A multi-domain homogenization method is proposed and developed in this thesis based on a two-scale technique. The method is capable of analyzing composite structures with several periodic distributions by partitioning the entire domain of the composite into substructures making use of the classical homogenization theory following a first-order standard continuum mechanics formulation. The need to develop the multi-domain homogenization method arose because current homogenization methods are based on the assumption that the entire domain of the composite is represented by one periodic or quasi-periodic distribution. However, in some cases the structure or composite may be formed by more than one type of periodic domain distribution, making the existing homogenization techniques not suitable to analyze this type of cases in which more than one recurrent configuration appears. The theoretical principles used in the multi-domain homogenization method were applied to assemble a computational tool based on two nested boundary value problems represented by a finite element code in two scales: a) one global scale, which treats the composite as an homogeneous material and deals with the boundary conditions, the loads applied and the different periodic (or quasi-periodic) subdomains that may exist in the composite; and b) one local scale, which obtains the homogenized response of the representative volume element or unit cell, that deals with the geometry distribution and with the material properties of the constituents. The method is based on the local periodicity hypothesis arising from the periodicity of the internal structure of the composite. The numerical implementation of the restrictions on the displacements and forces corresponding to the degrees of freedom of the domain's boundary derived from the periodicity was performed by means of the Lagrange multipliers method. The formulation included a method to compute the homogenized non-linear tangent constitutive tensor once the threshold of nonlinearity of any of the unit cells has been surpassed. The procedure is based in performing a numerical derivation applying a perturbation technique. The tangent constitutive tensor is computed for each load increment and for each iteration of the analysis once the structure has entered in the non-linear range. The perturbation method was applied at the global and local scales in order to analyze the performance of the method at both scales. A simple average method of the constitutive tensors of the elements of the cell was also explored for comparison purposes. A parallelization process was implemented on the multi-domain homogenization method in order to speed-up the computational process due to the huge computational cost that the nested incremental-iterative solution embraces. The effect of softening in two-scale homogenization was investigated following a smeared cracked approach. Mesh objectivity was discussed first within the classical one-scale FE formulation and then the concepts exposed were extrapolated into the two-scale homogenization framework. The importance of the element characteristic length in a multi-scale analysis was highlighted in the computation of the specific dissipated energy when strain-softening occurs. Various examples were presented to evaluate and explore the capabilities of the computational approach developed in this research. Several aspects were studied, such as analyzing different composite arrangements that include different types of materials, composites that present softening after the yield point is reached (e.g. damage and plasticity) and composites with zones that present high strain gradients. The examples were carried out in composites with one and with several periodic domains using different unit cell configurations. The examples are compared to benchmark solutions obtained with the classical one-scale FE method. / En esta tesis se propone y desarrolla un método de homogeneización multi-dominio basado en una técnica en dos escalas. El método es capaz de analizar estructuras de materiales compuestos con varias distribuciones periódicas dentro de un mismo continuo mediante la partición de todo el dominio del material compuesto en subestructuras utilizando la teoría clásica de homogeneización a través de una formulación estándar de mecánica de medios continuos de primer orden. La necesidad de desarrollar este método multi-dominio surgió porque los métodos actuales de homogeneización se basan en el supuesto de que todo el dominio del material está representado por solo una distribución periódica o cuasi-periódica. Sin embargo, en algunos casos, la estructura puede estar formada por más de un tipo de distribución de dominio periódico. Los principios teóricos desarrollados en el método de homogeneización multi-dominio se aplicaron para ensamblar una herramienta computacional basada en dos problemas de valores de contorno anidados, los cuales son representados por un código de elementos finitos (FE) en dos escalas: a) una escala global, que trata el material compuesto como un material homogéneo. Esta escala se ocupa de las condiciones de contorno, las cargas aplicadas y los diferentes subdominios periódicos (o cuasi-periódicos) que puedan existir en el material compuesto; y b) una escala local, que obtiene la respuesta homogenizada de un volumen representativo o celda unitaria. Esta escala se ocupa de la geometría, y de la distribución espacial de los constituyentes del compuesto así como de sus propiedades constitutivas. El método se basa en la hipótesis de periodicidad local derivada de la periodicidad de la estructura interna del material. La implementación numérica de las restricciones de los desplazamientos y las fuerzas derivadas de la periodicidad se realizaron por medio del método de multiplicadores de Lagrange. La formulación incluye un método para calcular el tensor constitutivo tangente no-lineal homogeneizado una vez que el umbral de la no-linealidad de cualquiera de las celdas unitarias ha sido superado. El procedimiento se basa en llevar a cabo una derivación numérica aplicando una técnica de perturbación. El tensor constitutivo tangente se calcula para cada incremento de carga y para cada iteración del análisis una vez que la estructura ha entrado en el rango no-lineal. El método de perturbación se aplicó tanto en la escala global como en la local con el fin de analizar la efectividad del método en ambas escalas. Se lleva a cabo un proceso de paralelización en el método con el fin de acelerar el proceso de cómputo debido al enorme coste computacional que requiere la solución iterativa incremental anidada. Se investiga el efecto de ablandamiento por deformación en el material usando el método de homogeneización en dos escalas a través de un enfoque de fractura discreta. Se estudió la objetividad en el mallado dentro de la formulación clásica de FE en una escala y luego los conceptos expuestos se extrapolaron en el marco de la homogeneización de dos escalas. Se enfatiza la importancia de la longitud característica del elemento en un análisis multi-escala en el cálculo de la energía específica disipada cuando se produce el efecto de ablandamiento. Se presentan varios ejemplos para evaluar la propuesta computacional desarrollada en esta investigación. Se estudiaron diferentes configuraciones de compuestos que incluyen diferentes tipos de materiales, así como compuestos que presentan ablandamiento después de que el punto de fluencia del material se alcanza (usando daño y plasticidad) y compuestos con zonas que presentan altos gradientes de deformación. Los ejemplos se llevaron a cabo en materiales compuestos con uno y con varios dominios periódicos utilizando diferentes configuraciones de células unitarias. Los ejemplos se comparan con soluciones de referencia obtenidas con el método clásico de elementos finitos en una escala.

