Portail HAL ONERA
Not a member yet
12842 research outputs found
Sort by
Understanding underlayer dynamics of a fire-prone Mediterranean Tree-Grass Ecosystems using In Situ Data, a 3D Radiative Transfer Model and multi-scale remote sensing data
No full textInternational audienceIn Mediterranean ecosystems, periods of severe heat and drought are frequent, and such extreme conditions are expected to become more common and intense in the coming years, thus enlarging fire-prone seasons [1], [2]. In Mediterranean tree grass ecosystems (TGE), two vegetation layers (trees and grass) compose the fuel source, with the herbaceous underlayer recognized as the most flammable living fuel [3]. Remote sensing provides useful data for mapping and monitoring the functional traits of vegetation, which can provide insightful information on its flammability potential such as fuel moisture content, and related proxies such as pigments, water and dry matter content. However, modeling herbaceous vegetation presents unique challenges due to high species diversity leading to overlapping phenological cycles and the coexistence of photosynthetic and non-photosynthetic materials. While the relationship between spectral variables and vegetation traits has been already studied for a long time in the case of photosynthetically active vegetation, the same is not true for senescent material [4]. Besides, the leaf and canopy radiative transfer models (RTMs) such as PROSPECT and PROSAIL commonly used to estimate plant traits, assume uniform optical properties within a vegetation layer, which limits its applicability to ecosystems where a mixture of photosynthetic and senescent plant material occurs. The 1D SenSCOPE RTM [5] partially addresses this problem by representing light interactions and physiology of photosynthetically active and senesced leaves in a same layer of vegetation, though it neglects geometric effects. Transitioning to 3D RTMs is essential to account for both spatial and optical heterogeneity, yet challenges persist in modeling mixed materials within complex canopies [6]. These challenges are further compounded by the temporal dynamics of phenological fractions and the physiological states of photosynthetic vegetation, which shift over time, making it difficult to develop a RTM parameterization that is sufficiently generalizable to capture these variations. Capturing seasonal variations in optical properties which reflect changes in vegetation flammability, throughout the phenological cycle is crucial for enhancing the accuracy and realism of fire prevention efforts This study aims to (1) design a more suitable 3D modeling of the herbaceous vegetation stratum in a Mediterranean TGE accounting for inherent variability of phenophases and then (2) validate the simulated spectra by comparing with measured spectra at multiple scales from ground to airborne and satellite, and along different phenophases. We employed the Discrete Anisotropic Radiative Transfer (DART) [7] RTM, incorporating site-specific structural and optical parameters to simulate reflectance of the herbaceous vegetation in a Mediterranean TGE located in Spain (Majadas de Tietar - latitude: 39.94° N; longitude: -5.77° W) during three phenological stages (biomass peak - BP, grass decay - GD and summer drought – SD [8]). A novel modeling approach is proposed that differentiates two phenological fractions within a 3D DART scene composed of 2 distinct vegetation volumes representing photosynthetic (PV) and non-photosynthetic (NPV) vegetation with individual optical properties, but sharing identical geometric parameters to ensure spatial overlap. The optical properties of the NPV volume are derived from in situ spectral libraries of senescent material acquired at the site, while the PV volume is modeled using PROSPECT leaf RTM. Leaf density within each vegetation volume is represented based on in situ Leaf Area Index (LAI) measurements. Other PROSPECT parameters, such as pigment content, water and dry matter, have been parameterized using in situ measurements, while DART model parameters, including structural and angular distributions, are based on the literature. As the proportions of the different vegetation fractions vary throughout the year as well as the physiological state of the PV vegetation, each phenophase was parameterized with specific ranges; thus, separate 3D mockups were constructed for each of the three studied phenophases. Simulated spectra were compared with reflectance datasets acquired at the study site at three different spatial scales: ground measurements from an ASD FieldSpec® 3 spectroradiometer, airborne images acquired by a Compact Airborne Spectrographic Imager CASI-1500i hyperspectral sensor, and multispectral satellite images fromSentinel-2 MSI. Spectral Angle Mapper (SAM) and Root Mean Squared Error (RMSE) were calculated to quantify shape and intensity similarity between the simulated and measured reflectance spectra. High levels of agreement between observed and simulated spectra across phenophases and spectral domains were found when comparing with ASD Fieldspec