Objectives In complex sea conditions, the roll motion of ships significantly increases navigation risks and directly affects the stability and safety of ships, developing ultra-short-term prediction techniques based on the ship’s current motion state to predict roll motion is therefore a key factor in ensuring safe navigation. To better predict ship roll motion and address the critical issue of safe navigation in the field of marine engineering, the application of the Long Short-Term Memory (LSTM) neural network model in the ultra-short-term prediction of ship roll motion was investigated, and the impact of different degrees of freedom coupling inputs on prediction accuracy was analyzed. Methods An LSTM model was established to conduct ultra-short-term prediction for the ship model DTMB5415, with a focus on analyzing the influence of the coupling between roll and heave on prediction accuracy. Results The results indicate that for the dual-feature input of roll and heave, the significant dynamic coupling effect between them enables the model to more accurately capture the characteristics of ship motion and predict its state. Consequently, the prediction error is significantly lower than that of other feature inputs. Conclusions This research provides important reference for enhancing the accuracy of ship motion prediction and has significant scientific research significance and engineering application prospects for improving the safe navigation and operational efficiency of ships in complex sea conditions.
Kong et al. (Sun,) studied this question.