Research On Ship Motion Prediction Based On Periodogram And Neural Network

The ship was disturbed by the environmental factors such as sea wave and ocean current. The swing existed inevitably, especially in the bad sea condition, which would bring about prodigious safe hidden trouble for ship's operation. If people could predict the ship motion trend advancing several seconds, not only improved the living and working environment for the crew at sea also improved the safety of ship motion. So the ship sea operation's task could be completed successfully. As a result, it was of great practise significance in the ship motion predictions research.This paper in-depth researched periodogram modeling and predictions. This paper researched periodogram predictions, improved periodogram predictions based on data overlaping section, improved periodogram predictions based on window functions and improved periodogram predictions based on data overlaping section with window functions. From a large number of simulation instances of the above four methods this paper summed up a certain pattern. So this paper improved periodogram predictions further. In improved periodogram predictions based on data overlaping section, this paper summed up that rate of data overlaping section taked 1/2 better. In improved periodogram predictions based on window functions, this paper summed up that window function was Hamming Window better. In improved periodogram predictions based on data overlaping section with window functions, this paper summed up that threshold taked about 1/50 better. Based on the above methods, in improved periodogram predictions based on data overlaping section with window functions this paper summed up that rate of data overlaping section taked 1/2,the window function was Hamming Window and threshold taked about 1/50 best. A large number of simulation results showed that improved periodogram predictions had higher predictions'accuracy for ship motion predictions and met the practise need for applicacitons of ship motion. This paper also researched the Neural Network and used in ship motion predictions. In the research, this paper selected BP Neural Network which was used in ship motion predictions. Firstly, this paper pre-treated ship motion attitude's data. Then, this paper determined the hidden layer neurons number of BP Neural Network by a large number of simulation instances. Thirdly, this paper adjusted the transfer functions of neurons to improve BP Neural Network ship motion predictions. Finally, regularization and introduction of momentum factor were used in improved BP Neural Network ship motion predictions. A large number of simulation instances showed that BP Neural Network ship motion predictions was better after regularization and BP Neural Network ship motion predictions was better after introduction of momentum factor. So improved BP Neural Network ship motion predictions met the practise need for applicacitons of ship motion further.