Verification Issues Related to CFD Simulations of Flow Around Circular Cylinders

Abstract

The classical problem of flow around a stationary smooth circular cylinder is used to evaluate Computational Fluid Dynamics (CFD) transient simulations using two approaches; Body Fitted Grid (BFG) and Immersed Boundary Method (IBM). BFG simulations were performed using a commercial CFD code ANSYS-FLUENT and IBM simulations using an in-house CFD code DREAM. Two dimensional simulations were performed at three different Reynolds numbers; 1 × 103, 1 × 105, and 5 × 105. Each of the cases was simulated using a coarse, medium and fine mesh. CFD simulations were evaluated using the following quantities; drag coefficient, lift coefficient, pressure coefficient, separation angle and the Strouhal number of the first harmonic of the lift coefficient. Average, and amplitude of the evaluation quantities are reported for every case. Simulations showed the grid dependence of the results, e.g. finer meshes captured higher harmonics of the drag coefficient which coarse meshes smeared due the large numerical viscosity. IBM simulations were also affected by the symmetry of the computational grid. Predicted quantities follow previously reported experimental trends fairly well except in the critical flow regime. Two dimensional calculations using turbulence models were performed for the case of Re = 1 × 105, and Re = 5 × 105. Turbulent results showed the importance of the grid resolution near the cylinder wall in capturing the physics of the problem. Three dimensional calculations were also performed and results are compared to those obtained from the two dimensional simulations. As may be expected, discretization error estimation methods using three grid calculations are not satisfactory for this highly unsteady flow problem, especially near the critical regime, 1 × 105 < Re < 5 × 105. This paper dwells on various issues related to verification of calculations for such highly unsteady flows.