A neural-corrector method for prediction of the vertical motions of a high-speed craft

A machine learning method is developed which capable of making predictions nonlinear ship motions in a range wave conditions when trained on response data from only single seaway. The formulated around the equations motion time domain, but are augmented with data-driven terms that act to correct force-balance equations. forcing modeled using Long Short-Term Memory (LSTM) recurrent neural networks. resulting hybrid governing solved numerically. Predictions compared test 2-DOF Generic Prismatic Planing Hull (GPPH) at forward speed head seas, histories given for both regular and irregular waves. training requirements classify specific seaway investigated quantified. over significant heights peak periods performed show effectiveness generalizing across different environmental conditions.