Search

We characterized the radiation response in the visible domain of a new multimode graded-index (GI) phosphosilicate optical fiber (GIMMF), exposed to the harsh environment (pulses of 14-MeV neutrons, X-rays, and γ -rays) associated with laser experiments at the OMEGA facility. The growth of permanent radiation-induced attenuation (RIA) was measured in situ after a series of laser shots involving a large production of 14-MeV neutrons (yields > 10^14 n per shot). RIA linearly increases with accumulated neutron fluence without recovery between shots. The obtained results allow a precise evaluation of this GIMMF vulnerability when implemented as part of laser or plasma diagnostics. Our work also reveals the potential of this class of optical fiber to serve as a radiation monitor in the radiation-rich mixed environments of megajoule class laser facilities and to provide a very fast and online estimation of the accumulated deposited dose at various locations of their experimental halls. In our experimental test configuration at OMEGA, 14-MeV neutrons are estimated to contribute to about 55% of the total deposited dose on the fibers, and the other optical losses are related to X-ray and γ -ray contributions. Those measurements could be, for example, benchmarked to the radiation maps obtained by Monte Carlo simulation tools, potentially facilitating the evaluation of the aging of diagnostics, components, and systems as well as their maintenance operations.

International audience; Identification of vibration damping in 3D-printed lattice structures

International audience; Estimation of the piezoelectric factor in nonlinear transducers

International audience; We propose an innovative approach to minimize the greenhouse gas impacts of additive manufactured structures over their entire life cycle. The novelty of our method lies in its simultaneous optimization of material selection, process selection, and design optimization. To fully leverage the potential benefits of additive manufacturing, we use topology optimization and compile a comprehensive database of printed materials and printing processes, which we share with the wider community. To account for the complex interdependence between materials and processes, our method employs a pairing system, which we efficiently reduce using topology optimization properties and a generalized form of Ashby indices. To enhance computational efficiency, we employ a meta-model. We validate our proposed method through successful testing on an aeronautical case and a pedestrian bridge, demonstrating its robustness even in the presence of environmental data uncertainty. The optimal material-process pair for the aeronautical structure is the cobalt-based super-alloy with the LENS process. Despite this pair having the highest material and processing emissions, the resulting lighter part lowers the use phase emissions. It appears that precise mechanical data is needed for the method to give accurate results: a 20% drop of Young's modulus totally disrupts the material-process pair ranking.

International audience; Aeronautical composite primary structures must evidence sufficient residual strength in the presence of damage for compliance with damage tolerance requirements. The study of stiffener debonding on panels subjected to compression after impact is performed in that scope. Compression leads to the buckling of the skin between the stiffeners, and thus a complex loading of the bonding between the skin and the stiffener. This paper describes the development of a stiffened specimen for the VERTEX multiaxial test rig as a first step towards the study of the damage tolerance evaluation of stiffened structures, under combined loadings and at the intermediate scale of the test pyramid. By using virtual testing, the specimen was designed to produce the phenomenology of interest as the first damage, i.e., the debonding of the stiffener from the centre. Three samples were manufactured and subjected to low velocity impacts at various locations and energies. Then the three samples were subjected to compression after impact, up to the stiffener debonding, under a post-buckling regime of the skin. Test loading evolution is described with force fluxes and global strains, obtained from in situ stereo-correlation. The different impacts were found to give different types of damage but similar residual strength to compression after impact.

International audience; Obsolescence is the fact that an entity (physical or logical) is becoming outdated or no longer possesses the required level of performance. The objectives of this article are twofold. First, it is intended to contribute to the understanding of obsolescence propagation. Secondly, two supporting approaches for the Identification and Assessment phases are proposed: the House of Obsolescence and the System Obsolescence Criticality Analysis. The former allows the mapping of obsolescence propagation via dependencies, whether imposed changes are desired or imposed, by external actors to the system architecture. Whereas, the objective of the latter is to assign an obsolescence criticality index to the identified risks in order to prioritize them for solution or mitigation determination during the analysis phase. The tools make extensive use of the modeled system knowledge through the application of Systems Engineering. The application of these approaches is presented through an illustrative study.

