No label defined (Q4244286)

The monograph under review is devoted to constructing effective algorithms for the numerical solution of the Cauchy problem in systems of ordinary differential equations with the ``mean'' stiffness. The increase of efficiency of explicit methods of Runge-Kutta type is achieved at the expense of more flexible management of an integration step, which is selected not only because of accuracy control but also with the help of an additional stability control of the numerical scheme. Moreover, the algorithms proposed are algorithms of variable order, step, and number of stages. Such algorithms expand the boundaries of applicability of explicit methods for stiff systems, as they allow us to select an optimal numerical scheme for each computational step. The monograph consists of the introduction, 5 chapters, applications, and the bibliography of 152 items. The first chapter is devoted to constructing inequalities for the accuracy and stability control of explicit methods, and gives the basic definitions. Two new ways of a global error estimation in one-step methods are developed which do not essentially increase computing costs. A new approach to constructing inequalities for the stability control of explicit methods is suggested. The chapter ends with the questions of a program realization of explicit methods with the accuracy and stability control. In Chapter 2, the author deals with the construction of integration algorithms with the accuracy control of the numerical scheme. Relations are obtained which connect parameters of explicit Runge-Kutta-type methods of stage \(m\) and coefficients of a stability polynomial of order \(m\). On the basis of two-, three-, and four-stage numerical formulas, the methods with extended regions of stability and with inequalities for the accuracy control are constructed. An estimation of the global error for the Merson method is obtained. The numerical results presented confirm reliability of the inequalities for the control of accuracy and efficiency of the algorithms constructed. Chapter 3 is devoted to constructing integration algorithms with the control of accuracy of calculations and stability of the numerical scheme. New methods are developed in which the step is selected depending on the accuracy and stability. Inequalities for the stability control of the known Merson and Fehlberg methods are obtained. The results of calculations which verify increase of the efficiency at the expense of additional stability control are presented. Chapter 4 is devoted to problems of constructing integration algorithms of variable order and step. Three new algorithms are developed. The methods included in each algorithm have different order of accuracy but are based on the same stages. Low-order methods have a broader stability region than that of high-order methods. If the integration step is limited on accuracy, the calculations are conducted under high-order schemata. If the step is limited on stability of the numerical scheme, the transition to low-order formulas with extended stability regions is implemented. At each step, the order of accuracy of a method is determined by means of inequalities for the accuracy and stability control. Two algorithms are based at the same stages as the Merson and Fehlberg methods. The calculation results for test stiff problems are presented which show a significant increase of efficiency at the expense of a combination of methods during calculations. In particular, the efficiency of the Merson and Fehlberg methods is increased more than in 10 times. In Chapter 5, an automation of generating integration algorithms of variable order, step, and number of stages is considered. The algorithm is developed and justified for determining such coefficients of stability polynomials that the stability region of the corresponding method has a prescribed shape and size. For every \(m\), formulas are obtained for determining the coefficients of the explicit \(m\)-stage Runge-Kutta-type methods of the first, second, and third order of accuracy with a prescribed stability region. For these methods, inequalities for the accuracy and stability control are obtained. A program complex for generating integration algorithms is developed in which the structure of methods and their stability properties are determined by a computer operator. The numerical results are presented obtained by the algorithm constructed. In the applications, test examples and graphics images of stability regions are given.