Portail HAL ONERA
Not a member yet
12842 research outputs found
Sort by
Surrogate-Based Optimization of System Architectures Subject to Hidden Constraints
No full textInternational audienceSystem architecture optimization can support the design of novel architectures by formulating the architecting process as an optimization problem. The exploration of novel architectures requires physics-based simulation due to a lack of prior experience to start from, which introduces two specific challenges for optimization algorithms: evaluations become more expensive (in time) and evaluations might fail. The former challenge is addressed by Surrogate-Based Optimization (SBO) algorithms, in particular Bayesian Optimization (BO) using Gaussian Process (GP) models (Kriging). An overview is provided of how BO can deal with challenges specific to architecture optimization, such as design variable hierarchy and multiple objectives: specific measures include ensemble infills, a multi-step infill optimization, and hierarchical sampling and correction algorithms. Evaluations might fail due to non-convergence of underlying solvers or infeasible geometry in certain areas of the design space. Such failed evaluations, also known as hidden constraints, pose a particular challenge to SBO/BO, as the surrogate model cannot be trained on empty results. This work investigates various strategies for satisfying hidden constraints in BO algorithms. Three high-level strategies are identified: rejection of failed points from the training set, replacing failed points based on non-failed points, and predicting the failed region. Through investigations on a set of test problems including a jet engine architecture optimization problem, it is shown that best performance is achieved with a mixed-discrete GP to predict the Probability of Viability (PoV), and by ensuring selected infill points satisfy some minimum PoV threshold. The jet engine architecture problem features a 50% failure rate and could not previously be solved by a BO algorithm. The developed BO algorithm and used test problems are available in the open-source Python library SBArchOpt
Correction de front d'onde plan focal par apprentissage par renforcement sans modèle
No full textInternational audienceOptical aberrations prevent telescopes from reaching their theoretical diffraction limit. Once estimated, these aberrations can be compensated for using deformable mirrors in a closed loop. Focal plane wavefront sensing enables the estimation of the aberrations on the complete optical path, directly from the images taken by the scientific sensor. However, current focal plane wavefront sensing methods rely on physical models whose inaccuracies may limit the overall performance of the correction. The aim of this study is to develop a data-driven method using model-free reinforcement learning to automatically perform the estimation and correction of the aberrations, using only phase diversity images acquired around the focal plane as inputs. We formulate the correction problem within the framework of reinforcement learning and train an agent on simulated data. We show that the method is able to reliably learn an efficient control strategy for various realistic conditions. Our method also demonstrates robustness to a wide range of noise levels
Optimisation Bayésienne avec des contraintes cachées pour la conception avion
No full textInternational audienceA challenge in aircraft design optimization is the presence of non-computable, so-called hidden, constraints that do not return a value in certain regions of the design space. In this paper, we present a novel method to handle hidden constraints in aircraft conceptual design using Bayesian optimization. The method entails modifying a portion of the acquisition function of a Bayesian optimization formulation using supervised machine learning classifiers. The proposed approach reduces the effect of classifiers on exploration, therefore allowing the optimization algorithm to consider regions of the design space where previous information is not available. In addition, we consider different classifiers for handling hidden constraints. We demonstrate the proposed method using two simulation-based aircraft design optimization problems related to landing gear sizing and aircraft performance. The obtained results show an improvement of the objective function with fewer function evaluations
Improved 0D Sizing of High Speed Civil Transportation Aircrafts
No full textInternational audienceThis paper aims at presenting an improved version of a MDAO (Multi-Disciplinary Analysis and Optimization) tool developed by ONERA in order to support the designer during the initial conceptual design phase of high-speed civil transportation aircrafts. This tool helps to evaluate the long-term sustainability of operative concepts by providing a 0D sizing (maximum take-off mass, fuel mass, dry mass, plan surface, volume, lift and thrust requirements, ticket price per passenger…) corresponding to a specific mission in terms of range (3000 to 18000 km), number of passengers (10 to 300), fuel (biokerosene, LCH4, LH2). The cruise Mach number (3 to 8), supersonic dynamic pressure (20 to 100 kPa), Küchemann parameter are design variables that could be frozen or optimized for each mission. This improved version corrects some weaknesses of the previous version thanks to new models of aerodynamics, propulsion, masses, ascent and descent phases, and it introduces the supersonic dynamic pressure as an additional parameter to optimize. After presenting the sizing tool and the physical and implemented cost models, this paper provides the results and especially discusses the optimized designs obtained for different missions
Reliability based optimisation of composite plates under aeroelastic constraints via adapted surrogate modelling and genetic algorithms
