Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
La stratégie de développement technologique de Google est-elle résiliente ?
No full textInternational audienc
Thermometric investigations for the characterization of fatigue crack initiation and propagation in Wire and Arc Additively Manufactured parts with as‐built surfaces
No full textInternational audienceMetal Additive Manufacturing (AM) allows for the fabrication of complex shapes with high added value at low costs. Indeed, as-built structures are near net shape: they require few to no finishing operations. However, as-built AM parts present significant roughness caused by the layer discretization. In the case of the Wire and Arc Additive Manufacturing (WAAM) process, used for large-scale structures, the as-built roughness is estimated to several hundreds of micrometers. For complex geometries, a complete machining of the surfaces is not necessarily possible. In this study, an experimental method is proposed, relying on thermoelastic stress analysis, to characterize the effect of as-built WAAM surface roughness on high-cycle fatigue properties. Using an infrared camera, multiple cracks can be detected and monitored over a large surface on rough WAAM samples under cyclic bending. The collected data constitutes valuable information for the identification of a fatigue model dedicated to as-built WAAM structures
Les filles qui, l codent l créent et équilibre : des dispositifs pour valoriser le numérique et les sciences au féminin du primaire au lycée
No full textInternational audienc
A generalized recursive Vogler algorithm for multiple bridged knife-edge diffraction
No full textInternational audienceIn this paper, we consider problem of estimating diffraction attenuation from approximation of a terrain using multiple bridged knife-edge model. This model can be considered as a generalization of well-known multiple knife-edge one where spaces between knife-edge are bridged by reflecting surfaces. A series based-standard solution is presented in the literature but suffers from its very high computational complexity. We, thus, propose to generalize the recursive Vogler algorithm developed for multiple knife-edge model to tackle this problem. To reduce complexity, the proposed algorithm exploit the recursive form to avoid repeated calculations of the existing solution. Moreover, we provide a complexity analysis of both the standard algorithm and the proposed algorithm based on number of computed integrals. Both theoretical and numerical results show that our proposed algorithm is faster than the original one while having identical accuracy, hence proving the effectiveness of our solution
Le MT180® ou les infortunes de l’excellence
No full textInternational audienceTen years ago, My Thesis in 180 Seconds ® (MT180®) appeared in the world Francophone higher education. Modelled on Australia's Three Minute Thesis (3MT), this communication competition for doctoral students engages students in a public speaking competition in which they have three minutes to pitch their research projects. In this sense, MT180 might be understood as a sort of popularization contest aiming to shed light on the "research report." Yet when one considers what is in play more closely (Corsi et al), this interpretation becomes questionable. Indeed, once one begins to ask former participants about their opinions regarding their experiences over the course of the competition, the idea that My Thesis primarily serves to bring science out of the laboratory and to raise interest in researchers becomes ever more clearly in need of being re-examined. This rethinking of what is at stake in MT180 is what we present here.Il y a dix ans maintenant, Ma thèse en 180 secondes® faisait son entrée dans les mondes francophones de la recherche et de l’enseignement supérieur. Inspiré d’un dispositif Australien, Three Minutes Thesis (3MT), cette compétition de communications pour doctorants porte les candidats à s’affronter les uns aux autres à coups de speechs de trois minutes censés rendre compte de leurs travaux de recherche. Il pourrait, en somme, s’agir d’une sorte de concours de vulgarisation visant à mettre en lumière « la relève de la recherche ». Or, à étudier précisément ce qui se dit et se joue au MT180® (Corsi et al.), il est permis d’en douter. Et, à demander à d’anciens lauréats ce qu’ils pensent de leur passage sur les planches du concours, l’analyse selon laquelle Ma thèse en 180 secondes® servirait premièrement à faire sortir les sciences des laboratoires et à valoriser les chercheurs en devenir demande à être réexaminée de près. Ce que nous proposons de faire ici
Self-heating measurements under cyclic loading to identify history effects on HCF properties of high-strength low-alloy steel: Part I—Experimental investigations
No full textInternational audienceHigh cycle fatigue properties of steels are usually characterized in the as-delivered state of the material. However, forming process, variable amplitude and accidental loading, called load history, could affect HCF properties. Given that investigating those history effect of fatigue properties by means of regular fatigue test would be very time consuming, self-heating measurements under cyclic loading were considered. Two load history effects were studied, a macroscopic plastic pre-strain and a cyclic pre-loading, by means of self-heating measurements under cyclic loading. All of the findings were used to propose a modification of a probabilistic two-scale model, which links self-heating measurements to HCF properties, given in the accompanying paper
Loading classification proposal for fatigue design of automotive chassis-parts: A relevant process for variable amplitude and multi-input load cases
No full textInternational audienceIt is essential for car manufacturers to ensure their vehicle robustness under a daily usage. As no detailed method nor standard prevails, each manufacturer is expected to setup its own methodology for safety parts life-assessment. The early stages of a renewed methodology are proposed in this study. This is based on proving-grounds measurements analysis to quantify the impact of various loads. The method, developed from the load measured at the vehicle wheels, considers the multi-input load case and its variable amplitude content. Measured time-series are analyzed and partitioned into two main categories depending on the time-correlation between the load axes. The driver-induced loads, Driven-Road, (DR) loads, are separated from the effects of random road conditions, Random-Road, (RR) loads. Then, for partition validation, a suitable life quantification method is applied in parallel to the two types of loading. On one hand, an assumption based on strong physical meaning eases the Rainflow Counting method application on the DR load. On the other hand, the use of spectral methods is explored to handle the random loads. This paper details the partition process of the time-series and each of the subsequent validation steps of the method to apply such a partition. Such a validation is performed via a comparison of the pseudo-damage between a referenced signal and the one issued from the partition
