Portail HAL ONERA
Not a member yet
12842 research outputs found
Sort by
Transfer learning in surrogate modeling with emphasis on aircraft design
No full textInternational audienceSurrogate modeling with insufficient data can lead to high prediction uncertainty and errors. A promising remedy to address this issue is the use of transfer learning techniques that leverage models built using data from other problems that are implicitly related to the problem of interest. We present an algorithm that uses transfer learning and mixtures of experts across different design space regions to improve the predictive capability of surrogate models. The algorithm uses existing data to divide a problem’s design space into clusters and build ensembles of surrogate models in each cluster using a multi-criteria weighting method. The proposed algorithm is shown to be both accurate and flexible, allowing for automated transfer learning with tuning parameters that cater for different problem types. The multi-criteria approach enables transfer learning in constrained Bayesian optimization by weighing models based on their shape, accuracy, and variance. The proposed method is demonstrated using aircraft conceptual design examples and showed up to 10% reduction in prediction errors
Prioritized multi-objective optimization of an aircraft flight performance based on Nash games from preponderant Pareto-optimal points
No full textInternational audienceA hierarchical multi-objective optimization process is applied to the performance of an Airbus-A320-type aircraft. Six design variables are used to calibrate wing geometry, mean aerodynamic chord and take-off thrust potential. Functional constraints are enforced (interval bound on static margin and upper bound on wing span). The system is subject to the classical Breguet laws of Flight Mechanics and to the numerical integration of the ascent phase. It is thus an Ordinary-Differential-Equation-constrained system and it is evaluated by the open-source software FAST-OAD. Three cost functionals are minimized: fuel mass at take-off, operational empty weight, and ascent duration. The first two, considered of preponderant importance are prioritized, and the third only secondary. The method permits, at moderate numerical cost, to identify quasi-Pareto-optimal by Nash games originating from Pareto-optimal solutions of the sole prioritized cost functionsUn processus d'optimisation hiérarchique multi-objectifs est appliqué aux performances d'un avion de type Airbus-A320. Six variables de conception sont utilisées pour calibrer la géométrie de l'aile, la corde aérodynamique moyenne et le potentiel de poussée au décollage. Des contraintes fonctionnelles sont appliquées (limite d'intervalle sur la marge statique et limite supérieure sur l'envergure de l'aile). Le système est soumis aux lois classiques de Breguet de la mécanique du vol et à l'intégration numérique de la phase d'ascension. Il s'agit donc d'un système d'équations différentielles ordinaires. Il est évalué par le logiciel libre FASTOAD. Trois fonctions de coût sont minimisées : la masse de carburant au décollage, le poids à vide opérationnel et la durée de l'ascension. Les deux premières, considérées comme prépondérantes, sont prioritaires et la troisième n'est que secondaire. La méthode permet, à un coût numérique modéré, d'identifier des solutions Pareto optimales des jeux de Nash à partir des solutions Pareto-optimales des seules fonctions de coût hiérarchisées
On Operational Diagnosis for Ground Stations: A Model-Based Approach
No full textInternational audienceChallenges faced in the field of operational diagnosis have grown a deal in the last decade, especially for complex, time-critical systems. In the meantime, although model-based techniques are being used widely to address the design of complex systems, but are not extended to their operation—in particular when dealing with faults: operators can benefit from the use of formal models for system monitoring and diagnostics. We thus propose a methodology to create a new type of Operations Dedicated Model (ODM) from the already existing system design models (functional and dysfunctional), with the use of the Behaviour Tree (BT) formalism. With the assumption that Safety Analysis (SA) models describe dysfunctional aspects of the system—notably via Fault Tree Analysis (FTA), we us Fault Trees (FTs) as an input for the ODM construction. We also demonstrate our proposed approach on a Satellite Ground Station example, and discuss how ODMs can improve systems’ operations
Effect of dilution and Lewis number of the fuel in the fuel-air mixture on the heat flux during flame wall interaction (FWI)
