HAL-MINES ParisTech
Not a member yet
27348 research outputs found
Sort by
Activité quaternaire de la faille de Marsanne révélée par des données paléosismologiques (SE France)
No full textInternational audienceLe séisme du Teil (Mw 4,9) a mis en évidence la possibilité d'une activité sismogène de failles auparavant considérées comme peu ou non actives. Dans ce contexte, de nouvelles données ont permis de mieux caractériser la trace en surface ainsi que la géométrie de la faille de Marsanne, identifiée comme potentiellement active par Jomard et al. (2017). Afin d'évaluer son activité quaternaire, une approche paléosismologique a été mise en œuvre. En s'appuyant sur un modèle structural du système de failles nord-cévenol issu d'imagerie sismique, trois sites d'étude ont été sélectionnés à l'interface entre des formations sédimentaires plio-quaternaires et la projection en surface de la faille.L'analyse des tranchées ouvertes sur ces sites a permis d'identifier une déformation tectonique compatible avec un paléoséisme inverse, caractérisé par un rejet vertical d'environ 50 cm, daté entre 320 ± 24 ka et 271 ± 39 ka par ESR. La quantification des déplacements cumulés, après correction des effets post-sismiques superficiels (reptation, fauchage), conduit à une estimation de magnitude minimum de Mw 5,9 pour un à deux événements. Plusieurs scénarios de rupture sont porposés suite à la compilation des observations en tranchée. En revanche, les tranchées réalisées plus au nord, dans les dépôts holocènes de la vallée du Rhône, n'ont révélé aucun indice de réactivation de la faille au cours des derniers 11 700 ans. La longueur de cette rupture est éstimée à 8 kilomètres, ce qui est cohérent avec la segmentation de la faille de Marsanne proposé par le modèle.Ces résultats, obtenus dans le cadre du projet FremTEIL en lien avec le programme FACT, permettent une réévaluation de la cartographie du système de failles des Cévennes, contribuent à affiner la compréhension de la dynamique tectonique du Bassin du Sud-Est, et fournissent des contraintes nouvelles sur l'aléa de rupture de surface dans la vallée du Rhône
Optimisation Multi-Étapes sous Incertitude : de la Réduction d'Arbres de Scénarios par Transport Optimal à l'Optimisation Robuste
No full textThis thesis addresses decision-making under uncertainty in multistage stochasticoptimization problems and proposes new methods for efficiently reducing scenario trees.Building on the Wasserstein distance, it introduces an exact approach for computingWasserstein barycenters using the Douglas–Rachford decomposition scheme, achievinga balance between numerical efficiency and accuracy. Extensions to unbalanced optimaltransport and constrained barycenters are also developed, providing greater flexibility forpractical applications such as scenario reduction and image processing. Finally, theseadvances are integrated into an enhanced tree-reduction algorithm and a completesoftware framework for multistage energy management, enabling comparison betweendifferent optimization paradigms, including stochastic, robust, and reinforcement learningapproaches. These software tools have been developed and implemented for industrialapplications at IFPEN.Cette thèse traite de la prise de décision sous incertitude dans les problèmes d’optimisationstochastique multi-étapes et propose de nouvelles méthodes pour réduire efficacement lesarbres de scénarios. En s’appuyant sur la distance de Wasserstein, elle introduit uneapproche exacte du calcul de barycentres de Wasserstein via le schéma de décompositionde Douglas–Rachford, conciliant précision et efficacité numérique. Une extension autransport optimal déséquilibré et au barycentre contraint est également développée, offrantune plus grande flexibilité pour des applications pratiques, notamment la réduction descénarios et le traitement d’images. Enfin, la thèse intègre ces avancées dans un algorithmeamélioré de réduction d’arbres et dans un logiciel complet de gestion d’énergie multi-étapes,comparant différentes approches d’optimisation (stochastique, robuste et par apprentissage).Ces outils logiciels ont été développés et déployés pour des applications industrielles à l’IFPEnergies nouvelles
Analyse et approximation de quelques équations différentielles stochastiques pour la modélisation de phénomènes non diffusifs : application au transport turbulent
