Distributed Finite-Element Analysis on Network of Workstations—Algorithms

Advanced Analysis of Steel Frames Using Parallel Processing and Vectorization

“…However, excellent reviews of past research efforts in applications of substructuring, parallel processing, and vectorization in computational mechanics are available. 2,6,15,26,28,60 Much of the past research in applications of substructuring and parallel processing have emphasized large-structure design optimization, 3,7,49,51 the elastic analysis of large structures, 4,5,8,9,31,32 nonlinear transient analysis, 35,53 nonlinear finite-element analysis, 36 and implementation of parallel equation solvers. The neural dynamics model has been implemented recently in the optimal design of large structures.…”

“…8,9 Furthermore, at the time of this writing, multiprocessor computers are available to the public (not just researchers at supercomputing centers). In addition, there are C, C++, and Fortran compilers capable of creating multithreaded code for multiprocessor PCs.…”

Comparison of Preconditioned Conjugate Gradient Method and LDLT Factorization for Structural Analysis

“…5,6 The PCG method also has become the most popular method for parallel-processing implementation. 1,11,14 For problems with multiple right-hand-side (RHS) vectors, theoretically, the LDL T becomes comparatively more competitive, since we only need to factorize the matrix K once and only forward and backward substitutions need to be performed as many as the number of RHS vectors. These theoretical comparisons and the increasing popularity of the PCG method are the motivation to compare the efficiency of the PCG method and a sparse LDL T factorization scheme for solving a sparse system of linear equations in large-scale structural analysis with both single and multiple RHS vectors.…”

“…K; for k = 1, ..., n do begin1. Select coefficients based on the updated matrix K k−1 :•K = K k−1 + E k−a vector v;• Store the rejected entries in T k−1 * k , so thatK k−1 * k = v + T k−1 * k ;…”

Possibilities of Workstation Clusters for Parallel Finite–Element Analysis

“…Since the computing environment with many workstations and networks is widespread, the realization of parallel finite-element analysis is expected using workstation clusters. In the last several years, some articles 1,2,5,10,12,15 have been devoted to the study of parallel finite-element analysis using several workstations. We have developed some parallel finite-element analysis systems 8,9 on a workstation cluster, and it has been found that the critical problem for workstation cluster systems is the limitation of data transmission capacity through the network for large-scale finite-element problems.…”