The Finite Element Method (FEM) is a powerful numerical method to solve wave propagation problems for open-region electromagnetic radiation/scattering problems involving objects with arbitrary geometry and constitutive parameters. In high-frequency applications, the FEM requires an electrically large computational domain, implying a large number of unknowns, such that the numerical solution of the problem is not feasible even on state-of-the-art computers. An appealing way to solve a large FEM problem is to employ a Domain Decomposition Method (DDM) that allows the decomposition of a large problem into several coupled subproblems which can be solved independently, thus reducing considerably the memory storage requirements. In this thesis, two new domain decomposition algorithms (FB-DDM and ILF-DDM) are implemented for the finite element solution of electromagnetic radiation/scattering problems. For this purpose, a nodal FEM code (FEMS2D) employing triangular elements and a vector FEM code (FEMS3D) employing tetrahedral edge elements have been developed for 2D and 3D problems, respectively. The unbounded domain of the radiation/scattering problem, as well as the boundaries of the subdomains in the DDMs, are truncated by the Perfectly Matched Layer (PML) absorber. The PML is implemented using two new approaches: Locally-conformal PML and Multi-center PML. These approaches are based on a locally-defined complex coordinate transformation which makes possible to handle challenging PML geometries, especially with curvature discontinuities. In order to implement these PML methods, we also introduce the concept of complex space FEM using elements with complex nodal coordinates. The performances of the DDMs and the PML methods are investigated numerically in several applications.

Supercritical fluid extraction has been used to recover biomaterials from natural matrices. Mathematical modeling of the extraction is required for process design and scale up. Existing models in literature are correlative and dependent upon the experimental data. Construction of predictive models giving reliable results in the lack of experimental data is precious. The long term objective of this study was to construct a predictive mass transfer model, representing supercritical fluid extraction of biomaterials in packed beds by the method of volume averaging. In order to develop mass transfer equations in terms of volume averaged variables, velocity and velocity deviation fields, closure variables were solved for a specific case and the coefficients of volume averaged mass transfer equation for the specific case were computed using one and two-dimensional geometries via analytical and numerical solutions, respectively. Spectral Element method with Domain Decomposition technique, Preconditioned Conjugate Gradient algorithm and Uzawa method were used for the numerical solution. The coefficients of convective term with additional terms of volume averaged mass transfer equation were similar to superficial velocity. The coefficients of dispersion term were close to diffusivity of oil in supercritical carbon dioxide. The coefficients of interphase mass transfer term were overestimated in both geometries. Modifications in boundary conditions, change in geometry of particles and use of three-dimensional computations would improve the value of the coefficient of interphase mass transfer term.

From the viewpoint of the adaptive solution of partial differential equations a finit e element method on hierarchical triangular meshes is developed permitting hanging nodes arising from nonuniform hierarchical refinement. Construction, extension and restriction of the nonuniform hierarchical basis and the accompanying mesh are described by graphs. The corresponding FE basis is generated by hierarchical transformation. The characteristic feature of the generalizable concept is the combination of the conforming hierarchical basis for easily defining and changing the FE space with an accompanying nonconforming FE basis for the easy assembly of a FE equations system. For an elliptic model the conforming FEM problem is solved by an iterative method applied to this nonconforming FEM equations system and modified by projection into the subspace of conforming basis functions. The iterative method used is the Yserentant- or BPX-preconditioned conjugate gradient algorithm. On a MIMD computer system the parallelization by domain decomposition is easy and efficient to organize both for the generation and solution of the equations system and for the change of basis and mesh.

