les agents de la stratégie distribuée résultante tendent chacun à capturer ou cristalliser une partie de l’algorithme exécuté par le processus de décision centralisé. Le résultat est alors un nouveau algorithme de prise de décision distribué, qui repose sur de l’apprentissage automatique. Dans cette thèse, une telle procédure de distribution de stratégie centralisée est établie, puis concrètement mise en œuvre en utilisant certaines architectures de réseaux de neurones.Ainsi, après avoir exposé le contexte et les motivations, nous posons la problématique étudiée. Les principales stratégies multiagent élaborées jusqu'à présent dans le cadre de la patrouille multiagent sont ensuite décrites, en particulier une stratégie centralisée à haute performance qui est la stratégie centralisée à distribuer ici étudiée, ainsi qu’une stratégie décentralisée assez simple qui est utilisée comme référence pour les stratégies décentralisées. Entre autres, quelques stratégies basées sur de l’apprentissage supervisé sont aussi décrites. Ensuite, le modèle ainsi que certains concept fondamentaux du problème de la patrouille multiagent sont définis. Nous définissons également la méthodologie établie pour aborder et étudier la problématique de cette thèse; méthodologie permettant de décentraliser toute stratégie centralisée par le biais de l’apprentissage supervisé. L'écosystème informatique que nous avons développé pour les besoins de ce travail est également décrit, notamment PyTrol, un simulateur à temps discret dédié à la patrouille multiagent développé dans le but d'effectuer des simulations de patrouille, d'évaluer des stratégies et de générer des données, et MAPTrainer, un framework s'appuyant sur la bibliothèque d’apprentissage automatique PyTorch, dédié à l’étude de l’apprentissage automatique dans le cadre de la patrouille multiagent. Deux stratégies s'appuyant sur des réseaux Long Short-Term Memory (LSTM) sont ensuite définies : RLPM et RAMPAGER. Dans ces stratégies, un réseau LSTM est utilisé comme prédicteur que les agents utilisent pour sélectionner de manière stochastique la prochaine place à visiter sur la zone à patrouiller. Le réseau LSTM est entraîné sur des données générées par la stratégie centralisée. RAMPAGER, qui s'appuie sur une initialisation analytique du réseau LSTM — initialisation guidée par la structure de la zone à patrouiller — s’avère être la meilleure stratégie décentralisée basée sur des réseaux LSTM. Nous présentons ensuite un nouveau type générique de stratégie, appelé Idleness Estimator, qui repose sur de l'estimation d’état. Dans les stratégies de ce type, chaque agent embarque un estimateur pour estimer le temps écoulé depuis la dernière visite sur chaque centre d’intérêt à surveiller. Cet estimateur est entraîné à partir de données générées par une stratégie centralisée très performante, comme précédemment. Différentes stratégies peuvent alors être obtenues en fonction de l'estimateur utilisé. Dans cette thèse, nous étudions trois types d'estimateurs : des réseaux de neurones artificiels, et particulièrement les réseaux de type perceptron multicouche, un modèle linéaire et la moyenne. Enfin, un système d'interaction est mis en place pour permettre aux agents de communiquer et d'améliorer leur estimation individuelle au moyen de l'interaction.Finalement, les stratégies de type Idleness Estimator, que ce soit avec et sans interaction, s'avèrent être les meilleures stratégies décentralisées étudiées dans cette thèse.

We consider the sum of the coordinates of a simple random walk on the K-dimensional hypercube, and prove a double asymptotic of this process, as both the time parameter n and the space parameter K tend to infinity. Depending on the asymptotic ratio of the two parameters, they converge towards either a Brownian motion, an Ornstein-Uhlenbeck process or an i.i.d. collection of Gaussian variables., Journal of Theoretical Probability, Springer, In press

