Investigations into the Impact of the Equivalence Ratio on Turbulent

Mar 24, 2016 - significantly influence the flow field in the wake region, the stabilization location of the flame, and the flame intensity. Specifical...
0 downloads 0 Views 9MB Size
Article pubs.acs.org/EF

Investigations into the Impact of the Equivalence Ratio on Turbulent Premixed Combustion Using Particle Image Velocimetry and Large Eddy Simulation Techniques: “V” and “M” Flame Configurations in a Swirl Combustor Gaurav Kewlani, Santosh Shanbhogue, and Ahmed Ghoniem* Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States ABSTRACT: Turbulent premixed combustion is studied using experiments and numerical simulations in an acoustically uncoupled cylindrical sudden-expansion swirl combustor, and the impact of the equivalence ratio on the flame−flow characteristics is analyzed. In order to numerically capture the inherent unsteadiness exhibited in the flow, the large eddy simulation (LES) technique based on the artificial flame thickening combustion model is employed. The experimental data are obtained using particle image velocimetry. It is observed that changes in heat loading, in the presence of wall confinement, significantly influence the flow field in the wake region, the stabilization location of the flame, and the flame intensity. Specifically, increasing the equivalence ratio drastically reduces the average inner recirculation zone size and causes transition of the flame macrostructure from the “V” configuration to the “M” configuration. In other words, while the flame stabilizes along the inner shear layer for the V flame, a persistent diffuse reaction zone is also manifested along the outer shear layer for the M flame. The average chemiluminescence intensity increases in the case of the M flame macrostructure, while the axial span of the reaction zone within the combustion chamber decreases. The predictions of the numerical approach resemble the experimental observations, suggesting that the LES framework can be an effective tool for examining the effect of heat loading on flame−flow interactions and the mechanism of transition of the flame macrostructure with a corresponding change in the equivalence ratio. the equivalence ratio.9,10 At high but lean equivalence ratios, the flame is unstable and oscillates strongly, while at intermediate equivalence ratios, weakly oscillating quasi-stable flames are observed. Near the lean blowout limit, long, stable flames are formed. Therefore, significant progress has been made in understanding the influence of varying the equivalence ratio on flame−flow interactions using experimental and numerical approaches. However, it is important to continue the research efforts in order to further investigate the fundamental physics underlying the flame−flow interactions as the heat loading is changed in combustor configurations, beginning with the analysis of acoustically uncoupled systems. Previously, researchers have attempted to study the impact of modifying the equivalence ratio on the dynamic flow features for wake-stabilized flames. Experiments corresponding to an axisymmetric bluff-body system11 have shown that increasing the equivalence ratio dramatically decreases the size of the recirculation zone, while the turbulent kinetic energy is somewhat suppressed as a result of dilatation. Combustion near the lean extinction limit, on the other hand, tends to result in rapid growth of temperature fluctuations. In the case of a swirl combustor, changes in the flame stabilization location and vortex breakdown bubble topology12−14 with increasing heat release rate have been identified. The present work focuses on a laboratory-scale cylindrical sudden-expansion swirl combustor configuration, and investigates the influence of the change in

I. INTRODUCTION Modern low-emission gas turbine systems often employ swirl injectors that produce a central recirculation zone to serve as the dominant flame stabilization mechanism. As a result, extensive efforts have been made by the research community to understand the dynamics of swirling flows and the stability of premixed swirl-stabilized combustion.1−3 For example, turbulent swirling flows injected into a coaxial dump chamber at different swirl numbers have been studied by large eddy simulation (LES) to gain insight into several salient features of swirling flows (including vortex breakdown and shear-layer instability), and the results have been validated against experimental data in terms of mean flow velocity and turbulence properties.4,5 Numerical studies have also been performed to investigate the combustion dynamics in swirlstabilized combustors, and the physical processes responsible for driving combustion instabilities (including the coupling between acoustic wave motions, vortex shedding, and flame oscillations) have been identified.6,7 In another study, the unsteady flame dynamics and transition of the flame structure from a stable state to an unstable state in a lean-premixed swirlstabilized combustor were analyzed. The inlet temperature and equivalence ratio were identified as important variables determining the stability characteristics of the combustor.8 It is now well-known that changes in the equivalence ratio can significantly influence flame stability, and this has been an active area of research. For example, recent experiments for an acoustically coupled backward-step combustor and for a cylindrical sudden-expansion swirl combustor have shown distinct dynamic regimes based on the operating range for © XXXX American Chemical Society

Received: December 14, 2015 Revised: March 10, 2016

A

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels equivalence ratio on the flame characteristics and the flow features, using state-of-the-art experimental (PIV) as well as numerical (LES) techniques. The next section provides a brief description of the combustor configuration and the particle image velocimetry (PIV) setup, which is followed by an overview of the combustion model and the simulation setup. The data obtained from the experimental PIV system and numerical LES code corresponding to the “V” and “M” flame macrostructures are then presented. It is observed that varying the equivalence ratio can appreciably influence the large-scale coherent flow features as well as the dynamics of the flow. Furthermore, close agreement between the numerical predictions and the experimental measurements is observed for the flame−flow features corresponding to each flame configuration.