Emergence of a Symbolic Goal Representation with an Intelligent Tutoring System based on Intrinsic Motivation
No full textInternational audienceGoal representation affects the performance of Hierarchical Reinforcement Learning (HRL) algorithms by decomposing complex problems into easier subtasks. Recent studies show that representations that preserve temporally abstract environment dynamics are successful in solving difficult problems with theoretical guarantees for optimality. These methods however cannot scale to tasks where environment dynamics increase in complexity. On the other hand, other efforts have tried to use spatial abstraction to mitigate the previous issues. Their limitations include scalability to high dimensional environments and dependency on prior knowledge.In this work, we propose a novel three-layer HRL algorithm that introduces, at different levels of the hierarchy, both a spatial and a temporal goal abstraction. We provide a theoretical study of the regret bounds of the learned policies. We evaluate the approach on complex continuous control tasks, demonstrating the effectiveness of spatial and temporal abstractions learned by this approach
Plasticité cyclique multiaxiale en zone confinée : approche théorique et expérimentale
No full textThe structural components exhibit zones of stress concentrations leading to heterogeneous stress states, that can lead to local plasticity and low cycle fatigue failure. In this work, two parts are proposed to study the local cyclic elasto-plastic behavior at the critical point. First, an analytical method is proposed to predict the local response only at the critical point of the structure. In the second step, experimental measurements are obtained through in-situ X-ray diffraction measurements to question the predictions of the simplified finite element method. The study of the behavior at the critical point of the structure is carried out using a method based on the use of an Adjustable Location Operator (ALO). Until now, two strategies have been used to identify this operator on a structure. Based on these methods, two lines of improvement are proposed to improve the numerical predictions. These two in the presence of a new interesting compromise between the identification cost and the quality of the predictions with respect to the numerical reference. In order to compare these numerical results with experimental results, a protocol for in-situ stress monitoring by X-ray diffraction was set up. Subsequently, tests were carried out to observe cyclic behavior, with a progressive and continuous relaxation of the average stress until the initiation of a crack.Les composants structurels présentent des zones de concentrations de contraintes conduisant à des états de contraintes hétérogènes, pouvant amener de la plasticité locale et à des ruptures en fatigue oligocyclique. Dans ce travail, deux volets sont proposés pour étudier le comportement local élasto-plastique cyclique au point critique. Dans un premier temps, une méthode analytique est proposée afin de prévoir la réponse locale uniquement au point critique de la structure. Dans un second temps, des mesures expérimentales sont obtenues grâce à des mesures in-situ par diffraction de rayons X afin de pouvoir questionner les prévisions de la méthode simplifiée et éléments finis. L’étude du comportement au point critique de la structure est réalisée à partir d’une méthode basée sur l’utilisation d’un Opérateur de Localisation Ajustable (OLA). Deux stratégies permettaient jusqu’alors d’identifier cet opérateur sur une structure. En s’inspirant de ces méthodes, deux axes d’améliorations sont proposés afin d’améliorer les prévisions numériques. Ces deux méthodes présentent un nouveau compromis intéressant entre le coût d’identification et la qualité des prévisions vis-à-vis de la référence numérique. Afin de confronter ces résultats numériques à des résultats expérimentaux, un protocole pour le suivi de contrainte in-situ par diffraction de rayons X est mis en place. Par la suite, des essais ont permis d’observer un comportement cyclique présentant notamment une relaxation progressive et continue de la contrainte moyenne jusqu’à l’amorçage d’une fissure
Defense and Artificial intelligence: study on the evolution of the use of artificial intelligence technologies in defense
No full textInternational audienceThis paper uses network analysis tools to study the evolution of the intertwining of defense and artificial intelligence (AI) technologies over the past two decades. AI’s development has brought about a profound change in the research sphere, in some cases altering the development of entire sectors. This disruptive power makes AI of pillar of the fourth industrial revolution. And indeed, AI has become a central component of many countries’ defense policies The technology is also a central element of the unprecedented export control measures of the U.S. administration of October 2022 against China, with the aim of hindering the development of “force-multiplying” technologies for defense capabilities. Focusing our analysis on the evolution of these technologies in defense appears particularly interesting for several reasons: technological edge and adaptability is highly valued for defense material; decision-making often needs to be quick in a complex and evolving environment; high-end and customized product necessitates evolving scientific and technological capacities.After analyzing the evolution of AI technologies in the defense sector, we study the evolution in other sectors of the economy to compare with defense. Network analysis tools are applied to patent data extracted from the Orbit Intelligence and PATSTAT (European Patent Office) databases. Preliminary results show that the defense sector was one of the first to introduce AI into its innovation process. The results show an increase over time in the number of combinations of AI and defense technologies, alongside an intensification of the usage of such links to produce new patents