Multimodal Problems, Premature Convergence versus Computation Effort in Dynamic Design Optimization

Dynamic Design Optimization (DDO) of various engineering problems exhibit multiple optima in the feasible domain. Such problems can be posed as complex multimodal optimization problems. Application of traditional optimization techniques to such problems is computationally expensive with a high risk of getting trapped into a local optimum. Similarly, Genetic Algorithms (GA) suffer from premature convergence and weak exploitation capabilities. In this paper, a Niche Hybrid Genetic Algorithm (NHGA) is proposed for optimizing continuous multimodal models. This architecture of Hybrid Algorithms (HAs) organically merges Niche Techniques and Nelder-Mead’s Simplex Method into GA. The NHGA is executed in the global exploitation and local exploration. In the former, a simplex search is performed in the potential niches for a quick evaluation of the promising search zones following the generation of dynamic niche sets by a Clearing Method (CM). A further simplex search (SS) is subsequently executed in the exploitation phase for a quick location of a global optimum in the located most promising zone and an inverse operator introduced to maintain population diversity. The proposed technique effectively alleviates premature convergence and improves the weak exploitation capacity of GAs. To emphasize the application of the algorithm, numerous multi-modal functions have been experimented with, and a 5degree of freedom vehicle suspension system optimized. Analytical results indicate the potential of the approach in DDO of mechanical systems. KeywordsDesign, Optimization, Convergence, Computation