Retrospective cost adaptive Reynolds-averaged Navier-Stokes k-ω model for data-driven unsteady turbulent simulations
Retrospective cost adaptive Reynolds-averaged Navier-Stokes k-ω model for data-driven unsteady turbulent simulations
复制标题
用于数据驱动的非定常湍流模拟的回顾性成本自适应雷诺平均纳维-斯托克斯 k-ω 模型
DOI：
10.1016/j.jcp.2017.11.037
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
S. Bailey
中科院分区：
文献类型：
--
作者：
Zhiyong Li;Jesse B. Hoagg;Alexandre Martin;S. Bailey
本文提出了一个数据驱动的计算模型，用于模拟非定常湍流，稀疏的测量数据是可用的。该模型采用回溯成本自适应（RCA）算法自动调整雷诺平均Navier-Stokes（RANS）k-ω湍流方程的封闭系数，以提高模拟流场与实测流场的一致性。RCA-RANS k-ω模型在定常流中使用管流试验用例进行验证，在非定常流中使用表面安装立方体试验用例进行验证。用于适应验证案例的测量值是从具有已知闭合系数的基线模拟中获得的。这些验证测试用例表明，RCA-RANS k-ω模型可以成功地调整闭合系数，以提高模拟流场与一组稀疏流场测量值之间的一致性。此外，RCA-RANS k-ω模型改善了模拟流量与不存在测量值的位置处的基线流量之间的一致性。RCA-RANS k-ω模型还通过两种试验情况的试验数据进行了验证：定常管流和非定常方柱绕流。在这两种测试情况下，与使用k-ω闭合系数标准值的非自适应RANS k-ω模型的结果相比，自适应提高了与实验数据的一致性。对于稳定的管流，自适应驱动的平均流向速度测量在24个位置沿着管道半径。RCA-RANS k-ω模型将这些位置的平均速度误差降低了35%以上。对于方柱上的非定常流，自适应由方柱上2个位置处的时变表面压力测量驱动。RCA-RANS k-ω模型将这些位置的平均表面压力误差降低了88.8%。
This paper presents a data-driven computational model for simulating unsteady turbulent flows, where sparse measurement data is available. The model uses the retrospective cost adaptation (RCA) algorithm to automatically adjust the closure coefficients of the Reynolds-averaged Navier–Stokes (RANS) k–ω turbulence equations to improve agreement between the simulated flow and the measurements. The RCA-RANS k–ω model is verified for steady flow using a pipe-flow test case and for unsteady flow using a surface-mounted-cube test case. Measurements used for adaptation of the verification cases are obtained from baseline simulations with known closure coefficients. These verification test cases demonstrate that the RCA-RANS k–ω model can successfully adapt the closure coefficients to improve agreement between the simulated flow field and a set of sparse flow-field measurements. Furthermore, the RCA-RANS k–ω model improves agreement between the simulated flow and the baseline flow at locations at which measurements do not exist. The RCA-RANS k–ω model is also validated with experimental data from 2 test cases: steady pipe flow, and unsteady flow past a square cylinder. In both test cases, the adaptation improves agreement with experimental data in comparison to the results from a non-adaptive RANS k–ω model that uses the standard values of the k–ω closure coefficients. For the steady pipe flow, adaptation is driven by mean stream-wise velocity measurements at 24 locations along the pipe radius. The RCA-RANS k–ω model reduces the average velocity error at these locations by over 35%. For the unsteady flow over a square cylinder, adaptation is driven by time-varying surface pressure measurements at 2 locations on the square cylinder. The RCA-RANS k–ω model reduces the average surface-pressure error at these locations by 88.8%.