measurements. The simulated spectra was validated in the spectral range from 400 to 2500 nm and good agreements were found for BP (SAM = 6.7°) and SD (SAM = 7.0°) phenophases, while GD presented higher discrepancies (SAM = 11°). The Short Wave InfraRed (SWIR) region (1450 to 2300 nm) exhibited the highest similarities between simulated and measured reflectance spectra, particularly for SD (SAMSWIR 1 = 0.3° and SAMSWIR 2 = 1.9°). In contrast, the VIS domain (400 to 700 nm) showed greater discrepancies, especially for BP and GD (SAM =SAM VIS BP = 5.7 and SAM VIS SD = 4.9). The Near Infrared (NIR) region (700 to 1350 nm) was generally well-simulated, though performance declined significantly for GD. Comparison with hyperspectral airborne data for BP and GD phenophases showed more accurate representation of observed data across most CASI range (400 to 1000 nm) with SAM ≃ 4.2°. However, differences emerged in the VIS region, where measured CASI spectra exhibited a distinct reflectance peak near 560 nm and an absorption feature around 500 nm not observed in the simulations. In the NIR region, simulated reflectance for SD was overestimated (~0.3) compared to observed values (~0.25), approaching the simulation lower bound. Finally, the simulated reflectance aligns well with measured Sentinel-2 reflectance across most spectral regions except in the SWIR 1 bands 11 for GD and SD phenophases, where simulated reflectance values were lower than observed. In the VIS region, notable differences were observed, with simulated reflectance values being higher than the Sentinel-2 measured reflectance spectra, approximately 1 compared to the observed 0.5). Despite the shape differences noticed in those results, the RMSE values across spectral regions, phenophases, and sensors remained low (0.01–0.06%), demonstrating good overall agreement between the simulated and measured median reflectance. To conclude, distinguishing PV and NPV vegetation volumes within the 3D model improves the precision and realism of the two phenological fractions by using respective in situ measurements as input parameters to define their optical properties and proportions. However, discrepancies detected in the comparison of the simulated and measured spectra could be related to unaccounted real-world elements within the 3D scene, such as flowers particularly when comparing across different measurement scales. Future prospects will target vegetation trait estimation from these DART-simulated synthetic datasets by training machine learning algorithms. [1] J. Piñol, J. Terradas, et F. Lloret, « Climate warming, wildfire hazard, and wildfire occurrence in coastal eastern Spain », Clim. Change, vol. 38, no 3, p. 345‑357, 1998. [2] J. E. Keeley et A. D. Syphard, « Climate change and future fire regimes: examples from California », Geosciences, vol. 6, no 3, p. 37, 2016. [3] M. Sánchez-Pinillos et al., « Spatial and temporal variations of overstory and understory fuels in Mediterranean landscapes », For. Ecol. Manag., vol. 490, p. 119094, juin 2021, doi: 10.1016/j.foreco.2021.119094. [4] B. Lu, C. Proctor, et Y. He, « Investigating different versions of PROSPECT and PROSAIL for estimating spectral and biophysical properties of photosynthetic and non-photosynthetic vegetation in mixed grasslands », GIScience Remote Sens., vol. 58, no 3, p. 354‑371, 2021. [5] J. Pacheco-Labrador et al., « senSCOPE: Modeling radiative transfer and biochemical processes in mixed canopies combining green and senescent leaves with SCOPE », BioRxiv, p. 2020‑02, 2021. [6] J. R. Melendo-Vega et al., « Improving the Performance of 3-D Radiative Transfer Model FLIGHT to Simulate Optical Properties of a Tree-Grass Ecosystem », Remote Sens., vol. 10, no 12, Art. no 12, déc. 2018, doi: 10.3390/rs10122061. [7] J.-P. Gastellu-Etchegorry et al., « Discrete anisotropic radiative transfer (DART 5) for modeling airborne and satellite spectroradiometer and LIDAR acquisitions of natural and urban landscapes », Remote Sens., vol. 7, no 2, p. 1667‑1701, 2015. [8] Y. Luo et al., « Using near-infrared-enabled digital repeat photography to track structural and physiological phenology in Mediterranean tree–grass ecosystems », Remote Sens., vol. 10, no 8, p. 1293, 2018
Analyse comparative des techniques de simulation des pièges poreux : des conditions limites de la paroi poreuse au raffinement automatique du maillage
