HAL Arts et Métiers
Not a member yet
14127 research outputs found
Sort by
Structural and Morphological Study of Thermally Fractionated P(VDF- co -TrFE) Ferroelectric Copolymers
No full textInternational audienceThermal fractionation of P(VDF-co-TrFE) copolymers with VDF/TrFE ratios ranging from 85/15 to 75/25 was achieved by applying Successive Self-nucleation and Annealing (SSA) protocol. This method, scarcely applied to VDF-based semi-crystalline polymers, involves multiple annealing isotherms at decreasing temperatures, initiated within the self-nucleation domain. Fractionation was observed for all the samples, but its expression on the Differential Scanning Calorimetry (DSC) scans varied depending on the copolymer chemical composition. The Curie transition was found to be significantly influenced by the fractionation of 80/20 and 75/25 samples. A structural study revealed an increase in both the crystallinity and the ferroelectric phase fraction after SSA, consistent with the enhancement of the remanent polarization in polarization-electric field loops. Furthermore, the formation of a distribution of crystalline lamellae with different thicknesses was highlighted by in situ Small-Angle X-ray Scattering (SAXS) analysis, which revealed the progressive melting of increasingly thicker crystalline lamellae during the heating ramp
Impact of MVDC Link Integration in Distribution Networks: A CIGRE Benchmark-based Approach
No full textInternational audienceIn recent years, the power generation and demand paradigm has evolved, driven by the increasing integration of distributed energy resources and renewable technologies. While these advancements introduce new functionalities to conventional electrical infrastructure, they also strain the system, exposing its limitations. In this context, Medium Voltage Direct Current (MVDC) technology is emerging as a promising solution, offering advantages in efficiency, scalability, and flexibility. This paper investigates the integration of an MVDC link within an existing distribution system, assessing its potential to address key grid challenges such as load balancing, cable loss reduction, and enhanced transmission efficiency
Thermal ageing of PP-EPDM thermoplastic vulcanizate: Multiscale study and kinetic modeling
No full textInternational audienceThis work presents a multi-scale study of the impact of thermal ageing of thermoplastic vulcanizate (TPV) obtained by blending a thermoplastic matrix (polypropylene), plasticizer, and dispersed crosslinked EPDM rubber particles. This study focuses on the case of TPV without pigments or stabilizers. The thermal ageing was mainly monitored by Fourier transform infrared and thermal analysis, to identify oxidation products and their kinetics of appearance, and GPC analysis performed on extracted PP fraction of the TPV. Data indicated that EPDM phase oxidizes first which drives degradation of the TPV. Appearance changes were characterized by colorimetry and gloss. Oxidation leads to strong changes of mechanical properties and aspect properties observed from the early hours of ageing, yellowing (change of Δb*) being also induced by the plasticizer. Yellowing reached unacceptable level (regarding the requirements for exterior automotive parts) before mechanical properties. The oxidation of EPDM and TPV was thus simulated by a kinetic model based on kinetic parameters already available in the literature. Its validity was verified by its capability to simulate the appearance of primary (hydroperoxides) and secondary (carbonyls) oxidation products
3D Printing Process and Local Preheating Technique Study Using a Laser‐Based Method
No full textInternational audienceThis study addresses the challenge of enhancing inter‐filament bonding in Fused Filament Fabrication (FFF) 3D printing, where conventional methods often result in suboptimal mechanical strength or structural distortions. A novel local preheating technique was introduced using a laser‐based approach to precisely control the temperature of previously deposited filaments. An experimental setup integrating a 3D printer with a 30 W diode laser and an optical system was developed to evaluate key parameters, including laser power, PLA color effects on laser absorptivity, and the relationship between preheating temperature and bonding strength. Results demonstrate that preheating to 180°C increases mechanical bonding strength by up to 35% without distortion, highlighting the technique's potential to improve FFF part quality
Méthode volumes-finis basée sur des techniques d’apprentissage automatique pour la dynamique des fluides
No full textNonlinear hyperbolic partial differential equations (PDEs) are ubiquitous in mathematical physics to represent complex phenomena like fluids mechanics, wave propagation, electromagnetism, etc. Solutions to nonlinear hyperbolic conservation laws are characterized by the propagation of wave-like structures at finite speed and whose solutions may develop discontinuities. Their approximation faces important issues associated to the low regularity of the solution, or to some underlying physical principles that need to be satisfied at the discrete level such as entropy stability.Machine Learning (ML) offers promising alternatives offering efficient scaling and generalization across parameters or boundary conditions for the approximation of solutions to PDE problems. Neural networks, as universal function approximators, have been leveraged in various forms to accelerate simulations, discover governing