OCR Text |
Show characteristics of the flame. The numerical simulations were performed using FLUENT V5. V5 is a fully unstructured general purpose CFD code using a collocated, finite volume scheme. For all simulations, the QUICK discretization scheme was used. Pressure-velocity coupling is handled using the SIMPLE algorithm. For all simulations, fuel and coflow air boundary conditions were set as 1/7,h powerlaw profiles. Atmospheric pressure boundaries were placed at the periphery. Standard wall functions were used for near-wall treatments. Consistent modeling of all other details was maintained between runs to isolate the effects of turbulence models. Solution convergence was accelerated using a variant of the multigrid procedure of Hutchinson and Raithby [17]. Simulations were performed in serial and parallel on a S U N ULTRA 60 workstation. SKE simulations were completed with the serial solver in approximately 5 hours. Nearly linear performance was seen using the V 5 parallel version. Simulation results Numerical simulations were performed using the SKE, R K E and R S M turbulence closure models with the equilibrium chemistry PDF combustion model. Sample temperature and velocity magnitude contours are shown in Figure 5. Radial profiles of mean axial velocity (u mean), axial velocity variance (u var), mean mixture fraction (f mean), mixture fraction rms (f rms) and mean temperature (temp mean) were compared at locations of 5, 20, 40, 60 and 80 fuel nozzle diameters downstream of the nozzle exit. Axial profiles were also examined. Figure 6 shows predictions for the mean axial velocity. The overall agreement is very good. At 5 diameters, the R K E and R SM models accurately predict the velocity profiles. The SKE model is seen to overpredict jet spreading. At 20 and 40 diameters, each model overpredicts the centerline mean velocity decay, with the R K E and R S M models showing slight improvements over the S K E model. By 60 diameters, the predicted axial velocity profiles begin to improve, with the RKE model actually overpredicting centerline velocities. The overprediction of the centerline velocity decay is due to the overly diffusive nature of the models (especially SKE). This trend appears to be compensated for at locations further downstream by buoyancy effects and the further evolution of large scale coherent Figure 5. Temperature (left) and velocity magnitude (right) contour plots for the SKE simulation. White denotes the maximum value (Tmax = 1750 K, Vmax = 42.7 m/s) and black the minimum {Tmr, = 300 K, Vmn = 0 m/s). |