No full textInternational audienceThis paper presents different numerical frameworks for simulating porous bleed systems. Despite progress in turbulence modelling and increased CPU resources, simulating such systems remains challenging, particularly for compressible turbulent flows, due to the significant scale disparity between the perforation holes and the main flow features. This study aims to evaluate four different methods to predict porous bleed effects on a supersonic air intake, with applications to industrial needs such as airbreathing cruise missile or combat aircraft design. The investigated approaches necessitate various levels of meshing effort, engineering time, computational resources and physical modelling. The four approaches are namely: (1) state-of-the-art bleed model, based on continuous suction, (2) classical multi-block structured body-fitted simulations where the holes are meshed (considered as reference), (3) IBC-based approach, which alleviates the need to mesh manually every hole, and (4) automatic mesh refinement, which allows for fully automated body-fitted workflow with minimal engineering time. A thorough comparison of quantities of significance for porous bleeds, like flow structure, bleed mass flux and velocity profiles, is undertaken. Despite a fair overall prediction capability, the porous bleed model approach suffers from the most significant discrepancies with respect to the reference data.Malgré les progrès réalisés dans la modélisation de la turbulence et l'augmentation des ressources informatiques, la simulation de ces systèmes reste difficile, en particulier pour les écoulements turbulents compressibles, en raison de l'écart d'échelle important entre les trous de perforation et les caractéristiques principales de l'écoulement. cette étude vise à évaluer quatre méthodes différentes pour prévoir les effets de parois poreuses sur une entrée d'air supersonique, avec des applications à des besoins industriels tels que la conception de missiles de croisière aérobies ou d'avions de combat. les approches étudiées nécessitent différents niveaux d'effort de maillage, de temps ingénieur, de ressources de calcul et de modélisation physique. les quatre approches sont les suivantes : (1) un modèle de piège poreux, basé sur une aspiration continue, (2) les simulations classiques multi-blocs structurées avec prise en compte des perforations dans le maillage (considérés comme référence), (3) l'approche basée sur les frontières immergées, qui évite d'avoir à mailler individuellement chaque trou, et (4) le raffinement automatique du maillage, qui permet une mise en données automatisé avec ajustement du corps avec un temps d'ingénierie minimal. une comparaison approfondie des quantités d'intérêt pour les parois poreuses, telles que la structure de l'écoulement, le flux de masse de la porosité et les profils de vitesse, est entreprise. malgré une capacité de prévision globale satisfaisante, l'approche du modèle de paroi poreuse montre des divergences les plus importantes par rapport aux données de référence
Wealkly supervised learning for snow cover segmentation in mountainous areas from Sentinel-1 SAR images using interpolated NDSI time series
No full textSnow cover plays a fundamental role in climate regulation and hydrological processes. Existing snow products are based on optical imagery. Yet, snow monitoring remains challenging in mountainous regions due to frequent cloud cover. Synthetic Aperture Radar (SAR) imagery, unaffected by clouds, enables regular wet snow observations. However dry snow remains mostly transparent to SAR. In this study, we propose a fully automated framework to transfer the snow detection capabilities of optical images, that does not differentiate dry and wet snow, to SAR images. A convolutional neural network is trained to predict a binary snow cover map from a Sentinel-1 Single Look Complex (SLC) dual-pol amplitude image and a snow-free reference image. We threshold the MODIS Normalized Difference Snow Index (NDSI) product to generate binary training labels. Our model is trained in a weakly supervised manner by filling the cloudinduced gaps via temporal interpolation. We first evaluate the influence of the input SAR channels configuration and show that concatenating the acquisition of the day with the reference image is preferable to more complex preprocessing. Then, we compare the Closest Neighbours Interpolation and the Kalman smoother to fill the cloud-induced gaps in the MODIS NDSI time series. We show that increasing the level of supervision improves the model performance.By removing all the gaps and the noise in the NDSI time series, the Kalman smoother yields the best model perfomance. However the regularization strength of the Kalman smoother is shown to be critical. To validate our method, we compare it to existing snow products. By comparing with the THEIA L2B Snow product, we show that our method gives comparable results to Sentinel-2 based snow cover maps. The comparison with the Copernicus Wet/Dry Snow product shows that our model can detect both wet and dry snow solely from Sentinel-1.</p
Temporal evolution of the morphology of contaminants deposits
No full textInternational audienceThe effects of contamination on space systems have been studied for many years. However, the effects especially due to the morphology of contaminant deposits have not been studied extensively. ONERA has developed a vacuum chamber for studying molecular contamination in which experiments aimed at studying changes in morphology. Ex-situ optical microscopy analyses revealed temporal maturation phenomena: coalescence and solidification of the deposit. A systematic study of contaminant/substrate/pressure conditions combinations allowed a first understanding of the influence of the pressure on changes in deposit morphology after contamination. These results show the importance of identifying the morphology of contaminant deposits within a short delay in order to anticipate changes in morphology
