Practical strategy for concurrent substructure analysis

Substructure modeling of general large-scale structures requires repetitive computations coming from various combinations of subcomponent or local structure modeling changes. Parallel algorithms for finite element analysis proposed in the literature have been targeted only toward high speedup and do...

Full description

Saved in:
Bibliographic Details
Published inComputers & structures Vol. 54; no. 5; pp. 939 - 944
Main Authors Synn, S.Y., Fulton, R.E.
Format Journal Article
LanguageEnglish
Published Oxford Elsevier Ltd 03.03.1995
Elsevier Science
Subjects
Exact sciences and technology
Fundamental areas of phenomenology (including applications)
Physics
Solid mechanics
Static elasticity
Static elasticity (thermoelasticity...)
Structural and continuum mechanics
Finite element method
Parallel programming
Partition method
Numerical method
Substructure
Domain decomposition
Structural analysis
Online AccessGet full text

Cover

More Information
Summary:Substructure modeling of general large-scale structures requires repetitive computations coming from various combinations of subcomponent or local structure modeling changes. Parallel algorithms for finite element analysis proposed in the literature have been targeted only toward high speedup and do not consider recomputations. In this paper, a practical computational strategy is proposed with the development of a theoretical formulation to predict the performance of a parallel matrix solver for two massively parallel machines (BBN/Butterfly, KSR1) and an efficient processor assignment algorithm to cover the initial computation stage and the recomputation stage of an irregularly shaped substructure model.
ISSN:0045-7949
1879-2243
DOI:10.1016/0045-7949(94)00385-G