Les métaheuristiques sont très utilisées dans le domaine de l'optimisation discrète. Elles permettent d’obtenir une solution de bonne qualité en un temps raisonnable, pour des problèmes qui sont de grande taille, complexes, et difficiles à résoudre. Souvent, les métaheuristiques ont beaucoup de paramètres que l’utilisateur doit ajuster manuellement pour un problème donné. L'objectif d'une métaheuristique adaptative est de permettre l'ajustement automatique de certains paramètres par la méthode, en se basant sur l’instance à résoudre. La métaheuristique adaptative, en utilisant les connaissances préalables dans la compréhension du problème, des notions de l'apprentissage machine et des domaines associés, crée une méthode plus générale et automatique pour résoudre des problèmes. L’optimisation globale des complexes miniers vise à établir les mouvements des matériaux dans les mines et les flux de traitement afin de maximiser la valeur économique du système. Souvent, en raison du grand nombre de variables entières dans le modèle, de la présence de contraintes complexes et de contraintes non-linéaires, il devient prohibitif de résoudre ces modèles en utilisant les optimiseurs disponibles dans l’industrie. Par conséquent, les métaheuristiques sont souvent utilisées pour l’optimisation de complexes miniers. Ce mémoire améliore un procédé de recuit simulé développé par Goodfellow & Dimitrakopoulos (2016) pour l’optimisation stochastique des complexes miniers stochastiques. La méthode développée par les auteurs nécessite beaucoup de paramètres pour fonctionner. Un de ceux-ci est de savoir comment la méthode de recuit simulé cherche dans le voisinage local de solutions. Ce mémoire implémente une méthode adaptative de recherche dans le voisinage pour améliorer la qualité d'une solution. Les résultats numériques montrent une augmentation jusqu'à 10% de la valeur de la fonction économique. / Metaheuristics are a useful tool within the field of discrete optimization that allow for large, complex, and difficult optimization problems to achieve a solution with a good quality in a reasonable amount of time. Often metaheuristics have many parameters that require a user to manually define and tune for a given problem. An adaptive metaheuristic aims to remove some parameters from being tuned or defined by the end user by allowing the method to specify and/or adapt a parameter or set of parameters based on the problem. The adaptive metaheuristic, using advancements in understanding of the problem being solved, machine learning, and related fields, aims to provide this more generalized and automatic toolkit for solving problems. Global optimization of mining complexes aims to schedule material movement in mines and processing streams to maximize the economic value of the system. Often due to the large number of integer variables within the model, complicated constraints, and non-linear constraints, it becomes prohibitive to solve these models using commercially available optimizers. Therefore, metaheuristics are often employed in solving mining complexes. This thesis builds upon a simulated annealing method developed by Goodfellow & Dimitrakopoulos (2016) to optimize the stochastic global mining complex. The method outlined by the authors requires many parameters to be defined to operate. One of these is how the simulated annealing algorithm searches the local neighborhood of solutions. This thesis illustrates and implements an adaptive way of searching the neighborhood for increasing the quality of a solution. Numerical results show up to a 10% increase in objective function value.

The presented master thesis is focused on the stock markets returns analysis using a new type of neural network. First chapter of the thesis describes the underlying theory of the financial time series prediction, Efficient Market Hypothesis and conventional forecasting models. Following part illustrates biological framework, basic principles, functioning of neural networks, their architecture and several well-known learning algorithms such as Gradient descent, Levenberg-Marquardt algorithm or Conjugate gradient. It also mentions certain disadvantages which influence the performance and effectiveness of neural networks. Third chapter is devoted to two applied metaheuristic techniques, i.e. genetic algorithms and simulated annealing that were integrated into neural networks framework to eliminate above mentioned drawbacks. Next chapter describes details of presented hybrid network, whereas the last section is aimed at evaluation of overall results of all models. It shows that on the examined sample hybrid network clearly outperformed standard techniques as well as ordinary neural networks and in most cases achieved the least mean squared error among all explored methods. Keywords: stock returns analysis, neural networks, genetic algorithms, simulated annealing, hybrid networks JEL classification:...

