Portail HAL ONERA
Not a member yet
12842 research outputs found
Sort by
Preuve de concept d'un lidar à courte portée et à haute résolution spectrale utilisant un laser à fibre compact à taux de répétition élevé
No full textInternational audienceIn recent years, several climate and air quality applications have required to understand the impact of aerosols close to their source, leading to the development of novel Short-Range Elastic Backscatter Lidars (SR-EBLs), which enable measuring the radiative properties of aerosols at high spatiotemporal resolutions (<10cm, 1s) in the short-range (3 to 500m). However, the elastic lidar equation is an ill-posed problem, having one equation for two atmospheric variables: the backscatter β(r) and extinction α(r) coefficients. Solving this equation requires assuming a value for the lidar ratio, i.e., a linear relationship between β and α, reducing the accuracy of the retrievals. Advanced lidar techniques, like the High Spectral Resolution Lidar (HSRL), measure molecular and particle scattering separately. Having a direct measurement of the molecular component allows for solving the lidar problem without assumptions about the lidar ratio. However, the existing atmospheric HSRLs cannot perform short-range measurements because i) they are usually blind in the first hundredths of meters (overlap restrictions), and ii) they prioritize spectral performance using ultranarrow band (and thus long-pulse) lasers, resulting in an insufficient spatiotemporal resolution.This work presents a proof-of-concept of a Short-Range High Spectral Resolution Lidar (SR-HSRL) optimized for aerosol characterization in the first kilometer of the atmosphere. This SR-HSRL uses a compact high-repetition rate fiber laser source with a 300 MHz linewidth and 5 ns pulse length. Since these two parameters are inversely proportional, and both are required for performing SR-HSRL measurements, a compromise had to be found to optimize the overall performance. The main challenge was to prove that, despite its relatively large linewidth, this laser has a satisfactory spectral performance so that it can be used for future implementations of the short-range HSRL. We chose this model after evaluating several laser sources because it has the right compromise between pulse length, linewidth, spectral stability, and size. The laser housing is 270 x 270 x 40 mm and weighs 2.9 kg, making it ideal for future integration on a portable short-range HSRL system.In the receiver part, a 10:90 beam splitter transmits 10% of the backscattered light to the total channel and reflects 90% of it to the HSR channel. A 40-cm-long iodine cell is used as the spectral filter for separating the Mie and Rayleigh aerosol components. We used two thermoelectrically cooled SiPM Multi-Pixel Photon Counter (MPPC) sensors and a 160MHz analog-to-digital converter to measure the signals. The spatiotemporal resolution, limited by the acquisition system, is 7.5 m and 1 s.To test the lidar, a two-day measurement campaign was performed at NIES in Tsukuba, Japan, in July 2024. We demonstrate that, despite having a relatively large laser linewidth, we can successfully remove the Mie aerosol component, retrieving aerosol backscatter coefficient profiles from as low as 80 m. We also compare the HSRL retrieval method to a non-conventional forward Fernald inversion method previously reported for SR-EBL. We found that the forward method normally sub-estimates β (up to 30% discrepancy) in aerosol layers and overestimates it in cloud zones (60 to >100% difference).Ces dernières années, plusieurs applications liées au climat et à la qualité de l'air ont nécessité de comprendre l'impact des aérosols à proximité de leur source, ce qui a conduit au développement de nouveaux lidars à rétrodiffusion élastique à courte portée (SR-EBL), qui permettent de mesurer les propriétés radiatives des aérosols à des résolutions spatiotemporelles élevées (<10 cm, 1 s) à courte portée (3 à 500 m). Cependant, l'équation élastique du lidar est un problème mal posé, ayant une équation pour deux variables atmosphériques : les coefficients de rétrodiffusion β(r) et d'extinction α(r). La résolution de cette équation nécessite de supposer une valeur pour le rapport lidar, c'est-à-dire une relation linéaire entre β et α, ce qui réduit la précision des récupérations. Les techniques lidar avancées, comme le lidar à haute résolution spectrale (HSRL), mesurent séparément la diffusion moléculaire et la diffusion des particules. Avoir une mesure directe de la composante moléculaire permet de résoudre le problème du lidar sans hypothèses sur le rapport lidar. Cependant, les HSRL atmosphériques existants ne peuvent pas effectuer de mesures à courte portée car i) ils sont généralement aveugles dans les premiers centièmes de mètres (restrictions de chevauchement), et ii) ils privilégient les performances spectrales en utilisant des lasers à bande ultra-étroite (et donc à impulsions longues), ce qui entraîne une résolution spatiotemporelle insuffisante.Ce travail présente une preuve de concept d'un lidar à haute résolution spectrale à courte portée (SR-HSRL) optimisé pour la caractérisation des aérosols dans le