Riverbed evolution simulation and application based on machine learning

In the process of riverbed evolution, the hydrodynamic model and sediment transport model are generally used to decouple riverbed deformation. Based on the numerical simulation results of long-sequence riverbed evolution by machine learning, this paper compares and analyzes the differences between machine learning and numerical simulation results for the average scouring and silting results of river channels and characteristic points, summarizes and proposes a calculation method for machine learning to predict riverbed evolution, which greatly reduces the time required for riverbed scouring and silting simulation. At the same time, the proportion of training data is optimized and compared with the numerical simulation results. Machine learning is applied to riverbed evolution simulation, and the overall scouring and silting calculation error does not exceed 10%.