Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
Méthodes d'inversion combinées pour les problèmes inverses de conductivité
No full textIn this thesis, we develop an inversion method that combines the use of a Kohn-Vogeliustype cost functional with a non-overlapping domain decomposition method as an iterativesolver. The idea behind this method is to iterate simultaneously on the solution of the directproblem using the domain decomposition method and on the unknown of the inverse problemusing gradient descent on the Kohn-Vogelius cost functional. This type of approach falls intothe category of ”one-shot inversion methods,” and its use has the potential to significantlyreduce the cost of inversion when the numerical solution of the direct problem is costly. Weare particularly interested in the case of geometric inverse problems where the unknown ofthe inverse problem is the support of a physical parameter’s discontinuity. The developmentsmade in this area were modeled on the inverse electrical conductivity problem, where the goalis to reconstruct the conductivity discontinuity interface from Cauchy data on the domainboundary. We prove the local convergence of the method in simplified cases and numericallyshow its efficiency for some two dimensional experiments with synthetic data. Additionally,we extend our approach to the more complex case where we also iterate on the value of con-ductivity. In this context, we have also developed an alternating inversion algorithm betweenthe geometry and the inner value of the conductivity, with an adaptive descent step.Dans cette thèse, nous développons une méthode d'inversion qui combine l'utilisation d'une fonctionnelle coût de type Kohn-Vogelius avec une méthode de décomposition de domaine sans recouvrement en tant que solveur itératif. L'idée derrière cette méthode est d'itérer simultanément sur la solution du problème direct via la méthode de décomposition de domaine et sur l'inconnue du problème inverse en utilisant une descente de gradient sur la fonctionnelle de Kohn-Vogelius. Ce type d'approche fait partie de la famille des méthodes dites "one-shot inversion methods", et son utilisation a le potentiel de réduire sensiblement le coût de l'inversion lorsque la résolution du problème direct est coûteuse. Nous nous intéressons plus particulièrement au cas des problèmes inverses géométriques où l'inconnue du problème inverse est le support d'une discontinuité d'un paramètre physique. Les développements réalisés sur cette thématique ont pris pour modèle inverse le problème inverse de conductivité électrique, où l'on cherche à reconstruire l'interface de discontinuité de la conductivité à partir de données de Cauchy sur la frontière du domaine. Nous prouvons un résultat de convergence locale de la méthode dans des cas simplifiés et l'avons validée numériquement pour certaines expériences bidimensionnelles avec des données synthétiques. De plus, nous étendons notre approche au cas plus complexe où l'on itère également sur la valeur de la conductivité. Dans ce contexte, nous avons également développé un algorithme d'inversion alternée entre la géométrie et la valeur intérieure de la conductivité, avec un pas de descente adaptatif
Beyond Presence: A No Compromises Approach to Re-Enchanting the Lunar Common
No full textInternational audienceThere is at present minimal popular outcry about the coming expropriation of the moon, and even less regarding the terrestrial implications of uneven lunar development.This paper explores this indifference and presents a no-compromises case for protecting the lunar commons. It focuses on the disenchantment of the moon, doing so by plunging into the longer history of our ideas regarding the value and significance of the extraterrestrial. It suggests that global accounts of the secularization of the world are misleading: they fail to account for the ultimate unsecularizability of the cosmos; and likewise fail to acknowledge the particular process of secularization and becoming worldly that has occurred with respect to the moon. Re-situating the secularization narrative beyond heaven and earth, this paper suggests that while deeper space has become a modernized version of the ancestral heaven, the moon has become the most disenchanted place in the cosmos, while Earth, thanks to the magical experiences generated by the moon missions, has begun to be re-enchanted. Drawing on work by Federici (2019) and Dardot and Laval (2020), the paper explores how practices of terrestrial “re-enchantment” can be used to protect the lunar commons, finding some possibilities, but also some dangers, including a tendency towards a fetishization of presence that comes from the earth and which may undermine efforts at lunar protection, and a belief in magical techno-solutionism that mirrors ideas derived from speculations on the deeper heavens
Enjeux environnementaux et sociétaux : des écosystèmes de formation, pour quelles transformations ?
