Prediction of three-dimensional flood-flow past bridge piers in a large-scale meandering river using convolutional neural networks

Abstract

The prediction of statistical properties of turbulent flow in large-scale rivers is important for river flow analysis. Large-eddy simulations (LESs) provide a powerful tool for such predictions, however, they require a very long sampling time and significant computing power to calculate the turbulence statistics of riverine flows. In this study, we developed encoder-decoder convolutional neural networks (CNNs) to predict the first- and second-order turbulent statistics of the turbulent flow of large-scale meandering rivers. We trained the CNNs using a dataset obtained from the LES of the flood flow in a large-scale river with three bridge piers, which formed the training testbed. Subsequently, we employed the trained CNNs to predict the turbulent statistics of the flood flow in a river with different bridge pier arrangements, which formed the validation testbed. The CNN predictions for the validation testbed river flow were compared with the simulation results of a separately performed LES to evaluate the performance of the developed CNNs. The results showed that the trained CNNs can successfully produce turbulent statistics of the flood flow in large-scale rivers, such as that chosen for the validation testbed.