Nonlinear wave evolution with data-driven breaking.

Area of Science:

• Fluid dynamics
• Oceanography
• Machine learning

Background:

• Wave breaking is the primary mechanism for ocean wave energy dissipation and influences sea state properties.
• Its turbulent nature makes direct numerical simulations computationally prohibitive for complex scenarios.
• Accurate prediction of wave breaking is vital for understanding ocean-atmosphere interactions, pollution, and rogue waves.

Purpose of the Study:

• To develop a computationally efficient and accurate method for predicting ocean wave breaking.
• To overcome the limitations of traditional numerical simulations for turbulent wave breaking.
• To enhance the understanding of wave evolution and its impact on marine environments.

Main Methods:

• A blended machine learning framework combining a physics-based nonlinear evolution model with a recurrent neural network (RNN).
• Utilized wave tank measurements for training data, avoiding computationally intensive simulations.
• Employed a long short-term memory (LSTM) neural network for finite-domain corrections to the evolution model.

Main Results:

• The blended framework accurately predicts wave breaking events and their effects on wave evolution.
• Demonstrated excellent predictive performance even with external, previously unseen data.
• Successfully integrated physics-based models with machine learning for complex fluid dynamics problems.

Conclusions:

• The developed machine learning framework offers a powerful tool for predicting ocean wave breaking.
• This approach significantly reduces computational demands compared to traditional methods.
• The findings have implications for oceanography, marine engineering, and climate modeling.