Portail HAL Ensta
Not a member yet
11080 research outputs found
Sort by
Influence of Carbon Nanotube Defects on the Elastic Modulus of Nanocomposite: Multiscale Simulation
No full textInternational audienceIn the present study, atomistic-based continuum modeling of single-walled carbon nanotube (SWCNT)-reinforced composite had been performed to predict the effective modulus of the nanocomposite through multiscale modeling. SWCNT of armchair type (10,10) was modeled with a three-dimensional beam element to estimate equivalent axial Young's modulus using finite element analysis (FEA) for pristine and defective CNTs. The equivalent modulus of CNT with 5% vacancy defects has been reduced by 35.35% in comparison to the pristine CNTs. In addition, molecular dynamics (MD) simulations were conducted on standalone CNT for pristine and defective CNTs. Further, long and short CNT-reinforced nanocomposites were modeled through interphase FE modeling using Representative Volume Element (RVE) which is comprised of three phases, i.e., SWCNT, interphase, and matrix. Finally, the effects of CNT defects, interphase property, modulus ratio, height fraction, and volume fraction on the overall elastic response of the nanocomposite were demonstrated by statistical analysis. The analytical expressions of effective elastic modulus of both short and long CNT-based nanocomposites have been derived through regression analysis
Modelling of the fatigue cracking resistance of grid reinforced asphalt concrete by coupling fast BEM and FEM
No full textInternational audienceWe present a computational modeling approach aimed at investigating the effect of fiber grid reinforcement on crack opening displacement and fatigue crack propagation. Grid reinforcements are modeled using elastic membrane finite elements, while the cracked concrete is treated using a symmetric boundary element method (BEM), which in particular allows easy geometrical modelling and meshing of cracks. The BEM is accelerated by the fast multipole method, allowing the handling of potentially large BEM models entailed by three-dimensional configurations hosting multiple cracks. Fatigue crack growth is modelled using the Paris law. The proposed computational approach is first verified on a reinforced cracked beam, and then applied to a three-dimensional configuration featuring a grid-reinforced asphalt pavement
Multimodal deep learning for cetacean distribution modeling of fin whales (Balaenoptera physalus) in the western Mediterranean Sea
No full textInternational audienceCetacean Distribution Modeling (CDM) is used to quantify mobile marine species distributions and densities. It is essential to better understand and protect whales and their relatives. Current CDM approaches often fail in capturing general species-environment relationships, which would be valid within a broader range of environmental conditions that characterize the surveyed regions. This paper aims at investigating the usefulness of deep learning based schemes, namely multi-task and transfer learning, in CDM. Co-training of a stochastic presence-background model on a classification task and a deterministic rule-based model on a regression task was performed. Whale presence-only records were used for the first task, and index outputs of a feeding habitat occurrence model for the second one. This new approach has been experimented through the study case of fin whales in the western Mediterranean Sea. To evaluate our approach, a new metric called True Positive rate per unit of Surface Habitat (TPSH) and an original multimodal fully-connected neural networks were developed. A Generalized Additive Model (GAM)—a standard CDM method—was also used as a reference for performance. Results show that our multi-task learning model improves both the feeding habitat model by 10.8% and data-driven models such as GAM by 16.5% on our TPSH metric in relative terms, revealing a higher accuracy of our approach in estimating whale presence. Such trends in results have been further supported by the use of two other independent datasets that forced models to generalize beyond their training dataset of species-environment relationships. Performance could be further improved by adopting more optimal thresholds as observed from Receiver Operating Characteristic curves, e.g. the multi-task learning model could reach absolute gains up to 10% in the median of the True Positive Rate while maintaining its habitat spatial spreading. Globally, our work confirmed our working hypothesis that expert information on whale behaviour represent a good knowledge base for model generalization. This result can be further improved by a concurrent learning of more local species-environment relationships from in-situ presence data
Probing strong-field QED in beam-plasma collisions
No full textInternational audienceOngoing progress in laser and accelerator technology opens new possibilities in high-field science, notably for the study of the largely unexplored strong-field QED regime where electron-positron pairs can be created directly from light-matter or even light-vacuum interactions. Laserless strategies such as beam-beam collisions have also been proposed with the prospect of pushing strong-field quantum electrodynamics (SFQED) in the nonpertubative regime. Here we report on an original concept to probe strong-field QED by harnessing the interaction between an electron beam and a solid target. When a high-density, ultrarelativistic beam impinges onto an even denser plasma, the beam self fields are reflected at the plasma boundary. In the rest frame of the beam electrons, these fields can exceed the Schwinger field, leading to strong-field QED effects such as quantum nonlinear inverse Compton scattering and nonlinear Breit-Wheeler electron-positron pair creation. We show that such beam-plasma collisions can produce results similar to beam-beam collisions with the advantage of a much simpler experimental setup including the automatic overlap between the beam and the reflected fields. This scenario opens the way to precision studies of strong-field QED, with measurable clear signatures in terms of gamma-ray photon and pair production, and thus is a very promising milestone on the path towards laserless studies of nonperturbative SFQED
Influence of infill density on the dynamic behavior of 3D-printed CF-PEKK composites using split Hopkinson’s pressure bars