Boundary layer flashback of H2/Air premixed flames in a swirling flow around a central body
No full textInternational audienceFast and thin premixed hydrogen flames can lead to flashback scenarios which are unusual, especially for swirled configurations. Flashback can occur far from all walls, in the bulk flow, if the flow speed is less than the flame speed: this is a scenario which is usually avoided by increasing flow rates. However, flashback can also occur near walls where the flow speed goes to zero. Injector walls boundary layers always contain a zone where the local flow speed is less than the flame speed, even if the bulk flow velocity is large. The size of this zone is controlled by the velocity gradient at the wall which is the classical parameter used to predict flashbacks in boundary layers.In this study, flashback of lean hydrogen-air flames is computed using DNS (Direct Numerical Simulation) (flame resolved). Without swirl, results are compared and validated against experimental measurements and usual flashback criteria based on wall velocity gradient. DNS are also performed with swirl in a sector of an annular chamber, providing maps of flashback occurrence as function of swirl number and wall velocity gradient. Results show that swirl enhances flashback propensity and that thermodiffusive effects must be accounted to build a flashback criteria, indeed very lean H 2 flames flashback for flow speeds higher than expected. Novelty and significanceAlmost all injection systems designed for hydrogen face a new, key issue in terms of operability: flashback. This study presents for the first time an analysis of the combined effects on flashback of the velocity gradients at the wall and of swirl. DNS of a swirling flow around a central body are performed and flashback maps are produced in a (swirl-velocity gradient) diagram of direct use for systems injecting lean premixed hydrogen-air mixtures
Une stratégie pour l'optimisation robuste et fiabiliste directe des empilements composites
No full textInternational audienceThe increasing demand for lightweight, high-performance materials has accelerated the adoption of composite structures, renowned for their high specific strength and stiffness, along with their ability to tailor structural behavior through stacking sequence optimization. This tailoring allows for precise control of mechanical properties, enabling advanced functionalities in aeronautics, automotive, and wind energy applications. However, the design of composite structures is inherently complex due to the high-dimensional design space introduced by stacking sequences and the uncertainties due to material properties and manufacturing processes. Genetic Algorithms (GAs) have proven effective in retrieving pseudo-optimal stacking sequences, leveraging their global search capability to navigate this complex design space. This work proposes a novel methodology for stacking sequence retrieval, integrating a reliability-based design framework to address these challenges. By incorporating uncertainties into the optimization process, the methodology facilitates mechanical tailoring of composite structures while ensuring performance reliability under manufacturing and material variability.La demande croissante de matériaux légers et performants a accéléré l’adoption de structures composites, réputées pour leur résistance et rigidité spécifiques élevées, ainsi que pour la possibilité d’ajuster leur comportement structurel grâce à l’optimisation des séquences d’empilement. Cet ajustement permet un contrôle précis des propriétés mécaniques, offrant des fonctionnalités avancées dans l’aéronautique, l’automobile et l’énergie éolienne. Cependant, la conception des structures composites est intrinsèquement complexe en raison de l’espace de conception de haute dimension introduit par les séquences d’empilement et des incertitudes liées aux procédés de fabrication. Les algorithmes génétiques (AG) se sont révélés efficaces pour trouver des séquences d’empilement pseudo-optimales, exploitant leur capacité de recherche globale pour naviguer dans cet espace complexe. Ce travail propose une méthodologie novatrice pour la récupération des séquences d’empilement, intégrant un cadre de conception basé sur la fiabilité pour relever ces défis. En intégrant les incertitudes dans le processus d’optimisation, la méthodologie facilite l’ajustement mécanique des structures composites tout en garantissant la fiabilité des performances face à la variabilité des matériaux et de la fabrication
Effets de l'ingestion de couche limite sur le bruit d'une hélice carénée
