Toward neural-network-based large eddy simulation: application to turbulent channel flow

Jonghwan Park,Haecheon Choi

DOI: https://doi.org/10.1017/jfm.2020.931

IF: 3.7

2021-03-05

Journal of Fluid Mechanics

Abstract:Abstract

mechanics,physics, fluids & plasmas

What problem does this paper attempt to address?

Constrained large-eddy simulation of laminar-turbulent transition in channel flow

Zhao Yaomin,Xia Zhenhua,Shi Yipeng,Xiao Zuoli,Chen Shiyi

DOI: https://doi.org/10.1063/1.4895589

IF: 4.6

2014-01-01

Physics of Fluids

Abstract:A constrained large-eddy simulation (CLES) of a laminar-turbulent transition in a temporally developing channel flow is performed. First, we confirm the capability of CLES to simulate this transition problem using the a priori Reynolds stress estimated from a direct numerical simulation. Based on the analysis of the Reynolds stress during the transition process, an intermittency factor is introduced in the Reynolds-averaged Navier-Stokes equation (RANS) model to account for the transition property. Two simple approaches are used to construct the intermittency factor. One is based on the shape factor, and the other is based on the coefficients of Smagorinsky models. The CLES results using the intermittency modified RANS model can accurately predict the onset of the transition and the basic transition process, in a manner similar to a large eddy simulation with dynamics Smagorinsky model (LES-DSM). Meanwhile, CLES preserves its advantage over LES-DSM in the turbulent state. The present work illustrates that CLES can be used to simulate transitional flows. (C) 2014 AIP Publishing LLC.
Large Eddy Simulation of Spanwise Rotating Turbulent Channel Flow with Dynamic Variants of Eddy Viscosity Model

Zhou Jiang,Zhenhua Xia,Yipeng Shi,Shiyi Chen

DOI: https://doi.org/10.1063/1.5011236

IF: 4.6

2018-01-01

Physics of Fluids

Abstract:A fully developed spanwise rotating turbulent channel flow has been numerically investigated utilizing large-eddy simulation. Our focus is to assess the performances of the dynamic variants of eddy viscosity models, including dynamic Vreman’s model (DVM), dynamic wall adapting local eddy viscosity (DWALE) model, dynamic σ (Dσ) model, and the dynamic volumetric strain-stretching (DVSS) model, in this canonical flow. The results with dynamic Smagorinsky model (DSM) and direct numerical simulations (DNS) are used as references. Our results show that the DVM has a wrong asymptotic behavior in the near wall region, while the other three models can correctly predict it. In the high rotation case, the DWALE can get reliable mean velocity profile, but the turbulence intensities in the wall-normal and spanwise directions show clear deviations from DNS data. DVSS exhibits poor predictions on both the mean velocity profile and turbulence intensities. In all three cases, Dσ performs the best.
Constrained Large Eddy Simulation of Wall-Bounded Turbulent Flows

Shiyi Chen,Yipeng Shi,Zuoli Xiao,Zhenhua Xia,Jianchun Wang

DOI: https://doi.org/10.1007/978-3-642-31818-4_10

IF: 4.6

2013-01-01

Physics of Fluids

Abstract:We present a novel simulation tool-constrained large eddy simulation (CLES), for numerical experiments on the wall-bounded turbulent flows. Different from the traditional large eddy simulation(LES) and the available hybrid RANS/LES approaches, the CLES method computes the whole flow domain by solving the LES equations with a Reynolds-stress-constrained (RSC) subgrid-scale (SGS) stress model in the near-wall region and a traditional SGS stress model in the rest.The CLES approach is validated by simulating the turbulent channel flow and flow around a circular cylinder. With the same grid resolutions, CLES can successfully simulate all these flow regimes as well as DES and other available methods. For the case of attached flows, CLES is able to eliminate the non-physical Log-Layer Mismatch problem in traditional hybrid RANS/LES methods successfully, and to predict mean velocity profile, turbulent stresses and skin friction coefficient more accurately compared with the DES. For the case of detached flows, the performance of CLES is comparable to DES.
Discrete large eddy simulation

lan tao,k r rajagopal,g q chen

DOI: https://doi.org/10.1016/S1007-5704(01)90023-1

IF: 4.186

2001-01-01

Communications in Nonlinear Science and Numerical Simulation

Abstract:Despite the intense effort expended towards obtaining a model for describing the turbulent flows of fluid, there is no model at hand that can do an adequate job. This leads us to look for a non-traditional approach to turbulence modeling. In this work we conjoin the notion of large eddy simulation with those of fuzzy sets and neural networks to describe a class of turbulent flows. In previous works we had discussed several issues concerning large eddy simulation such as filtering and averaging. Here, we discuss tile use of fuzzy sets to improve the filtering procedure.
A recursive neural-network-based subgrid-scale model for large eddy simulation: application to homogeneous isotropic turbulence

Chonghyuk Cho,Jonghwan Park,Haecheon Choi

DOI: https://doi.org/10.1017/jfm.2024.992

IF: 3.7

2024-11-30

Journal of Fluid Mechanics

Abstract:We introduce a novel recursive procedure to a neural-network-based subgrid-scale (NN-based SGS) model for large eddy simulation (LES) of high-Reynolds-number turbulent flow. This process is designed to allow an SGS model to be applicable to a hierarchy of different grid sizes without requiring expensive filtered direct numerical simulation (DNS) data: (1) train an NN-based SGS model with filtered DNS data at a low Reynolds number; (2) apply the trained SGS model to LES at a higher Reynolds number; (3) update this SGS model with training data augmented with filtered LES (fLES) data, accommodating coarser filter size; (4) apply the updated NN to LES at a further higher Reynolds number; (5) go back to Step (3) until a target (very coarse) filter size divided by the Kolmogorov length scale is reached. We also construct an NN-based SGS model using a dual NN architecture whose outputs are the SGS normal stresses for one NN and the SGS shear stresses for the other NN. The input is composed of the velocity gradient tensor and grid size. Furthermore, for the application of an NN-based SGS model trained with one flow to another flow, we modify the NN by eliminating bias and introducing a leaky rectified linear unit function as an activation function. The present recursive SGS model is applied to forced homogeneous isotropic turbulence (FHIT) and successfully predicts FHIT at high Reynolds numbers. The present model trained from FHIT is also applied to decaying homogeneous isotropic turbulence and shows an excellent prediction performance.

mechanics,physics, fluids & plasmas
Neural Ideal Large Eddy Simulation: Modeling Turbulence with Neural Stochastic Differential Equations

Anudhyan Boral,Zhong Yi Wan,Leonardo Zepeda-Núñez,James Lottes,Qing Wang,Yi-fan Chen,John Roberts Anderson,Fei Sha

DOI: https://doi.org/10.48550/arXiv.2306.01174

IF: 5.414

2023-06-01

Machine Learning

Abstract:We introduce a data-driven learning framework that assimilates two powerful ideas: ideal large eddy simulation (LES) from turbulence closure modeling and neural stochastic differential equations (SDE) for stochastic modeling. The ideal LES models the LES flow by treating each full-order trajectory as a random realization of the underlying dynamics, as such, the effect of small-scales is marginalized to obtain the deterministic evolution of the LES state. However, ideal LES is analytically intractable. In our work, we use a latent neural SDE to model the evolution of the stochastic process and an encoder-decoder pair for transforming between the latent space and the desired ideal flow field. This stands in sharp contrast to other types of neural parameterization of closure models where each trajectory is treated as a deterministic realization of the dynamics. We show the effectiveness of our approach (niLES - neural ideal LES) on a challenging chaotic dynamical system: Kolmogorov flow at a Reynolds number of 20,000. Compared to competing methods, our method can handle non-uniform geometries using unstructured meshes seamlessly. In particular, niLES leads to trajectories with more accurate statistics and enhances stability, particularly for long-horizon rollouts.
Wavelet-based adaptive implicit large-eddy simulation of turbulent channel flow

Giuliano De Stefano

DOI: https://doi.org/10.1016/j.compfluid.2024.106190

IF: 3.077

2024-02-05

Computers & Fluids

Abstract:The wavelet-based adaptive implicit large-eddy simulation approach to computational modeling of turbulent flows is introduced. The direct solution of the wavelet-filtered compressible Navier–Stokes equations is attained without employing any explicit modeling procedure for the unclosed terms. Instead, the dissipative properties of the numerical method using adaptive wavelet collocation are exploited for implicit subgrid-scale modeling. The performance of the newly proposed method was preliminarily assessed by carrying out adaptive simulations of fully developed turbulent supersonic channel flow, which represents a well-established benchmark for wall-bounded attached compressible turbulent flows. The comparison against higher-fidelity solutions proved both the feasibility and the effectiveness of the implicit modeling approach for wavelet-based large-eddy simulation. The consistency of the wavelet-based approach with turbulence physics guaranteeing the correct amount of implicit dissipation was demonstrated.

mechanics,computer science, interdisciplinary applications
Large Eddy Simulation of Sediment-Laden Turbulent Flow in an Open Channel

Zhanhong Wan,Zhilin Sun,Zhenjiang You,Qiyan Zhang

DOI: https://doi.org/10.1142/s0217979208039666

2008-01-01

Abstract:Sediment transport in fully developed turbulent open channel flow has been investigated using large eddy simulation (LES) of the incompressible Navier–Stokes equations. The scalar transport equation of the sediments concentration, which is based on the continuous-phase approach, is adopted. The settling process is taken into account with a modified settling velocity appearing in the sediment concentration equation. A Smagorinsky model allowing for the interaction between the fluid flow and the suspended sediment is used to simulate the unresolved, subgrid scale terms. The LES results are compared with the experimental data, and good general agreement is achieved.
Neural network-based closure models for large-eddy simulations with explicit filtering

Mark Benjamin,Gianluca Iaccarino

2024-08-31

Abstract:Data from direct numerical simulations of turbulent flows are commonly used to train neural network-based models as subgrid closures for large-eddy simulations; however, models with low a priori accuracy have been observed to fortuitously provide better a posteriori results than models with high a priori accuracy. This anomaly can be traced to a dataset shift in the learning problem, arising from inconsistent filtering in the training and testing stages. We propose a resolution to this issue that uses explicit filtering of the nonlinear advection term in the large-eddy simulation momentum equations to control aliasing errors. Within the context of explicitly-filtered large-eddy simulations, we develop neural network-based models for which a priori accuracy is a good predictor of a posteriori performance. We evaluate the proposed method in a large-eddy simulation of a turbulent flow in a plane channel at $Re_{\tau} = 180$. Our findings show that an explicitly-filtered large-eddy simulation with a filter-to-grid ratio of 2 sufficiently controls the numerical errors so as to allow for accurate and stable simulations.

Fluid Dynamics,Computational Physics
Large Eddy Simulation of Stratified Turbulent Channel Flow with a Dynamic Subgrid-Scale Model

Zhong Feng-quan,Liu Nan-sheng,Lu Xi-yun,Tong Bing-gang

DOI: https://doi.org/10.1007/s11741-000-0001-5

2000-01-01

Abstract:In this paper, large eddy simulation coupled with a dynamic subgrid-scale (SGS) model is used to study turbulent channel flows with heat transfer. Some fundamental flow behaviors are discussed. Based on our calculated results, the dynamic SGS model can reasonably predict some main characteristics of stratified turbulent channel flows. Our results are also in good agreement with theoretical analyses and previous calculated results.
Large-eddysimulation of startified channel flow

Liu Ningyu,Lu Xiyun,Wang Shanwu,Zhuang Lixiar

1997-01-01

Acta Mechanica Sinica/Lixue Xuebao

Abstract:An investigation of large-eddy sirnulation(LES) for turbulent channel flow with buoyancy effects was performed by solving the resolved incompressible Navier-Stokes equations under the Boussinesq approximation. The Smagorinsky eddy-viscosity model and Yoshizawa eddy-viscosity model were used to describe the unresolved subgrid scaie (SGS) fluctuations respectively. After some numerical testing, the latter was further simplified so that it can be used in the dynamic model closure. A LES code was developed for parallel computations by using the parallel technique, and was run on the Dawn-1000 parallel computer. To demonstrate the viability and accuracy of the code, our results are compared with and found in good agreement with available LES results.
Large eddy simulation of free surface turbulent channel flow with heat transfer

Fengquan Zhong,Nansheng Liu,Xiyun Lu,Lixian Zhuang

2002-01-01

Journal of Hydrodynamics

Abstract:The Large Eddy Simulation (LES) was used to study the free-surface turbulent channel flow with passive heat transfer. The three-dimensional filtered incompressible Navier-Stokes equations and energy equation were numerically solved with dynamic Subgrid Scale (SGS) models for modeling turbulent stresses and heat flux. To compare the turbulent behavior of the free-surface and two-walled channel flows, the LES of two-walled turbulent channel flow was performed. The statistical quantities and flow structures of the free-surface turbulence with heat transfer in the vicinity of the free-surface were investigated. The results are also in good agreement with theoretical analysis and available results by Direct Numerical Simulation (DNS).
Large Eddy Simulation of a Planar Jet Flow with Nanoparticle Coagulation

Mingzhou Yu,Jianzhong Lin,Lihua Chen,Tatleung Chan

DOI: https://doi.org/10.1007/s10409-006-0011-z

IF: 3.5

2006-01-01

Acta Mechanica Sinica

Abstract:Coagulation and growth of nanoparticles subject to large coherent structures in a planar jet has been explored by using large eddy simulation. The particle field is obtained by employing a moment method to approximate the nanoparticle general dynamic equation. An incompressible fluid containing particles of 1 nm in diameter is projected into a particle-free ambient. The results show that the coherent structures dominate the evolution of the nanoparticle number intensity diameter and polydispersity distributions as the jet develops. In addition, the coherent structures act to increase the diffusion of particles, and the vortex rolling-up makes the particles distributing more irregularly while the vortex pairing causes particle distributions to become uniform. As the jet travels downstream, the time-averaged particle number concentration becomes lower in the jet core and higher in the outskirts, whereas the time-averaged particle mass over the entire flow field maintains unaltered, and the time-averaged particle diameter and geometric standard deviations grow and reach their maximum on the interface of the jet region and the ambient.
Development of a large-eddy simulation subgrid model based on artificial neural networks: a case study of turbulent channel flow

Robin Stoffer,Caspar M. van Leeuwen,Damian Podareanu,Valeriu Codreanu,Menno A. Veerman,Martin Janssens,Oscar K. Hartogensis,Chiel C. van Heerwaarden

DOI: https://doi.org/10.5194/gmd-14-3769-2021

IF: 5.1

2021-06-24

Geoscientific Model Development

Abstract:Abstract. Atmospheric boundary layers and other wall-bounded flows are often simulated with the large-eddy simulation (LES) technique, which relies on subgrid-scale (SGS) models to parameterize the smallest scales. These SGS models often make strong simplifying assumptions. Also, they tend to interact with the discretization errors introduced by the popular LES approach where a staggered finite-volume grid acts as an implicit filter. We therefore developed an alternative LES SGS model based on artificial neural networks (ANNs) for the computational fluid dynamics MicroHH code (v2.0). We used a turbulent channel flow (with friction Reynolds number Reτ=590) as a test case. The developed SGS model has been designed to compensate for both the unresolved physics and instantaneous spatial discretization errors introduced by the staggered finite-volume grid. We trained the ANNs based on instantaneous flow fields from a direct numerical simulation (DNS) of the selected channel flow. In general, we found excellent agreement between the ANN-predicted SGS fluxes and the SGS fluxes derived from DNS for flow fields not used during training. In addition, we demonstrate that our ANN SGS model generalizes well towards other coarse horizontal resolutions, especially when these resolutions are located within the range of the training data. This shows that ANNs have potential to construct highly accurate SGS models that compensate for spatial discretization errors. We do highlight and discuss one important challenge still remaining before this potential can be successfully leveraged in actual LES simulations: we observed an artificial buildup of turbulence kinetic energy when we directly incorporated our ANN SGS model into a LES simulation of the selected channel flow, eventually resulting in numeric instability. We hypothesize that error accumulation and aliasing errors are both important contributors to the observed instability. We finally make several suggestions for future research that may alleviate the observed instability.

