Parallel implementation of macro-scale pseudo-particle simulation for particle–fluid systems

As a particle method, macro-scale pseudo-particle modeling (MaPPM) is an effective approach applied to micro-scale simulation of particle–fluid systems. In this paper, a parallel algorithm for macro-scale pseudo-particle modeling based on spatial decomposition (SD) is presented. The parallel impleme...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Computers & chemical engineering Ročník 29; číslo 7; s. 1543 - 1553
Hlavní autori: Wang, Xiaowei, Guo, Li, Ge, Wei, Tang, Dexiang, Ma, Jingsen, Yang, Zhangyuan, Li, Jinghai
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: Elsevier Ltd 01.06.2005
Predmet:
Dynamic load balancing
Macro-scale pseudo-particle modeling
Parallel algorithm
Particle–fluid systems
Parallel algorithm
Macro-scale pseudo-particle modeling
Dynamic load balancing
Particle–fluid systems
ISSN:0098-1354, 1873-4375
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:As a particle method, macro-scale pseudo-particle modeling (MaPPM) is an effective approach applied to micro-scale simulation of particle–fluid systems. In this paper, a parallel algorithm for macro-scale pseudo-particle modeling based on spatial decomposition (SD) is presented. The parallel implementation utilizes MPI as the programming environment. Due to movement of particles during simulation, the parallelization of MaPPM may suffer from load imbalance and attendant performance degradation. Recursive Coordinate Bisection (RCB) is adopted to partition the whole computational domain in a dynamic fashion to balance the workload in processors. The Shift scheme is modified to meet the communication requirement in the dynamic partition. The parallel approach was applied to simulation of bubble behavior in gas–solid fluidized beds with different system sizes to test its performance. The computations were conducted on cluster of workstations (COW). Experimental results show that the algorithm has a good scalability. With dynamic load balancing (DLB), the parallel efficiency can be improved by up to 8%. To sum up, it was a successful implementation for the parallelization of macro-scale pseudo-particle modeling.
Bibliografia:ObjectType-Article-2
SourceType-Scholarly Journals-1
ObjectType-Feature-1
content type line 23
ISSN:0098-1354
1873-4375
DOI:10.1016/j.compchemeng.2004.12.006