Non-equilibrium turbulent boundary layers in high reynolds number flow at incompressible conditions: effects of streamline curvature and three dimensionality

The physics and computational prediction of turbulent boundary layer flow over axisymmetric and three-dimensional bodies are examined. Three cases were considered for which extensive experimental results and companion Reynolds-averaged Navier Stokes (RANS) solutions were obtained and/or available in the open literature. These cases all have Reynolds numbers based upon the freestream velocity and body geometric scale on the order of 105 to 106, which is large for laboratory scales but small compared to the maximum scales observed for full-scale vehicles. Despite significant differences in approach flow fields and geometries for these three cases, some common themes emerged in the findings. All cases involved complications due to pressure gradients combined with streamwise curvature, and all exhibited regions of turbulence reduction due to accelerated flow. These complications led to discrepancies in computed results even in attached flow regions where it is often assumed that RANS models provide reliable predictions. The authors recommend further work on modelling approaches that can capture rapid distortion effects on turbulence transport that can be incorporated into industry-useful frameworks. Two cases involving laterally symmetric, three-dimensional wall-mounted hills with aft-body separation revealed that asymmetric mean flow fields are likely to result. This finding has been observed in experiments conducted in multiple facilities and in computations using multiple solvers and turbulence models. It is concluded that non-unique and asymmetric global flow solutions are fundamental to flow cases with lateral geometric symmetry involving turbulent boundary layer separation. Further work is also needed to accurately predict low-frequency unsteadiness due to geometries that produce non-unique mean flow fields. For such flows, it remains to be definitively determined whether experimentally observed modes of the mean flow are equivalent, or nearly equivalent, to asymmetric mean flow solutions obtained using RANS approaches.

本研究针对轴对称及三维物体表面湍流边界层流动的物理机制与计算预测方法展开系统探究。本次研究选取三个案例，其相关的大量实验结果及配套的雷诺平均纳维-斯托克斯（Reynolds-averaged Navier Stokes, RANS）数值求解结果均已通过公开学术文献获取或可从公开渠道获得。上述三个案例的雷诺数均以来流速度与物体几何特征尺度为基准，量级处于10^5至10^6之间；该量级在实验室尺度下属于较大值，但相较于实车尺度下观测到的最大流动尺度仍偏小。尽管这三个案例的来流流场与几何构型存在显著差异，但其研究结果仍呈现出若干共性特征：所有案例均存在由压力梯度与流向曲率共同引发的流动复杂性，且均出现了因流动加速导致湍流抑制的区域。即便在通常认为RANS模型可提供可靠预测的附着流区域，上述流动复杂性仍导致了计算结果与实际值的偏差。研究团队建议进一步开发能够捕捉湍流输运快速畸变效应的建模方法，并将其整合至工业可用的计算框架中。针对两个带有后体分离的横向对称三维贴体山丘构型案例的研究显示，其流动可能产生非对称的平均流场。该结论已在多套实验装置开展的实验以及采用多种求解器与湍流模型的数值计算中得到验证。综上，对于涉及湍流边界层分离的横向几何对称流动案例，非唯一且非对称的全局流动解是其固有特性。针对此类会产生非唯一平均流场的构型，仍需开展进一步研究以精准预测其低频非定常流动特性。针对此类流动，目前仍需明确：实验中观测到的平均流模态与采用RANS方法得到的非对称平均流解是否一致或近似一致。