Reconstruction of three-dimensional turbulent flow structures using surface measurements for free-surface flows based on a convolutional neural network. (English) Zbl 07668170
The authors present a Convolutional Neural Network (CNN) designed to reconstruct the subsurface flow velocity of a turbulent open-channel flow using the surface information and investigate the relationship of the reconstruction performance to the flow physics. The CNN model consists of an encoder that extracts information from a two-dimensional discrete grid of surface variables, including the surface elevation and the fluctuating surface velocities, and a decoder that reconstructs a three-dimensional velocity field from the extracted surface information. It is found that the CNN method can infer the near-surface velocities with high accuracy, as verified by both a visual inspection of instantaneous fields and quantitative measures using normalised mean squared errors. The amplitude spectra and the phase errors of different Fourier modes are also analysed to assess the scale-specific reconstruction performance. The performance assessments indicate that the proposed CNN model is an effective tool for inferring the subsurface flow field and by using this method, a considerable amount of subsurface flow information, including the three-dimensional velocities of certain large-scale flow structures in the lower half of the channel, can be inferred from free surfaces.
Reviewer: Abdallah Bradji (Annaba)
MSC:
| 76M21 | Inverse problems in fluid mechanics |
| 76F10 | Shear flows and turbulence |
| 76F65 | Direct numerical and large eddy simulation of turbulence |
| 68T05 | Learning and adaptive systems in artificial intelligence |
| 68T07 | Artificial neural networks and deep learning |
Keywords:
machine learning; open-channel flow; deformable free surface; normalised mean squared reconstruction error; amplitude spectrum; Fourier modeReferences:
| [1] | Adrian, R.J. & Moin, P.1988Stochastic estimation of organized turbulent structure: homogeneous shear flow. J. Fluid Mech.190, 531-559. · Zbl 0642.76070 |
| [2] | Azulay, A. & Weiss, Y.2019Why do deep convolutional networks generalize so poorly to small image transformations?J. Machine Learning Res.20 (184), 1-25. · Zbl 1433.68389 |
| [3] | Brocchini, M. & Peregrine, D.H.2001The dynamics of strong turbulence at free surfaces. Part 1. Description. J. Fluid Mech.449, 225-254. · Zbl 1007.76027 |
| [4] | Brunton, S.L., Noack, B.R. & Koumoutsakos, P.2020Machine learning for fluid mechanics. Annu. Rev. Fluid Mech.52, 477-508. · Zbl 1439.76138 |
| [5] | Christensen, K.T. & Adrian, R.J.2001Statistical evidence of hairpin vortex packets in wall turbulence. J. Fluid Mech.431, 433-443. · Zbl 1008.76029 |
| [6] | Colburn, C.H., Cessna, J.B. & Bewley, T.R.2011State estimation in wall-bounded flow systems. Part 3. The ensemble Kalman filter. J. Fluid Mech.682, 289-303. · Zbl 1241.76303 |
| [7] | Dolcetti, G., Horoshenkov, K.V., Krynkin, A. & Tait, S.J.2016Frequency-wavenumber spectrum of the free surface of shallow turbulent flows over a rough boundary. Phys. Fluids28 (10), 105105. |
| [8] | Du, Y. & Zaki, T.A.2021Evolutional deep neural network. Phys. Rev. E104 (4), 045303. |
| [9] | Duraisamy, K., Iaccarino, G. & Xiao, H.2019Turbulence modeling in the age of data. Annu. Rev. Fluid Mech.51, 357-377. · Zbl 1412.76040 |
| [10] | Encinar, M.P. & Jiménez, J.2019Logarithmic-layer turbulence: a view from the wall. Phys. Rev. Fluids4 (11), 114603. |
| [11] | Erichson, N.B., Mathelin, L., Yao, Z., Brunton, S.L., Mahoney, M.W. & Kutz, J.N.2020Shallow neural networks for fluid flow reconstruction with limited sensors. Proc. R. Soc. A476 (2238), 20200097. · Zbl 1472.68172 |
| [12] | Fukami, K., Fukagata, K. & Taira, K.2019Super-resolution reconstruction of turbulent flows with machine learning. J. Fluid Mech.870, 106-120. |
| [13] | Fukami, K., Fukagata, K. & Taira, K.2021Machine-learning-based spatio-temporal super resolution reconstruction of turbulent flows. J. Fluid Mech.909, A9. · Zbl 1461.76306 |
