Portail HAL ONERA
Not a member yet
12842 research outputs found
Sort by
Slalom-tailoring haptic piloting assistance: A comparison between a static force field-based design and a design inherited from human motion
No full textInternational audienceHaptic flight controls have been extensively studied across various applications over the last years. This study focuses on the slalom task as defined by ADS-33F, an obstacle avoidance scenario. We examine two haptic designs derived from a common model of obstacle repulsion and goal attraction. The first design is based on a force field, representing an information processing approach that adjusts force feedback according to the distance to obstacles. The second design follows an ecological approach, inspired by goal-steering dynamics in human walking. Both designs were tested using a Pilot-in-the-Loop simulator with various gate configurations to modulate geometrical difficulty, and two control modes to examine the impact of the Controlled Element's dynamics. The results are analyzed in terms of average Cycle Time, velocity, acceleration, and applied force. Sidestick activity is modeled using Mottet's model of human arm kinematics . Participants also completed a NASA-Task Load Index questionnaire after each trial to evaluate perceived task load. This paper presents initial findings from a study involving 10 participants. Table 4: Notations from human arm kinematics model [2
Unlocking 3D nanoparticle shapes from 2D HRTEM images: Deep Learning for classification at atomic resolution
No full textInternational audienc
Confidence Calibration of Classifiers with Many Classes
No full textInternational audienceFor classification models based on neural networks, the maximum predicted class probability is often used as a confidence score. This score rarely predicts well the probability of making a correct prediction and requires a post-processing calibration step. However, many confidence calibration methods fail for problems with many classes. To address this issue, we transform the problem of calibrating a multiclass classifier into calibrating a single surrogate binary classifier. This approach allows for more efficient use of standard calibration methods. We evaluate our approach on numerous neural networks used for image or text classification and show that it significantly enhances existing calibration methods. Our code can be accessed at the following link: https://github.com/allglc/tva-calibration
AltaRica Wizard : un environnement de modélisation et de simulation pour l'analyse de systèmes d'équations Booléennes stochastiques
No full textInternational audienceAltaRica Wizard is an integrated modeling environment dedicated to probabilistic risk and safety analyses of complex technical systems. The new version of AltaRica Wizard handles two modeling languages: AltaRica 3.0 and S2ML+SBE. These two languages are part of the S2ML+X family. S2ML stands for System Structure Modeling Language, a complete set of object oriented (classes and prototypes) constructs to structure models. The X represents any mathematical framework suitable to describe the stochastic behavior of systems. With S2ML+SBE, SBE means systems of Stochastic Boolean Equations, which is the underlying mathematical framework of the popular modeling formalisms such as fault trees or reliability bloc diagrams. S2ML+SBE generalizes these two formalisms without introducing more complexity to design and assess models, from a qualitative and quantitative point of view. The aim of this article is to present AltaRica Wizard as an integrated modeling environment dedicated to systems of Stochastic Boolean Equations. This includes the editor to design S2ML+SBE models, the associated compiler of Boolean equations (fault trees), and the integration of the XFTA engine.AltaRica Wizard est un environnement de modélisation intégré dédié aux analyses probabilistes de risque et de sécurité des systèmes techniques complexes. La nouvelle version d’AltaRica Wizard prend en charge deux langages de modélisation : AltaRica 3.0 et S2ML+SBE. Les deux langages appartiennent à la famille S2ML+X : S2ML signifie langage de modélisation de structure des systèmes (‘System Structure Modeling Language’ en anglais), c'est-à-dire un ensemble complet de constructions génériques orientées objet à classes et à prototypes pour structurer les modèles ; X représente tout cadre mathématique adapté pour décrire le comportement (stochastique) des systèmes. Dans S2ML+SBE, SBE signifie systèmes d'équations booléennes stochastiques (systems of Stochastic Boolean Equations), qui constituent le cadre mathématique sous-jacent des formalismes de modélisation populaires tels que les arbres de défaillances et les schémas bloc de fiabilité. S2ML+SBE généralise ces deux formalismes, sans introduire de complexité supplémentaire dans la création de modèles et le calcul des indicateurs de sécurité qualitatifs et quantitatifs. L'objectif de cet article est de présenter AltaRica Wizard, en tant qu'environnement de modélisation intégré pour l'évaluation de systèmes d'équations booléennes (stochastiques). Cela inclut l'éditeur pour créer les modèles S2ML+SBE, le compilateur d'équations booléennes (arbres de défaillance) et l'intégration du moteur de calcul XFTA
Optimization of a Harmonic Structured Ti-24Nb-4Zr-8Sn Alloy through Combined Microstructural and Mechanical Investigations
No full textInternational audienc
Thevenin black box model to characterize conducted electromagnetic interference of DC/AC converters for aerospace applications
No full textInternational audienceThis article focuses on behavioral modelling of power systems, specifically DC/AC converters to evaluate electromagnetic (EM) interferences as conducted emissions current, at both inputs and outputs. A linear equivalent model of the converter is proposed as a five-port black box model, which allows for interaction between the input and output ports. To implement it, we associated it with experimental setups for measuring impedance and currents in frequency domain. The experimental setup only provides the amplitude of the current, so three hypotheses were proposed to determine the phase shift between the currents. To validate our model, we compare the measurements and calculations of current in common mode in conducted emissions for these three hypotheses.Cet article propose une modélisation comportementale des systèmes d'alimentation, en particulier des convertisseurs DC/AC, afin d'évaluer les interférences électromagnétiques (EM) en tant que courant d'émission conduit, à la fois au niveau des entrées et des sorties. Un modèle équivalent linéaire du convertisseur est proposé sous la forme d'un modèle de boîte noire à cinq ports, qui permet l'interaction entre les ports d'entrée et de sortie. Pour le mettre en œuvre, nous l'avons associé à des dispositifs expérimentaux pour mesurer l'impédance et les courants dans le domaine des fréquences. Le dispositif expérimental ne fournissant que l'amplitude du courant, trois hypothèses ont été proposées pour déterminer le déphasage entre les courants. Pour valider notre modèle, nous comparons les mesures et les calculs du courant en mode commun dans les émissions conduites pour ces trois hypothèses
