Deep Learning for Data-Driven Turbulence Modeling in Flow over Periodic Hills
Abstract
Data-driven method has gained rapid growth in recent years. It is driven by the rise of big data in various fields. Nowadays, Deep learning is the most famous data-driven method used in wide range of applications such as in fluid mechanics. Turbulence modeling is an unsolved problem in fluid mechanics. Reynolds-averaged Navier-Stokes (RANS) is the most popular method for turbulence modeling in real-world problems. The objective of RANS turbulence modeling is to relate the Reynolds stress with the mean flow properties. The weakness of the RANS model has driven the research to develop another approach. The application of deep learning in turbulence modeling has shown promising results in recent years. In this work, deep learning is used to develop a model for turbulence closure modeling. The performance of this model is compared with RANS k-ω model as the classical turbulence model. From the results of this work, it is shown that the neural network model proposed by the author could give better performance on giving the closure relation for turbulent flow over periodic hills which gives 57% RMSE improvement from the RANS model and could capture the separation phenomenon when RANS model is struggling.
References
[2] Craft, T., Launder, B. & Suga, K., Prediction of turbulent transitional phenomena with a nonlinear eddy-viscosity model, International Journal of Heat and Fluid Flow, 18, pp. 15-28, 1997.
[3] Pope, S, Turbulent flows, ed. 1, Cambridge University Press, 2000.
[4] Parish, E.J. & Duraisamy, K., A paradigm for data-driven predictive modelling using field inversion and machine learning. Journal of Computational Physics, 305, pp. 758–774. 2016.
[5] Singh, A.P., Medida, S. & Duraisamy, K., Machine-learning-augmented predictive modeling of turbulent separated flows over airfoils. AIAA Journal, 55:7, pp. 2215-2227, Apr. 2017.
[6] Tracey, B., Duraisamy, K. & Alonso, J. J., A machine learning strategy to assist turbulence model development. 53rd AIAA Aerospace Sciences Meeting, 1287, 2015.
[7] Duraisamy, K., Iaccarino, G. & Xiao, H., Turbulence modeling in the age of data. Annual Review of Fluid Mechanics, 51, pp. 357–377, Jan. 2019.
[8] Ling, J., Kurzawski, A. & Templeton, J., Reynolds averaged turbulence modelling using deep neural networks with embedded invariance, Journal of Fluid Mechanics, 807, pp. 155-166, Oct. 2016.
[9] Kandoorp, M. & Dwight, R.P., Data-driven modelling of the Reynolds stress tensor using random forests with invariance, Computers & Fluids, 202, Apr. 2020.
[10] Pope, S. B., A more general effective-viscosity hypothesis, Journal of Fluid Mechanics, 72, pp. 331, Nov. 1975.
[11] Fang, R., Sondak, D., Protopapas, P. & Succi, S., Neural Network Models for the Anisotropic Reynolds Stress Tensor in Turbulent Channel Flow, Journal of Turbulence, 21, pp. 525-543, Feb. 2020.
[12] Xiao, H., Wu, J.L., Laizet, S. & Duan, L., Flows over periodic hills of parameterized geometries: A dataset for data-driven turbulence modeling from direct simulations. Computers and Fluids, 200, Mar. 2020.
[13] Wilcox, D.C., Reassessment of the Scale-Determining Equation for Advanced Turbulence Models, AIAA Journal, 26, pp. 1299, Nov. 1988.
[14] Kingma, D.P. & Jimmy, B., Adam: A Method for Stochastic Optimization. International Conference on Learning Representations, 2015.
[15] McConkey, R., Yee, E. & Lien, F.S., A curated dataset for data-driven turbulence modelling, Sci Data, 8, pp. 255. Sep. 2021.
