Portail HAL edf
Not a member yet
11029 research outputs found
Sort by
On the convergence of PINNs
No full textInternational audiencePhysics-informed neural networks (PINNs) are a promising approach that combines the power of neural networks with the interpretability of physical modeling. PINNs have shown good practical performance in solving partial differential equations (PDEs) and in hybrid modeling scenarios, where physical models enhance data-driven approaches. However, it is essential to establish their theoretical properties in order to fully understand their capabilities and limitations. In this study, we highlight that classical training of PINNs can suffer from systematic overfitting. This problem can be addressed by adding a ridge regularization to the empirical risk, which ensures that the resulting estimator is risk-consistent for both linear and nonlinear PDE systems. However, the strong convergence of PINNs to a solution satisfying the physical constraints requires a more involved analysis using tools from functional analysis and calculus of variations. In particular, for linear PDE systems, an implementable Sobolev-type regularization allows to reconstruct a solution that not only achieves statistical accuracy but also maintains consistency with the underlying physics
Certification of a wheat reference material for OBT measurement
No full textInternational audienceIt is difficult to measure the radioactivity of organically bound tritium in environmental samples. Over the last twenty years or so, many laboratories have been working to develop reliable methods for measuring tritium. Several interlaboratory comparisons have been organised to develop these methods and enable laboratories to compare themselves. However, the accuracy of the measurement methods has never been assessed due to the lack of certified reference materials available for use in the analyses. This presentation describes the process that led to the production, at the end of 2024, of the first wheat certified reference material for measuring organically bound tritium
A novel airflow zonal model for urban microclimate modelling at the block scale
No full textInternational audienc
Three-dimensional numerical modeling of sediment transport in a highly turbid estuary with pronounced seasonal variations
No full text(IF 3.5;Q2)International audienceSimulating sediment dynamics in a large and energetic estuary system remains challenging, primarily due to the spatial and temporal complexities of the interaction between flow and sediment transport, especially for sand-mud mixtures. This study uses a three-dimensional (3D) numerical model, based on the open TELEMAC system, to investigate the dynamics of suspended sediment concentration (SSC) in the Gironde Estuary, a complex estuarine environment characterized by an estuarine turbidity maximum (ETM) and significant variations in river discharge. The main contributions of this study include addressing the challenges of coupling bed friction with sediment transport of the sand-mud mixture for feedback on bed roughness and bottom depth changes and the ability of the model to capture the migration of ETM from high to low flow. Additionally, the current study analyzes the ability of the model to capture the migration of ETM from high to low flow, and it utilizes a calibration strategy that minimizes parameters by using in situ data and encompassing hydroemorpho-sedimentary interactions. A sensitivity analysis was done using different settling velocity approaches and sediment classes to establish an optimal model configuration and the uncertainty associated with the reduced model parameterization is discussed. The model satisfactorily reproduces the hydrodynamic features, particularly when the hydro-sedimentary feedbacks are taken into account, the seasonal trend of SSC, springneap variations, and the development of a well-defined ETM. The selection of a specific formulation for the settling velocity influences the location and magnitude of ETM. The van Leussen formula not only predicts a broad movement of ETM from high to low river flow, but also predicts high turbidity for extended periods during low river flow. Conversely, two empirical formulas from Le Hir and Defontaine predicted the highest turbidity during neap tides but sediment losses during prolonged simulations. The results of this study contribute to a deeper understanding of sediment dynamics in the Gironde Estuary, providing valuable information for future estuarine modeling and management
Spatio-temporal clustering and reconciliation for regional electricity demand forecasting
No full textThis paper proposes a three-stage approach to forecasting the electricity demand time series of several areas belonging to the same region. First, time series are aggregated on the basis of a spatio-temporal clustering approach, and a three-level hierarchy is build. Second, benchmark forecasts are generated for all series using generalized additive models. Finally, the forecasts are optimally projected onto a coherent subspace to ensure that the final forecasts satisfy the hierarchical constraints. Our approach is tested on Alberta electricity demand data; experimental results suggest that successive clustering and projection steps improve the benchmark forecasts both at the aggregate level and at the disaggregated level.</div
