Archive ouverte de Centrale Lyon
Not a member yet
32420 research outputs found
Sort by
Green Insulation System for Transformers in MVDC and HVDC Energy Networks
No full textInternational audienceThe feasibility of using natural ester liquids as insulation materials for transformers in MVDC and HVDC neworks is investigated. A comparative study with mineral oil based insulation system is presented. The comparison includes the dielectric characteristics such as the permittivity and electrical conductivity of the liquids and their impregnated pressboard. The characterizations have been made at 30 o C and 70 o C using Novocontrol Dielectric Spectroscopy and Polarization and Depolarization Current measurements. The AC and DC electric field stress on the insulation materials is studied using a finite element method based simulation model. It can be observed that the AC stress in the pressboard is higher for ester liquid than for mineral oil. However, the DC field stress on pressboard is lower for ester liquid than on mineral oil. The study shows that ester and ester impregnated pressboards are suitable substitutes for mineral oil system.</div
Système déductif sans coupure pour la logique continue intuitionniste
No full textWe introduce and develop propositional continuous intuitionistic logic and propositional continuous affine logic via complete algebraic semantics. Our approach centres on AC-algebras, which are algebras of sup-preserving functions from to an integral commutative residuated complete lattice (in the intuitionistic case, is a locale). We give an algebraic axiomatisation of AC-algebras in the language of continuous logic and prove, using the Macneille completion, that every Archimedean model embeds into some AC-algebra. We also show that (i) satisfies exactly when is a locale, (ii) involutiveness of negation in corresponds to that in , and that (iii) adding those conditions recovers classical continuous logic. For each variant —affine, intuitionistic, involutive, classical —we provide a sequent style deductive system and prove completeness and cut admissibility. This yields the first sequent style formulation of classical continuous logic enjoying cut admissibility
TracePath: Modeling and Analyzing Competency Trajectories with Graph-Based Learning Analytics over a Hybrid Polystore
No full textInternational audienceA competency-based approach supported by personalized learning paths and prompt feedback accelerates skill development by continuously adapting to learners' needs and maintaining high levels of engagement. Capturing and understanding learner competency development through interaction data offers the potential for early intervention and optimized educational design, yet introduces challenges related to scalability and complexity. We present TracePath, a novel graph-based framework that models learner trajectories as directed graphs, where nodes correspond to competencies or learner states and edges denote transitions such as validation or rejection events. This approach uncovers common learning pathways, identifies bottlenecks, and supports predictive analytics. At the core, a generic metamodel formalizes Competency Transition Graphs (CTGs), enabling comprehensive graph-based analytics implemented over a hybrid polystore architecture that integrates both relational and NoSQL databases. Our design decouples data extraction from graph exploration, allowing efficient querying, clustering, and pattern matching to deliver timely and explainable learning insights. Empirical validation using real-world data from the écri+ ecertification project demonstrates TracePath's effectiveness in providing scalable, dynamic, and low-latency learning analytics to support personalized education.</div
Advancements in Autonomous Space Robotics at ONERA : Control Frameworks and Co-Design Strategies for On-Orbit Servicing and Assembly
No full textInternational audienceSpace Manipulator Systems (SMS) are becoming key in space exploitation and exploration, offering a versatile range of solutions from space debris capture to structure assembly. However, recent missions involving manipulators aboard satellites and space structures must deal with lightweight and large ele- ments that exhibit flexible behaviors. Despite the challenges posed by flexible elements, enhancing the autonomy of SMS remains crucial to ensure their viability as solutions. For the pre-design of the SMS, path-planning applications, or controller design, there is a necessity for methods to assess the couplings between the manipulator, the SMS base, and any flexible elements manipulated by the manipulator or attached to the base. Moreover, recently proposed control strategies have demonstrated a keen interest in developing model-based controllers, which advantageously provide an efficient utilization of actua- tors and mitigation of internal disturbances within the system. The presentation focuses on recent research conducted at ONERA to design and validate control strate- gies for SMS performing On-Orbit Servicing (OOS) in the presence of flexible structures. These strate- gies are analyzed and validated across varying levels of detail and realism. The first part of the presentation addresses the derivation of kinematic and dynamic models for a free- floating SMS with a flexible body attached at the end of a kinematic chain, using a Lagrangian formal- ism. The second part presents ONERA’s real-time simulation platform, specifically developed to enable the rapid design, prototyping, and testing of model-based control solutions at fine time scales. The platform’s software and hardware architecture ensures high-fidelity modeling of space robot dynamics and flexible structures, integrating visual environment models for state observation, computer vision processing, virtual sensor data fusion, and full robot control. Finally, the presentation will demonstrate the platform’s value and versatility in addressing current challenges in space exploration and exploitation. Applications will include the On-Orbit deployment of large flexible space structures, co-design of assembly scenarios considering Guidance, Navigation, and Control (GNC) constraints, and strategies for the safe capture of tumbling target
Méthodes d'explication basées sur la théorie des jeux pour les réseaux de neurones sur graphes
