Machine Learning-Enhanced Optimization of Rotor Blades for Rotary-Wing Mars Uavs Through Coupled Cfd Simulation

Abstract

This study introduces a method for optimizing rotary-wing Mars UAV rotor blades by combining 2D airfoil CFD simulations with machine learning. The algorithms like ANN, Ada-Boost, SVM-L, and SVM-G were used to analyze key input and output parameters. The SVM-G algorithm showed high accuracy for Cl/Cd and both SVM-G and Ada-Boost for C1.5 l/Cd. The optimized blade design aimed to maximize C1.5 l/Cd spanwise and was validated through Martian atmospheric ground simulations. Results showed that at 100W, the UAV's rotor system could produce 5.02N thrust, with an aerodynamic efficiency of 0.7437 and power loading of 0.0487N/W. This approach demonstrates the potential of integrating machine learning with CFD for effective and accurate UAV rotor blade design in Martian environments.

Keywords: Mars UAV, machine learning, CFD simulation, rotor blade, aerodynamic performance.