Efficient Design of Coanda UAV Based on Surrogate Models

Abstract

The increased use of unmanned aerial vehicles (UAVs) in recent years has resulted in an uptick in related accidents. Accidents involving humans have been attributed to propeller laceration, highlighting the inherent risks associated with propellers. This study addressed the optimal design of nozzle shapes for Coanda UAVs without exposed propellers. Evaluating each nozzle candidate through actual implementation is a time consuming process; therefore, this study employed computational fluid dynamics (CFD) as an alternative. However, this method still requires a relatively long computational time. To solve this problem, a surrogate model constructed by support vector regression to evaluate nozzle shapes is proposed, which also evaluates the nozzle shapes not fully covered by the simulation. This approach reduced the number of CFD simulations required, thereby reducing the time required for optimization. Optimization using only CFD for nozzle shape evaluation was compared with the proposed method using a surrogate model. Under the condition of approximately 200 design candidates, the optimization accuracy was almost identical, whereas the optimization time was reduced by approximately one-third. Furthermore, measurements from the actual implementation of the nozzle optimized using the proposed method indicated an approximately 25% increase in the thrust-to-weight ratio of the nozzle.

Keywords

• Design optimization
• self-organizing maps
• surrogate model
• unmanned aerial vehicles (UAVs)