Unsupervised Machine Learning to Teach Fluid Dynamicists to Think in 15 Dimensions

An autoencoder is used to compress and then reconstruct three-dimensional stratified turbulence data in order to better understand fluid dynamics by studying the errors in the reconstruction. The original single data set is resolved on approximately $6.9\times10^{10}$ grid points, and 15 fluid variables in three spatial dimensions are used, for a total of about $10^{12}$ input quantities in three dimensions. The objective is to understand which of the input variables contains the most relevant information about the local turbulence regimes in stably stratified turbulence (SST). This is accomplished by observing flow features that appear in one input variable but then `bleed over' to multiple output variables. The bleed over is shown to be robust with respect to the number of layers in the autoencoder. In this proof of concept, the errors in the reconstruction include information about the spatial variation of vertical velocity in most of the components of the reconstructed rate-of-strain tensor and density gradient, which suggests that vertical velocity is an important marker for turbulence features of interest in SST. This result is consistent with what fluid dynamicists already understand about SST and, therefore, suggests an approach to understanding turbulence based on more detailed analyses of the reconstruction on errors in an autoencoding algorithm.

[1]  Pierre Baldi,et al.  Autoencoders, Unsupervised Learning, and Deep Architectures , 2011, ICML Unsupervised and Transfer Learning.

[2]  S. M. de Bruyn Kops,et al.  A mathematical framework for forcing turbulence applied to horizontally homogeneous stratified flow , 2011 .

[3]  E. Lindborg,et al.  The energy cascade in a strongly stratified fluid , 2006, Journal of Fluid Mechanics.

[4]  A. Mohan,et al.  Compressed Convolutional LSTM: An Efficient Deep Learning framework to Model High Fidelity 3D Turbulence , 2019, 1903.00033.

[5]  Koji Fukagata,et al.  Synthetic turbulent inflow generator using machine learning , 2018, Physical Review Fluids.

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

[7]  Vincent Dumoulin,et al.  Deconvolution and Checkerboard Artifacts , 2016 .

[8]  V. A. Krasil’nikov,et al.  Atmospheric turbulence and radio-wave propagation , 1962 .

[9]  Richard D. Sandberg,et al.  Turbulence Model Development using CFD-Driven Machine Learning , 2019 .

[10]  P. Davidson,et al.  Dynamics of stratified turbulence decaying from a high buoyancy Reynolds number , 2015, Journal of Fluid Mechanics.

[11]  J. Riley,et al.  The effects of stable stratification on the decay of initially isotropic homogeneous turbulence , 2018, Journal of Fluid Mechanics.

[12]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[13]  Douglas K. Lilly,et al.  Stratified Turbulence and the Mesoscale Variability of the Atmosphere , 1983 .

[14]  Charalambos Chrysostomou,et al.  Neural network-based modelling of unresolved stresses in a turbulent reacting flow with mean shear , 2019, 1904.08167.

[15]  S. Grossmann The Spectrum of Turbulence , 2003 .

[16]  S. D. B. Kops,et al.  Kinetic energy dynamics in forced, homogeneous, and axisymmetric stably stratified turbulence , 2012 .

[17]  J. Herring,et al.  Energy spectra of stably stratified turbulence , 2009, Journal of Fluid Mechanics.

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

[19]  Michael L. Waite Stratified turbulence at the buoyancy scale , 2011 .

[20]  J. Fourier Théorie analytique de la chaleur , 2009 .

[21]  Geoffrey E. Hinton,et al.  Deep Learning , 2015, Nature.

[22]  R. Mohan,et al.  Background Modeling Using Deep-Variational Autoencoder , 2018, ISDA.

[23]  Andrew Zisserman,et al.  Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps , 2013, ICLR.

[24]  Francisco Javier Gonzalez,et al.  Learning low-dimensional feature dynamics using convolutional recurrent autoencoders , 2018 .

[25]  Yoshua Bengio,et al.  Greedy Layer-Wise Training of Deep Networks , 2006, NIPS.

[26]  James J. Riley,et al.  Turbulent/non-turbulent interfaces in wakes in stably stratified fluids , 2016, Journal of Fluid Mechanics.

[27]  Noriyasu Omata,et al.  A novel method of low-dimensional representation for temporal behavior of flow fields using deep autoencoder , 2019, AIP Advances.

[28]  J. Riley,et al.  Dynamics of turbulence strongly influenced by buoyancy , 2003 .

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

[30]  Jean-Marc Chomaz,et al.  Self-similarity of strongly stratified inviscid flows , 2001 .

[31]  S. Pope,et al.  A deterministic forcing scheme for direct numerical simulations of turbulence , 1998 .

[32]  James J. Riley,et al.  Stratified Turbulence: A Possible Interpretation of Some Geophysical Turbulence Measurements , 2007 .

[33]  J. R. Taylor,et al.  Robust identification of dynamically distinct regions in stratified turbulence , 2016, Journal of Fluid Mechanics.

[34]  P. Bartello,et al.  Sensitivity of stratified turbulence to the buoyancy Reynolds number , 2013, Journal of Fluid Mechanics.

[35]  J. Chomaz,et al.  Scaling analysis and simulation of strongly stratified turbulent flows , 2007, Journal of Fluid Mechanics.

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

[37]  S. Corrsin,et al.  Free-Stream Boundaries of Turbulent Flows , 1955 .

[38]  Michael Chertkov,et al.  From Deep to Physics-Informed Learning of Turbulence: Diagnostics , 2018, ArXiv.

[39]  Claus-Dieter Munz,et al.  Deep Neural Networks for Data-Driven Turbulence Models , 2018, J. Comput. Phys..

[40]  Vladimir N. Vapnik,et al.  The Nature of Statistical Learning Theory , 2000, Statistics for Engineering and Information Science.

[41]  Atieh Alizadeh Moghaddam,et al.  A deep learning framework for turbulence modeling using data assimilation and feature extraction , 2018 .