Tracking Turbulent Coherent Structures by Means of Neural Networks

The behaviours of individual flow structures have become a relevant matter of study in turbulent flows as the computational power to allow their study feasible has become available. Especially, high instantaneous Reynolds Stress events have been found to dominate the behaviour of the logarithmic layer. In this work, we present a viability study where two machine learning solutions are proposed to reduce the computational cost of tracking such structures in large domains. The first one is a Multi-Layer Perceptron. The second one uses Long Short-Term Memory (LSTM). Both of the methods are developed with the objective of taking the the structures’ geometrical features as inputs from which to predict the structures’ geometrical features in future time steps. Some of the tested Multi-Layer Perceptron architectures proved to perform better and achieve higher accuracy than the LSTM architectures tested, providing lower errors on the predictions and achieving higher accuracy in relating the structures in the consecutive time steps.

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

[2]  Javier Jiménez,et al.  Time-resolved evolution of coherent structures in turbulent channels: characterization of eddies and cascades , 2012, Journal of Fluid Mechanics.

[3]  Kwan-Liu Ma,et al.  Interactive feature extraction and tracking by utilizing region coherency , 2009, 2009 IEEE Pacific Visualization Symposium.

[4]  R. Moser,et al.  Spectral methods for the Navier-Stokes equations with one infinite and two periodic directions , 1991 .

[5]  V. Avsarkisov,et al.  Turbulent plane Couette flow at moderately high Reynolds number , 2014, Journal of Fluid Mechanics.

[6]  James M. Wallace,et al.  The wall region in turbulent shear flow , 1972, Journal of Fluid Mechanics.

[7]  M. R. Head,et al.  New aspects of turbulent boundary-layer structure , 1981, Journal of Fluid Mechanics.

[8]  P. Moin,et al.  Turbulence statistics in fully developed channel flow at low Reynolds number , 1987, Journal of Fluid Mechanics.

[9]  Javier Jiménez,et al.  The three-dimensional structure of momentum transfer in turbulent channels , 2011, Journal of Fluid Mechanics.

[10]  Jonghwan Park,et al.  Machine-learning-based feedback control for drag reduction in a turbulent channel flow , 2020, Journal of Fluid Mechanics.

[11]  George Cybenko,et al.  Approximation by superpositions of a sigmoidal function , 1989, Math. Control. Signals Syst..

[12]  M. S. Chong,et al.  A general classification of three-dimensional flow fields , 1990 .

[13]  S. Hoyas,et al.  DNS of thermal channel flow up to Reτ=2000 for medium to low Prandtl numbers , 2018, International Journal of Heat and Mass Transfer.

[14]  Sergio Hoyas,et al.  Following Vortices in Turbulent Channel Flows , 2020, SOCO.

[15]  Hui Li,et al.  A Novel Algebraic Stress Model with Machine-Learning-Assisted Parameterization , 2020, Energies.

[16]  Seyed Amin Bagherzadeh,et al.  Develop 24 dissimilar ANNs by suitable architectures & training algorithms via sensitivity analysis to better statistical presentation: Measure MSEs between targets & ANN for Fe–CuO/Eg–Water nanofluid , 2019, Physica A: Statistical Mechanics and its Applications.

[17]  Javier Jiménez,et al.  A high-resolution code for turbulent boundary layers , 2009, J. Comput. Phys..

[18]  M. Tokarev,et al.  Deep Reinforcement Learning Control of Cylinder Flow Using Rotary Oscillations at Low Reynolds Number , 2020, Energies.

[19]  H. Alsulami,et al.  Influence of base fluid, temperature, and concentration on the thermophysical properties of hybrid nanofluids of alumina–ferrofluid: experimental data, modeling through enhanced ANN, ANFIS, and curve fitting , 2020, Journal of Thermal Analysis and Calorimetry.

[20]  J. Jiménez,et al.  Effect of the computational domain on direct simulations of turbulent channels up to Reτ = 4200 , 2014 .

[21]  Javier Jiménez,et al.  Near-wall turbulence , 2013 .

[22]  Javier Jiménez,et al.  Scaling of the velocity fluctuations in turbulent channels up to Reτ=2003 , 2006 .

[23]  S. Hoyas,et al.  DNS of a turbulent Couette flow at constant wall transpiration up to $Re_{\unicode[STIX]{x1D70F}}=1000$ , 2017, Journal of Fluid Mechanics.

[24]  Hossein Azizpour,et al.  Predictions of turbulent shear flows using deep neural networks , 2019, Physical Review Fluids.

[25]  Jürgen Schmidhuber,et al.  Long Short-Term Memory , 1997, Neural Computation.

[26]  S. Hoyas,et al.  Letter: The link between the Reynolds shear stress and the large structures of turbulent Couette-Poiseuille flow , 2018 .

[27]  Petros Koumoutsakos,et al.  Machine Learning for Fluid Mechanics , 2019, Annual Review of Fluid Mechanics.

[28]  Carl D. Meinhart,et al.  Vortex organization in the outer region of the turbulent boundary layer , 2000, Journal of Fluid Mechanics.

[29]  S. Lele Compact finite difference schemes with spectral-like resolution , 1992 .

[30]  S. Hoyas,et al.  Influence of the computational domain on DNS of turbulent heat transfer up to Reτ=2000 for Pr=0.71 , 2018, International Journal of Heat and Mass Transfer.

[31]  Jürgen Schmidhuber,et al.  Learning to Forget: Continual Prediction with LSTM , 2000, Neural Computation.

[32]  J. Jiménez Coherent structures in wall-bounded turbulence , 2017, Journal of Fluid Mechanics.

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

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

[35]  Mostafa Safdari Shadloo,et al.  Estimation of Pressure Drop of Two-Phase Flow in Horizontal Long Pipes Using Artificial Neural Networks , 2020 .

[36]  Jürgen Schmidhuber,et al.  Learning Precise Timing with LSTM Recurrent Networks , 2003, J. Mach. Learn. Res..

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

[38]  Javier Jiménez,et al.  Self-similar vortex clusters in the turbulent logarithmic region , 2006, Journal of Fluid Mechanics.

[39]  Javier Jiménez,et al.  Geometry and clustering of intense structures in isotropic turbulence , 2004, Journal of Fluid Mechanics.