气动弹性
跨音速
空气动力学
翼型
计算流体力学
非线性系统
空气动力
计算机科学
控制理论(社会学)
机械
结构工程
物理
工程类
量子力学
人工智能
控制(管理)
作者
Zihao Dou,Chuanqiang Gao,Weiwei Zhang,Yang Tao
出处
期刊:AIAA Journal
[American Institute of Aeronautics and Astronautics]
日期:2023-05-01
卷期号:61 (6): 2412-2429
摘要
Transonic buffet is an aerodynamic phenomenon of self-sustained shock oscillations. The aeroelastic problem caused by it is very complex, including two different dynamic modes: forced vibration and frequency lock-in. The vibration of the structure has a negative influence on the fatigue life of the aircraft. Especially in the region of frequency lock-in, the limit cycle oscillations occur due to the instability of the structural mode. Researchers have accurately predicted the region of frequency lock-in in transonic buffet and have clarified its mechanism by using a linear aerodynamic model. However, the nonlinear aeroelastic modeling and prediction of the transonic buffet remain to be solved. The long short-term memory (LSTM) deep neural network is suitable for predicting the time-delayed effects of unsteady aerodynamics. And it has achieved remarkable results in sequential data modeling. In the present work, a nonlinear model is developed for the aeroelastic system with NACA0012 airfoil in transonic buffeting flow and validated with the coupled computational fluid dynamics/computational structural dynamics (CFD/CSD) simulation. First, the data set and the loss function are specially designed. Then, the reduced-order model (ROM) based on the LSTM of the flow is built by using unsteady Reynolds-averaged Navier–Stokes computations data in a post-buffet state. By coupling the ROM and the single degree-of-freedom equation for the pitching angle, the nonlinear aeroelastic model is finally produced. The results show that the phenomenon of frequency lock-in and the self-sustained buffeting aerodynamics are precisely reconstructed. And the model has a strong generalization ability and can reproduce complex vibrations caused by competition between different modes. In short, the model can replace the CFD/CSD method in the current case with high efficiency and accuracy. The method can be used for modeling and prediction of other various complex aeroelastic systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI