Surrogate-Based Optimization of a High-Altitude Propeller

A surrogate-based, multipoint and multiobjective optimization was performed in this work for a High-Altitude Propeller. For each blade design, the performance was evaluated with 3D RANS at three advance ratios for cruise and one advance ratio for climb, while the ability to take off, was assessed with Vortex Theory. Moreover, a different sampling strategy for the DoE phase was tested as an alternative to the Latin Hypercube Sampling of the whole design space. In this optimization problem, two objective functions were defined, one for the aerodynamic performance and one for the propeller weight, while Kriging was used to find the response surface of each objective. The Expected Improvement Matrix with the Euclidean distance was used as an infill criterion and the Kriging Believer algorithm was adopted in order to parallelize the procedure. The Pareto front formation is discussed as, well as the characteristics of some specific designs. In the end, a variance based sensitivity analysis is performed on the evaluated designs to give insight into the importance of the different design variables.