• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2015, Vol. 51 ›› Issue (11): 133-141.doi: 10.3901/JME.2015.11.133

• 数字化设计与制造 • 上一篇    下一篇

基于设计空间缩减的多学科协同优化新方法

金霞1, 2, 段富海1, 辛大志2, 母刚1   

  1. 1.大连理工大学机械工程学院;
    2.朝阳师范高等专科学校
  • 出版日期:2015-06-15 发布日期:2015-06-15
  • 基金资助:
    国家自然科学基金(50905022)和航空科学基金(20130863006)资助项目

Novel Multidisciplinary Collaborative Optimization Method Based on Design Space Decrease

JIN Xia1, 2, DUAN Fuhai1, XIN Dazhi2, MU Gang1   

  1. 1.School of Mechanical Engineering, Dalian University of Technology;
    2.Chaoyang Teachers College
  • Online:2015-06-15 Published:2015-06-15

摘要: 提出一种多学科协同优化设计新方法—设计空间缩减协同优化 (Design space decrease collaborative optimization,DSDCO)。根据子系统级优化结果,确定分解设计空间的平面方程,将设计空间合理分解为多个子空间,去除其中不可行设计子空间,将缩减后的设计空间传递到系统级优化。根据系统级优化在各个子空间的优化信息,择优选取系统级优化结果和下一次优化计算的设计空间,循环进行优化迭代计算,直至系统级优化值符合收敛条件。该方法通过缩减优化求解空间,不断更新系统级优化模型,将传统协同优化(Collaborative optimization,CO)中系统级非线性等式约束变换为只含有变量边界的线性不等式约束,解决了传统CO系统级求解困难的问题。DSDCO在变量有界的多学科设计优化(Multidisciplinary design optimization, MDO)问题中,对原始问题约束函数的凸性无要求,对优化迭代起始点的位置无要求。分别利用数值算例、减速器设计和弹簧设计三个典型算例,验证了DSDCO方法的正确性。

关键词: 多学科优化设计, 非线性约束, 设计空间, 协同优化

Abstract: A novel multidisciplinary collaborative optimization method – Design space decrease collaborative optimization (DSDCO) is presented. According to the subsystem-level optimization results, the equations of planes used to decompose design space are determined. Then, the current design space is divided into several subspaces reasonably. Some infeasible design subspaces are deleted and the rests are passed to the system-level optimization as the elements of the system-level variables space. The system-level optimization value in each subspace is analyzed, then the preferred system-level optimization results and the next iteration design space are chosen. The iteration process is continued until the system-level optimization solution is converged. In DSDCO, as the design space is updated, the system-level optimized mathematical function is corrected. The system-level nonlinear equality constraints in conventional collaborative optimization(CO) are turned into linear inequality constraints, i.e. variable boundaries, so the computation difficult problem in the conventional CO is avoided. For MDO problem with variable boundaries, DSDCO has no limit to the start point and constraint function convexity. Three typical examples of a numerical test problem, a gear reducer design problem and a spring design problem illustrate the capabilities of the DSDCO method.

Key words: collaborative optimization(CO), design space, multidisciplinary design optimization(MDO), nonlinear constraint

中图分类号: