Application of MPI in Displacement Based Multilevel Structural Optimization

Abstract

The weight optimization of trusses using Displacement based Multilevel Structural Optimization (DMSO)^1-4 is investigated. In this approach, the optimization task is subdivided into a single system and multiple subsystems level optimizations. The approach is considered to be efficient for large structures, since parallel computing can be utilized in the different optimization levels. A 240-element truss was tested for efficiency improvement in DMSO on both, a network of SUN workstations using the MPICH implementation of the Message Passing Interface (MPI) and on a faster Beowulf cluster using the LAM implementation of MPI. Parallel processing was applied to the subsystems level, where the derivative verification feature of the optimizer NPSOL was first used in the optimizations. Even with parallelization, this resulted in large runtimes. Then, the optimizations were repeated without derivative verification, and the results were compared. They were consistent and showed no need for derivative verification, giving a large increase in efficiency in the DMSO algorithm. They also showed that very little improvement in performance was obtained by parallelization once the derivative verification feature in NPSOL was turned off.