基于高斯函数和信赖域更新策略的Kriging响应面法

摘要

In order to set up an accurate and efficient response surface model while satisfying engineering precision requirements, after analyzing basic features of existing response surface modeling methods, Kriging response surface method based on Gauss function and trust region update approach was put forward.Gauss function was adopted as the correlation function for different designs, and the trust region update approach was utilized to adjust the sampling region of experiment design in the optimization process.The proposed method was applied in four wellknown test functions and a thermal-structural optimization of a general turbine disc.The results demonstrated the effectiveness of this method.%为建立拟合效率高且能够满足工程精度要求的响应面近似模型,在分析已有响应面建模方法基本特性的基础上,提出一种基于高斯函数和信赖域更新策略的Kriging响应面法.该方法将高斯函数作为设计点之间的相关函数,并在优化过程中采用信赖域更新策略对试验设计的取样区域进行动态修正.通过在四个典型数值算例和某涡轮盘的热-结构优化设计中的应用,验证了所提方法的有效性.