A Speed Prediction Model Using A HIL Simulator and Neural Network for Cargo Ships

This paper introduces the research and construction of a speed prediction model for ships using an artificial neural network and a hardware-in-the-loop (HIL) simulator. The proposed model is an MLP feedforward neural network training based on the back-propagation algorithm. Instead of using historical data such as from the AIS identification system, noon report data, or directional wave spectrum data, this study uses a dataset generated from a HIL simulator, with the plant model being a dynamical model of the real ship. In particular, the 3D dynamic model is built based on the 6-degree-of-freedom (6-DOF) equations of the ship, mathematical models of environmental disturbances, and the propulsion system. The testing results show that the proposed model works reliably and accurately with the testing dataset. The model can be applied to find optimal routes for cargo ships to improve fuel efficiency.