The aim of the thesis is to propose a compact modeling of ionizing particles effects (SET/TID) on sub28nm MOS technologies. Validation of the models will be done through TCAD simulations and SET model will be included into the MUSCA SEP3 simulation platform for soft errors risk assessment purpose., L’effet des radiations naturelles sur l’électronique est une problématique majeure pour tout circuit électronique embarqué dans l’espace, aux altitudes avioniques ou au niveau du sol (automobile). Ces radiations incluent les particules cosmiques primaires pour le spatial (protons, ions lourds…), les particules secondaires au sein de l’atmosphère (neutrons, muons, protons..) ainsi que les particules α résultantes de la radioactivité inhérente à tout matériau (contamination naturelle en Uranium/Thorium). L’effet de ces particules à l’échelle du transistor est la génération de porteurs libres engendrant un pulse en courant transitoire aux électrodes du composant que l’on appelle « Single Event Transient » (SET). Ce pulse est alors capable de perturber le fonctionnement du circuit en générant des erreurs dans le traitement/stockage de l’information : « soft errors ». Par ailleurs, une exposition permanente des circuits aux radiations peut engendrer des dérives dans les caractéristiques électriques des transistors, que l’on nomme effets de dose (« Total Ionizing Dose effects » (TID). La modélisation de ces phénomènes permet d’évaluer le risque d’erreurs des fonctions assurées par un circuit en incluant ces modèles dans les simulateurs de circuit (simulateurs utilisant le langage Verilog-A pour décrire les modèles des composants du circuit considéré).Le modèle SET développé à l’ONERA, dit de « diffusion/collection », permet de bien décrire la phénoménologie du pulse SET pour les transistors BULK. Ce n’est plus le cas pour les technologies plus intégrées (nœuds sub-28nm) telles que les technologies SOI qui font intervenir des phénomènes plus fins. Par ailleurs, ce modèle, bien que parvenant à décrire le pulse SET sur un transistor isolé, ne rend pas compte de la rétroaction dynamique du circuit environnant sur ce pulse : un modèle codé en Verilog-A apparait alors comme l’unique moyen de prendre en compte cette rétroaction de la manière la plus simple possible. Par ailleurs, une implémentation convenable en Verilog-A rendra optimale la robustesse et la rapidité d’exécution du modèle lors de la simulation de circuits faisant intervenir des millions de transistors. Un modèle physique, robuste et rapide d’exécution, et implémenté en Verilog-A dans les simulateurs de circuits est dit compact.

International audience; The ESD cubesat mission aims at demonstrating the efficiency of a new generation of instruments designed to assess and prevent anomalies related to electrostatic charging issues. To meet its objectives this Earth polar orbit mission shall sustain harsh environmental conditions. This paper presents a multi-physics numerical analysis of space environment effects. Auroral electrons injected during geomagnetic substorms are shown to cause severe differential charging on solar cells and on insulating materials. The galactic cosmic rays are shown to be responsible for a single event rate of 122 upsets / day, while protons of the radiation belt contribute only for 0.4 upsets / day on a 28-nm FPGA SRAM based in LEO orbit. The atomic oxygen fluence analysis requires sensitive equipment to be located in the wake of the spacecraft. The impacts on the ESD mission development are discussed.