Cette étude prend place dans le cadre du projet ANR Inxilicium visant à la réalisation de cellules solaires en couches minces de silicium par jet d’encre. Les nanoparticules de silicium sont des matériaux à fort potentiel pour la levée de verrous technologiques grâce à leurs propriétés spécifiques. Des encres de nanoparticules de Si issues de diverses méthodes de synthèse ont été imprimées par jet d’encre sur différents substrats : quartz, électrodes métalliques (aluminium, molybdène) et transparente conductrice (ZnO:Al). L’optimisation du procédé d’impression, de l’interaction encre/substrat (via la modulation de l’énergie de surface des substrats) et de l’étape de séchage a permis l’obtention de couches minces homogènes et continues (plusieurs centaines de nm à quelques µm d’épaisseur)A posteriori, une étape de recuit est nécessaire pour recouvrer des propriétés fonctionnelles. L’utilisation de nanoparticules à la physico-chimie de surface contrôlée fait décroître les températures de frittage de 1100 °C à environ 600 °C. En complément, des recuits sélectifs (micro-ondes et photonique) ont été évalués pour leur application sur des substrats flexibles et bas coûts.Les propriétés optiques et les interfaces électrode/silicium ont été examinées afin d’intégrer ces couches dans des dispositifs (cellule solaire…). La formation de transitions métallurgiques Al-Si et Mo-Si a été étudiées par DRX-in situ. L’ensemble de ces travaux a permis la réalisation d’une jonction PN montrant un comportement photovoltaïque à fort champ grâce aussi à la mise au point d’une méthode innovante de collage ouvrant la voie à une réduction du bilan thermique des procédés de fabrication. / This study takes place in the frame of the Inxilicium project from the National Research Agency, which targets the fabrication of silicon thin film solar cells by inkjet-printing. Thanks to their specific properties, silicon nanoparticles are materials with strong potential for technological breakthroughs. Silicon nanoparticle-based inks made by different synthesis routes have been inkjet-printed on different substrates: quartz, metallic electrodes (aluminum, molybdenum) and transparent electrodes (ZnO:Al). Homogeneous and continuous thin films (from several hundreds of nm to some µm thick) have been obtained through optimization of the printing process, the ink/substrate interaction (via substrates surface energy tuning) and the drying step.A posteriori, an annealing step is mandatory for recovering of functional properties. By using nanoparticles with tailored surface physical chemistry, the sintering temperature decreases from 1100 °C to 600 °C. In order to allow the use of this material on flexible and low cost substrates, selective sintering (microwave and photonics) have been also evaluated.Thin film optical properties and electrode/silicon interfaces have been investigated with the purpose to integrate those layers into devices (solar cells…). Metallurgical evolution of Al-Si and Mo-Si physical interfaces has been studied by in situ XRD.This work allowed the fabrication of a PN junction with a photovoltaic behaviour under strong polarization voltage thanks to the development of an innovative thermal pasting process, which opens the way to the reduction of process thermal budget.

