Simulations of vitiated bluff body stabilized flames
Smith, Andrew Gerard
MetadataShow full item record
Bluff bodies have a wide range of applications where low-cost, light weight methods are needed to stabilize flames in high-speed flow. The principles of bluff body flame stabilization are straightforward, but many details are not understood; this is especially true in vitiated environments where measurements are difficult to obtain. Most work has focused on premixed flames but changing application requirements are now driving studies on non-premixed gaseous and spray flames. This thesis aims to improve the understanding of vitiated, bluff body stabilized flames, specifically on non-premixed, spray flames, through the use of Large Eddy Simulation (LES). The single flameholder facility at Georgia Tech was chosen as the basis for the simulations in this thesis. The flameholder was a rectangular bluff body with an aerodynamic leading edge with discrete liquid fuel injectors embedded just upstream of the trailing edge in a configuration described as “close-coupled.” The liquid phase was modeled using a Lagrangian particle approach where discrete fuel droplets were injected into the domain. Experimental data was used to tune model parameters as well as the stripped droplet velocities and sizes. The discharge coefficient needed to be taken into account to achieve the correct fuel jet penetration. The experiments were conducted over a range of global equivalence ratios; lean equivalence ratios, φ global ≈ 0.5, exhibited symmetric flame shedding and conversely large scale sinusoidal B ́ernard/von-K ́arm ́an shedding was observed when the equiva- lence ratio was near unity. Reacting flow LES were computed at these two fuel flow rates to improve understanding of the different flame dynamics. LES were first com- pleted using a quasi-laminar subgrid turbulence-chemistry interaction model. Span- wise averaging of instantaneous and time-averaged LES results were compared with experimental high- and low-speed imaging and showed the LES was in qualitative agreement at both fuel flow rates. At phi_global ≈ 0.5, the fuel jet did not penetrate as far into the crossflow compared to phi_global ≈ 0.95 and thus more fuel was delivered to the shear layers of the bluff body resulting in higher heat release in the shear layers for the low fuel flow rate. The heat release damped the large sinusoidal structures via gas expansion and baroclinic torque generation. Higher fuel jet penetration in the phi_global ≈ 0.95 case meant less fuel was delivered to the shear layers and so less heat release occurred directly behind the bluff body so the large scale sinusoidal shedding was not damped. The impact of the subgrid turbulence-chemistry interaction model on the flame dynamics was tested by comparing the quasi-laminar LES with LES using the subgrid linear eddy model (LEMLES). The flame structure predicted with LEMLES matched that of the quasi-laminar LES, at both fuel flow rates in the near- field behind the bluff body but deviated farther downstream. A flame edge analysis showed little sensitivity to the choice of subgrid model in the region x < 4D. A high-order hybrid finite-difference solver with consisting of a WENO upwind method and compact central scheme was implemented to assess the effects of the numerical method. A series of test cases was used to verify, validate and compare several of the available spatial and temporal methods before the high fuel flow rate bluff body case was run. For the simple test cases the higher-order methods were clearly more efficient but for more complex cases the differences between the second- order and high-order methods are smaller. To test the hypothesis that the fuel jet penetration was the main factor in the flame dynamics another configuration with a modified fuel injector diameter was simulated. The injector size was chosen to match the spray penetration of phi_global ≈ 0.5 case while maintaining the fuel flow rate of the phi_global ≈ 0.95 case. The results confirmed the hypothesis as the flame dynamics of this configuration match the original low fuel flow rate case.