航空学报 > 2024, Vol. 45 Issue (8): 228982-228982   doi: 10.7527/S1000-6893.2023.28982

主动学习基自适应PC⁃Kriging模型的复合材料结构可靠度算法

龚煜廉1, 张建国1(), 吴志刚2, 褚光远3, 范晓铎1, 黄赢1   

  1. 1.北京航空航天大学 可靠性与系统工程学院, 北京 100191
    2.北京航空航天大学 航空科学与工程学院, 北京 100191
    3.中国运载火箭技术研究院, 北京 100076
  • 收稿日期:2023-05-09 修回日期:2023-06-12 接受日期:2023-06-14 出版日期:2024-04-25 发布日期:2023-06-16
  • 通讯作者: 张建国 E-mail:zjg@buaa.edu.cn
  • 基金资助:
    国家重点研发计划(2021YFB1715000)

Reliability algorithm of composite structure based on active learning basis-adaptive PC-Kriging model

Yulian GONG1, Jianguo ZHANG1(), Zhigang WU2, Guangyuan CHU3, Xiaoduo FAN1, Ying HUANG1   

  1. 1.School of Reliability and Systems Engineering,Beihang University,Beijing 100191,China
    2.School of Aeronautic Science and Engineering,Beihang University,Beijing 100191,China
    3.China Academy of Launch Vehicle Technology,Beijing 100076,China
  • Received:2023-05-09 Revised:2023-06-12 Accepted:2023-06-14 Online:2024-04-25 Published:2023-06-16
  • Contact: Jianguo ZHANG E-mail:zjg@buaa.edu.cn
  • Supported by:
    National Key R & D Program of China(2021YFB1715000)

摘要:

针对复合材料机翼随机固有频率可靠性分析复杂、高维、高度非线性和计算时间长的问题,本文提出了一种主动学习基自适应PC-Kriging模型的可靠度算法。在基自适应PC-Kriging模型中,采用一种基自适应策略来确定多项式混沌展开部分的正交多项式基,以近似数值模型的全局响应,Kriging用于高阶非线性插值以近似数值模型的局部响应。在主动学习可靠度计算框架中,引入加权K均值聚类,在一次迭代中添加K个对失效概率贡献较大的候选样本点以减少迭代次数和加快收敛速度。通过一个高度非线性的数值算例分析,证明了所提方法的有效性和准确性。将该方法应用于复合材料板和复合材料机翼的随机固有频率可靠性分析,获得了准确高效的可靠度计算结果。

关键词: 可靠性分析, 基自适应PC-Kriging, 主动学习, 加权K均值聚类, 复合材料机翼, 随机固有频率

Abstract:

To address the complex, high dimensional, highly nonlinear, and long computing time-consuming problems of random natural frequency reliability analysis of composite wings, a reliability algorithm based on active learning basis-adaptive PC-Kriging model is proposed in this paper. A basis-adaptive strategy is used in this model to determine the orthogonal polynomial basis of the polynomial chaos expansion to approximate the global response of the numerical model, and Kriging is used for higher-order nonlinear interpolation to approximate the local response of the numerical model. In the framework of active learning reliability calculation, weighted K mean clustering is introduced, which means that K candidate sample points with greater contribution to failure probability are added in one iteration to reduce the number of iterations and accelerate the convergence rate. The effectiveness and accuracy of the proposed method are proved by a highly nonlinear numerical example. The proposed method is applied to the random natural frequency reliability analysis of composite plate and composite wing, and the accurate and efficient reliability calculation results are obtained.

Key words: reliability analysis, Basis-adaptive PC-Kriging, active learning, weighted K mean clustering, composite wings, random natural frequency

中图分类号: