Data-driven deconvolution for large eddy simulations of Kraichnan turbulence

In this article, we demonstrate the use of artificial neural networks as optimal maps which are utilized for convolution and deconvolution of coarse-grained fields to account for sub-grid scale turbulence effects. We demonstrate that an effective eddy-viscosity is predicted by our purely data-driven large eddy simulation framework without explicit utilization of phenomenological arguments. In addition, our data-driven framework precludes the knowledge of true sub-grid stress information during the training phase due to its focus on estimating an effective filter and its inverse so that grid-resolved variables may be related to direct numerical simulation data statistically. The proposed predictive framework is also combined with a statistical truncation mechanism for ensuring numerical realizability in an explicit formulation. Through this, we seek to unite structural and functional modeling strategies for modeling non-linear partial differential equations using reduced degrees of freedom. Both a priori and a posteriori results are shown for a two-dimensional decaying turbulence case in addition to a detailed description of validation and testing. A hyperparameter sensitivity study also shows that the proposed dual network framework simplifies learning complexity and is viable with exceedingly simple network architectures. Our findings indicate that the proposed framework approximates a robust and stable sub-grid closure which compares favorably to the Smagorinsky and Leith hypotheses for capturing the theoretical k−3 scaling in Kraichnan turbulence.In this article, we demonstrate the use of artificial neural networks as optimal maps which are utilized for convolution and deconvolution of coarse-grained fields to account for sub-grid scale turbulence effects. We demonstrate that an effective eddy-viscosity is predicted by our purely data-driven large eddy simulation framework without explicit utilization of phenomenological arguments. In addition, our data-driven framework precludes the knowledge of true sub-grid stress information during the training phase due to its focus on estimating an effective filter and its inverse so that grid-resolved variables may be related to direct numerical simulation data statistically. The proposed predictive framework is also combined with a statistical truncation mechanism for ensuring numerical realizability in an explicit formulation. Through this, we seek to unite structural and functional modeling strategies for modeling non-linear partial differential equations using reduced degrees of freedom. Both a priori a...

[1]  Petros Koumoutsakos,et al.  Data-assisted reduced-order modeling of extreme events in complex dynamical systems , 2018, PloS one.

[2]  Paris Perdikaris,et al.  Machine learning of linear differential equations using Gaussian processes , 2017, J. Comput. Phys..

[3]  Steven L. Brunton,et al.  Deep learning of dynamics and signal-noise decomposition with time-stepping constraints , 2018, J. Comput. Phys..

[4]  Karthik Duraisamy,et al.  Turbulence Modeling in the Age of Data , 2018, Annual Review of Fluid Mechanics.

[5]  Jean-Luc Aider,et al.  Closed-loop separation control using machine learning , 2014, Journal of Fluid Mechanics.

[6]  Hod Lipson,et al.  Distilling Free-Form Natural Laws from Experimental Data , 2009, Science.

[7]  Yuji Hattori,et al.  Searching for turbulence models by artificial neural network , 2016, 1607.01042.

[8]  J. Ferziger,et al.  Improved subgrid-scale models for large-eddy simulation , 1980 .

[9]  V. Canuto,et al.  Determination of the Smagorinsky–Lilly constant CS , 1997 .

[10]  D. Whiteson,et al.  Deep Learning and Its Application to LHC Physics , 2018, Annual Review of Nuclear and Particle Science.

[11]  F. Sarghini,et al.  Neural networks based subgrid scale modeling in large eddy simulations , 2003 .

[12]  Richard D. Sandberg,et al.  Hybrid Reynolds-Averaged/Large-Eddy Simulation Methodology from Symbolic Regression: Formulation and Application , 2017 .

[13]  P. Moin,et al.  A dynamic subgrid‐scale eddy viscosity model , 1990 .

[14]  J. Smagorinsky,et al.  GENERAL CIRCULATION EXPERIMENTS WITH THE PRIMITIVE EQUATIONS , 1963 .

[15]  J. Templeton,et al.  Reynolds averaged turbulence modelling using deep neural networks with embedded invariance , 2016, Journal of Fluid Mechanics.

[16]  Omer San,et al.  A neural network approach for the blind deconvolution of turbulent flows , 2017, Journal of Fluid Mechanics.

[17]  Mahdi Saeedipour,et al.  Approximate deconvolution model for the simulation of turbulent gas-solid flows: An a priori analysis , 2018 .

[18]  Michele Milano,et al.  Neural network modeling for near wall turbulent flow , 2002 .

[19]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[20]  M. Oberlack,et al.  Analysis and development of subgrid turbulence models preserving the symmetry properties of the Navier–Stokes equations , 2007 .

[21]  K. Taira,et al.  Super-resolution reconstruction of turbulent flows with machine learning , 2018, Journal of Fluid Mechanics.

[22]  J. Kutz,et al.  Compressive Sensing Based Machine Learning Strategy For Characterizing The Flow Around A Cylinder With Limited Pressure Measurements , 2013 .

[23]  O. San,et al.  High-order methods for decaying two-dimensional homogeneous isotropic turbulence , 2012, 1212.0920.

[24]  Ryan N King,et al.  Autonomic closure for turbulence simulations. , 2016, Physical review. E.

[25]  C. Pain,et al.  Non‐intrusive reduced‐order modelling of the Navier–Stokes equations based on RBF interpolation , 2015 .

