Search

• A novel adaptive control method based on DSN and RL is proposed to solve a problem of controlling the morphing trailing edge for drag reduction. • Due to the construction of network based on the physical prior knowledge, DSN has better accuracy in predicting lift and pitching moment than the traditional surrogate models. • After training, the controller can solve multi-dimensional and nonlinear complex problems. Under the condition of maintaining pitching moment balance, the lift-to-drag ratio at the start point can be increased by 1.36%, and can be increased by 2.61% at the end point. The morphing trailing-edge wing can adaptively change its shape according to the different flight conditions, improving cruising efficiency. The control of the morphing trailing edge is a high-dimensional variable and nonlinear problem, facing the difficulty of the high training cost and the difficulty in real-time continuous control. To overcome these difficulties, a novel adaptive control method based on the couple of Deep Supervision Net (DSN) and Deep Reinforcement Learning (DRL) is proposed to solve a control problem of the morphing trailing-edge wing. DSN uses distributed loads to improve aerodynamic prediction accuracy compared to the traditional surrogate models, while DRL conducts real-time control and improves the performance at non-design points compared to the traditional multi-design points optimization. After control, the lift-to-drag ratio at the start point can be increased by 1.36%, and can be increased by 2.61% at the end point tested on a commercial wide-body aircraft model. [ABSTRACT FROM AUTHOR]

The computational fluid dynamics is the main method to evaluate the aerodynamic performance for the optimization of high-lift devices. Currently, the direct three-dimensional (3D) optimization requires significant computational resources. Additionally, the commonly used heuristic algorithms can not extract the experience of two-dimensional (2D) optimization to accelerate the 3D optimization process. In order to resolve these issues, a novel 2D-to-3D optimization method based on the coupling of Deep Reinforcement Learning (DRL) and Transfer Learning (TL) is proposed to conduct the aerodynamic optimization of the 3D high-lift devices and tested on the NASA Trap Wing model. The 2D optimization is first carried out, and then its neural networks in DRL and the optimal configuration are extracted by TL to turn into the 3D optimization. Compared with the direct 3D optimization, the proposed 2D-to-3D optimization method can result in improved aerodynamic performance for the same computational cost, or can save 51%-81% of the computational cost to obtain a similar performance. [ABSTRACT FROM AUTHOR]

In a propeller blade optimization, both aerodynamic and aeroacoustic performance were considered simultaneously. A multi-fidelity sampling scheme was adopted by Transfer Learning (TL) to improve the overall optimization efficiency. A Deep Neural Network (DNN) was selected to map the non-linear relationship between the blade parameters and the aerodynamic/aeroacoustic performance, with the optimization being achieved by implementing a deep reinforcement learning algorithm, namely, Deep Deterministic Policy Gradient (DDPG), upon which a Multi-fidelity DNN based surrogate model (TL-MFDNN) was introduced with Transfer Learning between pre-trained and retrained processes. It was found that, by comparing the TL-MFDNN surrogate model based optimization with DDPG optimization using direct CFD simulation, the overall computing cost can be saved up to 77.3% and the optimized propeller has maximum noise reduction of up to 1.69 dB, with a negligible penalty on propulsive performance. [ABSTRACT FROM AUTHOR]