<p>In this work, we proposed a spectral integral method (SIM)-spectral element method (SEM)- finite element method (FEM) domain decomposition method (DDM) for solving inhomogeneous multi-scale problems. The proposed SIM-SEM-FEM domain decomposition algorithm can efficiently handle problems with multi-scale structures, </p><p>by using FEM to model electrically small sub-domains and using SEM to model electrically large and smooth sub-domains. The SIM is utilized as an efficient boundary condition. This combination can reduce the total number of elements used in solving multi-scale problems, thus it is more efficient than conventional FEM or conventional FEM domain decomposition method. Another merit of the proposed method is that it is capable of handling arbitrary non-conforming elements. Both geometry modeling and mesh generation are totally independent for different sub-domains, thus the geometry modeling and mesh generation are highly flexible for the proposed SEM-FEM domain decomposition method. As a result, the proposed SIM-SEM-FEM DDM algorithm is very suitable for solving inhomogeneous multi-scale problems.</p> / Dissertation

The high-density integrated circuit (IC) gives rise to geometrically complex multiscale chip-package structures whose electromagnetic performance is difficult to predict. This motivates this dissertation to work on an efficient full-wave transient solver that is capable of capturing all the electromagnetic behaviors of such structures with high accuracy and reduced computational complexity compared to the existing methods. In this work, the unconditionally stable Laguerre-FDTD method is adopted as the core algorithm for the transient full-wave solver. As part of this research, skin-effect is rigorously incorporated into the solver which avoids dense meshing inside conductor structures and significantly increases computational efficiency. Moreover, as an alternative to typical planar interconnects for next generation high-speed ICs, substrate integrated waveguide, is investigated. Conductor surface roughness is efficiently modeled to accurately capture its high-frequency loss behavior. To further improve the computational performance of chip-package co-simulation, a novel transient non-conformal domain decomposition method has been proposed. Large-scale chip-package structure can be efficiently simulated by decomposing the computational domain into subdomains with independent meshing strategy. Numerical results demonstrate the capability, accuracy and efficiency of the proposed methods.

Simulation of the system behavior of hydrate-bearing geologic media involves solving fully coupled mass- and heat-balance equations. In this study, we develop a domain decomposition approach for large-scale gas hydrate simulations with coarse-granularity parallel computation. This approach partitions a simulation domain into small subdomains. The full model domain, consisting of discrete subdomains, is still simulated simultaneously by using multiple processes/processors. Each processor is dedicated to following tasks of the partitioned subdomain: updating thermophysical properties, assembling mass- and energy-balance equations, solving linear equation systems, and performing various other local computations. The linearized equation systems are solved in parallel with a parallel linear solver, using an efficient interprocess communication scheme. This new domain decomposition approach has been implemented into the TOUGH+HYDRATE code and has demonstrated excellent speedup and good scalability. In this paper, we will demonstrate applications for the new approach in simulating field-scale models for gas production from gas-hydrate deposits.