No full textInternational audienceThe dynamic mechanical behavior of 3D-printed CF-PEKK composites has been studied using the split Hopkinson pressure bar (SHPB). Three sets of samples with different infill densities (20, 50, and 100%) were impacted at different impact pressures (1.4, 1.7, 2.0, 2.4 bar) to characterize their dynamic behavior in terms of strain rate, stress, and strain performance. Moreover, the samples’ dynamic response and failure behavior are cross-referenced with each infill density sample's microstructural behavior and physical parameters, which can give an excellent dynamic response under different strain rate ranges. The results showed that dynamic stress and strain behavior are proportional to the infill density with a maximum dynamic stress of approximately 90 MPa with 100% infill density. In addition, the behavior of the transmitted wave during the dynamic impact showed that a sample with less infill density could attenuate/absorb dynamic impact. These results obtained under this impact pressure indicate the high precision and repeatability of the SHPB approach for the tested 3D composites under dynamic loading and prove that the striker bar velocity significantly impacts the amplitudes of different recorded waves to understand the behavior in detail. In addition, high-speed camera images also showed that the 3D-printed composite sample showed that using a higher infill density enhanced both the damage performance and the impact resistance of the three-dimensionally printed composites at each impact pressure because of the minimized intensity and the amount of final macro damage. This study is useful and helpful for design engineers to study the influence of the process parameter of infill on the dynamic behavior of printed composite materials and confirms the growing interest in 3D printing of composite materials to be employed in different engineering applications which are not limited to static or quasi-static loading circumstances but mainly include dynamic loading conditions
Piezoresistive Behavior of a Conductive Polyurethane Based-Foam for Real-Time Structural Monitoring
No full textInternational audienceSmart flexible materials with piezoresistive property are increasingly used in the field of sensors. When embedded in structures, they would allow for in situ structural health monitoring and damage assessment of impact loading, such as crash, bird strikes and ballistic impacts; however, this could not be achieved without a deep characterization of the relation between piezoresistivity and mechanical behavior. The aim of this paper is to study the potential use of the piezoresistivity effect of a conductive foam made of a flexible polyurethane matrix filled with activated carbon for integrated structural health monitoring (SHM) and low-energy impact detection. To do so, polyurethane foam filled with activated carbon, namely PUF-AC, is tested under quasi-static compressions and under a dynamic mechanical analyzer (DMA) with in situ measurements of its electrical resistance. A new relation is proposed for describing the evolution of the resistivity versus strain rate showing that a link exists between electrical sensitivity and viscoelasticity. In addition, a first demonstrative experiment of feasibility of an SHM application using piezoresistive foam embedded in a composite sandwich structure is realized by a low-energy impact (2 J) test
Robust capacitated Steiner trees and networks with uniform demands
No full textInternational audienceWe are interested in the design of robust (or resilient) capacitated rooted Steiner networks in the case of terminals with uniform demands. Formally, we are given a graph, capacity, and cost functions on the edges, a root, a subset of vertices called terminals, and a bound k on the number of possible edge failures. We first study the problem where k=1 and the network that we want to design must be a tree covering the root and the terminals: we give complexity results and propose models to optimize both the cost of the tree and the number of terminals disconnected from the root in the worst case of an edge failure, while respecting the capacity constraints on the edges. Secondly, we consider the problem of computing a minimum-cost survivable network, that is, a network that covers the root and terminals even after the removal of any k edges, while still respecting the capacity constraints on the edges. We also consider the possibility of protecting a given number of edges. We propose three different formulations: a bilevel formulation (with an attacker and a defender), a cutset-based formulation and a flow-based one. We compare the formulations from a theoretical point of view, and we propose algorithms to solve them and compare their efficiency in practice
Estimation efficace des incertitudes avec Packed-Ensembles
No full textInternational audienceDeep Ensembles (DE) are a prominent approach for achieving excellent performance on key metrics such as accuracy, calibration, uncertainty estimation, and out-of-distribution detection. However, hardware limitations of real-world systems constrain to smaller ensembles and lower-capacity networks, significantlydeteriorating their performance and properties. We introduce Packed-Ensembles(PE), a strategy to design and train lightweight structured ensembles by carefully modulating the dimension of their encoding space. We leverage grouped convolu-tions to parallelize the ensemble into a single shared backbone and forward pass to improve training and inference speeds. PE is designed to operate within the memory limits of a standard neural network. Our extensive research indicates that PE accurately preserves the properties of DE, such as diversity, and performs equally well in terms of accuracy, calibration, out-of-distribution detection, and robustness to distribution shift. We make our code available at github.com/ENSTA-U2IS/torch-uncertainty.Deep Ensembles (DE) est une des approches principales pour obtenir d’excellentes performances sur des tâches de classification et de détection de données hors distribution. En pratique cependant, ces ensembles et leurs membres sont généralement petits, et donc peu performants, car limités par les contraintes matérielles. Nous proposons Packed-Ensembles (PE), une méthode pour définir et entraîner des ensembles plus légers en modulant la dimension des espaces de représentation. Nous utilisons des convolutions groupées pour paralléliser l’ensemble en une unique structure et ainsi améliorer les temps d'entraînement et d’inférence. PE a la capacité de fonctionner avec une empreinte mémoire équivalente à celle d’un réseau de neurones classique. Au travers d’études approfondies, nous montrons que PE préserve les propriétés de DE, telles que la diversité, et atteint des résultats similaires sur toutes les métriques de précision et de quantification d’incertitude. Le code associé au papier est disponible sur \url{https://github.com/ENSTA-U2IS/torch-uncertainty}
Analysis of sampling methods for imaging a periodic layer and its defects
No full textInternational audienceWe revisit the differential sampling method introduced in [9] for the identification of a periodic domain and some local perturbation. We provide a theoretical justification of the method that avoids assuming that the local perturbation is also periodic. Our theoretical framework uses functional spaces with continuous dependence with respect to the Floquet-Bloch variable. The corner stone of the analysis is the justification of the Generalized Linear Sampling Method in this setting for a single Floquet-Bloch mode
The Knowledge-based approach to defense
No full textInternational audienceThe knowledge-based approach to innovation makes possible to consider the proximities between companies from the original angle of knowledge rather than from the angle of the industrial sector, which is too restrictive to study defense. Indeed, the production of armaments is diverse, ranging from ships and submarines to missiles and satellites