Large eddy simulations of forced heat convection in a pin-fin array with a priori examination of an eddy-viscosity turbulence model

Abstract

Reynolds-averaged Navier–Stokes (RANS) modeling remains the predominant technique for engineering simulations of turbulent flow and heat transfer. Most simulations employ the linear-eddy viscosity assumption to model the Reynolds stresses, which is known to produce inaccurate results in complex turbulent flows. The objective of this study is to provide new insight on the inadequacy of LEV models in complex engineering turbulence by performing an a priori examination of a typical LEV model, the k−ω shear-stress transport (SST) model, for flow predictions in a pin-fin array. First, LES results obtained from simulations with three levels of mesh resolution are presented and validated against published experimental data. Second, an a priori examination of the SST model is performed, considering discrepancies between the shape and orientation of the Reynolds stress tensors obtained from the SST model and the LES. The results illustrate that the LES predicted Reynolds stresses have significantly different characteristics in regions of developing versus fully developed turbulence, and in the bulk-flow versus near-wall regions. These differences are not reflected in the Reynolds stresses predicted by the SST model, which predicts very similar Reynolds stress shapes and orientations in these different flow regimes. The analysis provides some insight on correcting the shape of the Reynolds stress tensor in particular; this information will be used in future work to examine practical methods to quantify turbulence model form uncertainties.