Airfoil buffet aerodynamics at plunge and pitch excitation based on long short-term memory neural network prediction

Abstract In the present study, a nonlinear system identification approach based on long short-term memory (LSTM) neural network is applied for prediction of transonic buffet aerodynamics. The as reduced-order modeling (ROM) technique an efficient computation time-varying integral quantities such aerodynamic force and moment coefficients. Therefore, procedure well generalization ROM are presented. training data set LSTM–ROM provided by performing forced-motion unsteady Reynolds-averaged Navier–Stokes simulations. Subsequent to process, associated with buffet. performance trained demonstrated computing loads NACA0012 airfoil investigated at freestream conditions. contrast previous studies considering only pitching excitation, both pitch plunge degrees freedom individually simultaneously excited means user-defined signal. strong effects considered ROM. By comparing results full-order computational fluid dynamics solution, good capability presented method indicated. However, compared including slightly reduced shown.