| Wing is the main lifting surface and a major component of a plane. Airfoil is a key part of wingsand empennage.The quality of airfoil design directly affects a plane’s aerodynamic performance andflying qualities. Traditional methods of the airfoil design, such as wind tunnel and CFD-basedapproaches, are time-consuming and costly, and highly rely on the designers’ experience.A fast aerofoil design method based on Gaussian process regression has being proposed recently.It is computationally fast and accurate. However this method uses NACA parameters as designvariables, and the number of design varivables is relatively small. This easily leads to poorgeneralization capability. To overcome the shortcomings of previous work, this paper combines CSTparameterization and Gaussian process regression to efficiently predict airfoil aerodynamicperformance and shape parameters of unknown airfoil. The thesis selets group of four-digit NACAairfoils, and calculates their CST parameters, aerodynamic performance and pressure distribution asthe experimental data. Using this data, Gaussian processes are trained for the fast airfoil aerodynamicperformance evaluation and airfoil reverse design. The obtained results are compared with thoseobtained from NACA parameterization.Experimental results indicate that the CST-based prediction approach is more accurate andcomputationally faster.The propused method can be applied to aid airfoil design in practice. |