No full textInternational audienceMost combustion systems where flame is bounded by walls encounter Flame-wall interactions (FWI). Heat losses during FWI are in the magnitude of MW/m2 leading to concern about inefficiency, and thermal stresses. Hence, FWI is a topic of significant interest. With the goals of climate change pushing us toward renewable fuels with least carbon emissions, it is imperative to understand the FWI of renewable fuels like hydrogen-air mixture. In this study, FWI in a head-on-quenching configuration is carried out in a constant volume chamber. High-speed surface temperature measurement is carried out to compute instantaneous wall heat flux. Adding inert diluents to methane-air mixtures, the effect of dilution on the peak of heat flux during FWI is studied. Dilution affects the peak of heat flux predominantly by changing the adiabatic flame power of the fuel-air mixture. Furthermore, a large variation in flame power is carried out by changing the dilution fraction in the methane-air mixture to quantify its effect. It is found that the peak of heat flux and non-dimensional heat flux follows a nonlinear decrease with an increase in flame power. Knowing this effect, FWI of three different fuels, that is, methane, hydrogen, and a mixture of acetylene and hydrogen, having different Lewis numbers (Le) of fuel in a fuel-oxidizer mixture are studied at equivalent flame power. This comparison shows that FWI of hydrogen-air mixtures in HOQ configuration have a lower peak of heat flux compared to other fuels which may be due to the preferential diffusion of hydrogen-air mixture
Détermination de l'énergie de rupture d'interface d'une barrière environnementale d'un composite à matrice céramique
No full textInternational audienceThe present work concerns the study of the interface fracture energy between a SiC/SiC Ceramic Matrix Composite and an environmental barrier coating. Four-point flexural tests with no precrack were conducted. These tests enable for the stable propagation of two interfacial cracks. They were carried out at room temperature and were instrumented with visible light cameras. This instrumentation allowed for the analysis of the tests thanks to digital image correlation as well as comparisons between experimental and numerical results to locate crack tips and to calculate the interface fracture energy using numerical methods based on linear elastic fracture mechanics. The limits of the method as well as the uncertainties associated with the crack length and the fracture energy were assessed.Ce travail concerne l’étude de la ténacité d’interface entre un composite à matrice céramique SiC/SiC (CMC) et sa barrière environnementale. Des essais de flexion à quatre points sans préfissure ont été réalisés. Ces essais permettent la propagation stable de deux fissures interfaciales. Ils ont été effectués à température ambiante et ont été instrumentés avec des caméras optiques. Cette instrumentation a permis l’analyse des essais grâce à la corrélation d’images numériques, ainsi que des comparaisons entre les résultats expérimentaux et numériques pour localiser les extrémités des fissures et calculer la ténacité d’interface en utilisant des méthodes numériques basées sur la mécanique de la rupture élastique linéaire. Les limites de la méthode ainsi que les incertitudes associées à la longueur de la fissure et à la ténacité à la fracture ont été évaluée
Experimental classification of dynamic speckle regimes : insights from controlled rotational diffuser measurements
No full textInternational audienceDynamic speckle phenomena, arising from coherent light scattering on moving diffuse surfaces, are widely used for motion analysis in fields such as biomedical imaging and industrial inspection. However, the classification of dynamic speckle regimes remains challenging, particularly in understanding their distinct temporal and spatial behaviors across various velocity ranges. In this study, we introduce a comprehensive framework for categorizing speckle dynamics into three regimes: frozen, intermediate, and fully decorrelated. Each exhibits distinct temporal decorrelation properties, with direct consequences for motion quantification. To validate this framework, we designed an experimental setup comprising a coherent laser source, a controlled rotational diffuser as the moving scattering surface, and a high-resolution imaging system. This configuration enables precise control of speckle motion and systematic sampling of a wide velocity range. The experimental results reveal consistent and reproducible transitions between the three regimes, in strong agreement with the predicted contrast–velocity relationships. Our findings underscore the practical significance of this classification. In particular, they demonstrate that system performance depends critically on the regime in which measurements are made. Accurate velocity estimation requires adapting the acquisition strategy to the specific characteristics of each regime, including frame rate and exposure time. The intermediate regime, where contrast varies only weakly with velocity, should be avoided in system design due to its poor sensitivity. In addition to clarifying speckle dynamics, this framework enables the optimization of imaging system parameters by ensuring that measurements are performed in regimes where contrast is most sensitive to motion
Méthodes de diffusion et apprentissage par score-matching appliqués aux équations de Navier–Stokes stochastiques
