Dynamic mesh re‐partitioning considering iterative convergence rate for multiphase flows in OpenFOAM

Abstract

For parallel multiphase flows, the core procedure is solving linear systems using preconditioned iterative methods, and the iterative convergence rate is crucial to the overall efficiency. How the mesh is partitioned influences the iterative convergence rate. However, the numerical characteristics of the linear systems vary significantly along with the time steps because of the dramatic change of the flow fields. Traditional static mesh partition cannot guarantee a good convergence feature throughout the simulation. A dynamic mesh re-partitioning scheme MDM-RPar (Multiphase Dynamic Mesh Re-Partitioning) is proposed and implemented in OpenFOAM (Open Field Operation And Manipulation). MDMRPar employs a new surface field to record the linear system information in every time step. For simple multiphase problems with topo-invariant mesh, MDMRPar periodically adopts the numerical information from the previous time step and calculates a weighted graph from the mesh topology in a straightforward way. For multiphase flows using adaptive mesh refinement method, a coarse graph with vertex weights is built based on the mesh topology and refinement history. The numerical information on the surface field is integrated into the coarse graph as edge weights. In both cases, weighted graphs are re-partitioned by ParMetis, a general multi-level parallel graph re-partitioning package. Experimental results on three multiphase flows show that MDMRPar significantly outperforms the traditional partitioning/re-partitioning schemes both in the iterative convergence rate and the total simulation time. KEYWORDS adaptive mesh refinement, dynamic mesh re-partitioning, iterative convergence rate, multiphase flows, OpenFOAM 1 INTRODUCTION OpenFOAM 1 stands for Open Field Operation And Manipulation. It is an open source code using C++ libraries and serves as a finite-volume-based computational fluid dynamics (CFD) platform for the research community. Multiphase flows are classical and important problems in CFD research field and have been simulated a lot with OpenFOAM. As large-scale, parallel computers have become more widely available and numerical models and algorithms have advanced, parallel simulations help expand the range of physical phenomenon dramatically. In a parallel multiphase problem,the mesh needs to be partitioned and distributed among the processors. How the mesh is partitioned has a significant influence on the overall parallel computing efficiency.For a simple multiphase problem, a static mesh with an appropriate size is enough to exhibit the fluid features. Traditionally, the partitioning is conducted only once before the simulation starts, and the partition then remains unchanged during the simulation. The partitioning goals include the balance of load and the minimization of communication volume. For a complex multiphase problem, AMR (adaptive mesh refinement) is usually adopted to resolve key local features, especially near the interface. 2 The refined mesh needs to be re-partitioned in parallel during the simulation.Tradi...

Estudo e Análise de Processamento Paralelo em Simulações de Fluidodinâmica Computacional com OpenFOAM

“…Alguns deles indicam que o gargalo na escalabilidade dos solucionadores OpenFOAM está ligadoà massiva comunicação MPI entre processadores relacionada tanto a forma de particionamento da malha [Wang et al 2012] quanto ao solucionador iterativo selecionado [Rivera and Furlinger 2011]. Nesse contexto, existem autores que propuseram melhorias através de um novo método de particionamento estático a fim de melhorar desempenho e, também, convergência de fluxos multifásicos [Xu et al 2019]. O trabalho em [Lin et al 2018] propõe otimizações na comunicação de solucionadores iterativos.…”

Foreword to the Special Section on Parallel and Distributed Computing, Applications and Technologies

“…In the first paper, the authors proposed a multiphase dynamic mesh re‐partitioning scheme called MDMRPar in OpenFOAM for solving in a linear system, which takes the numerical characteristics of the linear system from the previous time step into consideration and adopts it in the re‐partitioning to speed up the convergence of iterative solvers. They evaluated the performance of the proposed MDMRPar on two damBreak simulations with topoinvariant meshes and a falling drop simulation using adaptive mesh refinement and showed an improvement of 18%‐32% on the total solution time than the traditional partitioning/re‐partitioning schemes.…”