基于替代模型的高超声速前体/进气道一体化优化

采用基于替代模型的渐进优化策略对二维高超声速前体/进气道进行一体化设计优化,采用拉丁超立方试验设计法选择样本点,采用二维粘性CFD方法计算进气道流场来建立样本数据库,综合运用了多项式响应面、Kriging模型、BP神经网络和径向基神经网络等替代模型.相对于基准构型,前体/进气道的优化构型在设计态时提高了流量捕获与来流压缩能力,提高了总压恢复性能,同时减小了阻力系数,综合性能提高了5.3%;在非设计态时优化构型的综合性能也有不同程度的改善. 

An integrated optimization of hypersonic forebody/inlet based on surrogate models

A successive optimization strategy based on surrogate models was applied to integrated optimization of a hypersonic 2D forebody/inlet.The sample data points were selected by Latin Hypercubes design of experiment technique,and the sample database was constructed by forebody/inlet flowfield numerical calculation using computational fluid dynamics(CFD) method based on 2D viscous Navier-Stokes equation.Four types of surrogate models including polynomial response surface,Kriging model,BP neural network and radial bases function neural network were adopted.As compared with baseline forebody/inlet at on-design state,the optimal forebody/inlet has better flow capture and compression and total pressure recover capability,and has lower drag force coefficient,with 5.3% improvement of overall performance.The performance of optimal forebody/inlet at off-design states has also improved to some extent.