FlowSRNet: A multi-scale integration network for super-resolution reconstruction of fluid flows

Abstract

A wide range of research problems in physics and engineering involve the acquisition of high-resolution data. Recently, deep learning has proved to be a prospective technique for super-resolution (SR) reconstruction of fluid flows. General deep learning methods develop temporal multi-branch networks to improve SR accuracy while ignoring computational efficiency. Further, the generalization ability of the deep learning model in different fluid flow scenarios is still an unstudied issue. In this article, we propose an efficient multi-scale integration network called FlowSRNet to reconstruct the high-resolution flow fields. Specifically, we elaborately design a lightweight multi-scale aggregation block (LMAB) to capture multi-scale features of fluid data, which contains a parallel cascading architecture (PCA) and feature aggregation module (FAM). The residual backbone of FlowSRNet is built by cascading the lightweight multi-scale aggregation blocks (LMABs) (cascaded blocks number N=8) in a serial manner. Also, we present a small architecture LiteFlowSRNet (cascaded blocks number N=2) for comparison. In addition, a corresponding SR dataset is constructed to train and test the proposed model, which contains different kinds of fluid flows. Finally, extensive experiments are performed on different fluid data to evaluate the performance of the proposed model. The results demonstrate that our approach achieve state-of-the-art SR performance on various fluid flow fields. Notably, our method enjoys merit of lightweight, which facilitates the development of the complicated calculation in computational fluid dynamics.

A machine learning model for reconstructing skin-friction drag over ocean surface waves

“…2019, 2021; Bi et al. 2022). Therefore, incorporating a multiscale CNN approach for skin-friction drag reconstructions could potentially enhance the model's performance and appropriately address the complex interactions prevalent across diverse scales.…”

“…Moreover, a series of studies have validated the effectiveness of multiscale CNNs in turbulence enrichment (e.g. Fukami et al 2019Fukami et al , 2021Bi et al 2022). Therefore, incorporating a multiscale CNN approach for skin-friction drag reconstructions could potentially enhance the model's performance and appropriately address the complex interactions prevalent across diverse scales.…”

“…In addition to architectural considerations, we investigated the effect of different batch sizes, explicitly evaluating the performance using batch sizes 32 and 64; these are either the same as or greater than commonly adopted sizes for ML tasks involving spatio-temporal data (see, e.g. Du et al 2018;Guastoni et al 2021;Kim et al 2021;Bi et al 2022). By exploring this parameter, we aimed to strike a balance between computational efficiency and model convergence, ensuring optimal training dynamics for our specific problem (LeCun et al 2002;Keskar et al 2016).…”

“…2021; Bi et al. 2022). By exploring this parameter, we aimed to strike a balance between computational efficiency and model convergence, ensuring optimal training dynamics for our specific problem (LeCun et al.…”

Spatial prediction of the turbulent unsteady von Kármán vortex street using echo state networks

“…In the past years, similar efforts continued by mainly applying different types of deep learning methods to achieve super-resolution in the cylinder wake, channel flow, isotropic turbulence, and turbulent convection. [16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33] Another approach closely related to super-resolution is flow reconstruction from sparse measurements Phys. Fluids 35, 115141 (2023); doi: 10.1063/5.0172722…”