Portail HAL ONERA
Not a member yet
12842 research outputs found
Sort by
RAICE: Real-time Autoencoder for Indoor Coverage Estimation
No full textInternational audienceIndoor environments present significant challenges for electromagnetic propagation prediction due to the complexity of their geometric and physical modeling, as well as the difficulties in accurately reproducing the numerous interactions of electromagnetic waves. Complementing statistical and physical simulation tools, recent years have seen the emergence of approaches leveraging advancements in machine learning, particularly deep learning. This paper introduces a new framework for predicting electromagnetic coverage maps in unknown indoor environments, targeting frequencies from UHF to C-band.Les environnements intérieurs présentent des défis importants pour la prévision de la propagation électromagnétique en raison de la complexité de leur modélisation géométrique et physique, ainsi que des difficultés à reproduire avec précision les nombreuses interactions des ondes électromagnétiques. En complément des outils de simulation statistique et physique, les dernières années ont vu l'émergence d'approches tirant parti des progrès de l'apprentissage automatique, en particulier de l'apprentissage profond. Cet article présente un nouveau cadre pour prédire les cartes de couverture électromagnétique dans des environnements intérieurs inconnus, pour des fréquences allant de l'UHF à la bande C
Dynamique stochastique de la post-combustion dans un panache de fusée riche en carburant et supersonique
No full textInternational audienceThis study explores the large-scale stochastic dynamics of the afterburning in a supersonic, fuel-rich plume exhausting from a model rocket engine tested at the MASCOTTE facility of ONERA. High-speed OH* chemiluminescence imaging reveals significant fluctuations in the axial position of the flame's leading edge, spanning approximately 25 nozzle diameters. This contrasts with conventional camcorder observations, which suggest steady afterburning. The propagation speed of the flame front, relative to the flow velocity (obtained from a RANS non-reactive simulation), is well framed within the Chapman-Jouguet detonation and deflagration velocities. Hence, the flame dynamics follows a stochastic, aperiodic cycle, including phases of autoignition, deflagration propagation, deflagration-to-detonation transition (DDT), detonation propagation, and failure. The mechanism of turbulence-assisted DDT proposed by [A. Poludnenko, T. Gardiner, and E. Oran, Phys. Rev. Lett., 107, (2011)] could favor repeated detonation initiations in the absence of confining effects and obstacles. The experimental case constitutes a more challenging benchmark than the classical steadily lifted supersonic flame of Cheng et al. for validating CFD simulations of supersonic combustion applications
Improvement of mountain natural risks analysis: assessment of reach, seasonal expo- sure and presence probabilities
No full textInternational audienceMountain natural phenomena threaten people and infrastructures. Risk-informed decision making to select risk reduction measures always starts with risk analysis. Natural risks are assessed through a combination of hazard, exposure and vulnerability (equivalent to severity and probability in industrial technological contexts). In practice, characterizing the exposure is indeed not that easy since for a given magnitude, a phenomenon can have several possible trajectories, each of them corresponding to a sub-scenario with a given conditional probability. Seasonal mountain phenomena occurrence and human touristic occupation are highly variable inducing peaks in occupancy rates. This paper addresses the issue of operational assessment of assets exposure considering their seasonal reach and presence probability for different phenomenon sub-scenarios. Simplified and practical methodologies are proposed to first calculate risk based on seasonal phenomenon occurrence and exposure and secondly calculate the reach probabilities of their spatial extent. Simple examples are given for a first single phenomenon (torrential flood) and demonstrate the influence of seasonal occurrence and presence hypothesis on calculated risks. Methodologies can be extended to deal with multi-risk contexts
Fluorescence Thermography for High-Resolution Characterization of Transmitarray Antenna in X Band
No full textInternational audienceRapid advances in microwave technologies necessitate accurate electromagnetic (EM) field measurement, especially for optimizing communication systems. Visualizing entire field distributions with high spatial fidelity is vital for enhancing antenna performance in satellite communications and 5G networks. This study investigates fluorescence thermography using rhodamine B (RhB) coated films for non-intrusive, highresolution characterization of EM fields, focusing on an Xband Transmitting Array (TA) antenna. Experimental results exhibit efficient field focusing, distinct high-intensity regions, and sidelobe suppression, aligning with theoretical expectations. Fluorescence thermography could therefore be a valuable tool for antenna designers, enabling understanding of complex field interactions without disturbing the reactive environment. This work validates the method as a complement to conventional EM field characterization and underscores its potential for driving next-generation telecommunication technologies
