Improving accuracy of turbulence models by neural network

Satoshi Miyazaki,Yuji Hattori

DOI: https://doi.org/10.48550/arXiv.2012.01723

2020-12-03

Abstract:Neural networks of simple structures are used to construct a turbulence model for large-eddy simulation (LES). Data obtained by direct numerical simulation (DNS) of homogeneous isotropic turbulence are used to train neural networks. It is shown that two methods are effective for improvement of accuracy of the model: weighting data for training and addition of the second-order derivatives of velocity to the input variables. As a result, high correlation between the exact subgrid scale stress and the prediction by the neural network is obtained for large filter width; the correlation coefficient is about 0:9 and 0:8 for filter widths 48:8\eta and 97:4\eta, respectively, where \eta is the Kolmogorov scale. The models established by neural networks are close to but not identical with the gradient models. LES with the neural network model is performed for the homogeneous isotropic turbulence and the initial-value problem of the Taylor-Green vortices. The results obtained with the neural network model are in reasonable agreement with those of the filtered DNS. However, symmetry in the latter problem is broken since the neural network model does not possess rigorous symmetry under orthogonal transformations.

Fluid Dynamics

What problem does this paper attempt to address?

This paper attempts to improve the accuracy of turbulence models through neural networks. Specifically, researchers use a simple neural network structure to construct turbulence models in large - eddy simulation (LES). They train the neural network with data obtained from direct numerical simulation (DNS) and present two effective methods to improve the accuracy of the model: weighting the training data and adding the second - order derivative of velocity to the input variables. The results show that for larger filter widths, there is a high correlation between the neural network predictions and the exact sub - grid - scale stresses; when the filter widths are \(48.8\eta\) and \(97.4\eta\) respectively, the correlation coefficients are approximately 0.9 and 0.8 respectively, where \(\eta\) is the Kolmogorov scale. In addition, the study also finds that the model established by the neural network is close to but not exactly equivalent to the gradient model. The researchers also conducted LES experiments to verify the performance of the neural network model in homogeneous isotropic turbulence and the Taylor - Green vortex initial - value problem. The results show that the results of the neural network model are basically consistent with those of the filtered DNS, but in the latter problem, due to the lack of symmetry under strict orthogonal transformation in the neural network model, symmetry is broken.