Abstract
The application of machine learning (ML) algorithms to turbulence modeling has shown promise over the last few years, but their application has been restricted to eddy viscosity based closure approaches. In this article, we discuss the rationale for the application of machine learning with high-fidelity turbulence data to develop models at the level of Reynolds stress transport modeling. Based on these rationales, we compare different machine learning algorithms to determine their efficacy and robustness at modeling the different transport processes in the Reynolds stress transport equations. Those data-driven algorithms include Random forests, gradient boosted trees, and neural networks. The direct numerical simulation (DNS) data for flow in channels are used both as training and testing of the ML models. The optimal hyper-parameters of the ML algorithms are determined using Bayesian optimization. The efficacy of the above-mentioned algorithms is assessed in the modeling and prediction of the terms in the Reynolds stress transport equations. It was observed that all three algorithms predict the turbulence parameters with an acceptable level of accuracy. These ML models are then applied for the prediction of the pressure strain correlation of flow cases that are different from the flows used for training, to assess their robustness and generalizability. This explores the assertion that ML-based data-driven turbulence models can overcome the modeling limitations associated with the traditional turbulence models and ML models trained with large amounts of data with different classes of flows can predict flow field with reasonable accuracy for unknown flows with similar flow physics. In addition to this verification, we carry out validation for the final ML models by assessing the importance of different input features for prediction.
摘要
近年来, 机器学习(ML)算法在湍流建模中的应用显示出了希望, 但它们的应用仅限于基于涡流黏度的封闭方法. 本文讨论了 应用机器学习与高保真湍流数据的基本原理, 以开发雷诺应力传输建模水平的模型. 基于这些理论, 本文比较了不同的机器学习算法, 以确定它们在雷诺应力输运方程中建模不同输运过程的有效性和鲁棒性. 这些数据驱动的算法包括随机森林、梯度增强树和神经网 络. 采用直接数值模拟(DNS)数据作为ML模型的训练和测试, 利用贝叶斯优化方法确定了ML算法的最优超参数. 在雷诺应力输运方程 的建模和预测中, 评估了上述算法的有效性. 观察到三种算法都以可接受的精度水平预测湍流参数. 再将这些模型应用在不同于训练 的流动情况的压力应变相关性的预测, 以评估其鲁棒性和通用性. 这探讨了基于ML的数据驱动湍流模型可以克服传统湍流模型的建 模局限性, 并且ML模型用大量不同类型的流数据训练, 可以对具有相似流物理的未知流以合理的精度预测流场. 除此之外, 本文通过 评估不同输入特征的重要性来验证最终的ML模型.
Similar content being viewed by others
References
H. K. Versteeg, and W. Malalasekera, An introduction to computational fluid dynamics: the finite volume method, Pearson education (2007).
S. Pope, Turbulent Flows (Cambridge University Press, New York, 2000).
Z. X. Sun, M. Y. Wang, L. Y. Wei, F. B. Kong, and G. W. Yang, Aerodynamic shape optimization of an urban maglev train, Acta Mech. Sin. 37, 954 (2021).
L. Chen, and Y. Dong, Numerical investigation on fluid forces of piggyback circular cylinders in tandem arrangement at low Reynolds numbers, Acta Mech. Sin. 37, 599 (2021).
M. Y. Wang, S. A. Hashmi, Z. X. Sun, D. L. Guo, G. Vita, G. W. Yang, and H. Hemida, Effect of surface roughness on the aerodynamics of a high-speed train subjected to crosswinds, Acta Mech. Sin. 37, 1090 (2021).
S. B. Pope, A more general effective-viscosity hypothesis, J. Fluid Mech. 72, 331 (1975).
J. P. Panda, A reliable pressure strain correlation model for complex turbulent flows, JAFM 13, 1167 (2020).
A. A. Mishra, and S. S. Girimaji, Toward approximating non-local dynamics in single-point pressure-strain correlation closures, J. Fluid Mech. 811, 168 (2017).
J. P. Panda, and H. V. Warrior, A representation theory-based model for the rapid pressure strain correlation of turbulence, J. Fluids Eng. 140, 081101 (2018).
M. Lee, and R. D. Moser, Direct numerical simulation of turbulent channel flow up to, J. Fluid Mech. 774, 395 (2015), arXiv: 1410.7809.
