In this study, we investigate wave-induced nonlinear ship motions using the impulse response function–long short-term memory (IRF-LSTM) hybrid approach. To achieve this, we first calculate the linear ship motions using the classical IRF method. Additionally, we introduce the LSTM technique to account for the nonlinear properties of dynamic systems. For the development of the LSTM machine learning model, the two different types of networks are incorporated so that the influences of the wave excitation and memory effects are considered simultaneously. The integrated LSTM model is trained on the discrepancies between nonlinear and linear responses and complementarily introduced with the linear IRF method in the inference stage. The applied hybrid model demonstrates the potential of machine learning for real-time prediction of nonlinear ship motions in waves.

This content is only available via PDF.
You can access this article if you purchase or spend a download.