Portail HAL edf
Not a member yet
11029 research outputs found
Sort by
Méthode à noyau informée par la physique
No full textInternational audiencePhysics-informed machine learning typically integrates physical priors into the learning process by minimizing a loss function that includes both a data-driven term and a partial differential equation (PDE) regularization. Building on the formulation of the problem as a kernel regression task, we use Fourier methods to approximate the associated kernel, and propose a tractable estimator that minimizes the physics-informed risk function. We refer to this approach as physics-informed kernel learning (PIKL). This framework provides theoretical guarantees, enabling the quantification of the physical prior’s impact on convergence speed. We demonstrate the numerical performance of the PIKL estimator through simulations, both in the context of hybrid modeling and in solving PDEs. In particular, we show that PIKL can outperform physics-informed neural networks in terms of both accuracy and computation time. Additionally, we identify cases where PIKL surpasses traditional PDE solvers, particularly in scenarios with noisy boundary conditions.L'apprentissage informé par la physique consiste à intégrer un a priori physique dans l'entrainement d'algorithmes d'apprentissage automatique en minimisant une fonction de coût qui comprend à la fois un terme d'attache aux données et une régularisation par équation différentielle partielle (EDP). En se basant sur la formulation du problème comme une méthode à noyau, nous utilisons des méthodes de Fourier pour approximer le noyau associé, et proposons un estimateur implémentable sur ordinateur qui minimise la fonction de risque informée par la physique. Nous appelons cette approche le Physics-Informed Kernel Learning (PIKL). Ce cadre vient avec des garanties théoriques, permettant notamment de quantifier l'impact de la physique sur la vitesse de convergence. Nous démontrons la performance numérique de l'estimateur PIKL par des simulations, à la fois dans le contexte de la modélisation hybride et de résolution d'EDPs. En particulier, nous montrons que le PIKL peut surpasser les réseaux neuronaux informés par la physique en termes de précision et de temps de calcul. En outre, nous identifions des cas où PIKL a de meilleures performances que les solveurs d'EDP traditionnels, en particulier dans des scénarios avec des conditions aux limites bruitées
Computational homogenization of a physically-based crystal plasticity law for irradiated bainitic steels
No full textInternational audienceThe elasto-viscoplastic response of irradiated bainitic steels for pressure vessels of light water reactors is described by a multiscale micromechanical model. The model relies on a simplified set of complex constitutive equations describing intragranular flow under a wide range of temperatures, strain rates, and irradiation levels. These equations were themselves partially calibrated by multiscale analyses based on dislocation dynamics calculations, atomistic calculations, and experimental measurements. They include the contribution of jog drag, lattice friction, evolution of dislocation microstructures, and irradiation hardening. The scaling up of these intragranular laws to polycrystalline samples relies on a computational homogenization method which solves the field equations within periodic representative volume elements by means of Fast Fourier Transforms. This computational method proves advantageous relative to the finite element method in handling the complex microstructural morphology of the model required to achieve overall constitutive isotropy. Macroscopic simulations for uniaxial curves under different irradiation levels are first confronted to experimental curves to identify certain microscopic material parameters employed to describe the evolution of the mean-free path of dislocations with deformation. Subsequent comparisons for the evolution of the yield stress, irradiation hardening and the response to sudden strain-rate variations are then reported for a class of steels with various chemical compositions under wide ranges of temperature, loading rate and irradiation level. Good agreement is obtained in all cases. Finally, simulations are employed to explore the influence of the initial dislocation density on the intragranular stress and strain fields. An appreciable influence on the fields is observed during the elasto-viscoplastic transition but not deep in the plastic range
WindDragon: automated deep learning for regional wind power forecasting