Approved for public release, distribution is unlimited / The extreme strain, strain rate and temperature gradients during Friction Stir Processing (FSP) render measurement of key parameters in the stir zone infeasible with common methods. The objective of this research was to separate the effects that temperature and deformation in an experimental study of the microstructure and mechanical properties of Ni-AL bronze (NAB). This was accomplished by subjecting as-cast NAB material to several isothermal annealing and quenching treatments as well as isothermal hot rolling processes. Sufficient material was generated to provide results and data for subsequent optical microscopy, tensile, & hardness tests. All results were then compared to similar data collected from previous works completed here at Naval Postgraduate School and with other DARPA FSP program participants. During the course of this work correlations were drawn between FSP material and the material subjected to isothermal hotworking, which may enhance our understanding of the roles that various FSP process parameters have on the microstructural transformation sequence within this material. The hot-rolling study conducted here suggests that FSP process parameters leading to severe deformation at temperatures between 950-1000 C in the NAB material provides high ductility (elongation approximately 28%) with moderate strengths. / Lieutenant, United States Coast Guard

L’allégement des véhicules est l’un des objectifs prioritaires des constructeurs automobile pour répondre aux directives environnementales d’émission de CO2. Le développement des aciers multiphasés à très haute résistance mécanique est l’une des solutions communément adoptées pour réduire l’épaisseur des tôles dans les véhicules tout en conservant leur capacité à assurer la sécurité des passagers. La présente étude porte sur l’optimisation des paramètres du procédé de production industrielle de l’une des ces familles d’aciers : les aciers Dual Phase microalliés au titane et au niobium de haut grade ; c'est-à-dire possédant une résistance à la rupture supérieure à 800MPa.A partir d’une microstructure initiale bainite + martensite laminée à froid, les différents phénomènes se produisant au cours du recuit, de la chauffe jusqu’à la fin du maintien intercritique, sont caractérisés. L’influence des paramètres du recuit comme la vitesse de chauffe, la température et le temps de maintien est étudiée. Le système d’interactions triple entre la précipitation des éléments de microalliage, la recristallisation et la formation de l’austénite est au cœur du problème. Un scénario des évolutions microstructurales a été établi à partir de la caractérisation des divers phénomènes. La finesse de la microstructure étudiée (sub-micrométrique) a nécessité l’emploi combiné de techniques de caractérisation multi-échelles : MEB, MET, sonde atomique tomographique, nano-SIMS. / Lightening the weight of vehicles is one of the main challenging objectives of the automotive industry to reach the environmental regulation in term of CO2 emissions. The development of multiphase high strength steels is a common solution to reduce the thickness of sheet steel used in vehicles while keeping the same level of passenger’s safety requirements. The present study deals with the optimization of industrial process parameters applied to obtain one of these steels: the high strength microalloyed Dual Phase steels; i.e. with ultimate tensile strength superior to 800MPa.From an initial cold rolled microstructure made of bainite + martensite, the phenomena occurring during the annealing are characterized since the heating up to the end of the intercritical holding. The influence of process parameters as the heating rate, the holding temperature and the holding time are studied. The triple interactions system between the precipitation of microalloying elements, the recrystallization and the austenite formation is the core of the problem. A scenario of microstructural evolutions has been established based on the characterized phenomena. The studied fine microstructure (sub-microns) requires the combination of multiscale characterization techniques: SEM, TEM, atom probe tomography, nano-SIMS.

Les aciers multiphasés à très haute résistance mécanique destinés à l’industrie automobile font l’objet d’importantes activités de recherche et développement dans le contexte de l’allègement des structures. L’obtention de meilleurs compromis entre résistance et ductilité nécessite de comprendre l’influence du couple composition chimique nominale – paramètres du procédé d’élaboration sur la formation des microstructures.La présente étude s’inscrit dans cette démarche de compréhension et porte en particulier sur les mécanismes se produisant lors du recuit des nuances d’aciers Dual Phase de haut grade microalliés au Ti et au Nb, dont la microstructure initiale laminée à froid est composée de bainite et de martensite. Ces mécanismes (précipitation des éléments de microalliage Ti, Nb et Mo, recristallisation de la ferrite et formation de l’austénite) présentent des interactions complexes. Le scénario des évolutions microstructurales lors du recuit est caractérisé à l’aide d’une étude expérimentale s’appuyant sur des techniques d’analyses complémentaires à différentes échelles (DRX in situ, MEB, MET, SAT). L’influence respective des éléments de microalliage Ti, Nb et Mo et des paramètres du recuit sur ce scénario est clarifiée à l’aide d’une caractérisation systématique des évolutions microstructurales couplée à la mise en œuvre d’une démarche de modélisation des mécanismes et de leurs interactions. / To meet the need for weight reductions in the automotive industry, new advanced high-strength steels are being developed. The achievement of a better balance between high strength and high formability requires a deep understanding of both the effect of chemical composition and processing parameters on the formation of microstructures. The present work contributes to such an objective and deals with the mechanisms occurring during annealing of Dual Phase steels microalloyed with Ti and Nb. Microstructural changes during this stage include precipitation of microalloying elements, ferrite recrystallization and austenite formation. These mechanisms are investigated using complementary experimental techniques at different scales such as in situ XRD, SEM, TEM and APT in various Dual Phase steel grades having the same bainite-martensite initial cold-rolled microstructure. Using combined experimental and modeling approaches, the present work clarifies the separate influence of microalloying elements Ti, Nb and Mo and heating rate on the mechanisms occurring during annealing and their interactions.

