东北大学学报(自然科学版) ›› 2023, Vol. 44 ›› Issue (1): 76-82.DOI: 10.12068/j.issn.1005-3026.2023.01.011

• 机械工程 • 上一篇    下一篇

面向结构可靠性分析的并行自适应加点策略

查从燚1, 孙志礼1, 潘陈蓉2, 王健1   

  1. (1.东北大学 机械工程与自动化学院, 辽宁 沈阳110819; 2.安徽新华学院 通识教育部, 安徽 合肥230088)
  • 发布日期:2023-01-30
  • 通讯作者: 查从燚
  • 作者简介:查从燚(1993-),男,安徽六安人,东北大学博士研究生; 孙志礼(1957-),男,山东巨野人,东北大学教授,博士生导师.
  • 基金资助:
    国家自然科学基金资助项目(51775097,51875095).

Parallel Adaptive Sampling Strategy for Structural Reliability Analysis

ZHA Cong-yi1, SUN Zhi-li1, PAN Chen-rong2, WANG Jian1   

  1. 1. School of Mechanical Engineering & Automation, Northeastern University, Shenyang 110819, China; 2. Department of General Education, Anhui Xinhua University, Hefei 230088, China.
  • Published:2023-01-30
  • Contact: ZHA Cong-yi
  • About author:-
  • Supported by:
    -

摘要: 现有的自适应加点策略多局限于Kriging模型,或在每次迭代过程中只能选取一个最佳样本点,效率较低.为解决上述问题,本文提出了一种通用的并行自适应加点策略CF-K.该方法考虑了样本点的局部不确定性并确保所选样本点分布在极限状态函数附近;此外,结合k-means算法以实现并行计算,即利用多台计算机在每次迭代的同时进行多个样本的仿真.算例分析表明,与其他方法相比,所提方法在满足精度要求的条件下具有更少的迭代次数,更节省时间.基于所提方法的结构可靠性分析不仅在计算效率和精度之间取得了较好的平衡,在理论上还可用于任何现有的代理模型.

关键词: 代理模型;自适应加点策略;k-means算法;并行;结构可靠性

Abstract: Many existing adaptive sampling strategies are limited to the Kriging models, or are with low efficiency for only selecting one best sample point at each iteration. To solve the above issues, a general parallel adaptive sampling strategy CF-K is proposed. The proposed method considers the local uncertainty of the sample points and ensures that the selected sample points distribute around the limit-state function. Furthermore, the k-means algorithm is incorporated to achieve parallel computation, which means that several sample points can be simulated simultaneously at each iteration on several computers. The numerical cases show that the proposed method has less iterations and saves more time than other methods under the condition of satisfying the precision. The structural reliability analysis based on the proposed method not only achieves a good balance between computational efficiency and accuracy, but also be available for any existing surrogate models in principle.

Key words: surrogate model; adaptive sampling strategy; k-means algorithm; parallel; structural reliability

中图分类号: