Computational fluid dynamics applications on parallel-vector computers: Computations of stirred vessel flows
Publication date
2002
Document type
Research article
Author
Organisational unit
Universität Erlangen-Nürnberg
Scopus ID
ISSN
Series or journal
Computers & fluids : an international journal
Periodical volume
31
Periodical issue
1
First page
69
Last page
97
Part of the university bibliography
Nein
Abstract
Advances in parallel-vector computers have resulted in a computer architecture that is able to provide the computer power needed for large-scale flow predictions. This is demonstrated in the present paper by applying two parallel-vector computers, one with shared memory and the other with distributed memory, to the computation of stirred vessel flows for a wide range of Reynolds numbers, including laminar and turbulent regimes. The governing equations for unsteady fluid flows, together with appropriate boundary conditions, and the adopted solution procedure are summarized. The implementation of the numerical algorithm into a computer program is outlined. The strategies employed for vectorization and parallelization are described and emphasis is placed on four different parallelization strategies partially taking advantage of a shared-memory architecture. Results of flow predictions inside a vessel stirred by a Rushton turbine are presented. The power characteristic expressed as Newton number versus Reynolds number of the stirrer is well predicted. It is shown that computational fluid dynamics simulations provide reliable results and yields a detailed and accurate picture of the complex flow phenomena observed in stirred-tank reactors. The performance of the simulation code on the two parallel-vector computers was measured and the reasons for the differences in the performances of the two architectures are discussed in detail. © 2001 Elsevier Science Ltd. All rights reserved.
Version
Not applicable (or unknown)
Access right on openHSU
Metadata only access