Direct numerical simulation of turbulent pipe flow using the lattice Boltzmann method
Direct numerical simulation of turbulent pipe flow using the lattice Boltzmann method
复制标题
DOI：
10.1016/j.jcp.2017.11.040
复制
发表时间：
2018-03
期刊：
影响因子：
--
通讯作者：
Cheng Peng;N. Geneva;Zhaoli Guo;Lian-Ping Wang
中科院分区：
文献类型：
--
作者：
Cheng Peng;N. Geneva;Zhaoli Guo;Lian-Ping Wang
本文采用介观晶格玻尔兹曼方法（LBM）在D3Q19晶格网格和D3Q27晶格网格上首次对湍流管道流动进行了直接数值模拟。以前从未报道过使用LBM对湍流管道流动进行DNS，可能是由于以前在弯曲固体表面下使用LBM实现的不准确性和数值稳定性。事实上，甚至有人推测D3Q19晶格可能不适合作为湍流管道流动的DNS工具。在本文中，我们表明，通过仔细的实现，可以使用D3Q19和D3Q27晶格网格获得准确的湍流统计。在D3Q19晶格的模拟中，暴露了与模拟数值稳定性有关的几个问题。对这些问题进行了讨论并提出了解决办法。另一方面，D3Q27晶格的模拟比D3Q19晶格更稳定。在求解Navier-Stokes方程的基础上，将得到的摩擦雷诺数Re τ= 180时的湍流统计数据与已发表的实验结果和其他DNS结果进行了系统的比较。比较内容包括平均流量、平均速度和平均涡度、平均压力和平均压力、速度偏度和平整度、速度和涡度的空间相关性和能量谱。总的来说，我们得出结论，D3Q19和D3Q27模拟都产生了准确的湍流统计数据。使用D3Q27晶格可以抑制由于数值伪影导致的平均流中的弱二次流型。
In this paper, we present a first direct numerical simulation (DNS) of a turbulent pipe flow using the mesoscopic lattice Boltzmann method (LBM) on both a D3Q19 lattice grid and a D3Q27 lattice grid. DNS of turbulent pipe flows using LBM has never been reported previously, perhaps due to inaccuracy and numerical stability associated with the previous implementations of LBM in the presence of a curved solid surface. In fact, it was even speculated that the D3Q19 lattice might be inappropriate as a DNS tool for turbulent pipe flows. In this paper, we show, through careful implementation, accurate turbulent statistics can be obtained using both D3Q19 and D3Q27 lattice grids. In the simulation with D3Q19 lattice, a few problems related to the numerical stability of the simulation are exposed. Discussions and solutions for those problems are provided. The simulation with D3Q27 lattice, on the other hand, is found to be more stable than its D3Q19 counterpart. The resulting turbulent flow statistics at a friction Reynolds number of Re τ= 180 are compared systematically with both published experimental and other DNS results based on solving the Navier–Stokes equations. The comparisons cover the mean-flow profile, the rms velocity and vorticity profiles, the mean and rms pressure profiles, the velocity skewness and flatness, and spatial correlations and energy spectra of velocity and vorticity. Overall, we conclude that both D3Q19 and D3Q27 simulations yield accurate turbulent flow statistics. The use of the D3Q27 lattice is shown to suppress the weak secondary flow pattern in the mean flow due to numerical artifacts.