AI-Based Optimization and Hydrodynamic Performance of Floating Pontoon Breakwaters through Experimental and Numerical Modeling

Abstract

Ensuring effective coastal protection requires not only attenuating incoming wave energy but also preserving the operational reliability of marine infrastructure. This study investigates the hydrodynamic behavior of two pontoon-type floating breakwaters—simple and stepped—through a combined program of new laboratory wave-flume experiments and high-fidelity numerical modeling in ANSYS AQWA. A 1:15 physical model was constructed and tested under regular waves with incident heights of 2.2–3.3 cm and periods of 0.44–0.84 s. The experiments showed that the simple pontoon achieved optimum energy dissipation at shorter wave periods (0.44 s), whereas the stepped configuration performed more efficiently in longer waves (0.84 s). Transmission coefficients (Kt) ranged from 0.2 to 1.5, with the stepped design consistently exhibiting stronger wave-energy reduction. Numerical simulations reproduced the laboratory findings with a maximum deviation of about 8%. To further enhance performance, an artificial-intelligence framework was developed that integrates artificial neural networks (ANNs) for rapid prediction of wave-induced loads, genetic algorithms (GAs) for multi-objective structural optimization, and reinforcement learning (RL) for adaptive real-time control. Together these methods lowered the mean transmission coefficient from 0.52 to 0.40 and improved computational efficiency by 88%. In addition, a prototype digital-twin environment, linked to an IoT sensor network, was established to enable continuous structural monitoring and to support condition-based maintenance. The results highlight the dominant influence of breakwater geometry, draft, and mooring characteristics on hydrodynamic behavior and demonstrate that AI-guided optimization provides a practical, scalable, and cost-effective pathway for designing resilient and low-carbon coastal infrastructure.