III. NUMERICAL CONFIGURATION LES techniques with appropriate subgrid-scale (SGS) turbulence models, combustion models, and reaction mechanisms are considered as one of the promising numerical simulation approaches, balancing computational complexity and predictive accuracy. While direct numerical simulations resolve all of the turbulence length and time scales, they are computationally expensive and impractical for high Reynolds number, large-scale applications. Solving the Reynoldsaveraged Navier−Stokes equations, on the other hand, models the influence of turbulence on the mean flow and hence cannot capture the unsteady flow. In LES, rather than averaging the effect of turbulence, the equations are filtered, enabling the larger scales of turbulence to be explicitly resolved while the effect of the smallest ones on the large scales is modeled. Resolving the large scales enables part of the inherent unsteadiness in the flow to be captured, which is particularly important when the dynamics on the large scales plays an important role, while modeling the SGS effects ensures that the approach is computationally manageable. Modeling of the filtered reaction rates in the species transport equations presents a major challenge in simulating turbulent premixed combustion using LES. The reaction rates in such scenarios are highly nonlinear functions of temperature and species mass fractions, and chemical reactions are confined to thin reacting layers on small scales that cannot be resolved on typical LES grids. As a consequence, the turbulence−chemistry interactions must be modeled, and a number of combustion models and approaches have been suggested,15,16 including the artificial flame thickening combustion model, which is discussed in the next section. Artificial Flame Thickening Approach. The artificial flame thickening technique, a finite-rate chemistry combustion modeling approach, essentially involves artificial thickening of the flame front, allowing it to be resolved on the LES grid, while maintaining the same laminar flame speed and turbulence−flame interaction. Increasing the flame thickness by a factor F while maintaining a constant flame speed can be achieved by suitably modifying the diffusivity (D) and the mean reaction rate (ω̅ ) by replacing D with FD and ω̅ with ω̅ /F). If F is sufficiently large, the thickened flame front can then be resolved on the LES computational grid.15 The filtered species transport equation can be written as

II. EXPERIMENTAL CONFIGURATION A canonical swirl geometry14 that is acoustically uncoupled is employed to investigate the isothermal flow and reacting flow phenomena. In this section, details about the combustor configuration and the corresponding PIV setup are provided. Experimental Configuration. The cylindrical combustor setup, shown in Figure 1, is designed to stabilize combustion using a

Figure 1. Experimental setup for premixed combustion in a swirl combustor.

̃ ∂(ρ ̅ Yu ∂ρ ̅ Yĩ ∂ ⎛⎜ ∂Y ̃ ⎞⎟ ωi̇ i j̃ ) ρ + = FD ̅ i ⎟+ ∂t ∂xj ∂xj ⎜⎝ ∂xj ⎠ F

combination of swirl and sudden expansion and comprises an inlet pipe with a diameter (Dinlet) of 38 mm and a combustion chamber with a radius (R) of 38 mm downstream of the expansion plane. A mixture of air and methane is introduced at a bulk inlet velocity (UCD) of 8.4 m/s (macroscopic Reynolds number (ReCD) of ∼20 000 based on the inlet pipe radius Dinlet/2 and the bulk fluid velocity UCD). The mixture inlet temperature (TC) is 300 K, and the nominal pressure (P) is 101 kPa. The swirler, located 50 mm upstream of the expansion plane, has eight blades, each inclined at an angle (θ) of 45° with respect to the cylinder cross section, with an estimated swirl number (S) of 0.7. The section downstream of the expansion plane, where the flame is anchored, consists of a 40 cm quartz tube for optical access. Particle Image Velocimetry System. PIV is widely used in experimental research to study the spatial flow features in turbulent nonreacting and reacting flows. The technique has distinct advantages over conventional point-velocity measurement approaches such as laser doppler velocimetry and hot wire anemometry since the entire flow field can be resolved to study the interaction between the flame surface and the flow. Furthermore, recent developments in high-speed lasers and cameras have also enabled high-repetition-rate measurements that allow the flow structures to be resolved over a wider range of time and length scales. In this work, two-dimensional (2D) velocity maps have been obtained using a high-speed CMOS camera mounted above the combustor, recording images at 1−2 kHz. The PIV measurements were processed using the LaVision DaVis 7.2 software. Further details about the setup can be found in ref 15.

(1)

where ρ̅ is the filtered density, Ỹ i is the species mass fraction, ũ is the filtered velocity vector, and ω̇i is the filtered species reaction rate. The thickening of the flame front, however, leads to a modified interaction between turbulence and chemistry since the Damkohler number is decreased by the factor F. The flame becomes less sensitive to turbulence, and wrinkling of the flame front is reduced. To account for this, an efficiency function (E) is introduced5 that recovers the underestimation of the flame front wrinkling due to the thickening approach. The balance equation for the chemical species then takes the form

̃ ∂(ρ ̅ Yu ∂ρ ̅ Yĩ Eωi̇ ∂ ⎛⎜ ∂Y ̃ ⎞ i j̃ ) ρ ̅ EFDi ⎟⎟ + + = ⎜ ∂t ∂xj ∂xj ⎝ ∂xj ⎠ F

(2)

An important aspect of the thickened flame approach is the evaluation of the efficiency function E in order to appropriately account for the reduced wrinkling of the thickened flame front due to turbulence, and different models have been proposed to define the efficiency function. Further details about the LES framework and the artificial flame thickening combustion modeling approach employed in this work can be found in refs 16 and 17. Simulation Setup. The choice of the numerical grid is governed by estimates of the physical length scales associated with the flow configuration. In order to capture the bulk of the energy-containing structures and to resolve at least 80% of the turbulent kinetic energy, a filter width (Δ) to integral length scale (LI) ratio of approximately B

