Dynamic local coupling for multiphase flow: A compromise between efficiency and stability

Abstract

Complex models involving numerous coupled physical processes create substantial computational challenge. This paper introduces a solver algorithm that maintains a locally fully coupled system in the subdomains in which individual physical process interacts with other processes strongly. Meanwhile, the solutions in the other regions are treated in a decoupled fashion. The fully coupled regions are updated dynamically by either different timesteps or iterations. Global coupling of multiple physics generally results in systems with large number of unknowns that are computationally unfeasible. On the other hand, decoupling strategies alleviates the computational load, but results in stability issues, especially for nonlinear problems and sometimes encounters divergence during the solution process. The local coupling strategy applies error estimators to determine the strength of interaction between the physical processes in subdomains. By maintaining the system in fully coupled form within critical regions, the stability issue from the decoupling strategies is avoided while the computational load is significantly reduced as compared to a global coupling strategy.