1698

Energy & Fuels 2008, 22, 16981706

Computational Fluid Dynamics for Simulation of Wind-Tunnel Experiments on Flare Combustion Systems
David Castieira and Thomas F. Edgar*
Department of Chemical Engineering, UniVersity of Texas at Austin, 1 UniVersity Station C0400, Austin, Texas 78712-0231 ReceiVed September 11, 2007. ReVised Manuscript ReceiVed January 10, 2008

Flaring is used extensively in the energy and petrochemical industries to dispose of unwanted combustion gases by burning them in an open ame. However, these units may represent an important source of gas emissions due to inefcient operation under certain conditions such as high crosswind velocities. Several experimental studies have previously focused on ames burning in a xed volume by using wind tunnels. In these experiments, the entire plume of combustion products was collected, sampled, and analyzed to calculate the combustion efciency. Present work simulates these wind-tunnel experiments by using the commercial computational uid dynamics (CFD) software package Fluent 6.2. Several three-dimensional (3D) computational models are developed, and suitable turbulence and chemistry models are applied to simulate the complex combustion phenomena and ame downwash. The computational work was greatly reduced by applying the laminar amelet model, which assumes that a turbulent ame is an ensemble of small laminar structures called amelets. Inefcient combustion is observed at high crosswinds, and simulation results are in very good agreement with experimental data. These results show that CFD can successfully simulate these wind-tunnel are experiments. The resulting simulation models could be used to estimate the hydrocarbon emissions from chemical and petrochemical ares at crosswind conditions, an environmental issue of great importance in air pollution models.

1. Introduction Industrial ares are combustion systems designed to safely dispose of waste gases from chemical and petrochemical plants. Theoretically, these units convert effectively all hydrocarbons to CO2 and H2O through high temperature oxidation reactions. When operating properly, industrial ares achieve a 9899% combustion efciency, so only a few percent of unburned hydrocarbons would be released to the atmosphere under these conditions. Unfortunately, there are many situations in which this high combustion efciency may be compromised. For example, steam-assisted ares are designed to provide smokeless combustion by adding steam into the combustion zone. However, some experimental studies have shown that addition of steam into the are may affect the resulting combustion efciency.1,2 While adding turbulence and promoting benecial chemical interactions with the carbon particles, excessive steam addition can decrease the temperature of the ame to the point where inefciency becomes a concern. Another variable that affects ame destruction efciency is the wind speed surrounding the are. Flames at crosswind conditions may exhibit a stripping away and dilution of part of the fuel stream before it encounters a source of ignition to begin its reaction.3 Moreover, crosswind reduces the ame size,
* To whom correspondence should be addressed. Telephone: (512) 4713080. Fax: (512) 471-7060. E-mail: edgar@che.utexas.edu. (1) McDaniel, M. Flare Efciency Study; EPA-600/2-83-052; July 1983. (2) Pohl, J. H.; Soelberg, N. R. EValuation of the Efciency of Industrial Flares: H2S Gas Mixtures and Pilot Assisted Flares; EPA-600/2-86-080; September, 1986. (3) Johnson, M. R.; Kostiuk, L. W. Combust. Flame 2000, 123, 189 200.

resulting in lower combustion efciency because less oxygen is entrained into smaller ames, and are combustion efciency may decrease rapidly as wind speed increases from 1 to 6 m/s. Experimental evidence suggests that are combustion efciencies typically may be in the range of 70% at low wind speeds and could be even lower at higher wind speeds. Unfortunately, there is still great uncertainty about are efciency and the resulting emissions. Published research on large-scale jet diffusion ames burning in an open atmosphere is limited.4,5 However, collecting representative samples from ames burning in an open atmosphere is very difcult. In addition, online measurements on these systems is complicated by the size and turbulence of the ames, which are typically located at the tip of stacks anywhere from 10 to over 100 m tall to prevent dangerous conditions at ground level. For all these reasons, two different approaches have been proposed to analyze the effect of crosswind velocity on industrial are performance and the resulting emissions. One of these approaches involves the experimental study of reduced scale, turbulent diffusion ames located within a wind-tunnel facility. Typically, a closed-circuit wind tunnel is used so samples can be easily collected and analyzed. The second approach to understand industrial are behavior is to use computational uid dynamics (CFD). CFD is based on the application of fundamental physics for the prediction of reacting ow phenomena, and it relies on the numerical solution
(4) Kuipers, E. W.; Jarvis, B.; Bullman, S. J.; Cook, D. K.; McHugh, D. R. Combustion efciency of natural gas ares; effect of wind speed, ow rate, and pilots; Internal report; Shell Research and Technology Thorton and British Gas Research Centre, 1996. (5) Strosher, M. InVestigations of Flare Gas Emissions in Alberta; Environmental Technologies, Alberta Research Council: Calgary, Alberta, Canada, 1996.

