단일파로 인해 생성된 바닥 근처 흐름 특성 예측

Abstract

Accurately predicting near-bed flow in coastal environments is essential for understanding wave-induced bottom dynamics. Near- bed flow in regions adjacent to the shoreline is a key driver of geomorphological changes such as erosion and deposition. Quantitatively predicting this flow is considered as a key task for coastal protection and disaster response. This study employs SedFoam, a one-dimensional two-phase flow model based on OpenFOAM, to tackle the measurement challenges that arise near the bed under dynamic wave conditions. The governing equations for the fluid and sediment phases were solved with SedFoam, accounting for inter-phase momentum exchange, particle collisions, and shear stress, to generate high-resolution vertical flow profiles under single-wave conditions. While SedFoam offers detailed and physically consistent results, its high computational cost limits its use for real-time applications. Consequently, the convolutional neural network (CNN) model was developed using outputs from a number of SedFoam simulations involving various wave periods, flow velocities, and sediment properties. The CNN was trained to predict key near-bed flow features including bed shear stress, maximum sheet flow layer thickness, and vertical velocity profiles. By leveraging its ability to recognize spatial patterns, the CNN reduces computation time while maintaining prediction accuracy. The CNN model predicted time-varying bed shear stress and velocity profiles. The model also successfully estimated the maximum thickness of the sheet flow layer, and comparisons with existing empirical formulas for bed shear stress further demonstrated the reliability of the model. This study highlights the potential of combining physics-based simulations with machine learning to enable fast and accurate prediction of near-bed flow characteristics in coastal environments.