Helicopter fuselage aerodynamic data fitting using multivariate smoothing thin plate splines

Abstract

Smoothing thin plate splines, a nonparametric statistical technique for multivariate data fitting, were investigated to predict the aerodynamic performance (output variables) of a generic 3D helicopter fuselage as functions of the pitch angle and of some geometric parameters describing their shape (input variables). In order for the smoothing thin plate splines to be properly applied, a database needed to be constructed containing pairs of input–output variables. To this purpose, a sample helicopter fuselage was chosen and 14 variants were generated modifying the geometric parameters; then, the pertinent lift, drag and pitching moment coefficients were obtained via computational fluid dynamics. The smoothing thin plate splines model was built excluding from the database one fuselage at a time and was then used to determine the aerodynamic performance of the left out configuration: finally, the obtained results were compared with those coming from direct computational fluid dynamics simulations over the same fuselage. The prediction capability of the smoothing thin plate splines models has been confirmed for all the analyzed fuselage geometries.