Effets de l'ondulation hors-plan sur le comportement mécanique de cornières composites sous flexion 4-point : comparaison essai-calcul
No full textInternational audienceThis study proposes a numerical methodology to integrate into finite element simulations the effect of out-ofplanewaviness defects on the mechanical behavior in pure unfolding (4-point bending) of thermoset L-angle laminates.The defect physically observed by micrography or tomography is described using a parametric model via an optimizationprocess. The defect is then explicitly introduced into a ply-by-ply mesh that faithfully represents the original image. Linearelastic simulations associated with an out-of-plane failure criteria (used in the ONERA's progressive damage model) arethen carried out and compared with tests conducted at ONERA, or tests shared by Airbus Operations. The prediction ofthe failure strength by delamination and its localization are satisfactory in comparison with experimental tests. This showsthe interest of the explicit modeling approach of the defect in the mesh to predict the mechanical knock down factorsrelated to the presence of out-of-plane undulations.Cette étude propose une méthodologie numérique pour intégrer dans des simulations éléments finis l'effet de défauts d'ondulation hors-plan sur le comportement mécanique en dépliage pur (flexion 4-point) de cornières composites stratifiées à matrice thermodurcissable. Le défaut physiquement observé par micrographie ou tomographie est décrit à l'aide d'un modèle paramétrique via un processus d'optimisation. Le défaut est ensuite explicitement introduit dans un maillage pli à pli représentant fidèlement l'image originale. Des simulations linéaires élastiques avec un critère de rupture hors-plan (Onera Progressive Failure Model -OPFM) sont ensuite réalisées et comparées à des essais menés à l'ONERA, ou des essais partagés par Airbus Operations. Les prévisions de la force à rupture par du délaminage et la localisation de la ruine sont satisfaisantes en comparaison avec les essais. Ceci montre l'intérêt de l'approche de modélisation explicite du défaut dans le maillage pour prévoir l'abattement des propriétés mécaniques liées à la présence d'ondulation horsplan
Méthode Galerkin Discontinu basée sur une Reconstruction SIAC Compacte
No full textInternational audienc
Considérations sur la précisions d'algorithmes d'assimilation de données pour la construction de champs d'écoulements denses
No full textInternational audienceWithin the framework of the European Union Horizon 2020 project HOMER (Holistic Optical Metrology for Aero-Elastic Research), data assimilation (DA) algorithms for dense flow field reconstructions developed by different research teams, hereafter referred to as the participants, were comparatively assessed. The assessment is performed using a synthetic database that reproduces the turbulent flow in the wake of a cylinder in ground effect, placed at the distance of one diameter from a lower wall. Downstream of the cylinder, this wall continues either in the form of a flat steady wall, or of a flexible panel undergoing periodic oscillations; these two situations correspond to two different test cases, the latter being introduced to extend the evaluation to fluid–structure interaction problems. The input data for the data assimilation algorithms were datasets containing the particle locations and their trajectories identification numbers, at increasing tracer concentrations from 0.04 to 1.4 particles/mm3 (equivalent image density values between 0.005 and 0.16 particles per pixel, ppp). The outputs of the DA algorithms considered for the assessment were the three components of the velocity, the nine components of the velocity gradient tensor and the static pressure, defined in the flow field on a Cartesian grid, as well as the static pressure on the wall surface, and its position in the deformable wall case. The results were analysed in terms of errors of the output quantities with respect to the ground-truth values and their distributions. Additionally, the performances of the different DA algorithms were compared with that of a standard linear interpolation approach. The velocity errors were found in the range between 3 and 11% of the bulk velocity; furthermore, the use of the DA algorithms enabled an increase of the measurement spatial resolution by a factor between 3 and 4. The errors of the velocity gradients were of the order of 10–15% of the peak vorticity magnitude. Accurate pressure reconstruction was achieved in the flow field, whereas the evaluation of the surface pressure revealed more challenging. As expected, lower errors were obtained for increasing seeding concentration. The difference of accuracy among the results of the different data assimilation algorithms was noticeable especially for the pressure field