<p>In this thesis, inverse problems related to microwave imaging and microwave component design are investigated. Our contribution in microwave imaging for breast tumor detection can be divided into two parts. In the first part, a vectorial 3D near-field microwave holography is proposed which is an improvement over the existing holography algorithms. In the second part, a simple and fast post-processing algorithm based on the principle of blind de-convolution is proposed for removing the integration effect of the antenna aperture. This allows for the data collected by the antennas to be used in 3D holography reconstruction. The blind deconvolution algorithm is a well-known algorithm in signal processing and our contribution here is its adaptation to microwave data processing.</p> <p>Second, a procedure for accelerating the space-mapping optimization process is presented. By exploiting both fine- and surrogate-model sensitivity information, a good mapping between the two model spaces is efficiently obtained. This results in a significant speed-up over direct gradient-based optimization of the original fine model and enhanced performance compared with other space-mapping approaches. Our approach utilizes commercially available software with adjoint-sensitivity analysis capabilities.</p> / Thesis / Master of Applied Science (MASc)

<p>This investigation is motivated by the problem of optimal design of cooling elements in modern battery systems used in hybrid/electric vehicles. We consider a simple model of two-dimensional steady-state heat conduction generated by a prescribed distribution of heat sources and involving a one-dimensional cooling element represented by a closed contour. The problem consists in finding an optimal shape of the cooling element which will ensure that the temperature in a given region is close (in the least squares sense) to some prescribed distribution. We formulate this problem as PDE-constrained optimization and use methods of the shape-differential calculus to obtain the first-order optimality conditions characterizing the locally optimal shapes of the contour. These optimal shapes are then found numerically using the conjugate gradient method where the shape gradients are conveniently computed based on adjoint equations. A number of computational aspects of the proposed approach is discussed and optimization results obtained in several test problems are presented.</p> / Master of Science (MSc)

Le progrès technologique nécessite des techniques de plus en plus sophistiquées et précises de traitement de matériaux. Nous étudions le traitement de matériaux par faisceaux de haute énergie : un jet d’eau abrasif, une sonde ionique focalisée, un laser. L’évolution de la surface du matériau sous l’action du faisceau de haute énergie est modélisée par une EDP. Cette équation contient l’ensemble des coefficients inconnus - les paramètres de calibration de mo- dèle. Les paramètres inconnus peuvent être calibrés par minimisation de la fonction coût, c’est-à-dire, la fonction qui décrit la différence entre le résultat de la modélisation et les données expérimentales. Comme la surface modélisée est une solution du problème d’EDP, cela rentre dans le cadre de l’optimisation sous contrainte d’EDP. L’identification a été rendue bien posée par la régularisation du type Tikhonov. Le gradient de la fonction coût a été obtenu en utilisant les deux méthodes : l’approche adjointe et la différen- ciation automatique. Une fois la fonction coût et son gradient obtenus, nous avons utilisé un minimiseur L-BFGS pour réaliser la minimisation.Le problème de la non-unicité de la solution a été résolu pour le problème de traitement par le jet d’eau abrasif. Des effets secondaires ne sont pas inclus dans le modèle. Leur impact sur le procédé de calibration a été évité. Ensuite, le procédé de calibration a été validé pour les données synthétiques et expérimentales. Enfin, nous avons proposé un critère pour distinguer facilement entre le régime thermique et non- thermique d’ablation par laser. / The technological progress demands more and more sophisticated and precise techniques of the treatment of materials. We study the machining of the material with the high energy beams: the abrasive waterjet, the focused ion beam and the laser. Although the physics governing the energy beam interaction with material is very different for different application, we can use the same approach to the mathematical modeling of these processes.The evolution of the material surface under the energy beam impact is modeled by PDE equation. This equation contains a set of unknown parameters - the calibration parameters of the model. The unknown parameters can be identified by minimization of the cost function, i.e., function that describes the differ- ence between the result of modeling and the corresponding experimental data. As the modeled surface is a solution of the PDE problem, this minimization is an example of PDE-constrained optimization problem. The identification problem was regularized using Tikhonov regularization. The gradient of the cost function was obtained both by using the variational approach and by means of the automatic differentiation. Once the cost function and its gradient calculated, the minimization was performed using L-BFGS minimizer.For the abrasive waterjet application the problem of non-uniqueness of numerical solution is solved. The impact of the secondary effects non included into the model is avoided as well. The calibration procedure is validated on both synthetic and experimental data.For the laser application, we presented a simple criterion that allows to distinguish between the thermal and non-thermal laser ablation regimes.

