Ship Rolling Time Series Analysis, Modeling And Prediction

When sailing in the sea, due to the impact of storms and other interference, ship will have various movements. Serious roll motion has played an important role for reducing ship behaviors in waves, deteriorating operational life-span of ship, reducing work conditions of equipment and reducing the ship's stability. Therefore, stabilizing the ship rolling motion has been the important subject. Reducing ship roll motion is one of the important tasks in ship motion control. Ship rolling as a research object, the paper mainly studies the ship roll motion's simulation and time series prediction.To study the ship rolling motion, It needs to study the impact of the waves on the ship. Waves is the main reason in the ship roll motions, it is necessary to build an effective mathematic model of waves in the research of the ship rolling. Due to randomicity and nonlinear of the ship roll motion, it is extraordinary difficulty to gain exact roll motion model. Roll motion prediction has become a very important issue. It is of great significance to enhance the ship's seaworthiness. The paper is based on the above the issue to study.In this paper, with varieties of methods comparing, for example ARMA model, bi-directional differential, grey cycle extension, and the cycle stack extraction cycle -ANFIS combination modality, a compositive prediction approach of the roll motion time series forecasting, based on extraction cycle and ANFIS model is presented in the thesis. Firstly, based on cycle of extraction methods, the original time series are decomposed of the cycle characteristic signals. Secondly, It can take extract the periodic wave as ANFIS model input. Lastly, it reuses ANFIS multi-step model prediction. This method can not only gain the rolling cycle characteristic signals, but also reduce the difficulty of forecasting, and achieve a long-term forecast. The simulation results illustrate the effectiveness of the method, high accuracy is acquired. It also will be able to overcome some shortcomings of the traditional ARMA time series prediction and bi-directional differential.