HAL-Ecole des Ponts ParisTech
Not a member yet
42743 research outputs found
Sort by
Can Foundation Language Models Predict Fluid Dynamics?
No full textThe application of deep learning-based, data-driven methods to fluid dynamics problems has attracted significant interest in recent years, due to their potential for faster flow predictions (model inferences) and reduced computational requirements compared to traditional Computational Fluid Dynamics (CFD) methods. Despite their success, existing supervised deep learning approaches typically require extensive model design and large amounts of high-quality training data to achieve satisfactory performance for each specific fluid flow problem.The emergence of foundation models---pre-trained on large-scale, multidisciplinary datasets and capable of solving a wide range of downstream tasks---raises the question of whether their capabilities can extend to scientific domains such as fluid dynamics. Most foundation models to date have been rooted in language tasks. This study investigates a specific instantiation of a foundation model, the Llama~3 large language model, for predicting fluid flows with varying dynamical complexities.Results show that the Llama~3 foundation model, although originally designed for natural language processing, can be applied to fluid dynamics problems with simple engineering adaptations and without fine-tuning its pre-trained weights. This approach achieves improved accuracy and robustness compared to conventional MLP or LSTM models with comparable capacities and equivalent training data. Furthermore, fine-tuning the model by injecting problem-specific knowledge into the pre-trained weights further enhances its performance. These findings suggest that foundation models hold promise for fast inference and deployment of solutions to fluid dynamics problems. Whether unified with models from other domains (e.g., language models) or developed specifically for fluid dynamics, such models could become powerful tools for improving predictive accuracy and computational efficiency in engineering and scientific applications
La régénération des infrastructures ferroviaires, une transition sans fin ?
No full textNational audienceQuelques recherches sur les moteurs internet qui mesurent le degré d’utilisation d’un terme le confirment, l’usage du terme « régénération » a fait un véritable bond à partir du tournant des années 2000. Si sa diffusion se tasse puis s’affaiblit doucement dans le monde anglophone, son utilisation poursuit une pente croissante dans le monde francophone. Le terme émane des sciences de la nature et il possède une définition précise en médecine. La médecine régénérative (ou régénératrice) vise la réparation ou le remplacement de cellules, tissus ou organes pour restaurer une fonction altérée du corps humain. Sorti de ce domaine disciplinaire, le terme est certes moins ubiquiste que celui, encore plus invasif, de résilience. Cependant, il a néanmoins envahi tous les champs de l’action sur la ville et ses équipements. Les architectes, urbanistes, gestionnaires d’infrastructures et de réseaux l’emploient couramment, et de proche en proche la société et le monde politique le reprennent, au point que sa signification précise semble de plus en plus floue, comme le dénonce dans une tribune récente du journal Le Monde une association spécialisée dans la production agricole biologique . Quelles pratiques et quelles significations sont donc associées à la notion de régénération lorsque celle-ci concerne des infrastructures, c’est-à-dire des quantités d’artefacts physiques hautement complexes enchevêtrées avec des structures de savoir et de pouvoir qui soutiennent, notamment dans nos sociétés avancées, toutes les superstructures économiques, environnementales et sociales qui font notre quotidien ? Le besoin d’utiliser ce terme spécifique plutôt que ceux de réparation, restauration ou modernisation nous dit-il quelque chose de la manière dont les parties prenantes, des chercheurs les plus avancés aux techniciens de maintenance les plus humbles, se représentent les capacités d’évolution de notre environnement physique
GI 725A b: A potential super-Earth detected with SOPHIE and SPIRou in an M dwarf binary system at 3.5 pc
