Relative motion prediction of pontoon bridge module offshore connection based on deep learning.

Improving the safety and efficiency of offshore connection of pontoon bridge modules and avoiding connection failures or collision from their relative motion due to waves is currently an important study for landing operations. Thus, this paper proposes an online prediction method for the relative motion of the offshore pontoon bridge module connection based on a long short-term memory (LSTM) deep learning architecture. The developed scheme processes the motion response data from the wave tank to de-noise and segment them, employs the sample data obtained for training and testing, and generates a prediction model operating under various working conditions. Through extensive experiments, we verify that without requiring any information on the module and waves, our method attains a high forecast accuracy and provides a decision basis for the offshore connection of the pontoon bridge modules. • This paper proposes an online prediction method using a deep learning architecture for the connection of pontoon bridge modules. • The wavelet denoising method is implemented for data treatment of the pontoon bridge module connection with the measured data. • According to the difference in prediction accuracy in different stages, the required sample data is analyzed. • The proposed prediction model achieves satisfactory pontoon bridge module offshore Connection prediction performance. [ABSTRACT FROM AUTHOR]