Chaotic Dynamics Analysis Of Ship Sway And Its Prediction In Time Domain

Ship motion prediction in time domain, namely prediction of real time value and amplitude of ship motion in six degrees of freedom, is still an open question which is much concerned by the domain of shipping and ship engineering especially by the navies of the world. But the data available shows that the prediction length of real ship motion is sill less than 10 seconds which limits its application seriously. The reason is the lack of clear understanding of the ship motion mechanism in waves. This paper therefore aims to find out the nonlinearity dynamics held in the ship motion and to present the corresponding prediction methods to prolong the prediction length which can be used in the practice of navigation.(1) This paper analyzes the nonlinearity mechanism of ship motion which shows the large amplitude nonlinear rolling motion and its coupling motion in regular waves have inherent chaotic characteristic and the ship motion in irregular waves can only be analyzed through ship motion time series due to the effect of wind and waves and the complexity of the ship motion.(2) This paper collects a large quantity of real ship swaying motion data. On the base of preprocessing of sampling and filtering, this paper analyzes all the data selected according to the criterion judging the chaotic characteristic. The results show that a certain chaotic characteristic is held in the ship sway motion which presents a new approach for ship motion prediction.(3) This paper presents the Add-weighted One-rank Local-region Model and the add-weighted predicting model based on the largest Lyapunov exponent. The technique of phase space reconstruction is used to optimizing the structure of RBF neural network and thus improves the prediction performance to the chaotic time series. Prediction results on the real ship motion show that the effective prediction length can amount to above 10 seconds.(4) The prediction methods based on chaotic theory will be less effective due to the nonstationarity of ship motion time series. This paper thus uses the Empirical Mode Decomposition (EMD) to reduce the non-stationary and presents the prediction method based on EMD-RBF. Prediction results on the real ship motion show that this method can obviously reduce the effect of nonstationarity and get better prediction precision.