No full textInternational audienceDans le cadre déterministe, les équations de Navier–Stokes incompressibles décrivent les fluides visqueux et sont, en deux dimensions, globalement bien posées sous des hypothèses classiques. Cependant, elles ne capturent pas la variabilité et les phénomènes multi-échelles de la turbulence, liés aux dynamiques non résolues et aux incertitudes de modélisation. Les formulations stochastiques pallient cette limite en introduisant un bruit modélisant les fluctuations et l’incertitude, con- duisant à des équations différentielles partielles stochastiques, reliant mécanique des fluides et physique statistique. La solution du système stochastique devient alors un processus aléatoire dont la densité évolue selon une équation de Fokker–Planck, établissant un lien entre modélisation physique et inférence statistique. Ce cadre ouvre la voie aux modèles génératifs fondés sur le score-matching, qui apprennent le gradient du logarithme de la densité de probabilité. En paramé- trant ce champ par un réseau de neurones, on restitue la géométrie statistique des champs de vitesse sans expliciter la den- sité, permettant ensuite de reconstruire ou d’échantillonner les statistiques du flot via des processus de diffusion inverses. Ces méthodes s’appuient sur l’essor des modèles génératifs de diffusion, une classe de modèles particulièrement puissants pour l’apprentissage de distributions complexes
Hierarchical Coordination of UAVs for Dynamic Task Assignment in Large-Scale Traffic Surveillance Missions
No full textInternational audienceThis paper presents an hierarchical coordination architecture for a fleet of UAVs dedicated to road traffic surveillance over large urban areas. The system is built around a central drone, acting as a coordinator, which is responsible for monitoring the status of the fleet and dynamically assigning surveillance tasks in response to reported traffic events. To ensure scalability and responsiveness, our architecture combines a spatial clustering mechanism to partition mission area and distribute drones accordingly, with a receding horizon task assignment (RHTA) strategy within each sub-region. The fleet coordination requires designing specific trajectories for the central drone to ensure communication within the fleet and periodic updates of the surveillance information. This hybrid approach enables adaptive, region-based task allocation while preserving a global overview through the coordinator. Simulation results highlight the relevance and flexibility of the proposed coordination sc heme when addressing dynamic and large-scale surveillance scenarios
Identification of a full-scale aircraft by phase-locked loop experiment
No full textInternational audienc
Optical Thermometry for Dynamic Imaging of Heat Transport in Analog Porous Media
No full textInternational audienceUnderstanding the interplay between thermal advection by fluid flow, conduction, and the structural complexity of structures bearing flow is critical for geothermal energy, aquifer thermal energy storage, and characterization of hydrologic systems. Yet capturing and quantifying this interplay at relevant scales remains challenging. Here we analyze the thermal dynamics in analogous porous media from laboratory experiments that allow pore‑scale measurements of the fluid's temperature field, using a novel optical thermometry approach.Experiments are conducted in a flow cell containing a quasi-2D porous medium (15 × 6 cm, porosity 0.36, minimum pore size 0.5 mm) consisting of cylindrical grains whose thermal properties mimic those of natural aquifers. The resident fluid at room temperature is displaced by injection of the same fluid heated at 50 °C while temperature fields of the liquid phase are imaged at regular time intervals. The fluid's temperature is inferred from the thermally activated delayed fluorescence decay of Zr(MesIPDPt‑BuPh)2 colloids, a bright and photostable zirconium(IV) complex with sensitivity >1.9 %/K. High‑speed imaging under modulated LED excitation captures the decay, which is calibrated to the temperature and provides non‑invasive, spatially distributed thermometry. Temperature fields are thus computed every 2.5 s with a ±0.3 K precision, a 131 ms temporal, and a 0.3 mm spatial resolution.The resulting temperature maps reveal thermal front propagation influenced by the existence of preferential flow paths and gradients indicative of Local Thermal Non‑Equilibrium (LTNE), where fluid and solid phases maintain distinct temperatures. These patterns, inaccessible to point-wise methods such as thermocouples, demonstrate the potential of optical thermometry for quantitative characterization of heat transport in heterogeneous porous media in the laboratory. Comparisons with numerical simulations reproducing the experimental geometry and conditions (medium porosity, grain sizes and positions, Darcy flux) validate the observations and encourage ongoing extensions to other geological settings and flow regimes, as well as reactive processes