A tomografia por impedância elétrica é uma modalidade de imageamento médico recente, com diversas vantagens sobre as demais modalidades já consolidadas. O recozimento simulado é um algoritmo que apresentada qualidade de solução, mesmo com a utilização de uma regularização simples e sem informação a priori. Entretanto, existe a necessidade de reduzir o tempo de processamento. Este trabalho avança nessa direção, com a apresentação de um método de reconstrução que utiliza o recozimento simulado e paralelização massiva em GPU. A paralelização das operações matriciais em GPU é explicada, com uma estratégia de agendamento de threads que permite a paralelização efetiva de algoritmos, até então, considerados não paralelizáveis. Técnicas para sua aceleração são discutidas, como a heurística de fora para dentro. É proposta uma nova representação de matrizes esparsas voltada para as características da arquitetura CUDA, visando um melhor acesso à memória global do dispositivo e melhor utilização das threads. Esta nova representação de matriz mostrou-se vantajosa em relação aos formatos mais utilizados. Em seguida, a paralelização massiva do problema inverso da TIE, utilizando recozimento simulado, é estudada, com uma proposta de abordagem híbrida com paralelização tanto em CPU quanto GPU. Os resultados obtidos para a paralelização do problema inverso são superiores aos do problema direto. A GPU satura em aproximadamente 7.000 nós, a partir do qual o ganho em desempenho é de aproximadamente 5 vezes. A utilização de GPUs é viável para a reconstrução de imagens de tomografia por impedância elétrica. / Electrical impedance tomography is a new medical imaging modality with remarkable advatanges over other stablished modalities. Simulated annealing is an algorithm that renders quality solutions despite the use of simple regularization methods and the absence of a priori information. However, it remains the need to reduce its processing time. This work takes a step in this direction, presenting a method for the reconstruction of EIT images using simulated annealing and GPU parallelization. The parallelization of matrix operations in GPU is explained, with a thread scheduling strategy that allows the effective parallelization of not-yet effectively parallelized algorithms. There are strategies for improving its performance, such as the presented outside-in heuristic. It is proposed a new sparse matrix representation focused on the CUDA architecture characteristics, with improved global memory access patterns and thread efficiency. This new matrix representation showed several advantages over the most common formats. The massive parallelization of the TIE\'s inverse problem using simulated annealing is studied, with a proposed hybrid approach that uses parallelization in both CPU and GPU. Results showed that the performance gain for the inverse problem is higher than the one obtained for the forward problem. The GPU device saturates with meshes of size of approximately 7,000 nodes, with a performance gain around 5 times faster than serial implementations. GPU parallelization may be used for the reconstruction of electrical impedance tomography images.

Materiais magnetostrictivos, como o TERFENOL-D (Tb0,27-0,30Dy0,73-0,70Fe2) e o GALFENOL (Fe72-82Ga18-28), são aplicados em diversos tipos de atuadores, sensores e coletores de energia. Contudo, existe a necessidade de novos materiais que sejam mais baratos, ambientalmente amigáveis e com melhores propriedades mecânicas. Por essa razão, as ligas de Fe-Al são uma alternativa, já que o alumínio é mais abundante na natureza e é sete vezes mais barato que o gálio. O objetivo desse trabalho foi estudar a influência de dois tratamentos termomecânicos nas propriedades magnéticas das ligas (Fe1-xAlx)98,4B1,6, onde x = 0,18; 0,13 e 0,21. No primeiro processamento, as ligas foram submetidas a um tratamento térmico sob compressão para introduzir uma anisotropia magnetocristalina extrínseca. Para compressões de até 180 MPa, as ligas tratadas sob tensão obtiveram valores de magnetostricção e dos coeficientes piezomagnéticos maiores que o das ligas sem tratamento. Ou seja, este processamento aumentou a sensibilidade das ligas como atuador (d33) e sensor (d33 *). No segundo processamento termomecânico, amostras com formato de chapa foram laminadas e tratadas termicamente para induzir um crescimento anormal de grãos e consequentemente uma textura. Embora não tenha ocorrido crescimento de grão anormal durante o tratamento térmico, os valores de magnetostricção foram maiores que o das amostras como fundidas. Ambos processamentos termomecânicos foram efetivos para melhorar as propriedades magnéticas e a liga (Fe0,87Al0,13)98,4B1,6 foi a que obteve os maiores valores de magnetostricção e dos coeficientes piezomagnéticos. Em outras palavras, dentre as ligas estudadas, esta composição foi mais promissora para ser aplicada em atuadores e sensores. / Magnetostrictive materials, like TERFENOL-D (Tb0.27-0.30Dy0.73-0.70Fe2) and GALFENOL (Fe72-82Ga18-28), are applied in several types of sensors, actuators and energy harvesting. Nevertheless, there is the necessity of new materials that are cheaper, environmentally friendly and with good mechanical properties. For that reason, the Fe-Al alloys are an alternative, since the aluminum is more abundant in nature and is seven times cheaper than gallium. The goal of this work is to study the influence of two thermomechanical proceeding at magnetic properties of the (Fe1- xAlx)98.4B1.6 alloys, which x = 0.18; 0.13 e 0.21. At the first proceeding, the alloys were submitted to a stress annealing to introduce an extrinsic magnetocrystalline anisotropy. Up to 180 MPa of compression stress, stress annealed alloys have magnetostriction and piezomagnetic coefficient values higher than the alloys without annealing. In other words, this proceeding increased the alloys sensibility to be applied like actuators (d33) and sensors (d33 *). At the second thermomechanical proceeding, samples with sheet format were rolled and annealed to induce an abnormal grain growth and, consequently, a texture. Although the annealing not produced an abnormal grain growth, the magnetostriction values are higher than for the as-cast samples. Both proceedings were efficient to improve the magnetic properties and the (Fe0.87Al0.13)98.4B1.6 alloy achieved the higher values of magnetostriction and piezomagnetic coefficient. In another words, between the studied alloys, this composition was the most promising to be applied like actuators and sensors.

Aujourd'hui l'utilisation de l'échographie 3D en cardiologie est limitée car l'imagerie de la totalité du myocarde sur un cycle cardiaque, sans apnée, reste un défi technologique. Une solution consiste à réduire le nombre de capteurs dans les sondes échographiques matricielles afin d'alléger le procédé d'acquisition: ces sondes sont dites parcimonieuses. Le but de cette thèse est de proposer les meilleures dispositions d'un nombre réduit de capteurs piézo-électriques répartis sur la surface active de la sonde afin d'optimiser leur capacité à produire des images homogènes en termes de contraste et résolution dans tout le volume d'intérêt. Ce travail présente l'intégration de simulations acoustiques réalistes élaborées au sein d'un processus d'optimisation stochastique (algorithme de recuit simulé). La structure proposée pour le design des sondes parcimonieuse est suffisamment générale pour être appliquée aux sondes régulières (éléments actifs disposés selon une grille) et non-régulières (positionnement arbitraire des éléments actifs). L'introduction d'une fonction d'énergie innovante permet de sculpter en 3D le diagramme optimal de rayonnement de la sonde. Les résultats de sondes optimisées obtenues possèdent 128, 192 ou 256 éléments pour favoriser leur compatibilité avec les échographes commercialisés à ce jour, ce qui permettrait de déployer l'échographie 3D à moindre coût et à très large échelle / Today, the use of 3D ultrasound imaging in cardiology is limited because imaging the entire myocardium on a single heartbeat, without apnea, remains a technological challenge. A solution consists in reducing the number of active elements in the 2D ultrasound probes to lighten the acquisition process: this approach leads to sparse arrays. The aim of this thesis is to propose the best configuration of a given number of active elements distributed on the probe active surface in order to maximize their ability to produce images with homogeneous contrast and resolution over the entire volume of interest. This work presents the integration of realistic acoustic simulations performed in a stochastic optimization process (simulated annealing algorithm). The proposed sparse array design framework is general enough to be applied on both on-grid (active elements located on a regular grid) and non-grid (arbitrary positioning of the active elements) arrays. The introduction of an innovative energy function sculpts the optimal 3D beam pattern radiated by the array. The obtained optimized results have 128, 192 or 256 active elements to help their compatibility with currently commercialized ultrasound scanners, potentially allowing a large scale development of 3D ultrasound imaging with low cost systems

O resíduo estudado neste trabalho é originado do processo de obtenção de alumínio metálico, de grande interesse no Brasil pelo fato do país ser detentor de algumas das maiores reservas do mineral bauxita no mundo, utilizado como fonte de alumínio. Tendo como estratégia a geração de resíduo zero, colaborando para as tecnologias ambientalmente amigáveis, este trabalho estuda a incorporação de um resíduo resultante da recuperação de alumínio presente na escória gerada durante o processo de produção primária do alumínio metálico, por plasma térmico. Utilizando-se o diagrama de equilíbrio de fases do sistema Al2O3-CaO-SiO2, fez-se a adequação das composições visando a incorporação de resíduo no produto cerâmico sem alterar as características de processamento do material. A obtenção de vidros e de fritas vitrocerâmicas com o resíduo borra branca foi realizada fazendo-se a fusão das composições calculadas e, para os vitrocerâmicos, tratamento térmico posterior de devitrificação. Os produtos obtidos foram caracterizados utilizando-se técnicas de análise tais como difração de raios X (DRX), microscopia eletrônica de varredura (MEV) e espectroscopia do infravermelho (FTIR). Foi possível obter material vitrocerâmico com até 30% de resíduo de alumínio, após a fusão a 1300°C e devitrificação a 900°C. Em adição, o resíduo demonstrou ser um promissor material auxiliar na formação de fases cristalinas em baixos tempos de tratamento térmico. / This work studies the residue obtained from the process of aluminum metal extraction activities, a great interest process, because of Brazil own some of the biggest bauxite mineral reserves in all the world. As a useful choice for no residue generation, and a support for environmentally friendly technologies, this work studies the white dross residue (WDR), from the process of aluminum metal reduction by thermal plasma. The phase equilibrium diagram of Al2O3-Ca O-SiO2 system was used to calculate the compositions. The WDR were incorporated in a ceramic product without modifying its principal characteristics. The fusion and devitrification treatments were studied. XRD (X-ray diffractometry), SEM (scanning electron microscopy) and FTIR (transformed Fourier infrared) were used to investigate the glass and glassceramic samples. These techniques showed that is possible to get glassceramic with up to 30 mass% of WDR after molten at 1300 deg C and annealed at 900 deg C. In addition, the WDR showed to be a promising material in attainment of crystalline phases in less times of heat treatment for annealing.