Abstract:

Solving the Reynolds Averaged Navier-Stokes (RANS) equations is essential for an accurate estimation of the aerodynamic loads on the helicopter rotor blades. This is highly time consuming process, especially, in the case of blade shape optimization in which several RANS solutions are obtained using Computational Fluid Dynamics (CFD) solvers. A recent approach to diminish the duration of computation is to employ metamodels like machine learning techniques. A well-built metamodel is expected to successfully mimic the CFD solutions. In this study, the Support Vector Regression (SVR) method based on a set of CFD solutions is used as the metamodel. CFD solutions are computed employing the commercial solver, FINE/Turbo by NUMECA International. The support vector regression model is built to give a functional relationship between the spanwise twist and the generated thrust and torque. The smooth variation of twist is defined using a 3-knot cubic spline. A total of 5 parameters are considered as input for the spline definition. The optimum twist distribution is determined according to a baseline blade: the Caradonna-Tung helicopter rotor blade. The following optimization cases are studied: maximum thrust at the experimental torque value and minimum torque at the experimental thrust value. The optimum cases are easily obtained since the SVR provides an analytical model. Results show that it is possible to enhance the blade performance by redistributing the spanwise twist without carrying out a full geometry optimization of the blade shape. Moreover, it has been observed that the twist distribution for both cases show similar behavior from the midspan to the tip while the distribution from the root to the midspan for both cases is totally in contrast.