International audience; Nowadays, drones can be developed for a wide range of use cases, from infrastructure monitoring to sea rescue, urban mobility or military purposes. Which drone design is best suited for a specific mission? To answer this question, we need to solve a constrained optimization problem based on a multidisciplinary design model that takes the mission into account. The model generally being a computationally expensive numerical model whose gradients are not available all the time encourages us to consider a Bayesian optimization approach. Such strategy is well known to achieve a trade-off between exploitation and exploration in order to find interesting minimal area with a reduced number of function evaluations. A multi-fidelity approach can improve even more the computational efficiency of the Bayesian optimization strategy. In this work, we aim at designing a fixed-wing drone (fully electric) for long range surveillance mission. Two fidelity level electric drone models are developed. For a given mission requirement, the final battery state of charge is optimized with respect to drone design variables. Optimizations are performed on several missions using both a mono and a multi-fidelity Bayesian optimization strategy. The interest of using a multi-fidelity method for overall drone design has been assessed. The multi-fidelity super-efficient global optimization algorithm (MFSEGO) appeared to need less budget to reach convergence than the mono-fidelity algorithm and to be more robust to the initial design of experiments.

International audience; Graph Neural Networks have been applied to learn the flight and structural dynamics of an High Altitude Long Endurance aircraft in discrete gust. The graph network methodology allows building a model for structural displacements, loads and aircraft flight dynamics leveraging on the inductive bias given by the physical connections. The results show promising capabilities in model approximation and potential for symbolic identification of aerodynamics and structural forces.

International audience; It is sometimes desirable to delve further into how the inputs affect the output in design optimization and uncertainty analysis. Surrogate models such as Gaussian Process Regression and support vector regression are useful for such tasks and can be further enhanced by introducing advanced post-processing methods. This paper investigates Shapley Additive Explanation (SHAP) as a tool to aid surrogate-assisted data-driven analysis. In particular, surrogate-enabled SHAP analysis allows visualization of the input-output relationship in a meaningful way using summary SHAP plots and SHAP dependence plots. Some importantinformation that can be extracted and visualized from SHAP includes the importance of input variables (i.e., global sensitivity analysis), nonlinearity level, and level of interactions. The benefits of Shapley values for engineering analysis using surrogate models are demonstrated in three engineering test problems.

International audience; Many engineering problems are governed by complex governing equations that are difficult and typically require high computational costs to solve. Machine learning and surrogate modelling aid such an endeavour by providing a cheap-to-evaluate prediction model that acts as a replacement of the original model. While most research focuses on predicting scalar values (e.g., lift and drag), predicting the solution field is also of interest in many practical engineering and scientific applications. This paper proposes a Physics-Informed Proper Orthogonal Decomposition (POD) technique that improves the solution field prediction by enforcing governing equations as a loss penalty. The proposed idea utilizes a reduced-order modeling technique based on POD to decompose solution snapshots into singular vectors and values. A Gaussian Process Regression is then utilized to predict the singular values from variable parameters. The predicted singular values from the data of the problem are then adjusted via optimization to minimize the physics-informed loss and achieve better prediction. In this paper, we illustrate the efficacy of the proposed method on simple two-dimensional partial differential equations. The result clearly shows that the proposed physics-informed POD outperforms the conventional POD in terms of approximation error.

International audience; A low-computational-cost method is proposed in this paper to predict steady-state nonlinear aeroelastic response for high-aspect-ratio wings. This fast nonlinear static aeroelasticity method relies on a simple but robust mathematical approach. The work presented in this paper evaluates the accuracy of a nonlinear aerodynamic method on a large range of Mach numbers for a geometry representative of a typical industrial high-aspect-ratio-wing aircraft. The robustness of the method relies on the accurate estimation of the local pressure field based on an aerodynamic database using a local incidence estimated with a vortex lattice method. The database can be filled with flight test, wind tunnel test, or high-fidelity numerical simulation. The structural deformation is provided by a coupled aeroelastic high-fidelity numerical simulation, and so this paper focuses mainly on the development and validation of the aerodynamic model. Then, the flexible coefficients of the wing are compared to the high-fidelity aeroelastic numerical simulations for a set of Mach numbers ranging from subsonic to transonic conditions. The results presented in this paper show that the present method is very accurate for low-Mach-number regimes (error is lower than 1% on lift), and it is also adapted to transonic flow regimes because the error on lift is lower than 5%. For high Mach numbers, the current solution commits larger errors on drag and pitching moment coefficients.