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

intensity upstream of the swirler is reduced as a result of flow straightening between the choke plate and the swirler and also because of the presence of the swirler vanes, which have a dominating influence on the flow as it enters the combustion chamber. At the exit, zero Neumann conditions are specified for all of the variables except the pressure, for which wave-transmissive conditions are used. No-slip conditions are applied for the flow at the walls, and heat transfer at the walls is considered in order to account for any thermal losses by utilizing an enthalpy-based boundary condition and providing a suitable thermal diffusivity value; zero Neumann conditions are specified for the other variables. In order to maintain reasonable computational efficiency, appropriate wall functions are utilized to resolve the flow features in the wall boundary layer. Numerical computations start from quiescent conditions, and the unsteady flow characteristics evolve naturally. In order to initiate the flame, a hightemperature pulse is applied at the inlet section to ignite the fuel; the reacting mixture convects downstream and eventually stabilizes as a flame in the wake of the swirler. Averaging is performed over nearly 10 flow-through cycles once the flow is established in the computational domain. To allow suitable comparison between the numerical predictions and the experimental measurements and to examine the influence of heat release on the flow for the reacting flow scenarios, the ranges of the magnitudes of the contours of some of the flame−flow variables are fixed. Additionally, for clarity, the radial direction is represented as the y axis in the diametric cross-sectional plane for which the contours of the flame−flow variables are depicted.

0.083 should be maintained.18 Using the inlet pipe radius (Dinlet/2), which approximates the typical annular jet thickness, as the value of LI gives an estimate of 1.6 mm for Δ. Appropriate grid clustering may be required to resolve structures near the walls and in the shear layer region for reacting flow. Thus, the nonuniform mesh utilized is composed of 0.55 million hexahedral cells and gets coarser along the streamwise direction. The numbers of cells in the axial, radial, and tangential directions in the combustion chamber are 145, 35, and 96, respectively; therefore, the average Δx and Δr values in the wake region are approximately 1.35 and 1.00 mm respectively, while the minimum Δx and Δr values, in the shear layer region, are approximately 0.75 and 0.50 mm, respectively. Figure 2 depicts a

Figure 2. (a) CAD model of truncated combustion chamber showing the swirler geometry and sudden-expansion section. (b) LES mesh for numerical computations in the truncated combustor section.

IV. RESULTS AND DISCUSSION: NONREACTING FLOW The prominent flow features corresponding to the isothermal flow include the inner recirculation zone (IRZ), the precessing vortex layer surrounding the IRZ, shear layers at the boundaries of the annular jet, and the outer recirculation zones (ORZs) at the corners. These flow structures are also evident from the LES contours of the average vorticity magnitude (depicted in Figure 3a) along with the streamlines. As the flow evolves downstream of the expansion plane, strong shear layers develop as a result of the velocity difference between the annular jet and the surrounding fluid, and largescale structures are generated in the region, as indicated by the instantaneous out-of-plane vorticity contours (Figure 3b). Further downstream, these vortices break down into smaller eddies that eventually dissipate. One of the important flow characteristics in a swirl configuration is the vortex breakdown phenomenon, which manifests itself as an abrupt change in the core of a slender vortex and develops into an IRZ, as indicated by the mean streamlines in Figure 3a. As a result of the rapid flow expansion resulting from the strong swirling motion and of the geometric confinement, ORZs are also formed. High vorticity is observed in these outer recirculating vortex cells, as depicted in Figure 3a. Aside from leading to the formation of the ORZs, the presence

computer-aided design (CAD) model of a truncated section of the combustion chamber showing the swirler geometry and the suddenexpansion region along with the corresponding LES mesh employed for the numerical computations. The temporal resolution is determined on the basis of physical time scale estimates (in order to prevent excessive numerical dissipation and numerical instability) as well as on chemical time scales. For the nonreacting flow, a precessing vortex core (PVC) frequency (f PVC) of 120 Hz is reported in the experiments. On the basis of this observation, a PVC-based time scale (1/f PVC) of 8.3 ms can thus be estimated. On the basis of the Courant−Friedrichs−Lewy (CFL) criterion, Cmax = (UCD + Us)Δt/Δx (with Cmax = 1, Us = 340 m/s, and Δx = 1 mm), the time step for the simulations (Δt) is estimated to be approximately 2.2 μs. A conservative value of 1 μs is used in the reacting flow simulations to account for local refinement and acceleration of the fluid above the bulk inlet velocity and to adequately incorporate the chemical time scales. At the inlet, Dirichlet conditions are used for all of the variables except the pressure, for which zero Neumann conditions are specified. The inlet velocity (specified upstream of the swirler) is considered to have a flat profile on which random fluctuations of 5% turbulence intensity are imposed, as experimental data are available only downstream of the expansion plane, which is optically accessible. This is a suitable approximation because the free stream turbulence

Figure 3. (a) Average vorticity magnitude (s−1). (b) Instantaneous out-of-plane vorticity (s−1). The corresponding 2D velocity vectors and streamlines are also shown (LES). C

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 4. Average axial velocity along with the 2D velocity vectors and streamlines: (a) PIV; (b) LES. The cross sections at which velocity data are compared are also shown.

Figure 5. rms axial velocity along with the 2D velocity vectors and streamlines: (a) PIV; (b) LES.

