A Parallel Contact Algorithm for Finite Element Analysis of Shell Structures

A parallel contact algorithm has been developed for nonlinear dynamic finite element analysis of shell structures. The contact algorithm accounts for sliding with friction. The contact algorithm has been designed to enable individual processors to dynamically build local contact domains during execution and to efficiently exchange data between processors as the contact problem evolves. The performance of the parallel contact algorithm has been studied by executing the program on 128 processors of a Symult Series 2010 distributed-memory multiprocessor computer: it has been found that computational load imbalance rather than interprocessor communication overhead is the primary obstacle to achieving high efficiency.