Hydrocarbon production or injection of fluids in the reservoir can produce changes in the rock stresses and in-situ geomechanics, potentially leading to compaction and subsidence with harmful effects in wells, cap-rock, faults, and the surrounding environment as well. In order to tackle these changes and their impact, accurate simulations are essential. The Mortar Finite Element Method (MFEM) has been demonstrated to be a powerful technique in order to formulate a weak continuity condition at the interface of sub-domains in which different meshes, i.e. non-conforming or hybrid, and / or variational approximations are used. This is particularly suitable when coupling different physics on different domains, such as elasticity and poroelasticity, in the context of coupled flow and geomechanics. In this dissertation, popular Domain Decomposition Methods (DDM) are implemented in order to carry large simulations by taking full advantage of current parallel computer architectures. Different solution schemes can be defined depending upon the way information is exchanged between sub-domain interfaces. Three different schemes, i.e. Dirichlet-Neumann (DN), Neumann-Neumann (NN) and MFEM, are tested and the advantages and disadvantages of each of them are identified. As a first contribution, the MFEM is extended to deal with curve interfaces represented by Non-Uniform Rational B-Splines (NURBS) curves and surfaces. The goal is to have a more robust geometrical representation for mortar spaces, which allows gluing non-conforming interfaces on realistic geometries. The resulting mortar saddle-point problem will be decoupled by means of the DN- and NN-DDM. Additionally, a reservoir geometry reconstruction procedure based on NURBS surfaces is presented as well. The technique builds a robust piecewise continuous geometrical representation that can be exploited by MFEM in order to tackle realistic problems, which is a second contribution. Tensor product meshes are usually propagated from the reservoir in a conforming way into its surroundings, which makes non-matching interfaces highly attractive in this case. In the context of reservoir compaction and subsidence estimation, it is common to deal with serial legacy codes for flow. Indeed, major reservoir simulators such as compositional codes lack parallelism. Another issue is the fact that, generally speaking, flow and mechanics domains are different. To overcome this limitation, a serial-parallel approach is proposed in order to couple serial flow codes with our parallel mechanics code by means of iterative coupling. Concrete results in loosely coupling are presented as a third contribution. As a final contribution, the DN-DDM is applied to couple elasticity and plasticity, which seems very promising in order to speed up computations involving poroplasticity. Several examples of coupling of elasticity, poroelasticity, and plasticity ranging from near-wellbore applications to field level subsidence computations help to show that the proposed methodology can handle problems of practical interest. In order to facilitate the implementation of complex workflows, an advanced Python wrapper interface that allows programming capabilities have been implemented. The proposed serial-parallel approach seems to be appropriate to handle geomechanical problems involving different meshes for flow and mechanics as well as coupling parallel mechanistic codes with legacy flow simulators. / text

A surface reconstruction and image enhancement non-linear finite element technique based on minimization of L1 norm of the total variation of the gradient is introduced. Since minimization in the L1 norm is computationally expensive, we seek to improve the performance of this algorithm in two fronts: first, local L1- minimization, which allows parallel implementation; second, application of the Augmented Lagrangian method to solve the minimization problem. We show that local solution of the minimization problem is feasible. Furthermore, the Augmented Lagrangian method can successfully be used to solve the L1 minimization problem. This result is expected to be useful for improving algorithms computing digital elevation maps for natural and urban terrain, fitting surfaces to point-cloud data, and image super-resolution.

Domain decomposition and convex optimization play fundamental roles in current computation and analysis in many areas of science and engineering. These methods have been well developed and studied in the past thirty years, but they still require further study and improving not only in mathematics but in actual engineering computation with exponential increase of computational complexity and scale. The main goal of this thesis is to develop some efficient and powerful algorithms based on domain decomposition method and convex optimization. The topicsstudied in this thesis mainly include two classes of convex optimization problems: optimal control problems governed by time-dependent partial differential equations and general structured convex optimization problems. These problems have acquired a wide range of applications in engineering and also demand a very high computational complexity. The main contributions are as follows: In Chapter 2, the relevance of an adequate inner loop starting point (as opposed to a sufficient inner loop stopping rule) is discussed in the context of a numerical optimization algorithm consisting of nested primal-dual proximal-gradient iterations. To study the optimal control problem, we obtain second order domain decomposition methods by combining Crank-Nicolson scheme with implicit Galerkin method in the sub-domains and explicit flux approximation along inner boundaries in Chapter 3. Parallelism can be easily achieved for these explicit/implicit methods. Time step constraints are proved to be less severe than that of fully explicit Galerkin finite element method. Based on the domain decomposition method in Chapter 3, we propose an iterative algorithm to solve an optimal control problem associated with the corresponding partial differential equation with pointwise constraint for the control variable in Chapter 4. In Chapter 5, overlapping domain decomposition methods are designed for the wave equation on account of prediction-correction" strategy. A family of unit decomposition functions allow reasonable residual distribution or corrections. No iteration is needed in each time step. This dissertation also covers convergence analysis from the point of view of mathematics for each algorithm we present. The main discretization strategy we adopt is finite element method. Moreover, numerical results are provided respectivelyto verify the theory in each chapter. / Doctorat en Sciences / info:eu-repo/semantics/nonPublished

