摘要
与其它直接处理法相比 ,约束随机搜索方向法和变量轮换法的算法具有简单明了、编程简单、容易实现等特点 ,特别适用于目标函数的梯度计算复杂或不能计算的场合。本文综合这两种方法的优点 ,提出了一种新型算法 ,其搜索方向在设计空间完全均布 ,最大搜索方向数与设计变量空间的维数成幂函数关系 ,并介绍了这种新型算法的迭代步骤。通过实例分析 。
Compared to other widely used direct search method,random search methods and cyclic variable method have the advantages of simpleness and clear,easily programmed and smple implementation,especially when the gradient is burdensome or impossible to calculate.In this paper,a new direct algorithm is proposed to enhance the performance of two methods,its search direction cover uniformly the whole design space,the maximal number of search,directions is a power function of the dimensions of design space,and the iterative step of new method was mentioned.Several examples indicated that this new algorithm has more calculated precision and stability.
出处
《机械设计与研究》
EI
CSCD
2001年第1期13-15,36,共4页
Machine Design And Research
关键词
优化设计
随机搜索法
变量轮换法
约束优化
Optimal design
optimization
random search method
cyclic variable method
