OCR Text |
Show Discussion/Conclusions • 0 <-' i M RKE HM EXP b/x "•** Figure 11. Predicted and measured jet spreading rates. momentum dominated regime for a variety of turbulent propane jet flames, consistent with other's observations in isothermal jets. Reasonable agreement with these observations is seen here. The experimental data shows a distinct momentum dominated regime in the jet far field. Because of the overpredicted decay of centerline velocity in the simulations, a constant M F D region is less evident. The influence of buoyancy (i.e., a linearly increasing MFD) can clearly be seen for downstream locations beyond the predicted transition position (£, = 2). In fact, the transition location and slope of the M FD profiles for both the experimental data and simulation results in this regime agree very well with the 0.5 slope observed in [6]. 2 - I SKE --P.KE • - -RSM • Eg Figure 12. Momentum flux density versus nondimensional downstream location for experimental data and numerical results. Particular attention has been paid in this investigation to the choice of turbulence model on simulation results. The numerical simulations performed here have provided validation of the SKE, RKE and RSM turbulence models for buoyant, turbulent jet diffusion flames over the range of data available. Overall agreement amongst the different models is good, with mean profiles demonstrating strong agreement with the experimental measurements. As expected, the R K E model shows improved jet spread predictions over the SKE model in the momentum dominated regime. The second order closure RSM showed slight improvements over the SKE model in several instances as well. Another important aspect to consider is computational cost. The S K E model is the least expensive of the models examined. The R K E model requires only very slightly more effort than SKE. RSM requires additional memory and C P U time due to the increased number of transport equations solved (6). Though efficient programming in FLUENT minimizes the computational overhead, on average, the R S M in F L U E NT V5 still requires 50-60% more C P U time per iteration compared to the k-e models. Furthermore, 15-20% more memory is used. Based on these considerations, the RKE model appears to be the most accurate, cost-effective approach to turbulence modeling for practical engineering analyses of flare flames in the momentum dominated and transition regimes. Though the experimental measurements used in this investigation contained data into the transition region from momentum to buoyancy dominated flow, the full effects of buoyancy on the flow field, and the ability of R A N S turbulence models to accurately predict this plume regime of buoyant, turbulent jet flames, could not be accurately assessed. Authors have described the importance of large scale, coherent structures in nonreacting turbulent flows [18] as well as in the nature of flames in the buoyant regime [6,19]. Conventional approaches for turbulence modeling, like those examined here, involve Reynolds averaged transport equations with first or second order turbulence closure models. |