Time series prediction of ship maneuvering motion at sea based on higher order dynamic mode decomposition

Ship maneuvering motion at sea is a complex dynamic process with multi-frequency and time-varying characteristics, making it hard to be predicted accurately. In this paper, a novel method based on Higher Order Dynamic Mode Decomposition (HODMD) is applied to predict the maneuvering motion of a full-scale ship, the teaching-training vessel YUKUN, at sea. The sea trial data of the vessel is used as training data, and the time series predictions of 35° turning circle and 10°/10° zigzag maneuvers of the vessel are carried out. The dynamic characteristics of the ship maneuvering motion at sea are revealed by modal analysis, and the ship maneuvering motion at sea is predicted by using the dominant modes to reconstruct the snapshots. A study is conducted to investigate the impact of the number of time-lagged snapshots in HODMD on prediction accuracy and efficiency. It is shown that the prediction results by HODMD are more consistent with the sea trial data than those by the standard DMD, demonstrating that the proposed method based on HODMD is promising for predicting the ship maneuvering motion of full-scale ships at sea.