航空学报 > 2024, Vol. 45 Issue (S1): 730583-730583   doi: 10.7527/S1000-6893.2024.30583

基于多重保真度的复杂空间缩减与优化方法

张澍1,2, 田阔1,2(), 郭聪1,2   

  1. 1.大连理工大学 工业装备结构分析优化与CAE软件全国重点实验室,大连 116024
    2.大连理工大学 工程力学系,大连 116024
  • 收稿日期:2024-04-01 修回日期:2024-04-23 接受日期:2024-05-20 出版日期:2024-12-25 发布日期:2024-07-05
  • 通讯作者: 田阔 E-mail:tiankuo@dlut.edu.cn
  • 基金资助:
    辽宁省人工智能领域科技重大专项(2023020702-JH26/101);国家重点研发计划(2022YFB3404700)

Complex space reduction and optimization method based on multiple fidelity model

Shu ZHANG1,2, Kuo TIAN1,2(), Cong GUO1,2   

  1. 1.State Key Laboratory of Structural Analysis,Optimization and CAE Software for Industrial Equipment,Dalian University of Technology,Dalian 116024,China
    2.Department of Engineering Mechanics,Dalian University of Technology,Dalian 116024,China
  • Received:2024-04-01 Revised:2024-04-23 Accepted:2024-05-20 Online:2024-12-25 Published:2024-07-05
  • Contact: Kuo TIAN E-mail:tiankuo@dlut.edu.cn
  • Supported by:
    Major Science and Technology Projects in the Field of Artificial Intelligence of Liaoning Province(2023020702-JH26/101);National Key Research and Development Program Project of China(2022YFB3404700)

摘要:

航天结构优化中常面临设计空间复杂、非线性程度高等问题,基于高保真度模型优化耗时长,难以在紧迫的研发周期内完成优化设计。对此提出了一种基于多重保真度的复杂空间缩减与优化方法。首先,针对优化问题建立高、中、低保真度的分析模型;进而,在初始设计空间中采样并基于低、中保真度数据构建代理模型,基于模糊C均值方法聚类,得到初始缩减空间;然后,在初始缩减空间内采样高保真度样本点,基于中、高保真度数据构建代理模型;最终,开展优化并根据响应值最优样本点的位置移动缩减空间,优化迭代得到最优解。为验证所提方法的有效性,针对测试函数算例和多级加筋壳工程算例开展研究。结果表明,对于测试函数,所提方法优化得到的最优解和全局最优解的相对误差分别降低了68.4%和44.4%;对于工程算例,所提方法优化得到的极限承载力相对传统单、变保真度优化方法分别提高了9.0%和14.7%。同时,相比前期工作提出的变保真度静态缩减空间方法,在达到相同极限承载力水平时计算耗时降低了41.5%。综上所述,所提优化方法可充分利用多重保真度数据,对复杂耗时的航天结构工程优化问题具有优化效率高、全局寻优能力强等优势。

关键词: 结构优化, 空间缩减, 多重保真度, 多级加筋壳, 模糊聚类

Abstract:

Aerospace structural optimization often faces the problems of complex design space and high nonlinearity characteristic. However, the optimization based on high-fidelity model is time-consuming, making it difficult to complete the optimization design in the tight research and development cycle. A complex space reduction and optimization method based on multiple fidelity model is proposed. Firstly, the high-fidelity, medium-fidelity analysis models are constructed. Then, the variable-fidelity surrogate model is constructed based on the low-fidelity and medium-fidelity data by sampling in the original design space, and the clustering is carried out based on Fuzzy-C means to obtain the initial reduction space. Then, the high-fidelity sample points are sampled in the initial reduction space, and the variable fidelity surrogate model is constructed based on the medium-fidelity and high-fidelity data. Finally, the optimization was carried out and the reduced space was moved according to the position of the sample points with the best response value, and the optimal solution was obtained through optimization iteration. To verify the effectiveness of the proposed method, the test function example and an engineering example of the hierarchical stiffened shells are carried out. The example results show that the relative errors of the optimal solution and the global optimal solution obtained by the proposed method is reduced by 68.4% and 44.4%, compared with the traditional high-fidelity and variable-fidelity surrogate optimization methods. For the engineering examples, the collapse load optimized by the proposed method is improved by 9.0% and 14.7%, compared with the high-fidelity and variable-fidelity surrogate optimization methods. At the same time, compared with the variable-fidelity static reduction space method proposed in previous work, the calculation time is reduced by 41.5% when reaching the same collapse load. In summary, the proposed optimization method can make full use of multiple fidelity data, and has the advantages of high optimization efficiency and strong global optimization ability for complex and time-consuming aerospace structural engineering optimization problems.

Key words: structural optimization, space reduction, multiple fidelity model, hierarchical stiffened shells, fuzzy clustering

中图分类号: