Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
Learning-based Integrated Cooperative Motion Planning and Control of Multi-AUVs
International audienc
Caractérisation et modélisation de la tenue à la fatigue et à la corrosion de l’acier inoxydable super duplex 2507 obtenu par fabrication additive LP-DED
The main objective of this thesis is to propose an in-depth study of the "Super duplex 2507 stainless steel - Laser Powder-Direct Energy Deposition (LP-DED) process" combination and to establish an approach for optimising the manufacturing process (with or without heat treatment) in terms of mechanical and electrochemical properties. Several steps are being taken to achieve this objective.Firstly, criteria for assessing the quality of the material deposited by additive manufacturing are defined. This led to the optimisation of printing parameters using the LP-DED BeAM Mobile-004 machine and SAF2507 super duplex stainless steel powder.Next, the impact of the process on the microstructure is assessed, as well as the impact of various heat treatments applied to the LP-DED raw material. A comparison with a forged steel of the same grade is also proposed, enabling the differences and similarities between conventional processes and metal additive manufacturing to be put into perspective.Uniaxial tensile tests, hardness tests, pitting corrosion tests and fatigue test campaigns are then carried out on the various materials obtained. This experimental phase enables the influence of microstructure to be studied and variations in the properties examined to be observed.Finally, a model is proposed to describe the fatigue behaviour observed in super duplex 2507 stainless steel. This model takes account of the two-phase nature of the material, with each phase exhibiting a specific behaviour. The proportion of each phase is integrated in an attempt to reflect the contrasts observed during self-heating fatigue testing campaigns as a function of the material's microstructure.This thesis contributes to a better understanding and optimisation of the LP-DED additive manufacturing process for super duplex 2507 stainless steel, by highlighting the importance of the microstructure on the mechanical and electrochemical properties of the material.L'objectif principal de cette thèse est de proposer une étude approfondie du couple "Acier inoxydable super duplex 2507 – procédé Laser Powder-Direct Energy Deposition (LP-DED)" et d'établir une démarche d'optimisation du processus de fabrication (avec ou sans traitement thermique) en termes de propriétés mécaniques et électrochimiques. Pour atteindre cet objectif, plusieurs étapes sont entreprises.Tout d'abord, des critères pour évaluer la qualité du matériau déposé par fabrication additive sont définis. Cela aboutit à l'optimisation des paramètres d'impression en utilisant la machine LP-DED BeAM Mobile-004 et la poudre d'acier inoxydable super duplex SAF2507.Ensuite, l'impact du procédé sur la microstructure est évalué ainsi que de divers traitements thermiques appliqués à la matière brute fabriquée par procédé LP-DED.Une comparaison avec un acier forgé de même nuance est également proposé et permet de mettre en perspective les différences et points communs entre les procédés conventionnels et la fabrication additive métallique.Des essais de traction uniaxiale, de dureté, de corrosion par piqûre et des campagnes d'essais de fatigue sont alors menés sur les différents matériaux obtenus. Cette phase expérimentale permet d'étudier l'influence de la microstructure et d'observer les variations des propriétés examinées.Enfin, en ouverture, un modèle est proposé pour décrire le comportement en fatigue observé sur l'acier inoxydable super duplex 2507. Ce modèle tient compte du caractère biphasique du matériau, chaque phase présentant un comportement spécifique. La proportion de chaque phase est intégrée pour tenter de refléter les contrastes observés lors des campagnes d'essais de fatigue par auto-échauffement en fonction de la microstructure du matériau.Cette thèse contribue ainsi à une meilleure compréhension et optimisation du procédé de fabrication additive LP-DED pour l’acier inoxydable super duplex 2507, en mettant en lumière l'importance de la microstructure sur les propriétés mécaniques et électrochimiques du matériau.Elle propose ainsi les procédés et leurs limites pour aboutir à un matériau imprimé avec les propriétés attendues
Learning on the Edge: Unlocking the Storage Bottleneck with a Divide and Conquer Approach
International audienc
Robust Deterministic Abstractions for Supervising Discrete-Time Continuous Systems
International audienceWe present a method for constructing discrete abstractions for discrete-time, continuous-state systems. Related approaches construct a discrete bisimulation, which leaves little room for non-determinism in the outputs and quickly leads to highly complex models since all concrete behavior is covered. Our approach is to relax these requirements and build a satisficing solution: a discrete abstraction that is deterministic, robust, and as complete as possible under the given parameters. This allows us to balance granularity and computational feasibility. We leverage linearization and linear feedback control to extend the approach from globally contractive systems to systems with contractive cycles. The resulting abstraction directly induces a supervisor policy. The approach is illustrated with numerical experiments and has potential applications in various domains where system safety and reversibility are essential
High-order numerical integration on self-affine sets
We construct an interpolatory high-order cubature rule to compute integrals of smooth functions over self-affine sets with respect to an invariant measure.The main difficulty is the computation of the cubature weights, which we characterize algebraically, by exploiting a self-similarity property of the integral.We propose an -version and a -version of the cubature, present an error analysis and conduct numerical experiments
La diversification technologique de Google sur la décennie 2010
International audienceLa stratégie de diversification technologique « liée de manière contrainte » de Google/Alphabet complexifie le paysage technologique de l’entreprise autour de son cœur de compétences en informatique. Nous produisons une spectrométrie de l’activité d’innovation technologique de Google/Alphabet sur la période 2011-2018 pour identifier les technologies qui structurent cette activité en utilisant les outils de l’analyse des réseaux. Cette période est marquée par une accélération du rythme de la diversification technologique de l’entreprise qui s’appuie sur une pratique intensive d’acquisitions et de dépôts de brevets. Nous proposons d’explorer deux hypothèses. La première considère que les diversifications récentes restent contraintes, mais de façon indirecte, au cœur informatique par le biais de diversifications passées. La deuxième considère les opportunités croissantes d’innovations originales permises par le « dé-silotage » des domaines technologiques que l’entreprise maîtrise
Sim-to-real transfer of adaptive control parameters for AUV stabilisation under current disturbance
International audienceLearning-based adaptive control methods hold the potential to empower autonomous agents in mitigating the impact of process variations with minimal human intervention. However, their application to autonomous underwater vehicles (AUVs) has been constrained by two main challenges: (1) the presence of unknown dynamics in the form of sea current disturbances, which cannot be modelled or measured due to limited sensor capability, particularly on smaller low-cost AUVs, and (2) the nonlinearity of AUV tasks, where the controller response at certain operating points must be excessively conservative to meet specifications at other points. Deep Reinforcement Learning (DRL) offers a solution to these challenges by training versatile neural network policies. Nevertheless, the application of DRL algorithms to AUVs has been predominantly limited to simulated environments due to their inherent high sample complexity and the distribution shift problem. This paper introduces a novel approach by combining the Maximum Entropy Deep Reinforcement Learning framework with a classic model-based control architecture to formulate an adaptive controller. In this framework, we propose a Sim-to-Real transfer strategy, incorporating a bio-inspired experience replay mechanism, an enhanced domain randomisation technique, and an evaluation protocol executed on a physical platform. Our experimental assessments demonstrate the effectiveness of this method in learning proficient policies from suboptimal simulated models of the AUV. When transferred to a real-world vehicle, the approach exhibits a control performance three times higher compared to its model-based nonadaptive but optimal counterpart
Engineering Education challenges in France and in Taiwan: The French model of Engineering Education along with main issues and a report on a research about Engineering Education in Taiwan. Conférence invitée
Conférence invitée. National Science and Technology Council (NSTC), TaipeiWe propose to delve into the unique characteristics and contemporary challenges faced by engineering education in France and in Taiwan. These two educational systems differ greatly in structure and approach, but they also share common challenges and opportunities that are crucial for the development of future engineers. It would be interesting to explore these similarities and differences, and discuss how both systems can learn from each other to better prepare engineers for the global stage.First, I will present the engineering education in France and how it addresses some very current challenges. Then, I will present the first results of a research conducted on engineering education in Taiwan
Apprentissages, sociabilisation, politisation. Perspective diachronique sur les effets d'une controverse sociotechnique sur les acteurs locaux (2014-2024)
International audienc
La question de la technologie à l’ère des fusées en Chine et en Occident
DoctoralThis essay, a talk given on two occasions in the spring of 2024, explores the political and cultural implications of space expansionism, and in particular the notion of astrocolonialism, with respect to existing normative ideas concerning political order in China and the Western tradition. It engages with the work of Martin Heidegger as well as the contemporary Chinese philosopher Tingyang Zhao. It is arguably a contribution to the study of what Yuk Hui calls "Cosmotechnics," and more specifically a study of the implications of space expansionism for the discursive orientation of tradition ways of imagining the spatial relationship between technology and the cosmos.Cet essai, une conférence donnée à deux reprises au printemps 2024, explore les implications politiques et culturelles de l'expansionnisme spatial, et en particulier la notion d'astrocolonialisme, par rapport aux idées normatives existantes concernant l'ordre politique en Chine et la tradition occidentale. Il s'intéresse aux travaux de Martin Heidegger ainsi qu'au philosophe chinois contemporain Tingyang Zhao. Il s’agit sans doute d’une contribution à l’étude de ce que Yuk Hui appelle la « cosmotechnique », et plus spécifiquement d’une étude des implications de l’expansionnisme spatial pour l’orientation discursive des manières traditionnelles d’imaginer la relation spatiale entre la technologie et le cosmos