Session: 08-05-01 AI‑Enabled Marine Sensing, Monitoring & Digital Twins

Submission Number: 182360

Study on Ocean Wave Reconstruction Utilizing Deep Learning 

Accurate observation and prediction of wave parameters are essential during ship sea trials. Accordingly, wave inversion based on ship motion responses is of paramount importance. In this study, ship motion responses to irregular waves across various sea states were initially computed using the potential flow theory method, resulting in the development of a comprehensive motion response database. Leveraging this database, deep learning models were constructed utilizing time-series motion data. These models employ ship motion responses as inputs to predict critical wave parameters, including peak wave period, significant wave height, and wave direction. Specifically, deep learning architectures incorporating coupled degrees of freedom (DOF) were explored and established, encompassing: Single-DOF (Heave), Dual-DOF (Heave and Pitch), and Triple-DOF (Heave, Pitch, and Roll). Wave parameter prediction was conducted using each model variant on test datasets. The findings indicate that the triple-DOF coupled model significantly outperforms the others in prediction accuracy, achieving improvements of 20% under sea state 4 and 15% under sea state 5 in wave inversion accuracy. Utilizing a data fusion strategy, training samples derived from various ocean wave spectra were combined to form composite inputs. This approach increased the diversity of features within the training dataset, consequently improving the model’s generalization performance. Relative to the initial model, the enhanced framework demonstrated a 20% increase in wave inversion accuracy under heterogeneous spectral conditions. 

Presenting Author: Donglin Liu Huazhong University of Science and Technology

Presenting Author Biography: Donglin Liu is a doctoral student at the School of Naval Architecture and Ocean Engineering of Huazhong University of Science and Technology. His main research fields are computational fluid dynamics and wave loads.

Authors:

Cheng Chang Huazhong University of Science and Technology
Donglin Liu Huazhong University of Science and Technology
Liwei Liu China Ship Development and Design Center
Chaobang Yao Huazhong University of Science and Technology