Portail HAL ONERA
Not a member yet
12842 research outputs found
Sort by
Improved Hybrid Numerical Methodology for Fast Design of Reconfigurable Transmit-Arrays Antenna
No full textInternational audienceThis paper presents an improved hybrid semianalytical methodology for the fast and accurate simulation of reconfigurable Transmit-Arrays Antennas (TAA), addressing key challenges in high-frequency telecommunication standards such as 5G mmWave and Beyond 5G. Building upon our previous work, we refine the characterization of unit-cell radiation patterns by incorporating an embedded active pattern extraction approach. This enables more precise predictions of beam steering performance, cross-polarization discrimination (XPD), and sidelobe levels (SLL). A 196-cell (14×14) 1-bit reconfigurable TAA operating at 10.25 GHz is designed and experimentally validated. The results demonstrate a strong agreement between measured, full-wave simulated, and hybrid simulated radiation patterns, with the improved method successfully capturing previously unmodeled asymmetries in scanning behavior. This work significantly enhances the predictive accuracy of hybrid simulations, making it a powerful tool for the rapid analysis and design of next-generation reconfigurable metasurfaces
Calcul et analyse de la SER de cibles maritimes en bande HF
No full textNational audienceLe travail présenté ici concerne l’étude de la SER (Surface Équivalente Radar) de cibles maritimes en bande HF. Nous proposons une méthode de détermination d’un modèle équivalent d’une cible maritime permettant de prédéterminer le champ électromagnétique rétrodiffusé dans une scène d’intérêt. L’exploitation de ce modèle donne accès à l’analyse des couplages entre ondes de ciel et ondes de surface lors de leurs interactions avec la cible et avec le milieu (la mer)
AIRIS² : un algorithme de diversité intelligente de gateway pour des systèmes de communication par satellite à très haut débit
No full textInternational audienceSatellite communication systems are shifting to higher frequency bands (Ka, Q/V, W) to support more dataintensive services and alleviate spectral congestion. However, the use of Extremely High Frequencies, typically above 20 GHz, causes significant tropospheric impairments, such as rain attenuation, which can causes system outages. To mitigate these effects, Smart Gateway Diversity (SGD) has emerged as a promising method for maximizing feeder link availability through an adaptive site diversity scheme. However, implementing such technique requires a decision-making policy to dynamically select the optimal set of gateways and prevent outages. This paper introduces AIRIS 2 , a deep learning algorithm that anticipates short-term rain events from rain attenuation measurement to enable efficient gateway switching. The approach is validated from five years of measured time series collected at Ka and Q/V bands at various sites and climatic conditions
Shielded cable unified MTL model applications
No full textInternational audienceThis paper presents a unified transmission-line model for multiple shield and multiconductor cables. The model is based on a single reference to which all wire voltages are taken. Tt can be applied whatever the connection configurations at both shield extremities from ideal 360° shield connections to simple bonding wires. Two applications of the model are proposed and compared to measurements: the first one addresses the case of a single shielded cable link with various electrical bonding solutions. The second application deals with an overshielded cable
VerifIoU - Robustness of Object Detection to Perturbations
No full textInternational audienceWe introduce a novel Interval Bound Propagation (IBP) approach for the formal verification of object detection models, specifically targeting the Intersection over Union (IoU) metric. The approach has been implemented in an open source code, named IBP IoU, compatible with popular abstract interpretation based verification tools. The resulting verifier is evaluated on landing approach runway detection and handwritten digit recognition case studies. Comparisons against a baseline (Vanilla IBP IoU) highlight the superior performance of IBP IoU in ensuring accuracy and stability, contributing to more secure and robust machine learning applications
Laminar–turbulent transition experiment on the effect of surface imperfections on a natural laminar flow profile in compressible flow conditions