International audience; With the increasing availability of data for Prognostics and Health Management (PHM), Deep Learning (DL) techniques are now the subject of considerable attention for this application, often achieving more accurate Remaining Useful Life (RUL) predictions. However, one of the major challenges for DL techniques resides in the difficulty of obtaininglarge amounts of labelled data on industrial systems. To overcome this lack of labelled data, an emerging learning technique is considered in our work: Self-Supervised Learning, a sub-category of unsupervised learning approaches. This paper aims to investigate whether pre-training DL models in a self-supervised way on unlabelled sensors data can be useful for RUL estimation with only Few-Shots Learning, i.e. with scarce labelled data. In this research, a fatigue damage prognostics problem is addressed, through the estimation of the RUL of aluminum alloy panels (typical of aerospace structures) subject to fatigue cracks from strain gauge data. Synthetic datasets composed of strain data are used allowing to extensively investigate the influence of the dataset size on the predictive performance. Results show that the self-supervised pre-trained models are able to significantly outperform the non-pre-trained models in downstream RUL prediction task, and with less computational expense, showing promising results in prognostic tasks when only limited labelled data is available.

International audience; Bell's inequality violation experiments are becoming increasingly popular in the practical teaching of undergraduate and master's degree students. Bell's parameter S is obtained from 16 polarization correlation measurements performed on entangled photons pairs. We first report here a detailed analysis of the uncertainty u(S) of Bell's parameter taking into account coincidence count statistics and errors in polarizers' orientation. We show using both computational modeling and experimental measurement that the actual sequence of the polarizer settings has an unexpected and strong influence on the error budget. This result may also be relevant to measurements in other settings in which errors in parameters may have non-random effects in the measurement.

International audience; The development of the aeronautical industry has brought about certain demands like the reduction of fuel consumption and, consequently, the weight of used structures. Therefore, topology analysis of aircraft components has played an important role in meeting these demands. Another factor widely studied in this context is the type of materials used in the manufacture of such components, hence, by presenting lower densities, composites have gained ground in this industry sector.Besides, an increasingly sought-after characteristic is the sustainability associated with the entire life cycle of any system or part, from its design, through the production phase, to its end of life. The assessment of such sustainability is often carried out by analysing the life cycle’s CO2 footprint, which tends to decrease when the raw materials used are less processed. Materials of natural origin, also known as "bio-materials", have therefore been the target of study, being the basis of the concept of eco-design. The accelerated development of 3D printing technologies has allowed the evolution of structural design methods. Additive manufacturing emerged in the 1980s and has since been expanding in the aerospace field. One of its great advantages is that it can be combined with the use of bio-composites, such as flax or hemp.The topology optimisation of any given structure brings great benefits in terms of the mitigation of the CO2 impact of a given part, as it translates into a reduction of its mass and fuel consumption. However, the combination of this method with the optimisation of the fibre path, in the 3D printing process, and the CO2 footprint of the part’s life cycle simultaneously can bring even more evident benefits. In this paper modifications to existing methods are proposed. In MATLAB, strategies are developed to study the topology and fibre orientation of potentially 3D printed thin structures.

International audience; The paper presents the results of dynamic crushing of sandwich tubes that had skins made of carbon or glass fibres-with epoxy resin-and an I214 poplar ply core. By increasing the number of poplar plies from two to six, the absorbed energy is doubled, showing the significant contribution of the wood. The Specific Energy Absorption of sandwiches with carbon fibre skins oscillated between 49.4 and 60 J/g while that with glass fibre skins varied from 35.4 to 43.3 J/g.

International audience; Global Navigation Satellite System (GNSS) is the widely used technology when it comes to outdoor positioning. But it has severe limitations with regard to safety-critical applications involving unmanned autonomous systems. Namely, the positioning performance degrades in harsh propagation environment such as urban canyons. In this letter we propose a new algorithm for GNSS navigation in challenging environments based on robust statistics. M-estimators showed promising results in this context, but are limited by some fixed hyper-parameters. Our main idea is to adapt this parameter, for the Huber cost function, to the current environment in a data-driven manner. Doing so, we also present a simple yet efficient way of learning with satellite data, whose number may vary over time. Focusing the learning problem on a single parameter enables to efficiently learn with a lightweight neural network. The generalization capability and the positioning performance of the proposed method are evaluated in multiple contexts scenarios (open-sky, trees, urban and urban canyon), with two distinct GNSS receivers, and in an airplane ground inspection scenario. The maximum positioning error is reduced by up to 68% with respect to M-estimators.

