Modeling of aerodynamic properties of wing profiles of fixed wing unmanned aerial vehicles with artificial neural networks

In this study, the aerodynamic performance parameters of LA2573A, NACA 4415, NASA LRN 1015 and SD7037 airfoil geometries used in fixed-wing UAVs sampled with reference to their maximum take-off weights were determined experimentally and an Artificial Neural Network (ANN) model was created with the results obtained. Airfoil geometries were designed using SolidWorks software. After the design was made, airfoils were produced using a Fortus 250 mc brand 3D printer. Experimental analyzes of the produced models were carried out at various attack angles and speeds in the subsonic wind tunnel which has a control volume of 305 x 305 x 600 mm. As a result of the experimental studies, CL, CD, CM and CL/CD aerodynamic performance parameters of the airfoil geometries were obtained. When the data were examined, it was determined that the NASA LRN 1015 model gave the most suitable result for negative attack angles and all speed values compared to other wing profiles. It is seen that the most suitable model for positive angles of attack and low speed values is NACA 4415. It is clearly seen that all airfoils exhibit identical performance characteristics for positive angles of attack and high speed values. When the ANN model obtained for each model is examined, a airfoil geometry that can characterize itself in real time for the most optimum results by controlling the angle of attack at different wind speeds can be revealed for use in future studies.