premier kilomètre de l'atmosphère. Ce SR-HSRL utilise une source laser à fibre compacte à taux de répétition élevé avec une largeur de raie de 300 MHz et une longueur d'impulsion de 5 ns. Étant donné que ces deux paramètres sont inversement proportionnels et qu'ils sont tous deux nécessaires à la réalisation de mesures SR-HSRL, un compromis a dû être trouvé pour optimiser les performances globales. Le principal défi était de prouver que, malgré sa largeur de raie relativement importante, ce laser a des performances spectrales satisfaisantes afin qu'il puisse être utilisé pour les futures implémentations du HSRL à courte portée. Nous avons choisi ce modèle après avoir évalué plusieurs sources laser car il présente le bon compromis entre longueur d'impulsion, largeur de ligne, stabilité spectrale et taille. Le boîtier laser mesure 270 x 270 x 40 mm et pèse 2,9 kg, ce qui le rend idéal pour une future intégration sur un système HSRL portable à courte portée.Dans la partie réceptrice, un séparateur de faisceau 10:90 transmet 10 % de la lumière rétrodiffusée au canal total et en réfléchit 90 % au canal HSR. Une cellule à iode de 40 cm de long est utilisée comme filtre spectral pour séparer les composants des aérosols Mie et Rayleigh. Nous avons utilisé deux capteurs SiPM Multi-Pixel Photon Counter (MPPC) refroidis thermoélectriquement et un convertisseur analogique-numérique de 160 MHz pour mesurer les signaux. La résolution spatiotemporelle, limitée par le système d'acquisition, est de 7,5 m et 1 s.Pour tester le lidar, une campagne de mesure de deux jours a été réalisée au NIES à Tsukuba, au Japon, en juillet 2024. Nous démontrons que, malgré une largeur de ligne laser relativement importante, nous pouvons éliminer avec succès la composante d'aérosol de Mie, en récupérant des profils de coefficient de rétrodiffusion d'aérosol à partir de 80 m. Nous comparons également la méthode de récupération HSRL à une méthode d'inversion de Fernald directe non conventionnelle précédemment rapportée pour SR-EBL. Nous avons constaté que la méthode directe sous-estime normalement β (jusqu'à 30 % d'écart) dans les couches d'aérosols et le surestime dans les zones nuageuses (différence de 60 à > 100 %)
Concurrent Value-Driven Optimization Problems Addressing Aircraft Design, Manufacturing, and Supply Chain
No full textInternational audienceIn the last decades, several studies show how the concurrent evaluation of manufacturing and supply chain in the early design phase brings significant advantages to industries. In this frame, a value-driven methodology has been developed and applied in the aeronautical context to identify the best solution when considering design, manufacturing, and supply chain criteria at the same time. This study aims at including optimization algorithms in the methodology to face a new challenge: the identification of solutions simultaneously optimizing manufacturing, design, and supply chain variables. To achieve this objective, collaborative optimization problems are presented in this research activity in three multidisciplinary design and optimization (MDO) cases, which integrate domains first in pairs and then together. Shortly, MDO case I identifies a Pareto front of optimal supply chain combinations producing aircraft components among millions of possibilities after several hours of execution; MDO case II shows the optimal aircraft configuration made by considering hundreds of combinations of materials and processes in less than one hour; MDO case III reaches the Pareto front in a half-day considering millions of choices of materials, processes, and enterprises
Influence of Oxygen and Zirconium Additions on Oxidation Resistance and Mechanical Properties of Ti-Al and Ti-Al-Zr Alloys
No full textInternational audienc
Amélioration de modèles Reynolds-Averaged Navier-Stokes instationnaires à partir de données éparses par assimilation de données séquentielle et apprentissage automatique
No full textInternational audienceA Bayesian-based approach is developed to learn predictive turbulence-model corrections for unsteady flow simulations. A distinct feature of the present approach is its ability to perform such a learning task using limited data, which is characteristic of realistic configurations where full sampling can be difficult. Relying on the Expectation-Maximization formalism, the learning task is performed in two steps that optimally combine the strengths of data-assimilation and machine-learning techniques. In a first step, an Ensemble Kalman Filter is used to perform sequential state estimation, namely inferring full flow representations from the considered sparse unsteady data. In a second step, the thus-obtained full states are used to form a training dataset to build the turbulence-model corrections. The present methodology is employed to learn corrective terms for the unsteady Reynolds-Averaged Navier-Stokes (URANS) equations closed by the Spalart-Allmaras model. The sparse data that are used for training are given in the form of a limited number of spatially pointwise velocity observations that are extracted from a Direct Numerical Simulation of the flow past a circular cylinder at Re = 3900. It is