![]() |
Computational & Technology Resources
an online resource for computational,
engineering & technology publications |
Civil-Comp Proceedings
ISSN 1759-3433 CCP: 68
DEVELOPMENTS IN ENGINEERING COMPUTATIONAL TECHNOLOGY Edited by: B.H.V. Topping
Paper X.2
A Parallel Environment for Three-Dimensional Finite Element Analysis C.O. Moretti+, J.B. Cavalcante Neto#*, T.N. Bittencourt+ and L.F. Martha#
+Department of Structural and Foundation Engineering, University of Sao Paulo, Brazil
Full Bibliographic Reference for this paper
C.O. Moretti, J.B. Cavalcante Neto, T.N. Bittencourt, L.F. Martha, "A Parallel Environment for Three-Dimensional Finite Element Analysis", in B.H.V. Topping, (Editor), "Developments in Engineering Computational Technology", Civil-Comp Press, Edinburgh, UK, pp 283-287, 2000. doi:10.4203/ccp.68.10.2
Abstract
A three-dimensional parallel environment for finite element
method (FEM) analysis is presented. This environment is
comprised by integrated computer programs, each of one
responsible for a different task: pre-processing, mesh
generation, structural analysis and post-processing. In this
work, this entire system is presented, with more emphasis to
volumetric mesh generation and FEM analysis that could be
performed in a parallel way. A program named FRANC3D (3D Fracture Analysis Code) is used in the preprocessing
step. The volumetric mesh is generated using a algorithm that combines an advancing front technique with a
recursive spatial decomposition technique, in this case an
octree, to define the internal nodes, element sizes, and mesh
transition. A strategy to parallelize this algorithm is also
presented. An existing finite element method program, called
FEMOOP (Finite Element Method - Object Oriented Programming) has been adapted to implement the parallel features. The parallel analysis can be performed using two different techniques: a domain decomposition technique or an element-by-element scheme. Both are presented in this work. Finally, analyses of three-dimensional models are used to test the performance and reliability of this parallel system.
purchase the full-text of this paper (price £20)
go to the previous paper |
|