Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
Dynamic response of shallow foundations on reinforced soil with rigid inclusions
No full textInternational audienceThe soil-improvement technique by rigid inclusions represents today a practical, economical and time-saving solution that allows the use of shallow foundations on compressible soils. Several studies have recently focused on the static behavior of this reinforcement system (i.e. ASIRI French national project), however, its seismic performance has not yet received the same amount of attention. This paper presents a state of the art in the dynamic soil-structure interaction (SSI) analysis of structures supported on shallow foundations on reinforced soil with rigid inclusions. Special attention is given to the calculation methods that are currently being used in design offices. A parametric study is conducted by means of 3D numerical models and the hybrid BEM-FEM method in order to highlight the possibilities and limitations of these methods, especially for the study of inertial and kinematic interaction effects. Different configurations of rigid inclusions are studied, including several lengths, spacing and foundation layout. The results of this parametric study are compared to those obtained for equivalent shallow and deep foundations in terms of impedance functions and kinematic interaction factors. Finally, a discussion on different aspects that may control the dynamic linear and non-linear response of these foundation systems is undertaken
Nanotechnologie et assemblages collés : étude des performances mécaniques d'un adhésif dopé par des nanocharges
No full textAdhesively bonded joint is a joining technique that is increasingly in a request today in many fields such as the automotive, aerospace, and naval. The adhesive selection is an important parameter to guarantee the reliability and durability of an adhesively bonded joint. It is well known that the wrong design of a bonded joint, the wrong choice of adhesive, or even a poorly executed surface treatment can have dramatic consequences. In terms of performance, the incorporation of carbon-based nanoparticles into the adhesive improves the performance of bonded joints. In this study, DGEBA adhesive doped by three kinds of nanofillers is established. Indeed, the adhesive has been doped with carbon nanotubes (CNT), graphene nanoplatelets (GNP), and carbon black (CB) with different mass fractions (1wt.%, 2wt.%, and 5wt.%). The overall objective is to study the influence of the type, mass fraction, and dimensions of the nanofillers on the mechanical behavior of the adhesive and the bonded joint. The results showed that each type of adhesive reinforced with nanofillers has a good potential in terms of mechanical behavior under static and dynamic loadings. However, the presence of a high mass fraction of nanoparticles tends to lead to degradation compared to the neat adhesive due to the transformation in microstructure morphology and physicochemical interactions. In addition, to quantify the damage resistance of the nanofiller-doped adhesive and demonstrate the adhesion compatibility with various kinds of substrates; DCB and ENF tests were performed on aluminum/aluminum and composite/composite bonded joints. Besides, numerical models taking into account the damage of the doped adhesive were developed and validated with the experimental results. This study demonstrates that the type, mass fraction, size, and shape of nanofillers play an important role in improving the performance of the adhesively bonded joints.Le collage structural est une technique d’assemblage de plus en plus demandée aujourd’hui dans beaucoup de domaines comme l’automobile, l’aéronautique, l’aérospatial, l’industrie navale et ferroviaire. La qualité d’un adhésif est un important paramètre pour garantir la fiabilité et la durabilité d’un assemblage collé. On sait qu’une fausse conception d’un joint collé, un mauvais choix d’adhésif ou même un traitement de surface mal réalisé pourra avoir des conséquences spectaculaires voire dramatiques. En termes de performance, l'ajout de nanoparticules à base de carbone à l'adhésif peut contribuer à améliorer les performances des assemblages collés. Dans cette étude, des adhésifs renforcés par trois types de nanacharges sont développés. En effet, l'adhésif DGEBA a été renforcé avec des nanotubes de carbone (CNT), des nanoplaquettes de graphène (GNP) et du noir de carbone (CB) avec différentes fractions massiques (1wt.%, 2wt.% et 5wt.%). L'objectif global est l'étude de l'influence du type, de la fraction massique et des dimensions des nanocharges sur le comportement mécanique de la colle et de l’assemblage collé. Les résultats ont montré que chaque type d'adhésif renforcé par des nanocharges possède un bon potentiel en termes de comportement mécanique sous des sollicitations statiques et dynamiques. Cependant, la présence d'une fraction massique élevée de nanoparticules a tendance à entraîner une dégradation par rapport à l'adhésif pur en raison du changement de morphologie de la microstructure et des interactions physico-chimiques. Et afin de quantifier la résistance à l’endommagement de l’adhésif dopé par des nanocharges et démontrer la compatibilité de l'adhésion avec divers substrats, des essais DCB et ENF ont étaient réalisés sur des assemblages Aluminium/aluminium et Composite/composite. De plus, des modèles numériques avec prise en compte de l’endommagement de la colle dopée ont été développés et validés avec les résultats expérimentaux. Cette étude démontre que le type, la fraction massique, la taille et la forme des nanocharges jouent un rôle important dans l'amélioration des performances des assemblages collés
Accelerating Random Forest on Memory-Constrained Devices through Data Storage Reorganization
No full textInternational audienceRandom forests is a widely used classification algorithm. It consists of a set of decision trees each of which is a classifier built on the basis of a random subset of the training data-set. In an environment where the memory work-space is low in comparison to the data-set size, when training a decision tree, a large proportion of the execution time is related to I/O operations. These are caused by data blocks swap-in from the storage device to the memory work-space. In traditional methods, data blocks contain elements that will be accessed and processed at different moments, thus, the spatial locality is low. In this paper, we seek to reduce the number of data-blocks swap-in operations by enhancing spatial locality. The idea is to reorder dataset blocks such as each block contains elements that are likely to be accessed together. Our experiments show that this method reduces random forest build time by 55% to 89%