No description available.

Advances in numerical optimization have raised the possibility that efficient and novel aircraft configurations may be ``discovered'' by an algorithm. To begin exploring this possibility, a fast and robust set of tools for aerodynamic shape optimization is developed. Parameterization and mesh-movement are integrated to accommodate large changes in the geometry. This integrated approach uses a coarse B-spline control grid to represent the geometry and move the computational mesh; consequently, the mesh-movement algorithm is two to three orders faster than a node-based linear elasticity approach, without compromising mesh quality. Aerodynamic analysis is performed using a flow solver for the Euler equations. The governing equations are discretized using summation-by-parts finite-difference operators and simultaneous approximation terms, which permit nonsmooth mesh continuity at block interfaces. The discretization results in a set of nonlinear algebraic equations, which are solved using an efficient parallel Newton-Krylov-Schur strategy. A gradient-based optimization algorithm is adopted. The gradient is evaluated using adjoint variables for the flow and mesh equations in a sequential approach. The flow adjoint equations are solved using a novel variant of the Krylov solver GCROT. This variant of GCROT is flexible to take advantage of non-stationary preconditioners and is shown to outperform restarted flexible GMRES. The aerodynamic optimizer is applied to several studies of induced-drag minimization. An elliptical lift distribution is recovered by varying spanwise twist, thereby validating the algorithm. Planform optimization based on the Euler equations produces a nonelliptical lift distribution, in contrast with the predictions of lifting-line theory. A study of spanwise vertical shape optimization confirms that a winglet-up configuration is more efficient than a winglet-down configuration. A split-tip geometry is used to explore nonlinear wake-wing interactions: the optimized split-tip demonstrates a significant reduction in induced drag relative to a single-tip wing. Finally, the optimal spanwise loading for a box-wing configuration is investigated.

This thesis considers applications of gradient-based optimization algorithms to the design and control of some mechanics systems. The material distribution approach to topology optimization is applied to design two different acoustic devices, a reactive muffler and an acoustic horn, and optimization is used to control a ball pitching robot. Reactive mufflers are widely used to attenuate the exhaust noise of internal combustion engines by reflecting the acoustic energy back to the source. A material distribution optimization method is developed to design the layout of sound-hard material inside the expansion chamber of a reactive muffler. The objective is to minimize the acoustic energy at the muffler outlet. The presence or absence of material is represented by design variables that are mapped to varying coefficients in the governing equation. An anisotropic design filter is used to control the minimum thickness of materials separately in different directions. Numerical results demonstrate that the approach can produce mufflers with high transmission loss for a broad range of frequencies. For acoustic devices, it is possible to improve their performance, without adding extended volumes of materials, by an appropriate placement of thin structures with suitable material properties. We apply layout optimization of thin sound-hard material in the interior of an acoustic horn to improve its far-field directivity properties. Absence or presence of thin sound-hard material is modeled by a surface transmission impedance, and the optimization determines the distribution of materials along a “ground structure” in the form of a grid inside the horn. Horns provided with the optimized scatterers show a much improved angular coverage, compared to the initial configuration. The surface impedance is handled by a new finite element method developed for Helmholtz equation in the situation where an interface is embedded in the computational domain. A Nitschetype method, different from the standard one, weakly enforces the impedance conditions for transmission through the interface. As opposed to a standard finite-element discretization of the problem, our method seamlessly handles both vanishing and non-vanishing interface conditions. We show the stability of the method for a quite general class of surface impedance functions, provided that possible surface waves are sufficiently resolved by the mesh. The thesis also presents a method for optimal control of a two-link ball pitching robot with the aim of throwing a ball as far as possible. The pitching robot is connected to a motor via a non-linear torsional spring at the shoulder joint. Constraints on the motor torque, power, and angular velocity of the motor shaft are included in the model. The control problem is solved by an interior point method to determine the optimal motor torque profile and release position. Numerical experiments show the effectiveness of the method and the effect of the constraints on the performance.

