•  
  •  
 

Abstract

The present study establishes deep learning models to predict the tensions and inclinations of mooring lines of a floating barge and verifies applicability of these models. Hydrodynamic and mooring analyses are conducted on the dataset used for the deep learning models. Three types of neural network models include a deep neural network (DNN) with input data representing the six degrees of freedom motions of a floating barge, convolutional neural network (CNN) with input images of the floating barge and mooring lines on a horizontal plane, and hybrid neural network (HNN) that consolidates the characteristics of DNN and CNN models. The output labels for the deep learning models are the tension and inclination of each mooring line. The models are trained using various hyperparameters, and the effect of the hyperparameters on model performance is investigated. The accuracy of the established models is evaluated through testing. The test results indicate that the DNN and HNN models outperform the CNN model, which is constrained by its dependency on two-dimensional image input data. The DNN and HNN models are able to accurately predict the mooring line tension and inclination under various environmental conditions, except for several peaks in harsh environments. The present study shows that the established deep learning models have the potential to replace a mooring monitoring system with fast and accurate predictions. Further research is required to investigate their applicability in mooring systems with multiple mooring lines and turret mooring systems.

Download
COinS