Nonlinear Aeroelastic Prediction in Transonic Buffeting Flow by Deep Neural Network

Zihao Dou, Chuanqiang Gao, Weiwei Zhang, Yang Tao

科研成果: 期刊稿件文章同行评审

16 引用 (Scopus)

摘要

Transonic buffet is an aerodynamic phenomenon of self-sustained shock oscillations. The aeroelastic problemcaused by it is very complex, includingtwodifferent dynamicmodes: forcedvibrationandfrequency lock-in.The vibrationof 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 structuralmode. 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) deepneuralnetworkis suitable for predictingthe time-delayedeffects ofunsteady aerodynamics.Andit has achieved remarkable results in sequential datamodeling. In the presentwork, a nonlinearmodel is developed for the aeroelastic systemwithNACA0012 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-averagedNavier-Stokes computations data in a post-buffet state. By coupling theROMand the single degreeof- freedomequation for the pitching angle, the nonlinear aeroelasticmodel is finally produced. The results showthat 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, themodel 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.

源语言英语
页(从-至)2412-2429
页数18
期刊AIAA Journal
61
6
DOI
出版状态已出版 - 5月 2023

指纹

探究 'Nonlinear Aeroelastic Prediction in Transonic Buffeting Flow by Deep Neural Network' 的科研主题。它们共同构成独一无二的指纹。

引用此