No full textInternational audienceThis symposium focuses on three initiatives dedicated to the training of engineers to understandand act on environmental and societal issues of the 21st century, which are anchored in threedifferent engineering training institutions in France. The ambition of this work is to study thedesign and implementation of concrete systems to respond to these issues, to provide elementsof understanding and to stimulate constructive questioning and a reflective approach for theactors of this type of training at different levels of the curricula (macro - institutions, meso -programs and systems, micro - actors). Through a cross-presentation of these experimentalsystems, and by taking a step back on the practices and modalities via a systemic approach, thecontributors highlight the perceived contributions and limits, question the targetedtransformations and the effects obtained, and identify the questions and perspectives raised.This feedback and the analyses that result from it contribute to a broader scientific reflection onthe design of training as a complex ecosystem. A specific way of conceiving training around theparadigm of complexity is emerging. This study allows us to take a different look at the processesof (co-)construction of such systems and to reexamine the learning processes and teachingmethods, even though the objectives and goals are, by nature, evolving and often uncertain.Ce symposium s'intéresse à trois initiatives dédiées à la formation d'ingénieurs pour comprendre et agir vis-à-vis d'enjeux environnementaux et sociétaux du XXIème siècle, qui prennent ancrages dans trois institutions de formation d'ingénieurs différentes en France. L'ambition de ce travail est d'étudier la conception et la mise en œuvre de dispositifs concrets pour répondre à ces enjeux, d'apporter des éléments de compréhension et de susciter un questionnement constructif et une démarche réflexive pour les acteurs de ce type de formations aux différents niveaux des curricula (macro – les institutions, méso – les programmes et dispositifs, micro – les acteurs). Par une présentation croisée de ces dispositifs expérimentaux, et une prise de recul sur les pratiques et modalités via une approche systémique, les contributrices et contributeurs mettent en évidence les apports et limites perçus, interrogent les transformations visées et les effets obtenus, identifient les questions et perspectives soulevées. Ces retours d'expériences et les analyses qui en découlent contribuent à une réflexion scientifique plus vaste autour de la conception des formations en tant qu'écosystèmes complexes. Se dessine une manière spécifique de concevoir la formation autour du paradigme de la complexité. Cette étude permet alors de porter un regard autre sur les processus de (co-)construction de tels dispositifs et de réinterroger les processus d'apprentissage et les modalités d'enseignements, alors même que les objectifs et finalités visés y sont par nature en évolution, et souvent incertains
Controlling Magnesium Silicates Coprecipitation Conditions: A Tool to Tune Their Surface Acid–Base Reactivity
No full textInternational audienceMagnesium silicates combining acidic and basic surface properties are known to be interesting as heterogeneous catalysts. Nevertheless, their catalytic performances are highly dependent on the synthesis method used. In this study, a series of magnesium silicates was synthesized for the first time using a coprecipitation method with a micromixer. It is first shown that changes in synthesis/precipitation pH led to magnesium silicates with different Mg/Si ratios: the higher the synthesis pH, the higher the Mg/Si ratio. Moreover, prepared silicates with a final Mg/Si ratio greater than 0.7, thus prepared at high pH, exhibit negligible specific surface area, whereas relatively high values (>180 m2/g) have been obtained for lower Mg/Si ratios. A set of experimental characterization data obtained by N2 physisorption, SEM, XRD, TGA-DTA as well as Raman and 29Si NMR spectroscopies are presented and discussed. They show the existence of two distinct families with a similar Magnesium Silicate Hydrate (MSH) phase, but they reveal different aggregation states and textural properties. Finally, the surface acid–base reactivity of the co-precipitated magnesium silicates was determined using the model reaction of 2-methylbut-3-yn-2-ol (MBOH) conversion. The results obtained suggest that it is possible to prepare silicates with a wide range of surface acid–base properties, from purely basic solids to those with both acidic and basic properties, by adjusting the final Mg/Si ratio via the control of the synthesis parameters
Supporting Qualitative Analysis with Large Language Models: Combining Codebook with GPT-3 for Deductive Coding
No full textInternational audienceQualitative analysis of textual contents unpacks rich and valuable information by assigning labels to the data. However, this process is often labor-intensive, particularly when working with large datasets. While recent AI-based tools demonstrate utility, researchers may not have readily available AI resources and expertise, let alone be challenged by the limited generalizability of those task-specific models. In this study, we explored the use of large language models (LLMs) in supporting deductive coding, a major category of qualitative analysis where researchers use pre-determined codebooks to label the data into a fixed set of codes. Instead of training task-specific models, a pre-trained LLM could be used directly for various tasks without fine-tuning through prompt learning. Using a curiosity-driven questions coding task as a case study, we found, by combining GPT-3 with expert-drafted codebooks, our proposed approach achieved fair to substantial agreements with expert-coded results. We lay out challenges and opportunities in using LLMs to support qualitative coding and beyond