Figure 6. Normalized profiles of (top) average axial velocity and (bottom) rms axial velocity: red circles, PIV; blue solid lines, LES.

of the chamber wall restricts the radially outward spread of the annular jet, resulting in a narrow IRZ. Another principal feature of swirling flows is the precessing vortex core, which develops when the central vortex core precesses around the symmetry axis at a specific frequency. For the scenario under consideration, the PVC frequency in the isothermal case is approximately 120 Hz, as estimated by sampling the radial velocity component downstream of the sudden expansion and performing a fast Fourier transform analysis of the signal.15 Figure 4 depicts the experimentally and numerically obtained contours of the average axial velocity along with the corresponding streamlines. The results show the presence of the vortex breakdown bubble (VBB), consisting of a single recirculating cell of fluid, forming the IRZ. Smaller vortices

(actually vortex rings) at the corners of the chamber are also evident, forming the ORZs. In Figure 5, the contours of the root-mean-square (rms) axial velocity field exhibit high-turbulence intensity in the annular region around the periphery of the IRZ. This is associated with the development of a strong shear layer in the region along with the existence of the PVC. Fluctuations are also high near the location where the annular jet impinges on the chamber walls. In the downstream section, the large-scale eddies subsequently break down into smaller structures, which dissipate through viscous effects, and the turbulence intensity of the flow field decreases. The normalized profiles for the average axial velocity and rms axial velocity obtained computationally and experimentally at cross sections downstream of the expansion plane of the D

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

V Flame Configuration. The Abel-transformed flame image for the V flame configuration, displayed in Figure 7(III), suggests the presence of a conical flame that stabilizes in the inner shear layer (ISL) and is anchored on the swirler (optically inaccessible and not shown). Ignition of incoming reactants takes place inside the inlet duct, and a thin, intense flame is observed in this region. While no reaction zone is observed in the interior of the IRZ, within the ORZ, or in the high-strain outer shear layer (OSL) region (Figure 8a), the flame front extends along the ISL downstream of the expansion plane, giving rise to a conical flame structure. The strongly burning flame envelops the IRZ and crosses the annular jet, continuing along the wall shear layer. In this downstream section, the average flame brush is observed to be thick with high CH chemiluminescence intensity. The reaction zone does not extend into the exhaust tube, indicating that combustion is complete before the flow exits the chamber. In Figure 8b, the contour of the average vorticity magnitude obtained from LES indicates significant vortex activity in the inner and outer shear layers due to the convective vortices and the combustion-induced baroclinic vorticity production. For the V flame scenario, the PVC is absent, which may be associated with the propagation of the flame into the inlet section.15 Further downstream, the breakdown of the vortices into smaller eddies due to heat release and thermal expansion in the vortex cores leads to a diffuse vorticity field. The small-scale eddies subsequently dissipate as a result of vortex stretching and viscosity effects as the flow convects toward the exhaust. In Figure 9, the contours of the average axial velocity obtained experimentally and numerically are displayed. The prominent flow features, including the annular jet, the shear layers at the boundaries of the jet, and the inner and outer recirculation zones, are suitably resolved in the simulations and closely resemble the flow structures from the experimental measurements; the topological flow difference from isothermal flow is principally in the structure of the IRZ. Examination of the shape of the experimental IRZ shows the presence of a toroidal, split-cell-type VBB. A secondary recirculating cell exists within the central core of the toroidal primary vortex structure, and the axial velocity along the centerline axis reaches a positive value as well. However, further examination of the flow field is required in order to elucidate the underlying physics behind the development of this interior region of flow. The numerical simulations do not predict this flow feature and indicate the presence of a single-cell-type IRZ with a high peak reverse velocity along the centerline, although the size and shape of the primary vortex is well-reproduced.

cylindrical chamber, are shown in Figure 6. The axial velocity and the flow fluctuations peak in the annular jet region that surrounds the IRZ. The flow turbulence decreases further downstream (x/R > 2) as the influence of the PVC and the vortices generated in the shear layer diminishes because of viscous dissipation. The LES predictions in the near-wall regions may be improved by appropriate mesh refinement and choice of a suitable wall function, while adjustments to the PIV setup for the measurement of higher-resolution flow fluctuation data at downstream locations and near the chamber walls may also be investigated.

V. RESULTS AND DISCUSSION: REACTING FLOW As depicted in Figure 7, the turbulent premixed flame in the swirl combustor exhibits different configurations depending on

Figure 7. Flame macrostructures in a swirl combustor (Abeltransformed flame images): (I) columnar wake flame (φ ≈ 0.51); (II) bubble-type wake flame (φ ≈ 0.55); (III) inner shear layer V flame (φ ≈ 0.60); (IV) inner shear layer/outer shear layer/outer recirculation zone M flame (φ ≈ 0.65).

the equivalence ratio. This is based upon whether the flame stabilizes near the front stagnation region of the VBB, within the IRZ (or near the rear stagnation region of the VBB), or in the low-velocity regimes of the inner and/or outer shear layers. Fundamental flow field characteristics, such as the structure of the inner and outer recirculation zones, the radial spread of the annular jet and its thickness, and the turbulent flow field intensity corresponding to these flame macrostructures, can differ substantially. In this work, the high but lean equivalence ratio scenarios (V and M flame configurations) are presented, describing the dominant flame stabilization mechanism, the spatial topological features for the mean and fluctuating flow fields in the combustion chamber, and pointwise velocity statistics at axial locations downstream of the expansion plane. It is observed that the numerical simulations adequately capture the heat release and flow features observed in the experiments and allow a closer examination of the mechanism of transition of the flame macrostructure with a corresponding change in the equivalence ratio.

Figure 8. (a) Magnitude of the strain rate tensor. (b) Average vorticity magnitude (s−1). The corresponding 2D velocity vectors and streamlines are also shown (LES). E

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 9. Average axial velocity along with the 2D velocity vectors and streamlines: (a) PIV; (b) LES. The cross sections at which velocity data are compared are also shown.

Figure 10. rms axial velocity along with the 2D velocity vectors and streamlines: (a) PIV; (b) LES.

Figure 11. Normalized profiles of (top) average axial velocity and (bottom) rms axial velocity: red circles, PIV; solid blue lines, LES.

