Spatiotemporal dam-break flood impact loads: Experimental study and deep learning analysis

The spatiotemporal distribution of pressure impact on a downstream vertical wall under two dam-break wave modes was investigated. A modification of Cumberbatch's theory predicted the forces exerted by dam-break waves on the downstream vertical wall. The maximum force acting on the downstream vertical wall was determined by predictive modelling. A grey wolf optimizer (GWO)-long short-term memory (LSTM) model was applied to forecast the evolution of the pressure curve under different wave impact modes. Increasing the upstream water depth huintensified the force exerted on the downstream vertical wall in the direction of the incoming flow for a bore-like wave, but triggered a double-peak distribution in the pressure evolution curve for an undular wave. As the downstream water depth hd increased further, the impact pressure generated both borelike and undular wave pressure, leading to pressure fluctuations. Increasing slope i enhanced surge wave energy, elevating the impulsive pressure for a bore-like wave, while the pressure evolution curve transformed from a single to double peak. The bore-like wave forces on the downstream vertical wall during the quasi-hydrostatic phase were accurately predicted. For the same slope i, a power function linked the dimensionless maximum force F ' max and the upstream-downstream water depth ratio h ', with F ' max proportional to ah ' b. The GWO-LSTM accurately modeled pressure evolution over time.