Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
Comparaison de conditions de surface libre linéarisées pour l'étude d'un hydrofoil submergé à l'aide d'une approche potentielle
No full textInternational audienc
Analyse expérimentale, théorique et numérique des matériaux à mémoire de forme NiTi poreux
No full textExperiments show that porosity brings new features to NiTi shape memory alloys such as low stiffness and density. The bio-compatible porous NiTi, especially with the pseudoelastic behavior, has a great potential for implant applications in the human body. Although different techniques are provided in the literature for manufacturing NiTi samples, the subject is still open for further investigation to achieve superior shape memory characteristics from the material. Based on this, the aim of the PhD thesis is to analyze and model the mechanical behavior of porous NiTi shape memory alloys.Porous NiTi compacts were produced using spark plasma sintering (SPS), and tested under uniaxial compression to observe the pseudoelasticity and shape memory effect. After sintering, the samples did not show the expected pseudoelastic behavior, and they were subjected to further heat treatment and characterization. A wide range of homogenization and/or aging were applied at different temperatures and duration, and the effect of heat treatment on the pseudoelastic behavior of SPSed NiTi was investigated systematically. Transformation behavior was observed using differential scanningcalorimetry, and phase composition was identified using X-ray diffraction (XRD). These characterizations gave an insight into the micro-structure of the SPSed samples after heat treatment. Then the mechanical behavior of the samples was studied. Instrumented micro-indentation was carried out to measure the hardness altered by aging. Selected samples were tested under uniaxial compression and differences in their pseudoelastic behavior were shown on stress-strain diagrams. At the end, an enhancement in the pseudoelastic response of the SPSed NiTi samples was observed, and the effect ofheat-treatment was discussed.In the modeling part, a macro-scale phenomenological model was proposed for the mechanical behavior of porous SMAs by using a poromechanics approach. The new phenomenological model considers the porous medium as a skeleton consisting of a solid matrix and connected porous space. The model was built starting from a dense SMA model (Zaki-Moumni model) in which the porosity was included as an internal state variable. Both the pseudoelastic and plastic deformations were considered, and the model was implemented into Abaqus through a user-defined material subroutine (UMAT) by using the implicit backward-Euler integration scheme. The phenomenological model was validated using experimental results available in the literature, and the numerical results obtained from the unit cell (UC) technique used in this study. The uni-axial stress-strain response was captured to a great extent with a significant reduction in numerical cost when compared to the UC approach. It isshown that the proposed model represents the mechanical behavior of porous SMAs with reasonable accuracy and could be especially useful in possible biomedical applications of SMAs.Les expériences montrent que la porosité apporte de nouvelles caractéristiques aux alliages à mémoire de forme NiTi, comme une faible rigidité et une faible densité. Le NiTi poreux biocompatible, notamment grâce à son comportement pseudo-élastique, présente un grand potentiel pour les applications d'implants dans le corps humain. Bien que différentes techniques soient fournies dans la littérature pour la fabrication d'échantillons de NiTi, le sujet reste ouvert pour des recherches supplémentaires afin d'optimiser es caractéristiques de mémoire de forme. Sur cette base, l'objectif de la thèse est d'analyser et de modéliser le comportement mécanique des alliages à mémoire de forme poreux en NiTi.Des échantillons poreux en NiTi ont été produits en utilisant le frittage par plasma à étincelle (SPS), et testés sous compression uniaxiale pour observer la pseudoélasticité et l'effet de mémoire de forme. Après le frittage, les échantillons n'ont pas montré le comportement pseudoélastique attendu, et ils ont été soumis à un traitement thermique supplémentaire et à une caractérisation. Une large gamme d'homogénéisation et/ou de vieillissement a été appliquée à différentes températures et durées, et l'effet du traitement thermique sur le comportement pseudoélastique du NiTi a été étudié systématiquement. Le comportement de transformation a été observé parcalorimétrie à balayage différentiel, et la composition des phases a été identifiée par diffraction des rayons X (XRD). Ces caractérisations ont donné un aperçu de la micro-structure des échantillons après traitement thermique. Ensuite, le comportement mécanique des échantillons a été étudié. Une micro-indentation a été réalisée pour mesurer la dureté altérée par le vieillissement. Les échantillons sélectionnés ont été testés sous compression uniaxiale et les différences dans leur comportement pseudo-élastique ont été montrées sur les diagrammes contrainte-déformation. A la fin, une amélioration de la réponse pseudo-élastique des échantillons de NiTi traités par SPS a été observée, et l'effet du traitement thermique a été discuté.Dans la partie modélisation, un modèle phénoménologique à l'échelle macro a été proposé pour le comportement mécanique des AMF poreux en utilisant une approche poro-mécanique. Le nouveau modèle phénoménologique considère le milieu poreux comme un squelette constitué d'une matrice solide et d'un espace poreux connectés. Le modèle a été construit à partir d'un modèle de l'AMF dense (modèle de Zaki-Moumni) dans lequel la porosité a été incluse comme une variable d'état interne. Les déformations pseudo-élastiques et plastiques ont été prises en compte, et le modèle a été implémenté dans Abaqus par le biais d'un sous-programme (UMAT) en utilisant le schéma d'intégration implicite d'Euler. Le modèle phénoménologique a été validé en utilisant les résultats expérimentaux disponibles dans la littérature, et les résultats numériques obtenus à partir de la technique de cellule unitaire (UC) utilisée dans cette étude. La réponse contrainte-déformation uni-axiale a été capturée dans une large mesure avec une réduction significative du coût numérique par rapport à l'approche UC. Il estIl est démontré que le modèle proposé représente le comportement mécanique des AMF poreux avec une précision raisonnable et pourrait être particulièrement utile dans d'éventuelles applications biomédicales des AMF
