基于代理模型的高超声速气动热模型降阶研究
A Reduced Order Modeling for Aerothermodynamic of Hypersonic Vehicles Based on Surrogate Method
-
摘要: 高超声速飞行器气动热的快速准确预测是当前高超声速气动热弹性分析的重要前提. 针对当前高超声速气动热工程计算、高精度数值计算和实验研究均不能很好适应工程应用的问题,结合代理模型的基本思想,提出了基于代理模型的高超声速气动热模型降阶方法,建立了一种高超声速气动热模型降阶框架. 以典型高超声速三维翼面为例,对比拉丁超立方采样方法lhsdesign函数和改进的逐次枚举的拉丁超立方方法SLE,利用相同的设计样本点和代理模型构造方法,SLE方法构造的降阶模型预测翼面温度平均误差、L∞和eNRSME均小于lhsdesign方法,SLE采样方法有助于提高降阶模型的精度;对比Kriging和RBF两种代理模型构造方法,Kriging方法构造降阶模型优于RBF方法. 针对典型的高超声速三维翼面气动热预测表明,本文高超声速气动热降阶方法具有较高的精度和效率.Abstract: Accurate and efficient estimation of aerodynamic heating is the basic prerequisite of theaerothermo-elasticity analysis. In order to solve the engineering adaptability problems of hypersonic aerothermodynamics calculation, numerical superprecision calculation and experimental investigation, a novel estimation method based on surrogate method for aerothermdynamic was proposed. Furthermore, a reduced order modeling (ROM) framework for aerothermodynamic was developed. Test results for the three-dimensional aerothermodynamics over a typical hypersonic control surface indicate that the average absolute errors, L∞ and NRSME by SLE are all smaller than those by lhsdesign using the same samlping points and the same sorrogate methods. So the SLE sampling method can help to improve the precision of ROM. Comparing the results of ROMs for the three-dimensional aerothermodynamics over a hypersonic control surface, it is indicated that the precision of Kriging is better than that of RBF. The developed ROMs for the three-dimensional surface show higher precision and efficiency for hypersonic aerothermodynamic.
下载: