Parallel algorithms for CFD-DEM modeling of dense particulate flows

TY - JOUR

T1 - Parallel algorithms for CFD-DEM modeling of dense particulate flows

AU - Wu, C. L.

AU - Ayeni, Oladapo

AU - Berrouk, A. S.

AU - Nandakumar, K.

PY - 2014/10/18

Y1 - 2014/10/18

N2 - Most of the CFD codes are parallelized under distributed-memory parallel computing environments. For this reason, many attempts have been made to develop parallel DPM/DEM algorithms for dense particulate flows under the same parallel architecture. For such a development, it is very difficult to achieve efficient load balancing of processors due to the heterogeneous particle spatial distribution that often characterizes dense particulate systems. In this work, parallel CFD-DEM algorithms have been developed under the distributed memory environment with a fluid flow solver based on finite volume method and arbitrary 3D unstructured meshes. An implicit two-phase coupling scheme was proposed to enhance numerical stability for complex dense particulate flow problems. Parallelization-generated numerical difficulties such as void fraction calculation, two-phase momentum exchange, and contact force calculations for particles at irregular and arbitrary partition boundaries were efficiently addressed. The load-balancing difficulty due to heterogeneous particle distribution was partly overcome by the introduction of multi-threading. An efficient algorithm is proposed to handle data-exclusive access of the shared-memory by multi-threads on a compute node. The developed parallel DPM model has been successfully used to simulate many important applications such as bubbling fluidized bed, granular Rayleigh-Taylor instability, and particle swarm dynamics.

AB - Most of the CFD codes are parallelized under distributed-memory parallel computing environments. For this reason, many attempts have been made to develop parallel DPM/DEM algorithms for dense particulate flows under the same parallel architecture. For such a development, it is very difficult to achieve efficient load balancing of processors due to the heterogeneous particle spatial distribution that often characterizes dense particulate systems. In this work, parallel CFD-DEM algorithms have been developed under the distributed memory environment with a fluid flow solver based on finite volume method and arbitrary 3D unstructured meshes. An implicit two-phase coupling scheme was proposed to enhance numerical stability for complex dense particulate flow problems. Parallelization-generated numerical difficulties such as void fraction calculation, two-phase momentum exchange, and contact force calculations for particles at irregular and arbitrary partition boundaries were efficiently addressed. The load-balancing difficulty due to heterogeneous particle distribution was partly overcome by the introduction of multi-threading. An efficient algorithm is proposed to handle data-exclusive access of the shared-memory by multi-threads on a compute node. The developed parallel DPM model has been successfully used to simulate many important applications such as bubbling fluidized bed, granular Rayleigh-Taylor instability, and particle swarm dynamics.

KW - Computational fluid dynamics

KW - Dense particulate flow

KW - Discrete element method

KW - Load balancing

KW - Parallel algorithm

UR - http://www.scopus.com/inward/record.url?scp=84906043914&partnerID=8YFLogxK

U2 - 10.1016/j.ces.2014.07.043

DO - 10.1016/j.ces.2014.07.043

M3 - 文章

AN - SCOPUS:84906043914

SN - 0009-2509

VL - 118

SP - 221

EP - 244

JO - Chemical Engineering Science

JF - Chemical Engineering Science

ER -