摘要
针对协同优化的计算结构及计算困难 ,提出了学科间不一致信息概念 ,并据此形成系统级优化的超球设计空间。有效地解决了原始系统级优化问题不满足 Kuhn- Tucker条件造成的计算困难 ,并克服了已有梯度方法计算易发散、响应面方法计算量大的缺点。利用文中提出的方法对一齿轮减速箱进行了设计 ,实现了并行设计的目的 。
Collaborative Optimization divides the optimization problem into two levels: one system level optimization and N parallel disciplinary level optimizations, where N is the number of disciplines. The difficulty in collaborative optimization is that the system level consistency constraints, in general, make it impossible to satisfy the Kuhn Tucker condition, making the system level optimization unsolvable. There exist already two methods——gradient method and response surface method——for dealing with the difficulty of system level consistency constraints, but these two methods, in our opinion, are still not satisfactory. We propose a new concept of inter disciplinary inconsistency information for dealing with the above mentioned difficulty. Sections 1 and 2 discuss in much detail the two existing methods as compared with our new method based on inter disciplinary inconsistency information. The original system level consistency constraints are the N equalities in expression (2). Section 2 takes N =2 to simplify the description of our new method. When N =2, the system level constraints based on interdisciplinary inconsistency information are the two inequalities in expression (6). In expression (6), s is the radius of the hyper ball in the hyper ball system level design space constructed with the new concept of inter disciplinary information as its base. Using expression (6) to take into consideration system level constraints, we can guarantee the satisfaction of Kuhn Tucker condition and thus the solvability of system level optimization. Section 2 explains why our new method is better than the two existing methods; it is better than the gradient method essentially because it is straightforward and does not need to bother about the trouble some finite difference calculations needed by the gradient method; it is better than the response surface method essentially because the amount of computation does not show the very rapid increase with increasing N as is the case with r
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2003年第4期461-464,共4页
Journal of Northwestern Polytechnical University
关键词
协同优化
学科间不一致信息
超球子空间
collaborative optimization, inter disciplinary inconsistency information, hyper ball system level design space