No full textThis work focuses on the study of different classes of stochastic differential equations (SDEs) arising in models of particle transport - in particular fibers - in turbulent flows. We develop approximation methods for these SDEs and establish associated convergence results. The first part is devoted to SDEs driven by a random Poisson measure. We generalize the numerical scheme of Asmussen-Rosiński, which approximates small jumps by Gaussian random variables, to a framework not necessarily driven by Lévy processes. Building on the work of N. Fournier, we prove convergence in Lp norm (rate 1/p) and in weak error (rate 1), and illustrate these results with simulations and an application to the orientation of rods in turbulence. We then establish an Alekseev-Gröbner type formula to compare two jump SDEs, by constructing a Poisson-Skorohod integral that preserves the additivity property. The second part concerns integrated Volterra processes. We prove a weak error comparison result, in terms of the L1 norm of the difference between kernels, which allows us to quantify various approximations, in particular the so-called Markovian approximation. The proof combines the functional Itô formula from Viens-Zhang and the path-dependent Feynman-Kac formula from Bonesini-Jacquier-Pannier. These tools are applied to the modeling of turbulent intermittency via a Riemann-Liouville type kernel, when the Hurst parameter tends to zero. Finally, we introduce a stochastic partial differential equation (SPDE) describing the dynamics of a flexible fiber in a turbulent flow modeled by incompressible Kraichnan noise. In the simplified case of a rigid fiber, we prove existence and uniqueness of the solution, and show that the SPDE is equivalent in law to an SDE, thereby establishing a connection with the first part.Ce travail porte sur l'étude de différentes classes d'équations différentielles stochastiques (EDS) apparaissant dans des modèles de transport de particules — en particulier de fibres — dans un écoulement turbulent. Nous développons des méthodes d'approximation pour ces EDS et établissons des résultats de convergence associés. La première partie est consacrée aux EDS dirigées par une mesure de Poisson aléatoire. Nous généralisons le schéma numérique d'Asmussen-Rosiński, qui approxime les petits sauts par des variables gaussiennes, à un cadre non nécessairement dirigé par un processus de Lévy. Prolongeant les travaux de N. Fournier, nous montrons la convergence en norme Lp (vitesse 1/p) et en erreur faible (vitesse 1), et illustrons ces résultats par des simulations ainsi qu'une application à un modèle d'orientation de bâtonnets en turbulence. Nous démontrons ensuite une formule de type Alekseev-Gröbner pour comparer deux EDS à sauts, en construisant une intégrale de Poisson-Skorohod qui respecte la relation de Chasles. La seconde partie s'intéresse aux processus de Volterra intégrés. Nous établissons un résultat de comparaison en erreur faible, en fonction de la norme L1 de la différence de noyaux, qui permet de quantifier diverses approximations, en particulier l'approximation dite markovienne. La démonstration combine la formule d'Itô fonctionnelle de Viens-Zhang et la formule de Feynman-Kac dépendante de chemin de Bonesini-Jacquier-Pannier. Ces outils sont appliqués à la modélisation de l'intermittence turbulente via un noyau de type Riemann-Liouville, lorsque le paramètre de Hurst tend vers zéro. Enfin, nous introduisons une équation aux dérivées partielles stochastique (EDPS) décrivant la dynamique d'une fibre flexible dans un écoulement turbulent modélisé par le bruit de Kraichnan incompressible. Dans le cas simplifié d'une fibre rigide, nous prouvons existence et unicité de la solution, et montrons que l'EDPS est équivalente en loi à une EDS, établissant ainsi un lien avec la première partie
Geophysics and geo-statistics for a 3D estimation of physical properties withinthe critical zone sub-surface
