乘波体MDO子学科动态代理建模方法  

Modeling Method of Subsystem Dynamic Surrogate in Waverider MDO Progress

作　　者：王烁[1] 姜欢[2] 李萍[1] 陈万春[1] 

机构地区：[1]北京航空航天大学宇航学院,北京100191 [2]北京临近空间飞行器系统工程研究所,北京100076

出　　处：《战术导弹技术》2011年第6期11-17,59,共8页Tactical Missile Technology

摘　　要：多学科设计优化中应用代理模型可以较好地解决系统分析精度和效率的矛盾,动态代理技术可以根据需要在优化的过程中不断提高代理模型精度.利用三种动态代理方法——最小响应面、最大期望提高和多点加点准则,分别对一维、二维函数和一个高维的协同优化问题中外形/气动学科构造了动态Kriging代理模型,对比了三种加点准则的动态代理效果,并分析了其特点和应用条件.结果表明,最小响应面加点准则是最适合于协同优化中学科优化的动态代理方法.The application of surrogate models in Multidisciplinary Design Optimization（MDO） can mediate contradictions between precision and efficiency of calculations.The technology of dynamic surrogate iteratively enhances the precision of surrogate models as required.Three methods of dynamic surrogate,the minimum response surface,the maximum expected improvement and multiple-points criterion are applied to one and two dimensional functions and a high-dimensional MDO progress managed in Collaborative Optimization（CO） strategy,in which the configuration/aerodynamic subsystem are dynamically surrogated based on Kriging model.The effects of application of all criterions are compared as well as their applicability discussed.Results show that the minimum response surface criterion is the best one suited with dynamic surrogate of subsystem.

关 键 词：多学科设计优化 协同优化 KRIGING 动态代理 加点准则 

分 类 号：V423.9[航空宇航科学与技术—飞行器设计]