Portail HAL UHA (Université de Haute-Alsace)
Not a member yet
17715 research outputs found
Sort by
Numerical modeling and simulation prediction of the forming process of 3D-tubular braided composite reinforcements
No full textInternational audienceTubular braids with a hollow structure are considered to be ideal reinforcements for manufacturing composite pipes and cylindrical structures. The mechanical properties of the fabric are critical in the forming process. A new simulation approach based on the non-orthogonal hyperelastic constitutive model for predicting the mechanical behavior of tubular braided reinforcements during forming was proposed. The validity and accuracy of this approach were certified by the uniaxial tensile test of tubular braided fabrics and the maximum errors are <15 %. In addition, preforming of fabrics on the tetrahedrons and cylinders has been investigated. The predicted shear angle and elongation of a single yarn of the tubular fabric after forming showed good agreement with experimental results. The maximum errors of shear angle and elongation of a single yarn are 10.4 % and 9.8 %, respectively. Furthermore, the formability of the fabric was also predicted to achieve damage-free preforms. This work thus gives a novelty perspective to further guide tubular braided fabric processing during the forming process
“This is a strange thing as e’er I look’d on”: Gustave Doré (1832-1880) et The Tempest »
No full textInternational audienc
Recent achievements in the synthesis and reactivity of pentafluorosulfanyl-alkynes
No full textInternational audienceIn the dynamic field of pentafluorosulfanyl (SF5) chemistry, the SF5-alkynes have emerged as essential, readily accessible, and modular building blocks for the construction of a wide variety of SF5-containing molecules. This polarized platform has been used to perform highly regio-, chemo-, and stereoselective transformations such as heterocycle synthesis, cycloaddition, and hydroelementation reactions. This brief review provides an overview of recent developments in the synthesis and reactivity of SF5-alkynes, which are fascinating building blocks that have yet to reveal their full synthetic potential
Cutting parameters optimization in edge trimming of UD-GFRP composites with diamond coated burr tools
No full textInternational audienceEdge trimming operations in the manufacturing process of fiber-reinforced polymers (FRP) may lead to various defects on machined parts. Burr tools, which consist of many pyramidal teeth, contribute to reduce such defects occurrence. Previous studies dealing with cutting parameters optimization in FRP milling with this kind of tool mainly focused on multidirectional composites. However, fiber orientation has a critical impact on FRP cutting mechanisms. Thus, this study investigates cutting parameters effect on cutting forces, machined surface temperature and machining quality for the two main FRP cutting mechanisms, using unidirectional composites. The results show that the impact of cutting parameters is affected by the cutting mechanism. Overall, radial engagement has the greatest effect on all the responses studied in this work, while feed per revolution and cutting speed have a much smaller impact
3D‐Printed Eosin Y‐Based Heterogeneous Photocatalyst for Organic Reactions
No full textInternational audienceHeterogenization of Eosin Y by 3D‐printing and its application in photocatalysis are reported. The approach allows a fine tuning of the photocatalyst morphology and its rapid preparation. Photocatalytic activity was evaluated through model organic reactions involving oxidation, reduction, and photosensitization pathways. The efficiency, recyclability and stability of 3D printed EY is remarkable paving the way to new generation of heterogeneous photocatalysts with a perfect control of their shape and adaptable to any photoreactors
Conception d'un système de perception robuste pour les véhicules autonomes associant l'apprentissage profond et la fusion de données
No full textThe thesis investigates the combination between deep learning and Evidence theory, an extended formalism for uncertainty representation. Its application is environment perception in autonomous vehicles. A baseline camera-LiDAR fusion architecture called cross-fusion is chosen for the task of road detection. It is an encoder-decoder deep learning network. In a first step, the thesis proposes an architecture reduction to satisfy power constraints in embedded system applications. The reduction is guided by the weight analysis inside the network. Several reduction strategies leading to two optimized networks are highlighted. The reduced networks show comparable performance with the baseline while saving computational space. In a second step, these models are coupled with Evidence theory. The evidential formulation is based on a distance to prototype approach. The evidential models are evaluated on real data considering several driving situations from the KITTI dataset and show better performance than the corresponding probabilistic ones: evidential models that are smaller, faster, more informative and better in performance than the baseline are obtained. However, prototype-based basic belief assignments show a high degree of conflict in some region of the road scene. The source of the conflict is identified and a rectification is proposed.La thèse étudie la combinaison entre les architectures d'apprentissage profond et la théorie des fonctions de croyance, formalisme approprié à la représentation étendue des imperfections des données. Son application est la perception de l'environnement pour les véhicules autonomes. Dans ce contexte, un réseau de référence de fusion caméra-LiDAR appelé fusion croisée est choisi pour la détection de routes. Il s'agit d'un réseau d'apprentissage profond basé sur un encodeur-décodeur. Dans un premier temps, l'architecture est réduite afin de répondre à des contraintes de puissance de calcul dans les applications de systèmes embarqués. Les méthodes de réduction sont basées sur l'analyse des poids internes au réseau. Les architectures réduites montrent des performances comparables à celles du réseau initial tout en réduisant la compléxité combinatoire. Dans un second temps, ces réseaux sont couplés à la théorie des fonctions de croyances par l’intégration de couches évidentielles. Les fonctions de masses sont déterminées par l’approche de la distance à un prototype. Les réseaux évidentiels sont ensuite évalués sur des données réelles du jeux de données KITTI et montrent de meilleures performances que les modèles probabilistes correspondants : ces réseaux sont à la fois plus petits, plus rapides, plus informatifs et plus performants que l’architecture initiale. Cependant, les fonctions de masses associées aux prototypes montrent un degré élevé de conflit dans certaines régions des scènes routières. La source du conflit est identifiée et une rectification est proposée
Synthesis of Pentafluorosulfanylated Ynamides and Further Functionalizations
No full textInternational audienceHerein is described the first synthesis of SF5-ynamides, versatile building blocks featuring the pentafluorosulfanyl motif. This synthesis proceeds through a two-step sequence of radical SF5-addition onto the π-system of a wide range of terminal ynamides and derivatives substituted with various nitrogen fragments followed by a dehydrochlorination reaction. A selection of downstream functionalization reactions, including nucleophilic additions, cycloadditions, and sulfur ylide synthesis, highlights the wide-ranging applications of this novel type of functionalized SF5-building block