Iterative regularization methods for ill-posed problems

This Ph.D thesis focuses on iterative regularization methods for regularizing linear and nonlinear ill-posed problems. Regarding linear problems, three new stopping rules for the Conjugate Gradient method applied to the normal equations are proposed and tested in many numerical simulations, including some tomographic images reconstruction problems. Regarding nonlinear problems, convergence and convergence rate results are provided for a Newton-type method with a modified version of Landweber iteration as an inner iteration in a Banach space setting.

Regularization methods for the solution of a nonlinear least-squares problem in tomography

In this work we study a polyenergetic and multimaterial model for the breast image reconstruction in Digital Tomosynthesis, taking into consideration the variety of the materials forming the object and the polyenergetic nature of the X-rays beam. The modelling of the problem leads to the resolution of a high-dimensional nonlinear least-squares problem that, due to its nature of inverse ill-posed problem, needs some kind of regularization. We test two main classes of methods: the Levenberg-Marquardt method (together with the Conjugate Gradient method for the computation of the descent direction) and two limited-memory BFGS-like methods (L-BFGS). We perform some experiments for different values of the regularization parameter (constant or varying at each iteration), tolerances and stop conditions. Finally, we analyse the performance of the several methods comparing relative errors, iterations number, times and the qualities of the reconstructed images.

Reconstruction of a stationary flow from boundary data using iterative methods

In this paper, three iterative procedures (Landweber-Fridman, conjugate gradient and minimal error methods) for obtaining a stable solution to the Cauchy problem in slow viscous flows are presented and compared. A section is devoted to the numerical investigations of these algorithms. There, we use the boundary element method together with efficient stopping criteria for ceasing the iteration process in order to obtain stable solutions.

Gradient-dependent density functionals of the Perdew-Burke-Ernzerhof type for atoms, molecules, and solids

One of the standard generalized-gradient approximations (GGAs) in use in modern electronic-structure theory [Perdew-Burke-Ernzerhof (PBE) GGA] and a recently proposed modification designed specifically for solids (PBEsol) are identified as particular members of a family of functionals taking their parameters from different properties of homogeneous or inhomogeneous electron liquids. Three further members of this family are constructed and tested, together with the original PBE and PBEsol, for atoms, molecules, and solids. We find that PBE, in spite of its popularity in solid-state physics and quantum chemistry, is not always the best performing member of the family and that PBEsol, in spite of having been constructed specifically for solids, is not the best for solids. The performance of GGAs for finite systems is found to sensitively depend on the choice of constraints stemming from infinite systems. Guidelines both for users and for developers of density functionals emerge from this work.

Recycling Krylov subspaces for efficient large-scale electrical impedance tomography

Electrical impedance tomography (EIT) captures images of internal features of a body. Electrodes are attached to the boundary of the body, low intensity alternating currents are applied, and the resulting electric potentials are measured. Then, based on the measurements, an estimation algorithm obtains the three-dimensional internal admittivity distribution that corresponds to the image. One of the main goals of medical EIT is to achieve high resolution and an accurate result at low computational cost. However, when the finite element method (FEM) is employed and the corresponding mesh is refined to increase resolution and accuracy, the computational cost increases substantially, especially in the estimation of absolute admittivity distributions. Therefore, we consider in this work a fast iterative solver for the forward problem, which was previously reported in the context of structural optimization. We propose several improvements to this solver to increase its performance in the EIT context. The solver is based on the recycling of approximate invariant subspaces, and it is applied to reduce the EIT computation time for a constant and high resolution finite element mesh. In addition, we consider a powerful preconditioner and provide a detailed pseudocode for the improved iterative solver. The numerical results show the effectiveness of our approach: the proposed algorithm is faster than the preconditioned conjugate gradient (CG) algorithm. The results also show that even on a standard PC without parallelization, a high mesh resolution (more than 150,000 degrees of freedom) can be used for image estimation at a relatively low computational cost. (C) 2010 Elsevier B.V. All rights reserved.

Lanczos algorithm on the grassmann manifold

5th. European Congress on Computational Methods in Applied Sciences and Engineering (ECCOMAS 2008) 8th. World Congress on Computational Mechanics (WCCM8)

Étude numérique des origines hémodynamiques des oscillations dans des réseaux de capillaires

