Asynchronous Swarm Structural Optimization of Satellite Adapter Ring

Abstract

The shape and size of a satellite adapter ring are optimized in this paper using particle swarm optimization and an in-house finite element program. For a realistic high-fidelity modeling of the adapter ring, a commercial preprocessor is used. However, optimization of a high-fidelity model inevitably involves longer computation time; therefore, parallelization is essential to shorten the total elapsed time. Furthermore, fitness evaluation time by the finite element simulation differs from case to case, which unavoidably causes idling of some of parallel computing resources. To overcome the inefficiency induced by the irregularities of individual fitness evaluation, an asynchronous parallel swarm algorithm is developed and applied to the structural optimization of the adapter ring with 19 design variables. The adapter ring’s total mass is reduced to 33.4 kg from 35.9 kg within 202 min. The asynchronous version also performs almost twice better compared with the synchronous one. The methodology proposed in this paper offers an attractive advantage for complex structural design, both in aspect of computational time and global search.