International audience; Predictive maintenance techniques attempt to trigger a maintenance intervention at the right moment by estimating the life expectation. Predictive maintenance is increasingly implemented by automated approaches able to perform diagnostics and prognostics. The main part of recent research in these approaches is focused in machinelearning structures whose reasoning is implicit and cannot be easily explained. This poses a problem for theirimplementation in highly constrained area such as aeronautics. To overcome this constraint, explicit reasoningapproaches such as the Similarity-Based Model (SBM) can be implemented. The SBM has been widely used for fault diagnostics and the remaining useful life (RUL) estimation, but the development of SBM includes tasks that often rely on high skilled experts. For instance, data reduction techniques required for SBM are often performed by experts judgment whose outcomes are not always consistent. The produced features from these techniques are used to build the Health Index that can be used to create the degradation trends that serve as a reference for the SBM. To overcome these difficulties, an automatic and unsupervised approach based on the Kernel Principal Component Analysis is proposed to enhance the Health Index creation. It preserves as much of the sensor information as possible improving the similarity-based RUL estimation. Additionally, when estimating the RUL of a system, the most similar degradation trends stored in the SBM library are used to compute individual RULs, the final RUL is obtained by a fusion rule technique that combines all these individual RULs into a consolidated value. For the fusion rule techniques, a self-adaptive method that does not rely on human expertize is proposed. This fusion rule can benefit of the accumulated knowledge over the SBM operation. This unsupervised approach to develop a SBM is validated with promising results against an equivalent and supervised algorithm that came out best in the 2008 prognostic health management challenge.

International audience; Mechanical and electro-mechanical de-icing systems are low-energy ice protection solutions based on fracturemechanisms. It can, however, be difficult to obtain the protection of an entire surface due to the limited propagation of fractures for some mechanisms. This article shows how it is possible to reshape the substrate in orderto favor the propagation of adhesive fracture at the ice/substrate interface. The first part of the paper introducesan analytical beam theory approach for running computations quickly, making it possible to achieve parametricoptimization of the substrate thickness and maximize the propagation length. The optimization results werevalidated using FEM software and tests on an aluminum prototype. A second method is also studied in this paper,topology optimization is used on a 2D finite element model to minimize the substrate mass of the proposedsolution and adhesive crack propagation is assessed in comparison with the mass impact. For different boundaryconditions, propagation ranges can be increased by up to 150% with a mass increase limited to 50%. Usingtopology optimization, the additional mass could be reduced by 60% while maintaining the sameperformances.

International audience; In recent years, unmanned aerial vehicles applications have grown in various fields. New design configurations have emerged to address these needs. This paper focuses on the multidisciplinary design optimization of three popular concepts of drones under a common framework: multicopters, fixed-wings, and the more recent fixed-wing VTOLs. Low-fidelity but computationally efficient models for drone sizing are proposed, some of them being shared by the different concepts. The models are integrated with a design optimization approach, and an efficient sizing methodology allows for reducing the problem complexity. This approach is instrumental because of the large number of design variables and constraints involved in the multidisciplinary design process. Examples of sizing results are provided to validate the methodology for fixed wing UAVs. In addition, a study case demonstrates the benefit of a common framework for comparing the performances of different drone concepts. A key finding is that the VTOL propulsion of the fixed-wing VTOLs must be carefully optimized to avoid an undesirable mass penalty.

International audience; Piezoelectric materials are employed in several different aerospace applications thanks to their actuating and sensing abilities. They can be adopted to control vibration, stress distribution, aeroelastic stability, shape changes, and in structural health monitoring systems. Given the level of reliability required in aeronautics, it is of vital importance to ensure their correct functioning during all aircraft life and for critical events such as impacts. This paper proposes a novel test standard method to measure and qualify the damage resistance of smart PZT structures to a drop-weight impact event. Procedures for the quantification of damage through measurements of electromechanical characteristics are introduced and verified by impact tests.

International audience; Increasing the aspect ratio of transonic aircraft wings represent a promising path to allow aviation to reduce its carbon footprint. The objective of this study is to develop sufficiently accurate aero-structural models in order to fully exploit High Aspect Ratio (HAR) wings drag reduction and performance improvements potential at preliminary design stage. The use of a multifidelity approach makes it possible to reduce computational costs by mainly resorting to low-fidelity computations and only running high-fidelity computations when necessary. The first step is to develop high and low fidelity models to perform static aeroelastic analysis on the NASA CRM wing. These models will be used as a baseline to be compared with other configurations of interest. Afterwards, models are developed for a modified version of the CRM wing with a higher aspect ratio, and static aeroelastic analysis are carried out. Finally, a Strut-Braced Wing (SBW) configuration is studied to evaluate its mass reduction potential, in addition to the drag reduction provided by the increased aspect ratio. These aero-structural models are implemented within an in-house aeroelastic analysis and optimization framework. This works focuses on aeroelastic analysis for now, but is to be extended to aeroelastic optimization, using a multifidelity approach.