No full textInternational audienceAchieving net-zero carbon emissions by 2050 necessitates the integration of substantial wind power capacity into national power grids. However, the inherent variability and uncertainty of wind energy present significant challenges for grid operators, particularly in maintaining system stability and balance. Accurate short-term forecasting of wind power is therefore essential. This article introduces an innovative framework for regional wind power forecasting over short-term horizons (1-6 h), employing a novel Automated Deep Learning regression framework called WindDragon. Specifically designed to process wind speed maps, WindDragon automatically creates Deep Learning models leveraging Numerical Weather Prediction (NWP) data to deliver state-of-the-art wind power forecasts. We conduct extensive evaluations on data from France for the year 2020, benchmarking WindDragon against a diverse set of baselines, including both deep learning and traditional methods. The results demonstrate that WindDragon achieves substantial improvements in forecast accuracy over the considered baselines, highlighting its potential for enhancing grid reliability in the face of increased wind power integration. Impact StatementThis article presents an optimization tool to automatically find efficient deep neural networks to forecast aggregated wind power generation at the level of a region or a country. These models are based on wind speed maps from numerical weather prediction (NWP) forecasts and take advantage of their spatio-temporal aspect. These methods could play a crucial role in the smooth operation of power grids in the context of massive renewable energy integration.</div
Enhancing Biocide Safety of Milk Using Biosensors Based on Cholinesterase Inhibition
No full textInternational audienceA sensitive and reliable electrochemical biosensor for the detection of benzalkonium chloride (BAC) and didecyldimethylammonium chloride (DDAC), the most commonly used disinfectant biocides in the agri-food industry, is described. Acetylcholinesterase from Drosophila melanogaster (DM AChE) and butyrylcholinesterase from horse serum (BChE) were immobilized by entrapment in a photocrosslinkable polymer on the surface of carbon screen-printed electrodes. Preliminary tests conducted in phosphate buffer showed limits of detection (LODs) of 0.26 µM for BAC using the BChE-based sensor and 0.04 µM for DDAC using the DM AChE sensor. These performances comply with the European regulation for dairy products, which sets a maximum allowable concentration of 0.28 µM for biocides. However, when tested directly in milk samples, a dramatic decrease in the sensitivity of both sensors towards BAC and DDAC biocides was reported. To overcome this problem, a simple liquid-liquid extraction was necessary prior to biosensor measurements, ensuring that the biosensors met European regulatory standards and provided an unbiased response
Assessment of a novel alcohol-in-biopolymer emulsion for enhanced remediation of diesel-contaminated soils
No full textInternational audienceConventional pump-and-treat technologies have demonstrated limited effectiveness in remediating soils contaminated with light non-aqueous phase liquids (LNAPLs), such as petroleum hydrocarbons. Non-conventional in-situ flushing with shear-thinning fluids, such as polymers, offers a promising alternative. However, even with polymer flushing, residual LNAPL ganglia may remain trapped in porous media, requiring further improvement of the flushing fluid to enhance remediation efficiency.In this study, we present a novel alcohol-in-biopolymer emulsion developed to enhance the recovery of residual diesel oil from porous media. Batch experiments were conducted to evaluate the partitioning behavior of fifteen different alcohols between the aqueous and diesel phases. The results revealed that 1-pentanol preferentially partitions into the diesel phase rather than the aqueous phase, leading to an increase in diesel oil volume via a swelling mechanism. Furthermore, 1-pentanol forms a stable and homogeneous emulsion when combined with an aqueous solution of the biopolymer xanthan gum, and the surfactant sodium dodecyl sulfate. The emulsion demonstrated stability over 30 d, ensuring its suitability for prolonged remediation processes. Rheological experiments confirmed the emulsion's shear-thinning behavior, which ensures stable and uniform displacement within porous media.A two-dimensional cell packed with silica sand was used to evaluate the efficiency of the emulsion in removing residual diesel oil. The results demonstrated that the emulsion propagates uniformly throughout the porous media, effectively achieving complete removal of residual diesel within 1.15 pore volumes of injection. Pore-scale visualizations revealed the swelling and subsequent mobilization of entrapped diesel ganglia induced by the emulsion, further confirming its efficacy. These findings highlight the potential of this novel alcohol-in-biopolymer emulsion to significantly improve diesel oil recovery from contaminated soils
Effect of a cement-bentonite grout on AVM glass alteration and C-steel corrosion at nanometer scale
