约束处理策略对遗传算法优化性能的影响  

Effect of constraint handling strategies on optimum property of genetic algorithm

作　　者：孙丕忠[1] 夏智勋[1] 赵建民[1] 

机构地区：[1]国防科技大学航天与材料工程学院,长沙410073

出　　处：《固体火箭技术》2005年第4期235-237,共3页Journal of Solid Rocket Technology

摘　　要：根据遗传算法处理带约束优化问题时常用的拒绝策略或惩罚策略的优势与不足,提出一种“自适应拒绝+惩罚”相结合的约束处理策略,分别采用3种不同约束处理策略完成了空中发射运载火箭的优化设计,并对优化结果进行了比较。结果表明,对于约束严的优化问题,拒绝策略收敛速度较慢;惩罚策略中罚函数的形式对优化结果影响较大;“自适应拒绝+惩罚”策略拒绝所有约束破坏程度超过容许范围的非可行解,并对约束破坏程度在容许范围内的非可行解进行惩罚,可大大改善遗传算法处理严约束优化问题的优化性能。Based on the advantages and disadvantages of refusal strategy or penalty strategy used to handle constraint problem by genetic algorithm,a hybrid strategy combining adaptive refusal strategy and penalty strategy was put forward. The optimum design for the air-launched vehicle was finished by using three strategies mentioned above ,and the optimum results were compared. The results show that refusal strategy used for strict constraint problem has a disadvantage of slow convergence velocity, and penalty function form has great effect on the optimum results. The hybrid strategy combining adaptive refusal strategy and penalty strategy refuses all infeasible solutions when constraint failure extent is beyond admissible range, and punishes the infeasible solutions when constraint failure extent is within the admissible range. So the hybrid strategy can significantly improve optimum property of genetic algorithm for strict constraint problem.

关 键 词：固体推进剂火箭 优化设计 遗传算法 约束 

分 类 号：V421.1[航空宇航科学与技术—飞行器设计]