Based On Wavelet Neural Network Modeling And Prediction Of Ship Motion

There are a lot of nonlinear systems in practice, so the research of nonlinear system is very important. The modeling and prediction of nonlinear system is one of its applications. According to the characteristics of historical data estimate the ship motion after some time. Adjust ship posture ahead to keep the ship balance; this is conducive to ship navigation and crew's habitability.In this dissertation, a wavelet and neural network(WNN) is constructed based on the advantage of wavelet and neural network. Around WNN modeling and prediction and accord to the characteristics of historical data, select the wavelet base which fit for ship motion short time modeling and predicting. The parameters of WNN are identified based on conjugate gradient algorithm. The method and the procedure are given in this dissertation. According to the short time ship motion modeling and prediction, the method of WNN is researched. The forward one step and the multi-step prediction algorithm are derived. Finally, take the Ship longitudinal swing motion for example, the simulation result is compared with other method, result improves the prediction accuracy and the prediction time is lengthened.