| [14] | Fukami, K., Nakamura, T. & Fukagata, K.2020Convolutional neural network based hierarchical autoencoder for nonlinear mode decomposition of fluid field data. Phys. Fluids32 (9), 095110. |
| [15] | Gakhar, S., Koseff, J.R. & Ouellette, N.T.2020On the surface expression of bottom features in free-surface flow. J. Fluid Mech.900, A41. |
| [16] | 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. |
| [17] | Guastoni, L., Güemes, A., Ianiro, A., Discetti, S., Schlatter, P., Azizpour, H. & Vinuesa, R.2021Convolutional-network models to predict wall-bounded turbulence from wall quantities. J. Fluid Mech.928, A27. |
| [18] | Güemes, A., Discetti, S. & Ianiro, A.2019Sensing the turbulent large-scale motions with their wall signature. Phys. Fluids31 (12), 125112. |
| [19] | Güemes, A., Discetti, S., Ianiro, A., Sirmacek, B., Azizpour, H. & Vinuesa, R.2021From coarse wall measurements to turbulent velocity fields through deep learning. Phys. Fluids33 (7), 075121. |
| [20] | Guo, X. & Shen, L.2010Interaction of a deformable free surface with statistically steady homogeneous turbulence. J. Fluid Mech.658, 33-62. · Zbl 1205.76125 |
| [21] | He, K., Zhang, X., Ren, S. & Sun, J.2016 Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, pp. 770-778. IEEE. |
| [22] | Hu, J., Shen, L. & Sun, G.2018 Squeeze-and-Excitation networks. In 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, pp. 7132-7141. IEEE. |
| [23] | Jagodinski, E., Zhu, X. & Verma, S.2020 Uncovering dynamically critical regions in near-wall turbulence using 3D convolutional neural networks. arXiv:2004.06187. |
| [24] | Jeong, J. & Hussain, F.1995On the identification of a vortex. J. Fluid Mech.285, 69-94. · Zbl 0847.76007 |
| [25] | Jiménez, J.2018Machine-aided turbulence theory. J. Fluid Mech.854, R1. · Zbl 1415.76249 |
| [26] | Kauderer-Abrams, E.2017 Quantifying translation-invariance in convolutional neural networks. arXiv:1801.01450. |
| [27] | Kim, H., Kim, J., Won, S. & Lee, C.2021Unsupervised deep learning for super-resolution reconstruction of turbulence. J. Fluid Mech.910, A29. · Zbl 1461.76308 |
| [28] | Koltakov, S.2013 Bathymetry inference from free-surface flow features using large-eddy simulation. PhD thesis, Stanford University, Stanford, CA. |
| [29] | Komori, S., Murakami, Y. & Ueda, H.1989The relationship between surface-renewal and bursting motions in an open-channel flow. J. Fluid Mech.203, 103-123. |
| [30] | Komori, S., Nagaosa, R., Murakami, Y., Chiba, S., Ishii, K. & Kuwahara, K.1993Direct numerical simulation of three-dimensional open-channel flow with zero-shear gas-liquid interface. Phys. Fluids A5 (1), 115-125. · Zbl 0825.76572 |
| [31] | Kumar, S., Gupta, R. & Banerjee, S.1998An experimental investigation of the characteristics of free-surface turbulence in channel flow. Phys. Fluids10 (2), 437-456. |
| [32] | Lecun, Y., Bottou, L., Bengio, Y. & Haffner, P.1998Gradient-based learning applied to document recognition. Proc. IEEE86 (11), 2278-2324. |
| [33] | 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 |
| [34] | Li, B., Yang, Z., Zhang, X., He, G., Deng, B.-Q. & Shen, L.2020Using machine learning to detect the turbulent region in flow past a circular cylinder. J. Fluid Mech.905, A10. · Zbl 1460.76389 |
| [35] | 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 |
| [36] | Liu, B., Tang, J., Huang, H. & Lu, X.-Y.2020Deep learning methods for super-resolution reconstruction of turbulent flows. Phys. Fluids32 (2), 025105. |
| [37] | Long, J., Shelhamer, E. & Darrell, T.2015 Fully convolutional networks for semantic segmentation. In 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, MA, USA, pp. 3431-3440. IEEE. |
| [38] | Loshchilov, I. & Hutter, F.2019 Decoupled weight decay regularization. arXiv:1711.05101. |
| [39] | Lozano-Durán, A. & Jiménez, J.2014Time-resolved evolution of coherent structures in turbulent channels: characterization of eddies and cascades. J. Fluid Mech.759, 432-471. |
| [40] | Lund, B., Graber, H.C., Hessner, K. & Williams, N.J.2015On shipboard marine X-band radar near-surface current “calibration”. J. Atmos. Ocean. Technol.32 (10), 1928-1944. |
| [41] | Mandel, T.L., Rosenzweig, I., Chung, H., Ouellette, N.T. & Koseff, J.R.2017Characterizing free-surface expressions of flow instabilities by tracking submerged features. Exp. Fluids58 (11), 153. |
| [42] | Matsuo, M., Nakamura, T., Morimoto, M., Fukami, K. & Fukagata, K.2021 Supervised convolutional network for three-dimensional fluid data reconstruction from sectional flow fields with adaptive super-resolution assistance. arXiv:2103.09020. |
| [43] | Metoyer, S., Barzegar, M., Bogucki, D., Haus, B.K. & Shao, M.2021Measurement of small-scale surface velocity and turbulent kinetic energy dissipation rates using infrared imaging. J. Atmos. Ocean. Technol.38 (2), 269-282. |
| [44] | Milano, M. & Koumoutsakos, P.2002Neural network modeling for near wall turbulent flow. J. Comput. Phys.182 (1), 1-26. · Zbl 1090.76535 |
| [45] | Muraro, F., Dolcetti, G., Nichols, A., Tait, S.J. & Horoshenkov, K.V.2021Free-surface behaviour of shallow turbulent flows. J. Hydraul. Res.59 (1), 1-20. |
| [46] | Murata, T., Fukami, K. & Fukagata, K.2020Nonlinear mode decomposition with convolutional neural networks for fluid dynamics. J. Fluid Mech.882, A13. · Zbl 1430.76395 |
| [47] | Nagaosa, R.1999Direct numerical simulation of vortex structures and turbulent scalar transfer across a free surface in a fully developed turbulence. Phys. Fluids11 (6), 1581-1595. · Zbl 1147.76465 |
| [48] | Nagaosa, R. & Handler, R.A.2003Statistical analysis of coherent vortices near a free surface in a fully developed turbulence. Phys. Fluids15 (2), 375-394. · Zbl 1185.76272 |
| [49] | Odena, A., Dumoulin, V. & Olah, C.2016Deconvolution and checkerboard artifacts. Distill1 (10), e3. |
| [50] | Page, J., Brenner, M.P. & Kerswell, R.R.2021Revealing the state space of turbulence using machine learning. Phys. Rev. Fluids6 (3), 034402. |
| [51] | Pan, S.J. & Yang, Q.2010A survey on transfer learning. IEEE Trans. Knowl. Data Engng22 (10), 1345-1359. |
| [52] | Pan, Y. & Banerjee, S.1995A numerical study of free-surface turbulence in channel flow. Phys. Fluids7 (7), 1649-1664. · Zbl 1023.76564 |
| [53] | 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 |
| [54] | Park, J. & Choi, H.2020Machine-learning-based feedback control for drag reduction in a turbulent channel flow. J. Fluid Mech.904, A24. |
| [55] | Pinelli, M., Herlina, H., Wissink, J.G. & Uhlmann, M.2022Direct numerical simulation of turbulent mass transfer at the surface of an open channel flow. J. Fluid Mech.933, A49. |
| [56] | 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 |
| [57] | 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 |
| [58] | Rashidi, M.1997Burst-interface interactions in free surface turbulent flows. Phys. Fluids9 (11), 3485-3501. |
| [59] | Rumelhart, D.E., Hinton, G.E. & Williams, R.J.1986Learning representations by back-propagating errors. Nature323 (6088), 533-536. · Zbl 1369.68284 |
| [60] | Sanjou, M. & Nezu, I.2011Turbulence structure and coherent vortices in open-channel flows with wind-induced water waves. Environ. Fluid Mech.11 (2), 113-131. |
| [61] | Sarpkaya, T.1996Vorticity, free surface, and surfactants. Annu. Rev. Fluid Mech.28, 83-128. |
| [62] | Savelsberg, R. & Van De Water, W.2008Turbulence of a free surface. Phys. Rev. Lett.100 (3), 034501. |
| [63] | Shen, L., Zhang, X., Yue, D.K.P. & Triantafyllou, G.S.1999The surface layer for free-surface turbulent flows. J. Fluid Mech.386, 167-212. · Zbl 0938.76044 |
| [64] | Shorten, C. & Khoshgoftaar, T.M.2019A survey on image data augmentation for deep learning. J. Big Data6 (1), 60. |
| [65] | Simonyan, K., Vedaldi, A. & Zisserman, A.2014 Deep inside convolutional networks: visualising image classification models and saliency maps. arXiv:1312.6034. |
| [66] | Smilkov, D., Thorat, N., Kim, B., Viégas, F. & Wattenberg, M.2017 SmoothGrad: removing noise by adding noise. In Workshop on Visualization for Deep Learning. ICML. arXiv:1706.03825. |
| [67] | 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. |
| [68] | Suzuki, T. & Hasegawa, Y.2017Estimation of turbulent channel flow at \(Re_{\tau } = 100\) based on the wall measurement using a simple sequential approach. J. Fluid Mech.830, 760-796. · Zbl 1421.76139 |
