Modeling Ship Equations of Roll Motion Using Neural Networks

A neural network‐based approach is applied to fit ship rolling models to experimental data and the initial reporting of this methodology is presented in the work of Xing and McCue. Two multivariable nonlinear models are used to describe the nonlinear forced roll motion of a ship at sea. One, a more traditional model, is based on ordinary differential equations, and the other is based on fractional differential equations (FDEs), which introduced a fractional derivative term to present added hydrodynamic inertia and traditional damping terms. The neural network method is tested using experimental data. The statistical analysis of 20 cases results showed that the FDEs appeared to better approximate the physics of the system.