Transient Flow Dynamics over a Bed Drop Integrating the CFD and Machine Learning Models

Channel bed singularities, including variations in channel geometry due to infrastructure, can significantly influence the dynamics of floods resulting from dam failure. This study aimed to examine the impact of bed singularity (drop) on transient flow dynamics in a prismatic channel. It employed numerical and machine learning models refined through experimental results presented in the literature. The three-dimensional Navier-Stokes equations were coupled with the volume of fluid (N-St-VOF) for fluid volume tracking. Subsequently, machine learning (ML) methods were used to predict the flow-structure interactions. The results showed that N-St-VOF and ML methods could satisfactorily predict the transitional flow characteristics by preserving the interface's sharpness. The transitional flow showed distinctive physical characteristics, including a jet-like evolution around the drop. The drop geometry and position along the channel changed the transient flow pattern, velocity, and impact load. The jet-like flow was more violent and elongated with a pronounced hydraulic jump, and kinetic energy was lost by increasing the drop dimensions.