No full textInternational audienceComposite materials are vital in aerospace for their exceptional strength-to-weight ratio. This study delves into reliability-based optimisation of composite plates within aeroelastic constraints, employing an efficient global optimisation method and genetic algorithms. Initial analysis focuses on aeroelastic responses such as limit flutter speed, gust response, and static loads, emphasising maximum strain assessment. To tackle optimisation challenges of composite stacking sequences, a homogenisation technique with lamination parameters is applied. We then formulate a constrained optimisation problem to minimise gust response while meeting flutter and maximum strain constraints. Surrogate models based on conditioned Gaussian Processes are developed for each aeroelastic response, facilitating optimisation within the composite design space. These models, with potential for local refinement, expedite optimal solution identification. Further, we integrate reliability-based optimisation into the framework to determine a robust stacking sequence using genetic algorithms, accounting for random fibre orientation variations. This holistic approach integrates aeroelastic analysis, constrained optimisation, surrogate modelling, and reliability-based optimisation, proving effective in designing reliable, efficient composite structures for aerospace, thus enhancing performance and safety
Gap engineering and wave function symmetry in C and BN armchair nanoribbons
No full textInternational audienceThere are many ways of engineering the band gap of nanoribbons, including application of stress, electric field, and functionalization of the edges. In this paper, we investigate separately the effects of these methods on armchair graphene and boron nitride nanoribbons. By means of density functional theory calculations, we show that, despite their similar structures, the two materials respond in opposite ways to these stimuli. By treating them as perturbations of a heteroatomic ladder model based on the tight-binding formalism, we connect the two behaviors to the different symmetries of the top valence and bottom conduction wave functions. These results indicate that opposite and complementary strategies are preferable to engineer the gap width of armchair graphene and boron nitride nanoribbons
Sense of agency in joint action: a critical review of we-agency
No full textInternational audienceThe sense of agency refers to the experience of control over voluntary actions and their effects. There is growing interest in the notion of we-agency, whereby individual sense of agency is supplanted by a collective agentic experience. The existence of this unique agentic state would have profound implications for human responsibility, and, as such, warrants further scrutiny. In this paper, we review the concept of we-agency and examine whether evidence supports it. We argue that this concept entails multiplying hypothetical agentic states associated with joint action, thus ending up with an entangled phenomenology that appears somewhat speculative when weighted against the available evidence. In light of this, we suggest that the concept of we-agency should be abandoned in favor of a more parsimonious framework for the sense of agency in joint action
A Deep Reinforcement Learning Approach to Wavefront Control for Exoplanet Imaging
No full textInternational audienceExoplanet imaging uses coronagraphs to block out the bright light from a star, allowing astronomers to observe the much fainter light from planets orbiting the star. However, these instruments are heavily impacted by small wavefront aberrations and require the minimization of starlight residuals directly in the focal plane. State-of-the art wavefront control methods suffer from errors in the underlying physical models, and often require several iterations to minimize the intensity in the dark hole, limiting performance and reducing effective observation time. This study aims at developing a data-driven method to create a dark hole in post-coronagraphic images. For this purpose, we leverage the model-free capabilities of reinforcement learning to train an agent to learn a control strategy directly from phase diversity images acquired around the focal plane.Initial findings demonstrate successful aberration correction in non-coronagraphic simulations and promising results for dark hole creation in post-coronagraphic scenarios. These results highlight the potential of model-free reinforcement learning for dark-hole creation, justifying further investigation and eventually experimental validation on a dedicated testbed
Self-supervised Rare Visual Token Enhancement
No full textThis paper focuses on rare visual token enhancement: given an image the model is trained to select few visual tokens with the property of being largely distant from all the others.This idea, which is structurally more frugal than classical self-supervised methods, could be an interesting pretext task for both fast classification based on few tokens, or, interesting point selection
Synchronisation d'horloge pour l'imagerie SAR bistatique aéroportée
No full textInternational audienceONERA experimented bistatic aircraft to aircraft SAR imaging with its business jet carried system Sethi as transmitter at X-band and its motor-glider carried system BuSARd as receiver with two channels for direct and terrain scattered paths. Redundant clock synchronisation techniques were implemented: GPS disciplining, direct channel digitising, active reference transponder, bistatic autofocus and cross-platform auto-registration. Experiments with varied configurations (linear or circular acquisitions, incident angle, bistatic angle) allow to draw synchronisation technique recommendations for both experimental and operational airborne bistatic SAR