No full textInternational audienceTwo experimental mockups were operated for one year at 70°C to study the corrosion of C-steel and alteration of AVM glass surrounded by Cox claystone. Only one system included cement-bentonite grout (CBG). Nanometer-scale analyses (TEM, XANES) examined the glass/C-steel interface to assess CBG's effects. The most notable difference was the presence of a nanometric magnetite layer on the C-steel surface in the CBG system. This layer, promoted by the slightly alkaline pH (8–10) solution influenced by CBG, could act as a passivating barrier, potentially mitigating corrosion, although corrosion rates showed no significant differences over one year. A 5 µm-thick sodium-depleted gel layer of disordered SiO 2 formed on AVM glass in both systems, with similar porosity (14–20 nm). The open porosity may limit the gel's protective capacity. However, glass alteration rates decreased over time due to passivation mechanism primarily driven by reduced diffusion through the gel layer, with chemical affinity playing a lesser role. Secondary phases (Si-Fe-O, Si-Mg-O) were detected only in the CBG system, likely originating from the claystone or CBG rather than the glass itself. These findings indicate that CBG had little effect on AVM glass alteration but may enhance long-term C-steel corrosion resistance through magnetite layer formation
Multidecadal trends in brown trout populations in France reveal a decline in adult abundance concomitant with environmental changes
No full textInternational audienceMost studies of brown trout (Salmo trutta) populations in headwater streams have focused on the year-to-year variability in recruitment and survival, but only few analyzed long-term trends in trout densities, under both control and regulated flow conditions. Here, we conducted trend analyses of brown trout age class densities on 36 stream reaches over the 1990-2020 period, including reaches located in a bypassed section. We also investigated long-term trends in a panel of key environmental variables (water temperature, stream flow, current velocity and habitat suitability). We found that annual water temperatures significantly increased by a median of +0.21 °C per decade. Analyses of stream flow revealed only a few significant trends, including a general increase in median values in spring and a general decrease in fall. A significant general decline in adult trout densities was observed, although disparities between geographic areas were highlighted. This decline is likely due to multifactorial effects, including possible interacting factors. Our results highlight the need to maintain and extend long-term monitoring of trout populations, which should be combined with extensive environmental monitoring
Meteorological Conditions Influence the Migration of a Marine Dune Field in the Southern North Sea
No full textInternational audienceA field of marine dunes has been studied in the Southern Bight of the North Sea. These large dunes, 1–5 m in height and several hundred meters in length, are highly mobile: migration rates of up to 30 m/year have been observed in places. The area is dominated by tides and is characterized by strong currents. Winds are predominantly from the southwest and, to a lesser extent, from the north. A large‐scale 3D numerical model was used to simulate the migration of this dune field over time. It is based on the process‐based openTELEMAC system. The model has been calibrated and validated against in situ bathymetric data and is therefore suited to our objective: to explore the contribution of weather (wind and atmospheric pressure) to the propagation of large marine dunes, in relation to that of tidal currents. To do this, a 4‐month period was simulated, with and without meteorological effects being taken into account in the numerical model. The results highlight the fundamental role of wind conditions in an accurate representation of seabed changes over time. They also show how meteorological events that are different from the prevailing conditions influence the short‐term evolution of the dune field
Maxwell's equations with hypersingularities at a negative index material conical tip
No full textInternational audienceWe study a transmission problem for the time harmonic Maxwell's equations between a classical positive material and a so-called negative index material in which both the permittivity ε and the permeability µ take negative values. Additionally, we assume that the interface between the two domains is smooth everywhere except at a point where it coincides locally with a conical tip. In this context, it is known that for certain critical values of the contrasts in ε and in µ, the corresponding scalar operators are not of Fredholm type in the usual H^1 spaces. In this work, we show that in these situations, the Maxwell's equations are not well-posed in the classical L^2 framework due to existence of hypersingular fields which are of infinite energy at the tip. By combining the T-coercivity approach and the Kondratiev theory, we explain how to construct new functional frameworks to recover well-posedness of the Maxwell's problem. We also explain how to select the setting which is consistent with the limiting absorption principle. From a technical point of view, the fields as well as their curls decompose as the sum of an explicit singular part, related to the black hole singularities of the scalar operators, and a smooth part belonging to some weighted spaces. The analysis we propose rely in particular on the proof of new key results of scalar and vector potential representations of singular fields