geosciences, multidisciplinary
Deconvolutional artificial neural network models for large eddy simulation of turbulence

Zelong Yuan,Chenyue Xie,Jianchun Wang

DOI: https://doi.org/10.1063/5.0027146

IF: 4.6

2020-11-01

Physics of Fluids

Abstract:Deconvolutional artificial neural network (DANN) models are developed for subgrid-scale (SGS) stress in large eddy simulation (LES) of turbulence. The filtered velocities at different spatial points are used as input features of the DANN models to reconstruct the unfiltered velocity. The grid width of the DANN models is chosen to be smaller than the filter width in order to accurately model the effects of SGS dynamics. The DANN models can predict the SGS stress more accurately than the conventional approximate deconvolution method and velocity gradient model in the <i>a priori</i> study: the correlation coefficients can be made larger than 99% and the relative errors can be made less than 15% for the DANN model. In an <i>a posteriori</i> study, a comprehensive comparison of the DANN model, the implicit LES (ILES), the dynamic Smagorinsky model (DSM), and the dynamic mixed model (DMM) shows that the DANN model is superior to the ILES, DSM, and DMM models in the prediction of the velocity spectrum, various statistics of velocity, and the instantaneous coherent structures without increasing the considerable computational cost; the time for the DANN model to calculate the SGS stress is about 1.3 times that of the DMM model. In addition, the trained DANN models without any fine-tuning can predict the velocity statistics well for different filter widths. These results indicate that the DANN framework with the consideration of SGS spatial features is a promising approach to develop advanced SGS models in the LES of turbulence.

mechanics,physics, fluids & plasmas
Large-eddy simulation of large-scale structures in long channel flow

D. Chung,B. J. McKeon

DOI: https://doi.org/10.1017/S0022112010002995

2010-01-19

Abstract:We investigate statistics of large-scale structures from large-eddy simulation (LES) of turbulent channel flow at friction Reynolds numbers $Re_\tau = 2 {\rm k}$ and $200 {\rm k}$. To properly capture the behaviour of large-scale structures, the channel length is chosen to be 96 times the channel half-height. In agreement with experiments, these large-scale structures are found to give rise to an apparent amplitude modulation of the underlying small-scale fluctuations. This effect is explained in terms of the phase relationship between the large- and small-scale activity. The shape of the dominant large-scale structure is investigated by conditional averages based on the large-scale velocity, determined using a filter width equal to the channel half-height. The conditioned field demonstrates coherence on a scale of several times the filter width, and the small-scale--large-scale relative phase difference increases away from the wall, passing through $\pi/2$ in the overlap region of the mean velocity before approaching $\pi$ further from the wall. We also found that, near the wall, the convection velocity of the large-scales departs slightly, but unequivocally, from the mean velocity.

Fluid Dynamics
Large-Eddy Simulation of Stratified Channel Flow

Liu Ningyu,Lu Xiyun,Wang Shanwu,Zhuang Lixian

DOI: https://doi.org/10.1007/bf02487192

1997-01-01

Abstract:An investigation of large-eddy simulation（LES） for turbulent chan-nel flow with buoyancy effects was performed by solving the resolved incompress-ible Navier-Stokes equations under the Boussinesq approximation. The Smagorinskyeddy-viscosity model and Yoshizawa eddy-viscosity model were used to describe theunresolved subgrid scale （SGS） fluctuations respectively. After some numerical test-ing, the latter was further simplified so that it can be used in the dynamic modelclosure. A LES code was developed for parallel computations by using the paralleltechnique, and was run on the Dawn-1000 parallel computer. To demonstrate theviability and accuracy of the code, our results are compared with and found in goodagreement with available LES results.
Modelling the flow around and wake behind net panels using large eddy simulations