No full textInternational audienceEstimating groundwater fluxes and exchanges between groundwater and surface water still remains one ofthe major challenges of near-surface critical zone studies. The roles of these water reservoirs are crucial forthe ecosystem and biodiversity, for ensuring the water supply of fresh water and irrigation water. However,these water fluxes are not directly measured in the field. They are estimated by hydrological and hydroge-ological modeling that necessitates a calibration, i.e., estimating parameters governing physical equationssuch as medium porosity, medium initial saturation, or water table level.Near-surface geophysics, petrophysics, and hydrogeophysics are often used to infer these parameters or soilproperties. For instance, electrical resistivity tomography is sensitive to the lithology and fluid content of thenear-surface, while the combination of seismic refraction tomography and surface-wave dispersion inversionallows to infer the water table level and saturation. These geophysical surveys are rather carried out along2D profiles than 3D grids for which data acquisition and post-processing remain time-consuming and havehigh computation costs in general.To overcome these limitations, inferring soil properties in 3D from a set of 2D sections delineating aninvestigation area could be done with the help of geostatistics. Based on six electrical resistivity profiles(five profiles of 72 electrodes and one of 48 electrodes, each with an electrode spacing of 0.40 m, 10 716 mea-surements) and six seismic profiles (same configuration as electrical resistivity profiles, 28 632 first arrivaltime measurements) located at the same place in a 30 m 30 m area in the Orgeval Critical Zone Observatory(France), we first deduce electrical resistivity, P-wave velocity, and S-wave velocity sections with classicalGauss-Newton regularized inversion algorithms (for resistivity and P-wave velocity), and inversion of dis-persion curves (for S-Wave velocity). Then, we estimate the Poisson’s ratio for a saturation assessment andthe water table level assessment. Finally, geostatistical tools such as GSTLearn (GeoSTatistics & MachineLearning Library) (Renard et al., 2025), allow us to interpolate our geophysical parameters in the study areaand obtain a 3D view. Calibration of a geostatistical model to results of geophysical inversion results allowthe use of kriging methods on the investigated area to estimate parameters values in any point. Beyond the3D distribution of parameters, we will present various tests aimed at translating the results into hydrofaciesrepresentation (hydraulic conductivity, porosity and vadoze zone paramters)
Navigation assistance method and device based on a kalman filter
No full textMethod for monitoring the navigation of a vehicle for the implementation of a navigation maneuver: the method including the steps of: estimating a magnitude by a Kalman filter from parameters, a digital model of the vehicle in its environment, and at least one measurement made by at least one sensor and associated with an uncertainty on this measurement, the magnitude serving to implement the navigation maneuver; determining at least one partial derivative value of the magnitude with respect to the at least one measurement; obtaining at least one criticality value of the at least one measurement for the estimation, from the at least one partial derivative
Learning Data-Driven Uncertainty Set Partitions for Robust and Adaptive Forecasting with Missing Data
No full textInternational audienc
Lessons for the United States from Pharmaceutical Regulation Abroad
No full textInternational audienceA American politicians from across the ideological spectrum have found a merican politicians from across the ideological spectrum have found a rare point of agreement: US drug prices are too high, and the government rare point of agreement: US drug prices are too high, and the government should do something about it. Figure 1 presents the results of a study of drug should do something about it. Figure 1 presents the results of a study of drug prices in different countries, which finds that overall US pharmaceutical prices, and prices in different countries, which finds that overall US pharmaceutical prices, and particularly prices of on-patent drugs, are several times higher than those in other particularly prices of on-patent drugs, are several times higher than those in other high-income countries such as France, Germany, Japan, and the United Kingdom high-income countries such as France, Germany, Japan, and the United Kingdom (Mulcahy, Schwam, and Lovejoy 2024). Such comparisons bolster the arguments of (Mulcahy, Schwam, and Lovejoy 2024). Such comparisons bolster the arguments of those advocating price controls or other government interventions in the United those advocating price controls or other government interventions in the United States. States.Why are policies around drug prices necessary? In other words, why do most developed countries rarely leave drug pricing to "the market" without government intervention? In all countries, pharmaceutical markets operate under