The average flow streamlines suggest that a closed recirculating vortex structure is formed, which extends to a length of approximately 2.5R downstream of the expansion plane. The radial spread and the thickness of the annular jet are noted to be different from that observed experimentally for the isothermal flow. On average, the spreading angle of the streamlines is enhanced, corresponding with the broadening of the IRZ, and the incoming fresh mixture behaves as a thin annular jet spreading into the combustor. Furthermore, exothermicity and gaseous expansion in the region increases the axial jet velocity for the reacting flow. In Figure 10, the contours of the rms axial velocity fluctuations indicate moderate turbulence activity in the shear layers and in the region where the annular jet impinges the wall. Typically, for reacting flows involving high turbulent burning

velocity and heat loading, the enhanced heat release augments the magnitude of the shear layer turbulence as well as turbulent dilatation. For the V flame configuration, while the absence of the PVC and the effects of dilatation influence the fluctuating flow field immediately downstream of the expansion plane, the turbulence intensity in the region surrounding the IRZ and in the downstream section remains moderately high as a result of the generation of combustion-induced turbulence.19 Significant flow fluctuations are also observed in the region where a secondary shear layer is formed between hot products and the reactant mixture. The transient eddies formed in the mixing layer that convect downstream and wrinkle the flame cause flow distortions along with the combustion-generated turbulence. The turbulence intensity is also high near the rear stagnation zone where flow reversal takes place. F

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 12. (a) Instantaneous CO mass fraction, (b) Instantaneous normalized heat release. The corresponding 2D velocity vectors and streamlines are also shown (LES).

Figure 13. (a) Average normalized deconvolved CH chemiluminescence intensity. (b) Average normalized heat release (LES). The corresponding 2D velocity vectors and streamlines are also shown.

Figure 14. (a) Average temperature (K). (b) rms temperature (K). The corresponding 2D velocity vectors and streamlines are also shown (LES).

pipe, upstream flow perturbations, and the grid resolution in the boundary layer of the inlet duct.20 Additionally, the predictions in the near-wall regions may be improved by appropriate mesh refinement and choice of suitable wall functions, while adjustments to the PIV setup for the measurement of higher-resolution flow fluctuation data in downstream locations could also be investigated. Figure 12 depicts the instantaneous LES contours of the CO mass fraction and the normalized heat release. It has been suggested for a similar geometry that the flame holding features are sensitive to the thermal condition on the walls and strain computations in the OSL.21 Specifically, when adiabatic computations are performed or when the influence of strain is inadequately represented, the flame is predicted to stabilize in the OSL as well, even though the experimental measurements indicate an ISL flame only. The LES framework based on the artificial flame thickening combustion model incorporates the influence of strain and the effect of heat loss from the walls and predicts the flame macrostructure corresponding to the V flame configuration with reasonable accuracy. Figure 13 shows the contours of the average normalized deconvolved CH chemiluminescence intensity from experi-

Figure 11 shows the normalized profiles for the average axial velocity and rms axial velocity obtained from numerical simulations and the experimental measurements. The thickness and radial spread of the annular jet, as well as the recirculation zone topology, are resolved with reasonable accuracy using LES. However, while the experiments suggest a secondary vortex in the core of the IRZ, the simulations predict a high peak reverse velocity along the centerline axis, requiring further investigation. The turbulence intensity is observed to be slightly lower than the isothermal flow in the section immediately downstream of the expansion plane, which is attributed to the absence of the precessing vortex.15 The presence of the reaction zone in the upstream section and the associated heat release, along with lowering of adverse pressure gradients and the Reynolds number, slow the eddy development in the shear layer, thereby reducing the turbulence.19 Further downstream, flow fluctuations peak near the wall, where the impact of the annular jet with the wall generates small-scale turbulence. It is observed that the simulations predict marginally higher axial velocity fluctuations than are measured experimentally, and these discrepancies may be attributed to the sensitivity of the resolved flow features to geometric features in the inlet G

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 15. (a) Magnitude of the strain rate tensor. (b) Average vorticity magnitude (s−1). The corresponding 2D velocity vectors and streamlines are also shown (LES).

Figure 16. Average axial velocity along with the 2D velocity vectors and streamlines: (a) PIV; (b) LES. The cross sections at which velocity data are compared are also shown.

of the flame in this region due to heat loss from the chamber walls.13 In Figure 15b, the contours of the average vorticity magnitude obtained from LES indicate significant vortex activity in the shear layers. While heat release and thermal expansion in the ISL, along with a reduction in the Reynolds number, suppress the development of the shear layer eddies to some extent, combustion-induced baroclinic vorticity production enhances the vorticity in the mixing layer.19 It is also observed that the flow is characterized by the absence of a precessing vortex, as noted in the V flame configuration.15 However, experimental measurements indicate the presence of small-scale fluctuations (at a characteristic frequency of 110 Hz) in the region, associated with transient eddies that are generated as a result of combustion-induced turbulence; the simulations also predict this flow behavior for the M flame configuration, suggesting that the high heat loading in this case may be responsible for the generation of these small-scale eddies.15 Figure 16 depicts the contours of the average axial velocity obtained experimentally and numerically. Significant changes in the mean flow field from the V flame configuration are evident, principally in the size and strength of the inner and outer recirculation zones. A decrease in the axial and radial span of the IRZ is observed, along with a reduction in the spreading angle of the annular jet. Additionally, the thickness of the annular jet increases, as does the velocity of the jet fluid. The vortex structure within the core of the IRZ that was observed experimentally in the V flame configuration is no longer present. Furthermore, the propagation of the flame along the OSL impacts the mean and fluctuating flow fields in the periphery of the IRZ. An elongation of the ORZ takes place as a result of heat release and gaseous expansion in the region and corresponds to a reduction in the spreading angle of the annular jet.