International audience; Tasks such as analysis, design optimization, and uncertainty quantification can be computationally expensive. Surrogate modeling is often the tool of choice for reducing the burden associated with such data-intensive tasks. However, even after years of intensive research, surrogate modeling still involves a struggle to achieve maximum accuracy within limited resources. This work summarizes various advanced, yet often straightforward, statistical tools that help. We focus on four techniques with increasing popularity in the surrogate modeling community: (i) variable screening and dimensionality reduction in both the input and the output spaces, (ii) data sampling techniques or design of experiments, (iii) simultaneous use of multiple surrogates, and (iv) sequential sampling. We close the paper with some suggestions for future research.

A bipolar charge transport model with two levels of traps (shallow and deep traps) is employed to describe and investigate the dynamic behavior of the space charge in low-density polyethylene (LDPE). This kind of model requires some initial experimental conditions alongside a set of parameters. An optimization method is used in order to provide the best precision for model parameters. Furthermore, the sensitivity of these parameters has been analyzed using an approach based on Sobol's method. In essence, this paper proposes a specific protocol that contributes to producing a global set of parameters able to provide the best approximate fit between experimental and simulated results. Net charge density is measured by the pulse electro-acoustic method (PEA) along with what is known as external charging current measurements. The simulated data are obtained by a bipolar charge transport model, which is developed for low-sensity polyethylene.

In this paper, mixed categorical structural optimization problems are investigated. The aim is to minimize the weight of a truss structure with respect to cross-section areas, materials and cross-section type. The proposed methodology consists of using a bi-level decomposition involving two problems: master and slave. The master problem is formulated as a mixed integer linear problem where the linear constraints are incrementally augmented using outer approximations of the slave problem solution. The slave problem addresses the continuous variables of the optimization problem. The proposed methodology is tested on three different structural optimization test cases with increasing complexity. The comparison to state-of-the-art algorithms emphasizes the efficiency of the proposed methodology in terms of the optimum quality, computation cost, as well as its scalability with respect to the problem dimension. A challenging 120-bar dome truss optimization problem with 90 categorical choices per bar is also tested. The obtained results showed that our method is able to solve efficiently large scale mixed categorical structural optimization problems., Accepted for publication in Structural and Multidisciplinary Optimization, to appear 2022

International audience; The ground movements induced by seismic waves create acoustic waves propagating upward in the atmosphere, thus providing a practical solution to perform remote sensing of planetary interiors. However, a terrestrial demonstration of a seismic network based on balloon-carried pressure sensors has not been provided. Here we present the first detection of seismic infrasound from a large magnitude quake on a balloon network. We demonstrate that quake's properties and planet's internal structure can be probed from balloon-borne pressure records alone because these are generated by the ground movements at the planet surface below the balloon. Various seismic waves are identified, thus allowing us to infer the quake magnitude and location, as well as planetary internal structure. The mechanical resonances of balloon system are also observed. This study demonstrates the interest of planetary geophysical mission concepts based on seismic remote sensing with balloon platforms, and their interest to complement terrestrial seismic networks.

International audience; Modal analyses are essential steps in structural design. Simple and quick numerical procedures and design tools for modal and dynamic analysis of bonded joints would be an attractive option to increase the efficiency of the design process in its early phases. This paper offers a simple and quick numerical tool dedicated to the modal analysis of single lap bonded joints with balanced and unbalanced adherends with various boundary conditions. Based on the macro-element technique developed with Taylor expansion in power series, this paper gives all the steps to develop the mass matrix associated with a macro-element. The mass matrix has been developed so far for 1D-bar and 1D-beam kinematic frameworks. The results obtained with the macro-element technique and a Finite Element analysis give similar results.

The present work was aimed at investigating the failure of quasi-isotropic fibers reinforced thermoplastic laminates solicited at different testing temperatures. Single-edge-notch bending (SENB) tests were conducted at room temperature (RT) and at a temperature higher than the glass transition temperature (Tg) to investigate the influence of failure mode (depending on initial notch orientation) as well as matrix ductility and toughness (depending on testing temperature) on: (1) damage mechanisms – (2) critical translaminar fracture toughness Gc and (3) G-R curves evolution. The mixed mode fracture toughness GI+II,c is dramatically lower than the mode I fracture toughness GIc, with −52% and −67% decreases, at RT and 150 °C respectively. In 45° notched specimens, most of the fracture energy (about 80%) stems from the mode I failure. Though they are not prominent on fracture, the failure mechanisms associated with mode II are instrumental in limiting the contribution of mode I on fracture behavior. Finally, though a temperature increase has very little influence on GIc, it significantly reduces the value of GI+II,c (−32%). This change primarily results from the formation of plastic-kink bands in compression that are promoted by both the ductility of the polyether ether ketone (PEEK) matrix at T > Tg and the mixed-mode failure.

