Identification of turbulence eddy viscosity coefficient in supersonic isolation section based on deep neural network
-
摘要:
雷诺平均Navier-Stokes(RANS)方程由于计算成本较低,当前仍然在工程设计领域广泛应用。为了进一步提升计算精度和减少时间,应用深度神经网络(deep neural networks, DNN)方法自适应辨识稳态湍流涡黏性系数。以隔离段的激波串前缘位置检测流场生成为例,使用Wilcox-2006
$k$ -$ \omega $ 湍流模型进行模拟。在不同的背压情况下,产生稳定状态的湍流涡黏性流场作为模型学习的训练数据集。最后在不同背压条件下开展了测试。结果表明:提出的DNN方法能够快速预测湍流涡黏性系数的值,预测值与计算流体力学数值模拟计算的参数值相比,方均根误差较小,可决系数大于99%,预测的流场结果与真实流场基本一致,进一步验证了深度学习技术在湍流模型参数辨识的可行性。Abstract:The Reynolds-averaged Navier-Stokes (RANS) equation is still widely used in engineering design due to its low computational cost. In order to further improve the calculation accuracy and reduce the time, a deep neural network (deep neural networks, DNN) method was applied to adaptively identify the steady-state turbulent eddy viscosity coefficient. Taking the detection flow field generated at the front edge of the shock train in the isolation section as an example, the Wilcox-2006
$k$ -$ \omega $ turbulence model was used for simulation. A steady-state turbulent eddy-viscous flow field was generated as a training dataset for model learning under different back pressure conditions. Finally, tests were carried out under different back pressure conditions. The results showed that the proposed DNN method can quickly predict the value of the turbulent eddy viscosity coefficient. The coefficient of determination was greater than 99%, and the predicted flow field results were basically consistent with the real flow field, which further verified the feasibility of deep learning technology in turbulence model parameter identification. -
表 1 参数的物理意义
Table 1. Physical meaning of parameters
参数 物理意义 $ {T_{\text{t}}} $ 总温 $ {p_{\text{t}}} $ 总压 $\,\rho$ 点的密度 $ {k_{{\text{turb}}}} $ 湍流模型$ k $的值 $x (x)$,$y (x)$, ${\textit{z}} (x)$ 流场中点的网格坐标 $Ma$ 马赫数 $t$ 温度 $ p $ 点的压力 ${v_{\text{t} } } (x)$ 涡黏性系数 $ {w_{{\text{turb}}}} $ 湍流模型$ w $的值 $u (x)$,$v (x)$, $w (x)$ 流场中点的速度分量 $ {C_{{\text{lim}}}} $ 应力限制系数 $ \kappa $ 卡门常数 -
[1] 陈植. 超燃冲压发动机隔离段流动机理及其控制的试验研究[D]. 合肥: 国防科学技术大学, 2015.CHEN Zhi. Experimental study on flow mechanism and control of isolation section of scramjet engine[D]. Hefei: National University of Defense Technology, 2015. (in Chinese) [2] DOLLING D S. Fifty years of shock-wave/boundary-layer interaction research: what next?[J]. AIAA Journal,2001,39(8): 1517-1531. doi: 10.2514/2.1476 [3] 贺武生. 超燃冲压发动机研究综述[J]. 火箭推进,2005(1): 29-32. doi: 10.3969/j.issn.1672-9374.2005.01.006HE Wusheng. A review of scramjet research[J]. Journal of Rocket Propulsion,2005(1): 29-32. (in Chinese) doi: 10.3969/j.issn.1672-9374.2005.01.006 [4] MATSUO K. Shock train and pseudo-shock phenomena in supersonic internal flows[J]. Journal of Thermal Science,2003(3): 204-208. [5] 詹王杰,严红. 隔离段内激波串受稳态和脉动背压影响的数值研究[J]. 推进技术,2021,42(5): 980-990. doi: 10.13675/j.cnki.tjjs.190671ZHAN Wangjie,YAN Hong. Numerical study on the influence of shock string on steady-state and pulsating back pressure in isolation section[J]. Journal of Propulsion Technology,2021,42(5): 980-990. (in Chinese) doi: 10.13675/j.cnki.tjjs.190671 [6] 吴正园,莫凡,高振勋,等. 湍流边界层与高温气体效应耦合的直接数值模拟[J]. 空气动力学学报,2020,38(6): 1111-1119. doi: 10.7638/kqdlxxb-2020.0132WU Zhengyuan,MO Fan,GAO Zhenxun,et al. Direct numerical simulation of coupling of turbulent boundary layer and hot gas effect[J]. Acta Aerodynamica Sinica,2020,38(6): 1111-1119. (in Chinese) doi: 10.7638/kqdlxxb-2020.0132 [7] XIAO Heng,CINNELLA P. Quantification of model uncertainty in RANS simulations: a review[J]. Progress in Aerospace Science,2019,108: 1-31. doi: 10.1016/j.paerosci.2018.10.001 [8] 尹宇辉,李浩然,张宇飞,等. 机器学习辅助湍流建模在分离流预测中的应用[J]. 