Machine learning-based prediction of flow field in combined wave-current flows

The flow field of combined wave-current flows is essential for coastal hydrodynamics and sediment transport. However, an accurate prediction of flow field in combined wave-current flows still remains a challenging task for coastal and ocean engineering. Artificial intelligence has become an important method for the prediction capabilities in recent years. In order to investigate the applicability of Artificial intelligence for the flow field forecast in combined wave-current flows, two machine learning methods were tested using the experimental data obtained from the PIV system. Both Nonlinear autoregressive neural network and Long Short-Term Memory approaches were adopted to forecast the flow field of combined wave-current flows. Good performance of predicting the velocities was observed for the NAR approach. Results suggest higher accuracy and faster speed for the NAR approach than the LSTM approach. This provides some guidance for further developments of data-driven ocean modelling. • The flow field in combined wave-current flows was predicted based on Artificial intelligence technique. • Both Nonlinear autoregressive neural network and Long Short-Term Memory approaches were adopted to forecast the velocity field. • Prediction results suggest higher accuracy and faster speed for Nonlinear autoregressive neural network, compared with the Long Short-Term Memory approach.