Study On Heave Motion Prediction Of Ships

The heave motion of ships excited by ocean waves seriously affect many marine applications, such as offshore operations, safe aircraft landing on vessels, etc. The influence will be worse during harsh sea conditions. To avoid these problems, controlling the heave motion of ships is necessary, for which the accurate prediction of heave motion of ships is significant.In order to realize the heave motion prediction of ships, the main work of this paper includes:(1) Heave motion prediction of ships based on Kalman filtering technique. Firstly, because the heave position of ships is difficult to be measured directly, an observer is constructed to estimate the heave position of ships according to the measured vertical acceleration by an inertial measurement unit (IMU). Then, heave motion equation of a ship is transformed into discretized state space model. At last, the heave position predictor is designed based on Kalman filtering technique on the basis of the heave motion equation of ships and the heave position estimate of ships. The heave position at the next moment is predicted based on the heave position estimate and the state estimate at current moment.(2) Heave motion prediction of ships based on radial basis function (RBF) neural networks. Firstly, the structure of RBF neural networks is determined. Then, the gradient descent method is used to train the RBF neural networks while the heave position estimates are used as the training samples. At last, the heave position at the next moment is predicted based on the trained RBF neural networks with heave position estimate of the current moment and three moments before it as input. Because of the fitting and generalization abilities of RBF neural networks, the heave motion is predicted based on RBF neural networks without parameters of ships.(3) Heave motion prediction of ships based on Elman neural networks. Procedures of heave motion prediction of ships based on Elman neural networks are the same as procedures of heave motion prediction of ships based on RBF neural networks. Because of the fitting and generalization abilities and dynamic memory characteristics of Elman neural networks, the heave motion is predicted based on Elman neural networks without parameters of ships.Finally, the three developed predicting methods are evaluated by simulation on a surface effect ship (SES), respectively. The simulation results indicate that the three predicting methods for the heave motion of ships are effective. The simulation results of the three methods are compared.