Optimizing avian flight dynamics with a synergetic bio-inspired and machine learning approach

This study presents a composite numerical and machine learning framework to enhance the aerodynamic performance of a bio-inspired flapping wing. The wing kinematics were extracted from biological flight data using a multi-step process: the DeepLabCut tool was applied to extract body point coordinates from avian flight videos, followed by data digitization in Google Colab and trajectory post-processing in Python. These kinematics were then prescribed to a three-dimensional wing model for high-fidelity unsteady Reynolds-Averaged Navier-Stokes (URANS) simulations of incompressible turbulent flow in ANSYS Fluent, utilizing a user-defined function (UDF) and a sliding mesh technique. Validated against existing experimental data, the numerical model serves as a reliable data generation tool. Subsequently, a machine learning model was developed to explore the design space and identify kinematic parameters that optimize aerodynamic efficiency. The results demonstrate that the proposed framework effectively bridges biological observation with computational optimization, offering a robust approach for the performance enhancement of bio-inspired flapping wings.