and the compliance with governing equations of fluid motion, and in particular mass conservation. The analysis of the flexible panel test case showed that the panel position could be reconstructed with micrometric accuracy, rather independently of the data assimilation algorithm and the seeding concentration. The accurate evaluation of the static pressure field and of the surface pressure proved to be a challenge, with typical errors between 3 and 20% of the free-stream dynamic pressure.Dans le cadre du projet européen Horizon 2020 HOMER (Holistic Optical Metrology for Aero-Elastic Research), des algorithmes d’assimilation de données (DA) pour la reconstruction de champs d’écoulement denses, développés par différentes équipes de recherche — ci-après désignées comme les participants — ont été évalués de manière comparative. L’évaluation a été réalisée à l’aide d’une base de données synthétique reproduisant l’écoulement turbulent dans le sillage d’un cylindre en effet de sol, placé à une distance égale à un diamètre d’une paroi inférieure. En aval du cylindre, cette paroi se prolonge soit sous la forme d’une paroi plane fixe, soit d’un panneau flexible soumis à des oscillations périodiques ; ces deux situations correspondent à deux cas tests distincts, le second ayant été introduit pour étendre l’évaluation aux problèmes d’interaction fluide–structure. Les données d’entrée des algorithmes d’assimilation étaient constituées d’ensembles de données contenant les positions des particules et leurs numéros d’identification de trajectoire, avec des concentrations de traceurs croissantes allant de 0,04 à 1,4 particules/mm³ (ce qui correspond à des densités d’image comprises entre 0,005 et 0,16 particules par pixel, ppp). Les sorties des algorithmes DA considérées pour l’évaluation comprenaient : les trois composantes de la vitesse, les neuf composantes du tenseur de gradient de vitesse, ainsi que la pression statique, définies dans le champ d’écoulement sur une grille cartésienne, de même que la pression statique à la paroi et, dans le cas de la paroi déformable, sa position. Les résultats ont été analysés en termes d’erreurs des grandeurs de sortie par rapport aux valeurs de référence et de leurs distributions. De plus, les performances des différents algorithmes DA ont été comparées à celles d’une approche classique d’interpolation linéaire. Les erreurs de vitesse se situaient entre 3 et 11 % de la vitesse moyenne ; en outre, l’utilisation des algorithmes DA a permis d’augmenter la résolution spatiale des mesures d’un facteur compris entre 3 et 4. Les erreurs sur les gradients de vitesse étaient de l’ordre de 10 à 15 % de l’intensité maximale de la vorticité. Une reconstruction précise de la pression a été obtenue dans le champ d’écoulement, tandis que l’évaluation de la pression de surface s’est révélée plus complexe. Comme attendu, des erreurs plus faibles ont été observées pour des concentrations de traceurs plus élevées. Les différences de précision entre les résultats des divers algorithmes d’assimilation de données se sont révélées particulièrement marquées pour le champ de pression et la conformité aux équations de la dynamique des fluides, en particulier la conservation de la masse. L’analyse du cas test du panneau flexible a montré que la position du panneau pouvait être reconstruite avec une précision micrométrique, de manière assez indépendante de l’algorithme DA utilisé et de la concentration en traceurs. L’évaluation précise du champ de pression statique et de la pression de surface s’est avérée être un défi, avec des erreurs typiques comprises entre 3 et 20 % de la pression dynamique de l’écoulement libre
Challenges of neural network accelerators for aeronautics—position paper
No full textInternational audienc
Dense building deformation monitoring from InSAR based on LiDAR height measurement
No full textInternational audienceCreating digital twins of cities is essential to detect urban heat islands, or assess the risk of flooding. Incorporating the temporal axis, changes like urban spreading or urban renovation, can be encapsulated in the digital twins. At smaller scales, digital twins can also be used to follow the deformations of buildings that can be engendered by terrain movement due to subsidence or underground works. These deformation measurements can be extracted from InSAR temporal stack together with the building height [Weissgerber2017]. Moreover, InSAR acquisitions allow monitoring large areas with a bi-monthly update. However, the interpretation of the InSAR phase may be hindered by geometrical effects, such as layovers and shadows that are very present in dense urban environments. On the other hand, 3D description of buildings can be extracted from LiDAR point clouds with their fine height resolution and the large density of measured points. However, LiDAR acquisitions have a small footprint, requiring a large number of flights to monitor an entire city, leading to infrequent updates. In this work, our goal is to use LiDAR cloud points as a reference that can help disentangle the InSAR information, in order to extract information about building deformation. To do so, two steps are needed: 1. The geometrical alignment between the LiDAR point clouds and the SAR image to get a LiDAR reference height for each pixel of the studied buildings. This includes the selection of LiDAR points that corresponds to scatterers visible in the sar image. 