HAL-CentraleSupelec
Not a member yet
77624 research outputs found
Sort by
Proceedings of the AI WILD West 2026
No full textInternational audienceArtificial Intelligence Workshop on Inverse problem, Learning, imaging & Data science in the West (AI WILD West) aims to bring together researchers working at the interface of inverse problems, artificial intelligence, and applied mathematics. The workshop focuses on methodological advances and their applications in translational research, imaging and data science. By fostering interactions between theory and practice, the workshop encourages cross-disciplinary exchanges between mathematics, machine learning, and application-driven research communities
Light-round ignition dynamics of a hydrogen-fueled annular combustor: Parametric effects and reduced-order modeling
No full textInternational audienceOne issue in the operation of annular combustors is to ensure a reliable light-round ignition that will establish flames on all injector units without excessive pressure excursion and in a relatively short period of time. This issue is here examined in the case where the annular system is fed with pure hydrogen by combing experimentation and reduced order modeling. Systematic experiments are carried out in a model scale configuration equipped with multiple injectors, in which pure hydrogen is delivered in cross-flow in a swirling stream of air. In this large set of experiments, ignition initiated by a single spark plug gives rise to a couple of flames traveling in clockwise and counterclockwise directions which at a later stage, propagate head-on and merge. It is found that the duration of this process is much shorter when the combustor is fed with pure H2 than when it is operated with gaseous propane-air mixtures or liquid sprays of heptane or dodecane. Systematic observations of the final stage before flame merging indicates that a layer of fresh reactants is formed between the two flame branches which slows down the flame propagation. This flame deceleration is here documented in the case of hydrogen flames. A reduced order model, that accounts for this final stage, is shown to suitably capture effects of global equivalence ratio and injection velocity on the light-round time delay. Experiments also provide indications on effects of injector swirl number and preheating of the chamber walls. In contrast with previous experiments with hydrocarbon flames, it is found that preheating has only a marginal effect on the light-round time. An examination of pressure records during ignition is finally carried out to quantify the amplitude of the ignition-induced pressure excursion. A novel scaling law is derived to estimate the corresponding pressure peaks and this model is shown to be consistent with experimental data. It is also found that under certain conditions, the ignition is followed by a cyclic regime corresponding to a thermoacoustic oscillation that is shown to be coupled by the first azimuthal mode of the chamber
On the evaluation of in-situ mechanical properties of an additively manufactured BCC lattice
No full textInternational audienceAdditive manufacturing enables for fabrication of complex lattice materials with superior strength-to-weight ratios. These lattice structures are made from small struts that are difficult to test individually and whose responses may be size-dependent. In the present work, the mechanical response of a BCC lattice was studied when subjected to in-situ compression, i.e., monitored via Digital Volume Correlation (DVC). Backtracking was implemented in the DVC procedure to account for geometric imperfections generated by the manufacturing process. Displacement fields were measured via DVC and then employed in a Finite Element Model Updating (FEMU) procedure to calibrate the parameters of the constitutive law. Numerical simulations with the calibrated parameters were found to be in good agreement with experiments. The lattice material had a lower strength than that reported in the literature and measured with ex-situ tensile tests. This difference is attributed to size effects; this work demonstrates that they can be accounted for and that a single in-situ test is sufficient to calibrate such material properties
Investigating the Effects of Augmented Reality on Message Credibility When Visualizing Environmental Impacts
No full textInternational audienceAugmented reality (AR) has increasingly been used to communicate environmental impacts, offering greater engagement than conventional displays. However, its effect on message credibility—how much people believe in the content of the communication—remains unclear. In a preregistered study, we compared the perceived credibility of environmental information presented via visualizations on an AR headset or a desktop display. We created display-specific visual encodings (3D concrete for AR, 2D bar charts for desktop) and added two control conditions to cross display and encoding. We found no difference in message credibility between AR and desktop, though concrete AR was rated most engaging. Supplementary material is available at https://osf.io/n4p5c/
A high-order numerical procedure to solve a ten-equation model for high-speed, monodisperse, fluid-particle flows
No full textIn this work, a ten-equation model is presented for the study of high-speed, monodisperse, fluidparticle flows. The model is well-posed for any fluid-particle material density ratio and includes a large range of physical modeling: collisional and frictional pressures in the dense regime, added mass, internal energy and heat transfer between particles, added mass and the fluid. The numerical methods associated with the hyperbolic system are designed to fulfill the main features of a compressible two-phase flow solver: capturing sharp particle fronts, preserving contact discontinuities, and obtaining stability in all flow regimes. This is done by employing a combination of an AUSM + up scheme for the particle phase, and a HLLC scheme for the fluid phase with limited WENO5 reconstructions. A necessary, but not sufficient, condition for stability is obtained by keeping the discrete consistency between spatial fluxes and non-conservative terms. Test cases combining a high-speed fluid interacting with heavy/light particles are used to demonstrate that the qualitative behavior of the flow dynamics over a wide range of density ratios is captured correctly by the ten-equation model
