Safety assessment of probe-and-drogue aerial refueling docking based on deep learning

Ensuring safety during aerial refueling is critical, especially during the docking phase. For increasing the safety of aerial refueling, this paper proposes a docking safety assessment method based on a deep neural network (DNN) to predict the docking success rate, in which particle swarm optimization (PSO) is adopted to solve the optimal network parameters. First, rich simulation data is generated based on the comprehensive aerial refueling docking simulation platform. Then, using the collected data, the relationship between the state of the aerial refueling system and the docking success rate is established based on the deep neural network after training, and the docking success rate can be predicted as a safety metric. Finally, the simulation comparison shows that the proposed PSO-DNN is better than the classical DNN prediction. In general, the deep learning network designed in this paper can obtain the internal information of the data well, and can evaluate the safety of aerial refueling according to the predicted docking success rate, which provides strong support for improving the safety of aerial refueling.