10.1021/ef700545j CCC: $40.75 2008 American Chemical Society Published on Web 04/18/2008

CFD on Flare Combustion Systems

Energy & Fuels, Vol. 22, No. 3, 2008 1699

of the governing transport equations for mass, energy, species, and momentum. Even though research has been done on the subjects of CFD and industrial combustion, there is very little discussion of industrial ares, except by Baukal et al.,6 which discusses CFD applications in industrial combustion. CFD has been recently applied to simulate the effect of steam addition and air addition for several laboratory-scale, turbulent nonpremixed ames.7 Detailed nite-rate chemistry models were applied to predict the species concentrations in the ame, while species mass balances were set up in order to compute the resulting ame combustion efciency. Simulation results showed that incomplete combustion of hydrocarbons may occur at high steam/fuel and air/fuel ratios up to the point where these ames become extinguished. The computational work was greatly reduced by assuming two-dimensional (2D) axisymmetric models. Reynolds numbers of these laboratory scale ames were comparable to those for industrial ares. CFD has also been used to develop 3D simulations of the effect of crosswind on a laboratory-scale, turbulent combustion ame.8 The ame was simulated at the exit of a vertical burner that was perpendicular to the air ow, a conguration that is relevant to continuous gas aring in the atmosphere. Simulations clearly showed that for high momentum ames moderate velocities may signicantly reduce the resulting combustion efciency. Unfortunately, direct application of CFD to simulate largescale industrial ares is very difcult. First of all, industrial ares are clearly turbulent, and direct numerical simulation of turbulent ows is not possible because of the wide range of time and length scales. Thus, some type of turbulence model must be applied. Second, realistic chemical mechanisms for hydrocarbon combustion cannot be described by a single reaction equation. Such models may include tens of species and hundreds of reactions that are known in detail for only a limited number of fuels. Hence, some chemistry simplication must be made. Furthermore, it is necessary to deal with complex turbulencechemistry interaction due to the sensitivity of reaction rates to local changes. It is very challenging to perform combustion simulations for large-scale ares. The total number of grids needed to capture all the combustion details makes the computational work almost prohibitive for large ares. Employing three-dimensional (3D) models signicantly increases the computational work. Even so, the lack of experimental data for industrial ares makes it very difcult to validate potential simulation results. Thus, CFD is restricted to the simulation of wind-tunnel experiments in this work, in order to compare model predictions with experimental data. This allows us to validate our results by direct comparison with experimental data. Validated simulation models could be used to estimate the actual hydrocarbon emissions from chemical and petrochemical plants. The commercial software Fluent 6.2 is used in this work. As a rst approach, we have selected the wind-tunnel experiments of low-momentum jet diffusion ames of Johnson and Kostiuk,3 where natural gas was burned at crosswind velocities ranging from 1.0 to 11.0 m/s in a 0.0221 m diameter burner. In addition, a new set of wind-tunnel experiments that have been experimentally studied at CANMET Energy Technology Centre, Ottawa, are used for simulation of larger scale ares. These
(6) Baukal, C. E.; Gershtein, V. Y.; Li, X. Computational Fluid Dynamics in Industrial Combustion; CRC Press LLC: Boca Raton, FL, 2000. (7) Castieira, D.; Edgar, T. F. Energy Fuels 2006, 20, 10441056. (8) Castieira, D.; Edgar, T. F. CFD for simulation of crosswind on the efciency of high momentum jet turbulent combustion ames. J. EnViron. Eng., submitted for publication.