International audience; An experimental study has been carried out to describe the evolution of the failure load relative to the maximum misalignment angle of curved carbon–epoxy laminates with out-of-plane wrinkle defects for a large range of misalignments. L-angle specimens with different levels of fibre wrinkling were manufactured and tested in four-point bending tests. Digital Image Correlation (DIC) and infrared thermography were used to monitor the tests. They allowed the stress concentrations induced by the wrinkles to be identified, together with their influences on the failure load of the sample. In most cases, increases in wrinkling levels led to decreases in failure loads.

International audience; Bayesian optimization is an advanced tool to perform efficient global optimization. It consists on enriching iteratively surrogate Kriging models of the objective and the constraints (both supposed to be computationally expensive) of the targeted optimization problem. Nowadays, efficient extensions of Bayesian optimization to solve expensive multi-objective problems are of high interest. The proposed method, in this paper, extends the super efficient global optimization with mixture of experts (SEGOMOE) to solve constrained multi-objective problems. To cope with the ill-posedness of the multi-objective infill criteria, different enrichment procedures using regularization techniques are proposed. The merit of the proposed approaches are shown on known multi-objective benchmark problems with and without constraints. The proposed methods are then used to solve a bi-objective application related to conceptual aircraft design with five unknown design variables and three nonlinear inequality constraints. The preliminary results show a reduction of the total cost in terms of function evaluations by a factor of 20 compared to the evolutionary algorithm NSGA-II.

