Data‐Driven Prediction of Turbulent Flow Statistics Past Bridge Piers in Large‐Scale Rivers Using Convolutional Neural Networks

• Published in: Water resources research 2022-01, Vol.58 (1), p.n/a
• Main Authors: Zhang, Zexia, Flora, Kevin, Kang, Seokkoo, Limaye, Ajay B., Khosronejad, Ali
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Bridge piers; Bridges; Coders; Computer applications; convolutional neural network; Current meandering; Flood flow; flood flow predictions; Flood management; Flood predictions; Floods; Flow characteristics; Fluid dynamics; Large eddy simulation; large‐scale rivers; Learning algorithms; Machine learning; Meandering; Neural networks; Physics; Piers; Predictions; River flow; River meanders; Rivers; Simulation; Statistical analysis; Statistical methods; Statistical prediction; Statistics; Stream flow; Test stands; Training; Turbulence; Turbulent flow
• ISSN: 0043-1397; 1944-7973
• DOI: 10.1029/2021WR030163

Prediction of statistical properties of the turbulent flow in large‐scale rivers is essential for river flow analysis. The large‐eddy simulation (LES) provides a powerful tool for such predictions; however, it requires a very long sampling time and demands 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 turbulence statistics of the turbulent flow of large‐scale meandering rivers using instantaneous LES results. We train the CNNs using a data set obtained from LES of the flood flow in a large‐scale river with three bridge piers—a training testbed. Subsequently, we employed the trained CNNs to predict the turbulence statistics of the flood flow in two different meandering rivers and bridge pier arrangements—validation testbed rivers. The CNN predictions for the validation testbed river flow were compared with the simulation results of a separately done LES to evaluate the performance of the developed CNNs. We show that the trained CNNs can successfully produce turbulence statistics of the flood flow in the large‐scale rivers, that is, the validation testbeds. Plain Language Summary It is vital to understand the physics of flood flow in rivers. Such understanding can help practicing engineers and researchers to (a) appropriately design infrastructures along and across rivers and (b) better protect the river environment. As a way to understand the flood flow in rivers, we are examining the possibility of a type of machine learning method to produce mean flow characteristics of the flood flow. We show that the machine learning tools can enable reliable prediction of the flood flows at a small fraction of the computational cost of the existing models. Key Points Convolutional neural network algorithms are developed to generate 3D realizations of the time‐averaged flood flow in large‐scale rivers Large‐eddy simulation results are employed to train and validate the developed convolutional neural network algorithms Validation studies show that convolutional neural network algorithms could successfully produce turbulence statistics of flood flows