A PARALLEL ALGORITHM FOR EIGENVALUE PROBLEMS OF LARGE-SCALE STRUCTURES
一种大型结构特征值问题的并行解法

In this paper, a parallel algorithm for calculating natural frequenciesand mode shapes of large structural systems is presented. In the process of subspace iteration, the method uses only substructure stiffness and mass matrices to carry out condensation concurrently and to get new iterative basis vector finally until the convergence is realized. The program of this algorithm has been realized on ELXSI-6400 parallel computer of Xian Jiao-tong university. The computational results show the computing time and memory space can be saved on large scale.