We present algorithms and implementations that overcome obstacles in the migration of the Fast Multipole Method (FMM), one of the most important algorithms in computational science and engineering, to exascale computing. Emerging architectural approaches to exascale computing are all characterized by data movement rates that are slow relative to the demand of aggregate floating point capability, resulting in performance that is bandwidth limited. Practical parallel applications of FMM are impeded in their scaling by irregularity of domains and dominance of collective tree communication, which is known not to scale well. We introduce novel ideas that improve partitioning of the N-body problem with boundary distribution through a sampling-based mechanism that hybridizes two well-known partitioning techniques, Hashed Octree (HOT) and Orthogonal Recursive Bisection (ORB). To reduce communication cost, we employ two methodologies. First, we directly utilize features available in parallel runtime systems to enable asynchronous computing and overlap it with communication. Second, we present Hierarchical Sparse Data Exchange (HSDX), a new all-to-all algorithm that inherently relieves communication by relaying sparse data in a few steps of neighbor exchanges. HSDX exhibits superior scalability and improves relative performance compared to the default MPI alltoall and other relevant literature implementations. We test this algorithm alongside others on a Cray XC40 tightly coupled with the Aries network and on Intel Many Integrated Core Architecture (MIC) represented by Intel Knights Corner (KNC) and Intel Knights Landing (KNL) as modern shared-memory CPU environments. Tests include comparisons of thoroughly tuned handwritten versus auto-vectorization of FMM Particle-to-Particle (P2P) and Multipole-to-Local (M2L) kernels. Scalability of task-based parallelism is assessed with FMM’s tree traversal kernel using different threading libraries. The MIC tests show large performance gains after adopting the prescribed techniques, which are inevitable in a world that is moving towards many-core parallelism.

La simulation numérique des systèmes multicorps en présence d'interactions complexes, dont le contact frottant, pose de nombreux défis, tant en terme de modélisation que de temps de calcul. Dans ce manuscrit de thèse, nous étudions deux familles de décomposition de domaine adaptées au formalisme de la dynamique non régulière des contacts (NSCD). Cette méthode d'intégration implicite en temps de l'évolution d'une collection de corps en interaction a pour caractéristique de prendre en compte le caractère discret et non régulier d'un tel milieu. Les techniques de décomposition de domaine classiques ne peuvent de ce fait être directement transposées. Deux méthodes de décomposition de domaine proches des formalismes des méthodes de Schwarz et de complément de Schur sont présentées. Ces méthodes se révèlent être de puissants outils pour la parallélisation en mémoire distribuée des simulations granulaires 2D et 3D sur un centre de calcul haute performance. Le comportement de structure des milieux granulaires denses est de plus exploité afin de propager rapidement l'information sur l'ensemble des sous-domaines via un schéma semi-implicite d'intégration en temps. / Numerical simulations of the dynamics of discrete structures in presence of numerous impacts and frictional contacts leads to CPU-intensive large time computations. To deal with such realistic assemblies, numerical tools have been developed, in particular the method called nonsmooth contact dynamics (NSCD). Such modeling has to deal with discreteness and nonsmoothness, such that domain decomposition approaches for regular continuum media has to be rethought. We present further two domain decomposition method linked to Schwarz and Schur formalism. Scalability and numerical performances of the methods for 2D and 3D granular media is studied, showing good parallel behavior on a supercomputer platform. The structural behavior of dense granular packing is herein used to introduce a spacial multilevel preconditioner with a coarse problem to improve convergence in a space-time approach.

