Search

L'oreille interne est un organe fascinant du corps humain. Elle contient des organes sensoriels très précis et hypersensibles, ce qui lui permet de jouer un rôle majeur dans la perception de nos mouvements et de notre orientation spatiale. Dans un premier temps, ce travail de thèse a porté sur la modélisation du fonctionnement des senseurs d'orientation de l'oreille interne. Un démonstrateur type « Réalité Virtuelle » a été développé sous Matlab/Simulink afin de visualiser en temps réel l'état de chaque senseur. Une modélisation plus détaillée par éléments finis et tenant compte d'interactions fluide/structure a permis d'étudier la dynamique des fluides au sein de chaque capteur ainsi que le déplacement de membranes - éléments clés permettant de coupler le déplacement du fluide avec la stimulation de cellules sensorielles. Dans un second temps, ce travail de thèse s'est orienté vers le développement de modèles non-linéaires et tridimensionnels de perception de l'orientation spatiale. Ces modèles supposent que notre cerveau estime/calcul nos perceptions d'orientation, de vitesse, et d'accélération de façon « optimale ». Par conséquent, les modèles développés se sont appuyés sur deux techniques d'estimation non-linéaires basées sur le filtre de Kalman (« Extended Kalman filter » & « Unscented Kalman filter »). En réponse à différent profils de stimulation, ces modèles permettent de prédire diverses illusions sensorielles connues dans le monde de l'aéronautique. En tant qu'applications potentielles, ces modèles pourraient être utilisés d'une part lors d'investigation de crash d'appareil afin de détecter si le pilote a été sujet à un phénomène de désorientation spatiale, et d'autre part pour le développement d'algorithmes de contrôle des simulateurs de vols. The non-auditory section of the human innner ear, the vestibular system, is recognized as the prime motion sensing center. The vestibular system is comprised of two primary sensory organs and represents an inertial measuring device which allows us to sense self-motion with respect to the six degrees of freedom in space. The scope of the work presented in this thesis concerns on one hand the modelling of the vestibular sensors, and on the other hand nonlinear models for human spatial orientation perception. First, a virtual reality model of the vestibular sensors has been developed so as to visualize in real time the state of each sensor in response to any kind of head motion. Second, a three-dimensional model of the entire set of canal using fluid-structural finite-elements simulations has been proposed. Using a strong coupling between the fluid flow and the structural displacements and also an Arbitrary Lagrangian Eulerian (ALE) approach for the moving mesh, we analyze displacements of the cupulae and fluid velocity during head rotation. Third, we developed a nonlinear model of human spatial orientation based on the Unscented Kalman Filter. This model successfully predicts the responses to a number of vestibular, visual and visual-vestibular motion paradigms. It turns out that the UKF yields more accurate and less oscillatory responses than Pommellet’s Extended Kalman Filter model. As a prospect, this model could be used to investigate aircraft crashes so as to detect whether or not pilots have experienced a phenomenon of spatial disorientation.

Experimental results are used to validate a 2D phenomenological model of the near wake based on Van Der Pol wake oscillators, which is the first one which describes the 2D motion of a cylinder in its transversal plan. In the case of a single cylinder, experimental and numerical results are in relatively good agreement. Those first results show that the proposed model can be used as a simple computation tool in the prediction of 2D VIV effects.

There are two possible approaches to solving Fluid-Structure Interaction (FSI) problems computationally. One approach is to link together preexisting (often commercial) solvers, usually with a finite-volume (FV) solver for the fluid flow, a finite-element (FE) solver for the structural analysis and a third, bridging code performing the tasks of coupling the solvers, data interpolation and simulation management. This has the benefit of utilising existing and well validated solver code. However it also imposes limitations on the mode of coupling and creates problems with the model setup. For example, it is quite difficult to achieve converged coupled solution due to difficulties with creating a coupled iteration loop across a fluid and a structural solver within a single time-step. The alternative approach is to undertake the whole solution problem entirely within a single code, which involves implementing the fluid computation in FE software or the structural computation in FV software. Within the Open Source OpenFOAM toolkit, this methodology has been implemented and tested on simple geometries such as flow past a cantilevered elastic square beam. The fluid flow is modelled using the Navier-Stokes equations for incompressible Newtonian fluid, modelled using the Arbitrary Lagrangian-Eulerian formulation (ALE), whilst the solid deformation is described using the geometrically nonlinear momentum equation in Lagrangian formulation. Spatial discretisation of both models is performed using second-order accurate unstructured FV method, with automatic mesh motion used to accommodate fluid-solid interface deformation. Coupling between the two models is performed using a loosely-coupled staggered solution algorithm, with the force transfered in one direction and the displacement in the other. One area in which FSI is important is biomedical simulation, particularly of flow in arteries. The compliance of arterial walls is an important aspect of the flow through the cardiac cycle, and can be of crucial importance in a number of medical conditions such as the development of aneurysms. Simulation of the flow in such cases is complicated by the geometry of the problem, which is complex across a range of length scales, with the artery changing direction and cross-section continuously, and also branching. Not only is this geometry complex to characterise and specify, but it also varies from individual to individual in ways which may be of interest ; simulation on idealised geometry can be shown to generate results which are qualitatively as well as quantitatively in error. Thus there has been a great deal of research into techniques for generating computational meshes from medical scans (MRI or CT scans), which enable the generation of patient-specific geometries and meshes. One approach to this task is the Voxel method, in which the scan pixels representing the volumes of interest are segmented out into separate masks ; these segmented regions are then meshed using the Marching Cubes algorithm, and truncated at the boundaries of each mask. This approach is implemented in the ScanIP/ScanFE package (Simpleware Ltd), which also includes image manipulation tools to preprocess the raw data, automatic and manual tools for segmentation, and subsequent mesh manipulation to smooth surfaces or improve mesh quality are also available. The output from this is a high quality tet or mixed tet/hex mesh for each segmented region, with guaranteed matching between the interface boundaries. In this work we present the results of a demonstration study involving FSI on arterial flow. Meshes have been generated for the flow domain and the artery wall using MRI scans of a femoral artery, and converted to OpenFOAM mesh format. Calculations have then been performed assuming the blood flow to be Newtonian and laminar, for a simple sinusoidal inlet flow variation and compliant walls, using the coupled FSI solver in OpenFOAM. Results are presented for wall shear stress and wall internal stress.