Gang Wang,Tobias Martin,Liuyi Huang,Hans Bihs

DOI: https://doi.org/10.1016/j.oceaneng.2021.109846

IF: 5

2021-11-01

Ocean Engineering

Abstract:The fluid flow through a fixed net panel is investigated using large eddy simulations. Computational meshing strategies suitable for smooth and fabric twines are proposed. Numerical validations are conducted by comparing with measurements of the flow around a circular cylinder and hydrodynamics of net panels. It is shown that the hydrodynamic simulation of a small portion of the net panel is sufficient to replicate the full-scale net. Hence, the effect of the incoming velocity, the diameter and the length of the twines on surrounding turbulence fields are studied. Unique flow patterns are revealed in contrast to existing side-by-side cylinder simulations. Distributions of turbulent kinetic energies in the wake region behind intersections are crucial. It is further shown that the net solidity has a more considerable impact on the flow patterns in the wake than the Reynolds number. Further, the velocity reduction through the net panel is determined by weighting the relative velocity over the net panel area in the far-field. The analysis reveals less velocity reduction in comparison to existing velocity reduction models based on theoretical derivations and good agreement with physical measurements. The comparison of smooth and fabric nets indicates additional shielding effects of the twisted configurations on the velocity reduction which is not addressed in previous studies.

engineering, civil, ocean, marine,oceanography
Large-eddy Simulation of the Turbulent Structure in Compound Open-Channel Flows

Zhihua Xie,Binliang Lin,Roger A. Falconer

DOI: https://doi.org/10.1016/j.advwatres.2012.10.009

IF: 5.361

2012-01-01

Advances in Water Resources

Abstract:A large-eddy simulation study has been undertaken to investigate the turbulent structure of open-channel flow in an asymmetric compound channel. The dynamic sub-grid scale model has been employed in the model, with the partial cell treatment being implemented using a Cartesian grid structure to deal with the floodplain. The numerical model was used to predict the: primary velocity and secondary currents, boundary shear stress, turbulence intensities, turbulent kinetic energy, and Reynolds stresses. These parameters were compared with experimental measurements published in the literature, with relatively close agreement being obtained between both sets of results. Furthermore, instantaneous flow fields and large-scale vortical structures were predicted and are presented herein. These vortical structures were found to be responsible for the significant lateral exchange of mass and momentum in compound channels. Crown Copyright (C) 2012 Published by Elsevier Ltd. All rights reserved.

Toward neural-network-based large eddy simulation: application to turbulent channel flow

Constrained large-eddy simulation of laminar-turbulent transition in channel flow

Large Eddy Simulation of Spanwise Rotating Turbulent Channel Flow with Dynamic Variants of Eddy Viscosity Model

Constrained Large Eddy Simulation of Wall-Bounded Turbulent Flows

Discrete large eddy simulation

A recursive neural-network-based subgrid-scale model for large eddy simulation: application to homogeneous isotropic turbulence

Neural Ideal Large Eddy Simulation: Modeling Turbulence with Neural Stochastic Differential Equations

Wavelet-based adaptive implicit large-eddy simulation of turbulent channel flow

Large Eddy Simulation of Sediment-Laden Turbulent Flow in an Open Channel

Neural network-based closure models for large-eddy simulations with explicit filtering

Large Eddy Simulation of Stratified Turbulent Channel Flow with a Dynamic Subgrid-Scale Model

Large-eddysimulation of startified channel flow

Large eddy simulation of free surface turbulent channel flow with heat transfer

Large Eddy Simulation of a Planar Jet Flow with Nanoparticle Coagulation

Development of a large-eddy simulation subgrid model based on artificial neural networks: a case study of turbulent channel flow

Deconvolutional artificial neural network models for large eddy simulation of turbulence

Large-eddy simulation of large-scale structures in long channel flow

Large-Eddy Simulation of Stratified Channel Flow

Modelling the flow around and wake behind net panels using large eddy simulations

Large-eddy Simulation of the Turbulent Structure in Compound Open-Channel Flows