Abstract
The presentation and modeling of turbulence anisotropy are crucial for studying large-scale turbulence structures and constructing turbulence models. However, accurately capturing anisotropic Reynolds stresses often relies on expensive direct numerical simulations (DNS). Recently, a hot topic in data-driven turbulence modeling is how to acquire accurate Reynolds stresses by the Reynolds-averaged Navier-Stokes (RANS) simulation and a limited amount of data from DNS. Many existing studies use mean flow characteristics as the input features of machine learning models to predict high-fidelity Reynolds stresses, but these approaches still lack robust generalization capabilities. In this paper, a deep neural network (DNN) is employed to build a model, map** from tensor invariants of RANS mean flow features to the anisotropy invariants of high-fidelity Reynolds stresses. From the aspects of tensor analysis and input-output feature design, we try to enhance the generalization of the model while preserving invariance. A functional framework of Reynolds stress anisotropy invariants is derived theoretically. Complete irreducible invariants are then constructed from a tensor group, serving as alternative input features for DNN. Additionally, we propose a feature selection method based on the Fourier transform of periodic flows. The results demonstrate that the data-driven model achieves a high level of accuracy in predicting turbulence anisotropy of flows over periodic hills and converging-diverging channels. Moreover, the well-trained model exhibits strong generalization capabilities concerning various shapes and higher Reynolds numbers. This approach can also provide valuable insights for feature selection and data generation for data-driven turbulence models.
![](http://media.springernature.com/lw685/springer-static/image/art%3A10.1007%2Fs10409-024-23629-x/MediaObjects/10409_2024_23629_Fig1_HTML.jpg)
摘要
湍流各向异性的表示和建模对于研究大尺度湍流结构和建立湍流模型至关重要. 然而, 准确计算雷诺应力的各向异性往往需要昂贵的直接数值模拟(DNS). **年来, 数据驱动湍流建模的一个热点是如何利用雷诺**均模拟方法(RANS)的数据和少量的DNS数据建立高精度的雷诺应力模型. 许多现有的研究实现了从**均流场特征到高精度的雷诺应力的预测模型, 但这些方法仍然缺乏**泛化能力. 本文采用深度神经网络(DNN), 以RANS的张量不变量作为输入映射高精度雷诺应力的各向异性不变量. 从张量分析和特征设计两方面, 在保持模型不变性的同时增**模型的泛化能力. 从理论上推导了雷诺应力各向异性不变量的泛函框架. 然后构造了完备不可约的 不变量, 作为DNN的备选输入特征. 此外, 我们提出了一种基于周期流动傅里叶变换的特征选择方法. 结果表明, 数据驱动模型在预测周期性流动和收缩扩张管道流动的湍流各向异性方面具有较高的精度, 而且训练的模型对外形和雷诺数具有很**的泛化能力. 本文的方法还可以为数据驱动湍流建模的特征选择和数据生成提供新的解决思路.
Data availability
The data that support the findings of this study are openly available in Kaggle at https://doi.org/10.34740/kaggle/dsv/2637500, Ref. [51], and within Refs. [36,46–49].
References
L. Biferale, and I. Procaccia, Anisotropy in turbulent flows and in turbulent transport, Phys. Rep. 414, 43 (2005).
J. L. Lumley, Computational modeling of turbulent flows, Adv. Appl. Mech. 18, 123 (1979).
J. L. Lumley, and G. R. Newman, The return to isotropy of homogeneous turbulence, J. Fluid Mech. 82, 161 (1977).
K.-S. Choi, A study of the return to isotropy of homogeneous turbulence (Cornell University, Ithaca, 1983).
K. S. Choi, and J. L. Lumley, The return to isotropy of homogeneous turbulence, J. Fluid Mech. 436, 59 (2001).
L. Terentiev, The turbulence closure model based on linear anisotropy invariant analysis, Dissertation for the Doctoral Degree (Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, 2006).
S. Banerjee, R. Krahl, F. Durst, and C. Zenger, Presentation of anisotropy properties of turbulence, invariants versus eigenvalue approaches, J. Turbul. 8, 1 (2007).
P. Å. Krogstad, and L. E. Torbergsen, Invariant analysis of turbulent pipe flow, Flow Turbul. Combust. 64, 161 (2000).
R. A. Antonia, L. W. B. Browne, and J. Kim, Some characteristics of small-scale turbulence in a turbulent duct flow, J. Fluid Mech. 233, 369 (1991).
