abstract: A moving overlapping mesh methodology that achieves spectral accuracy in space and up to second-order accuracy in time is developed for solution of unsteady incompressible flow equations in three-dimensional domains. The targeted applications are in aerospace and mechanical engineering domains and involve problems in turbomachinery, rotary aircrafts, wind turbines and others. The methodology is built within the dual-session communication framework initially developed for stationary overlapping meshes. The methodology employs semi-implicit spectral element discretization of equations in each subdomain and explicit treatment of subdomain interfaces with spectrally-accurate spatial interpolation and high-order accurate temporal extrapolation, and requires few, if any, iterations, yet maintains the global accuracy and stability of the underlying flow solver. Mesh movement is enabled through the Arbitrary Lagrangian-Eulerian formulation of the governing equations, which allows for prescription of arbitrary velocity values at discrete mesh points. The stationary and moving overlapping mesh methodologies are thoroughly validated using two- and three-dimensional benchmark problems in laminar and turbulent flows. The spatial and temporal global convergence, for both methods, is documented and is in agreement with the nominal order of accuracy of the underlying solver. Stationary overlapping mesh methodology was validated to assess the influence of long integration times and inflow-outflow global boundary conditions on the performance. In a turbulent benchmark of fully-developed turbulent pipe flow, the turbulent statistics are validated against the available data. Moving overlapping mesh simulations are validated on the problems of two-dimensional oscillating cylinder and a three-dimensional rotating sphere. The aerodynamic forces acting on these moving rigid bodies are determined, and all results are compared with published data. Scaling tests, with both methodologies, show near linear strong scaling, even for moderately large processor counts. The moving overlapping mesh methodology is utilized to investigate the effect of an upstream turbulent wake on a three-dimensional oscillating NACA0012 extruded airfoil. A direct numerical simulation (DNS) at Reynolds Number 44,000 is performed for steady inflow incident upon the airfoil oscillating between angle of attack 5.6 and 25 degrees with reduced frequency k=0.16. Results are contrasted with subsequent DNS of the same oscillating airfoil in a turbulent wake generated by a stationary upstream cylinder. / Dissertation/Thesis / Doctoral Dissertation Aerospace Engineering 2016

Il y a une décennie, le contenu numérique virtuel était limité à quelques applications – majoritairementles jeux vidéos, les films en 3D et la simulation numérique. Aujourd’hui, grâce à l’apparition de cartes graphiques performantes et bon marché, les objets 3D sont utilisés dans de nombreuses applications. A peu près tous les terminaux possédant des capacités d’affichage – des clusters de visualisation haute performance jusqu’aux smart phones – intègrent maintenant une puce graphique qui leur permet de faire du rendu 3D. Ainsi, les applications 3D sont bien plus variées qu’il y a quelques années. On citera par exemple la réalité virtuelle et augmentée en temps réel ou les mondes virtuels 3D. Dans ce contexte, le besoin de méthodes efficaces pour la transmission et la visualisation des données 3D est toujours plus pressant. De plus, la taille des maillages 3D ne cesse de s’accroître avec la précision de la représentation. Par exemple, les scanners 3D actuels sont capables de numériser des objets du monde réel avec une précision de seulement quelques micromètres, et génèrent des maillages contenant plusieurs centaines de millions d’´el´ements. D’un autre côté, une précision accrue en simulation numérique requiert des maillages plus fins, et les méthodes massivement parallèles actuelles sont capables de travailler avec des milliards de mailles. Dans ce contexte, la compression de ces données – en particulier la compression de maillages – est un enjeu important. Durant la décennie passée, de nombreuses méthodes ont été développées pour coder les maillages polygonaux. Néanmoins, ces techniques ne sont plus adaptées au contexte actuel, car elles supposentque la compression et la d´ecompression sont des processus sym´etriques qui ont lieu sur un mat´erielsimilaire. Dans le cadre actuel, au contraire, le contenu 3D se trouve cr´e´e, compressé et distribué par des machines de hautes performances, tandis que l’exploitation des données – par exemple, la visualisation – est effectuée à distance sur des périphériques de capacité plus modeste – éventuellement mobiles – qui ne peuvent traiter les maillages