DESIGN AND IMPLEMENTATION OF POLYNOMIAL-PRECONDITIONED EBE-PCG PARALLEL ALGORITHM IN CLUSTER

The computation aiming at large-scale structural dynamic problems, which is difficult and expensive to be implemented on serial computers, has become more and more popular. Based on the EBE scheme disassembling global computation to local elements and the SBS scheme allocating storage and tasks to the subdomains plotted by domain decomposition, a coarse granular EBE-PCG parallel algorithm is designed and then implemented in cluster. Jacobi-preconditioned matrix and polynomial-preconditioned matrix are introduced in PCG iterative method and their efficiencies are compared. The numerical examples of the dynamic response analysis of jeep frame and cantilever beam show that the EBE-PCG algorithm can not only greatly enlarge the solution scale but also obtain high parallel efficiency. It is an efficient distributed coarse granular parallel algorithm suitable for large-scale structural analysis and computing in cluster.