J. Jovanovic, The Statistical Dynamics of Turbulence (Springer, Berlin, Heidelberg, 2004).
L. Neuhaus, P. Gilge, J. Seume, and K. Mulleners, in Influence of surface roughness on the turbulent properties in the wake of a turbine blade: Proceedings of the 18th International Symposium on the Applications of Laser and Imaging Techniques to Fluid Mechanics, Lisbon, 2016, p. 4.
M. Andersson, and M. Karlsson, Characterization of anisotropic turbulence behavior in pulsatile blood flow, Biomech Model Mechan. 20, 491 (2021).
J. X. Wang, J. L. Wu, and H. **ao, Physics-informed machine learning approach for reconstructing Reynolds stress modeling discrepancies based on DNS data, Phys. Rev. Fluids 2, 034603 (2017).
J. P. Li, D. G. Tang, C. Yi, and C. Yan, Data-augmented turbulence modeling by reconstructing Reynolds stress discrepancies for adverse-pressure-gradient flows, Phys. Fluids 34, 045110 (2022).
M. Emory, J. Larsson, and G. Iaccarino, Modeling of structural uncertainties in Reynolds-averaged Navier-Stokes closures, Phys. Fluids 25, 110822 (2013).
C. Gorlé, and G. Iaccarino, A framework for epistemic uncertainty quantification of turbulent scalar flux models for Reynolds-averaged Navier-Stokes simulations, Phys. Fluids 25, 055105 (2013).
M. Emory, and G. Iaccarino, Componentality-based wall-blocking for RANS models, Center for Turbulence Research Annual Research Briefs (CTR, Palo Alto, 2014), p. 193.
M. A. Emory, Estimating model-form uncertainty in Reynolds-averaged navier-stokes closures, Dissertation for the Doctoral Degree (Stanford University, Palo Alto, 2014).
K. Duraisamy, G. Iaccarino, and H. **ao, Turbulence modeling in the age of data, Annu. Rev. Fluid Mech. 51, 357 (2019).
K. Duraisamy, Perspectives on machine learning-augmented Reynolds-averaged and large eddy simulation models of turbulence, Phys. Rev. Fluids 6, 050504 (2021).
A. P. Singh, S. Medida, and K. Duraisamy, Machine-learning-augmented predictive modeling of turbulent separated flows over airfoils, AIAA J. 55, 2215 (2017).
C. Wu, and Y. Zhang, Enhancing the shear-stress-transport turbulence model with symbolic regression: A generalizable and interpretable data-driven approach, Phys. Rev. Fluids 8, 084604 (2023).
X. L. Shan, Y. L. Liu, W. B. Cao, X. X. Sun, and W. W. Zhang, Turbulence modeling via data assimilation and machine learning for separated flows over airfoils, AIAA J. 61, 3883 (2023).
Z. Wang, and W. Zhang, A unified method of data assimilation and turbulence modeling for separated flows at high Reynolds numbers, Phys. Fluids 35, 025124 (2023).
L. Zhu, T. Wang, Z. Fan, X. **ang, W. Zhang, and X. Yuan, Physics-assisted recursive method for sample selection from wall-bounded turbulence data, Phys. Fluids 34, 085132 (2022).
L. Zhu, X. Sun, Y. Liu, and W. Zhang, One neural network approach for the surrogate turbulence model in transonic flows, Acta Mech. Sin. 38, 321187 (2022).
L. Hou, B. Zhu, and Y. Wang, kεNet: Discovering the turbulence model and applying for low Reynolds number turbulent channel flow, Acta Mech. Sin. 39, 322326 (2023).
L. Zhu, W. Zhang, and G. Tu, Generalization enhancement of artificial neural network for turbulence closure by feature selection, Adv. Aerodyn. 4, 1 (2022).
J. P. Panda, and H. V. Warrior, Evaluation of machine learning algorithms for predictive Reynolds stress transport modeling, Acta Mech. Sin. 38, 321544 (2022).
H. **e, Y. Zhao, and Y. Zhang, Data-driven nonlinear K-L turbulent mixing model via gene expression programming method, Acta Mech. Sin. 39, 322315 (2023).
Y. Yin, Z. Shen, Y. Zhang, H. Chen, and S. Fu, An iterative data-driven turbulence modeling framework based on Reynolds stress representation, Theor. Appl. Mech. Lett. 12, 100381 (2022).