shown that the corrected URANS model that is obtained via this strategy significantly outperforms the baseline model despite of the sparse nature of the considered data.Une approche bayésienne est développée pour apprendre des corrections prédictives à des modèles de turbulence pour des simulations instationnaires. Une caractéristique distincte de la présente approche est sa capacité à effectuer une telle tâche d’apprentissage en utilisant des données limitées, ce qui est caractéristique des configurations réalistes où un échantillonnage complet peut être difficile. S'appuyant sur le formalisme d'expectation-maximization, la tâche d'apprentissage est réalisée en deux étapes qui combinent de manière optimale les atouts de techniques d'assimilation de données et d'apprentissage automatique. Dans un premier temps, un filtre de Kalman d'ensemble est utilisé pour effectuer une estimation d'état séquentielle, c'est-à-dire déduire des représentations d'écoulements complètes à partir de données instationnaires clairsemées. Dans un deuxième temps, les états complets ainsi obtenus sont utilisés pour former un ensemble de données d'entraînement afin de construire des corrections à un modèle de turbulence. La présente méthodologie est utilisée pour apprendre des termes correctifs pour les équations Reynolds-Averaged Navier-Stokes instationnaires (URANS) fermées par le modèle Spalart-Allmaras. Les données éparses considérées correspondent à un nombre limité d'observations de vitesse spatialement ponctuelles qui sont extraites d'une simulation numérique directe de l'écoulement autour d'un cylindre circulaire à Re = 3900. Il est montré que le modèle URANS corrigé obtenu via cette stratégie surpasse considérablement le modèle de base malgré la nature clairsemée des données considérées
Flight Testing Active Flutter Suppression Technologies, the Euroean FliPASED Perspective
No full textInternational audienceFriction damping is one of the main mechanisms for dissipating vibration energy in turbomachinery. For assembled bladed disks, friction typically arises in the blade root — in the disk interface or, for shrouded blade configurations, in the interface between two consecutive shrouds. For integrally bladed disks, where no such interface naturally occurs, additional damping devices must be designed if energy is to be dissipated through friction. Split-ring dampers are such devices: they are located in a circumferential groove under the blisk platform, and pressed onto it by centrifugal loads. At low vibration levels and/or high rotating speeds, the split ring remains stuck to the disk, while at high levels, relative motion starts and dissipates vibrational energy. Such devices are often designed to mainly target a given nodal diameter mode of vibration. Nevertheless, designing these dampers is a challenging task. Indeed, their behavior depends on the design of the ring (dimensions, material and interface properties, etc.) and on the location of implementation; many different phenomena are involved such as centrifugal and thermal loading, mistuning, etc. Realistic contact modeling and simulation are also difficult because the dynamical behavior is non-linear due to the friction interfaces. In this work, the effect of a split ring damper is investigated experimentally on an industrial compressor blisk in a vacuum chamber. A traveling wave excitation with a prescribed nodal diameter index is produced using piezoelectric actuators. Blade vibrations are monitored through strain gauges. Forced responses are measured, first on a tuned configuration activating the split ring damping effect on the targeted specific nodal diameter. Then an intentional mistuning pattern is implemented, so that the split ring target nodal diameter appears in different modes of the new configuration. The experiments show that resonance peaks are damped for different modes from that targeted at the tuned design stage
Atomic‐Scale Insights Into the Thermal Stability of High‐Entropy Nanoalloys
No full textInternational audienceHigh entropy alloy nanoparticles bring hope to developing more efficient nanomaterials for high‐temperature applications. Nevertheless, the enhanced thermal stability of nearly equiatomic nanoalloys containing at least 5 metals is nothing more than theoretical speculation about the impact of thermodynamic contributions on their structural properties and remains to be proven. Here, in situ aberration‐corrected scanning transmission electron microscopy (STEM) and molecular dynamics simulations are combined to investigate at the atomic scale the thermal behavior of AuCoCuNiPt nanoparticles (NPs) from 298 to 973 K. Both in situ STEM heating and atomistic simulations reveal strong structural and chemical evolutions in the NPs with the formation and melting of an AuCu layer at the surface of NPs at high temperature. This phase separation that appears progressively with temperature is driven by pronounced atomic diffusion that is surprisingly more active in these quinary nanoalloys than in monometallic and bimetallic subsystems. Besides ruling out the existence of sluggish diffusion in AuCoCuNiPt nanoalloys and lowering their temperature range of application, the study allows distinguishing kinetic and thermodynamic effects on their structural properties, which is an essential prerequisite to better control the synthesis of complex nanomaterials
