Parameterization of waves interacting with homogenous permeable breakwaters: Integrating physical and numerical models with machine learning

Abstract

Breakwaters effectively reduce the incoming wave energy by wave dissipation and reflection. Thus, previous physical and numerical modelling studies have been devoted to improve the functionality and stability of these type of structures. However, less effort has been devoted to investigate the structural functionality of permeable breakwaters solely made of prefabricated elements. In this study, two-dimensional laboratory experiments on waves interacting with permeable breakwaters made of armor units were carried out in a wave flume. Analysis of laboratory data show that permeable breakwaters made of cubes presented larger dissipation and reflection coefficients than those made of tetrapods. A volume of fluid-type numerical model, based on the Volume-Averaged Reynolds-Averaged Navier-Stokes (VARANS) equations, was calibrated using the laboratory data. The VARANS model reproduces complex wave transformation processes including wave reflection, dissipation, and transmission. Hence, the numerical model was further employed to conduct a parametric study considering different breakwater geometry, water levels and wave conditions. New parametric formulations for estimating the wave reflection, transmission, and dissipation coefficients were developed using numerical modelling results and Machine Learning (ML). Hence, this study demonstrates the potential of combining physical and numerical modelling with ML for the study of wave-structure interaction.