Special adaptive grids and Runge - Richardson correction in problems with layers

При решении задач с пограничными и внутренними слоями на адаптивных сетках весьма желательно пользоваться разностными схемами, которые имеют достаточно хорошую точность и сходятся равномерно по малому параметру при стремлении шагов сетки к нулю. Однако эти требования оказываются противоречивыми: схемы высокой точности не сходятся равномерно, а равномерно сходящиеся схемы имеют обычно лишь первый порядок точности. Тем не менее существует уникальная возможность разрешить это противоречие, повышая порядок точности путем применения экстраполяционных поправок Рунге-Ричардсона, представляющих собой линейные комбинации разностных решений на вложенных сетках. В данной работе на примере нескольких употребительных разностных схем изучается эффективность такого подхода к расчетам, полученным на адаптивных сетках, явно задаваемых специальными координатными преобразованиями. Исследуются две схемы противопотокового типа с диагональным преобладанием, равномерно сходящиеся, в сравнении с аналогом схемы с центральной разностью, не имеющей диагонального преобладания и не сходящейся равномерно. Кроме простых поправок применяются также двукратные поправки, еще более повышающие порядок точности результирующих решений It is highly desirable using difference schemes with high accuracy and uniform convergence in a small parameter as the grid steps tend to zero for solving the problems with both boundary and interior layers. However, these requirements turn out to be contradictory: highly-accurate schemes may not converge uniformly, and uniformly converging schemes usually have only the first order of accuracy. Nevertheless, there is a unique opportunity to resolve this contradiction by increasing the order of accuracy by applying the Richardson-Runge extrapolation corrections, which are linear combinations of difference solutions on nested grids. In this paper, using the example of several common difference schemes, we study the efficiency of such approach for calculations obtained on adaptive grids that are explicitly specified by special coordinate transformations. Two diagonal-dominated upstream-type uniformly converging schemes are investigated. They are compared with an analogue of the scheme with central difference that does not have a diagonal dominance and does not converge uniformly. In addition to simple corrections, double corrections are also used, which further increase the order of accuracy of the resulting solutions

Numerical solutions of a system of singularly perturbed reaction-diffusion problems

This paper addresses the numerical approximation of solutions to a coupled system of singularly perturbed reaction-diffusion equations. The components of the solution exhibit overlapping boundary and interior layers. Sinc procedure can control the oscillations in computed solutions at boundary layer regions naturally because the distribution of Sinc points is denser at near the boundaries. Also the obtained results show that the proposed method is applicable even for small perturbation parameter as ? = 2-30. The convergence analysis of proposed technique is discussed, it is shown that the approximate solutions converge to the exact solutions at an exponential rate. Numerical experiments are carried out to demonstrate the accuracy and efficiency of the method.

The Parametrization of the Cauchy Problem for Nonlinear Differential Equations with Contrast Structures

Introduction. The paper provides an analysis of numerical methods for solving the Cauchy problem for nonlinear ordinary differential equations with contrast structures (interior layers). Similar equations simulate various applied problems of hydro- and aeromechanics, chemical kinetics, the theory of catalytic reactions, etc. An analytical solution to these problems is rarely obtained, and numerical procedure is related with significant difficulties associated with ill-conditionality in the neighborhoods of the boundary and interior layers. The aim of the paper is the scope analysis of traditional numerical methods for solving this class problems and approbation of alternative solution methods. Materials and methods. The traditional explicit Euler and fourth-order Runge-Kutta methods, as well as the implicit Euler method with constant and variable step sizes are used for the numerical solution of the Cauchy problem. The method of solution continuation with respect to the best argument is suggested as an alternative to use. The solution continuation method consists in replacing the original argument of the problem with a new one, measured along the integral curve of the problem. The transformation to the best argument allows obtaining the best conditioned Cauchy problem. Results. The computational difficulties arising when solving the equations with contrast structures by traditional explicit and implicit methods are shown on the example of the test problem solution. These difficulties are expressed in a significant decrease of the step size in the neighborhood of the boundary and interior layers. It leads to the increase of the computational time, as well as to the complication of the solving process for super stiff problems. The authenticity of the obtained results is confirmed by the comparison with the analytical solution and the works of other authors. Conclusions. The results of the computational experiment demonstrate the applicability of the traditional methods for solving the Cauchy problem for equations with contrast structures only at low stiffness. In other cases these methods are ineffective. It is shown that the method of solution continuation with respect to the best argument allows eliminating most of the disadvantages inherent to the original problem. It is reflected in decreasing the computational time and in increasing the solution accuracy. Keywords: contrast structures, method of solution continuation, the best argument, illconditionality, the Cauchy problem, ordinary differential equation For citation: Kuznetsov E. B., Leonov S. S., Tsapko E. D. The Parametrization of the Cauchy Problem for Nonlinear Differential Equations with Contrast Structures. Vestnik Mordovskogo universiteta = Mordovia University Bulletin. 2018; 28(4):486–510. DOI: https://doi.org/10.15507/0236-2910.028.201804.486-510 Acknowledgements: This work was supported by the Russian Science Foundation, project no. 18-19-00474.