M. Lee, and R. D. Moser, Extreme-scale motions in turbulent plane Couette flows, J. Fluid Mech. 842, 128 (2018), arXiv: 1706.09800.
X. Qian, H. Lu, C. Zou, H. Zhang, S. Shao, and H. Yao, Numerical investigation of the effects of turbulence on the ignition process in a turbulent MILD flame, Acta Mech. Sin. 37, 1299 (2021).
J. Panda, A review of pressure strain correlation modeling for Reynolds stress models, Proc. Institution Mech. Engineers Part C-J. Mech. Eng. Sci. 234, 1528 (2020).
J. P. Panda, A reliable pressure strain correlation model for complex turbulent flows, J. Appl. Fluid. Mech. 13, 1167 (2020).
K. Duraisamy, G. Iaccarino, and H. Xiao, Turbulence modeling in the age of data, Annu. Rev. Fluid Mech. 51, 357 (2019), arXiv: 1804.00183.
J. X. Wang, J. L. Wu, and H. Xiao, Physics-informed machine learning approach for reconstructing Reynolds stress modeling discrepancies based on DNS data, Phys. Rev. Fluids 2, 034603 (2017), arXiv: 1606.07987.
M. L. A. Kaandorp, and R. P. Dwight, Data-driven modelling of the Reynolds stress tensor using random forests with invariance, Comput. Fluids 202, 104497 (2020).
Y. Luan, and R. P. Dwight, Influence of turbulence anisotropy on RANS predictions of wind-turbine wakes, J. Phys.-Conf. Ser. 1618, 062059 (2020).
J. Wu, J. Li, X. Qiu, and Y. Liu, A comparative analysis of multi-machine learning algorithms for data-driven RANS turbulence modelling, J. Phys.-Conf. Ser. 1684, 012043 (2020).
Y. Zhao, H. D. Akolekar, J. Weatheritt, V. Michelassi, and R. D. Sandberg, RANS turbulence model development using CFD-driven machine learning, J. Comput. Phys. 411, 109413 (2020), arXiv: 1902.09075.
M. Schmelzer, R. P. Dwight, and P. Cinnella, Discovery of algebraic reynolds-stress models using sparse symbolic regression, Flow Turbul. Combust 104, 579 ((2020).
A. P. Singh, S. Medida, and K. Duraisamy, Machine-learning-augmented predictive modeling of turbulent separated flows over air-foils, AIAA J. 55, 2215 (2017).
B. D. Tracey, K. Duraisamy, and J.J. Alonso, in A machine learning strategy to assist turbulence model development: Procdeedings of the 53rd AIAA Aerospace Sciences Meeting, 2015, p. 1287.
E. J. Parish, and K. Duraisamy, A paradigm for data-driven predictive modeling using field inversion and machine learning, J. Comput. Phys. 305, 758 (2016).
R. Maulik, O. San, A. Rasheed, and P. Vedula, Data-driven deconvolution for large eddy simulations of Kraichnan turbulence, Phys. Fluids 30, 125109 (2018).
J. Ling, A. Kurzawski, and J. Templeton, Reynolds averaged turbulence modelling using deep neural networks with embedded invariance, J. Fluid Mech. 807, 155 (2016).
L. Zhu, W. Zhang, J. Kou, and Y. Liu, Machine learning methods for turbulence modeling in subsonic flows around airfoils, Phys. Fluids 31, 015105 (2019).
J. L. Wu, H. Xiao, and E. Paterson, Physics-informed machine learning approach for augmenting turbulence models: A comprehensive framework, Phys. Rev. Fluids 3, 074602 (2018), arXiv: 1801.02762.
J. Weatheritt, and R. Sandberg, A novel evolutionary algorithm applied to algebraic modifications of the RANS stress-strain relationship, J. Comput. Phys. 325, 22 (2016).
J. Weatheritt, and R. D. Sandberg, The development of algebraic stress models using a novel evolutionary algorithm, Int. J. Heat Fluid Flow 68, 298 (2017).
J. P. Huijing, R. P. Dwight, and M. Schmelzer, Data-driven RANS closures for three-dimensional flows around bluff bodies, Comput. Fluids 225, 104997 (2021).
R. Fang, D. Sondak, P. Protopapas, and S. Succi, Neural network models for the anisotropic Reynolds stress tensor in turbulent channel flow, J. Turbulence 21 525 (2020), arXiv: 1909.03591.