No full textInternational audienceIn the present study, the effect of various two-dimensional surface defects (forward-facing steps and ramps, backward-facing steps and ramps, gaps, and steps and gaps) on boundary layer transition was experimentally investigated in the compressible, subsonic regime. A laminar profile was specifically designed and manufactured by ONERA to allow for a maximum number of defects to be tested simultaneously, and to include resin pockets to accurately monitor laminar–turbulent transition using infrared thermography. Transition was also characterized using the ΔN model based on linear stability calculations. Relatively good agreement with existing ΔN models for forward-facing steps as well as gaps was found, indicating that these models, which were mostly developed for incompressible flows, can still be used as an initial estimate for compressible flows. One particular case of interest included a critical step and gap (for which transition occurred immediately downstream of the defect) where neither the gap nor the step component could be identified as mainly responsible for triggering transition. Steps and gaps should therefore be included whenever possible to the canonical shapes of defects investigated in transition experiments to further refine the different types of defect encountered in industrial application, and provide appropriate criteria for their allowable tolerances
Modèle de programmation par contraintes pour l'équilibrage et l'ordonnancement des lignes d'assemblage avec travailleurs mobiles et stations parallèles
No full textInternational audienceIn the context of aircraft assembly lines, increasing the production rate and decreasing the operating costs are two important, and sometimes contradictory, objectives. In small assembly lines, sharing production resources across workstations is a simple and efficient way to reduce operating costs. Therefore, workers are not assigned to a unique workstation but can walk between them. On the other side, paralleling workstations is an efficient way to increase the production rate. However, the combination of both strategies create complex conditions for tasks to access the production resources. This paper addresses the problem of allocating tasks to workstations and scheduling them in an assembly line where workers can walk across workstations, and where workstations can be organized in parallel. We model this problem with Constraint Programming. We evaluate it on real world industrial use cases coming from aircraft manufacturers, as well as synthetic use cases adapted from the literature
Aircraft Resource-Constrained Assembly Line Balancing with Learning Effect : a Constraint Programming Approach
No full textInternational audienceBalancing aeronautical assembly lines is a major challenge in modern aerospace manufacturing. Aircraft manufacturing plants typically have a predetermined production rate, but the production system requires a period of adaptation at start-up. This phenomenon, known as the learning effect, refers to the gradual improvement in efficiency through task repetition, thereby reducing task duration. However, the stability of an assembly line is also a critical factor, as any change in the production process incurs costs. In this study, Constraint Programming (CP) is used to optimise assembly line balancing, taking into account the learning effect to address the trade-off between achieving target production rates and minimising adjustments to the line
Adaptive Optics Rolling Slit Ophthalmoscope: High Resolution Camera-based Multimodal Retinal Imaging, Blood Flow Visualization and Neurovascular Coupling
No full textInternational audienc
Estimating Sea-ice drift using deep-learning optical flow algorithm
No full textInternational audienceSea ice has a substantial climatic impact, affecting heat exchanges between the ocean and atmosphere, nutrient distribution, and marine ecosystem health. Significant changes in sea ice behavior have been noted in recent years. Research shows that the ice’s thickness and extent have diminished, with multi-year ice significantly declining and freshly created ice increasing [1]. This development has increased the overall fragility of the ice, making it more sensitive to extreme fluctuations, premature break-up, and rapid drifting of the pack ice [2]. The increased variability of sea ice conditions poses problems for human activities in the polar regions. It also hinders efforts to preserve polar ecosystems and complicates accurate modeling of ocean and climate systems. In light of these circumstances, accurately measuring Sea Ice Drift (SID) is a cornerstone for comprehending the new dynamics of sea ice. Precise SID measurements will enable us to identify trends, model the intricate interactions between ice, oceans, and the atmosphere, and inform about the environmental and economic implications, particularly regarding navigation at sea and the exploitation of resources. Synthetic Aperture Radar (SAR) represents a promising solution for SID measurement since it operates independently of weather and illumination conditions, offering high spatial and temporal resolution. Consequently, SAR images are particularly well-suited to track ice sea movements on a fine scale with high temporal frequency. Indeed, emerging solutions derived from Sentinel-1 or RADARSAT-2 SAR images have been proposed, such as DTU-SID for the Arctic (10 km resolution) [3] or UTAS-SID for the Antarctic (kilometric resolution) [4]. While these products are useful for large-scale assessments, their coarser resolutions of existing products limit the ability to accurately represent the intricate interactions within ice structures, such as