ments and the average normalized heat release from LES. A thin reaction zone is observed in the ISL region, which extends upstream into the inlet channel (not shown). The flame front curves near the location where the annular jet impinges on the chamber walls, and the flame brush continues along the wall shear layer. The reaction zone is supported by entrainment phenomena brought about by the eddy structures that are convected downstream as well as by mixing processes augmented by the turbulence generated in the region. The increase in the burning velocity also allows the flamelets to persist in the region, in contrast to the lower-equivalence-ratio flame configurations (Figure 7). The flame begins to propagate toward the low-velocity rear stagnation zone, resulting in a thickened average flame surface in the region. The reaction process continues in the downstream section of the combustion chamber, and the fuel is consumed before the flow proceeds into the exhaust duct, as noted in Figure 7. Figure 14 shows the contours of the average and rms temperature fields obtained from the simulations. It is evident that no flame persists inside the ORZ/OSL region, and a conical flame that is stabilized in the ISL is observed. In contrast, an OSL/ORZ flame is manifested only for significantly high heat loading (φ ≥ 0.65), resulting in the M flame configuration, which is discussed in the next section. “M” Flame Configuration. The Abel-transformed flame image for the M flame configuration, displayed in Figure 7(IV), suggests the presence of a conical flame that is stabilized in the ISL, as in case of the V flame configuration, and anchored on the swirler (optically inaccessible and not shown). Additionally, a diffuse flame brush in the OSL region is also evident. As a result of the higher burning velocity and extinction strain rate, the flame overcomes the high-strain environment in the OSL section (Figure 15a) and is anchored on the periphery of the inlet tube. The higher heat loading also prevents any quenching H

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 17. (a) Combustor wall pressure gradients in confined isothermal and reacting flows, (b) Mean centerline velocity in confined isothermal and reacting flows.11

Figure 18. rms axial velocity along with the 2D velocity vectors and streamlines: (a) PIV; (b) LES.

Figure 19. Normalized profiles of (top) average axial velocity and (bottom) rms axial velocity: red circles, PIV; solid blue lines, LES.

Further inspection of the average flow field suggests the presence of a single-cell-type, closed recirculating vortex structure, with the front stagnation point located downstream of the expansion plane. The IRZ spans a length of approximately 2.0R downstream of the sudden expansion and is also quite narrow, with a radial span of approximately 1.0R. The change in the structure of the VBB from the V flame coonfiguration may be attributed to the influence of flame propagation on the mean flow field in this scenario. In the case of reacting flows, the acceleration of the flow (due to dilatation effects of combustion), together with restriction of significant mean streamline curvature by the confining walls, causes the flow to better overcome the adverse pressure gradients. Therefore, for low to moderate equivalence ratios, the fluid-

dynamic effects of heat release, in general, result in elongation of the IRZ compared with isothermal flow. However, with an increase in heat loading, the enhanced turbulent burning velocity causes the flamelets to propagate upstream, which tends to strongly impact the recirculation region topology, eventually reducing the axial span of the VBB (Figure 17). The influence of heat loading on the flow features as a function of equivalence ratio has been examined previously for an axisymmetric bluff-body configuration11 and corresponds with the observations noted for the swirl combustor. The contours of the rms axial velocity fluctuations (Figure 18) indicate high turbulence activity in the inner and outer shear layers around the periphery of the IRZ. Typically, for reacting flows involving high turbulent burning velocity and I

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 20. (a) Instantaneous CO mass fraction, (b) Instantaneous normalized heat release. The corresponding 2D velocity vectors and streamlines are also shown (LES).

Figure 21. (a) Normalized deconvolved CH chemiluminescence intensity (PIV). (b) Average normalized heat release (LES). The corresponding 2D velocity vectors and streamlines are also shown.

Figure 22. (a) Average temperature (K). (b) rms temperature (K). The corresponding 2D velocity vectors and streamlines are also shown (LES).

heat loading, the enhanced heat release augments the magnitude of the shear layer turbulence as well as turbulent dilatation. For the M flame configuration, the flow fluctuation intensity in the shear layer remains high as a result of the generation of combustion-induced turbulent kinetic energy. The narrowing of the IRZ also shifts the high-fluctuation region to within the recirculation zone. Additionally, turbulence is high in the region where the annular jet impinges the wall, extending further downstream along the wall shear layer, where combustion-generated turbulence augments the flow fluctuations. Figure 19 shows the normalized profiles for the average axial velocity and rms axial velocity obtained from the numerical simulations and the experimental measurements. A gradual spreading of the annular jet is observed, corresponding to a narrowing of the IRZ; the jet is also noted to be thicker in contrast to the V flame configuration. Thus, the shifts in the structure of the annular jet and the size of the IRZ are suitably resolved in the simulations. High axial flow oscillations are observed in the inner and outer shear layers around the periphery of the IRZ. The narrowing of the IRZ also results in an enhancement of the turbulent kinetic energy in the inward

locations of the recirculation zone. Significant turbulent activity is also noted near the wall at the axial location of maximum IRZ width (x/R ≈ 1), where the annular jet impinges the wall, extending downstream along the wall shear layer. The flow characteristics are suitably resolved numerically, although the simulated flow-field predictions differ slightly from the experimental measurements in the section immediately downstream of the sudden-expansion plane (x/R ≈ 0.5) because of a slight delay in the development of the VBB, as also noted in ref 13. This may be attributed to the sensitivity of the resolved flow features to geometric features in the inlet pipe and the grid resolution in the boundary layer of the inlet duct.20 Additionally, the predictions in the near-wall regions may be improved by appropriate mesh refinement and choice of a suitable wall function, while adjustments to the PIV setup for the measurement of higher-resolution flow fluctuation data in downstream locations and near the chamber walls could also be investigated. Figure 20 depicts the instantaneous LES contours of the CO mass fraction and normalized heat release. A thin flame front is present along the inner and outer shear layers and subsequently curves near the location where the annular jet impinges the J

