×

Unsupervised deep learning for super-resolution reconstruction of turbulence. (English) Zbl 1461.76308

Summary: Recent attempts to use deep learning for super-resolution reconstruction of turbulent flows have used supervised learning, which requires paired data for training. This limitation hinders more practical applications of super-resolution reconstruction. Therefore, we present an unsupervised learning model that adopts a cycle-consistent generative adversarial network (CycleGAN) that can be trained with unpaired turbulence data for super-resolution reconstruction. Our model is validated using three examples: (i) recovering the original flow field from filtered data using direct numerical simulation (DNS) of homogeneous isotropic turbulence; (ii) reconstructing full-resolution fields using partially measured data from the DNS of turbulent channel flows; and (iii) generating a DNS-resolution flow field from large-eddy simulation (LES) data for turbulent channel flows. In examples (i) and (ii), for which paired data are available for supervised learning, our unsupervised model demonstrates qualitatively and quantitatively similar performance as that of the best supervised learning model. More importantly, in example (iii), where supervised learning is impossible, our model successfully reconstructs the high-resolution flow field of statistical DNS quality from the LES data. Furthermore, we find that the present model has almost universal applicability to all values of Reynolds numbers within the tested range. This demonstrates that unsupervised learning of turbulence data is indeed possible, opening a new door for the wide application of super-resolution reconstruction of turbulent fields.

MSC:

76F65 Direct numerical and large eddy simulation of turbulence
PDFBibTeX XMLCite
Full Text: DOI arXiv

References:

