Computational Fluid Dynamics of the flow in a diffuser : - like geometry

Abstract: Simulations were performed to investigate flow separation of an asymmetricdiffuser - like geometry. The geometry used for the simulations was modeledafter an experimental setup with recorded flow data, which was compared tothe simulated data. For all simulations, steady state flow at the inlet was usedwith the assumption of a 2D flow.A grid convergence study consisting of three different grids was performed.From this study no apparent change in simulation results were observed forfiner grids. This is caused by the fact that the coarse grid had a high enoughresolution to fully capture the flow, meaning that the higher resolution gridsyielded small improvements.Additionally, two different turbulence models RN G k − ε and SST k − ωwere used for evaluating which model was best suited to model flow separation.The simulations showed that the RN G k − ε model could not capture the flowseparation and had a poor accuracy when predicting the turbulent kinetic energy(TKE). Simulation results from SST k − ω gave good results in capturing flowseparation and predicting both the velocity and TKE when compared to theexperimental data.Finally, a turbulence intensity study was made for the mid grid with theSST k − ω model. The turbulent intensity was set to 5%, 10%, 15% and 20%at the inlet. This resulted in the point of separation moving further down thegeometry to x/H ≈ [17.68, 18.71, 19.58, 20.72] for respective intensity. The pointof reattachment also moves to x/H ≈ [44.85, 43.60, 42.67, 41.67] for respectiveintensity.In summary for simulating flow separation in turbulent flows the SST k − ωmodel is optimal and an increase in turbulent intensity reduces the recirculationzone.