Research on Navigation Attitude of Air Cushion Vehicle Based on BP Neural Network

Abstract

Listen

Translate article

Air cushion vehicle (ACV) is a kind of highperformance ship type, which uses skirt lifting system to form air cushion between the bottom and the water surface, the air cushion pressure supports the ship to sail on the water surface, and is propelled by air propeller. ACV should keep a proper attitude when sailing, especially in the stage of high-speed operation. This not only affects the overall performance of its rapidity, manoeuvrability and stability, but also is important for its navigation safety. Therefore, it is necessary to study the navigation attitude of ACV. Compared with the displacement ships, the ACV is mainly supported by air cushion pressure, and its navigation attitude is more easily disturbed by various factors. The operation environment of water-air coupling is special, which forms a complex system with many factors and nonlinearity. In view of the limited similarity condition of model test and the low accuracy of theoretical calculation or numerical simulation, it is very difficult to accurately predict the navigation attitude of an ACV in various situations. In this paper, based on the navigation data of an ACV, the factors related to the navigation attitude are selected as the input vectors, and the navigation trimming angle and heeling angle are taken as the output vectors. The number of hidden layer nodes is reasonably selected, and the BP neural network model which can be used to predict the navigation attitude of the ACV is trained and established. The prediction accuracy of the neural network is further verified by the test data of start acceleration, high-speed straight navigation, braking, rotation, sideslip and other special states. The neural network model is used to analyse the change of ACV's navigation attitude caused by the change of single factor, and the influence rules of the factors such as speed, drift angle, wind direction, rudder angle and pitch difference on the navigation attitude of ACV are summarized. Compared with the displacement ships, the ACV is mainly supported by air cushion pressure, and its navigation attitude is more easily disturbed by various factors. The operation environment of water-air coupling is special, which forms a complex system with many factors and nonlinearity. In view of the limited similarity condition of model test and the low accuracy of theoretical calculation or numerical simulation, it is very difficult to accurately predict the navigation attitude of an ACV in various situations. In this paper, based on the navigation data of an ACV, the factors related to the navigation attitude are selected as the input vectors, and the navigation trimming angle and heeling angle are taken as the output vectors. The number of hidden layer nodes is reasonably selected, and the BP neural network model which can be used to predict the navigation attitude of the ACV is trained and established. The prediction accuracy of the neural network is further verified by the test data of start acceleration, high-speed straight navigation, braking, rotation, sideslip and other special states. The neural network model is used to analyse the change of ACV's navigation attitude caused by the change of single factor, and the influence rules of the factors such as speed, drift angle, wind direction, rudder angle and pitch difference on the navigation attitude of ACV are summarized. In this paper, based on the navigation data of an ACV, the factors related to the navigation attitude are selected as the input vectors, and the navigation trimming angle and heeling angle are taken as the output vectors. The number of hidden layer nodes is reasonably selected, and the BP neural network model which can be used to predict the navigation attitude of the ACV is trained and established. The prediction accuracy of the neural network is further verified by the test data of start acceleration, high-speed straight navigation, braking, rotation, sideslip and other special states. The neural network model is used to analyse the change of ACV's navigation attitude caused by the change of single factor, and the influence rules of the factors such as speed, drift angle, wind direction, rudder angle and pitch difference on the navigation attitude of ACV are summarized.