Analysis of formation flight aerodynamics using vortex lattice method and convolutional neural networks
Aerospace Science and Technology, cilt.179, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 179
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.ast.2026.113590
- Dergi Adı: Aerospace Science and Technology
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC, zbMATH, Academic Search Ultimate (EBSCO), Engineering Source (EBSCO)
- Anahtar Kelimeler: Formation flight, Machine learning, Sweep angle, UAV, Vortex lattice method
- Orta Doğu Teknik Üniversitesi Adresli: Evet
Özet
Significant fuel savings can be obtained in formation flight by taking advantage of wake vortices if the relative position of aircraft can be effectively optimized. In this paper, formation flight of a leader and a follower unmanned aerial vehicle (UAV) is investigated. The location of the follower UAV is varied relative to the leader while flying at various angles of attack, with the wing sweep angle varied between positive and negative values corresponding to forward-swept and swept-back configurations. An in-house code is developed based on the Vortex Lattice Method (VLM) that incorporates such variations in formation flight and is used in conjunction with machine learning to calculate the resulting aerodynamic coefficients of forces and moments. For doing so, the relative spacing and angles are encoded into a structured input representation and the calculated coefficients by the VLM are used to train a Convolutional Neural Network (CNN) model. Results show that the maximum drag reduction increases as the angle of attack and sweep angle are increased. For negative sweep angles, however, drag tends to increase due to downwash at certain locations behind the leader UAV, resulting in no drag reduction possibility at those locations. Among all sweep angles studied, the forward-swept configuration is observed to cause pronounced differences in lift and rolling moment coefficients. Similarly, the swept-back configuration with the highest sweep angle gives the largest difference in pitching moment. Comparison of the results from the trained CNN model with those from the VLM code shows that the CNN is capable of providing a quick and satisfactory assessment of the aerodynamic performance of formation flights compared to the VLM and that it can successfully predict the location of the sweet spot for different flight configurations.