[1] Abadi, M., et al.. 2015 Tensorflow: large-scale machine learning on heterogeneous systems. Software available from tensorflow.org.
[2] Beck, A., Flad, D. & Munz, C.-D.2019Deep neural networks for data-driven LES closure models. J. Comput. Phys.398, 108910.
[3] Berkooz, G., Holmes, P. & Lumley, J.L.1993The proper orthogonal decomposition in the analysis of turbulent flows. Annu. Rev. Fluid Mech.25 (1), 539-575.
[4] Brenner, M.P., Eldredge, J.D. & Freund, J.B.2019Perspective on machine learning for advancing fluid mechanics. Phys. Rev. Fluids4 (10), 100501.
[5] Brunton, S.L., Noack, B.R. & Koumoutsakos, P.2020Machine learning for fluid mechanics. Annu. Rev. Fluid Mech.52 (1), 477-508. · Zbl 1439.76138
[6] Buzzicotti, M., Bonaccorso, F., Clark Di Leoni, P. & Biferale, L.2020 Reconstruction of turbulent data with deep generative models for semantic inpainting from TURB-Rot database. arXiv:2006.09179v1.
[7] Cai, S., Liang, J., Gao, Q., Xu, C. & Wei, R.2020Particle image velocimetry based on a deep learning motion estimator. IEEE Trans. Instrum. Meas.69 (6), 3538-3554.
[8] Deng, Z., He, C., Liu, Y. & Kim, K.2019Super-resolution reconstruction of turbulent velocity fields using a generative adversarial network-based artificial intelligence framework. Phys. Fluids31 (12), 125111.
[9] Duraisamy, K., Iaccarino, G. & Xiao, H.2019Turbulence modeling in the age of data. Annu. Rev. Fluid Mech.51 (1), 357-377. · Zbl 1412.76040
[10] Fukami, K., Fukagata, K. & Taira, K.2019aSuper-resolution reconstruction of turbulent flows with machine learning. J. Fluid Mech.870, 106-120.
[11] Fukami, K., Fukagata, K. & Taira, K.2020aAssessment of supervised machine learning methods for fluid flows. Theor. Comput. Fluid Dyn.34 (4), 497-519.
[12] Fukami, K., Fukagata, K. & Taira, K.2020b Machine learning based spatio-temporal super resolution reconstruction of turbulent flows. arXiv:2004.11566v1.
[13] Fukami, K., Nabae, Y., Kawai, K. & Fukagata, K.2019bSynthetic turbulent inflow generator using machine learning. Phys. Rev. Fluids4 (6), 064603.
[14] Gamahara, M. & Hattori, Y.2017Searching for turbulence models by artificial neural network. Phys. Rev. Fluids2 (5), 054604.
[15] Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A. & Bengio, Y.2014 Generative adversarial nets. In NIPS, pp. 2672-2680.
[16] Graham, J., et al.. 2015A web services accessible database of turbulent channel flow and its use for testing a new integral wall model for LES. J. Turbul.17 (2), 181-215.
[17] Guastoni, L., Encinar, M.P., Schlatter, P., Azizpour, H. & Vinuesa, R.2020Prediction of wall-bounded turbulence from wall quantities using convolutional neural networks. J. Phys.: Conf. Ser.1522, 012022.
[18] Gulrajani, I., Ahmed, F., Arjovsky, M., Dumoulin, V. & Courville, A.2017 Improved training of Wasserstein GANs. In NIPS, pp. 5767-5777.
[19] He, K., Zhang, X., Ren, S. & Sun, J.2015a Deep residual learning for image recognition. arXiv:1512.03385v1.
[20] He, K., Zhang, X., Ren, S. & Sun, J.2015b Delving deep into rectifiers: surpassing human-level performance on ImageNet classification. In ICCV. IEEE.
[21] Hornik, K., Stinchcombe, M. & White, H.1989Multilayer feedforward networks are universal approximators. Neural Networks2 (5), 359-366. · Zbl 1383.92015
[22] Ioffe, S. & Szegedy, C.2015 Batch normalization: accelerating deep network training by reducing internal covariate shift. arXiv:1502.03167.
[23] Jiang, C.M., Esmaeilzadeh, S., Azizzadenesheli, K., Kashinath, K., Mustafa, M., Tchelepi, H.A., Marcus, P., Prabhat, & Anandkumar, A.2020 MeshfreeFlowNet: a physics-constrained deep continuous space-time super-resolution framework. arXiv:2005.01463v2.
[24] Karras, T., Laine, S. & Aila, T.2018 A style-based generator architecture for generative adversarial networks. arXiv:1812.04948v3.
[25] Kim, J. & Lee, C.2020aDeep unsupervised learning of turbulence for inflow generation at various Reynolds numbers. J. Comput. Phys.406, 109216.
[26] Kim, J. & Lee, C.2020bPrediction of turbulent heat transfer using convolutional neural networks. J. Fluid Mech.882, A18. · Zbl 1430.76246
[27] Kim, J., Moin, P. & Moser, R.1987Turbulence statistics in fully developed channel flow at low Reynolds number. J. Fluid Mech.177, 133-166. · Zbl 0616.76071
[28] Kingma, D.P. & Ba, J.L.2014 Adam: a method for stochastic optimization. arXiv:1412.6980.
[29] Kipf, T.N. & Welling, Max2016 Semi-supervised classification with graph convolutional networks. arXiv:1609.02907v4.
[30] Kutz, J.N.2017Deep learning in fluid dynamics. J. Fluid Mech.814, 1-4. · Zbl 1383.76380
[31] Lecun, Y., Bengio, Y. & Hinton, G2015Deep learning. Nature521 (7553), 436-444.
[32] Lee, C., Kim, J., Babcock, D. & Goodman, R.1997Application of neural networks to turbulence control for drag reduction. Phys. Fluids9 (6), 1740-1747.
[33] Lee, M. & Moser, R.D.2015Direct numerical simulation of turbulent channel flow up to \({R}e_\tau =5200\). J. Fluid Mech.774, 395-415.
