![]() | Only 14 pages are availabe for public view |
Abstract Summary Thesis title: “Block Iterative Methods and Acceleration Techniques”. The main objective in this work is based on the fact that the rate of convergence of the block Jacobi method is approximately the same as that of the point Gauss Seidel method. The aim of this thesis is to introduce a new version of group iterative methods for the algebraic systems arising from the discretization of boundary value problems in the plane. Some special forms of a relatively new iterative block technique are introduced, the KSLOR, KS2LOR, and GKSOR. The thesis consists of four chapters, Arabic summary, and English summary. Chapter One: Basic Concepts The basic concepts required in investigating iterative techniques for solving linear systems are introduced in a simple form. Chapter Two: Block Iterative Methods Standard Block Iterative methods (Block Jacobi, Block Gauss- Seidel, Block Successive over-relaxation) are studied, moreover the Block KSOR method is introduced. The Block relaxation schemes are generalizations of the point relaxation schemes. They update a whole set of components of unknown vector at each time, typically a sub-vector of the solution vector, instead of only one component. Special block iterative methods are discussed, namely the Line iterative methods (Line Jacobi, Line Gauss-Seidel, Line SOR, and Line KSOR). Also, the two Line iterative methods with the natural ordering as well as the red black ordering are discussed. Chapter Three: group Iterative Methods Different group iterative methods are considered, through the treatment of Poisson equation defined on a unit square with 8 8 64 internal mesh points. The novel approach of using groups of fixed size is considered. I.e. Groups of a particular number of individual equations (mesh points), each group is dealt in the same way as a single point. Our fundamental concern is to develop new grouping of the mesh points into small size groups of 2, 4, 8 and 16 points and to investigate their advantages as well as choosing the most efficient group for solving such linear systems which arise from the discretization of PDE’s in two-space dimensions. Chapter Four: Preconditioned and Convergence Acceleration Preconditioning is one of the most usable acceleration techniques. The approach of preconditioning is demonstrated through simple forms of well known types. It is proved that the preconditioning gives considerable improvement in the rate of convergence for the iterative method (Jacobi, Gauss-Seidel, SOR, and KSOR). It is demonstrated that the line SOR method is a more efficient method than preconditioning treatment. Moreover, the 2 line is more efficient than the line method. • The objective of the thesis is to introduce a new versions of block iterative methods. • The motivation of the thesis is the Discretization techniques of partial differential equations usually produce matrices which are banded, or block banded. It is worth to mention that: - All calculations were done by using Matlab R2012b in the Mathematics Department, Faculty of Science, Ain Shams University. - The results of chapter two are published in the Journal of “Applied and Computational Mathematics”. |