Session: 08-06-01 non-presentations

Paper Number: 129679

129679 - Limited-Sensors Based High-Reynolds-Number Turbulent Flow Field Reconstruction Around a Square Cylinder by Deep Learning Methods 

It is of significant importance to reconstruct the high-dimensional global flow field using limited sensor data, considering that sensors are often local, they are limited in full-scale measurements and the global distribution of physical quantities is more relevant in many engineering applications. In the literature, the reconstruction of laminar flows at low Reynolds number (Re) was mostly investigated. This study focuses on high-Re turbulent flows around a body, and aims to systematically investigate the reconstruction performance of deep learning models, including multilayer perceptron (MLP), convolutional neural network (CNN), and generative adversarial neural network (GAN). The extensive assessment encompasses the prediction accuracy of the instantaneous, phase-averaged, and mean flow fields are compared to comprehensively assess the reconstruction accuracy in different scales of separated turbulent flows. We also compared training efficiency. Results shows that all three models have low reconsturction errors in the instantaneous and mean flow fields for very small sensor inputs, and all of them decrease with the increase of sensors. GAN demonstrates a obvious advantage in reconstructing small-scale vortex structures. MLP is better at reconstructing time-averaged flow fields and reconstruction efficiency. Additionally, sensitivity of sensor number and training data size are studied to further augment the accuracy of flow field reconstruction.

Presenting Author: Yong Cao Shanghai Jiaotong University

Presenting Author Biography: Associate Professor of Shanghai Jiaotong University and PhD supervisor of School of Shipbuilding, Oceanic and Architectural Engineering. He is mainly engaged in the research of structural wind engineering, blunt body aerodynamics and computational fluid dynamics, and has published more than 30 papers in J. Fluid Mech. and Phys. Fluids and other major journals in the field (15 first author/correspondence papers). He is a member of the Wind Engineering and Industrial Aerodynamics Committee of the Chinese Society of Aerodynamics, and has been selected as a member of the Shanghai Overseas High-level Talent Introduction Programme and the Young Scientific and Technological Talents Sailing Programme under the Shanghai Science and Technology Innovation Action Plan. He was awarded the Society Award (Thesis Award) by the Japan Wind Engineering Association, the Best CFD Picture Award by the Japan Society of Fluid Mechanics, the Young Excellent Paper Award by the China Aerodynamics Conference, and the Guest Editor of two international journals.

Authors:

Yong Cao Shanghai Jiaotong University
Rui Li Shanghai Jiaotong University