Skip to main content Accessibility help

Reynolds averaged turbulence modelling using deep neural networks with embedded invariance

  • Julia Ling (a1), Andrew Kurzawski (a2) and Jeremy Templeton (a1)

There exists significant demand for improved Reynolds-averaged Navier–Stokes (RANS) turbulence models that are informed by and can represent a richer set of turbulence physics. This paper presents a method of using deep neural networks to learn a model for the Reynolds stress anisotropy tensor from high-fidelity simulation data. A novel neural network architecture is proposed which uses a multiplicative layer with an invariant tensor basis to embed Galilean invariance into the predicted anisotropy tensor. It is demonstrated that this neural network architecture provides improved prediction accuracy compared with a generic neural network architecture that does not embed this invariance property. The Reynolds stress anisotropy predictions of this invariant neural network are propagated through to the velocity field for two test cases. For both test cases, significant improvement versus baseline RANS linear eddy viscosity and nonlinear eddy viscosity models is demonstrated.

Corresponding author
Email address for correspondence:
Hide All
Banerjee, S., Kraal, R., Durst, F. & Zenger, C. H. 2007 Presentation of anisotropy properties of turbulence invariants versus eigenvalue approaches. J. Turbul. 8, 127.
Craft, T. J., Launder, B. E. & Suga, K. 1996 Development and application of a cubic eddy-viscosity model of turbulence. Intl J. Heat Fluid Flow 17, 108115.
Silver, D. et al. 2016 Mastering the game of Go with deep neural networks and tree search. Nature 529, 484489.
Domino, S. P., Moen, C. D., Burns, S. P. & Evans, G. H.2003 SIERRA/Fuego: a multi-mechanics fire environment simulation tool. AIAA Paper 2003-149.
Hinton, G. et al. 2012 Deep neural networks for acoustic modeling in speech recognition: the shared views of four research groups. IEEE Signal Process. Mag. 29, 8297.
Gatski, T. B. & Speziale, C. G. 1993 On explicit algebraic stress models for complex turbulent flows. J. Fluid Mech. 254, 5978.
Karpathy, A., Toderici, G., Shetty, S., Leung, T., Sukthankar, R. & Li, F. F. 2014 Large-scale video classification with convolutional neural networks. In Proc. IEEE Conf. on Computer Vision and Pattern Recognition, pp. 17251732. IEEE.
Lecun, Y., Bengio, Y. & Hinton, G. 2015 Deep learning. Nature 521, 436444.
Ling, J., Jones, R. & Templeton, J. 2016a Machine learning strategies for systems with invariance properties. J. Comput. Phys. 318, 2235.
Ling, J., Ruiz, A., Lacaze, G. & Oefelein, J. 2016b Uncertainty analysis and data-driven model advances for a jet-in-crossflow. In ASME Turbo Expo 2016, ASME.
Ling, J., Ryan, K. J., Bodart, J. & Eaton, J. K. 2016c Analysis of turbulent scalar flux models for a discrete hole film cooling flow. J. Turbomach. 138, 011006.
Ling, J. & Templeton, J. A. 2015 Evaluation of machine learning algorithms for prediction of regions of high RANS uncertainty. Phys. Fluids 27, 085103.
Maas, A., Hannun, A. & Ng, A. 2013 Rectifier nonlinearities improve neural network acoustic models. Proc. ICML 30, 16.
Marquillie, M., Ehrenstein, U. & Laval, J. P. 2011 Instability of streaks in wall turbulence with adverse pressure gradient. J. Fluid Mech. 681, 205240.
Milano, M. & Koumoutsakos, P. 2002 Neural network modeling for near wall turbulent flow. J. Comput. Phys. 182, 126.
Moser, R. D., Kim, J. & Mansour, N. N. 1999 Direct numerical simulation of turbulent channel flow up to Re 𝜏 = 590. Phys. Fluids 11, 943945.
Parish, E. & Duraisamy, K. 2016 A paradigm for data-driven predictive modeling using field inversion and machine learning. J. Comput. Phys. 305, 758774.
Pinelli, A., Uhlmann, M., Sekimoto, A. & Kawahara, G. 2010 Reynolds number dependence of mean flow structure in square duct turbulence. J. Fluid Mech. 644, 107122.
Pope, S. B. 1975 A more general effective-viscosity hypothesis. J. Fluid Mech. 72, 331340.
Ray, J., Lefantzi, S., Arunajatesan, S. & Dechant, L.2014 Bayesian calibration of a $k{-}\unicode[STIX]{x1D716}$ turbulence model for predictive jet-in-crossflow simulations. AIAA Paper 2014-2085.
Rossi, R., Philips, D. & Iaccarino, G. 2010 A numerical study of scalar dispersion downstream of a wall-mounted cube using direct simulations and algebraic flux models. Intl J. Heat Fluid Flow 31, 805819.
Ruiz, A. M., Oefelein, J. C. & Lacaze, G. 2015 Flow topologies and turbulence scales in a jet-in-cross-flow. Phys. Fluids 27, 045101.
Smith, G. F. 1965 On isotropic integrity bases. Arch. Rational Mech. Anal. 18, 282292.
Snoek, J., Larochelle, H. & Adams, R. P. 2012 Practical Bayesian optimization of machine learning algorithms. Adv. Neural Inform. Proc. Syst. 25, 29512959.
Tracey, B., Duraisamy, K. & Alonso, J. J.2013 Application of supervised learning to quantify uncertainties in turbulence and combustion modeling. AIAA Aerospace Sciences Meeting 2013-0259.
Tracey, B., Duraisamy, K. & Alonso, J. J.2015 A machine learning strategy to assist turbulence model development. AIAA Aerospace Sciences Meeting 2015-1287.
Wallin, S. & Johansson, A. 2000 An explicit algebraic Reynolds stress model for incompressible and compressible turbulent flows. J. Fluid Mech. 403, 89132.
Zhang, Z. J. & Duraisamy, K.2015 Machine learning methods for data-driven turbulence modeling. AIAA Computational Fluid Dynamics Conf. 2015-2460.
Recommend this journal

Email your librarian or administrator to recommend adding this journal to your organisation's collection.

Journal of Fluid Mechanics
  • ISSN: 0022-1120
  • EISSN: 1469-7645
  • URL: /core/journals/journal-of-fluid-mechanics
Please enter your name
Please enter a valid email address
Who would you like to send this to? *

JFM classification


Full text views

Total number of HTML views: 0
Total number of PDF views: 0 *
Loading metrics...

Abstract views

Total abstract views: 0 *
Loading metrics...

* Views captured on Cambridge Core between <date>. This data will be updated every 24 hours.

Usage data cannot currently be displayed