Short-term Prediction And Simulation Of Ship's Motion Based On LSTM

The short-term prediction of ship motion posture is of great significance to the safety of marine navigation and operations.Affected by the chaos characteristic,variable periodicity and noise signal of ship motion time series,it is difficult to obtain precise forecasting results of ship motion.In order to improve the forecast precision,researchers have spent a lot of efforts on the establishment of complex ship motion mathematical models.However,most of the models do not have portability.Considering the problems and the advantage of BLSTM model in data forecasting,this paper adopts the advanced BLSTM model to apply the model to ship motion prediction research for the first time.Virtual simulation technology is used to build ship real-time motion simulation platform,which can save cost,shorten the development cycle and enhance the human-computer interaction,so as to ensure the safety of the ship navigation without losing the real feeling of the ocean field and reduce the cost of the ship operator training.The main works of this thesis is as follows:1.In this paper,we study the time series characteristics of of ship motion and the theory of LSTM.Based on the LSTM model,which can explore the prediction of future information fully,we use LSTM to train the ship motion data.2.Based on LSTM network model,some useful optimization algorithms are adopted to increase the prediction accuracy from weight initialization to weight updating.For example,Xavier initial method,ReLU network activation function and dropout method is adopted to prevent over fitting when training model.Experiments on LSTM model and BLSTM model are carried out with Tensorflow environment,while the traditional AR model and RNN model are used to compare the experimental results of ship motion prediction.After this,the experimental results are analyzed.The analysis results show that the BLSTM model algorithm has a longer prediction time and higher prediction accuracy than other methods,so it is reasonable and effective to adopt this method.3.In this paper,the OpenGL library is configured by using Microsoft MFC framework to add the data to realize virtual reality simulation platform for real-time ship motion.Simulation results show that the system satisfies the characteristics of virtual reality,and is able to achieve the interactive effect between human and machines.So it instructs the operator steer the rudder rapidly in order to ensure the safety of shipnavigation.