equations, or hybridize with physics-based models. Yet these models suffers stability issues when applied to hyperbolic PDEs, and often lacks hard-constraints to meet physical constraints.Recent research explores hybrid solvers embedding Machine Learning within existing methods like the finite volume method (FVM) specifically for hyperbolic PDEs. Applications include shock capturing, flux limiting, artificial viscosity, and flux prediction, with extensions to unstructured meshes. While progress is evident, ML-enhanced finite volume solvers face key limitations balancing accuracy, stability, and computational efficiency while guaranteeing conservation laws.This thesis addresses these gaps by developing an interpretable and entropy-consistent sub-grid model with super-resolution properties within finite volume schemes. The approach integrates a learned spatial discretization into a limited second-order FV methods, augmented with regularizers to ensure stability. The method reproduces fine-grid results on coarse meshes while maintaining physical constraints, thus advancing ML-augmented solvers toward practical robustness.In the context of a collaboration, a hard constrained ML solver for Hamiltonian PDEs is also presented. In this work, the known self-attention mechanism of the transformer architecture is adapted to be volume-preserving, a wanted property for divergence-free systems.Les équations aux dérivées partielles (EDPs) hyperboliques non linéaires jouent un rôle prépondérant en physique pour modéliser différents phénomènes complexes, tels que les fluides ou encore l'électromagnétisme. Dans le cadre de l'étude de systémes hyperboliques, il est important de noter que, quelle que soit la régularité des conditons initiales ou aux limites utilisées, des discontinuités peuvent apparaître après un temps fini. Par conséquent, il est essentiel d'étudier ces EDPs au sens des distributions, où l'on recherche des solutions faibles. Les solutions dites faibles ne sont pas uniques. Il est alors nécessaire de compléter ces équations par des conditions d'admissibilité supplémentaires afin de sélectionner l'unique solution physique. Ces conditions peuvent prendre la forme d'une inégalité entropique.Tandis que les calculs numériques visent à supplanter différentes démarches expérimentales, ils deviennent de plus en plus lourds en termes de temps de calcul afin de capturer toutes les différentes échelles. L'apprentissage automatique propose des alternatives prometteuses permettant une mise à l'échelle et une généralisation efficace pour l'approximation de solutions EDP. Les réseaux neuronaux, en tant qu'approximateurs de fonctions universels, ont été utilisés sous diverses formes afin d'accélérer les simulations, découvrir des équations régissant certains phénomènes ou s'hybrider avec des modèles basés sur la physique. Cependant, ces modèles sont sujets à des problèmes de stabilité, attribuables à la présence de discontinuités dans les solutions de systémes hyperboliques. De plus, ces méthodes sont souvent dépourvues de contraintes strictes pouvant répondre à certaines contraintes physiques.Des recherches récentes explorent des solveurs hybrides intégrant l'apprentissage automatique dans des schémas numériques existants telles que la méthode des volumes finis (FVM), en particulier pour les équations différentielles aux dérivées partielles hyperboliques. Lesdites applications englobent la capture des chocs, la limitation des flux, la viscosité artificielle ou la prédiction de flux, avec des extensions aux maillages non structurés. Bien que des progrès notables aient été réalisés, les solveurs volumes finis améliorés par l'apprentissage automatique sont confrontés à des limitations majeures lorsqu'il s'agit d'équilibrer précision, stabilité et efficacité computationnelle tout en garantissant certaines lois de conservation.Cette thèse vise à combler ces lacunes en développant un modèle interprétable de sous-maille cohérent avec une inégalité d'entropie et doté de propriétés de super-résolution dans le cadre de schémas à volumes finis. Cette approche méthodologique intégre une discrétisation spatiale apprise dans un schéma volumes finis d'ordre deux, afin de garantir la stabilité du système. La méthode en question reproduit les résultats obtenus sur des maillages fins sur des maillages grossiers, tout en conservant les contraintes physiques.Dans le cadre d'une collaboration, un solveur machine learning pour les EDPs hamiltoniennes est également présenté. Lors de cette étude, le mécanisme d'auto-attention, venant de l'architecture du transformer, est adapté afin de préserver le volume, une propriété recherchée dans le contexte des systèmes hamiltoniens
Mapping Existing Modelling Approaches to Maritime Decarbonisation Using Latent Dirichlet Allocation
No full textInternational audienceWhile for a long time reluctant to take action over the climate emergency at hand, the maritime shipping industry is now addressing the pressing need to decarbonise. Within this context, numerous modelling approaches and associated tools have emerged, with the aim of either reducing shipping emissions directly or facilitating decision-making around the sector's transition. This paper explores the use of topic modelling-specifically Latent Dirichlet Allocation (LDA)-as a means of identifying the trends in these existing modelling approaches to maritime decarbonisation. The use of topic modelling is proposed as a means of overcoming challenges inherent to both the chosen field of study and wider shipping industry, namely significant heterogeneity and fragmentation. LDA is shown to provide an effective means of mapping this particular research field, with four topics identified as principal thematic trends. The results obtained may serve to ascertain where future research in sustainable shipping can most effectively intervene