空气动力学学报,2021,39(2): 23-32. doi: 10.7638/kqdlxxb-2020.0155YIN Yuhui,LI Haoran,ZHANG Yufei,et al. Application of machine learning-assisted turbulence modeling in separated flow prediction[J]. Acta Aerodynamica Sinica,2021,39(2): 23-32. (in Chinese) doi: 10.7638/kqdlxxb-2020.0155 [9] CRAFT T J,LAUNDER B E,SUGA K. Development and application of a cubic eddy-viscosity model of turbulence[J]. International Journal of Heat and Fluid Flow,1996,17(2): 108-115. doi: 10.1016/0142-727X(95)00079-6 [10] SPALART P R. Strategies for turbulence modelling and simulations[J]. International Journal of Heat and Fluid Flow,2000,21(3): 252-263. doi: 10.1016/S0142-727X(00)00007-2 [11] STEFANO M,HOSDER S,BAURLE R A. Effect of turbulence model uncertainty on scramjet isolator flowfield analysis[J]. Journal of Propulsion and Power,2019,36(1): 1-14. [12] SCHAEFER J,HOSDER S,WEST T,et al. Uncertainty quantification of turbulence model closure coefficients for transonic wall-bounded flows[J]. AIAA Journal,2017,55(1): 195-213. doi: 10.2514/1.J054902 [13] XIAO H,WU J L,WANG J X,et al. Quantifying and reducing model form uncertainties in Reynolds-averaged Navier-Stokes simulations: a data-driven, physics informed Bayesian approach[J]. Journal of Computational Physics,2016,324: 115-136. doi: 10.1016/j.jcp.2016.07.038 [14] MAULIK R,SAN O,RASHEED A,et al. Data-driven deconvolution for large eddy simulations of Kraichnan turbulence[J]. Physics of Fluids,2018,30(12): 125109.1-125109.15. [15] YARLANKI S, RAJENDRAN B, HAMANN H. Estimation of turbulence closure coefficients for data centers using machine learning algorithms[R]. San Diego, Canada: 13th Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems, 2012. [16] ZHANG Zejia, DURAISAM K. Machine learning methods for data-driven turbulence modeling[R]. Dallas, US: 22nd AIAA Computational Fluid Dynamics Conference, 2015. [17] KAANDORP M. Machine learning for data-driven RANS turbulence modelling[D]. Delft, US: Delft University of Technology, 2018. [18] SAAKAAR B,YASER A,SHAOWU P,et al. Prediction of aerodynamic flow fields using convolutional neural networks[J]. Computational Mechanics,2019,64(2): 525-545. doi: 10.1007/s00466-019-01740-0 [19] ZHU Linyang,ZHANG Weiwei,KOU Jiaqing,et al. Machine learning methods for turbulence modeling in subsonic flows around airfoils[J]. Physics of Fluids,2019,31(1): 015105.1-015105.14. [20] PANDA J P,WARRIOR H V. Data-driven prediction of complex flow field over an axisymmetric body of revolution using machine learning[J]. Fluid Dynamics (preprint),2021,arXiv: 2111.07937. [21] SHUVAYAN B,ANANTHAKRISHNAN B,HIDEAKI O. Fast estimation of internal flowfields in scramjet intakes via reduced-order modeling and machine learning[J]. Physics of Fluids,2021,33(10): 106110.1-106110.24. [22] SUN Xuxiang,CAO Wenbo,LIU Yilang,et al. High Reynolds number airfoil turbulence modeling method based on machine learning technique[J]. Computers and Fluids,2022,236: 105298.1-105298.30. [23] 何创新,邓志文,刘应征. 湍流数据同化技术及应用[J]. 航空学报,2021,42(4): 167-184.HE Chuangxin,DENG Zhiwen,LIU Yingzheng. Turbulent data assimilation technology and its application[J]. Chinese Journal of Aeronautics,2021,42(4): 167-184. (in Chinese) [24] LI Yunfei,CHANG Juntao,KONG Chen,et al. Recent progress of machine learning in flow modeling and active flow control[J]. Chinese Journal of Aeronautics,2021,35(4): 14-44. [25] KONG Chen,CHANG Juntao,LI Yunfei,et al. Flowfield reconstruction and shock train leading edge detection in scramjet isolators[J]. AIAA Journal,2020,58(9): 4068-4080. doi: 10.2514/1.J059302 [26] KONG Chen,CHANG Juntao,LI Yunfei,et al. A deep learning approach for the velocity field prediction in a scramjet isolator[J]. Physics of Fluids,2021,33(2): 026103.1-026103.18. [27] KONG Chen,CHANG Juntao,WANG Ziao,et al. Prediction model of flow field in an isolator over various operating conditions[J]. Aerospace Science and Technology,2021,111: 106576.1-106576.14. [28] LI Nan,CHANG Juntao,XU Kejing,et al. Prediction dynamic model of shock train with complex background waves[J]. Physics of Fluids,2017,29: 116103.1-116103.16. [29] KUKREJA H,BHARATH N,SIDDESH C S,et al. An introduction to artificial neural network[J]. International Journal of Advance Research and Innovative Ideas in Education,2016,1: 27-30.