摘要
目标层解分析是一种层次化、多层系统设计优化方法。为了确保求解多学科设计优化各子问题的可行性,提高求解效率,应用增广拉格朗日惩罚函数松弛化方法,对目标层解分析的内外层嵌套式求解策略进行改进,通过对内层循环的惩罚函数松弛化来减少内层循环病态子问题的求解计算时间,当内层循环获得收敛之后,外层循环更新惩罚权重来获得可行解。并置设计次数由10到1000的具体实例来对比分析各种惩罚函数对求解效率的影响。由实验可得,应用增广拉格朗日惩罚函数松弛化方法求解,计算权重得到减小,迭代次数减少到二次惩罚函数法的2%。
Analytical target cascading is a hierarchical method used for multilevel-system design optimization. To ensure better feasibility and higher efficiency of the sub problems brought out in the process of muhidisciplinary design optimization, an augmented Lagrangian relaxation method is used. By using this method, a typical nested solution strategy of analytical target cascading has been modified. The computational time for solving the inner loop illconditioning sub problems is reduced through the relaxation of penalty function of inner loop. After the convergence of inner loop is gained, the outer loop will acquire feasible solution by updating the penalty weight. Further, an experiment with one example ranging from 10 to 1000 is designed to compare the impacts on solving efficiency which are brought out by different penalty functions. The result shows that the augmented Lagrangian relaxation method gains less penalty weight and the iteration times will be reduced to 2 percent of that of Quadratic penalty function method.
出处
《计算机仿真》
CSCD
2008年第11期195-199,共5页
Computer Simulation
基金
面向航天行业的可定制PLM系统(2007AA040601)
关键词
多学科优化
目标层解分析
增广拉格朗日松弛化
惩罚函数
Muhidisciplinary design optimization
Analytical target cascading
Augmented lagrangian relaxation
Penalty function