Prediction Of Ship Roll Motion Based On Improved NARX Neural Network

In the actual ship navigation,affected by external factors such as wind and waves,the ship experiences roll and roll during the swing,which seriously threatens navigation safety.Intense rollover motion can cause crew seasickness,resulting in reduced work efficiency;increased risk of loss of goods;In severe cases,it may cause the ship to capsize,etc.The accurate prediction of ship rollover can provide a reference basis for ship design and improvement,improve the control accuracy of the equipment used by the ship and the safety and stability of ship navigation,which is of great significance to the ship navigation and transportation and the operation efficiency of the ship in the wind and wave.In this thesis,a nonlinear Auto Regressive models with Exogenous Inputs(NARX)prediction model is proposed to predict ship roll motion,and the relevant data used in this thesis are obtained when the Dalian Maritime University practice ship "Yukun" is used to conduct direct navigation experiments at sea.This method considers the influence of external factors during the operation of the real ship,takes the measured historical data of the ship’s roll motion as input,and also adds the data affected by wind direction,wind speed,water temperature and air pressure factors as external input,to improve the prediction accuracy of the ship’s roll motion to a certain extent,on this basis,the particle swarm algorithm(PSO)is used to optimize the weight of the input layer to the hidden layer,the weight of the hidden layer to the output layer,the bias of the hidden layer neurons and the bias of the output layer neurons,the process is as follows.First,initialize the number of swarms,population and maximum number of iterations,secondly,calculate the fitness function value of each particle,and then compare the fitness function value of each particle with the individual extreme value and global extreme value again,update the speed and position of each particle according to the comparison results,and finally repeat the previous steps until the maximum number of iterations to obtain the parameters of the NARX neural network,use the optimized parameters to initialize the NARX ship roll motion prediction model,and train the model with BP algorithm.Finally,the prediction model of rollover motion of PSO-NARX ships is established.This method can avoid the shortcomings of the neural network convergence to the local optimal and the initial position sensitivity of the weight parameters caused by only the BP learning algorithm,improve the robustness of the network model,and predict the roll motion of the ship more accurately.In this thesis,five kinds of ship rollover prediction models(BP prediction model,PSO-BP prediction model,RBF ship prediction model,NARX prediction model and PSO-NARX prediction model)are used for prediction experiments,and their prediction results are compared,and it can be seen from the comparison results that the prediction value of PSO-NARX ship rollover motion prediction model has been greatly improved compared with the other four models It is also proved that the PSO-NARX ship roll motion prediction model is a good model with practical significance.