Study On Extremely Short-term Prediction Of Ship Sway Motion

In the course of sea navigation,ships will be affected by the influence of sea breeze,wave,ocean currents and other factors to produce six degrees of freedom of the swaying movement,the ship’s sea operations and ship-borne weapons and equipment caused disturbance.If the ship’s swaying motion can be predicted,it will provide effective basic information for the motion compensation of the ship-borne stable platform,so as to improve the compensation accuracy.Therefore,this paper aims at the ship’s swaying motion,and studies the prediction method of the ship’s extremely short-term swaying motion.The main research work is as follows:Firstly,a dual-model ship sway prediction model is constructed for ship-borne stabilization platform,and by predicting the trend of ship swing movement,the ship’s motion prediction information is provided for the movement compensation of the stability platform,thus enhancing the compensation capacity of the stability platform.In view of the high-precision requirement of ship-borne stability platform compensation,the online learning model is designed in the dual model frame,the network parameters are adjusted according to the navigation environment,and the prediction accuracy is improved.In view of the high real-time requirements of ship-borne stability platform,the offline prediction is constructed in the dual-model prediction framework,and the network parameters of online learning training are used to make a prediction of the future ship’s swing trend.Secondly,from the frequency characteristics of ship swing motion,a multi-scale Long Short-Term Memory prediction algorithm is proposed.The method first makes a multi-scale analysis of the ship’s swing motion signal,decomposes the swinging motion signal into the scale components of different frequencies,then establishes the Long Short-Term Memory model set for the characteristics of different scale components,uses multiple Long Short-Term Memory models to study and predict the signals of different scale components,and finally,completes the overall prediction of ship sway movement by integrating the signal prediction of different scales.The experimental results show that the multi-scale Long Short-Term Memory prediction algorithm has better prediction ability than the Long Short-Term Memory model,which improves the prediction accuracy of ship sway movement.Finally,in view of the lack of learning ability of non-stable ship swing data in the Long Short-Term Memory,which causes the problem of large prediction deviation,this paper proposes an EMD-LSTM combination prediction algorithm.This method combines the non-smooth processing ability of the Experience Mode Decomposition and the time-series signal learning ability of the Long Short-Term Memory model,decomposes the non-smooth ship sway data into several stable components,uses the Long Short-Term Memory model to learn the volatility characteristics of each stable component,and finally reconstructs the prediction results of each component.The experimental results show that the combination prediction algorithm based on EMD-LSTM shows good prediction performance in dealing with non-smooth ship swing motion data compared with the Long Short-Term Memory model.Therefore,the algorithm is effective for the extreme short-term prediction of ship rocking motion,and provides a reference method to improve the prediction accuracy of ship’s extreme short-term swing movement.