In many applications, no physical description of the plant is easily available and the control law has to be designed on the basis of input-output measurements only. Two control strategies can then be considered : one can either identify a model of the plant and then use any kind of model-based technique (indirect methods) to obtain a control law, or use a data-driven strategy that directly compute the controller from the experimental data (direct methods). This work focuses on data-driven techniques : the objective of this thesis is to propose a new data-driven control technique based on frequency-domain data collected from the system to be controlled. After recalling some basics in feedback control, an overview of data-driven control is given. Then, the proposed method is introduced. It is a model reference technique where the identification problem is moved from the plant to the controller. In this work, two identification techniques are used to that purpose: the Loewner framework and the subspace approach. In addition, a preliminary analysis of the available frequency-domain data allows determining the performance limitations and selecting achievable specifications.Finally, a stability condition, already known in data-driven control, is used during the reduction of the controller to ensure closed-loop internal stability. Along this thesis, the different steps of the method are progressively applied on two numercial examples. In the end, the proposed technique is applied on two irrational systems described by partial differential equations: a continuous crystallizer and an open-channel for hydroelectricity generation. These two examples illustrate the type of applications for which using a data-driven control method is indicated.; Dans de nombreuses applications, aucun modèle physique du système n'est facilement disponible, il s'agit alors de contrôler le système uniquement à partir de mesures entrées-sorties. Deux types d'approches sont envisageables : identifier un modèle du système puis l'utiliser afin de synthétiser un contrôleur, ce sont les méthodes indirectes, ou identifier le contrôleur directement à partir des données du système, ce sont les méthodes directes. Cette thèse se concentre sur les méthodes directes : l'objectif du travail présenté est de mettre en place une nouvelle méthode directe basée sur des données fréquentielles du système à contrôler. Après un tour d’horizon des méthodes directes existantes, la méthode proposée est introduite. Il s’agit de résoudre un problème de suivi de modèle de référence dans lequel le problème d’identification est déporté du système vers le contrôleur. Dans ce cadre, deux techniques d’identification sont considérées dans cette thèse : l’interpolation de Loewner et l’approche des sous-espaces. De plus, une analyse préliminaire des données fréquentielles disponibles permet de connaître les limites en performances du système et, par conséquent, de choisir des spécifications atteignables. Enfin, une analyse de la stabilité en boucle fermée permet d’obtenir un contrôleur stabilisant d’ordre réduit. Tout au long de ce travail, les différentes étapes de la méthode sont appliquées progressivement sur des exemples numériques.Pour finir, la méthode proposée est appliquée sur deux systèmes irrationnels, décrits par des équations aux dérivées partielles: un cristalliseur continu et un canal de génération hydroélectrique. Ces deux exemples sont représentatifs de la catégorie de systèmes pour lesquels utiliser une méthode de contrôle directe est plus pertinent.

International audience; There is a large design variety of power buses, with voltage levels typically ranging from 28 to 100V. This state-of-art is well adapted to past and current needs in term of power conditioning and disctribution for science and telecommunication satellites. NEvertheless, a short-term need is rising for higher operating voltages, especially for the new electric propulsion systems and high-power payloads. The currently available solutions are to add DC/DC converters inisde echa user equipement to generate all its necessary internal supply lines for the satellite primary power bus. For high power/high voltage loads, this DC/DC stage leads to power dissipation and lowers the overall efficiency of the power chain. A major step forward would be to increase the voltage directly at the level of the primary bus in order to remove some voltage conversion stages leading to lower mass, cost, volume and power dissipation. The work performed within the European Union H2020 project "HV-EPSA" was aimed to study benefits and impact of the implementation of a bus voltage from 300V to 600V, including solar array, solar array drive mechnism, power conditioning and distribution, Hall effect thruster with direct drive topology, battery and harnesses. The main problematic to solve were arcing at high voltage/low pressure (Paschen law), interaction between plasma (natural and from plasmic propulsion) and solar arrays, and distribution function using GaN mosfet. Several test campaigns were performed and the results are presented in the paper.; Il existe une grande variété de conception de bus d'alimentation, avec des niveaux de tension allant généralement de 28 à 100 V. Cet état de l'art est bien adapté aux besoins passés et actuels en termes de conditionnement et de distribution de puissance pour les satellites scientifiques et de télécommunications. Néanmoins, le besoin à court terme augmente pour des tensions de fonctionnement plus élevées, en particulier pour les nouveaux systèmes de propulsion électrique et les charges utiles à haute puissance. Les solutions actuellement disponibles consistent à ajouter des convertisseurs DC / DC à l'intérieur de chaque équipement utilisateur afin de générer toutes ses lignes d'alimentation interne nécessaires pour le bus d'alimentation primaire du satellite. Pour les charges haute puissance / haute tension, cet étage DC / DC conduit à une dissipation de puissance et diminue l'efficacité globale de la chaîne de puissance. Un grand pas en avant serait d'augmenter la tension directement au niveau du bus primaire afin de supprimer certains étages de conversion de tension conduisant à une diminution de la masse, du coût, du volume et de la dissipation de puissance. Les travaux réalisés dans le cadre du projet H2020 de l'Union européenne "HV-EPSA" visaient à étudier les avantages et l'impact de la mise en œuvre d'une tension de bus de 300V à 600V, y compris les panneaux solaires, les systèmes d'entraînement de panneaux solaires, le conditionnement et la distribution de l'énergie, le propulseur à effet Hall avec topologie à entraînement direct, batterie et cablâge. Les principaux problèmes à résoudre étaient les arcs à haute tension / basse pression (loi de Paschen), l'interaction entre le plasma (naturel et de propulsion plasmique) et les panneaux solaires, et la fonction de distribution à l'aide du mosfet GaN. Plusieurs campagnes de tests ont été réalisées et les résultats sont présentés dans l'article.

