A grid-adaptive simulation model for turbulent flow predictions

Abstract

Hybrid Reynolds-averaged Navier–Stokes (RANS) and large eddy simulation (LES) methods, abbreviated as hybrid RANS-LES, have been rapidly developed and increasingly used for predicting complex turbulent flows. In this study, a new high-fidelity hybrid RANS-LES strategy that modifies the turbulent viscosity equation using the ratio of grid length scale to turbulent integral length scale based on the Kolmogorov energy spectrum, termed the grid-adaptive simulation (GAS) model, is proposed to achieve high accuracy for turbulent flows using different grid resolutions. Using the shear-stress transport (SST) k–ω model as the baseline turbulence model, the GAS-SST model is validated by predicting three typical turbulent flows with coarse and fine meshes, including periodic hill flow, circular cylinder flow, and simplified tip leakage flow. As a reference, the scale-adaptive simulation (SAS) and delayed detached-eddy simulation (DDES) models are also employed to predict the above three turbulent flows. Solutions of GAS-SST, SAS-SST, and DDES-SST are compared against the high-fidelity data from the experiments or LES solutions. Detailed comparisons show that the GAS-SST model could achieve high accuracy with different grid resolutions for all three validation cases, which means that the GAS model has strong grid-adaptive ability. The results predicted by the GAS-SST model using coarse meshes are usually much more in agreement with the high-fidelity data than those predicted by SAS-SST and DDES-SST models. The GAS model demonstrates the potential to address the accuracy and computational efficiency requirements for predicting turbulent flows.