Aerodynamic Design of Prop-Rotor Blades Using Neural Networks

We present a methodology for neural-network-based airfoil performance prediction for rapid blade airfoil section design and prop-rotor blade shape optimization. XFOIL was used to evaluate 1620 airfoils in the University of Illinois at Urbana-Champaign (UIUC) database to generate training data. Airfoil upper and lower surface geometries were parameterized using a modified class shape transformation. Three separate networks were trained: one each for lift, drag, and pitching moment. Isolated airfoil shape optimization was performed using the NSGA2 algorithm, targeting minimum average drag over an operating lift coefficient range for 10%-, 12%-, and 16%-thick airfoils. Prop-rotor aerodynamic optimization was carried out by designating airfoil shape parameters, blade twist distribution, and blade chord distribution as simultaneous design variables for two conflicting objectives: hover figure of merit and cruise propeller efficiency. Blade element momentum theory is used to predict rotor performance in hover and cruise using airfoil tables generated by the neural network. Pareto frontiers and an analysis of the resulting designs are presented. Using a neural network is advantageous for isolated airfoil optimization and rotor shape design because it decouples the computational cost of shape optimization from that of the flow solver.