Reducing computational expense associated with evolutionary detailed design

This paper provides a comparison of Adaptive Search methods for the reduction in computational expense associated with the optimisation of highly dimensional structural design problems. A sequential method for Dynamic Shape Refinement is presented. This technique utilises problem representations of varying complexity, from a coarse representation which is gradually refined until the desired level is achieved. This paper also presents the Injection Island Genetic Algorithm (iiGA). The technique utilises a number of levels of varying resolution grids, each evolving as a separate process. The iiGA method is further extended by the utilisation of a dynamic method of plate representation which performs on-line replacement of the lower resolution subpopulations with finer representations. The results show a significant improvement over those obtained utilising single level representations in terms of a reduction in computational expense, whilst also achieving significant improvements in design performance.