Toxicological evaluation of primary particulate matter emitted from combustion of aviation fuel
No full textInternational audienceRecently, Sustainable Aviation Fuel (SAF) blends and novel combustion technologies have been introduced to reduce aircraft engine emissions. However, there is limited knowledge about the impact of combustion technology and fuel composition on toxicity of primary Particulate Matter (PM) emissions, comparable to regulated non-volatile PM (nvPM). In this study, primary PM was collected on filters using a standardised approach, from both a Rich-Quench-Lean (RQL) combustion rig and a bespoke liquid fuelled Combustion Aerosol Standard (CAST) Generator burning 12 aviation fuels including conventional Jet-A, SAFs, and blends thereof. The fuels varied in aromatics (0–25.2%), sulphur (0–3000 ppm) and hydrogen (13.43–15.31%) contents. Toxicity of the collected primary PM was studied in vitro utilising Air-Liquid Interface (ALI) exposure of lung epithelial cells (Calu-3) in monoculture and co-culture with macrophages (differentiated THP-1 cells). Cells were exposed to PM extracted from filters and nebulised from suspensions using a cloud-based ALI exposure system. Toxicity readout parameters were analysed 24 h after exposure.Results showed presence of genotoxicity and changes in gene expression at dose levels which did not induce cytotoxicity. DNA damage was detected through Comet assay in cells exposed to CAST generated samples. Real-Time PCR performed to investigate the expression profile of genes involved in oxidative stress and DNA repair pathways showed different behaviours after exposure to the various PM samples. No differences were found in pro-inflammatory interleukin-8 secretion. This study indicates that primary PM toxicity is driven by wider factors than fuel composition, highlighting that further work is needed to substantiate the full toxicity of aircraft exhaust PM inclusive of secondary PM emanating from numerous engine technologies across the power range burning conventional Jet-A and SAF
Effets de confinement du bruit d'Open Fan en soufflerie à veine fermée : méthodologie numérique et évaluation
No full textInternational audienceThis work investigates, through numerical simulations, the acoustic confinement effects of an open fan test bench installed in a wind-tunnel with closed test section. The present activity makes use of the full IBM methodology developed in previous work, which consists in computing with a CAA solver, the propagation of the acoustic waves generated by an Open Fan acoustic source model injected in the CAA domain, in the presence of complex geometries and a realistic non-uniform mean flow previously obtained with a CFD solver. The CAA solver uses the Immersed Boundary Method (IBM) to deal with the bench and wind-tunnel geometries, which greatly simplifies the grid generation process and also drastically reduces the overall computational time by maximizing the explicit time step while propagating acoustic wave over long distances. Two simulations are performed for an operating point corresponding to the take-off condition: a free-field and a confined configuration with rigid walls. Despite strong reflected waves from the wind-tunnel walls, the free-field directivity can still be directly retrieved for the main directions of radiation.Cette étude traite, à l'aide de simulations numériques, des effets de confinement acoustique d'un banc d'Open Fan installé dans une soufflerie à veine fermée. Cette étude tire partie de la méthodologie entièrement IBM développée lors de travaux précédents et qui consiste à calculer avec un code CAA la propagation des ondes acoustiques générées par un modèle acoustique d'Open Fan injecté dans le domaine CAA, en présence de géométries complexes et d'un écoulement porteur non uniforme réaliste obtenu par ailleurs. Le code CAA utilise la méthode aux frontière immergées (IBM pour Immersed Boundary Method) permettant ainsi de traiter les géométries du banc ainsi que de la soufflerie, simplifiant ainsi grandement l'étape de maillage et réduisant ainsi le temps global de calcul
Transmitarrays à polarisation circulaire hautement efficaces et à large bande opérant dans les fréquences millimétriques
No full textInternational audienceThis paper presents a novel unit cell for linear-to-circular polarization conversion (LP-CP) used for transmitarray antennas. We propose an excitation technique based on transmission line modeling, in which a centralized via excitation is divided into two striplines designed to balance the excitations and ensure phase quadrature. To broaden the bandwidth and transmission efficiency, a stacked patch configuration is used, enabling high polarization conversion and transmission efficiencies. Experimental results from a 70×70-cell, broadband, highly efficient Ka-band transmitarray show high gain of 39.8 dB and aperture efficiency of 55% at 29 GHz, while maintaining an axial ratio of less than 1 dB over a bandwidth in excess of 23%.Cet article présente une cellule unitaire novatrice pour la conversion de la polarisation linéaire à la polarisation circulaire (LP-CP) utilisée pour les antennes de type transmitarray. Nous proposons une technique d'excitation basée sur la modélisation des lignes de transmission, dans laquelle une excitation par via centralisée est divisée en deux striplines conçues pour équilibrer les excitations et assurer une quadrature de phase. Pour élargir la bande et l'efficacité de transmission, une configuration de patchs empilés est utilisée, permettant des efficacités de conversion de polarisation et de transmission élevées. Les résultats expérimentaux d'un transmitarray de 70×70 cellules, à large bande et hautement efficace dans la bande Ka démontrent un gain élevé de 39,8 dB et une efficacité d'ouverture de 55% à 29 GHz, tout en maintenant un rapport axial inférieur à 1 dB sur une largeur de bande dépassant 23%