摘要
为了在复杂产品早期开发阶段更好的使用解析目标分流法(ATC),分析了基于不同罚函数的ATC方法的计算特性。给出了ATC方法的一般公式,比较了3种ATC方法的优缺点,以常用的几何规划算例为例,研究了初值、初始罚参数、收敛准则等因素对计算精度和计算花费的影响。计算结果表明,ATC是一种稳定性好、计算效率高的多学科设计优化(MDO)方法,并且基于增广拉格拉日函数的ATC方法的总体性能比其它两种方法要好一些。
In order to use Analytical target cascading(ATC) better at the early stage of the development of complex products,the ATC based on different penalty functions are analysed.The general ATC formulation is presented first.Then the advantages and disadvantages of three ATC methods are compared.In the end,a commen geometry programming problem is used to study how some factors such as initial values,initial penalty parameters and convergence criterion influence the precision and cost of compution.Results show that ATC is a MDO method that has good robustness and high computing efficiency and the ATC based on augmented Lagrangian function has better performance than the other two ATC methods.
出处
《计算机工程与设计》
CSCD
北大核心
2010年第11期2564-2567,共4页
Computer Engineering and Design
基金
国家863高技术研究发展计划基金项目(2006AA04Z119)
关键词
多学科设计优化
解析目标分流法
计算特性
罚函数
复杂系统
multidisciplinary design optimization
analytical target cascading
calculation characteristics
penalty functions
complex system