new experiments are characterized by larger burner diameters, in some cases comparable to industrial are diameters. 2. Governing Equations CFD relies on solving conservation or transport equations for mass, momentum, energy, and participating species. If the ow is turbulent, model equations for specic turbulent quantities have to be solved in addition. Since even with todays super computers resolving turbulent length scales directly results in tremendous effort, Reynolds averaged equations are typically applied to include the physics of turbulence. Hence, the basic model equations for a uid in turbulent ow are the Reynoldsaveraged NavierStokes (RANS) equations. For steady state, these equations are given below: 2.1. Continuity Equation. (F) ) 0 (1) where F is the density of the uid and is its ensemble-averaged j velocity vector, dened on a 3D domain. 2.2. Momentum Conservation Equation. (F) ) - p + (( + ()T) - F) (2)

where is the turbulent uctuation of the velocity vector, is the dynamic molecular viscosity of the uid, and F is the pressure. The overbars denote mean values. The Reynolds stresses, F, are extra terms that stem from decomposing jj solution turbulent variables into the mean and uctuating components; these terms must be modeled in order to close eq 2. A common approach employs the Boussinesq hypothesis to relate the Reynolds stresses to the mean velocity gradients: - ) t

i j 2 + - (t() + k)ij xj xi 3

(3)

where the Einstein summation notation is being used; that is, ij is the Krnecker delta, t is the eddy kinematic viscosity, and k is the kinetic energy of turbulence, dened by 1 (4) k ) ii 2 The Boussinesq approach is used in the k- model, which is the turbulence model applied in this work. The advantage of this approach is the relatively low computational cost associated with the computation of the kinematic viscosity. For the k- model,theeddyviscosityisobtainedfromthePrandtl-Kolmogorov relation: t ) Ck2 (5)

where is the rate of turbulent kinetic energy dissipation. Robustness, economy, and reasonable accuracy for a wide range of turbulent ows explain the popularity of this model in industrial ow and heat transfer simulations. 2.3. Energy Conservation Equation. When considering heat transfer within the uid and/or solid regions of the problem, Fluent also solves the energy equation. This equation is given below in a very general form: (FE) + ((FE + p)) ) (keff T b t J h b + (
i i j

b effV)) + Sh (6)

where keff is the effective conductivity, Ji is the diffusion ux of species i, and h is enthalpy. Notice that radiation effects are

1700 Energy & Fuels, Vol. 22, No. 3, 2008

Castieira and Edgar

Table 1. Cases Used for Simulation and Corresponding Jet Exit Velocity (Vj), Crosswind Velocity (U), and Experimental Fuel Jet to Crosswind Momentum Flux Ratio (R)
case B C D E F G Vj (m/s) 2.08 2.10 2.09 2.11 2.11 2.09 U (m/s) 1.33 2.76 4.09 5.49 8.27 11.05 R 1.43 0.34 0.15 0.085 0.038 0.021

Figure 1. Schematic of a closed-loop wind tunnel facility (all dimension in meters) at the University of Alberta. Reprinted by permission of Elsevier Science from Johnson and Kostiuk,3 Copyright 2000 by The Combustion Institute.

Figure 2. Sketch of ow structures in a low-momentum jet diffusion ame in a crosswind.

not considered in this work, so the corresponding term has been removed from eq 6. 2.4. Species Transport Equations. Finally, for reacting systems the species transport equations must be solved. In general form, this equation is given by (FY ) + (FYi) ) - Ji + Ri t i (7)

where Ri is the net rate of production of species i by chemical reaction. Fluent applies the nite volume method to discretize and solve the governing ow equations described above. 3. Wind Tunnel Conguration A set of low-momentum, natural gas diffusion ames located in a closed-loop wind tunnel were used for simulation. These ames were studied experimentally at the University of Alberta and the National Research Council of Canada. Measurement of experimental combustion efciencies was reported by Johnson and Kostiuk.3 A detailed description of the experimental setup and results was given as a nal report by Kostiuk et al.9 A schematic representation of the wind tunnel is given in Figure 1. The basic information about the experimental setup is given below; refer to refs 3 and 9 for a more detailed description. The experimental ames were established at the exit of a burner tube mounted vertically in the wind tunnel and perpendicular to the airow. In the vertical section downstream of the ame and in the upper section of the tunnel, a series of
(9) Kostiuk, L.; Johnson, M.; Thomas, G. Flare research project, nal report, University of Alberta, September, 2004.

