VECPAR'06 - Seventh International Meeting on High Performance Computing for Computational Science |
Performance Assessment for a Family of Continuum-based Shell Elements for Nonlinear Finite Element Analysis in Shared-memory Systems
Elizabeth Silva (LAMCE-PEC-COPPE/UFRJ)
This work presents an assessment of some proposals for improvements on members of the C0 quadrilateral continuum-based (CB) shell elements originally developed by Belytschko et al., 1984, 1989, 1992, for explicit type nonlinear finite element analysis (NL-FEA). These elements are extensively available in high performance commercial codes that implement nonlinear finite element analysis, notably for industrial applications where computational models of practical interest may typically comprise hundred thousand elements as a rough measure of problem dimensions. Reportedly, members of this element family perform very differently, considering overall computational performance, accuracy and robustness. In this work, proposals for new members in this family are derived and each one of these members is assessed for solution quality and computational costs involved while addressing several benchmark problems. A special implementation technique, namely the “lamina-by-lamina” approach, is introduced for members of this family. This approach enables improvements and robustness for every element in the family at marginal additional cost. Computational performance was evaluated in shared memory PVP (parallel vector processor) systems, such as T90 and SV1, installed respectively at CESUP/UFRGS and NACAD/COPPE/UFRJ.
Large Scale Simulations in all areas of Engineering and Science, Numerical Methods (PDE), finite element implementation in hpc environments for large scale nonlinear dynamic analysis
Logos Universidade Federal do Rio de Janeiro - Coordenação dos Programas de Pós-graduação de Engenharia Instituto Nacional de Matemática Pura e Aplicada Rio de Janeiro | Brazil | 2006 | July | 10 11 12 13