En simulant l’écoulement du sang dans un réseau de capillaires (en l’absence de contrôle biologique), il est possible d’observer la présence d’oscillations de certains paramètres comme le débit volumique, la pression et l’hématocrite (volume des globules rouges par rapport au volume du sang total). Ce comportement semble être en concordance avec certaines expériences in vivo. Malgré cet accord, il faut se demander si les fluctuations observées lors des simulations de l’écoulement sont physiques, numériques ou un artefact de modèles irréalistes puisqu’il existe toujours des différences entre des modélisations et des expériences in vivo. Pour répondre à cette question de façon satisfaisante, nous étudierons et analyserons l’écoulement du sang ainsi que la nature des oscillations observées dans quelques réseaux de capillaires utilisant un modèle convectif et un modèle moyenné pour décrire les équations de conservation de masse des globules rouges. Ces modèles tiennent compte de deux effets rhéologiques importants : l’effet Fåhraeus-Lindqvist décrivant la viscosité apparente dans un vaisseau et l’effet de séparation de phase schématisant la distribution des globules rouges aux points de bifurcation. Pour décrire ce dernier effet, deux lois de séparation de phase (les lois de Pries et al. et de Fenton et al.) seront étudiées et comparées. Dans ce mémoire, nous présenterons une description du problème physiologique (rhéologie du sang). Nous montrerons les modèles mathématiques employés (moyenné et convectif) ainsi que les lois de séparation de phase (Pries et al. et Fenton et al.) accompagnés d’une analyse des schémas numériques implémentés. Pour le modèle moyenné, nous employons le schéma numérique explicite traditionnel d’Euler ainsi qu’un nouveau schéma implicite qui permet de résoudre ce problème d’une manière efficace. Ceci est fait en utilisant une méthode de Newton- Krylov avec gradient conjugué préconditionné et la méthode de GMRES pour les itérations intérieures ainsi qu’une méthode quasi-Newton (la méthode de Broyden). Cette méthode inclura le schéma implicite d’Euler et la méthode des trapèzes. Pour le schéma convectif, la méthode explicite de Kiani et al. sera implémentée ainsi qu’une nouvelle approche implicite. La stabilité des deux modèles sera également explorée. À l’aide de trois différentes topologies, nous comparerons les résultats de ces deux modèles mathématiques ainsi que les lois de séparation de phase afin de déterminer dans quelle mesure les oscillations observées peuvent être attribuables au choix des modèles mathématiques ou au choix des méthodes numériques.

Implicit-factorization preconditioning and iterative solvers for regularized saddle-point systems

We consider conjugate-gradient like methods for solving block symmetric indefinite linear systems that arise from saddle-point problems or, in particular, regularizations thereof. Such methods require preconditioners that preserve certain sub-blocks from the original systems but allow considerable flexibility for the remaining blocks. We construct a number of families of implicit factorizations that are capable of reproducing the required sub-blocks and (some) of the remainder. These generalize known implicit factorizations for the unregularized case. Improved eigenvalue clustering is possible if additionally some of the noncrucial blocks are reproduced. Numerical experiments confirm that these implicit-factorization preconditioners can be very effective in practice.

Reduced gradient method combined with augmented Lagrangian and barrier for the optimal power flow problem

A new approach for solving the optimal power flow (OPF) problem is established by combining the reduced gradient method and the augmented Lagrangian method with barriers and exploring specific characteristics of the relations between the variables of the OPF problem. Computer simulations on IEEE 14-bus and IEEE 30-bus test systems illustrate the method. (c) 2007 Elsevier Inc. All rights reserved.

An unstructured CVFEM and moving interface algorithm for non-Newtonian Hele-Shaw flows in injection molding

Purpose - The purpose of this paper is to develop a novel unstructured simulation approach for injection molding processes described by the Hele-Shaw model. Design/methodology/approach - The scheme involves dual dynamic meshes with active and inactive cells determined from an initial background pointset. The quasi-static pressure solution in each timestep for this evolving unstructured mesh system is approximated using a control volume finite element method formulation coupled to a corresponding modified volume of fluid method. The flow is considered to be isothermal and non-Newtonian. Findings - Supporting numerical tests and performance studies for polystyrene described by Carreau, Cross, Ellis and Power-law fluid models are conducted. Results for the present method are shown to be comparable to those from other methods for both Newtonian fluid and polystyrene fluid injected in different mold geometries. Research limitations/implications - With respect to the methodology, the background pointset infers a mesh that is dynamically reconstructed here, and there are a number of efficiency issues and improvements that would be relevant to industrial applications. For instance, one can use the pointset to construct special bases and invoke a so-called ""meshless"" scheme using the basis. This would require some interesting strategies to deal with the dynamic point enrichment of the moving front that could benefit from the present front treatment strategy. There are also issues related to mass conservation and fill-time errors that might be addressed by introducing suitable projections. The general question of ""rate of convergence"" of these schemes requires analysis. Numerical results here suggest first-order accuracy and are consistent with the approximations made, but theoretical results are not available yet for these methods. Originality/value - This novel unstructured simulation approach involves dual meshes with active and inactive cells determined from an initial background pointset: local active dual patches are constructed ""on-the-fly"" for each ""active point"" to form a dynamic virtual mesh of active elements that evolves with the moving interface.