supplementary fans were used to ensure that the plume of combustion products was fully mixed into the wind-tunnel air before sampling. The wind tunnel was sufciently large that, during a typical 5-10 min test, the concentration of hydrocarbons in the tunnel remained small, and the effects of reburning were completely negligible. From a simulation point of view, only the test section around the burner (e.g., the box that followed the contraction section) is strictly relevant. The dimensions of this box are 2.44 m in width by 1.22 m in height by 11.8 m in length. However, a 5 m long box is enough to capture the ame behavior, so unnecessary computation work could be avoided by reducing the simulation domain. The oor of the wind tunnel was constructed with 19 mm thick plywood, while downstream of the are the tunnel was covered with 30 gauge aluminum sheeting to protect it from possible direct ame impingement. The walls along the test section were primarily Plexiglas. The ceiling upstream of the are was constructed with 19 mm thick plywood, but downstream of the are the ceiling was made of 19 mm thick ceramic panels that could safely resist the accidental impingement of the ame or hot combustion products. The diffusion ames were established at the exit of a 24.6 mm o.d. (22.1 mm i.d.) pipe that extended 47 cm into the wind tunnel. The experimental setup also included a 65% blockage ratio perforated plate turbulence plug with 3 mm diameter holes, which was placed inside the pipe three diameters upstream of the exit. The purpose of this plug was to create velocity proles similar to the turbulent pipe ow expected in full-scale industrial ares, independent of the actual ow velocity in the laboratory-scale ares. However, computer simulation of that perforated plate is very difcult to perform and may introduce numerical errors in the solution process. Thus, the turbulence of the fuel gas was adjusted in our simulation to match the experimental turbulence intensity measured 5 mm above the exit plane of the burner tube (see Kostiuk et al.9). For these experiments, the jet exit velocity of the fuel, Vj, was held approximately constant at 2 m/s, and the crosswind speed, U, was varied from 1 to 11 m/s. The external cold-ow Reynolds number (Re) ranged from 1570 to 17 270 as the crosswind increased from 1 to 11 m/s. Under these conditions, the ow regime on the outside of the pipe are could be considered to be in the regime of having a laminar boundary layer separation. The turbulent uctuation in the core ow of the tunnel was found to be consistently less than 0.4% except at low wind speeds (<2 m/s), where the intensity rises to about 1.8%. The fuel gas used was sales grade natural gas (95.2% CH4, 2.1% C2H6, 1.7% N2, 0.8% CO2, and 0.2% other, by volume). In order to guarantee ignition, the experiments used a retractable hydrogen jet diffusion ame that was previously ignited by using a manual high-voltage spark system. Once the hydrogen ame was correctly positioned, the ow of are gas was easily ignited. After the are was ignited, the ow of hydrogen was turned off and the ignition system was retracted. Ignition can be

CFD on Flare Combustion Systems

Energy & Fuels, Vol. 22, No. 3, 2008 1701

Figure 3. Schematic representation of the middle plane used to plot results.

Figure 4. Long-exposure photographs of the experimental ame (left side) and simulation results for the temperature contours along the middle plane (right side). Moving downward, the crosswind velocity is increased from 1.33 m/s (case B) to 11.05 m/s (case G). Reprinted by permission of Elsevier Science from Johnson and Kostiuk,3 Copyright 2000 by The Combustion Institute.

simulated in Fluent 6.2 by patching a high-temperature region near the burner tip, which helps to switch on the combustion reactions in the same way as a spark helps ignites an actual ame. 4. Simulation Results For this study, the commercial software Fluent 6.2 was used for simulation. Gambit 2.0 was used to create the mesh and the

grid. The nal 3D grid had 321 040 cells and 990 713 faces, and it was successfully checked for skewness. The simulations were performed by using the 3D segregated solver incorporated in Fluent. Implicit formulation was chosen; only steady state solutions were obtained. A second order upwind scheme was used to solve all equations. The SIMPLE algorithm10 was used
(10) Patankar, S. V. Numerical Heat Transfer and Fluid Flow; McGraw Hill: New York, 1980.

1702 Energy & Fuels, Vol. 22, No. 3, 2008

Castieira and Edgar

Figure 5. Simulation results for the contours of CO2 mass fraction at the middle plane.

Figure 6. Simulation of the three zones for a ame under a crosswind (case E).