[26]  N. Adams,et al.  An approximate deconvolution procedure for large-eddy simulation , 1999 .

[27]  Jinlong Wu,et al.  Physics-informed machine learning approach for reconstructing Reynolds stress modeling discrepancies based on DNS data , 2016, 1606.07987.

[28]  Frank O. Bryan,et al.  Evaluation of scale-aware subgrid mesoscale eddy models in a global eddy-rich model , 2017 .

[29]  K. Duraisamy,et al.  Using field inversion to quantify functional errors in turbulence closures , 2016 .

[30]  J. Nathan Kutz,et al.  Deep learning in fluid dynamics , 2017, Journal of Fluid Mechanics.

[31]  S. Brunton,et al.  Discovering governing equations from data by sparse identification of nonlinear dynamical systems , 2015, Proceedings of the National Academy of Sciences.

[32]  William J. Layton,et al.  A simple and stable scale-similarity model for large Eddy simulation: Energy balance and existence of weak solutions , 2003, Appl. Math. Lett..

[33]  Prakash Vedula,et al.  Subgrid modelling for two-dimensional turbulence using neural networks , 2018, Journal of Fluid Mechanics.

[34]  I. Oseledets,et al.  Application of machine learning to viscoplastic flow modeling , 2018, Physics of Fluids.

[35]  Michelle Girvan,et al.  Hybrid Forecasting of Chaotic Processes: Using Machine Learning in Conjunction with a Knowledge-Based Model , 2018, Chaos.

[36]  Baylor Fox-Kemper,et al.  Log-Normal Turbulence Dissipation in Global Ocean Models. , 2018, Physical review letters.

[37]  R. Kraichnan.  Inertial Ranges in Two‐Dimensional Turbulence , 1967 .

[38]  R. Moser,et al.  Theoretically based optimal large-eddy simulation , 2009 .

[39]  Jean-Luc Guermond,et al.  Mathematical Perspectives on Large Eddy Simulation Models for Turbulent Flows , 2004 .

[40]  A. Vorobev,et al.  Smagorinsky constant in LES modeling of anisotropic MHD turbulence , 2008 .

[41]  C. Leith.  Diffusion Approximation for Two‐Dimensional Turbulence , 1968 .

[42]  Hui Li,et al.  Prediction model of velocity field around circular cylinder over various Reynolds numbers by fusion convolutional neural networks based on pressure on the cylinder , 2018 .

[43]  Richard D. Sandberg,et al.  The development of algebraic stress models using a novel evolutionary algorithm , 2017 .

[44]  Prakash Vedula,et al.  A framework for large eddy simulation of Burgers turbulence based upon spatial and temporal statistical information , 2015 .

[45]  Stefan Hickel,et al.  Subgrid-scale modeling for implicit large eddy simulation of compressible flows and shock-turbulence interaction , 2014 .

[46]  Richard Sandberg,et al.  A novel evolutionary algorithm applied to algebraic modifications of the RANS stress-strain relationship , 2016, J. Comput. Phys..

[47]  R. Moser,et al.  Optimal LES formulations for isotropic turbulence , 1999, Journal of Fluid Mechanics.

[48]  David A. Landgrebe,et al.  A survey of decision tree classifier methodology , 1991, IEEE Trans. Syst. Man Cybern..

[49]  Claus-Dieter Munz,et al.  Neural Networks for Data-Based Turbulence Models , 2018 .

[50]  Omer San,et al.  Neural network closures for nonlinear model order reduction , 2017, Adv. Comput. Math..

[51]  Heng Xiao,et al.  Quantifying and reducing model-form uncertainties in Reynolds-averaged Navier-Stokes simulations: A data-driven, physics-informed Bayesian approach , 2015, J. Comput. Phys..

[52]  R. Maulik,et al.  A stable and scale-aware dynamic modeling framework for subgrid-scale parameterizations of two-dimensional turbulence , 2017 .

[53]  Karthik Duraisamy,et al.  Machine Learning-augmented Predictive Modeling of Turbulent Separated Flows over Airfoils , 2016, ArXiv.

[54]  G. Tryggvason,et al.  Using statistical learning to close two-fluid multiphase flow equations for a simple bubbly system , 2015 .

[55]  Christophe Eric Corre,et al.  Subgrid-scale scalar flux modelling based on optimal estimation theory and machine-learning procedures , 2017 .

[56]  M. Germano.  The similarity subgrid stresses associated to the approximate Van Cittert deconvolutions , 2015 .

[57]  Heung-Il Suk,et al.  Deep Learning in Medical Image Analysis. , 2017, Annual review of biomedical engineering.

[58]  J. Templeton.  Evaluation of machine learning algorithms for prediction of regions of high Reynolds averaged Navier Stokes uncertainty , 2015 .

[59]  Zuoli Xiao,et al.  Dynamic optimization methodology based on subgrid-scale dissipation for large eddy simulation , 2016 .

[60]  Traian Iliescu,et al.  A posteriori analysis of low-pass spatial filters for approximate deconvolution large eddy simulations of homogeneous incompressible flows , 2014, 1401.6217.

[61]  George E. Karniadakis,et al.  Hidden physics models: Machine learning of nonlinear partial differential equations , 2017, J. Comput. Phys..

[62]  H. Schaeffer,et al.  Learning partial differential equations via data discovery and sparse optimization , 2017, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences.