基于本征正交分解和代理模型的高超声速气动热模型降阶研究

气动热弹性分析是高超声速飞行器设计的关键技术之一。高超声速飞行器气动热的准确快速预测是气动热弹性分析的重要前提。针对当前气动加热工程计算、数值计算和实验研究均不能很好满足设计要求的问题,采用本征正交分解(POD)与代理模型(Surrogate)技术结合的模型降阶(POD-Surrogate)方法,建立了一种快速高效的高超声速气动热降阶模型框架。针对典型高超声速三维翼面气动热预测研究结果表明:当保留的POD基模态个数大于20时,PODKriging方法和POD-RBF(Radial Basis Function)方法的降阶模型得到的翼面温度分布与计算流体力学(CFD)计算温度L∞平均误差分别达到6%和14%,相对均方根误差(NRMSE)平均误差分别达到4%和12%,继续增加POD的基模态并不能提高降阶模型的预测精度;针对高超声速机翼气动热计算,POD-Kriging方法比POD-RBF方法具有更高的精度;针对典型的高超声速三维翼面气动热预测表明:基于POD-Surrogate方法的气动热降阶模型具有较高的精度和效率。

Reduced order aerothermodynamic modeling research for hypersonic vehicles based on proper orthogonal decomposition and surrogate method

Aerothermoelasticity is one of the key technologies for hypersonic vehicles. Accurate and efficient estimation of aerodynamic heating is the basis of the aerothermoelasticity. Aimed at solving the shortcomings of engineering calculation and computational fluid dynamics, a novel estimation method based on proper orthogonal decomposition (POD) and surrogate(POD-Surrogate) method is proposed. Furthermore, a reduced order modeling framework for aerothermodynamic is developed. Test results for the three-dimensional aerothermodynamic over a hypersonic control surface indicate that the average L_∞ errors for POD-Kriging method and POD-RBF (Radial Basis Function) method can reach 6% and 14%, and the average normalized root mean square error (NRMSE) errors can reach 4% and 12%. Using more POD basis modes would not obviously improve the estimation precision when the basis modes reach 20. Reduced order models for the three-dimensional aerothermodynamics over a hypersonic control surface indicate that the precision of POD-Kriging method is better than that of POD-RBF method. In a word, the reduced order modeling for three-dimensional aerothermodynamics has good precision and efficiency.