Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.638432
Title: Parallel processing procedures for finite element analysis
Author: Paw, G. F.
Awarding Body: University College of Swansea
Current Institution: Swansea University
Date of Award: 1995
Availability of Full Text:
Access from EThOS:
Abstract:
The objective of the research is to improve the solution capability of a finite element system by the use of parallel processing techniques. Initial research began with the transputer system at a time when the transputer was considered the ultimate parallel machine. The transputer based parallel program LINSUB was developed to solve linear static stress analysis problems using a modified Cholesky algorithm for substruturing. Further developments in this field were halted when it became clear that other parallel systems, namely, those concerned with workstation clusters, were moving rapidly ahead of the transputer in terms of performance and the transputer was rapidly becoming obsolete. Further work concentrated on parallel processing on workstation clusters in a local area network and the effectiveness of the Parallel Virtual Machine (PVM) used as the supportive parallel environment was investigated. This work was then extended onto a shared memory - MIMD system on the Silicon Graphics Power Challenge workstation. The program PARFEI (PARallel Finite Element Implementation) was developed for large linear static stress analyses and nonlinear analyses of elasto-plastic problems in particular, using a multifrontal approach based on domain decomposition techniques. An implementation based on a specially designed share memory (SM) synchronization concept was incorporated into the same program, so that PARFEI can run sequentially or in parallel using either PVM or SM. Performance tests were carried out on the workstation clusters using PVM and on the shared memory machine using PVM and SM. The results revealed that, in general, PVM performed better for the linear system and SM was the more efficient of the two when an incremental/iterative process was involved.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.638432  DOI: Not available
Share: