Abstract <p>We consider several coupled biophysical problems. These are the problems in moving domains, including blood flow, two-phase flow of the blood with the clot, and blood flow and coagulation process. The corresponding mathematical models are the Navier–Stokes equations, the Navier–Stokes–Cahn–Hilliard equations, and the Navier–Stokes–Brinkman equations with coagulation chemical kinetics and advection–diffusion of blood factors, respectively. Due to problem stiffness, the fully coupled implicit collocated discretization is considered and discussed in [<CitationRef CitationID="CR1">1</CitationRef>]. The present paper is the second in the series presenting our approach to the simulation of such complex processes. The nonlinear problems require the solution of huge, coupled, block-structured saddle-point systems. In this work, we study the applicability of various open-source linear solvers, choose the most robust, and analyze 700-cores performance of the simulations with the chosen linear solver.</p>

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

Solving Coupled Problems of Blood Flow and Coagulation in Moving Domains, II: Algebraic Solvers and Their Parallel Performance

  • I. Konshin,
  • K. Terekhov,
  • Yu. Vassilevski

摘要

Abstract

We consider several coupled biophysical problems. These are the problems in moving domains, including blood flow, two-phase flow of the blood with the clot, and blood flow and coagulation process. The corresponding mathematical models are the Navier–Stokes equations, the Navier–Stokes–Cahn–Hilliard equations, and the Navier–Stokes–Brinkman equations with coagulation chemical kinetics and advection–diffusion of blood factors, respectively. Due to problem stiffness, the fully coupled implicit collocated discretization is considered and discussed in [1]. The present paper is the second in the series presenting our approach to the simulation of such complex processes. The nonlinear problems require the solution of huge, coupled, block-structured saddle-point systems. In this work, we study the applicability of various open-source linear solvers, choose the most robust, and analyze 700-cores performance of the simulations with the chosen linear solver.