Sub-Cycle Control of Relativistic Plasma Mirror Dynamics
No full textInternational audienceWe report on carrier-envelope phase (CEP) effects on the emission of high-order harmonics and electron beams from plasma mirrors driven by relativistic-intensity near-single-cycle laser pulses at 1 kHz repetition rate
Near-single-cycle pulses generated through post-compression on FAB1 laser at ATTOLAB-Orme facility
No full textEOS Annual Meeting (EOSAM 2021): Topical Meeting (TOM) 13- Ultrafast Optical Technologies and Applications, 13-17 September 201, Rome (Italy). Virtual meetingInternational audienceGenerating high-energy few-cycle pulses is key in the study of light-matter interaction in the regime of high field physics. Attosecond science possess the necessary time resolution to study the underlying fundamental processes but requires repetitions rates on the order the kilohertz and stabilization of the Carrier-Envelope Phase. We present here a post-compression stage delivering 3.8fs pulses with 2.5mJ coupled to a Ti: Sa based 1 kHz TW-class laser which can deliver 17.8fs pulses with 350mrad shot to shot CEP noise. This is the first step towards high-energy few-cycle post-compression of the FAB laser at ATTOLAB-Orme
Variational Methods for Acoustic Radiation in a Duct with a Shear Flow and an Absorbing Boundary
No full textInternational audienceThe well-posedness of the acoustic radiation in a 2D duct in presence of both a shear flow and an absorbing wall described by the Myers boundary condition is studied thanks to variational methods. Without flow the problem is found well-posed for any impedance value. The presence of a flow complicates the results. With a uniform flow the problem is proven to be always of the Fredholm type but is found well-posed only when considering a dissipative radiation problem. With a general shear flow, the Fredholm property is recovered for a weak enough shear and the dissipative radiation problem requires to introduce extra conditions to be well-posed: enough dissipation, a large enough frequency and non-intuitive conditions on the impedance value
Offuscation transitoire pour la sécurité HLS : application à la sécurité du cloud, aux filigranes numériques et à la défense contre les virus type chevaux de Troie
No full textThe growing globalization of the semiconductor supply chain, as well as the increasing complexity and diversity of hardware design flows, have lead to a surge in security threats: risks of intellectual property theft and reselling, reverse-engineering and malicious code insertion in the form of hardware Trojans during manufacturing and at design time have been a growing research focus in the past years. However, threats during highlevel synthesis (HLS), where an algorithmic description is transformed into a lower level hardware implementation, have only recently been considered, and few solutions have been given so far. In this thesis, we focus on how to secure designs during behavioral synthesis using either a cloud-based or an internal but untrusted HLS tool. We introduce a novel design time protection method called transient obfuscation, where the high-level source code is obfuscated using key-based techniques, and deobfuscated after HLS at register-transfer level. This two-step method ensures correct design functionality and low design overhead. We propose three ways to integrate transient obfuscation in different security mechanisms. First, we show how it can be used to prevent intellectual property theft and illegal reuse in a cloud-based HLS scenario. Then, we extend this work to watermarking, by exploiting the side-effects of transient obfuscation on HLS tools to identify stolen designs. Finally, we show how this method can also be used against hardware Trojans, both by preventing insertion and by facilitating detection.La mondialisation croissante de la chaîne d'approvisionnement des semi-conducteurs, ainsi que la complexité et la diversité croissantes des flux de conception de matériel, ont entraîné une recrudescence des menaces de sécurité : risques de vol et de revente de propriété intellectuelle, de rétro-ingénierie et d'insertion de code malveillant sous la forme de chevaux de Troie pendant la fabrication et au moment de la conception ont fait l'objet d'une recherche croissante ces dernières années. Cependant, les menaces lors de la synthèse de haut niveau (HLS), où une description algorithmique est transformée en une implémentation matérielle de niveau inférieur, n'ont été envisagées que récemment, et peu de solutions ont été proposées jusqu'à présent. Dans cette thèse, nous nous concentrons sur la sécurisation des conceptions lors de la synthèse comportementale à l'aide d'un outil HLS basé sur le cloud ou interne, mais non fiable. Nous introduisons une nouvelle méthode de protection au moment de la conception appelée offuscation, où le code source de haut niveau est obscurci à l'aide de techniques basées sur des clés, et désobscurci après HLS au niveau du transfert de registre. Cette méthode en deux étapes garantit une fonctionnalité de conception correcte et une faible surcharge de conception. Nous proposons trois façons d'intégrer l'offuscation transitoire dans différents mécanismes de sécurité. Tout d'abord, nous montrons comment il peut être utilisé pour empêcher le vol de propriété intellectuelle et la réutilisation illégale dans un scénario HLS basé sur le cloud. Ensuite, nous étendons ce travail au filigranes numériques, en exploitant les effets secondaires de l'offuscation transitoire sur les outils HLS pour identifier les conceptions volées. Enfin, nous montrons comment cette méthode peut également être utilisée contre les chevaux de Troie matériels, à la fois en empêchant l'insertion et en facilitant la détection
Caractérisation objective de la transparence cornéenne par analyse d’images d’OCT clinique
No full textInternational audienc