No full textInternational audienceWe report the discovery of a super-Earth candidate orbiting the nearby mid-M dwarf Gl 725A using the radial velocity (RV) method. The planetary signal has been independently identified using high-precision RVs from the SOPHIE and SPIRou spectrographs, in the optical and near-infrared (NIR) domains, respectively. We modelled the stellar activity signal jointly with the planet using two Gaussian processes, one for each instrument to account for the chromaticity of the stellar activity and instrumental systematics, along with a Keplerian model. The signal was significantly detected with a RV semi-amplitude of 1.67 ± 0.20 m/s. The planet Gl725A b is found to be in an orbit compatible with circular with a period of 11.2201 ± 0.0051 days. We analysed 27 sectors of TESS photometry, for which no transit event was found. We determined a minimum mass of Mp sin i = 2.78 ± 0.35 M⊕, which places the planet in the super-Earth regime. Using mass-radius relationships, we predict the planetary radius to be between 1.2 and 2.0 R⊕. The proximity of Gl 725A (at only 3.5 pc) makes this new exoplanet one of the closest to Earth and joins the group of S-type low-mass planets in short orbits (P < 15 days) around close M dwarfs
Endogenous clustering and analogy-based expectation equilibrium
No full textInternational audienceNormal-form two-player games are categorized by players into K analogy classes so as to minimize the prediction error about the behavior of the opponent. This results in Clustered Analogy-Based Expectation Equilibria in which strategies are analogy-based expectation equilibria given the analogy partitions and analogy partitions minimize the prediction errors given the strategies. We distinguish between environments with self-repelling analogy partitions in which some mixing over partitions is required and environments with self-attractive partitions in which several analogy partitions can arise, thereby suggesting new channels of belief heterogeneity and equilibrium multiplicity. Various economic applications are discussed
Open Review of "Expressing general constitutive models in FEniCSx using external operators and algorithmic automatic differentiation"
No full textOpen Review of "Expressing general constitutive models in FEniCSx using external operators and algorithmic automatic differentiation" published in JTCAM with doi 10.46298/jtcam.14449International audienc
Deposition of microplastics on the sedimentation active band of rivers
No full textInternational audienc
A coherent index for dichotomy in version-controlled repositories
No full textInternational audienceVersion Control Systems such as Git and Mercurial model their repositories as collections of Merkle directed acyclic graphs. New versions of the code base are added as sources in local graphs and then shared with other agents in the lifetime of the project. In the largest graphs, which can grow to several millions of revisions, sub-linear algorithms become necessary for recurring tasks. A common solution is to use a precomputed index that can grow dynamically along with the graphs. In this paper, we propose a versatile and compact index (a few bytes per node in practice) for dichotomy operations on Merkle DAGs. Furthermore, our index is coherent, in the sense that all agents participating in a repository have the same information for each node they know about.In order to define our index, we introduce the notion of range, a small-sized representation of a set of nodes which can easily be partitioned into smaller ranges. We show how it can be used for the problems of reachability and label discovery, and compare its performance with existing indices for reachability
MatMat v1.0.0 – A modular Input-Output framework: Assessing country-driven environmental and socio-economic impacts of forward-looking scenarios along global supply chains
No full textInternational audienc
« L’autoconsommation collective d’électricité en France. Emergence d’une innovation contrariée »
No full textInternational audiencePrésentation de l'ouvrage et actualisation de la réflexio
Multi-physics modelling of 3D-printed concrete evolution in environmental conditions
No full textInternational audienceExtrusion-based 3D-printed cementitious structures have high water loss after printing provoking significant plastic shrinkage. In this study, we propose a thermo-poro-mechanical model of printed cementitious materials, driven by the experimental observation of a positive correlation between the printed wall thickness and compressive strength at the hardened state. The model is developed to represent evaporation at free surfaces, water consumption associated to the cement hydration and water flow within the material, accounting for their effect on temperature variations, strains and on the evolution of stiffness and compressive strength. Comparisons of compressive strength and plastic shrinkage with experiments are presented, demonstrating the validity of the proposed model. In the absence of protective measures, wall thickness is positively correlated with compressive strength and negatively correlated with shrinkage. When preventing evaporation by putting printed specimens in water, plastic shrinkage is significantly reduced and the compressive strength is increased, reaching similar values as cast samples