fractures, compression zones, or local movements at the floe scale, often observable at a scale of hundreds of meters. Furthermore, these products use conventional maximum cross-correlation techniques, which might not be appropriate for identifying low-scale and moving discontinuities. Therefore, higher-resolution SID products are required to give precise information about small-scale movements. In this context, optical flow methods stand out as a promising alternative to MCC methods. Unlike MCC, optical flow methods can estimate dense displacement fields at a sub-pixel level [5]. However, conventional optical flow methods often rely on naive assumptions, such as intensity conservation, which can limit their accuracy in noisy or sparsely textured environments, typical of SAR images. Deep learning-based optical flow techniques overcome this restriction by making it possible to learn intricate representations, which greatly improves performance. These cutting-edge techniques have shown to be quite successful in motion estimation for natural picture processing [6]. However, they have yet to be applied to SID estimation using SAR images, mainly due to the Lack of a dataset with the necessary resolution. This work aims to propose a novel optical flow deep learning-based method for estimating SID that will handle two main challenges: 1. Lack of labels and data scarcity at the desired resolution for model training: Labeled data for sea ice drift (SID) estimation at a fine resolution of 100 meters presents a significant challenge, primarily because deep learning approaches currently require labels for SID at this level of resolution that are not available. We can develop an artificially annotated database by creating pairs of synthetic SAR images with diverse displacement scenarios. The dataset will include a variety of sea ice drift patterns and different SAR acquisition scenarios, carefully balancing the complexity and representativeness of the simulations with the volume of data needed for effective model training. However, simulated data may still be insufficient for training complex deep learning networks that generalize well to real-world situations. Adapted training strategies, such as semi-supervised learning or fine-tuning methods, will be employed to leverage unlabeled data. These strategies enable the pre-training of the network on synthetic data, followed by gradual adaptation to real data to minimize discrepancies between the distributions. 2. Multimodality for high temporal resolution (≤6h): To achieve SID measurement at a temporal resolution of less than 6 hours, exploiting multimodal data by combining SAR images (Sentinel-1) with optical images (Sentinel-2) is necessary. However, this approach introduces a first challenge linked to the heterogeneity of the modalities since the characteristics of the two types of images differ significantly (spatial and spectral resolution and type of signal). A multimodal network will be developed to meet this challenge, integrating specific encoders for each modality to extract relevant characteristics while harmonizing the representations in a shared space that enables the integration of radar and optical information. The second challenge is the quantity of data available for this task. The dependence of optical images on the lighting and weather conditions significantly limits the availability of SAR/optical image pairs. We propose a database enrichment model that generates SAR/optical pairs through a sensor change simulation.This approach enables the conversion of a SAR image into an equivalent optical image, mimicking what could have been captured by optical sensors or inversely, using optical or neural methods, allowing the creation of additional data to train the SID measurement model accurately. This work will be evaluated by benchmarking our SID estimation framework against existing products to validate the proposed methodology. This evaluation will consider the accuracy of SID retrieval, the reliability of uncertainty estimates, and the computational efficiency of the pipeline. References [1] R Kwok. Arctic sea ice thickness, volume, and multiyear ice coverage: losses and coupled variability (1958–2018). 13(10):105005. [2] Jack C. Landy, Geoffrey J. Dawson, Michel Tsamados, Mitchell Bushuk, Julienne C. Stroeve, Stephen E. L. Howell, Thomas Krumpen, David G. Babb, Alexander S. Komarov, Harry D. B. S. Heorton, H. Jakob Belter, and Yevgeny Aksenov. A year-round satellite sea-ice thickness record from CryoSat-2. 609(7927):517–522. [3] European Union-Copernicus Marine Service. Global ocean - high resolution SAR sea ice drift. [4] HEIL, PETRA and HYLAND, GLENN. Satellite-derived high-resolution sea-ice motion in the southern ocean, 2015-2019. [5] Zisis I. Petrou, Yang Xian, and YingLi Tian. Towards breaking the spatial resolution barriers: An optical flow and super-resolution approach for sea ice motion estimation. 138:164–175. [6] Deqing Sun, Xiaodong Yang, Ming-Yu Liu, and Jan Kautz. PWC-net: CNNs for optical flow using pyramid, warping, and cost volume. Version Number: 3