International audience; Neutron spectrometers are operated simultaneously in the framework of a remoted network. The CHINSTRAP, supported by IPEV, aims at recording neutron spectra at the Concordia station (+3233m) in Antarctica since December 2105. The neutron spectrometer measures the neutron spectrum over a wide energy range from meV up to tens of GeV with a short time resolution. Several parameters can influence the measurement: the atmospheric pressure, the hydrometric environment close to the instrument (snowfall) and the atmospheric water vapor. The atmospheric water vapor dominates the ground cosmic ray variations within the first few hundred meters above the ground. The HAMSTRAD polar project is dedicated to measure of the trends in water vapor and temperature profiles from the lower part of the troposphere to the lower part of the stratosphere, using a radiometer. This paper proposes to present data analyses, including short and long-term analyses (first four years) and taking into account environmental and systematic effects. In particular, study focusing on Forbush Decreases and Ground Level Enhancement occurred in September 2017 is proposed.

Context. Earth and outer planets are known to produce intense non-thermal radio emissions through a mechanism known as cyclotron maser instability (CMI), requiring the presence of accelerated electrons generally arising from magnetospheric current systems. In return, radio emissions are a good probe of these current systems and acceleration processes. The CMI generates highly anisotropic emissions and leads to important visibility effects, which have to be taken into account when interpreting the data. Several studies have shown that modelling the radio source anisotropic beaming pattern can reveal a wealth of physical information about the planetary or exoplanetary magnetospheres that produce these emissions. Aims. We present a numerical tool, called ExPRES (Exoplanetary and Planetary Radio Emission Simulator), which is able to reproduce the occurrence in a time-frequency plane of R−X CMI-generated radio emissions from planetary magnetospheres, exoplanets, or star–planet interacting systems. Special attention is given to the computation of the radio emission beaming at and near its source. Methods. We explain what physical information about the system can be drawn from such radio observations, and how it is obtained. This information may include the location and dynamics of the radio sources, the type of current system leading to electron acceleration and their energy, and, for exoplanetary systems, the orbital period of the emitting body and the strength, rotation period, tilt, and the offset of the planetary magnetic field. Most of these parameters can only be remotely measured via radio observations. Results. The ExPRES code provides the proper framework of analysis and interpretation for past, current, and future observations of planetary radio emissions, as well as for future detection of radio emissions from exoplanetary systems (or magnetic, white dwarf–planet or white dwarf–brown dwarf systems). Our methodology can be easily adapted to simulate specific observations once effective detection is achieved.

International audience; A numerical study of the parameters controlling the viscous penalty method is investigated to better set up Particle-Resolved Direct Numerical Simulations (PR-DNS) of particulate flows. Based on this analysis, improvements of the methods are proposed in order to reach an almost second order convergence in space. The viscous penalty method is validated in Stokes regime by simulating a uniform flow past a fixed isolated cylinder. Moreover, it is also utilized in moderate Reynolds number regime for a uniform flow past a square configuration of cylinder and compared in terms of friction factor to the well-known Ergun correlation.

International audience; Characterized by a very fast growth of the electron population in vacuum of Radio-Frequency (RF) devices, the multipactor effect has been widely studied during the past decades. As it limits the transmitted RF power and may degrade RF devices, its understanding is primordial. Multipactor simulation tools give accurate results in the simplest cases, but are less accurate for advanced configurations: complex geometries, dielectric materials, presence of magnetic fields, etc. In such cases, an accurate modelling of the electron emission phenomena becomes essential. We extended a one-dimension electron emission (EE) model to three dimensions. The obtained model is compared to measured electron emission yields. The impact of this new model on the simulated multipactor threshold of parallel plane wave-guide is also investigated.