to couple pressure and velocities. Radiation effects were not considered. Proper boundary conditions were specied. Burner walls and wind-tunnel walls were dened as walls in Fluent. For the gas inlet surface and the wind inlet surface, the velocity inlet boundary condition was applied. For the wind outlet surface, the pressure outlet boundary condition was used, and this pressure was 1 atm. In order to compute the combustion efciency of a particular ame, a material balance was set up for the ame domain to compute the mass ux of the important species. Boundary cells at the wind outlet surface were rened to correctly calculate the integral values in that zone. The laminar amelet combustion model was applied because it greatly reduces the computational work of other complex models such as the eddy dissipation concept model (EDC) or

the probability density function model (composition-PDF), while still keeping the necessary degree of accuracy. Moreover, the most difcult part of modeling these particular ames arises from the turbulence ow patterns rather than very complex chemistry reactions, so the laminar amelet model seems very appropriate in this case. By using a detailed chemical mechanism, laminar opposed-ow diffusion amelets were calculated in Fluent. The chemical mechanism implemented in this work was GRI 3.0,11 which contained 325 reactions and 53 species for natural gas combustion simulation. The laminar amelets were then embedded in a turbulent ame by using statistical PDF methods. At very low jet-to-wind momentum ux ratios, the jet uid issuing from the burner tip may be severely deected by the transverse stream. Under these conditions, a signicant downwash may occur as a portion of the combusting gases verge on being drawn into the low-pressure region on the downwind side of the stack.12 In fact, at high crosswind, industrial ares are observed to present a standing vortex on the downstream side of the burner tube. Notice that ame downwash creates an ignition source that helps to stabilize the ame, so this phenomenon is not necessarily associated with very low combustion efciencies, as discussed later in this paper. A more detailed explanation for the fuel stripping mechanism under these conditions is given by Johnson et al.13 A sketch of the common ow structures at low-momentum jet diffusion ames in a crosswind is shown in Figure 2. This gure illustrates the three regions that are typically identied in these types of ames.14 Simulation of ame downwash is complicated by the aerodynamic interactions of the transverse air ow, burner tube, and deected fuel. Moreover, ame downwash is associated with boundary layer formation in the burner tube.15 That is why the k- turbulence model by itself is not capable of predicting this phenomenon. More sophisticated models such as large eddy simulation (LES) can solve the ne details of the ow, but LES is computationally very expensive, even on pc clusters, and it might be infeasible for practical applications. Fluent 6.2 offers two approaches to simulate the boundarylayer formation attached to the wall. In one approach, the viscosity-affected inner region is not resolved. Instead, semiempirical formulas called wall functions are used to bridge the viscosity-affected region between the wall and the fully-turbulent region. In another approach, the turbulence models are modied to enable the viscosity-affected region to be resolved with a mesh all the way to the wall, including the viscous sublayer. The latter is called enhanced wall-treatment in Fluent, and it is the option applied to this work. In any case, a very ne mesh near the burner tube must be dened to correctly simulate boundary layer formation. The cells attached to the burner walls in this work were 0.5 mm long with a growth rate of 1.1, which was shown to be sufcient to capture the ame downwash phenomenon. The realizable k- model was selected. More information about the simulation models can be found in ref 16.
(11) Smith, G. P.; Golden, D. M.; Frenklach, M.; Moriarty, N. W.; Eiteneer, B.; Goldenberg, M.; Bowman, C. T.; Hanson, R. K.; Song, S.; Gardiner, W. C.; Lissianski, V. V.; Qin, Z. http://www.me.berkeley.edu/ gri_mech/, 2005. (12) Huang, R. F.; Chang, J. M. Combust. Flame 1994, 98, 267272. (13) Johnson, M. R.; Wilson, D. J.; Kostiuk, L. W. Combust. Sci. Technol. 2001, 169, 155174. (14) Gollahalli, S. R.; Brzustowski, T. A.; Sullivan, H. F. Trans. Can. Soc. Mech. Eng. 1975, 3, 205214. (15) Huang, R. F.; Wang, S. M. Combust. Flame 1999, 117, 5977.

CFD on Flare Combustion Systems

Energy & Fuels, Vol. 22, No. 3, 2008 1703

Figure 7. Velocity vector eld in the middle plane for cases B and G. Table 2. Computed Species Mass Flows (10-5 kg/s) and Resultant Combustion Efciencies for the Simulated Casesa
case B C D E F G
a

(CH4)in 49.947 49.947 49.947 49.947 49.947 49.947

(CO2)in 1.343 1.343 1.343 1.343 1.343 1.343

(C2H6)in 2.063 2.063 2.063 2.063 2.063 2.063

(CH4)out 0.00 0.00 0.00 0.00 0.00 0.00

(CO)

out

(CO2)out 142.66 145.18 146.72 141.23 137.94 132.38

(C2H6)out 0.00 0.00 0.00 0.00 0.00 0.00

c (%) 98.5 100.3 101.4 97.5 95.3 91.4

1.130 0.003 0.002 0.002 2.541 7.317

Cases C and D predict efciencies slightly over 100% due to numerical errors.