Bowyer-Watson Algorithm for Delaunay Triangulation on Hyperbolic Surfaces
No full textThe well known Bowyer-Watson algorithm constructs the Delaunay triangulation of a point set in the plane incrementally. The restoration of the Delaunay property in the conflict zone of an inserted point is non-ambiguous since this zone is a topological disk. This property no longer holds on a surface with non-trivial topology. We propose an extension of the Bowyer-Watson algorithm to hyperbolic surfaces. Our algorithm uses a thin/thick decomposition of the hyperbolic surface. It first computes an initial point set together with its Delaunay triangulation via an epsilon-net sampling of the surface. The insertion procedure in the thick part of the surface is as in the original algorithm in the plane. The insertion in the thin part differs to overcome the possibly non-trivial topology of the conflict zone
Deterministic Root Finding over Finite Fields with Elliptic Curves
No full textWe study deterministic root-finding for univariate polynomials over finite fields of large characteristic. Despite efficient randomized algorithms, no general deterministic polynomial-time algorithm is known. Classical deterministic approaches exploit the structure of the multiplicative group , and achieve good complexity when is sufficiently smooth, a restrictive condition that is satisfied only rarely.We introduce a new deterministic root-finding framework that replaces the multiplicative group by the group of rational points of an elliptic curve and its quadratic twist, both defined over . This removes the dependence on the smoothness of . Given an elliptic curve with Frobenius trace , the resulting algorithm requires arithmetic operations in , where is the largest prime factor of , and where is a feasible exponent for matrix multiplication.We argue that elliptic curves with favorable group structure are abundant: under a natural smoothness heuristic analogous to classical conjectures on the distribution of smooth numbers and supported by numerical evidence, there exists for every a curve such that .Under this heuristic, and with a precomputation that depends only on , we obtain for each finite field of sufficiently large characteristic a deterministic root-finding algorithm whose complexity is polynomial in and .Finally, we complement the theoretical analysis with an implementation in Magma, illustrating practical feasibility on general 64-bit prime fields
Exploring Non-Euclidean Geometry Through Virtual and Real Experiences to Unlock Non-Experts' Intuition
No full textInternational audienceThree-dimensional non-Euclidean geometries are highly abstractmathematical constructs, typically introduced only at advanced levelsof mathematical education due to their conceptual and technicalcomplexity. This paper presents a study exploring whether intuitiveunderstanding of one such geometry – S2 × E, the product of a 2-sphere and a line – can be facilitated through an interactive virtualreality (VR) experience without relying on formal mathematical instruction.We designed an immersive VR environment that allowsusers to explore S2 × E as inhabitants of this space. Participants,who were neither mathematics experts nor students specializing inmathematics, engaged with the system after a short introduction usinga learn-by-doing approach. Results show that participants wereable to infer and articulate key properties of S2 × E without knowledgeprior to the experiment, and that their performance was independentof their educational background or attitude towards mathematics.Our findings demonstrate the potential of VR as a mediumfor conveying abstract geometric concepts through embodied learning,making them accessible to non-expert audiences
STATE-FEEDBACK STABILIZATION OF 2 × 2 HYPERBOLIC SYSTEMS WITH DISTRIBUTED ACTUATION
No full textThis paper addresses the stabilization of a 2 × 2 system of heterodirectional linear first-order hyperbolic partial differential equations using distributed in-domain actuation. By applying a backstepping Volterra transformation, the PDE system is reformulated as an integral difference equation. We then propose a controller with an auto-regressive structure, formulated as a distributed delayed feedback involving both the system state and the control input. The controller gains are computed by solving a set of Fredholm integral equations, whose solvability is ensured via an approximate controllability condition. The proposed methodology is further extended to a mixed actuation setting, in which the same control input also acts at one boundary of the domain. Numerical simulations are provided to illustrate and validate the theoretical results
Decentralized Coalition Formation of Infrastructure Providers for Resource Provisioning in Coverage Constrained Virtualized Mobile Networks
No full textInternational audienceThe concept of wireless virtualized networks allowsthe utilization of resources provided by several infrastructureproviders (InP) to set up a service over some geographical area.Resource provisioning in such situation is challenging due tothe reluctance of InPs to share private information. This paperproposes a decentralized solution allowing several InPs to forma coalition so as to satisfy the constraints imposed by a MobileVitrual Network Operator, while sharing limited information.The proposed solution adopts a hedonic approach for coalitionformation. The coalition decisions relies on satisfaction of servicerequirements and on the profit of InPs. The evaluation resultdemonstrates the applicability and performance of the proposedsolution