Statistical characterization of the microstructure of microcellular polyurethane foams
No full textInternational audienceMicrocellular polyurethane (MCU) foams are used in many industrial applications. Their effective properties are closely related to their microstructure. Being able to characterize their microstructure is therefore particularly useful for designing them or improving the quality control of parts on production lines. This paper presents tools developed to quantitatively characterize the cellular microstructure of MCU. The goal is to propose a method that is accessible to academic and industry laboratories. The paper investigates the possibility of using scanning electron microscope images to perform this characterization automatically, quantitatively, and with sufficient statistics. To evaluate the relevance and limitations of the approach, the results are compared to data obtained by analysing X-ray micro-computed tomography images obtained on the same samples. To challenge the methodology, the study is conducted on twelve samples with densities between 400 and 600 kg m−3 and different cellular microstructures. A relevant image processing procedure is developed to detect the cells from the acquired images, automatically and objectively. The fraction, size, shape and spatial distribution of the cells are analysed. The difference between 2D and 3D measurements is investigated. The minimum size of the area to analyse, which guarantees the representativeness of the results from a statistical point of view, is determined
Positron Acceleration in Linear, Moderately Non-Linear and Non-Linear Plasma Wakefields
No full textInternational audienceAccelerating particles to high energies with high efficiency and beam quality is crucial in developing accelerator technologies. The plasma acceleration technique, providing unprecedented high gradients, is considered as a promising future technology. While important progress has been made in plasma-based electron acceleration in recent years, identifying a reliable acceleration technique for the positron counterpart would pave the way to a linear e⁺e⁻ collider for high-energy physics applications. In this work, we show further studies of positron beam quality in moderately non-linear (MNL)* plasma wakefields. With a positron bunch of initial energy 1 GeV, emittance preservation can be achieved in optimised scenarios at 2.38 mm’mrad. In parallel, asymmetric beam collisions at the interaction point (IP) are studied to evaluate the current luminosity reach and provide insight to improvements required for positron acceleration in plasma. It is necessary to scale down the emittance of the positron bunch. In the MNL regime, a positron beam with 238 ’m’mrad level emittance implies compromise in charge or necessity for ultra-short bunches
Asking for Knowledge : Training RL Agents to Query External Knowledge Using Language
No full textInternational audienceTo solve difficult tasks, humans ask questions to acquire knowledge from external sources. In contrast, classical reinforcement learning agents lack such an ability and often resort to exploratory behavior. This is exacerbated as few present-day environments support querying for knowledge. In order to study how agents can be taught to query external knowledge via language, we first introduce two new environments: the grid-world-based Q-BabyAI and the text-based Q-TextWorld. In addition to physical interactions, an agent can query an external knowledge source specialized for these environments to gather information. Second, we propose the 'Asking for Knowledge' (AFK) agent, which learns to generate language commands to query for meaningful knowledge that helps solve the tasks. AFK leverages a non-parametric memory, a pointer mechanism and an episodic exploration bonus to tackle (1) irrelevant information, (2) a large query language space, (3) delayed reward for making meaningful queries. Extensive experiments demonstrate that the AFK agent outperforms recent baselines on the challenging Q-BabyAI and Q-TextWorld environments. The code of the environments and agents are available at https://ioujenliu.github.io/AFK
Explaining Aha! moments in artificial agents through IKE-XAI: Implicit Knowledge Extraction for eXplainable AI
No full textInternational audienceDuring the learning process, a child develops a mental representation of the task he or she is learning. A Machine Learning algorithm develops a latent representation of the task it learns. We investigate the development of the knowledge construction of an artificial agent (AA) by getting inspiration from the one of children, through the analysis of its behavior, i.e., its sequences of moves. We focus on the Tower of Hanoï (TOH) task, a well-known transformation problem in the field of problem-solving as well as one of the fundamental tasks to study children’s knowledge construction about their world and the Aha! phenomenon. We define knowledge here as a set of facts, information, and skills acquired through experience by the AA that contribute to gaining a theoretical or practical understanding of a subject or the world. The main contribution of our work is to propose a 3-step end-to-end methodology for knowledge extraction from AA named Implicit Knowledge Extraction with eXplainable Artificial Intelligence (IKE-XAI). IKE-XAI extracts the AA implicit knowledge in form of an automaton, encoded