In recent decades, designers in aeronautics have been specifically interested in the interaction between the various experts involved in the modeling of the performance of aircraft. These experts use so-called disciplinary numerical solvers to model their discipline (for example aerodynamic, structure or propulsion) and exchange information to take into account the influence of other disciplines. The system coupled by this information is called multidisciplinary analysis (MDA). The search of the best design, called multidisciplinary optimization, consists in identifying the design variables optimizing a performance function derived fromthe MDA solution. Solving this optimization problem raises several challenges. First, the evaluation of each disciplinary solver is numerically expensive andminimizing the calls to these solvers through the optimization represents one of the main objectives of the thesis. Secondly, it is assumed that the data exchanged by the disciplinary solvers is of high dimensionality (typically discretized fieldwhose dimension exceeds 1000). Finally, the resolution method must be as generic as possible : non-intrusive in order to use industrial disciplinary solvers and adaptable to MDA involving various number of disciplines. Several formulations have been proposed for the resolution of this type of problem, but none of them can fully overcome these challenges. One of the most promising ideas is to replace each disciplinary solver by Gaussian processes (GP) and couple these surrogate models to approximate the MDA. The idea is then to use the variance of the GP to estimate the uncertainties on the MDA and to enrich the GP if necessary. However, this method cannot be applied when the dimension of the exchanged information is too high. Another way of reducing the numerical cost is based on model order reduction methods by projection. To do so, the outputs of the models are projected onto a reduced basis of lower dimension. One of the limitations of this approximation is the evaluation of the error made by the projection. The objective of the thesis is the coupling of GP and model order reduction to propose a method for solving a multidisciplinary optimization problem overcoming the proposed challenges. To this end, the idea is to replace each disciplinary solver by an approximation combining projection and interpolation by GP. However, in order to certify that the approximation is accurate, the error made shall be estimated in order to refine the approximation if needed. A first part of our work consists in using the Petrov-Galerkin projection as a model order reduction technique and estimating the projection error by statistical analysis and preconditioning. An application is proposed on the optimization of a quantity of interest using Bayesian optimization. More precisely, the error is estimated to determine if the approximated solution shall be used instead of the exact solution. This coupling is tested on the preliminary design of an aircraft wing and shows a strong potential for reducing computational cost. Secondly, we use GP to interpolate the coefficients on the reduced basis and estimate the error made using the variance of the GP. A training method for the construction of the reduced basis and the GP is proposed in the context of multidisciplinary analysis. The errors caused by the approximation of each solver are then propagated to obtain an estimate of the error on the MDA. An MDA solver is proposed by enrichment of the disciplinary GP and reduced basis until the uncertainties caused by the interpolation are considered sufficiently low. This algorithm was tested in an aeroelastic context modeling an airplane wing. Subsequently, several couplings with multidisciplinary optimization algorithms showed the interest of the method by greatly lowering the optimization budget.; Ces dernières décennies, les concepteurs avant-projet en aéronautique ont porté un intérêt particulier sur l’interaction entre les différents experts impliqués dans la modélisation des performances d’un avion. Ces experts utilisent des codes de simulation numériques dits disciplinaires pour modéliser leur discipline (par exemple l’aérodynamique, la structure ou la propulsion) et échangent des informations afin de prendre en compte l’influence des autres disciplines sur la leur. Le système couplé via ces informations est nommé analyse multidisciplinaire (MDA). La recherche de la meilleure configuration, nommée optimisation multidisciplinaire, consiste à identifier les variables de conception optimisant une fonction de performance calculée à partir de la solution de la MDA. La résolution de ce problème d’optimisation soulève plusieurs défis. Premièrement, l’évaluation de chaque modèle disciplinaire est coûteuse et minimiser les appels à ces solveurs lors de l’optimisation représente un des principaux objectifs de la thèse. Deuxièmement, nous considérons que les informations échangées entre les solveurs disciplinaires sont de grande dimension (typiquement des champs discrétisés dont la dimension excède 1000). Finalement, la méthode de résolution doit être la plus générique possible : non intrusive afin de pouvoir utiliser des solveurs disciplinaires industriels et pouvant s’adapter à des MDA faisant intervenir un nombre variable de disciplines impliquées dans le processus de conception. Plusieurs formulations ont été proposées pour la résolution de ce type de problème mais aucune ne permet de relever entièrement ces défis. Une des propositions les plus prometteuses est de remplacer chacun des solveurs disciplinaires par des processus gaussiens (GP) et de coupler ces solveurs approchés pour obtenir une approximation de la MDA. L’idée est alors d’utiliser la variance des GP pour estimer les incertitudes sur la MDA et d’enrichir les GP lorsque cela s’avère nécessaire. Cependant, cette méthode ne peut être appliquée lorsque les informations échangées sont de grande dimension. Une autre voie de réduction du coût numérique est basée sur les méthodes de réduction d’ordre de modèle par projection. Pour cela, les sorties des modèles sont projetées sur une base réduite de faible dimension. Une des limitations de cette approximation est l’évaluation de l’erreur commise par la projection. L’objectif de la thèse est alors de coupler GP et réduction d’ordre de modèle afin de proposer une méthode de résolution d’un problème d’optimisation multidisciplinaire permettant de relever les défis proposés. Pour cela, l’idée est de remplacer chaque solveur disciplinaire par une approximation combinant projection et interpolation par GP. Cependant, afin de certifier que l’approximation est fidèle, l’erreur commise devra être estimée afin d’affiner l’approximation si celle-ci est trop importante. Une première partie de nos travaux consiste à utiliser la projection de Petrov-Galerkin comme technique de réduction d’ordre de modèle et d’estimer l’erreur de projection par analyse statistique et préconditionnement. Une application est proposée sur l’optimisation d’une quantité d’intérêt via un couplage avec un algorithme bayésien. Plus précisément, l’estimation de l’erreur est utilisée pour déterminer si la solution réduite peut être utilisée à la place de la solution exacte. Ce couplage testé sur le dimensionnement d’une aile d’avion montre un fort potentiel de réduction des temps de calcul. Dans un second temps, nous avons utilisé les GP pour interpoler les coefficients de la base réduite et estimé l’erreur d’approximation grâce à la variance des GP. Une méthode d’entraînement spécifique au contexte multidisciplinaire est proposée pour la construction des bases réduites et des GP. Les erreurs induites par l’approximation de chaque solveur sont alors propagées afin d’obtenir une estimation de l’erreur sur la MDA. Un algorithme de résolution de MDA est proposé par enrichissement des GP disciplinaires et des bases réduites jusqu’à ce que les incertitudes liées à l’interpolation soient considérées comme suffisamment faibles. Cet algorithme a été testé dans un contexte aéroélastique modélisant une aile d’avion. Par la suite, plusieurs couplages avec des algorithmes de résolution d’optimisation multidisciplinaire ont montré l’intérêt de la méthode en permettant de fortement réduire le budget de l’optimisation.

