Practical high-velocity impact simulation for ceramic and metal multiple plate with load balancing parallel algorithm

摘要

We successfully demonstrated the fracture of brittle materials with nodes separation scheme [1]. However, the node separation scheme which generates new nodes and contact surfaces continuously results in the huge amount of calculation load such that a simulation usually takes more than a week. In this paper, the MPI (Message Passing Interface) parallelization method of non-linear Lagrangian finite element especially for the brittle materials is described. Particular methods to enhance parallel performance are described in details. Some of those are the efficient data structure and communication method to handle new nodes and contact surfaces, static and dynamic load balancing technique for balancing the calculation load of each processor. To evaluate the parallel performance and accuracy of the program, two applications are demonstrated with 32 processors: glass-metal bar impact problem and a long rod impact into ceramic-metal multiple layered plates.