In dieser Arbeit entwickeln wir einen Löser für elliptische Randwertprobleme. Dazu diskretisieren wir ein Randwertproblem mittels der Methode der finiten Elemente und erhalten ein Gleichungssystem. Mittels Gebietszerlegungstechniken unterteilen wir das Gebiet der Differentialgleichung und können Teilprobleme des Randwertproblems definieren. Durch die Gebietszerlegung wird eine Hierarchie von Zerlegungen definiert, die wir mittels eines Gebietszerlegungsbaumes festhalten. Anhand dieses Baumes definieren wir nun einen Löser für das Randwertproblem. Dabei berechnen wir die verschiedenen Matrizen des Lösers durch den sogenannten HDD-Algorithmus (engl. hierarchical domain decomposition). Die meisten der zu erstellenden Matrizen sind dabei vollbesetzt, weshalb wir sie mittels Hierarchischer Matrizen approximieren. Mit Hilfe der Hierarchischen Matrizen können wir die Matrizen mit einem fast linearen Aufwand erstellen und speichern. Der Aufwand der Matrixoperationen ist ebenfalls fast linear. Damit wir die Hierarchischen Matrizen für den HDD-Algorithmus verwenden können, müssen wir die Technik der Hierarchischen Matrizen erweitern. Unter anderem führen wir eingeschränkte Clusterbäume, eingeschränkte Blockclusterbäume und die verallgemeinerte Addition für Hierarchische Matrizen ein. Zusätzlich führen wir eine neue Clusterbaum-Konstruktion ein, die auf den HDD-Algorithmus zugeschnitten ist. Die Kombination des HDD-Algorithmus mit Hierarchischen Matrizen liefert einen Löser, den wir mit einem fast linearen Aufwand berechnen können. Der Aufwand zur Berechnung einer Lösung sowie der Speicheraufwand ist ebenfalls fast linear. Des Weiteren geben wir noch einige Modifizierungen des HDD-Algorithmus für weitere Anwendungsmöglichkeiten an. Zusätzlich diskutieren wir die Möglichkeiten der Parallelisierung, denn durch die Verwendung der Gebietszerlegung wird das Randwertproblem in unabhängige Teilprobleme aufgeteilt, die sich sehr gut parallelisieren lassen. Wir schließen die Arbeit mit numerischen Tests ab, die die theoretischen Aussagen bestätigen.

Submitted by Marcio Filho (marcio.kleber@ufba.br) on 2017-06-02T13:18:48Z No. of bitstreams: 1 Dissertacao - Daniel Correia.pdf: 2103439 bytes, checksum: 21998777742f6bb17677f65b32158289 (MD5) / Approved for entry into archive by Vanessa Reis (vanessa.jamile@ufba.br) on 2017-06-08T11:07:45Z (GMT) No. of bitstreams: 1 Dissertacao - Daniel Correia.pdf: 2103439 bytes, checksum: 21998777742f6bb17677f65b32158289 (MD5) / Made available in DSpace on 2017-06-08T11:07:45Z (GMT). No. of bitstreams: 1 Dissertacao - Daniel Correia.pdf: 2103439 bytes, checksum: 21998777742f6bb17677f65b32158289 (MD5) / O objetivo deste trabalho é desenvolver e aplicar uma metodologia para a realização da análise modal experimental em um conjunto de bombeio por cavidades progressivas (BCP), em condições de instalação similares às de poços reais. A técnica utilizada neste trabalho, conhecida como Enhanced Frequency Domain Decomposition (EFDD), é capaz de estimar os parâmetros modais com boa acurácia, mesmo na presença de ruído. Ela pode estimar as frequências naturais e formas modais através da decomposição em valores singulares (SVD - Singular value decomposition) da matriz densidade espectral de potência (PSD - Power spectral density) da resposta do sistema a uma excitação de entrada. A taxa de amortecimento pode ser obtida através do cálculo do decremento logarítmico nas funções de autocorrelação. A primeira etapa do estudo consiste no desenvolvimento de algoritmos para a implementação de alguns métodos de análise modal experimental comuns na literatura. Foram implementados os métodos do Peak picking (PP), do Frequency Domain Decomposition (FDD), EFDD e do Least Square Complex Exponential (LSCE). É utilizado um sinal sintético de parâmetros conhecidos para comparar os resultados obtidos entre tais técnicas. Além disso, este sinal teve alguns parâmetros alterados, a exemplo da razão sinal-ruído, e o comportamento da estimação com estas técnicas foi avaliado. Na etapa seguinte, são analisadas estruturas relacionadas ao conjunto BCP, como a coluna de revestimento, onde o conjunto será instalado, além de um dos seus principais subsistemas, o conjunto rotor-estator. Por fim, é realizada a análise modal experimental do conjunto BCP, no interior do poço de testes, em condições de instalação similares àquelas encontradas em um poço real.