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 23. Average wall heat flux (W/m2) from LES: (a) V flame configuration; (b) M flame configuration.

configuration for the swirl combustor (Figure 20b). Thus, the flame holding features are sensitive to the thermal condition on the walls and strain computations in the OSL, and an OSL flame is manifested only for significantly high heat loading, resulting in the M flame macrostructure, as observed in the experiments and predicted in the simulations.

wall. The reaction zone is thick in this section, and the high rate of mixing in the region augments heat release, resulting in a high-intensity flame front. The flame continues along the wall shear layer and exhibits a compact envelope that extends to a distance of approximately 3.0R downstream of the expansion plane. Thus, combustion is complete before the flow exits the chamber. In Figure 21, the contours of the average normalized deconvolved CH chemiluminescence intensity from the experiments and the average normalized heat release from LES are shown. With an increase in heat loading, the flame− flow field is impacted by combustion-induced turbulence as well as dilatation effects. For the M flame scenario, the turbulence is slightly enhanced in the mixing layer, which along with the augmented heat release results in a high-intensity flame in the ISL (Figure 21). Downstream of this region, the temperature remains fairly constant within the recirculation zone (Figure 22a), while a slight thickening of the flame front takes place in the wall shear layer region, as observed in the V flame scenario. Additionally, a weakly burning flame propagates along the OSL, resulting in a moderate level of temperature fluctuations in the region. This is also evident from the contours of the rms temperature field shown in Figure 22b. The transition of the flame macrostructure from the V configuration to the M configuration can be attributed to the increase in the extinction strain rate as well as the burning velocity, allowing the flame brush to stabilize and persist in the high-strain OSL region (Figure 15a). Furthermore, the presence of the combustor wall significantly influences the flame behavior and the temperature field, which can also be inferred from the wall heat flux contours, as depicted in Figure 23 for the V and M flame configurations. The average heat flux over the circumference of the combustion chamber is projected onto a diametrical cross section, and significant heat losses are observed for each case. For the lower heat loading case, the wall heat flux is predominantly significant in the region beyond the location where the flame impinges on the wall (x/R > 1). In the case of the higher-equivalence-ratio flow, the presence of the flame in the OSL results in a hightemperature region in the ORZ, and the wall heat flux is particularly high in this section of the combustion chamber. The presence of the wall and the ensuing heat loss can also impact the stabilization of the flame and provoke extinction, particularly with a reduction in the equivalence ratio. It has been suggested for cases involving moderate heat loading in a similar combustor geometry that the heat losses to the combustor wall can rapidly quench the flame; once the specific heat release rate crosses a threshold value, a persistent flame can stabilize in the OSL,13,21 such as in the case of the M flame

VI. CONCLUSIONS In the case of confined reacting flows, the mean and fluctuating flow fields can influence the stabilization of the flame and its intensity. Likewise, the heat release also substantially impacts the mean and fluctuating flows. Besides influencing the vorticity field and the flow field topology, the change in the heat loading also influences the flame speed, specific heat release, and extinction strain rate as well as the wall heat flux. This can significantly impact the flame stabilization location as well as the locations of intense heat release within the combustion chamber. Depending on the flame−flow interactions, the flame in a confined swirl combustor exhibits different configurations as the equivalence ratio is varied. Wall confinement can also influence the flame−flow dynamics; in general, it prevents any significant deflection of mean flow streamlines by restricting the radially outward spread of the annular jet, thereby narrowing the inner recirculation zone in contrast to an open flame. It also results in the generation of turbulence in the region where the annular jet impinges on the wall and produces acceleration of the flow that typically elongates the inner recirculation zone. Additionally, the combustor wall also results in heat losses and can appreciably influence the stabilization of the flame in the outer shear layer. Consequently, the average and fluctuating flow fields as well as the reaction zone intensity are critically impacted by the presence of the wall. Therefore, in the case of confined swirling reactive flows, increasing the equivalence ratio can result in an alteration of the dominant flame stabilization mechanism, thereby causing transitions across distinct flame configurations, while also modifying the recirculation zone topology significantly. In this work, the transition of the flame macrostructure with changes in the equivalence ratio has been investigated using experiments and numerical simulations for a laboratory-scale cylindrical sudden-expansion swirl combustor. It has been observed that changes in heat loading in the presence of wall confinement significantly influence the flow field in the wake region, the stabilization location of the flame, and the flame intensity. Specifically, increasing the equivalence ratio drastically reduces the average size of the inner recirculation zone and causes a transition of the flame macrostructure from the “V” configuration to the “M” configuration. While the flame is stabilized along the inner shear layer for the V flame, a persistent diffuse reaction zone is also manifested along the K

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels outer shear layer for the M flame. The average chemiluminescence intensity increases in the case of the M flame macrostructure, while the axial span of the reaction zone within the combustion chamber decrease. The LES predictions correspond with the experimental measurements, and the insights obtained highlight the effect of heat loading variations on the flame−flow interactions in a wall-confined swirl combustor configuration, which can significantly impact combustion stability in acoustically coupled systems.