| [69] | Tan, M. & Le, Q.2019 EfficientNet: rethinking model scaling for convolutional neural networks. Proceedings of the 36th International Conference on Machine Learning. Proc. Machine Learning Research97, 6105-6114. |
| [70] | Vahdat, A. & Kautz, J.2020 NVAE: a deep hierarchical variational autoencoder. In Proceedings of the 34th International Conference on Neural Information Processing Systems, (NeurIPS 2020), Vancouver, BC, Canada, (ed. H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan & H. Lin) pp. 19667-19679. Curran Associates. |
| [71] | Verleysen, M. & François, D.2005 The curse of dimensionality in data mining and time series prediction. In Computational Intelligence and Bioinspired Systems (ed. J. Cabestany, A. Prieto & F. Sandoval), pp. 758-770. Springer. |
| [72] | Wang, C., Gao, Q., Wang, B., Pan, C. & Wang, J.2021Vortex-to-velocity reconstruction for wall-bounded turbulence via the field-based linear stochastic estimation. J. Fluid Mech.922, A18. · Zbl 1486.76056 |
| [73] | Wang, G., Park, H.J. & Richter, D.H.2020Effect of computational domain size on inertial particle one-point statistics in open channel flow. Intl J. Multiphase Flow125, 103195. |
| [74] | 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. |
| [75] | Wang, M. & Zaki, T.A.2021State estimation in turbulent channel flow from limited observations. J. Fluid Mech.917, A9. · Zbl 1469.76053 |
| [76] | Wang, Q., Wang, M. & Zaki, T.A.2022What is observable from wall data in turbulent channel flow?J. Fluid Mech.941, A48. · Zbl 1503.76052 |
| [77] | 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), 10:1-10:21. |
| [78] | Xie, Y., Franz, E., Chu, M. & Thuerey, N.2018tempoGAN: a temporally coherent, volumetric GAN for super-resolution fluid flow. ACM Trans. Graph.37 (4), 95:1-95:15. |
| [79] | Xuan, A., Deng, B.-Q. & Shen, L.2019Study of wave effect on vorticity in Langmuir turbulence using wave-phase-resolved large-eddy simulation. J. Fluid Mech.875, 173-224. · Zbl 1419.76386 |
| [80] | Xuan, A., Deng, B.-Q. & Shen, L.2020Numerical study of effect of wave phase on Reynolds stresses and turbulent kinetic energy in Langmuir turbulence. J. Fluid Mech.904, A17. · Zbl 1460.76561 |
| [81] | Xuan, A. & Shen, L.2019A conservative scheme for simulation of free-surface turbulent and wave flows. J. Comput. Phys.378, 18-43. · Zbl 1416.76195 |
| [82] | Xuan, A. & Shen, L.2022Analyses of wave-phase variation of Reynolds shear stress underneath surface wave using streamline coordinates. J. Fluid Mech.931, A32. |
| [83] | Yamamoto, Y. & Kunugi, T.2011Direct numerical simulation of a high-Froude-number turbulent open-channel flow. Phys. Fluids23 (12), 125108. |
| [84] | Yang, D. & Shen, L.2011Simulation of viscous flows with undulatory boundaries. Part I. Basic solver. J. Comput. Phys.230 (14), 5488-5509. · Zbl 1419.76485 |
| [85] | Yao, J., Chen, X. & Hussain, F.2022Direct numerical simulation of turbulent open channel flows at moderately high Reynolds numbers. J. Fluid Mech.953, A19. |
| [86] | Ying, X.2019An overview of overfitting and its solutions. J. Phys.: Conf. Ser.1168, 022022. |
| [87] | Yokojima, S. & Nakayama, A.2002Direct numerical simulation of open-channel flow with free-surface fluctuations. J. Hydraul. Coast. Environ. JSCE2002 (712), 57-72. |
| [88] | Yoshimura, H. & Fujita, I.2020Investigation of free-surface dynamics in an open-channel flow. J. Hydraul. Res.58 (2), 231-247. |
| [89] | Zhang, C., Shen, L. & Yue, D.K.P.1999The mechanism of vortex connection at a free surface. J. Fluid Mech.384, 207-241. · Zbl 0936.76016 |
| [90] | Zhang, Q. & Zhu, S.2018Visual interpretability for deep learning: a survey. Front. Inform. Tech. El.19 (1), 27-39. |
| [91] | Zhang, R.2019 Making convolutional networks shift-invariant again. Proceedings of the 36th International Conference on Machine Learning. Proc. Machine Learning Research97, 7324-7334. |
| [92] | Zhang, Y., Tiňo, P., Leonardis, A. & Tang, K.2021A survey on neural network interpretability. IEEE Trans. Emerg. Top. Comput. Intell.5 (5), 726-742. |
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.