Numerical estimation of ultrasonic phase velocity and attenuation for longitudinal and shear waves in polycrystalline materials
No full textInternational audienceFinite element computations offer ways to study the behavior of ultrasonic waves in polycrystals. In particular, the simulation of plane waves propagation through small representative elementary volumes of a microstructure allows estimating velocities and scattering-induced attenuation for an effective homogenenous material. Existing works on this topic have focused mainly on longitudinal waves. The approach presented here relies on generating periodic samples of microstructures in order to accommodate both longitudinal and shear waves. After some discussion on the parametrization of the simulations and the numerical errors, results are shown for several materials. These results are compared to an established theoretical attenuation model that has been adapted to use a fully analytical expression of the two-point correlation function for the polycrystals of interest, and to use velocities corresponding to different reference media. Promising comparisons are obtained for both longitudinal and shear waves when using more representative media, obtained through Hill averaging or a self-consistent approach. This illustrates how the numerical method can assist in developing and validating analytical models for elastic wave propagation in heterogeneous media
A hackathon for flood map prediction from geospatial data with parsimonious machine learning models
No full textInternational audienceFlooding poses significant risks across various sectors in France. This paper presents the outcomes of a machine learning hackathon focused on predicting the extent of various types of floods by leveraging a combination of geospatial and climate data. A Convolutional Neural Network (CNN) emerged as the most effective model, achieving strong performance in predicting the temporal evolution of flood risk maps. The evaluation not only includes prediction accuracy but also incorporates robustness, frugality, and explainability, in line with the principles of trustworthy AI principles. A key feature of this challenge was the absence of streamflow data, allowing the models to predict floods in regions where such data is unavailable. This highlights the potential of machine learning to improve flood forecasting in data-scarce environments.2 Hackathon setup 2.1 Geospatial Data Labels: Flood maps were extracted from the Sen1Floods11 dataset [2] which can be visualized on the Global Flood Database. Sen1Floods11 provides surface water data, including raw Sentinel-1 WIP paper.</div
airGRteaching: Teaching Hydrological Modelling with the GR Rainfall-Runoff Models ('Shiny' Interface Included). Manual of the R package version 0.3.5
No full textManuelAdd-on package to the 'airGR' package that simplifies its use and is aimed at being used for teaching hydrology. The package provides 1) three functions that allow to complete very simply a hydrological modelling exercise 2) plotting functions to help students to explore observed data and to interpret the results of calibration and simulation of the GR ('Génie rural') models 3) a 'Shiny' graphical interface that allows for displaying the impact of model parameters on hydrographs and models internal variables
Influence of stress variation on radar wave propagation in concrete: Application to the monitoring of nuclear containment building
No full textInternational audienceContainment buildings of nuclear power plants are made of prestressed concrete. Maintaining the tension in prestressing tendons is essential for the safety of these structures. However, steel and concrete can be subject to ageing and pathology (creep, shrinkage, corrosion…), this can lead to prestressing losses. It is then crucial to be able to know the tension state of tendons. Current techniques used to evaluate tension in cables are semi-destructive and are then unusable on containment buildings. This is why it is relevant to develop non-destructive techniques to evaluate potential prestressing losses. Based on the inaccessibility of the cables, this study suggests evaluating concrete stress rather than cable tension. This work particularly focuses on the influence of stress on radar wave propagation in concrete. Tests show that a compressive stress increase on concrete slabs leads to an isotropic delay in the electromagnetic signal and an anisotropic decrease in amplitude, more pronounced in the direction normal to the stress direction. These variations highly depend on concrete hydric state, as it is observed that an increase in stress on a near-saturation or an oven-dried concrete does not induce a signal variation. Finally, in-situ tests performed onto a mock-up of a nuclear containment building built by EDF confirm the sensitivity of radar waves to a variation of concrete stress