La simulation numérique est largement utilisée pour évaluer les performances aérodynamiques des aéronefs ainsi qu'en optimisation de forme. Ainsi l'objectif de ces simulations est souvent le calcul de fonctions aérodynamiques. L'objet de cette thèse est d'étudier des méthodes d'adaptation de maillages basées sur la dérivée totale de ces fonctions par rapport aux coordonnées du maillage (notée dJ/dX). Celle-ci pouvant être calculée par la méthode adjointe discrète. La première partie de cette étude concerne l'application de méthodes d'adaptation de maillages appliquées à des écoulements de fluides parfaits. Le senseur qui détecte les zones de maillage à raffiner s'appuie sur la norme de cette dérivée pour adapter des maillages pour le calcul d'une fonction J. La seconde partie du travail est la construction et l'étude de critères plus fiables basés sur dJ/dX pour d'une part adapter des maillages et d'autre part estimer si un maillage est bien adapté ou non pour le calcul de la fonction J. De plus une méthode de remaillage plus efficace basée sur une EDP elliptique est aussi présentée. Cette nouvelle méthode est appliquée pour des écoulements bidimensionnels de fluides parfaits ainsi que pour un écoulement décrit par les équations RANS. La dernière partie de l'étude est consacrée à l'application de la méthode proposée à des cas tridimensionnels d'écoulement RANS sur des géométries d'intérêt industriel. / Numerical simulation is widely used for the assessment of aircraft aerodynamic performances and shape optimizations. Hence the objective of these simulations is often to compute aerodynamic outputs. The purpose of this thesis is to study mesh adaptation methods based on the total derivative of the outputs with respect to mesh coordinates (denoted dJ/dX). This derivative can be computed using the discrete adjoint method. The first part of this study is about the application of mesh adaptation methods applied for Eulerian flows. The mesh locations to refine are detected using a sensor based on the norm of the derivative dJ/dX. This study confirmed that this derivative is relevant in order to adapt a mesh for the computation of the output J. The second part of this work is the construction and the study of reliable criteria based on dJ/dX for both mesh adaptation and the quality assessment of a given mesh for the computation of the output J. Moreover a more efficient remeshing method based on an elliptic PDE is presented too. This new method is applied for both two-dimensional Eulerian flows and a flow described by the RANS equations. The last part of the study is devoted to the application of the proposed method to three-dimensional RANS flows on geometries of industrial interest.

<div>The rapid development of topology optimization in photonics has greatly expanded the number of photonic structures with extraordinary performance. The optimization is usually solved by using a gradient-based optimization algorithm, where gradients are evaluated by the adjoint sensitivity analysis. While the adjoint sensitivity analysis has been demonstrated to provide reliable gradients for designs of dielectrics, there has not been too much success in plasmonics. The difficulty of obtaining accurate field solutions near sharp edges and corners in plasmonic structures, and the strong field enhancement jointly increase the numerical error of gradients, leading to failure of convergence for any gradient-based algorithm. </div><div> </div><div>We present a new method of calculating accurate sensitivity with the FDTD method by direct differentiation of the time-marching system in frequency domain. This new method supports general frequency-domain objective functions, does not relay on implementation details of the FDTD method, works with general isotropic materials and can be easily incorporated into both level-set-based and density-based topology optimizations. The method is demonstrated to have superior accuracy compared to the traditional continuous sensitivity. Next, we present a framework to carry out density-based topology optimization using our new sensitivity formula. We use the non-linear material interpolation to counter the rough landscape of plasmonics, adopt the filteringand-projection regularization to ensure manufacturability and perform the optimization with a continuation scheme to improve convergence. </div><div> </div><div>We give two examples involving reconstruction of near fields of plasmonic structures to illustrate the robustness of the sensitivity formula and the optimization framework. In the end, we apply our method to generate a rectangular temperature profile in the recording medium of the HAMR system. </div>