Data-Driven Models for Alert Management in Infrastructure Monitoring Within Their Environment
No full textInternational audienceThis study investigates the implementation of alert systems for critical infrastructures within their operational environment. The challenge arises from the extensive scale of these infrastructures, the heterogeneity and complexity of the data sources, and the imperative for real-time responsiveness. To address these challenges, we propose data-driven methodologies that demonstrate the feasibility of such systems. Specifically, we examine two complementary case studies: 1) the detection of sinkholes in railway networks and 2) the development of a metamodel for seismic risk assessment in dam infrastructure. These case studies illustrate the pivotal role of artificial intelligence and predictive analytics in enhancing alert systems and facilitating proactive risk management. The proposed approach contributes to the field of critical infrastructure monitoring by integrating advanced data-driven strategies, thereby reinforcing its scientific relevance and novelt
A Review of Methods and Data on the Recycling of Plastics from the European Waste Stream of Electric and Electronic Equipment
No full textInternational audiencePlastics make up a significant proportion of the stream of the European Waste of Electric and Electronic Equipment (WEEE), yet the use of recycled plastic materials is very low in new manufactured products. A description of the WEEE waste stream in Europe is given, with a focus on the plastic materials commonly found in WEEE that include four principal polymers: polypropylene (PP), polycarbonate (PC), acrylonitrile-butadiene-styrene (ABS) and polystyrene (PS). Furthermore, the legislative aspects related to WEEE and plastics recycling in Europe are complex, and numerous norms have been dictated by the European Commission. These norms are crucial to the sector of polymer recycling and production in Europe. Moreover, an overview of the entire treatment chain is presented. More specifically, each step of a typical recycling chain is introduced, with a focus on the sorting of plastics and the separation of polymers. Lastly, the influence of contaminants in the plastic fraction is discussed, both in terms of polymer particles and unwanted additives. By showing the impact of the purity rate on the mechanical properties of recycled plastics, the consequences of inadequate end-of-life treatment for WEEE-plastics is highlighted, hence linking the quality of recycled plastics to the separation step and the re-compounding of recycled granulates
Détection numérique de la délamination dans les matériaux composites stratifiés par analyse des vibrations modales
No full textce papier de conférence entre dans la thématique des matériaux composites. on essai de trouver un outil numérique (analyse modale) pour contrôler les échantillons du fibres de verre et détecter les défauts internesInternational audienceDétection numérique de la délamination dans les matériaux composites stratifiés par analyse des vibrations modale
Screen Printed Piezoelectric Transducers for Structural Health Monitoring of Curved Thick Composite Panels
No full textSHM : Structural Health Monitoring; GAPSUS - Acoustique Physique, Sous-Marine et Ultra-SonoreNational audienceThis research focuses on the development and experimental validation of a novel printed piezoelectric transducers network employed on a foreign object damage panel substructure of an aircraft engine fan blade. The main goal of the work is to leverage the screen printing technology to fabricate arrays of piezoelectric transducers and ultimately employ these transducers for operations, enabling the development of structural health monitoring methods for the panel. The printed transducer is made up of a piezoelectric layer sandwiched between two silver electrodes, each printed in a controlled manner. Upon printing and drying of the layers, the transducers undergo polarization. The electromechanical behaviour of the printed transducers, characterized using impedance measurements, exhibits high repeatability, thus indicating its potential for large scale industrial deployment. Following this, it is demonstrated that the transducers are capable of accurately sensing impact, which is one the most common yet critical sources of damage to an engine fan blade. It is also shown that the printed transducers are able to detect acoustic emission events. The ability of the printed transducers to actuate and sense guided wave signals over a range of ultrasonic frequencies is also demonstrated. Furthermore, apart from the noticeable advantages of the non-intrusive nature, and negligible weight as compared to their traditional ceramic counterparts, the printed piezoelectric transducers can potentially be integrated into the manufacturing process in the future, and the presence of transducer arrays ensures the availability of other transducers in case of an individual failure during service. This innovative printing technology for PZT transducer networks thus holds significant promise in bridging the gap between research advancements and the industrial implementation of SHM technology