International audience; Nowadays, experimental analyses of fluid-structure interactions on flexible rotating blades are few widespread. For this purpose, an experimental rotor test bench for blades with variable flexibility is under development. Different sets of blades have been fabricated with different composite materials and variable thickness, depending on the number of layers considered.The main objective of the present work is to estimate which sets of blades are more sensitive to flutter occurrence based on their Campbell diagrams. For this purpose, a finite element approach is developed to model the dynamic behavior of the blades assumed as a uniform, slender, isotropic and homogeneous beam with out-of-plane bending and torsion deformations, submitted to inertial and centrifugal forces due the rotation effects. Two numerical models were developed by considering Timoshenko and Euler-Bernoulli beam theories; both are validated by means of the software COMSOLr. Later, the mechanical properties of each set of blades have been identified from the frequency response functions (FRFs) obtained through experimental vibration tests. The frequencies and eigenmodes of blades determined experimentally are compared with those obtained with FEM model. Then, rotation effects on the blade frequencies are studied by using Campbell diagrams for all set of blades studied. It is evidenced for some of the sets the presence of intersections between bending and torsional frequencies in the interval of the operational rotating speed of the test bench. This fact might be considered as an indicator of potential flutter occurrence.Aerodynamic load will be consider in the future work.

International audience; Aeroelasticity plays an important role in the design and certification of aircrafts. Whether steady aerodynamics is key for static methods, unsteady phenomenon is key for dynamic approaches. Nevertheless, in both aeroelastic problematics, it is basically aerodynamics who monopolizes the CPU cost. Thus, fast methods are interesting as alternative to the high cost of CFDCSM simulations.The aim of this study is to present an unsteady method that predicts the aerodynamics of aircrafts encountering a wind gust. First of all, the process that brought to the formulation of the static aeroelasticity from Wind Tunnel Tests (WTT) is refreshed. It is a rapid and robust method based on the principle of the local incidence shift due to flexible effects. This difference in the angle of attack is used to interpolate the pressure coefficient from aerodynamics lookup tables.Secondly, the static method (called Fast Nonlinear Static Aeroelasticity –FNSA–) is validated when analyzing a high aspect ratio wing which assures a flexible structure. The method deals with nonlinear, flexible and compressible effects in steady aerodynamics.Finally, an algorithm is proposed to integrate the FNSA formulation to the unsteady problem of gust load computation. The methodology uses a quasi-steady approach of the static solver together with Wagner delay function modeling the unsteadiness of the flow. Restricting our aeroelastic equation of motion to a simple mass-stiffness system, the airstream directed downward due to the gust is added to the downwash of each section. The formulation allows, then, a reaction to the gust at each time step that yields to flexible results with an accuracy comparable to the Unsteady Vortex Lattice Method but around 3600 times faster.

International audience; Prognostics and Health Management (PHM) relies on the availability of large amounts of data for a given system and allows to analyse this data and to draw conclusions as to the health state of the system, the identification of faults and failures, as well as the calculation of the remaining useful life time. Often, it is expected that this data is labelled, i.e. that the data has been pre-analysed and that for each data point an exact information is available as to what it is about, when it was measured, etc. In reality, this is not always easy and this labeled data is not always available. For example, on aerospace structures, complete labeled data until the end of their lifetime are not usually available. This may hamper for example the use of Deep Learning (DL) techniques for Predictive Maintenance, as they rely on the availability of large amounts of labeled sensor data. In this paper a framework and associated code r is proposed to generate high dimensional data sets for a realistic fatigue damage prognostics problem, representative of fatigue cracks propagation in aeronautical fuselage panels. With this data, DL techniques can be trained, and we will illustrate this with a case study involving several of the most commonly used DL models to address failure prognostics.

International audience; In this paper, a closed-form analytical solution procedure is proposed to solve the coupled fluid-structure interaction (FSI) equations that involve the first-order Doubly Asymptotic Approximation (DAA1) formulation. An efficient method comprised of the nonstandard finite difference (NSFD) scheme is applied. First of all, analytical equations are developed to analyze the response of a rigid mass-spring oscillator in an air-backed condition when it is subjected to a plane shock exponential wave coming from a far-field underwater explosion. After validating the results with LS-DYNA/USA (DAA1), these equations are extended to determine the response of a two-dimensional, simply-supported rectangular plate, and then tested on the isotropic and orthotropic plates within a small deflection regime. Parametric studies are also performed by varying the load decay times, peak pressures, as well as the aspect ratios of the plate. Finally, the advantages and limitations of the proposed formulae are exposed along with suggestions for the future work.