Boundary Element Method Analysis of 3D Effects and Free-Surface Proximity on Hydrofoil Lift and Drag Coefficients in Varied Operating Conditions
No full textInternational audienceThe use of hydrofoils to enhance ship performance raises the scientific issue of free-surface proximity, which is important to consider during the design stage, to feed velocity prediction programs, for instance. Typically, the flow over a shallowly submerged hydrofoil is characterized by the Froude number, the submergence depth-to-chord ratio, the angle of attack, and geometric parameters of the lifting surface. Among these parameters, the present paper investigates the influence of the wing aspect ratio on the lift and drag coefficients of hydrofoils operating near a free surface. For this purpose, rectangular wings with an H105 profile at 2° angle of attack and aspect ratios ranging from 4 to 20 are systematically analyzed using a 3D boundary element method. The free surface is modeled using a linearized Neumann-Kelvin boundary condition. Chord-based Froude numbers of 0.5, 1.1, and 6.3 are studied. The submersion depth is swept between 0.1 and 30 times the foil chord length. The evolution of the normalized lift and drag coefficients with respect to the foil submersion and the aspect ratio is discussed in detail. Flow velocity is shown to play a significant role in the evolution of the lift and drag coefficients with submersion depth, close to the free surface, for all the aspect ratios. Its influence gets reduced by moving away from the free surface. The critical submersion depth, where the free-surface effects cease, is found to increase with higher flow velocity and aspect ratio. Furthermore, both positive and negative correlations between the force coefficients and the aspect ratio are identified, depending on the operating conditions. It is found that when the proximity to the free surface either enhances or impairs a force coefficient relative to its value in unbounded flow, increasing the aspect ratio amplifies this effect. Overall, this study confirms the effectiveness of steady boundary element methods for simulating the flow around hydrofoil wings in the vicinity of a free surface and contributes to further understanding the influence of geometric parameters on hydrofoil performance
Improving Knee Osteoarthritis Classification with Markerless Pose Estimation and STGCN Model
No full textInternational audienceKnee osteoarthritis (KOA) is a debilitating disease that greatly impacts the quality of life, particularly among the elderly population. Conventional subjective assessment methods for KOA have limitations in terms of accuracy and objective diagnosis. This paper proposes an innovative approach by integrating advanced technologies, specifically the Spatio-Temporal Graph Convolutional Network (STGCN), applied to gait analysis from markerless videos, for precise and quantitative assessment of KOA. The STGCN network is applied to normalized data obtained from Blazepose, a markerless pose estimation technique. Evaluated on an academic dataset of 80 RGB videos, it provides an accuracy of 93.75%. By leveraging the capabilities of the STGCN network, this study significantly enhances the classification of KOA based on gait patterns, offering promising prospects for improved diagnosis and treatment strategies for individuals with KOA
A Framework to Design and Build a Height Controllable Eversion Robot
No full textInternational audienceEversion robots are used in many different environments for a variety of applications.However, the vast majority of these environments are land based and minimal work has been done on environments that are 3-dimensional (3D) such as in air or underwater. To be able to successfully operate in such environments, eversion robots need to be capable of operatingin3D,including adjusting their configuration along the vertical axis. This vertical motion for eversion robots has not been developed before. This paper proposes a design for an eversion robot that, when fully extended, is able to float in air, using varying amounts of helium to change its buoyancy and hence, the amount of lift force in the vertical direction. The robot in this paper consists of a fixed base, one main continuous everting chamber with a constriction at half its length to form two links and two actuation pouches, which are externally attached. Furthermore, the calculations for determining the quantity of helium required to generate a specific force to lift the robot a specified distance, for position changes is shown. Finally, a kinematic model of the robot shows how the robots link tip positions affect the shape and position of the robot
Attosecond Science: an emerging field brought to light by the Physics Nobel Prize
No full textInternational audienceThe 2023 Nobel Prize in Physics has been awarded to Anne L’Huillier, Pierre Agostini and Ferenc Krausz “for experimental methods that generate attosecond pulses of light for the study of electron dynamics in matter”. In this article, we contextualize their seminal works and their central roles that led to the emergence of a global ultrafast community exploring the frontiers of electron dynamics in dilute and condensed matter