Diversity enhanced particle swarm optimization algorithm and its application in vehicle lightweight design

Abstract

Particle swarm optimization, a widely used metaheuristic algorithm, mimics the cooperation behavior among species. The PSO algorithm has become a new trend owing to its simplicity and strong optimization capacity. However, premature convergence problem is also a serious issue for PSO comparable with other evolutionary algorithms. Diversity loss is generally known as one of the major causes. For enhancing the diversity of swarms during optimization procedure, an improved PSO algorithm named OLAR-PSO-d is proposed, which incorporates design of experiment technique as well as adaptive reset operator into standard PSO. The OLAR-PSO-d algorithm is compared with other 10 heuristic algorithms. The numerical experiments’ results demonstrate the priority of OLAR-PSO-d both in optimization ability and algorithm stability. The proposed algorithm is also used in a vehicle lightweight design problem. The auto-body achieves 20.25 kg weight reduction with meeting all the performance requirements of crashworthiness.