conditions that differ substantially from the assumptions behind a competitive market that would equilibrate supply and demand based on price. These conditions reflect information problems, the specificities of healthcare markets, and the cost structure of drug development.First, pharmaceuticals are "credence goods": not only is their quality difficult for consumers to determine prior to consumption, but their effects are challenging to identify even after consumption. Regulation of entry, administered by the US Food and Drug Administration (FDA) and similar regulators like the European Medicines Agency (EMA), aims to reduce this information asymmetry. Consumers can be Lessons for the United States from Pharmaceutical Regulation Abroad</div
Correlations between blowability and uniaxial stretchability of PET before and after mechanical recycling
No full textInternational audienceIn the literature, previous studies have linked the behaviour of PET under stretch-blowing and under uniaxialstretching thanks to the time-temperature equivalence principle. Other works showed variations of micro-structural, thermal and mechanical properties of PET submitted to mechanical recycling, but blowability was notmeasured. In this article, the behaviour during free stretch-blowing tests was investigated for a virgin commercialPET before and after mechanical recycling with a step of solid-state post-condensation. For the same physicalstate, the lower stretch-blown volumes obtained for the recycled PET were correlated with a sooner nucleationleading to a lower Natural Draw Ratio, as observed during uniaxial tensile tests. Our results confirm that theuniaxial stretching tests used in the past to discriminate the behaviours of neat PET resins in blowing, can also beused to predict variations of blowability under mechanical recycling
HOW RHEOLOGY HELPS MAKING 3D PRINTED ADDITIVE-AND CROSSLINKER-FREE BIO-AEROGELS
No full textInternational audienceBio-aerogels are nanostructured low-density and high-specific surface materials, and the majority are based on polysaccharides. Two main pathways are used to make bio-aerogels: either by polymer dissolution or from a suspension of “nanoparticles” (nanocellulose, nanochitin), followed by solvent exchange and drying with supercritical CO2. In our work, we consider the first pathway. Till now, the majority of bio-aerogels made by dissolution/solvent exchange route are prepared in “simple” shapes (cylinders, disks) by molding [1], or by making particles (dropping, prilling, emulsification) [2, 3]. However, certain applications, for example, biomedical, may require complex shapes and hierarchical porosity. This can be achieved by 3D printing. The goal of our work was to use Direct Ink Writing to print self-standing solutions with no crosslinking and no additives, as they may alter sensitive applications. Two examples will be demonstrated: cellulose/ionic liquid solutions and aqueous carboxymethyl cellulose (CMC) solutions. In both cases we adjusted the rheological properties of solutions to obtain yield-stress shear thinning fluids. For cellulose, this was done by tuning its solubility. After obtaining various complex-shaped structures based on dissolved polysaccharides, aerogels were made and characterized. The density of solid parts was around 0.1 g/cm3 and the specific surface area was 140 m2/g for CMC aerogels [4] and 250 m2/g for cellulose aerogels. Figure 1. Printed CMC solutions and corresponding aerogels [4]AcknowledgementsWe acknowledge the financial support from CARNOT MINES and China Scholarship Council, IFF for providing CMC samples, and Région Sud – Provence-Alpes-Côte d’Azur for supporting the acquisition of scanning electron microscope. We warmly thank J. Jaxel (PERSEE, Mines Paris, France) for supercritical drying and G. L. Tian (Ecole Polytechnique, France) for coding printing pressure calibration.[1] T. Budtova, “Cellulose II aerogels: a review”, Cellulose 2019, 26, 81–121[2] M. Negrier, E. El Ahmar, R. Sescousse, M. Sauceau, G. Bouet, D. Eglin, T. Budtova “Upcycling cellulose waste textile into aerogel beads via prilling technique”, Cellulose 2024, 31, 823–839[3] L. Druel, A. Kenkel, V. Baudron, S. Buwalda, T. Budtova, « Cellulose Aerogel Microparticles via Emulsion-Coagulation Technique », Biomacromolecules 2020, 21, 1824−1831[4] S. Yu, R. Castellani, A. Pereira, R. Valette, T. Budtova, “Additive-free 3D-printed nanostructured carboxymethyl cellulose aerogels”, Int J Biol Macromol, 2025, 300, 14027