Fast and accurate prediction of airflow and drag force for duct ventilation using wall-modeled large-eddy simulation

Abstract Rectangular ducts are commonly adopted in ventilation systems to push airflow through pipes, with Reynolds number reaching up to 500000. Computational cost for flows with such a Reynolds number is prohibitive. Moreover, accurate prediction of airflow and drag in duct flows is important for ventilation duct design to remove dust particles and reduce turbulence, which will be beneficial for pipes cleaning and energy consumption. Hence, this work proposed a fast and accurate technique to simulate high-Reynolds-number turbulent flows in a square duct, using improved wall-modeled large-eddy simulation (LES) models. The improved wall-modeled LES approach features an improved wall-stress model and a self-adaptive subgrid-scale (SGS) model, which can dynamically adjust the mixing length to better predict wall-bounded flows. A new filtering strategy (filtering in time and space) is proposed as well. Wall-modeled LES are then conducted for the duct flows at a Reynolds number of 250000 with y + of the first point (away from the wall) larger than 300. Existing experiments were applied for validations. In the result, the computational time is greatly reduced (over 90%) compared with wall-resolved LES. The improved wall-modeled LES shows a better prediction of wall friction (drag) with acceptable mean velocity profiles. Moreover, the self-adaptive model captures the acceleration of the mean streamwise velocity near the corner but not for the conventional Smagorinsky approach. The methodology developed in this study provides a fast and accurate wall-modeled LES approach for the simulation of high-Reynolds-numbers ventilation duct flows.