de grande taille dans leur int´egralité. Ceci fait de lacompression de maillage un processus intrinsèquement asymétrique.Dans cette thèse, notre objectif est d’étudier et de proposer des méthodes pour la compression de maillages de grande taille. Nous nous intéressons plus particulièrement aux méthodes d’accès aléatoire, qui voient la compression comme un problème intrinsèquement asymétrique. Dans ce modèle, le codeur a accès à des ressources informatiques importantes, tandis que la décompression estun processus temps réel (souple) qui se fait avec du matériel de plus faible puissance. Nous décrivons un algorithme de ce type et l’appliquons au cas de la visualisation interactive. Nous proposons aussi un algorithme streaming pour compresser des maillages hexaèdriques de très grande taille utilisés dans le contexte de la simulation numérique. Nous sommes ainsi capables decompresser des maillages comportant de l’ordre de 50 millions de mailles en moins de deux minutes, et en n’utilisant que quelques mégaoctets de mémoire vive. Enfin, nous proposons, indépendamment de ces deux algorithmes, un cadre théorique général pour améliorer la compression de géométrie. Cet algorithme peut être utilisé pour développer des méthodes de prédiction pour n’importe quel algorithme basé sur un paradigme prédictif – ce qui est la cas dela majorité des méthodes existantes. Nous dérivons ainsi des schémas de prédictions compatibles avec plusieurs méthodes de la littérature. Ces schémas augmentent les taux de compression de 9% enmoyenne. Sous des hypothèses usuelles, nous utilisons aussi ces résultats pour prouver l’optimalité de certains algorithmes existants. / A decade ago, 3D content was restricted to a few applications – mainly games, 3D graphics andscientific simulations. Nowadays, thanks to the development cheap and efficient specialized renderingdevices, 3D objects are ubiquitous. Virtually all devices with a display – from a large visualizationclusters to smart phones – now integrate 3D rendering capabilities. Therefore, 3D applications arenow far more diverse than a few years ago, and include for example real-time virtual and augmentedreality, as well as 3D virtual worlds. In this context, there is an ever increasing need for efficient toolsto transmit and visualize 3D content.In addition, the size of 3D meshes always increases with accuracy of representation. On one hand,recent 3D scanners are able to digitalize real-world objects with a precision of a few micrometers, andgenerate meshes with several hundred million elements. On the other hand, numerical simulationsalways require finer meshes for better accuracy, and massively parallel simulation methods now generatemeshes with billions of elements. In this context, 3D data compression – in particular 3D meshcompression – services are of strategic importance.The previous decade has seen the development of many efficient methods for encoding polygonalmeshes. However, these techniques are no longer adapted to the current context, because they supposethat encoding and decoding are symmetric processes that take place on the same kind of hardware.In contrast, remote 3D content will typically be created, compressed and served by high-performancemachines, while exploitation (e.g. visualization) will be carried out remotely on smaller – possiblyhand held – devices that cannot handle large meshes as a whole. This makes mesh compression anintrinsically asymmetric process.Our objective in this dissertation is to address the compression of these large meshes. In particularwe study random-accessible compression schemes, that consider mesh compression as an asymmetricproblem where the compressor is an off-line process and has access to a large amount of resources,while decompression is a time-critical process with limited resources. We design such a compressionscheme and apply it to interactive visualization.In addition, we propose a streaming compression algorithm that targets the very large hexahedralmeshes that are common in the context of scientific numerical simulation. Using this scheme, we areable to compress meshes of 50 million hexahedra in less than two minutes using a few megabytes ofmemory.Independently from these two specific algorithms, we develop a generic theoretical framework toaddress mesh geometry compression. This framework can be used to derive geometry compressionschemes for any mesh compression algorithm based on a predictive paradigm – which is the case of thelarge majority of compression schemes. Using this framework, we derive new geometry compressionschemes that are compatible with existing mesh compression algorithms but improve compressionratios – by approximately 9% on average. We also prove the optimality of some other schemes underusual smoothness assumptions.

