Application of an hybrid optimization approach in the design of long endurance airfoils

Future Re-Configurable Multi Mission Unmanned Aerial Vehicle (RC-MM-UAV) Design concepts are expected to comprise of morphing wings with mission segment based airfoils. The Direct Numerical Optimization (DNO) methodology for airfoil shape optimization is established. The PARSEC shape function is used for airfoil geometry parameterization, coupled with a low-fidelity solver and Particle Swarm Optimizer (PSO) in the design analysis of a long endurance airfoil. A single-point airfoil design study through the DNO approach was executed. Results indicate that the methodology is computationally demanding thus, Artificial Neural Networks (ANN) are introduced to address this issue. A relationship between PARSEC airfoil geometry variables as inputs and the equating aerodynamic coefficients as outputs are used in network training and validation. The effect of varying training sample size was evaluated to establish a network with acceptable generalization capabilities. The PSO model is integrated to the ANN model for airfoil design optimization. The hybrid PSO/ANN structure required 38% fewer solver calls in comparison to the direct search approach. The proposed hybrid methodology is applicable for future multi-point airfoil design optimizations.