U. Schumann, Realizability of Reynolds-stress turbulence models, Phys. Fluids 20, 721 (1977).
A. J. Simonsen, and P. Å. Krogstad, Turbulent stress invariant analysis: Clarification of existing terminology, Phys. Fluids 17, 088103 (2005).
S. Banerjee, O. Ertunc, and F. Durst, in Anisotropy properties of turbulence: Proceedings of the 13th WSEAS International Conference on Applied Mathematics, Wisconsin, 2008, pp. 26–57.
M. Emory, and G. Iaccarino, Visualizing turbulence anisotropy in the spatial domain with componentality contours, Center for Turbulence Research Annual Research Briefs (CTR, Palo Alto, 2014), p. 123.
H. **ao, J. L. Wu, S. Laizet, and L. Duan, Flows over periodic hills of parameterized geometries: A dataset for data-driven turbulence modeling from direct simulations, Comput. Fluids 200, 104431 (2020).
S. B. Pope, A more general effective-viscosity hypothesis, J. Fluid Mech. 72, 331 (1975).
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).
C. G. Speziale, and T. B. Gatski, Analysis and modelling of anisotropies in the dissipation rate of turbulence, J. Fluid Mech. 344, 155 (1997).
M. Frewer, More clarity on the concept of material frame-indifference in classical continuum mechanics, Acta Mech 202, 213 (2009).
M. Itskov, Tensor Algebra and Tensor Analysis for Engineers (Springer, Cham, 2007).
J. L. Wu, H. **ao, and E. Paterson, Physics-informed machine learning approach for augmenting turbulence models: A comprehensive framework, Phys. Rev. Fluids 3, 074602 (2018).
E. L. Peters, R. Balin, K. E. Jansen, A. Doostan, and J. A. Evans, S-frame discrepancy correction models for data-informed Reynolds stress closure, J. Comput. Phys. 448, 110717 (2022).
A. J. M. Spencer, and R. S. Rivlin, Isotropic integrity bases for vectors and second-order tensors, Arch. Rational Mech. Anal. 9, 45 (1962).
X. Wang, J. Kou, W. Zhang, and Z. Liu, Incorporating physical models for dynamic stall prediction based on machine learning, AIAA J. 60, 4428 (2022).
R. McConkey, E. Yee, and F. S. Lien, A curated dataset for data-driven turbulence modelling, Sci. Data 8, 255 (2021).
M. Marquillie, J.-P. Laval, and R. Dolganov, Direct numerical simulation of a separated channel flow with a smooth profile, J. Turbul. 9, N1 (2008).
J.-P. Laval, and M. Marquillie, Direct numerical simulations of converging-diverging channel flow, in: Progress in Wall Turbulence: Understanding and Modeling (Springer, Dordrecht, 2009), pp. 203–209.
L. A. C. A. Schiavo, A. B. Jesus, J. L. F. Azevedo, and W. R. Wolf, Large Eddy Simulations of convergent-divergent channel flows at moderate Reynolds numbers, Int. J. Heat Fluid Flow 56, 137 (2015).
K. E. Taylor, Summarizing multiple aspects of model performance in a single diagram, J. Geophys. Res. 106, 7183 (2001).
R. McConkey, E. Yee, and F. S. Lien, A curated dataset for data-driven turbulence modeling, Sci. Data 8, 255 (2021).
Acknowledgements
This work was supported by the National Natural Science Foundation of China (Grant No. 92152301).
Author information
Authors and Affiliations
Contributions
Author contributions Weiwei Zhang and **anglin Shan designed the research. Weiwei Zhang acquired the financial support for the project leading to this publication. **anglin Shan investigated the methodology, designed the computational framework, constructed the model, and wrote the manuscript. Xuxiang Sun and Wenbo Cao help to formulate overarching research goals and analyze the results. Zhenhua **a and Weiwei Zhang helped organize the manuscript, revised, and edited the final version.
Corresponding author
Ethics declarations
Conflict of interest On behalf of all authors, the corresponding author states that there is no conflict of interest.
Rights and permissions
About this article
Cite this article
Shan, X., Sun, X., Cao, W. et al. Modeling Reynolds stress anisotropy invariants via machine learning. Acta Mech. Sin. 40, 323629 (2024). https://doi.org/10.1007/s10409-024-23629-x
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10409-024-23629-x