欢迎光临《计算力学学报》官方网站!
基于主动学习Kriging模型的结构多失效模式可靠度计算
Structural reliability calculation for multiple failure modes based on an active learning Kriging model
投稿时间:2019-01-24  修订日期:2019-03-29
DOI:
中文关键词:  可靠度  多失效模式  代理模型  多输出Kriging  学习函数
英文关键词:reliability  multiple failure modes  surrogate model  multi-output Kriging  learning functions
基金项目:
作者单位E-mail
赵维涛 沈阳航空航天大学 zhwt201@163.com 
摘要点击次数: 78
全文下载次数: 0
中文摘要:
      对于具有多失效模式的结构可靠度计算问题,文中利用多输出Kriging模型作为代理模型。该代理模型只需对所有功能函数进行一次建模,无需对每个功能函数建立各自的代理模型,且在建模过程中能够考虑各失效模式之间的相关性。本文方法设定的初始样本点不仅对随机变量均值附近区域给予足够重视,而且能够兼顾设计空间的边缘区域,进而确保初始代理模型在全局空间内具有较好精度,以减少后续利用学习函数更新代理模型的次数。数值算例表明,本文方法具有较好的计算精度和较高的计算效率,当失效模式较多时,本文方法的计算效率大幅提升。
英文摘要:
      A multi-output Kriging model is used as the surrogate model to solve a problem of structural reliability calculation with multiple failure modes. In this study, the surrogate model is constructed only once for all performance functions, without having to construct a separate surrogate model for each function, and the correlation between failure modes can be considered in the modeling process. The initial sample points given by the proposed method not only consider the region near the mean of the random variables, but also can consider the edge region of the design space, thus the initial surrogate model has better accuracy in the global space, so that the number of updating surrogate model by using learning functions is reduced. Numerical examples show that the proposed method can obtain the satisfying accuracy, and the proposed method can greatly improve efficiency especially for a large number of failure modes.
  查看/发表评论  下载PDF阅读器
您是第4803211位访问者
版权所有:《计算力学学报》编辑部
本系统由 北京勤云科技发展有限公司设计