Estruturas de dados (ED) topológicas oferecem diversas vantagens quando se deseja executarumadeformação sobreumamalha. Essas EDs permitem movimentar os nós da malha sem modificar sua topologia, são relativamente simples de seremimplementadas e também são passíveis de serem incorporadas a um ciclo simulação/deformação de forma completamente automática e eficiente. O primeiro objetivo deste trabalho é a concepção de uma ED topológica para representação de malhas elásticas. Tais malhas podem ser do tipo superficial ou volumétrica, e ainda simples ou mista. Para melhor desempenho, confiabilidade e menor consumo de memória, deseja-se que a ED seja implícita quanto à representação de componentes incidentes e adjacentes dos elementos presentes na malha. Outro objetivo deste trabalho é abordar o problema de geração de malhas em domínios arbitrários definidos por uma função implícita. O método proposto é uma extensão do algoritmo de Partição da Unidade Implícita (PUI). Para isso, o método proposto é baseado numa abordagem de preenchimento de superfícies. Este método proposto gera adaptativamente tetraedros em diferentes níveis de refinamento de acordo com o nível de detalhe presente na região do domínio. Diferentemente de trabalhos anteriores, esta característica é feita naturalmente sem necessitar de uma estrutura auxiliar. Para este fim, usa-se uma estrutura algébrica chamada de triangulação Ja1 que é capaz de lidar com tais refinamentos. Além do mais, a triangulação Ja1 permite que se percorra a estrutura simplesmente através de regras algébricas que é uma outra vantagem do método proposto / Topological data Structures (DS) oer several advantages when performing a deformation on a mesh. These DSs allow movement throughout the mesh without modifying its topology, are relatively simple to implement, and there is always the possibility of merging it to simulation/deformation cycle on a completely automatic and ecient form. The main goal of this work is to design and implement a topological DS to represent elastic meshes. These meshes can be either of surface or volume kind, and even simple or mixed. For better performance, more reliability and lower memory consumption, it is necessary that the DSs allow the representation of incident and adjacent components of a given element to be implicit. The second objective to this work is to tackle the problem of mesh generation on arbitrary domains defined by implicit functions. The proposed method is an extension to the algorithm of Partition of Unity Implicits (PUI). For this the proposed method is based on an isosurface stung approach. It adaptively generates the tetrahedra in dierent levels of refinement accordingly to the level of detail presented by the regions of the domain. Dierently to previous work, this feature is done naturally without the aid of an auxiliary data structure. To this end, we use an algebraic structure, named Ja1 triangulation, which is capable of dealing with such refinements. In addition, the Ja1 triangulation permits traversing through the mesh by simply using algebraic rules which is another advantage to the proposed method

Simulações de escoamentos multifásicos são de grande interesse em aplicações práticas na indústria, em particular na indústria petrolífera, entre outras. Vários processos dependem do entendimento físico de escoamentos envolvendo iteração com partículas, sedimentação e separação de fluidos. Dos muitos métodos existentes para a simulação dos processos acima descritos, há um crescente interesse no aumento de precisão, o que levou ao desenvolvimento de estratégias que utilizam esquemas de elementos finitos discretizados em malhas dinâmicas e adaptativas, usando uma formulação ALE (do inglês, Arbitrary Lagrangian-Eulerian), juntamente com uma representação geométrica da interface. Neste sentido, este trabalho tem o objetivo de estudar e implementar estratégias robustas de controle e adaptação de malhas, em situações onde a malha dinâmica é sujeita a grandes deformações. Uma biblioteca de algoritmos e rotinas foi então desenvolvida para este fim, implementando técnicas de controle e otimização da qualidade dos elementos da malha, técnicas de adaptação da interface entre fluidos com esquemas de conservação de massa, técnicas de mudanças topológicas e preservação de propriedades materiais, além de uma comunicação facilitada destas rotinas com códigos de simulação numérica de escoamentos multifásicos existentes / Multiphase flow simulations are of great interest in practical applications, particularly in the oil industry. Several processes depend on understanding physical aspects of flows with particle interaction, sedimentation and fluid separation. Among the several existing methods to simulate the processes described above, theres a growing interest in achieving higher precision, which led to the development of strategies that use finite element discretization in adaptive, dynamic meshes, using the ALE formatulation along with a geometrical representation of the interface. In this context, this thesis aims to study and implement robust strategies for mesh adaptation, for cases where the dynamic mesh is subject to large deformations. A library of routines and algorithms was developed, implementing mesh elements control and quality optimization techniques, fluid interface adaptation techniques with a mass conservation scheme, topological modifications and material properties preservation techniques, and also a decoupled, simplified communication between these routines with existing multiphase flow numerical simulation code

The paper is concerned with the Nitsche mortaring in the framework of domain decomposition where non-matching meshes and weak continuity of the finite element approximation at the interface are admitted. The approach is applied to singularly perturbed reaction-diffusion problems in 2D. Non-matching meshes of triangles being anisotropic in the boundary layers are applied. Some properties as well as error estimates of the Nitsche mortar finite element schemes are proved. In particular, using a suitable degree of anisotropy of triangles in the boundary layers of a rectangle, we derive convergence rates as known for the conforming finite element method in presence of regular solutions. Numerical examples illustrate the approach and the results.

L'imagerie électrique est un outil de plus en plus important pour un large domaine d'applications relatives à la caractérisation de la subsurface proche. D'importants développements ont été réalisés au cours des vingt dernières années pour l'amélioration des systèmes d'acquisitions et des algorithmes d'inversions. L'acquisition et le traitement de gros jeux de données reste toutefois une tâche délicate, en particulier en présence de topographie. Afin d'améliorer la gestion de la topographie, nous avons développé un nouvel algorithme d'inversion électrique 2.5D et 3D. Nous avons proposé deux nouvelles formulations pour supprimer la singularité à la source. Le problème direct est résolu en utilisant la méthode des Différences Finies Généralisées et des maillages non structurés, permettant une représentation précise de la topographie. Le code d'inversion utilise la méthode de l'état adjoint pour calculer le gradient de la fonction objective de manière économique. Cette approche a donné de bons résultats avec des données synthétiques. Les premiers résultats sur des données réelles ont permis de retrouver les principales structures de la subsurface, ainsi que plusieurs zones de faibles résistivités pouvant correspondre à des zones fracturées. / DC resistivity imaging plays an important role for a wide range of applications related to the characterization of the shallow subsurface. Major developments have been made over the last two decades to improve acquisition systems as well as resistivity inversion. Nevertheless, large-scale data sets still represent a challenging task, in particular with arbitrary topography. In order to better take into account topography, we have developed a new 2.5D/3D tomographic inversion code. Two new formulations for the singularity removal have been proposed. The direct problem is solved in the framework of the Generalized Finite Differences Method, that allows the use of unstructured meshes yielding a fine represention of topography. The inversion code uses the adjoint state method to compute the gradient of the misfit function in a numerically efficient way, giving goodresults on synthetic data. First results on real data have shown the main subsurface structures, as well as several low resistivity zones possibly corresponding to fractured areas.

[pt] Este trabalho trata da solução numérica das equações de Navier-Stokes, na forma vorticidade-função corrente, via método das Diferenças Finitas e técnicas de aceleração baseadas no uso de malhas múltiplas. Embora outras opções tenham sido consideradas, a que melhor funcionou tratou o problema de forma não acoplada: a solução da equação de vorticidade foi obtida pela uso desta aceleração e a solução da equação de função corrente, uma equação puramente elíptica, foi resolvida via método das relaxações sucessivas. O código desenvolvido foi aplicado a diversos problemas, inclusive ao problema da cavidade com tampa móvel, em diversos números de Reynolds, típico no teste de simuladores em Dinâmica dos Fluidos. Foram testados um método clássico (armazenamento da correção) e o método FAZ (Full Approximation Storage). Os resultados obtidos mostram claramente os ganhos computacionais obtidos na formulação escolhida. Expressando em percentual, valores com 80 por cento de ganho foram obtidos se comparados os resultados do método multigrid com o método iterativo básico utilizado (S.O.R.), indicando o potencial do uso desta técnica para problemas mais complexo incluindo aqueles em coordenadas generalizadas. / [en] This works deals with the numerical solution of the Navier-Stokes equations, written in the stream function-vorticity form, by the finite difference method and acceleration techniques using multiple meshes. Although other solution schemes have been investigated, best results were obtained by treating the problem in a non-coupled form: the solution for the vorticity equation was obtained by the multigrid method and the solution of the streamfunction equation, which is purely elliptic, was solved by the S.O.R. (Successive over relaxation method). The computer code was applied to several problems, including the wall driven problem considering a wide range of Reynolds numbers, which is a typical benchmark problem for testing fluid-dynamic simulations. The classical method (storage of the correction) and the methos FAS (Full Approximation Storage) have been tested. The results obtained clearly show that a very efficient computational scheme has been achieved with the multigrid method. For example, when comparing this method with the basic S.O.R. method, relative gains in the order of 80 per cent have been obtained. This indicates that the present technique has potential use in more complicated fluid dynamics problems including those involving generalized coordinates.

Les opérations chirurgicales pour le traitement des défauts de soutènement d'organes dans les pays industrialisés sont des actes de plus en plus fréquents, et requièrent l'implantation de plus d'un million de prothèses par an. Même si l'usage des prothèses de soutènement a permis de diminuer les complications postopératoires, les taux de réinterventions restent très élevés. Ces travaux présentent différentes stratégies permettant d'apporter de nouvelles propriétés à des treillis en polypropylène, afin d'améliorer leur résistance à l'infection et de permettre leur suivi postopératoire en IRM. Des treillis anti-infectieux sont développés en enrobant les filaments de treillis d'un réservoir de polymère dégradable et biocompatible contenant des agents anti-infectieux, par une technique de pulvérisation à l'aide d'un aérographe. L'association ofloxacine – rifampicine incorporée présente une excellente activité antibactérienne in vitro, et la cinétique de libération prolongée des agents actifs permet d'inhiber la contamination des treillis pendant au moins 72 heures. Les techniques de stérilisation par rayonnement Gamma n'impactent ni la stabilité des agents actifs, ni l'efficacité antibactérienne des treillis anti-infectieux.Afin de permettre une visualisation en IRM des treillis implantés, des polymères ont été synthétisés par greffage d'agent de contraste (DTPA-Gd) sur de la poly(ε-caprolactone) (dégradable) et du poly(acrylate de méthyle) (biostable). Après enrobage des prothèses par ces polymères, le rehaussement du signal induit par la présence du gadolinium permet de visualiser les prothèses à la fois in vitro et in vivo sur différents types d'appareils d'IRM. La stabilité de l'agent de contraste est suffisante pour pouvoir visualiser les treillis pendant plusieurs mois, quelle que soit la technique de stérilisation utilisée. / Soft tissue reinforcement surgical operations in industrial countries are common and require annually at least one million of prostheses for treating this problem. Even if meshes used for organ prolapse surgical procedures allow decreasing postoperatory complications, reinterventions ratio is still relevant. This work present several strategies to bring new properties to polypropylene meshes in order to improve their resistance to infection and enable their postoperative MRI follow-up. Anti-infective meshes are developed by coating the filaments of the meshes with an antibiotics drug reservoir based on degradable and biocompatible polymers, using an airbrush system. Dual ofloxacin-rifampicin antibiotics incorporation on these coated meshes shows an excellent antibacterial activity in vitro and sustained release of the drugs can inhibit meshes contamination for at least 72 hours. Sterilization procedures using Gamma-ray irradiation impact neither the drugs stability nor the anti-infective meshes activity. In order to visualize by Magnetic Resonance Imaging (MRI) meshes after implantation, a contrast agent (DTPA-Gd) is covalently grafted onto the polymeric backbone of poly(ε-caprolactone) (degradable) or poly(methyl acrylate) (biostable). Meshes were coated with these new polymers and MR signal enhancement induced by the presence of gadolinium allows the visualization of the meshes in vitro and in vivo with several MR equipments. Coated meshes are visible during several months, whatever the sterilization procedures, showing the stability of the contrast agent.