This thesis investigates the characteristics of a close proximity underwater explosion and its effect on a ship-like structure. Finite element model tests are conducted to verify and validate the propagation of a pressure wave generated by an underwater explosion through a fluid medium, and the transmission of the pressure wave in the fluid to a structure using the Multi-Material Arbitrary Lagrangian/Eulerian method. A one dimensional case modeling the detonation of a spherical TNT charge underwater is investigated. Three dimensional cases modeling the detonation of an underwater spherical TNT charge, and US Navy Blast Test cases modeling a shape charge and a circular steel plate, and a shape charge and a Sandwich Plate System (SPS) are also investigated. This thesis provides evidence that existing tools and methodologies have some capability for predicting early-time/close proximity underwater explosion effects, but are insufficient for analyses beyond the arrival of the initial shock wave. This thesis shows that a true infinite boundary condition, a modified Gruneisen equation of state near the charge, and the ability to capture shock without a very small element size is needed in order to provide a sufficient means for predicting early-time/close proximity underwater explosion effects beyond the arrival of the initial shock wave. Master of Science

In this paper, we consider a multi-disciplinary optimization problem where the initial shape of a wing is sought in order to cope, after elastic deformation by the flow, with some optimality conditions. We propose a medium-strong coupling which allows to consider different softwares communicating a small number of times. Applications to the optimization of the AGARD Wing 445.6 and a flexible supersonic aircraft wing are presented.

In order to provide complementary means for measurement of water levels in pressurized vessels of nuclear reactors, the possibilities of an immersed torsional wave sensor are being looked into. It has already been modeled, considering an incompressible fluid. Yet, because the fluid can turn into a two-phase fluid, we have investigated how to extend the existing model. As a first step, we have taken into account the compressibility of the surrounding fluid. We focus on a cylindrical waveguide with an elliptic cross-section. Its transverse dimensions are small compared with its length and the wavelengths in the fluid. We start with the elasticity equations for the waveguide. Then, from the exact expression of the pressure exerted by the fluid on the waveguide boundary, a long wavelength approximation is obtained. We end by applying Hamilton's principle of energy conservation, which leads to an approximate equation governing the fluid-loaded waveguide motion, and to an expression of the apparent phase velocity of the torsional wave in the immersed waveguide. Finally, fluid level measurement is possible. Some simulations are made, highlighting the influence of the compressibility.; Afin de renforcer l'instrumentation existante pour la mesure de niveau d'eau dans des enceintes de Réacteurs à Eau Pressurisée, on s'intéresse à une méthode fondée sur l'étude de la propagation d'une onde de torsion dans un guide d'ondes immergé. Antérieurement à cette thèse, ce dispositif avait été modélisé dans le contexte particulier d'un fluide environnant monophasique et incompressible. Notre modèle permet d'utiliser la sonde à onde de torsion pour mesurer des niveaux de fluides compressibles, situation qui apparaît en configuration accidentelle lorsque le fluide environnant devient diphasique. Nous considérons un guide d'ondes cylindrique de section elliptique, dont les dimensions transversales sont petites devant sa longueur et la longueur d'onde dans le fluide. Nous écrivons d'abord les équations d'élasticité du guide. Ensuite, nous exprimons la pression exercée par le fluide à la surface du guide, et nous en calculons une approximation "grandes longueurs d'ondes". Nous appliquons finalement le principe de conservation de l'énergie de Hamilton pour obtenir une équation approchée du mouvement du guide d'ondes, et une expression de la vitesse de phase apparente de l'onde de torsion dans le guide immergé. Nous pouvons alors utiliser ces résultats pour la mesure de niveau. Quelques simulations sont réalisées, qui montrent la pertinence de la prise en compte de la compressibilité.

Hydroelasticity is a phenomenon occurring during the interaction between water and a deformable structure. Structural deformations can modify the fluid motion, introducing difficulties in the calculation of the impact-induced stresses. Predicting the structural deformations and stresses during the water entry of flexible structures is a major challenge and suitable computer-aided design tools are currently being developed and validated. This work investigates the water impact of deformable wedges. Experimental and numerical results on the parameters affecting hydroelasticity are evaluated for deformable plates with different boundary conditions. The occurrence of hydroelastic effects is shown to depend on the ratio between the natural frequency of the structure and the characteristic wetting time. The conditions for which fluid-structure interactions are significant and must be accounted for in the analysis are identified, while outside this range the structure can be modeled as a rigid body.

International audience; The enhancement of high altitude drone endurance compels to design very flexible high-aspect-ratio composite airframe vulnerable to destructive fluid/structure interaction like flutter or torsional divergence. Extensive research has been conducted to increase critical speed without being at the expense of weight balance, one of the promising solutions is the aeroelastic tailoring which consists in a specific configuration of laminated composite layup. The present work presents an aeroelastic reduced order model suitable for the non linear anisotropic behavior of this kind of composite wing implemented in Fortran and Python using optimised open source solver. Along with validation test cases, a simple composite laminate specimen simulation is presented to assess aeroelastic tailoring effect.