A preliminary investigation of Large Eddy Simulation (LES) of the flow around a cylinder at ReD = 3900 using a commercial CFD code
Engineering fluid mechanics simulations at high Reynolds numbers have traditionally been performed using the Reynolds-Averaged Navier Stokes (RANS) equations and a turbulence model. The RANS methodology has well-documented shortcomings in the modeling of separated or bluff body wake flows that are characterized by unsteady vortex shedding. The resulting turbulence statistics are strongly influenced by the detailed structure and dynamics of the large eddies, which are poorly captured using RANS models (Rodi 1997; Krishnan et al. 2004). The Large Eddy Simulation (LES) methodology offers the potential to more accurately simulate these flows as it resolves the large-scale unsteady motions and entails modeling of only the smallest-scale turbulence structures. Commercial computational fluid dynamics products are beginning to offer LES capability, allowing practicing engineers an opportunity to apply this turbulence modeling technique to much wider array of problems than in dedicated research codes. Here, we present a preliminary evaluation of the LES capability in the commercial CFD solver StarCD by simulating the flow around a cylinder at a Reynolds number based on the cylinder diameter, D, of 3900 using the constant coefficient Smagorinsky LES model. The results are compared to both the experimental and computational results provided in Kravchenko & Moin (2000). We find that StarCD provides predictions of lift and drag coefficients that are within 15% of the experimental values. Reasonable agreement is obtained between the time-averaged velocity statistics and the published data. The differences in these metrics may be due to the use of a truncated domain in the spanwise direction and the short time-averaging period used for the statistics presented here. The instantaneous flow field visualizations show a coarser, larger-scale structure than the study of Kravchenko & Moin (2000), which may be a product of the LES implementation or of the domain and resolution used. Based on this preliminary study, we conclude that StarCD's LES implementation may useful for low Reynolds number LES computations if proper care is used in the problem and mesh definition.
- Research Organization:
- Lawrence Livermore National Lab. (LLNL), Livermore, CA (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 877787
- Report Number(s):
- UCRL-TR-219145; TRN: US200608%%689
- Country of Publication:
- United States
- Language:
- English
Similar Records
High-resolution time-resolved experiments on mixing and entrainment of buoyant jets in stratified environments
Combustion Modeling for Direct Fired Supercritical CO2 Power Cycles