变循环发动机外涵道气流掺混特性建模研究  

Modeling Research on Bypass Flow Mixing Characteristics of Variable Cycle Engine

作　　者：吴宋伟 田杰 张天宏[1] 李凌蔚 李佳翱 WU Song-wei;TIAN Jie;ZHANG Tian-hong;LI Ling-wei;LI Jia-ao(College of Energy and Power,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)

机构地区：[1]南京航空航天大学能源与动力学院,江苏南京210016

出　　处：《推进技术》2022年第12期143-151,共9页Journal of Propulsion Technology

基　　金：国家自然科学基金(51976089)。

摘　　要：模式选择活门和前涵道引射器作为变循环发动机的关键几何部件,用于改善发动机在稳态以及模式切换过程的部件匹配性能,针对两者开度对外涵道气流掺混总压损失的影响开展研究。采用二维数值模拟方法,对内部压缩系统流场进行了仿真分析,得到了气流掺混的总压恢复系数在模式选择活门和前涵道引射器不同开度下的变化规律。以数值仿真结果作为建模研究的数据集,提出一种基于贝叶斯优化的混合核极限学习机算法,建立了总压恢复系数映射模型,采用留一交叉验证方法对模型进行了评估。结果表明,本文所提出的核极限学习机算法的均方根误差为0.0094,泛化效果相比多元线性回归模型提升约70.6%以上,所建拟合模型预测的相对误差最大不超过1.57%。The mode selector valve and the forward bypass ejector are the key geometric components of the variable cycle engine,which are used to improve the engine’s component matching performance in the steadystate and the mode transitioning process. Research is carried out on the influence of the two openings on the total pressure loss of the bypass flow mixing. A two-dimensional numerical simulation method was used,and the flow field of the internal compression system was simulated and analyzed. The change rule of the total pressure recovery ratio of flow mixing under different openings of the mode selector valve and the forward bypass ejector was obtained. Numerical simulation results were used as the data set for modeling research. A hybrid kernel extreme learning machine algorithm based on Bayesian optimization was proposed,and the total pressure recovery ratio mapping model was established. The leave-one-out cross-validation method was used to evaluate the model. The root mean square error of the kernel extreme learning machine algorithm proposed in this paper is 0.0094,and the generalization effect is improved by more than 70.6% compared with the multiple linear regression model. The relative error predicted by the fitting model is no more than 1.57%.

关 键 词：变循环发动机 前涵道引射器 模式选择活门 气流掺混 总压恢复系数 贝叶斯优化 混合核极限学习机 

分 类 号：V233.7[航空宇航科学与技术—航空宇航推进理论与工程]