Aerodynamic predictions for unmanned aerial vehicle formation flight using vortex lattice method and machine learning

Date

2024-9

Author

Tatar, Umutcan

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

339
views

897
downloads

Cite This

Close formation flights of Unmanned Aerial Vehicles (UAVs) offer significant potential for enhancing efficiency and capabilities of UAV missions. However, the complexity and long durations of Computational Fluid Dynamics (CFD) often bring limitations to the design process. To adress this, the Vortex Lattice Method (VLM) based on incompressible potential flow can be utilized as a more efficient alternative. In this thesis, a method that combines the VLM and machine learning is demonstrated for estimating the aerodynamic performance of UAVs in formation. The VLM code calculates the aerodynamic coefficients and optimum point based on the relative three-dimensional positions of UAVs for different angles of attack and sweep angles of their wings. Results are used to train an Artificial Neural Network (ANN) and a Convolutional Neural Network (CNN) model in order to grasp the intricate relationships between these parameters. The CNN process Signed Distance Function (SDF) images, which helps to effectively capture positional information and model wing parameters such as sweep angle by recognizing the geometric changes across the entire domain. When compared to the VLM predictions, these neural network models are observed to be capable of capturing general aerodynamic trends, which demonstrates that they can be used for quick and satisfactory assessments of aerodynamic performance of close formation flight.

Subject Keywords

Unmanned aerial vehicle, Vortex lattice method, Artificial neural network, Convolutional neural network, Formation flight

URI

https://hdl.handle.net/11511/111069

Collections

Graduate School of Natural and Applied Sciences, Thesis