Predicting 3-DoF motions of a moored barge by machine learning
Yu Yang, Tao Peng, Shijun Liao
Abstract
The real-time prediction of a floating platform or a vessel is essential for motion-sensitive maritime activities. It can enhance the performance of motion compensation system and provide useful early-warning information. In this paper, we apply a machine learning technique to predict the surge, heave, and pitch motions of a moored rectangular barge excited by an irregular wave, which is purely based on the motion data. The dataset came from a model test performed in the deep-water ocean basin, at Shanghai Jiao Tong University, China. Using the trained machine learning model, the predictions of 3-DoF (degrees of freedom) motions can extend two to four wave cycles into the future with good accuracy. It shows great potential for applying the machine learning technique to forecast the motions of offshore platforms or vessels.