结合Kriging模型权重信息熵函数的结构可靠性研究  

Structural Reliability Study Combining the Kriging Model and the Weight Information Entropy Function

作　　者：李景奎 刘文琪 周岩 李战东 LI Jing-kui;LIU Wen-qi;ZHOU Yan;LI Zhan-dong(Civil Aviation College,Shenyang Aerospace University,Liaoning Shenyang 110136,China)

机构地区：[1]沈阳航空航天大学民用航空学院,辽宁沈阳110136

出　　处：《机械设计与制造》2024年第10期168-171,177,共5页Machinery Design & Manufacture

基　　金：辽宁省教育厅重点攻关和服务地方项目(JYT19003)。

摘　　要：为提高基于Kriging模型信息熵函数(Information Entropy Function,H)的可靠性计算效率,考虑样本点与极限状态曲面的空间距离和随机变量的概率密度函数,通过对样本点的信息熵赋予不同的权值,提出权重信息熵函数(Weight Information Entropy Function,WH)。该学习函数选择更接近极限状态曲面且概率密度函数值较大的样本点更新Kriging模型,从而减少对功能函数的调用次数,有效提高可靠性计算效率。通过算例表明:与其他学习函数相比,WH学习函数在建立Kriging模型过程中所需要的样本点更少,收敛速度更快,计算效率更高。To improve the reliability calculation efficiency of the Information Entropy Function(H)based on the Kriging model,the Weight Information Entropy Function(WH)is proposed.The WH learning function considers both locations of the sample points in the variable space and the probability density function of random variables.Different weights are given to the information entropy of the sample points.Sample points which approach not only to the limit state surface but also have large probability density function value,are selected to update the Kriging model.Consequently,it can reduce the number of calls to the performance function and improve the efficiency of the reliability computation.Specifically,the proposed method is verified by several examples.The results show that this method requires fewer sample points in the process of establishing the Kriging model.Finally,compared with other active learning functions,it has fast convergence speed and high calculation efficiency.

关 键 词：结构可靠性分析 KRIGING模型 学习函数 权重信息熵函数 概率密度函数 

分 类 号：TH16[机械工程—机械制造及自动化] TB114.3[理学—概率论与数理统计]