Orientador: Wu Shin-Ting / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-26T15:53:26Z (GMT). No. of bitstreams: 1 Loos_WallaceSouza_M.pdf: 21840648 bytes, checksum: 8bb81a3fcf4de4d44de658f543435dda (MD5) Previous issue date: 2014 / Resumo: A reformatação curvilínea é uma técnica computacional de exploração, portanto, não invasiva, que permite realizar cortes curvilíneos sobre as neuroimagens. Ela complementa as reformatações multiplanares na localização de displasia cortical focal, causa comumente associada a epilepsia refratária. Foi desenvolvido um algoritmo de reformatação curvilínea baseado em malhas de offset pelo nosso grupo de pesquisa. Dois problemas foram identificados no algoritmo: artefatos visuais e limitação da região a ser reformatada em áreas visíveis. Neste trabalho apresentamos soluções para contornar estes problemas. Mostramos que a causa dos artefatos são as auto-interseções locais e que um algoritmo de simplificação pode removê-las de forma eficiente sem comprometer a geometria de reformatação. Desenvolvemos um algoritmo de costura para juntar malhas obtidas em diferentes ângulos de vista numa única malha de corte. Desta forma, é possível realizar a reformatação simultânea nos dois hemisférios cerebrais propiciando um diagnóstico baseado em análise comparativa dos dois lados. Na implementação dos nossos algoritmos procuramos explorar as vantagens de uma estrutura de dados eficiente e do paralelismo das unidades de processamento gráfico (GPUs) para termos resultados em tempo interativo. Experimentos preliminares com as neuroimagens dos pacientes com crises epilépticas apontam que a ferramenta pode colaborar na identificação de lesões sutis / Abstract: Curvilinear Reformatting is a computational exploration technique, therefore noninvasive, that allows making curvilinear slices on 3D neuroimaging data. It complements multiplanar reformatting to find lesions of focal cortical dysplasia, a common cause of refractory epilepsy. Our research group has developed an algorithm for curvilinear reformatting based on offset meshes. Two problems were identified: visual artifacts and limitations of the visible area to be reformatted. In this work we presented two solutions to solve these problems. We show that the cause of the artifacts are the local self-intersections and that an simplification algorithm can remove them efficiently without compromising the geometry of reformatting. We developed an sewing algorithm to join meshes from different angles of view on a single mesh. In this way, it is possible to make the curvilinear reformatting in both hemispheres of the brain, providing a diagnosis based on comparative analysis of the two sides. In the implementation of our algorithms we tried to explore the advantages of efficient data structure and parallelism of graphic processor units (GPU) to achieve results at interactive rates. Preliminary results with neuroimages from patients with epileptic seizures indicate that the tool may collaborate in the finding of subtle lesions / Mestrado / Engenharia de Computação / Mestre em Engenharia Elétrica

Dois métodos distintos são descritos e implementados. O primeiro método, proposto por Ruppert, possui garantias teóricas de qualidade quando a fronteira do domínio obedece certas restrições. O segundo método, proposto por Persson, possibilita um maior controle na densidade dos elementos que discretizam o domínio. As vantagens, desvantagens e particularidades de cada um dos métodos são descritas e detalhadas / Two distinct methods are described and implemented. The first method, proposed by Ruppert, has theoretical guarantees on the quality of elements when the domain boundaries respect certain restrictions. The second method, proposed by Persson, makes it possible to have greater control over the density of the elements that make up the domain. The advantages, disadvantages and specific points about each method are described and detailed