Table 3. Specication of the Three CANMET Wind-Tunnel Experiments18

case 1 2 3 run label FTF030305_CE5 FTF030408_CE2 FTF030414_CE5 total mass pipe diameter pipe length fuel (kg/h) (in.) (in.) 15 15 10 1 4 6 24 36 36 c (%) 97.54 99.78 98.71

Figure 8. Comparison between experimental data and simulation results for combustion efciency at increasing crosswind velocities. Experimental data is obtained from Johnson and Kostiuk.3

Six different cases for natural gas combustion are simulated in this section. They are listed in Table 1, as identied by Johnson and Kostiuk.3 This table shows for each case the jet exit velocity (Vj), the crosswind velocity (U), and the experimental fuel jet to crosswind momentum ux ratio (R), dened as R) FjetVj FairU2 (8)

Notice that the temperature contours can be used as a good estimator of the photographs (e.g., high temperature can be associated with high luminosity), although they do not necessarily match completely. It is worth noting that the experimental ames, especially at low crosswinds, are slightly more vertical than the simulated ames. This visual discrepancy is likely due to buoyancy effects, which were not accounted for in the simulation work. Furthermore, these simulated temperature contours represent only the steady state solution of the problem, so in reality, turbulent uctuations would eventually affect the shape and position of the ames. Finally, experimental results show that, as the crosswind velocity is increased, downwash begins to occur as a portion of the combustion gases verge on being drawn into the low-pressure region on the leeward side of the stack. This phenomenon is also observed in the simulation results. Simulation results for contours of CO2 mass fraction are given in Figure 5. Clearly the amount of CO2 produced in the ame is affected by increasing wind speed, which in turn should affect
Table 4. Air Conditions at the CANMET Wind-Tunnel Experiments
case 1 2 3 P (kPa) 100.7 103.1 102.3 T (C) -5.8 -1.8 15.1 wind speed (m/s) 9.34 5.27 9.97 relative humidity (%) 70 76 29

In this paper, 2D contour plots are given at the middle plane of the wind tunnel, which is illustrated in Figure 3. In Figure 4 (left side), a series of long-exposure photographs of the overall position and size of the ame for the experiments are shown. On the right side of this gure, the corresponding simulation results for the contour plots of temperature are shown.

1704 Energy & Fuels, Vol. 22, No. 3, 2008

Castieira and Edgar

Figure 9. Representation of the meshed model box for case 3, top view.

Figure 10. Contours of temperature (left side) and CO2 mass fraction (right side) at the middle plane for the CANMET wind-tunnel experiments.

the resulting combustion efciency as shown in the next section. While the combustion efciency (c) can be dened in a number of ways, in this work combustion efciencies focus on the fully oxidized combustion products with the goal of completely oxidizing all of the fuel to CO2 (Bourguignon et al.17). Equation 9 shows how (c) is calculated: c ) mass flow rate of carbon in CO2 produced by flame mass flow rate of carbon in CxHy in the fuel gas stream (9)

(16) Castieira, D. Three-dimensional computational uid dynamics modeling of combustion ares. Ph.D. Thesis, Department of Chemical Engineering at the University of Texas, Austin, TX, 2006. (17) Bourguignon, E.; Johnson, M. R.; Kostiuk, L. W. Combust. Flame 1999, 119, 319334.

For both experimental ames and simulations, the ame structures described in Figure 2 become apparent. In fact, three different zones are observed in Figure 6, which corresponds to case E in Figure 4. A planar stationary vortex attached to the burner tube denes the rst zone, the long axisymmetric tail of the ame forms the third zone, and the junction that connects these two main parts of the ame denes the second zone. The size of the axisymmetric tail is reduced as the crosswind velocity increases, while the recirculation zone grows down the tube as more ow is drawn into the low-pressure region. Ultimately, at very high crosswinds, the main tail of the ame would be extinguished and only the recirculating vortex would remain. A better understanding of the downwash phenomenon can be obtained by analyzing the simulated velocity vectors in the middle plane. Figure 7 displays these velocity vectors, which indicate the magnitude and direction of the ow velocity at each point near the burner wall for cases B and G. A recirculation zone near the burner exit is clearly observed, so the ow turns backward in this zone. This recirculation zone enlarges as the crosswind velocity increases from 1 to 11 m/s. Hence, at high crosswind velocities, a large portion of the fuel is drawn downward. Thus, the shapes and positions of the ames observed in these simulations give a qualitative estimation of the ame behavior under a crosswind. However, it is important to compute the resulting combustion efciencies and to compare them to the experimental measurement. Table 2 shows the computed mass ows by Fluent at inlet and outlet boundaries for several species involved in the chemistry of the ame. The last column computes the combustion efciency based on eq 9. A graphical comparison between the experimental and simulated combustion efciencies is shown in Figure 8. As observed in Figure 8, the experimental data and simulation results are in very good agreement, considering the complexity