Augmented Lagrangian methods under the constant positive linear dependence constraint qualification

Two Augmented Lagrangian algorithms for solving KKT systems are introduced. The algorithms differ in the way in which penalty parameters are updated. Possibly infeasible accumulation points are characterized. It is proved that feasible limit points that satisfy the Constant Positive Linear Dependence constraint qualification are KKT solutions. Boundedness of the penalty parameters is proved under suitable assumptions. Numerical experiments are presented.

Uma investigação da influência do parâmetro de barreira na solução do problema de fluxo de potência ótimo

Pós-graduação em Engenharia Elétrica - FEB

Processamento interpretativo de dados magnetométricos e inversão de dados gravimétricos aplicados à prospecção de hidrocarbonetos

Apresentamos dois métodos de interpretação de dados de campos potenciais, aplicados à prospecção de hidrocarbonetos. O primeiro emprega dados aeromagnéticos para estimar o limite, no plano horizontal, entre a crosta continental e a crosta oceânica. Este método baseia-se na existência de feições geológicas magnéticas exclusivas da crosta continental, de modo que as estimativas das extremidades destas feições são usadas como estimativas dos limites da crosta continental. Para tanto, o sinal da anomalia aeromagnética na região da plataforma, do talude e da elevação continental é amplificado através do operador de continuação analítica para baixo usando duas implementações: o princípio da camada equivalente e a condição de fronteira de Dirichlet. A maior carga computacional no cálculo do campo continuado para baixo reside na resolução de um sistema de equações lineares de grande porte. Este esforço computacional é minimizado através do processamento por janelas e do emprego do método do gradiente conjugado na resolução do sistema de equações. Como a operação de continuação para baixo é instável, estabilizamos a solução através do funcional estabilizador de primeira ordem de Tikhonov. Testes em dados aeromagnéticos sintéticos contaminados com ruído pseudo-aleatório Gaussiano mostraram a eficiência de ambas as implementações para realçar os finais das feições magnéticas exclusivas da crosta continental, permitindo o delineamento do limite desta com a crosta oceânica. Aplicamos a metodologia em suas duas implementações a dados aeromagnéticos reais de duas regiões da costa brasileira: Foz do Amazonas e Bacia do Jequitinhonha. O segundo método delineia, simultaneamente, a topografia do embasamento de uma bacia sedimentar e a geometria de estruturas salinas contidas no pacote sedimentar. Os modelos interpretativos consistem de um conjunto de prismas bidimensionais verticais justapostos, para o pacote sedimentar e de prismas bidimensionais com seções verticais poligonais para as estruturas salinas. Estabilizamos a solução, incorporando características geométricas do relevo do embasamento e das estruturas salinas compatíveis com o ambiente geológico através dos estabilizadores da suavidade global, suavidade ponderada e da concentração de massa ao longo de direções preferenciais, além de vínculos de desigualdade nos parâmetros. Aplicamos o método a dados gravimétricos sintéticos produzidos por fontes 2D simulando bacias sedimentares intracratônicas e marginais apresentando densidade do pacote sedimentar variando com a profundidade segundo uma lei hiperbólica e abrigando domos e almofadas salinas. Os resultados mostraram que o método apresenta potencial para delinear, simultaneamente, as geometrias tanto de almofadas e domos salinos, como de relevos descontínuos do embasamento. Aplicamos o método, também, a dados reais ao longo de dois perfis gravimétricos sobre as Bacias de Campos e do Jequitinhonha e obtivemos interpretações compatíveis com a geologia da área.

Uso das rotações de givens modificadas como um método direto para obtenção e atualização das soluções em sistemas com acumulação seqüencial de dados

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Matrix-vector multiplication and triangular linear solver using GPGPU for symmetric positive definite matrices derived from elliptic equations

The modern GPUs are well suited for intensive computational tasks and massive parallel computation. Sparse matrix multiplication and linear triangular solver are the most important and heavily used kernels in scientific computation, and several challenges in developing a high performance kernel with the two modules is investigated. The main interest it to solve linear systems derived from the elliptic equations with triangular elements. The resulting linear system has a symmetric positive definite matrix. The sparse matrix is stored in the compressed sparse row (CSR) format. It is proposed a CUDA algorithm to execute the matrix vector multiplication using directly the CSR format. A dependence tree algorithm is used to determine which variables the linear triangular solver can determine in parallel. To increase the number of the parallel threads, a coloring graph algorithm is implemented to reorder the mesh numbering in a pre-processing phase. The proposed method is compared with parallel and serial available libraries. The results show that the proposed method improves the computation cost of the matrix vector multiplication. The pre-processing associated with the triangular solver needs to be executed just once in the proposed method. The conjugate gradient method was implemented and showed similar convergence rate for all the compared methods. The proposed method showed significant smaller execution time.