基于直接数值模拟的可压缩湍流模型评估和改进

ASSESSMENT OF THE COMPRESSIBLE TURBULENCE MODEL BY USING THE DNS DATA

对来流Mach数2.25和6的平板边界层湍流进行了直接数值模拟,并通过与理论、实验及他人计算结果的对比对数值结果进行了验证.基于直接数值模拟得到的湍流数据库,对常用的湍流模型进行了先验评估.评估的湍流模型有k-ε模型(包括标准k-ε模型、可实现的k-ε模型及低Reynolds数k-ε模型)、SA模型及BL模型.结果显示,对于Mach 2.25的平板边界层,可实现的k-ε模型及低Reynolds数k-ε模型具有较好的预测能力,而标准k-ε模型预测的湍流黏性系数偏高;SA模型在边界层内层预测准确度较高,而在外层预测值偏高.而对于Mach 6的平板边界层,k-ε模型及SA模型预测的湍流黏性系数均偏高,尤其是标准k-ε模型.对于Mach 6的平板边界层,BL模型低估了内-外层交界位置,造成湍流黏性系数预测值严重偏低.作者通过修改模型系数及内-外层交界位置对BL模型进行了修改,修改后模型预测的湍流黏性系数与DNS给出的值吻合较好.

Direct numerical simulation （DNS） of turbulent flat-plate boundary layer flows with free-stream Mach numbers of 2.25 and 6 are performed, and the data are validated by compared with theoretical, experimental and numerical results. Based on the DNS data, turbulence models （including κ-ε model, SA model and BL model） are assessed. For Much 2.25 case, numerical results show that the realizable κ-ε model agrees the DNS data very well and the turbulent viscous coefficient of the classical κ-ε model is higher than that of the DNS data. SA model gives quite good turbulent viscous in the inner layer of the boundary-layer, but it gives higher turbulent viscous in the outer layer. For high Mach number case, κ-ε models and SA model overestimate turbulent viscous coefficient. In Much 6 case, BL model cannot give good interface location between the inner and the outer layer. To improve the BL model, the authors modify the coefficients, and the new model＇s viscous coefficient agrees well with the DNS data.