• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2016, Vol. 52 ›› Issue (14): 79-105.doi: 10.3901/JME.2016.14.079

• 运载工程 • 上一篇    下一篇

基于计算试验设计与代理模型的飞行器 近似优化策略探讨*

龙腾12, 刘建2, WANGGGary3, 刘莉12, 史人赫2, 郭晓松2   

  1. 1. 北京理工大学飞行器动力学与控制教育部重点实验室 北京 100081,
    2. 北京理工大学宇航学院 北京 100081,
    3. 西门菲莎大学机电工程学院 素里 V3T 0A3 加拿大
  • 出版日期:2016-07-20 发布日期:2016-07-20
  • 作者简介:

    作者简介:龙腾(通信作者),男,1982年出生,博士,副教授。主要研究方向为飞行器总体设计、多学科设计优化理论与应用、飞行器协同控制与决策。

    E-mail:tenglong@bit.edu.cn;bitryu@gmail.com

  • 基金资助:
    * 国家自然科学基金(51105040, 11372036)、航空科学基金(2011ZA72003, 2015ZA72004)和北京理工大学基础研究基金(20130142008)资助项目; 20151028收到初稿,20160321收到修改稿;

Discuss on Approximate Optimization Strategies Using Design of Computer Experiments and Metamodels for Flight Vehicle Design

LONG Teng12, LIU Jian2, WANG G Gary3, LIU Li12, SHI Renhe2, GUO Xiaosong2   

  1. 1. Key Laboratory of Dynamics and Control of Flight Vehicle of Ministry of Education,Beijing Institute of Technology Beijing, Beijing 100081,
    2. School of Aerospace Engineering, Beijing Institute of Technology, Beijing, 100081,
    3. School of Mechatronic Engineering System, Simon Fraser University, Surrey, BC V3T 0A3, Canada
  • Online:2016-07-20 Published:2016-07-20

摘要:

现代飞行器设计优化中广泛应用高精度分析模型以提高设计可信度与综合性能,但是也带来了计算复杂性问题。为了有效缓解计算耗时的问题,基于计算试验设计与代理模型的飞行器近似优化策略成为研究热点。近似优化策略通过构造合理的近似模型引导优化过程快速收敛到最优解,从而达到降低计算成本,缩短设计周期的目的。根据广泛的文献调研,对飞行器近似优化策略的发展现状进行详细探讨。给出近似优化策略的定义、求解流程、特点以及关键技术。对计算试验设计方法、代理模型方法、精度校验与代理模型选择方法等技术进行综述。围绕静态与自适应两类近似优化策略,重点讨论典型的代理模型管理与更新策略与收敛准则。针对飞行器多学科设计优化问题,探讨近似优化策略与分解策略在求解效率与收敛性方面的技术特点。通过数值算例对各项关键技术的特点进行较详尽的对比分析与总结,并依托飞行器设计优化工程实例对近似优化策略的综合性能进行探讨,指出不同近似优化策略的适用范围。研究结果表明,飞行器近似优化策略在优化效率、全局收敛性以及鲁棒性等方面体现出较显著的优势,具有良好的工程应用前景。指出飞行器近似优化策略的未来研究方向。

关键词: 代理模型, 多学科设计优化优化, 计算试验设计, 近似, 全局优化, 优化策略, 飞行器设计

Abstract:

Although the wide use of high fidelity analysis models in modern flight vehicle design is beneficial to improving the design credibility and overall performance of flight vehicle systems, it also causes high computational cost and complexity. In order to alleviate the computational difficulty, approximate optimization strategies using design of computer experiments(DoCE) and metamodels for flight vehicles have become more and more popular, which construct reasonable approximation models to enable efficient convergence to the optimal solution with much less computational burden and shorter design cycles. An extensive literature survey of the state-of-the-art of approximate optimization strategies in the context of flight vehicle design is provided. The definition, solution process, features and key technologies of approximate optimization strategies are presented, and then the development of DoCE, metamodeling, accuracy assessment and metamodel selection, as well as corresponding typical methodologies, are reviewed. Moreover, metamodel management and updating schemes and termination criteria used in both static and adaptive approximate optimization strategies are specifically discussed. The efficiency and convergence behaviors of approximate optimization strategies for solving multidisciplinary design optimization (MDO) problems are analyzed via comparison with decomposition-based strategies. A number of well-known numerical benchmark problems are employed to discuss the characteristics of the aforementioned key technologies. Furthermore, the overall performance and applicability of different approximate optimization strategies are discussed through flight vehicle design applications. Comparative studies demonstrate that approximate optimization strategies show obvious advantages in optimization efficiency, convergence and robustness, which are important for engineering applications. Future research directions of approximate optimization strategies are given.

Key words: approximation, design of computer experiments, global optimization, metamodel, multidisciplinary design optimization, optimization strategy, flight vehicle design