No full textGraph Neural Networks (GNNs) have achieved remarkable success across a wide range of applications. However, their complex architectures make them difficult to interpret, limiting their adoption in critical domains where transparency is essential. Although many explanation methods have been proposed for GNNs, most lack theoretical guarantees regarding the faithfulness and reliability of their outputs. Among these methods, game-theoretic approaches stand out by offering formal guarantees, grounded in well-established axioms. Most game-theoretic methods define the input components (e.g., nodes or edges) as players and compute their contributions in a cooperative game using solution concepts such as Shapley values or HN-values. However, the exact computation of contribution values is exponential in the number of players, making the use of exact solutions computationally infeasible for large graphs. Consequently, approximation methods are often employed. Nevertheless, as the size of the input graph increases, the quality of these approximations deteriorates, undermining the reliability of the explanations. To address these challenges, this thesis proposes two complementary solutions. First, we introduce a method called INSIDE-SHAP that uses activation rules, derived from hidden representations of GNNs, as players in the cooperative game. INSIDE-SHAP decouples the size of the player set from the size of the input graph, enabling more efficient computation of Shapley-based contributions without sacrificing theoretical guarantees. Second, we present a novel game-theoretic framework (ESPAM) based on alternative axioms, distinct from those underpinning Shapley values. This framework permits exact polynomial-time computation of contribution values, thereby eliminating the need for approximations while preserving structural dependencies within the graph. Furthermore, the proposed method is model-agnostic, enhancing its applicability across diverse GNN architectures. Finally, we demonstrate the practical utility of ESPAM by applying it to a complex GNN model for molecule-drug interaction prediction. Together, these contributions advance the development of scalable, faithful, and theoretically grounded explanations for GNN models.Les réseaux de neurones graphiques (GNNs) ont connu un succès remarquable dans un large éventail d'applications. Cependant, la complexité de leurs architectures les rend difficiles à interpréter, ce qui limite leur adoption dans des domaines critiques où la transparence est essentielle. Bien que de nombreuses méthodes d'explication aient été proposées pour les GNNs, la plupart ne disposent pas de garanties théoriques quant à la fidélité et à la fiabilité de leurs résultats. Parmi ces méthodes, les approches basées sur la théorie des jeux se distinguent en offrant des garanties formelles, fondées sur des axiomes bien établis. La plupart des méthodes issues de la théorie des jeux considèrent les composantes d'entrée (par exemple, les nœuds ou les arêtes) comme des joueurs, et calculent leur contribution dans un jeu coopératif en utilisant des concepts de solution tels que les valeurs de Shapley ou les HN-values. Toutefois, le calcul exact de ces contributions est exponentiel en fonction du nombre de joueurs, ce qui rend les solutions exactes irréalisables d'un point de vue computationnel pour les grands graphes. Par conséquent, des méthodes d'approximation sont souvent employées. Néanmoins, à mesure que la taille du graphe d'entrée augmente, la qualité de ces approximations se dégrade, compromettant ainsi la fiabilité des explications. Pour relever ces défis, cette thèse propose deux solutions complémentaires. Premièrement, nous introduisons une méthode appelée INSIDE-SHAP, qui utilise des règles d'activation, dérivées des représentations cachées des GNNs, comme joueurs dans le jeu coopératif. INSIDE-SHAP découple la taille de l'ensemble des joueurs de celle du graphe d'entrée, permettant ainsi un calcul plus efficace des contributions basées sur les valeurs de Shapley, sans compromettre les garanties théoriques. Deuxièmement, nous présentons un nouveau cadre théorique, ESPAM, basé sur des axiomes alternatifs, distincts de ceux qui sous-tendent les valeurs de Shapley. Ce cadre permet un calcul exact des valeurs de contribution en temps polynomial, éliminant ainsi le besoin d'approximation tout en préservant les dépendances structurelles au sein du graphe. De plus, la méthode proposée est indépendante du modèle, ce qui renforce son applicabilité à diverses architectures de GNNs. Enfin, nous démontrons l'utilité pratique de ESPAM en l'appliquant à un modèle GNN complexe pour la prédiction des interactions molécule-médicament. Ensemble, ces contributions font progresser le développement de méthodes d'explication évolutives, fidèles et théoriquement fondées pour les modèles de GNN
Steady state large deviations for one-dimensional, symmetric exclusion processes in weak contact with reservoirs
No full textInternational audienceConsider the symmetric exclusion process evolving on an interval and weakly interacting at the end-points with reservoirs. Denote by its dynamical large deviations functional and by the associated quasi-potential, defined as V(\gamma) = \inf_{T>0} \inf_u I_{[0,T]} (u), where the infimum is carried over all trajectories such that , , and is the stationary density profile. We derive the partial differential equation which describes the evolution of the optimal trajectory, and deduce from this result the formula obtained by Derrida, Hirschberg and Sadhu \cite{DHS2021} for the quasi-potential through the representation of the steady state as a product of matrices
Schubert polynomial expansions revisited
No full text29 pages, 7 figuresInternational audienceWe give an elementary approach utilizing only the divided difference formalism for obtaining expansions of Schubert polynomials that are manifestly nonnegative, by studying solutions to the equation on polynomials with no constant term. This in particular recovers the pipe dream and slide polynomial expansions. We also show that slide polynomials satisfy an analogue of the divided difference formalisms for Schubert polynomials and forest polynomials, which gives a simple method for extracting the coefficients of slide polynomials in the slide polynomial decomposition of an arbitrary polynomial
Effect of the aspect ratio and wall heterogeneities on the mechanical behaviour of the aneurysm wall: Experimental investigation on phantom arteries
No full textInternational audienc
Infinite-Dimensional Flats in the Space of Positive Metrics on an Ample Line Bundle
No full textInternational audienceAbstract We show that any continuous positive metric on an ample line bundle lies at the apex of many infinite-dimensional Mabuchi-flat cones. More precisely, given any bounded graded filtration of the section ring of , the set of bounded decreasing convex functions on the support of the Duistermaat–Heckman measure of embeds -isometrically into the space of bounded positive metrics on with respect to Darvas’ distance for all , and in particular with respect to the Mabuchi metric ()