Made available in DSpace on 2014-06-11T19:27:14Z (GMT). No. of bitstreams: 0 Previous issue date: 2006-05-17Bitstream added on 2014-06-13T20:35:25Z : No. of bitstreams: 1 borges_as_me_ilha.pdf: 1848055 bytes, checksum: baddb0e3ae6ff7e75ac2a367efd5a7a3 (MD5) / Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) / O presente trabalho propõe o estudo e implementação de uma metodologia para a estimação dos parâmetros modais de estruturas utilizando uma técnica de identificação baseada apenas na resposta do modelo, denominada Decomposição no Domínio da Freqüência (DDF). Para tal são abordados os conceitos básicos envolvidos na análise modal, análise modal baseada apenas na resposta e métodos de identificação. A formulação do algoritmo é baseada na decomposição da matriz densidade espectral de potência utilizando a técnica da decomposição em valores singulares (SVD). A decomposição da matriz densidade espectral nas linhas de freqüências correspondentes aos picos de amplitude, permite a estimativa dos modos de vibrar do sistema. Tem-se ainda que, o primeiro vetor singular obtido com a decomposição da matriz densidade espectral, para cada linha de freqüência, na região em torno do modo, contém as respectivas informações daquele modo e o correspondente valor singular leva a uma estimativa da função densidade espectral de um sistema de um grau de liberdade (1GL) equivalente. Neste caso, a matriz densidade espectral de saída é decomposta em um conjunto de sistemas de 1 grau de liberdade. Posteriormente, esses dados são transformados para o domínio do tempo, utilizando a transformada inversa de Fourier, e as razões de amortecimento são estimadas utilizando o conceito de decremento logaritmo. A metodologia é avaliada, numa primeira etapa, utilizando dados simulados e posteriormente utilizando dados experimentais. / The present work proposes the study and implementation of a methodology for the estimating of the modal parameters of structures by using the output-only data. The technique called Frequency Domain Decomposition (DDF) identifies the modal parameters without knowing the input. For that, it is discussed the basic concepts involved in identification, modal analysis and output-only modal analysis. The formulation of the algorithm is based on the decomposition of the power spectral density matrix by using the singular values decomposition technique (SVD). The decomposition of the spectral density matrix for the lines of frequency corresponding to the amplitude peaks, allows the estimating of the modes shape of the system. Additionally, the first singular vector obtained with the decomposition of the spectral density matrix, for each line of frequency, in the area around of the peak, contains the respective information of that mode. The corresponding singular value leads to an estimating of the spectral density function of an equivalent system of one degree of freedom. Therefore, the output spectral density matrix is decomposed in a set of one degree of freedom system. Later on, those data are transformed for the time domain by using the inverse Fourier transform and the damping ratios estimated from the crossing times and the logarithm decrement of the corresponding single degree of freedom system correlation function. The methodology is evaluated using simulated and experimental data.

abstract: A domain decomposition method for analyzing very large FDTD domains, hundreds of thousands of wavelengths long, is demonstrated by application to the problem of radar scattering in the maritime environment. Success depends on the elimination of artificial scattering from the “sky” boundary and is ensured by an ultra-high-performance absorbing termination which eliminates this reflection at angles of incidence as shallow as 0.03 degrees off grazing. The two-dimensional (2D) problem is used to detail the features of the method. The results are cross-validated by comparison to a parabolic equation (PE) method and surface integral equation method on a 1.7km sea surface problem, and to a PE method on propagation through an inhomogeneous atmosphere in a 4km-long space, both at X-band. Additional comparisons are made against boundary integral equation and PE methods from the literature in a 3.6km space containing an inhomogeneous atmosphere above a flat sea at S-band. The applicability of the method to the three-dimensional (3D) problem is shown via comparison of a 2D solution to the 3D solution of a corridor of sea. As a technical proof of the scalability of the problem with computational power, a 5m-wide, 2m-tall, 1050m-long 3D corridor containing 321.8 billion FDTD cells has been simulated at X-band. A plane wave spectrum analysis of the (X-band) scattered fields produced by a 5m-wide, 225m-long realistic 3D sea surface, and the 2D analog surface obtained by extruding a 2D sea along the width of the corridor, reveals the existence of out-of-plane 3D phenomena missed by the traditional 2D analysis. The realistic sea introduces random strong flashes and nulls in addition to a significant amount of cross-polarized field. Spatial integration using a dispersion-corrected Green function is used to reconstruct the scattered fields outside of the computational FDTD space which would impinge on a 3D target at the end of the corridor. The proposed final approach is a hybrid method where 2D FDTD carries the signal for the first tens of kilometers and the last kilometer is analyzed in 3D. / Dissertation/Thesis / Doctoral Dissertation Electrical Engineering 2018