2. The analysis of the InSAR phase temporal stack to detect deformation. The first step is to separate the contribution of the elevation in the InSAR phase and the contribution of the deformation. This is done by converting the reference altitude of each pixel found in step one to a reference phase for each InSAR acquisition. However, a phase difference that is due to the discrepancy between the phase scattering center height and the LiDAR geometry can be detecting since it is stable through time which the deformation pattern that changes with the acquisition. Then for InSAR pairs where deformation has been detected, the dense deformation pattern can be analyzed to detect non-uniform deformation across the building façade. This study is based on a temporal stack of 96 images acquired by TerraSAR-X/TanDEM-X between 2007 and 2012, with the reference image being chosen as the 24/01/2009. The LiDAR reference is the LiDAR HD acquired by IGN in 2023. Given the large time gap between the TerraSAR-X/TanDEM-X acquisition and the LiDAR HD one, we selected buildings with no structural changes between the two sources of data. Mainly, we focus on three buildings previously studied in [Weissgerber2017]: The Mirabeau Tower, The Cristal Tower and the Keller Tower. The footprint of the building is also supposed to be known. It is collected from the BD TOPO also produced by IGN. To align geometrically the LiDAR reference and the SAR images, we use the co-registration algorithm described in [Weissgerber2022] that enables to project each LiDAR point in the reference SAR image. Given the size of the LiDAR point cloud, we project only the LiDAR points that are classified as buildings in the building footprint. This enables removing the ground, vegetation points or artifacts that are labeled as such in the LiDAR point cloud. To select the LiDAR points that correspond to visible scatterers in the SAR image, the points are projected in the azimuth/range plane, without considering their height difference. Then, the visibility criteria are computed separately for each azimuth by assuming that the building façades are opaque. If the roof is visible, the points are considered as visible if their height is above the shadow cast by the points with a smaller range than theirs. If the roof is not visible, only the points having the smallest range are considered as visible. After having selected the LiDAR point visible in the SAR acquisition, all the SAR pixels may not have a height reference due to the scarcity of the LiDAR point cloud. To fill the gap in the façade, the height of the LiDAR points is linearly regressed between the ground and the top of the building. To increase the robustness of the linear regression, we use all the points that are on the same side of the building, a side being defined by a segment in the shapefile of the building footprint. Once this reference height is computed, it is converted into the reference phase using the InSAR acquisition metadata. Then, it can be subtracted from the measure InSAR phase to analyze the phase residue. The phase residue can either be due to deformation or from a discrepancy between this reference phase and the scatterers phase center. To compensate the latest, the reference height is updated by estimating the height offset from the conversion of the InSAR phase residue to height for acquisition close to the reference image, and acquisition in the same month but different years. This choice is motivated by the results of [Weissgerber2017] that showed that in this dataset, the deformations were mainly due to thermal expansion and thus were very small between acquisitions one year apart, for which temperatures are close. Having compensated the reference height for the difference between LiDAR height and the phase center height, the observed deformation patterns are mostly linear with the façade elevation, aligned with deformation due to thermal expansion. A more in-depth analysis has to be carried out to detect small phase anomalies that could indicate non-homogenous thermal deformation or deformation from another origin. Moreover, the difference between the height and the phase center height could also be analyzed either to understand the scattering mechanism of these buildings façades or to update the 3D representation of buildings by details captured only in the InSAR phase. References [Weissgerber2017] Flora Weissgerber, Elise Colin-Koeniguer, Jean-Marie Nicolas et Nicolas Trouvé, “3D Monitoring of Buildings Using TerraSAR-X InSAR, DInSAR and PolSAR Capacities”, Remote Sens. 2017, 9(10), 1010; doi:10.3390/rs9101010 [Weissgerber2022] Flora Weissgerber, Laurane Charrier, Cyril Thomas , Jean-Marie Nicolas and Emmanuel Trouvé (2022) LabSAR, a one-GCP coregistration tool for SAR–InSAR local analysis in high-mountain regions. Front. Remote Sens. 3:935137. doi: 10.3389/frsen.2022.93513
