Vector and multithread computation of silencer performance prediction on a dual-processor PC workstation

G. Lou, T. W. Wu, P. Zhang, C. Y.R. Cheng

Research output: Contribution to journalArticlepeer-review

6 Scopus citations

Abstract

Silencers used in industry usually contain complex internal components such as thin baffles, internal tubes, perforated tubes, and bulk-reacting sound absorbing materials. The direct mixed-body boundary element method (BEM) is an ideal analysis tool because each component has its own surface attribute and can be meshed independently. A model can be easily created by assembling individual components together. However, the BEM computation over a broad frequency range is very time consuming, especially at high frequencies. To speed up the computation, a dynamic meshing scheme is adopted and the BEM computation at each frequency is vectorized as well as parallelized. The multi-processor x86architecture running Windows NT/2000 is selected as the computational platform mainly because of its popularity and low cost. Evaluation of the fundamental solutions is vectorized to speed up numerical integration. At the same time, parallelization is achieved via the multithread implementation. Both the matrix generation part and the matrix solution part are parallelized. Numerical results show that an overall parallel efficiency of around 0.9 can be achieved on a dual-processor PC workstation.

Original languageEnglish
Pages (from-to)61-70
Number of pages10
JournalEngineering Analysis with Boundary Elements
Volume26
Issue number1
DOIs
StatePublished - Jan 2002

Bibliographical note

Funding Information:
This research was supported by Nelson Industries, Inc.

Keywords

  • Acoustics
  • BEM
  • Muffler
  • Parallel computing

ASJC Scopus subject areas

  • Analysis
  • Engineering (all)
  • Computational Mathematics
  • Applied Mathematics

Fingerprint

Dive into the research topics of 'Vector and multithread computation of silencer performance prediction on a dual-processor PC workstation'. Together they form a unique fingerprint.

Cite this