[34] Lee, S. & You, D.2019Data-driven prediction of unsteady flow over a circular cylinder using deep learning. J. Fluid Mech.879, 217-254. · Zbl 1430.76311
[35] Leoni, P.C.Di, Mazzino, A. & Biferale, L.2020Synchronization to big data: nudging the Navier-Stokes equations for data assimilation of turbulent flows. Phys. Rev. X10 (1), 011023.
[36] Li, Y., Perlman, E., Wan, M., Yang, Y., Meneveau, C., Burns, R., Chen, S., Szalay, A. & Eyink, G.2008A public turbulence database cluster and applications to study Lagrangian evolution of velocity increments in turbulence. J. Turbul.9, N31. · Zbl 1273.76210
[37] Ling, J., Kurzawski, A. & Templeton, J.2016Reynolds averaged turbulence modelling using deep neural networks with embedded invariance. J. Fluid Mech.807, 155-166. · Zbl 1383.76175
[38] Liu, B., Tang, J., Huang, H. & Lu, X.-Y.2020Deep learning methods for super-resolution reconstruction of turbulent flows. Phys. Fluids32 (2), 025105.
[39] Maulik, R. & San, O.2017A neural network approach for the blind deconvolution of turbulent flows. J. Fluid Mech.831, 151-181. · Zbl 1421.76134
[40] Maulik, R., San, O., Rasheed, A. & Vedula, P.2019Subgrid modelling for two-dimensional turbulence using neural networks. J. Fluid Mech.858, 122-144. · Zbl 1415.76405
[41] Milano, M. & Koumoutsakos, P.2002Neural network modeling for near wall turbulent flow. J. Comput. Phys.182 (1), 1-26. · Zbl 1090.76535
[42] Mirza, M. & Osindero, S.2014 Conditional generative adversarial nets. arXiv:1411.1784v1.
[43] Mohan, A.T., Lubbers, N., Livescu, D. & Chertkov, M.2020 Embedding hard physical constraints in neural network coarse-graining of 3d turbulence. arXiv:2002.00021v2.
[44] Pandey, S., Schumacher, J. & Sreenivasan, K.R.2020A perspective on machine learning in turbulent flows. J. Turbul.21, 567-584.
[45] Parish, E.J. & Duraisamy, K.2016A paradigm for data-driven predictive modeling using field inversion and machine learning. J. Comput. Phys.305, 758-774. · Zbl 1349.76006
[46] Perlman, E., Burns, R., Li, Y. & Meneveau, C.2007 Data exploration of turbulence simulations using a database cluster. In Proceedings of the 2007 ACM/IEEE Conference on Supercomputing. ACM.
[47] Rabault, J., Kolaas, J. & Jensen, A.2017Performing particle image velocimetry using artificial neural networks: a proof-of-concept. Meas. Sci. Technol.28 (12), 125301.
[48] Rabault, J., Kuchta, M., Jensen, A., Réglade, U. & Cerardi, N.2019Artificial neural networks trained through deep reinforcement learning discover control strategies for active flow control. J. Fluid Mech.865, 281-302. · Zbl 1415.76222
[49] Rai, M.M. & Moin, P1991Direct simulations of turbulent flow using finite-difference schemes. J. Comput. Phys.96 (1), 15-53. · Zbl 0726.76072
[50] Raissi, M., Perdikaris, P. & Karniadakis, G.E.2019Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Comput. Phys.378, 686-707. · Zbl 1415.68175
[51] Raissi, M., Yazdani, A. & Karniadakis, G.E.2020Hidden fluid mechanics: learning velocity and pressure fields from flow visualizations. Science367 (6481), 1026-1030. · Zbl 1478.76057
[52] Scherl, I., Strom, B., Shang, J.K., Williams, O., Polagye, B.L. & Brunton, S.L.2020Robust principal component analysis for modal decomposition of corrupt fluid flows. Phys. Rev. Fluids5 (5), 054401.
[53] Smagorinsky, J.1963General circulation experiments with primitive equations. Mon. Weath. Rev.91 (3), 99-164.
[54] Srinivasan, P.A., Guastoni, L., Azizpour, H., Schlatter, P. & Vinuesa, R.2019Predictions of turbulent shear flows using deep neural networks. Phys. Rev. Fluids4 (5), 054603.
[55] Verma, S., Novati, G. & Koumoutsakos, P.2018Efficient collective swimming by harnessing vortices through deep reinforcement learning. Proc. Natl Acad. Sci. USA115 (23), 5849-5854.
[56] Vreman, A.W.2004An eddy-viscosity subgrid-scale model for turbulent shear flow: algebraic theory and applications. Phys. Fluids16 (10), 3670-3681. · Zbl 1187.76543
[57] Wang, J.-X., Wu, J.-L. & Xiao, H.2017Physics-informed machine learning approach for reconstructing Reynolds stress modeling discrepancies based on DNS data. Phys. Rev. Fluids2 (3), 034603.
[58] Werhahn, M., Xie, Y., Chu, M. & Thuerey, N.2019A multi-pass gan for fluid flow super-resolution. Proc. ACM Comput. Graph. Interact. Tech.2 (2).
[59] Xie, C., Wang, J., Li, K. & Ma, C.2019Artificial neural network approach to large-eddy simulation of compressible isotropic turbulence. Phys. Rev. E99 (5), 053113.
[60] Xie, Y., Franz, E., Chu, M. & Thuerey, N.2018TempoGAN: a temporally coherent, volumetric GAN for super-resolution fluid flow. ACM Trans. Graph.37 (4).
[61] Yeung, P.K., Donzis, D.A. & Sreenivasan, K.R.2012Dissipation, enstrophy and pressure statistics in turbulence simulations at high Reynolds numbers. J. Fluid Mech.700, 5-15. · Zbl 1248.76089
[62] Zhu, J.-Y., Park, T., Isola, P. & Efros, A.A.2017 Unpaired image-to-image translation using cycle-consistent adversarial networks. In ICCV. IEEE.
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.