AI-Augmented Computational Fluid Dynamics (CFD) Models for Predicting Nonlinear Wave-Structure Interactions

Accurately predicting nonlinear wave-structure interactions remains one of the most challenging problems in marine hydrodynamics due to turbulence complexity and transient boundary effects. This study presents an AI-augmented Computational Fluid Dynamics (CFD) framework that integrates a hybrid Convolutional-LSTM neural network with the Navier-Stokes solver to accelerate and enhance hydrodynamic predictions. High-fidelity CFD simulations generated through OpenFOAM form the foundation for training the AI surrogate model to approximate pressure, velocity, and free-surface dynamics under various sea states. The proposed hybrid framework achieves a ۶۲% reduction in computational cost while maintaining a correlation coefficient of R² > ۰.۹۷ compared to full CFD benchmarks. Validation against benchmark datasets demonstrates accurate reconstruction of vortex shedding, pressure peaks, and transient wave slamming. Validation against experimental wave basin data (MARIN benchmark) confirms that the model accurately reproduces vortex shedding and pressure peaks under nonlinear wave impacts. The model's capacity for real-time and cross-geometry generalization positions it as a promising solution for digital twin integration, offshore design optimization, and autonomous vessel control. These findings demonstrate the feasibility of AI-assisted CFD as a physically consistent and computationally efficient tool for marine hydrodynamics.

کلیدواژه ها:

Fluid Dynamics ، AI-augmented model ، deep neural networks ، Nonlinear Wave ، Marine AI systems