摘要
通过分析现有的多学科设计优化中任务分解方法(枚举法、聚类识别法和分支定界法)的特点,指出了现有方法的不足.提出了将遗传算法应用于优化任务的分解问题,给出了具体的算法描述和详细的任务分解算法流程,并分析总结给出了该算法的优点:①遗传算法对搜索空间没有任何要求,因此对函数关系矩阵(FDT,Function Dependence Table)也没有任何要求;②遗传算法是一种随机迭代方法,不需要估计初值;③遗传算法同时对一组解进行搜索,大大提高了搜索速度,在保证计算精度的基础上得到全局最优解.最后还以齿轮减速器优化问题为例,将遗传算法应用于上述问题的任务分解过程,得到了较为满意的分解结果,并从计算方案次数的角度定量地比较了所提出方法与现有方法的区别,从而证明了该方法的正确性和优越性.
Based on the analyzing of current methods of solving task decomposition problem in MDO, such as enumerate algorithm, clustering identification method, branch and bound method（BBM） , their disadvantages were pointed out. Then, the description and detailed flow were delivered. The advantages of GA were described as follows： first, there is no special request for search space and function dependence table （FDT） ; second, GA is a stochastic iterative method, so the initial value is unnecessary; third, GA searches from a group of solutions simultaneity, which improves the search speed and gets global optimal solution with higher accuracy. Finally, the gear reducer optimization problem was taking as an example and GA was used. The analyzing result was acceptable. The calculation quantity was reduced rapidly. The correctness and advantage of the GA were proved when compared the calculation quantity with the methods mentioned before.
出处
《北京航空航天大学学报》
EI
CAS
CSCD
北大核心
2009年第6期673-677,共5页
Journal of Beijing University of Aeronautics and Astronautics
关键词
多学科设计优化
遗传算法
任务分解
muhidisciplinary design optimization （MDO）
genetic algorithm
task decomposition
作者简介
陈云霞（1977-），女，安徽怀宁人，副教授，chengyunxia@buaa．edu．cn．