during its learning. We showcase this technique to solve and explain the TOH task when researchers have only access to moves that represent observational behavior as in human-machine interaction. The 3 steps of IKE-XAI are: first, a Q-learning agent that learns to perform the TOH task; second, a trained recurrent neural network with LSTM units that encodes an implicit representation of the TOH task; and third, an XAI process using a post-hoc implicit rule extraction algorithm to extract graph representations (Finite State Automata, FSA) as visual and explicit explanations of the behavior of the Q-learning agent. This methodology blends neural and symbolic (in our case FSA) components to provide more interpretable model outcomes.At the experimental level, we demonstrate that it is possible to extract the vision of the AA of a simple task (TOH with N=3 disks) and complex one (TOH with N=4 disks and N =6 disks), in the form of FSA that represents AA’s problem-solving strategies, for their explainability. In parallel to the decrease in the average number of movements required to complete a task, namely the acquisition of expertise, we also observed a change in the FSAs extracted at different moments of this acquisition of expertise. The analysis of the characteristics of the FSAs shows a change in the number of nodes and the weights of the transitions. Regarding the Aha! moment, in the 3 experimental contexts, the analyses carried out allowed us to conclude that the Aha! moment for an AA occurs when it changes its behavior in a noticeable way, which translates into a significant change in the extracted FSAs and a stabilization of these. Our experiments show that the IKE-XAI approach helps to understand the development of the Q-learning agent behavior by providing a global explanation of its knowledge evolution during learning. IKE-XAI also allows researchers to identify the agent's Aha! moment by determining from what moment the knowledge representation stabilizes and the agent no longer learns.As a conclusion, we showed that IKE-XAI makes it possible to elucidate the evolution of knowledge acquisition of a learning AA through the study of its behavior over time in terms of an extracted, synthesizing FSA. This allows us to convey, in a symbolic manner, a more explainable vision of AAs. This work also brings a light on the subject of the Aha! moment for autonomous agents and beyond, it leads to a reflection on the question of the definition of insight for an autonomous artificial agent. The convergence of models is thus interesting for the study of this phenomenon in autonomous artificial agents, and more globally for the question of explainability
Assessing marine mammal diversity in remote Indian Ocean regions, using an acoustic glider
No full textInternational audienceMany observations collected from whaling logbooks or more recent satellite tags and acoustic surveys report that the Indian Ocean is a very important place for large baleen whales. They undergo long seasonal migrations from Southern feeding grounds to tropical and subtropical mating and breeding grounds. However, whether and where they stop to rest or feed during their long travels are poorly known. The Indian Ocean is also home to many odontocete species such as sperm whales, killer whales and multiple delphinid species. In this paper, we analyze passive acoustic data collected by an electric glider around two steep bathymetric features located in the Western subtropical Indian Ocean (Walters Shoal) and in the mid subtropical Indian Ocean (St. Paul and Amsterdam islands), both included in Important Marine Mammal Areas (IMMAs). The acoustic data were manually reviewed and annotated by two analysts. The aim of this experiment was to improve the knowledge on marine mammal presence in these little studied IMMAs. We found that bioacoustic activity was quite high in both monitored areas with 40% of the records containing marine mammal sounds in Walters Shoal and 70% in St. Paul and Amsterdam islands. Calls from Antarctic blue whales, Southwestern and Southeastern Indian Ocean pygmy blue whales, fin whales and an unidentified baleen whale were detected at one or both sites. Odontocete clicks and whistles were also recorded at both sites. The discussion puts these marine mammal acoustic detections back into the context of their seasonal and geographical presence already described by other studies in the Indian Ocean and makes hypotheses about the role of the two studied areas for marine mammals
Long time behaviour for electromagnetic waves in dissipative Lorentz media
No full textInternational audienceA very general class of models for describing the propagation of waves in dispersive electromagnetic media is provided by generalized Lorentz models. In this work, we study the long time behaviour of the solutions of the dissipative version of these models
Kinematic interaction forces in rigid inclusions under seismic loading
No full textInternational audienceThe reinforcement technique with rigid inclusions is considered as a practical, economical, and time-saving foundation solution. Several studies have recently focused on the static and cyclic behaviour of this reinforced foundation system. However, its seismic behaviour has not yet been widely studied. This paper presents a state of the art in the dynamic soil-structure interaction studies on foundation reinforced with rigid inclusions. Special attention is given to the estimation of the kinematic bending moments of inclusions in different soil deposits comprising several stiffness contrasts. The kinematic bending moments calculated by a transient dynamic analysis are compared with those obtained by means of pseudo-static Winkler model approach. The results show that, compared to the dynamic approach, the pseudo-static approach is able to predict the location where the maximum bending moment develops but underestimates its amplitude for important values of the stiffness contrast at the corresponding interface between soil layers