Deep Learning Based Fast Prediction and Optimization of Aerodynamic Performance for a Propeller with Gurney Flap

High-altitude and long-endurance unmanned air vehicles have placed high demands for the performance of propellers under low Reynolds numbers. Conventional propeller design methods are less efficient, making it difficult to achieve a breakthrough in propulsion efficiency. This paper explores the possibility of extending the Gurney flap on low Reynolds number propellers to achieve efficiency breakthrough. A framework based on deep learning is established to fast predict and optimize the aerodynamic performance of a propeller with Gurney flap, which can quickly obtain the optimal propeller profile airfoil shape, distribution of twist angle, and chord length at different advance ratios. The results show that the surrogate model based on Multi-Task Learning has high prediction precision and efficiency, which can obtain performance parameters within 0.06 s. At different advance ratios, the optimized propellers with Gurney flap have a significant improvement in propulsive efficiency. In particular, the optimal propeller with Gurney flap consumes 35.65% less power in cruise at the advance ratio of 0.9, resulting in a propulsion efficiency of 80.74%, an improvement of 11.14% compared to the pre-optimization period. In conclusion, Gurney flaps are important components to enhance propulsion performance for propellers, and propellers with Gurney flaps have better aerodynamic efficiency than propellers without those. This work provides a reference for accurate and efficient design of low Reynolds numbers propellers.