No full textInternational audienceIn this paper, the effects of boundary layer ingestion on the noise emitted from an academic ducted propeller are evaluated using lattice Boltzmann simulations and wind tunnel experiments. The boundary layer is generated by an aggressive S-plate placed upstream of the propeller and its impact is assessed by comparison with an isolated ducted propeller configuration. Three operating conditions are studied, which correspond to a free-stream velocity of 32 m/s and rotational speeds of 6000, 8000, and 11,000 rpm. Aerodynamic results are compared between simulations and experiments and show a good agreement. They highlight the increase in flow asymmetry and turbulence ingested by the ducted propeller in the presence of the Splate. Direct acoustic results are extracted from the simulations at the position of the two experimental microphone arrays and compared against experiments in terms of tonal and broadband contributions. Similar trends are observed despite some quantitative discrepancies. In the presence of the S-plate, the stronger flow distortion is responsible for an increase of tonal noise at the blade passing frequencies, while the higher turbulence intensity generates increased broadband noise levels.Dans cet article, les effets de l'ingestion de la couche limite sur le bruit émis par une hélice carénée académique sont évalués en utilisant des simulations Boltzmann sur réseau et des expériences en soufflerie. La couche limite est générée par une plaque en S agressive placée en amont de l'hélice, et son impact est évalué par comparaison avec une configuration d'hélice carénée isolée. Trois conditions de fonctionnement sont étudiées, correspondant à une vitesse d'écoulement de 32 m/s et des vitesses de rotation de 6000, 8000 et 11 000 tr/min. Les résultats aérodynamiques sont comparés entre simulations et expériences, montrant un bon accord. Ils mettent en évidence l'augmentation de l'asymétrie de l'écoulement et de la turbulence ingérée par l'hélice carénée en présence de la plaque en S. Les résultats acoustiques directs sont extraits des simulations au niveau des deux réseaux de microphones expérimentaux et comparés aux expériences en termes de contributions tonales et large bande. Des tendances similaires sont observées malgré quelques écarts quantitatifs. En présence de la plaque en S, la distorsion du flux plus marquée est responsable d'une augmentation du bruit tonal aux fréquences de passage des pales, tandis que l'intensité de turbulence plus élevée génère des niveaux de bruit large bande accrus
Learning the Feasibility of Sets of Acquisition Tasks for Earth Observation Satellites
No full textInternational audiencePlanning Earth Observation satellite activities involves computing observation sequences while considering timedependent orientation maneuvers. We introduce a novel approach leveraging neural networks and image classification to learn scheduling feasibility for observation tasks. A 0/1 classifier takes an image of POIs along the satellite's orbit as input, estimating whether a schedulable sequence exists within temporal constraints. Learned offline, this classifier enables quick online selection of optimal POI subsets (among candidates) without requiring numerous costly maneuver computations. Experiments show our classifier achieves high accuracy in approximating scheduling feasibility, allowing for efficient optimization with minimal risk of misclassified solutions
Mesh Dispatching for Area Coverage using Several Earth Observation Systems
No full textInternational audienceIn this paper, we consider area coverage requests that can be fulfilled using a set of Earth observation systems, where each system disposes of its own satellites and its own mission center. The approach proposed involves a federation layer that dispatches meshes of various sizes to the different systems in a seamless way for the end-users. The dispatch strategy is optimized based on heuristic search and large neighborhood search. Weather and system workload uncertainties being hard to estimate, re-dispatch operations are also used to re-optimize the allocation of meshes to systems depending on the current progress. Experiments are performed on scenarios where the goal is to collect images over countries covering up to hundreds of thousands square kilometers, such as France
Port-Hamiltonian formulation and stabilizing controller for a liquid propelled rocket engine
No full textInternational audienceReusable technology in the field of space launchers requires complex maneuvers to land a launcher first stage, requiring variable thrust from the rocket engine. The developments in electrical actuators allowed the introduction of closed-loop controllers for liquid propelled rocket engines (LPRE). While closed-loop controllers have been suggested in the literature with robustness guarantees, few stability proofs have been given. The LPRE is a complex non-linear system, rendering a direct approach to determine a Lyapunov function complex. In this paper, a reformulation of the state-space equations into a model more adapted to stability analysis is proposed, and a passivity approach is derived to prove the stability. In addition, a closed-loop controller that enforces the passivity of the system is designed, with a new equilibrium assignment. Simulated results illustrate the performances of the closed-loop controlled engine