The ability of FLUENT to simulate base flows with and without ejection was assessed using two test cases: a duct containing a backwards-facing step and a NACA 0012 symmetric aerofoil modified to include trailing edge coolant ejection.
The standard k-ε, RNG k-ε, Reynolds Stress, Spalart-Allmaras and low Reynolds number k-ε turbulence models were used to simulate the backwards-facing step geometry. Constant velocity and turbulence profiles, experimental velocity profiles with constant turbulence profiles and experimental velocity and turbulence profiles were used as inlet boundary conditions. Six Reynolds numbers ranging from 5000 to 64000 were modeled. The numerical solution to the backwards-facing step geometry was compared to experimental velocity and turbulence intensity profiles at various locations relative to the step. The Launder-Sharma low Reynolds number k-ε model, RNG k-ε model and Reynolds Stress model each performed well within certain range of Reynolds numbers. No single turbulence model performed well over the entire range of Reynolds numbers. It was noted that the velocity profiles in the inlet region and downstream of the step agreed well with experimental results when the turbulence intensity profiles also agreed well with the measurements. Hence, the ability of CFD to simulate a turbulent flow is strongly dependent on its ability to capture the effects of turbulence.
The simulations of the isolated aerofoil, with trailing-edge coolant ejection, were run to match experiments in which the free stream Reynolds number was 0.2 ×105 based on trailing edge thickness. The numerical solution was compared to experimental velocity profiles at half a chord downstream of the trailing edge for eight blowing rates. The RNG k-ε and Reynolds stress turbulence models as well as an entirely laminar viscous solution procedures were used. A satisfactory level of grid independence was not achieved for any of the simulations. This is attributed partly to the grid generation algorithm and problems associated with further grid refinements. The lack of grid independence introduces significant uncertainty into the assessment of the models performance. The models did not accurately predict the mixing of the coolant and free stream flows downstream of the aerofoil nor did they capture the asymmetry evident in the experimental measurements. The influence of increasing the distance between the outlet and the trailing edge, the assumption of chord wise coolant ejection and the sensitivity of the solution to the free stream and coolant turbulence inlet boundary conditions have not been investigated.