This paper investigates methods for composing difference approximations to form one-step and many-step parallel difference schemes of a given order, with a focus on numerically solving the Cauchy problem for both ordinary differential equations and evolutionary partial differential equations. The proposed discrete approximations allow for varying the order of error while moving away from the maximum possible on a fixed set of knots, while ensuring the absolute or A- \(\alpha \) stability of the numerical solutions. It is shown that one-step schemes of the maximum approximation order remain absolutely stable as the dimensionality of computational blocks increases. Fulfilment of the condition of absolute or A- \(\alpha \) stability for multistep methods is provided by reducing the maximum possible order of approximation.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Parallel Methods for Composing Difference Approximations in Simulation Dynamic Processes

  • Olga Dmytriyeva,
  • Vira Huskova

摘要

This paper investigates methods for composing difference approximations to form one-step and many-step parallel difference schemes of a given order, with a focus on numerically solving the Cauchy problem for both ordinary differential equations and evolutionary partial differential equations. The proposed discrete approximations allow for varying the order of error while moving away from the maximum possible on a fixed set of knots, while ensuring the absolute or A- \(\alpha \) stability of the numerical solutions. It is shown that one-step schemes of the maximum approximation order remain absolutely stable as the dimensionality of computational blocks increases. Fulfilment of the condition of absolute or A- \(\alpha \) stability for multistep methods is provided by reducing the maximum possible order of approximation.