CFD on Flare Combustion Systems

Energy & Fuels, Vol. 22, No. 3, 2008 1705

Figure 11. Velocity vectors and recirculating region at the middle plane for the CANMET wind-tunnel simulations. Table 5. Computed Species Mass Flows (10-5 kg/s) and Resultant Combustion Efciency for the CANMET Wind-Tunnel Simulations
case 1 2 3 (CH4)in 380.229 380.229 253.471 (CO2)in 6.782 6.782 4.522 (C2H6)in 15.696 15.696 10.464 (CH4)out 0.002 0.000 0.000 (CO)out 19.183 0.000 0.000 (CO2)out 1075.988 1095.620 726.718 (C2H6)out 0.000 0.000 0.000 c experimental (%) 97.54 99.78 98.71 c simulation (%) 97.56 99.35 98.85

of these 3D simulations. Initially, the efciency is slightly higher because more air is entrained into the ame. However, as the crosswind velocity grows larger than 5.49 m/s (case E), the combustion efciency decreases and degrades signicantly with further increases in the wind speed (cases F and G). Both experiments and simulations conrm these results. Notice that simulated combustion efciencies exhibit the same trend as the experimental data, with a maximum efciency located around 4 m/s crosswind velocity. Mismatches between simulations and experiments are on the order of 1-2%, so these CFD models successfully simulate the wind-tunnel experiments. A ame bent over with the windward side of the ame detached from the stack or a ame base trapped in the recirculating ow in the wake of the stack is not sufcient to cause signicant inefciencies, which was also pointed out by Johnson and Kostiuk.3 In fact, this recirculating zone seems to help stabilize the ame so the resulting combustion efciency is higher than expected at high crosswinds. For high momentum ames where ame downwash does not occur, the combustion efciency may be more sensitive to high crosswinds, as shown by Castieira and Edgar.8 5. Larger-Scale Wind-Tunnel Simulation In this section, we apply CFD for the simulation of a set of wind-tunnel experiments that have been experimentally studied at the CANMET Energy Technology Centre, Ottawa. These experiments are part of a more extensive study of are combustion systems under crosswind conditions in closed-loop wind tunnels, but only three particular cases are considered here for simulation. The experimental setup was similar to the wind-tunnel experiments of Johnson and Kostiuk,3 so only relevant differences will be given here. First of all, the are test facility in this case was not closed loop, and the methodology for measuring efciency was different. Second, the basic dimensions for the box section (i.e., the section after the contraction of Figure 1) were 8.230 m in length by 1.219 m in width by 1.82 m in height. Natural gas was used as a fuel, and its composition was (by volume) 95.33% CH4, 2.1% C2H6, 0.13% C3H8, 1.8%

N2, 0.62% CO2, and 0.02% C4H10. Table 3 describes some physical parameters of the three ames studied in this work, including the case number, the run label, the total mass of fuel sent to the burner, the nominal pipe diameter (Schedule 40), and the pipe length. The last column lists the experimental combustion efciencies measured by the authors of the experiment. The variable air conditions in the wind tunnel are given in Table 4, including pressure, temperature, wind speed, and relative humidity. As observed in Table 3, the pipe dimensions for cases 2 and 3 are relatively large (the pipe diameter for case 3 is comparable to some industrial are diameters), which makes their computer simulation more challenging than that of the previous cases. The simulation procedure employed was essentially the same as that discussed. However, larger simulation domains have to be dened for cases 2 and 3 due to their larger scale. Hence, the length of the simulation box was set as 5.0, 6.0, and 7.0 m for cases 1, 2, and 3, respectively. This required a larger number of cells in the domain. Moreover, the grid concentration around the burner was increased for cases 2 and 3 in order to capture in detail the physics of these ames. A schematic representation of the grid model from a top view is presented in Figure 9 for case 3. The nal number of grid cells for cases 1, 2, and 3 were 255 520, 566 720, and 988 960 cells, respectively. The applied models, parameters, and procedures applied in Fluent were very similar to previous simulations. Only boundary conditions differ signicantly in order to t the new physical conditions for fuel and air. The simulation time increased signicantly due to the large number of cells, especially for case 3. Convergence was achieved after progressively increasing the under-relaxation factors for density from 0.7 to 1.0. Figure 10 shows the contour plots for temperature (left side) and CO2 (right side) at the middle plane of the simulation box. Unfortunately, photographs are not available for these experimental ames. A signicant downwash can be observed for
(18) Gogolek, P. Personal communication. 2005.

