Ship Motion Prediction And Wave Inversion Based On Deep Learning Method

When ship sailing at the ocean,the wave-induced ship motions always bring adverse effects on carrying out maritime operations.As a result,it is important to predict the ship motions a few seconds in advance.What is more,because of the difficulty of measuring the sea state,using ship motions as a method for determining the ocean waves is useful.In this study,deep machine learning is used to solve these two problems.For accurate ship motion prediction with the ocean wave given,the traditional methods are to solve the hydrodynamic equations,which are often based on linear theory.Different from the traditional methods,this thesis uses the long-short-term-memory(LSTM)neural network in deep machine learning to predict the ship’s pitch motions with the wave time history as input.In order to improve the quality of the motion prediction,three learning models are proposed and tested:(1)a refined learning model that predicts the difference between the linear model and the real records;(2)a learning model that depends on the wave heights and wave steepness at one point;and(3)a learning model that depends on the wave heights at multi-pointsFor determining the ocean waves from ship motions,the convolutional neural network(CNN)and the deep artificial neural network(ANN)are employed to solve the sea state classification problem and to calculate wave time series,respectively.During this process,the Hilbert transform method is used to map the 1D ship motion into 3D picture to classify the sea state and to gain information about the sea state.