基于高斯过程回归的机翼/短舱一体化气动优化  

Integrated aerodynamic optimization of wing/nacelle based on Gaussian process regression

作　　者：季廷炜[1] 莫邵昌 谢芳芳 张鑫帅 蒋逸阳 郑耀[1] JI Ting-wei;MO Shao-chang;XIE Fang-fang;ZHANG Xin-shuai;JIANG Yi-yang;ZHENG Yao(School of Aeronautics and Astronautics,Zhengjiang University,Hangzhou 310027,China)

机构地区：[1]浙江大学航空航天学院,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2023年第3期632-642,共11页Journal of Zhejiang University：Engineering Science

摘　　要：为了解决机翼/短舱一体化气动设计的高维非线性优化问题,基于高斯过程回归(GPR)模型提出新型优化设计方法.采用类别形状函数变换(CST)方法对机翼/短舱一体化构型中的翼型进行几何参数化建模;通过控制机翼形状参数、短舱形状参数和短舱安装参数实现机翼/短舱构型变形,该参数化建模过程共计包含50个设计参数.通过GPR模型构建机翼/短舱设计参数与气动性能之间的代理模型,并采用贝叶斯优化(BO)算法实现代理模型的自更新和最优气动外形的获取.结果表明:优化后一体化构型的阻力系数下降了10.95%,通过流场分析发现机翼外形和短舱外形的优化改善了表面流场结构,短舱安装位置的优化减弱了机翼和短舱间的气动干扰.A new optimization design method based on Gaussian process regression(GPR)model was proposed to resolve the high-dimensional nonlinear optimization problem of integrated aerodynamic design of wing/nacelle.The geometric parametric modeling of the airfoils in the integrated configuration of wing/nacelle was realized by the class and shape transformation(CST)method.The deformation of the integrated configuration of wing/nacelle was achieved by controlling the wing shape parameters,the nacelle shape parameters and the nacelle installation parameters.The parametric modeling process included 50 design parameters in total.The GPR model was used to construct a surrogate model between the design parameters and the aerodynamics performance of the integrated wing/nacelle geometry.Bayesian optimization(BO)algorithm was used to realize the self-update of the surrogate model and the acquisition of the optimal aerodynamic shape.Results showed that the drag coefficient of the integrated configuration was reduced by 10.95%after the optimization.The flow field analysis shows that the optimization of the wing shape and the nacelle shape improves the surface flow field structure,and the optimization of the nacelle’s installation position reduces the aerodynamic interference between wing and nacelle.

关 键 词：机翼/短舱 气动优化设计 参数化建模 高斯过程回归(GPR) 贝叶斯优化(BO) 

分 类 号：V221.3[航空宇航科学与技术—飞行器设计]