La réduction des nuisances sonores est un enjeu permanent pour les acteurs de l’aéronautique. L’optimisation de la réduction de bruit apportée par les traitements acoustiques tapissant la nacelle des réacteurs turbofan passe par une caractérisation précise des matériaux employés dans l’environnement aéroacoustique d’utilisation, qui met en jeu un écoulement rasant de vitesse importante combiné à de forts niveaux sonores. L’objectif de cette thèse est de développer une méthode inverse pour la détermination de l’impédance acoustique de liners soumis à un écoulement rasant, basée sur des mesures non intrusives du champ de vitesse acoustique au-dessus du matériau par Vélocimétrie Laser Doppler (LDV). L’impédance de liner est obtenue par minimisation de l’écart entre le champ de vitesse acoustique mesuré et le champ simulé numériquement en résolvant les équations d’Euler linéarisées bidimensionnelles harmoniques, discrétisées par un schéma Galerkin discontinu. Le gradient de la fonction objectif minimisée est calculé via la résolution, à chaque itération, des équations directes et adjointes. Une première étape de validation du solveur est effectuée sur des cas-tests académiques, puis sur des cas expérimentaux impliquant des mesures de pression acoustique en paroi rigide opposée au liner. Dans un second temps, la méthode est appliquée à des mesures de vitesse acoustique obtenues par LDV dans le banc B2A de l’ONERA en l’absence d’écoulement. La dernière étape consiste à prendre en compte l’effet d’un écoulement rasant de profil cisaillé. Les impédances identifiées à partir de mesures LDV en présence d’écoulement ont notamment permis de gagner en compréhension sur les phénomènes d’absorption intervenant dans le banc B2A. / While aircraft noise constraints become increasingly stringent, efficient duct treatment of turbofan engines requires an accurate knowledge of liner impedance with grazing flow at high acoustic levels. This thesis aims at developing an impedance eduction method in the presence of grazing flow. The inverse process is based on acoustic velocity fields acquired by Laser Doppler Velocimetry (LDV) above the liner. The liner acoustic impedance is obtained by minimization of the distance between the measured acoustic velocity field and the simulated one. Computations rely on the resolution of the 2D linearized Euler equations in the harmonic domain, spatially discretized by a discontinuous Galerkin scheme. The gradient of the objective function is achieved by the resolution, at each iteration on the liner impedance, of the direct and adjoint equations. The solver is first validated on academic test cases, then on experimental results of acoustic pressure measurements at the rigid wall opposite the liner. Secondly the method is applied to acoustic velocity measurements obtained by LDV above the liner without flow, in the ONERA B2A test bench. The last step consists in taking into account the effects of a sheared grazing flow. The impedances educed from LDV measurements in the presence of flow namely allowed to gain insight into the absorption phenomena occuring in the B2A test bench.

Dans cette thèse, nous nous sommes intéressés à l'estimation des paramètres de transport de solutés dans un milieu poreux peu perméable, d'un point de vue théorique et appliqué, en lien avec le stockage des déchets nucléaires à vie longue en couche géologique profonde. La problématique de la gestion des déchets nucléaires est devenue durant ces dernières années une préoccupation importante de la société. Les modèles mathématiques utilisés pour le stockage des déchets en site profond sont basés sur les équations d'écoulement et de transport en milieu poreux. Ces modèles nécessitent de connaître les paramètres physiques des différentes couches géologiques. Ces paramètres (porosité et diffusion) n'étant pas accessibles directement, nous avons dans un premier temps étudié l'aspect identification de paramètres à zones fixées. Pour résoudre notre problème inverse, nous avons utilisé des méthodes déterministes (ou variationnelles) basées sur la minimisation d'un critère (ou fonction coût) quadratique, l'objectif étant de trouver l'ensemble des paramètres qui minimisent le critère. Nous avons décomposé le milieu en zones homogènes dans lesquelles les coefficients ont été identifiés (technique de zonation). Pour cela, nous avons différentié le code de convection-diffusion Traces avec le logiciel de différentiation automatique Tapenade afin d'obtenir le code adjoint de Traces. Dans un second temps, nous avons utilisé ce travail pour résoudre le problème d'identification de zones. On a cherché, partant d'une configuration géologique modifiée, à retrouver la configuration géologique réelle en identifiant les paramètres qui ont pour support ces zones. En regroupant ensuite par isovaleurs les coefficients identifiés, nous avons obtenu la configuration recherchée.