Data-Driven Models For Levee Breach Flows

This study deals with data-driven models that can estimate the breach discharge and approach depth for levee breach flows. The ML models are trained from the data sourced from the validated 2D numerical models. The MLP model uses limited features to estimate the breach discharge non-dimensionalized as a ratio of breach discharge to the inflow discharge. The MLP models are validated with 15 different experimental observations. The results show that the MLP models can estimate the breach discharge with high accuracy.