Motion prediction of floating platforms based on convolutional neural networks

In complex marine environments, the irregular six-degree-of-freedom (6-DoF) motion responses of floating platforms significantly impact structural stability and energy production efficiency. Accurate prediction of platform motion trajectories enables proactive motion control strategies, reducing operational risks under harsh environmental conditions. To address motion response prediction across diverse wave conditions, this study employs Computational Fluid Dynamics (CFD) simulations to develop a numerical model of the floating platform and generate corresponding datasets. Using these datasets, we establish a Convolutional Neural Network (CNN)-based motion prediction model that integrates mooring line tension data with real-time motion parameters to forecast the platform's 5-second trajectory. The results demonstrate that the CNN model achieves high-precision predictions with an average computational latency below 2 ms on general-purpose computing hardware. Moreover, prediction errors remain consistently within a low tolerance range across various operational scenarios, demonstrating robust real-time performance and strong environmental adaptability.