中文摘要：

平均 Reynolds 的数字模拟(RANS ) 的预兆的能力当模特儿被在蜿蜒地流的开的隧道斜槽模仿流动并且把获得的结果与测量数据作比较调查。二个实验的流动结构是不同的以便得到更好的卓见。二个旋涡粘性骚乱模型和不同的墙处理方法被测试。比较证明没有必要差别在预言之中存在。骚乱模型的差别有有限效果，并且近的墙精炼稍微改进预言。结果证明当纵的速度很好通常被预言时，第二等的流动的预兆的能力被流动结构的复杂性大部分决定。在情况 1 一简单流动结构，第二等的流动速度相当被预言。在情况 2，由突然地弄弯的连续颠倒组成弯曲，流动结构在第一拐弯，和复杂流动结构以后变得复杂导致第二等的流动的差的预言。分析证明骚乱 anisotropy 的高水平与边界层分离，然而并非与在毫无疑问引起 RANS 模型的差的预言的中央区域的流动结构复杂性是相关的。骚乱模型修正和墙处理方法恰好在模仿复杂流动结构改进 RANS 模型的预兆的能力。

英文摘要：

The predictive capability of Reynolds-averaged numerical simulation （RANS） models is investigated by simulating the flow in meandering open channel flumes and comparing the obtained results with the measured data. The flow structures of the two experiments are much different in order to get better insights. Two eddy viscosity turbulence models and different wall treatment methods are tested. Comparisons show that no essential difference exists among the predictions. The difference of turbulence models has a limited effect, and the near wall refinement improves the predictions slightly. Results show that, while the longitudinal velo- cities are generally well predicted, the predictive capability of the secondary flow is largely determined by the complexity of the flow structure. In Case 1 of a simple flow structure, the secondary flow velocity is reasonably predicted. In Case 2, consisting of sharp curved consecutive reverse bends, the flow structure becomes complex after the first bend, and the complex flow structure leads to the poor prediction of the secondary flow. The analysis shows that the high level of turbulence anisotropy is related with the boundary layer separation, but not with the flow structure complexity in the central area which definitely causes the poor prediction of RANS models. The turbulence model modifications and the wall treatment methods barely improve the predictive capability of RANS models in simulating complex flow structures.