1706 Energy & Fuels, Vol. 22, No. 3, 2008

Castieira and Edgar

cases 2 and 3, which is likely due to the high crosswind velocity and the larger burner surface where the ame is attached. The velocity vectors for the three corresponding simulations are displayed in Figure 11. As expected, these vectors reect how the ow moves downward and becomes trapped in the recirculation zone. For case 3, most of the ow exiting the burner moves to this zone. Table 5 shows the computed mass ows by Fluent at inlet and outlet boundaries for several species involved in the chemistry of the ame. The last two columns show the experimental combustion efciencies (c) based on eq 9 and the simulated combustion efciencies (c) obtained with Fluent. High combustion efciencies are observed in both experiments and simulations in spite of the high crosswinds. This means that most of the fuel is burned within the recirculating ow in the wake of the stack. Clearly, the agreement between the experimental results and the simulation data is excellent, and this agreement is even better than in previous wind-tunnel cases. The reason for this may be that turbulence boundary conditions were known for these CANMET wind-tunnel experiments. In fact, for the wind-tunnel simulations studied in section 4, it was necessary to add articial turbulence because of the perforated plate at the experimental setup, which may introduce some errors in the computations. 6. Conclusions CFD analysis has been successfully applied to simulate a set of closed-loop wind-tunnel experiments in order to understand industrial are behavior and potential gas emissions from these units. The laminar amelet model was shown to be accurate enough to capture the most important details of the ame. A thin boundary layer was dened to simulate ame downwash, a phenomenon associated with low-momentum ares under high crosswind. Results show that high crosswind velocities affect the resulting combustion efciency (in terms of total combustion to CO2). However, for low momentum ares (e.g., low gas/ crosswind ratios) such as these ones, a signicant amount of fuel may be trapped in the recirculating zone near the burner wall, which helps stabilize the ame and avoids fuel stripping away from the burner exit without burning. This makes the resulting combustion efciencies higher than expected. For highmomentum ares (e.g., high gas/crosswind ratios), the ame is unlikely to attach to the stack, so lower combustion efciencies are observed even at high crosswinds.8 Both experimental results and simulations conrm this.

We also simulated larger diameter ares under a crosswind that were part of a different set of wind-tunnel experiments. In this case, the computational work increased due to the larger simulation domain and larger number of cells. Agreement between the experimental results and the simulations was very good in all cases. Computational uid dynamics is an excellent tool to simulate ares burning in wind-tunnel facilities. Use of simulation in the study of are units yields economic savings compared to expensive wind-tunnel experiments. Moreover, CFD allows for the possibility of detailed analysis of species concentration proles and turbulent ow patterns within the ame, which may not be available experimentally.
Acknowledgment. The authors thank the Texas Air Research Center for its support of this research and Dr. P. Gogolek of the CANMET Energy Technology Centre, Ottawa, for providing the selected data from the Flare Test Facility. They also want to thank Dr. David Schowalter (Fluent) for his help and recommendations in implementing the software.

Nomenclature
A ) area projections over the faces of the zone D ) burner diameter h ) enthalpy bi ) diffusion ux of species i J k ) kinetic energy of turbulence Keff ) effective conductivity LK ) Kolmogorov length scale LR ) reaction length scale mi ) mass ow rate of species i through a boundary p ) pressure R ) momentum ux ratio Ri ) net rate of production of species i by chemical reaction Vj ) jet exit velocity U) crosswind velocity ) ensemble-averaged velocity vector j ) turbulent uctuation of the velocity vector x ) axial position xi ) mass fraction of species i ) species mass fraction ij ) Krnecker delta ) rate of turbulent kinetic energy dissipation c ) combustion efciency ) dynamic molecular viscosity of the uid F ) density of the uid t ) eddy kinematic viscosity
EF700545J