A New Scalable Parallel Finite Element Approach for Contact-Impact Problems
Har, J.(1998) A New Scalable Parallel Finite Element Approach for Contact-Impact Problems, Doctoral Thesis, Georgia Institute of Technology, Atlanta.
The parallel processing using a multiple interaction multiple data (MIMD) huge parallel computer is a promising approach to the solution of contact-impact problems which requires several hundred hours of CPU time on pipelined vector supercomputers in order to simulate structural behaviors. The goal of this proposed research is to advance contact-impact analysis through improved nonlinear mechanics combined with the development of new parallel algorithms to achieve better accuracy and efficiency. The underlying focus of the method is on the effectiveness and efficiency for inclusion in future finite element systems implemented in parallel computers. For a scalable parallel processing, a domain decomposition algorithm and an interprocessor communication minimization are explored in the strategy. A prototype code, denoted GT-PARADYN, is developed and implemented to investigate the behavior of nonlinear shell structures using the IBM SP2, a distributed-memory multicomputer. A new and improved scalable parallel contact-impact finite element method is presented for implementation on future parallel computers.