LES study of turbulent flow fields over hilly terrains — Comparisons of inflow turbulence generation methods and SGS models

Abstract

The performance of inflow turbulence generation methods and SGS models are crucial for accurate prediction of turbulent flow fields over hilly terrains using large eddy simulation (LES). It is thus necessary to evaluate the capacity of different inflow turbulence generation methods and SGS models. In this study, two types of inflow turbulence generation methods, wind tunnel replication (WTR) method and consistent discrete random flow generation (CDRFG) method, are used to simulate the approaching flows of turbulent flow fields over simplified topographies, including a smooth two-dimensional (2-D) ridge and a smooth three-dimensional (3-D) hill. The advantages and limitations of aforementioned methods are discussed in the context of flow patterns and characteristics of turbulent flow. The ability of three different subgrid-scale (SGS) models is systematically investigated: (a) standard Smagorinsky model (SM), (b) coherent structure Smagorinsky model (CSM) and (c) dynamic Lagrangian Smagorinsky model (DLSM). The numerical results obtained from the CSM and DLSM are in relatively poor agreement with those of the experiment, which is mainly caused by inaccurate estimation of the characteristics of turbulent flow near the ground surface of approaching flow. The size of the recirculation bubble predicted by the CSM and DLSM are smaller than that of the SM for both 2-D ridge and 3-D hill. It is found that the closed-wake region is formed behind the 2-D ridge, while the open-wake region is established on the leeward side of 3-D hill due to the existence of the secondary flow. The open-wake region is well reproduced by the SM rather than the CSM and DLSM. The non-Gaussian characteristics of wake flow and the topography-driven twist effects are also examined.