Sputtering yield and induced contamination of Kapton, PEEK and PTFE under xenon ions bombardment
No full textInternational audienceIn this article, we present the results of sputtering yield measurements of Kapton, PEEK and PTFE due to xenon ions bombardment. We measured these yield using the weight loss technique, for energies ranging from \qtyrange[range-units = single]{350}{800}{\eV} and incidence angles from \qtyrange[range-units = single]{0}{75}{\degree}. Polymers sputtering yield dependence versus incidence angle appears to be much lower than what we usually observe on other materials (such as metals). Furthermore, PTFE exhibits a very high erosion rate, which is two orders of magnitude bigger than Kapton and PEEK one. We also performed collection measurements using a quartz crystal microbalance (QCM) apparatus. Such method provides differential sputtering yields data which are key pieces of information regarding satellites contamination issues. We also computed the total sputtering yields obtained by this second method, and it appeared that they were much smaller than the ones measured previously by weight loss, specially for PTFE. This suggests that erosion products may be of different types (atoms, molecules) which are not all able to stick on a facing surface
From Perception to Decision: Emergence and Variation of Trust During Decision Making
No full textInternational audienceTrust is central in human-AI collaboration. In order to understand how trust emerges towards automated systems, decisional trust at individual level needs to be deepened. While the emergence of trust can be regarded using metacognition and accumulation evidence theories, its variation can be likened to a temporal variable whose dynamics change over time and can be studied through learning theories. Here, we aim to assess the effect of learning and the impact of evaluation of confidence on trust variations for decision making. To this end, participants will perform a perceptual decision-making task of varying difficulty to account for evidence accumulation, confidence in their own decision and prediction emerging from learning, combined with feedback on their performance. Subjective, behavioral and electroencephalographic measures will be recorded during the task to unveil dynamics and markers of trust. The data collection of 40 participants will start next month, providing preliminary results for defining and understanding the interactions between these different processes. Eventually, the identified markers and features can be transposed to the comprehension of trust towards AI decision making
Measuring and Addressing Information Leakage in Concept Bottleneck Models
No full textInternational audienceConcept Bottleneck Models (CBMs) are transparent models that use a human-interpretable bottleneck layer to improve interpretability in tasks such as image classification. However, model transparency does not guarantee that decisions are supported by relevant information. When this is not the case, we speak of information leakage. To provide a quantitative measure of this property, we introduce a metric called Irrelevant Concept Contribution (ICC), the first metric that directly quantifies information leakage for concept bottleneck models. Using this metric, we propose a comprehensive comparison of bottleneck architectures, training strategies, and decision processes. In particular, we demonstrate that state-of-the-art performance in image classification can be achieved using only relevant concepts that contribute positively to the decision process on commonly used datasets. This makes the decision process relevant by design and in an explicit way
Learning to Solve Resource-Constrained Project Scheduling Problems with Duration Uncertainty Using Graph Neural Networks
No full textInternational audienceThe Resource-Constrained Project Scheduling Problem (RCPSP) is a classical scheduling problem that has received significant attention due to of its numerous applications in industry. However, in practice, task durations are subject to uncertainty that must be considered in order to propose resilient scheduling. In this paper, we address the RCPSP variant with uncertain tasks duration (modeled using known probabilities) and aim to minimize the overall expected project duration. Our objective is to produce a baseline schedule that can be reused multiple times in an industrial setting regardless of the actual duration scenario. We leverage Graph Neural Networks in conjunction with Deep Reinforcement Learning (DRL) to develop an effective policy for task scheduling. This policy operates similarly to a priority dispatch rule and is paired with a Serial Schedule Generation Scheme to produce a schedule. Our empirical evaluation on standard benchmarks demonstrates the approach's superiority in terms of performance and its ability to generalize. The developed framework, Wheatley, is made publicly available online to facilitate further research and reproducibility