A scalable parallel algorithm for the direct numerical simulation of three-dimensional incompressible particulate flow

TY - JOUR

T1 - A scalable parallel algorithm for the direct numerical simulation of three-dimensional incompressible particulate flow

AU - Jin, Shi

AU - Minev, Peter

AU - Nandakumar, Krishnaswamy

N1 - Funding Information: This project is partially supported by the Discovery Grants of the Natural Sciences and Engineering Research Council of Canada (NSERC). The authors thank WestGrid, AICT at University of Alberta and HPC at University of Calgary for computing facilities and support. S.J. also wants to thank Matthew Knepley and the PETSc team for development support.

PY - 2009/6

Y1 - 2009/6

N2 - Particulate flow is of great importance from both the scientific and engineering points of view. Owing to the complexity of particle-flow interactions, direct numerical simulations (DNS) of inertial particulate flow with finite-size particles have been limited to a very small number of particles, while the industrial applications involve larger numbers with many orders of magnitude. This article presents a parallel implementation of a fictitious domain method for the DNS of particulate flows. The method is thoroughly tested and its parallel performance on distributed memory clusters is evaluated on a large-scale problem. Finally, we present the results for the separation of 21,336 particles of two different densities in a viscous fluid. Although there is still a significant gap between DNS and the industrial applications, the present algorithm allows to simulate significantly large number of particles so that a meaningful statistical analysis can be performed. This will help in the development of new closure relations for the averaged models of multiphase flows.

AB - Particulate flow is of great importance from both the scientific and engineering points of view. Owing to the complexity of particle-flow interactions, direct numerical simulations (DNS) of inertial particulate flow with finite-size particles have been limited to a very small number of particles, while the industrial applications involve larger numbers with many orders of magnitude. This article presents a parallel implementation of a fictitious domain method for the DNS of particulate flows. The method is thoroughly tested and its parallel performance on distributed memory clusters is evaluated on a large-scale problem. Finally, we present the results for the separation of 21,336 particles of two different densities in a viscous fluid. Although there is still a significant gap between DNS and the industrial applications, the present algorithm allows to simulate significantly large number of particles so that a meaningful statistical analysis can be performed. This will help in the development of new closure relations for the averaged models of multiphase flows.

KW - DNS

KW - Multiphase flow

KW - Parallel computing

KW - Particulate flow

KW - Unstructured grid domain decomposition

UR - http://www.scopus.com/inward/record.url?scp=68949090043&partnerID=8YFLogxK

U2 - 10.1080/10618560902973748

DO - 10.1080/10618560902973748

M3 - 文章

AN - SCOPUS:68949090043

SN - 1061-8562

VL - 23

SP - 427

EP - 437

JO - International Journal of Computational Fluid Dynamics

JF - International Journal of Computational Fluid Dynamics

IS - 5

ER -