TEIXEIRA, Daniel Nascimento. Uma técnica de decomposição a priori para geração paralela de malhas bidimensionais. 2014. 95 f. Dissertação (Mestrado em ciência da computação)- Universidade Federal do Ceará, Fortaleza-CE, 2014. / Submitted by Elineudson Ribeiro (elineudsonr@gmail.com) on 2016-07-11T12:51:06Z No. of bitstreams: 1 2014_dis_dnteixeira.pdf: 17919971 bytes, checksum: 092ad12b33cf64a31552e6a839a5a5bc (MD5) / Approved for entry into archive by Rocilda Sales (rocilda@ufc.br) on 2016-07-15T12:49:53Z (GMT) No. of bitstreams: 1 2014_dis_dnteixeira.pdf: 17919971 bytes, checksum: 092ad12b33cf64a31552e6a839a5a5bc (MD5) / Made available in DSpace on 2016-07-15T12:49:53Z (GMT). No. of bitstreams: 1 2014_dis_dnteixeira.pdf: 17919971 bytes, checksum: 092ad12b33cf64a31552e6a839a5a5bc (MD5) Previous issue date: 2014 / This work describes a technique of two-dimensional domain decomposition for parallel mesh generation. This technique works for both distributed and shared memory and has the freedom to use any data structure that manages rectangular regions parallel to the axes to decompose the domain given as input, such as a quaternary tree (quadtree) or a binary space decomposition (bsp), for example. Any process of mesh generation that respects the prerequisites established can be used in the subdomains created, for instance, Delaunay or Advancing Front, among others. This technique is called a priori because the mesh on the interface of the subdomains is generated prior to the their internal meshes. The load estimation for each sub-domain in this work is performed with the aid of a refined quadtree, whose level of refinement guides the creation of edges that are defined from the bounderies of only inner cells. This way of estimate load produces results that accurately represent the number of elements to be generated in each subdomain. That contributes to a good partitioning of the domain, making the mesh generation in parallel be significantly faster than the serial generation. Furthermore, the quality of the generated mesh in parallel is qualitatively equivalent to that generated serially within acceptable limits. / Este trabalho descreve uma técnica de decomposição de domínios bidimensionais para geração em paralelo de malhas. Esta técnica funciona tanto para memória distribuída quanto compartilhada, além de permitir que se utilize qualquer estrutura de dados que gere regiões quadrangulares paralelas aos eixos para decompor o domínio dado como entrada. Pode se utilizar por exemplo, uma árvore quaternária (quadtree) ou uma partição binária do espaço (bsp). Além disso, qualquer processo de geração de malha que respeite os pré-requisitos estabelecidos pode ser empregado nos subdomínios criados, como as técnicas de Delaunay ou Avanço de Fronteira, dentre outras. A técnica proposta é dita a priori porque a malha de interface entre os subdomínios é gerada antes das suas malhas internas. A estimativa de carga de processamento associada a cada subdomínio é feita nesse trabalho com a ajuda de uma quadtree refinada, cujo nível de refinamento orienta a criação das arestas que são definidas a partir da discretização das fronteiras das células internas. Essa maneira de estimar carga produz resultados que representam, com boa precisão, o número de elementos a serem gerados em cada subdomínio. Isso contribui para um bom particionamento do domínio, fazendo com que a geração de malha em paralelo seja significativamente mais rápida do que a geração serial. Além disso, a qualidade da malha gerada em paralelo é qualitativamente equivalente àquela gerada serialmente, dentro de limites aceitáveis.