A. Beck, D. Flad, and C. D. Munz, Deep neural networks for data-driven LES closure models, J. Comput. Phys. 398, 108910 (2019), arXiv: 1806.04482.
N. Parashar, B. Srinivasan, and S. S. Sinha, Modeling the pressure-Hessian tensor using deep neural networks, Phys. Rev. Fluids 5, 114604 (2020).
A. A. Mishra, K. Duraisamy, and G. Iaccarino, Estimating uncertainty in homogeneous turbulence evolution due to coarse-graining, Phys. Fluids 31, 025106 (2019).
W. N. Edeling, G. Iaccarino, and P. Cinnella, Data-free and data-driven RANS predictions with quantified uncertainty, Flow Turbul. Combust 100, 593 (2018).
A. A. Mishra, and S. Girimaji, Linear analysis of non-local physics in homogeneous turbulent flows, Phys. Fluids 31, 035102 (2019).
T. B. Gatski, and C. G. Speziale, On explicit algebraic stress models for complex turbulent flows, J. Fluid Mech. 254, 59 (1993).
A. A. Mishra, and S. S. Girimaji, On the realizability of pressure-strain closures, J. Fluid Mech. 755, 535 (2014).
A. A. Mishra, and S. S. Girimaji, Intercomponent energy transfer in incompressible homogeneous turbulence: multi-point physics and amenability to one-point closures, J. Fluid Mech. 731, 639 (2013).
A. A. Mishra, and S. S. Girimaji, Pressure-strain correlation modeling: towards achieving consistency with rapid distortion theory, Flow Turbul. Combust 85, 593 (2010).
J. P. Panda, H. V. Warrior, S. Maity, A. Mitra, and K. Sasmal, An Improved Model Including Length Scale Anisotropy for the Pressure Strain Correlation of Turbulence, J. Fluids Eng. 139, 044503 (2017).
S. S. Girimaji, Pressure strain correlation modelling of complex turbulent flows, J. Fluid Mech. 422, 91 (2000).
C. G. Speziale, S. Sarkar, and T. B. Gatski, Modelling the pressure-strain correlation of turbulence—An invariant dynamical systems approach, J. Fluid Mech. 227, 245 (1991).
J. Panda, A review of pressure strain correlation modeling for Reynolds stress models, Proc Instit. Mech. Eng. Part C-J. Mech. Eng. Sci. 234, 1528 (2020).
A. A. Mishra, and S. S. Girimaji, Hydrodynamic stability of three-dimensional homogeneous flow topologies, Phys. Rev. E 92, 053001 (2015).
J. Rotta, Statistische theorie nichthomogener turbulenz. Z. Phys. 129, 547 (1951).
S. Sarkar, and C. G. Speziale, A simple nonlinear model for the return to isotropy in turbulence, Phys. Fluids A-Fluid Dyn. 2, 84 (1990).
H. Warrior, S. Mathews, S. Maity, and K. Sasmal, An Improved Model for the Return to Isotropy of Homogeneous Turbulence, J. Fluids Eng. 136, 034501 (2014).
J. Bardino, J. H. Ferziger, and W. C. Reynolds, Improved turbulence models based on large eddy simulation of homogeneous, incompressible turbulent flows, Report, Stanford University, 1983.
H. Chen, Y. Zeng, and Y. Li, Machine learning models for the secondary Bjerknes force between two insonated bubbles, Acta Mech. Sin. 37, 35 (2021), arXiv: 2001.08291.
K. Fukami, K. Fukagata, and K. Taira, Assessment of supervised machine learning methods for fluid flows, Theor. Comput. Fluid Dyn. 34, 497 (2020), arXiv: 2001.09618.
L. Breiman, Machine Learn. 45, 5 (2001).
F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay, Scikit-learn: machine learning in Python, J. Mach. Learn. Res. 12, 2825 (2011).
ANSYS: ANSYS FLuent 14.0 User’s Guide. ANSYS.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Panda, J.P., Warrior, H.V. Evaluation of machine learning algorithms for predictive Reynolds stress transport modeling. Acta Mech. Sin. 38, 321544 (2022). https://doi.org/10.1007/s10409-022-09001-w
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10409-022-09001-w