The Research On Parallel Computation Of Finite Element Structural Analysis Based On MPI Cluster

Network parallel computing is one of the most attractive forward position subjects for parallel computing in recent. The finite element method (FEM) is a very effective method used for structure analysis nowadays. The technology of parallel computing is introduced into the structure finite element analysis to multiply the size of problem to be solved, accelerate the analysis and facilitate the large-scale application in structural engineering. In combination with the key project of Shanghai Educational Committee, parallel algorithms for FEM under distributed networked workstation cluster parallel environment are studied and implemented in this dissertation and applied to analysis of steel gate structure.Based on the networked workstation cluster parallel environment of MPI, the main contents of this dissertation are as follows:(1) A networked workstation cluster parallel computing environment is set up. The basic concepts of parallel computing are described. MPI programming is presented.(2) The load-balanced multiplication of a large sparse matrix with vector on workstation cluster is considered. A method that combines fast load balancing with efficient message-passing techniques to alleviate computation and inter-node communications is presented. Moreover, I/O overhead can be efficiently amortized through I/O latency hiding and overall load balancing. The preconditioned conjugate gradient method (PCGM) is an efficient iterative method used to solve the finite element systems of equations with symmetric positive definite system matrices. The algorithm of PCGM is parallelized. The storage scheme is analyzed in detail. Optimization techniques for the sparse matrix vector multiplication are adopted in programming. The experiment result shows that the designed parallel algorithm has high speedup and good efficiency on the high performance workstation cluster.(3) A parallel preconditioned conjugate gradient iterative algorithm for finite element problems with coarse-mesh/fine-mesh formulation is presented. An efficient preconditioner is derived from the multigrid stiffness matrix. Example of solution of simple linear elastic problem on rectangular grids is presented and parallel performance is discussed. The results show higher speedup and efficiency. The algorithm is efficient for parallel computing.(4) The parallel finite element method using domain decomposition technique is adopted to the distributed parallel environment of workstation cluster. The parallel algorithm is presented for solving the conjugate gradient method with element based domain decomposition. Model of dam structural analysis problem is solved. The parallel performance is analyzed.(5) On the basis of mode synthesis analysis, parallel algorithm of solving large-scale structural eigenproblem is presented. The eigenmode of the structure is solved using subspace iterative parallel method. The substructure subspace iterative method is implemented using the stiffness matrix and mass matrix of the substructures without assembling the stiffness and mass matrix of whole structure. The numerical results show that this parallel algorithm is effective for large-scale structure eigenproblem.