Les besoins industriels en simulation numérique de milieux granulaires sont de plus en plus conséquents pour des systèmes de grande dimension. Le cas d'interactions entre grains de type contact unilatéral avec frottement présente des difficultés supplémentaires pour de telles simulations. Dans ce mémoire une approche par décomposition de domaine est proposée. Les méthodes de sous structuration ont été initialement développées pour des milieux continus généralement discrétisés en mécanique des solides en éléments finis. La plateforme LMGC90 (Logiciel de Mécanique Gérant le Contact en Fortran 90) constitue le cadre d'implantation d'algorithmes dédiés. Ainsi des algorithmes de décomposition de domaine, reposant sur les méthodes LArge Time Increment  et Gauss Seidel Non Linéaire, adaptés à un système de type granulaire sont définis, implantés et comparés.Pour exploiter le potentiel en calcul parallèle des méthodes ci-dessus, les procédures d'échange de message par MPI (Message Passing Interface) sont ajoutées au code. Ensuite,  l'amélioration de l'extensibilité des approches multidomaines par l'ajout d'une échelle macroscopique est testée.  Enfin, dans la perspective d'un dialogue entre modèles discret (échelle microscopique) et continu (échelle macroscopique), une version enrichie de la méthode GSNL-DD (Gauss Seidel Non Linéaire en Décomposition de domaine) est proposée. L'accélération de convergence attendue est ensuite étudiée théoriquement sur des exemples de taille réduite, avant quelques tests sur échantillons plus conséquents. / Industrial demand for numerical simulation of granular media is increasing for large systems. The case of interactions between grains such as unilateral contact with friction involves additional difficulties to these simulations. This study investigates a domain decomposition approach. The sub-structuration methods were originally developed for continuous media usually discretized by finite elements for solid mechanics. The LMGC90 platform (software to manage contact with distinct elements) provides a framework for the implementation of algorithms. Thus, domain decomposition algorithms, based on the LArge Time INcrement and Non Linear Gauss Seidel methods, ans suited to a granular problem are defined, implemented and compared. To exploit the potential for parallel computing of the aforementioned methods, the exchanging messages with MPI (Message Passing Interface) is added to the code. Then, the improvement of the scalability of multi-domain approaches through the addition of a macroscopic scale is tested. Finally, in order to implement a dialogue between the discrete (microscopic scale) and continuous (macroscopic scale) models, an enhanced version of the NLGS-DD method (Non Linear Gauss Seidel with domain decomposition) is proposed. The expected acceleration of the convergence is studied theoretically on reduced-size samples, prior to performing some tests on larger samples.

Nos travaux concernent la mise en oeuvre d’une méthode multiéchelle pour faciliter la simulation numérique de structures complexes, appliquée à la modélisation de composants aéronautiques (notamment pour les pièces tournantes de turboréacteur et des structures composites stratifiées). Ces développements sont basés autour de la méthode Arlequin qui permet d’enrichir des modélisations numériques, à l’aide de patchs, autour de zones d’intérêt où des phénomènes complexes se produisent. Cette méthode est mise en oeuvre dans un cadre général permettant la superposition de maillages incompatibles au sein du code de calcul Z-set{Zébulon, en utilisant une formulation optimale des opérateurs de couplage. La précision et la robustesse de cette approche ont été évaluées sur différents problèmes numériques. Afin d’accroître les performances de la méthode Arlequin, un solveur spécifique basé sur les techniques de décomposition de domaine a été développé pour bénéficier des capacités de calcul offertes par les machines à architectures parallèles. Ces performances ont été évaluées sur différents cas tests académiques et quasi-industriels. Enfin, ces développements ont été appliqué à la simulation de problèmes de structures composites stratifiées. / Our work concerns the implementation of a method for convenient multiscale numerical simulation of complex structures, applied to the modeling of aircraft components (including rotating parts made of jet engine from laminate composite structures). These developments are based on the Arlequin method which allows to enrich numerical modeling, using patches around areas of interest where complex phenomena occur. This method is implemented in a general framework in order to link made of incompatible meshes in the Z-set{Zébulon finite element code, using an optimal formulation of the coupling operators. The accuracy and robustness of this approach were evaluated on various numerical problems. To increase the performance of the Arlequin method, a specific solver based on domain decomposition techniques has been developed to take advantage of computing capabilities offered by parallel machine architectures. Its performance has been evaluated on different numerical assessments from academic to industrial tests. Finally, these developments have been applied to the simulation of problems made of laminate composite structures.

Cette thèse s’intéresse à la simulation acoustique de problèmes de grandes tailles. La parallélisation des méthodes numériques d’acoustique est le sujet principal de cette étude. Le manuscrit est composé de trois parties : lancé de rayon, méthodes de décomposition de domaines et algorithmes asynchrones. / This thesis studies numerical methods for large-scale acoustic problems. The parallelization of the numerical acoustic methods is the main focus. The manuscript is composed of three parts: ray-tracing, optimized interface conditions for domain decomposition methods and asynchronous iterative algorithms.