Research Information System
Aerospace Science and Technology, vol.167, 2025 (SCI-Expanded)
This paper focuses on mapping two-dimensional airfoil pressure characteristics over three-dimensional wing geometry. The primary objective is to generate quick and reliable fighter wing pressure distributions under transonic maneuvering conditions using deep learning algorithms. The proposed model, WingNet, employs an encoder-decoder architecture that outputs the ΔCp distribution between the wing and airfoil. High-fidelity wing datasets are generated using a 3-D Reynolds Averaged Navier Stokes (RANS) flow solver, while low-fidelity datasets are created with 2-D airfoil RANS simulations, focusing on transonic and moderate angle of attack (AOA) regimes. The results demonstrate accurate computation of three-dimensional wing Cp distributions using airfoil Cp datasets and wing planform inputs. The model achieves mean absolute errors of 0.0010, 0.0069, and 0.0026 in drag, lift, and pitch moment coefficients, respectively. The method effectively captures span-wise shock wave distributions, leading- and trailing-edge separations, crucial for fighter wing optimizations and aeroelastic analyses under transonic sustained turn conditions.