Proceedings of the 50th AIAA Aerospace Sciences Conference; American Institute of Aeronautics and Astronautics: Reston, VA, 2012; DOI: 10.2514/6.2012-450. (13) Chterev, I.; Foley, C. W.; Foti, D.; Kostka, S.; Caswell, A. W.; Jiang, N.; Lynch, A.; Noble, D. R.; Menon, S.; Seitzman, J. M.; Lieuwen, T. C. Flame and Flow Topologies in an Annular Swirling Flow. Combust. Sci. Technol. 2014, 186, 1041−1074. (14) LaBry, Z. A.; Taamallah, S.; Kewlani, G.; Shanbhogue, S. J.; Ghoniem, A. F. Mode Transition and Intermittency in an Acoustically Uncoupled Lean Premixed Swirl-Stabilized Combustor. In ASME Turbo Expo, Volume 4B: Combustion, Fuels and Emissions; American Society of Mechanical Engineers: New York, 2014; Paper GT201427266, DOI: 10.1115/GT2014-27266. (15) Kewlani, G., Large Eddy Simulations of Premixed Turbulent Flame Dynamics: Combustion Modelling, Validation and Analysis. Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 2014. (16) Kewlani, G.; LaBry, Z.; Abani, N.; Shanbhogue, S.; Ghoniem, A. Large Eddy Simulations and Experimental Investigation of Flow in a Swirl Stabilized Combustor. In Proceedings of the 50th AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition; American Institute of Aeronautics and Astronautics: Reston, VA, 2012; DOI: 10.2514/6.2012-178. (17) Kewlani, G.; Vogiatzaki, K.; Shanbhogue, S.; Ghoniem, A. Validation study of Large Eddy Simulations of Wake Stabilized Reacting Flows Using Artificial Flame Thickening Approaches. In Proceedings of the 51st AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition; American Institute of Aeronautics and Astronautics: Reston, VA, 2013; DOI: 10.2514/ 6.2013-169. (18) Pope, S. B. Turbulent Flows; Cambridge University Press: Cambridge, U.K., 2000. (19) Pan, J. C.; Vangsness, M. D.; Ballal, D. R. Aerodynamics of bluffbody stabilized confined turbulent premixed flames. J. Eng. Gas Turbines Power 1992, 114, 783−789. (20) Taamallah, S.; Vogiatzaki, K.; Kewlani, G.; Ghoniem, A. Influence of Boundary Layer Trip on Non-Swirling and Swirling Reacting Jets Using Large Eddy Simulation. Presented at the SIAM Numerical Combustion Conference, 2013; Paper CP6. (21) Polifke, W.; Hirsch, C.; Zellhuber, M.; Komarek, T.; Chong, L. Influence of Strain and Heat Loss on Flame Stabilization in a NonAdiabatic Combustor. In Proceedings of the Fourth European Combustion Meeting, 2009.

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This research was partially funded under Grant KUS-110-01001 from the King Abdullah University of Science and Technology. The contributions by Dr. Zachary LaBry, Dr. Konstantina Vogiatzaki, and Dr. Neerav Abani to the discussions are gratefully acknowledged.



REFERENCES

(1) Huang, Y.; Yang, V. Dynamics and stability of lean-premixed swirl-stabilized combustion. Prog. Energy Combust. Sci. 2009, 35, 293− 364. (2) Gicquel, L. Y. M.; Staffelbach, G.; Poinsot, T. Large Eddy Simulations of gaseous flames in gas turbine combustion chambers. Prog. Energy Combust. Sci. 2012, 38, 782−817. (3) Yoon, C.; Geiji, R.; Anderson, W. E.; Sankaran, V. Computational Investigation of Combustion Dynamics in a Lean Direct Injection Gas Turbine Combustor. In Proceedings of the 51st AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace Exposition; American Institute of Aeronautics and Astronautics: Reston, VA, 2013; DOI: 10.2514/6.2013-166. (4) Lu, X.; Wang, S.; Sung, H.-G.; Hsieh, S.-Y.; Yang, V. Large-eddy simulations of turbulent swirling flows injected into a dump chamber. J. Fluid Mech. 1999, 527, 171−195. (5) Huang, Y.; Yang, V. Effect of swirl on combustion dynamics in a lean-premixed swirl-stabilized combustor. Proc. Combust. Inst. 2005, 30, 1775−1782. (6) Huang, Y.; Sung, H.-G.; Hsieh, S.-Y.; Yang, V. Large Eddy Simulation of combustion dynamics of lean-premixed swirl-stabilized combustor. J. Propul. Power 2003, 19, 782−794. (7) Wang, S.; Hsieh, S.-Y.; Yang, V. Unsteady flow evolution in swirl injector with radial entry. I. Stationary conditions. Phys. Fluids 2005, 17, 045106. (8) Huang, Y.; Yang, V. Bifurcation of flame structure in a leanpremixed swirl-stabilized combustor: transition from stable to unstable flame. Combust. Flame 2004, 136, 383−389. (9) Altay, H. M.; Speth, R. L.; Hudgins, D. E.; Ghoniem, A. F. The impact of equivalence ratio oscillations on combustion dynamics in a backward facing step combustor. Combust. Flame 2009, 156, 2106− 2116. (10) Speth, R.; Altay, H.; Hudgins, D.; Annaswamy, A.; Ghoniem, A. Vortex-Driven Combustion Instabilities in Step and Swirl-Stabilized Combustors. In Proceedings of the 46th AIAA Aerospace Sciences Meeting; American Institute of Aeronautics and Astronautics: Reston, VA, 2008; DOI: 10.2514/6.2008-1053. (11) Ballal, D. R.; Heneghan, S. P.; Schmoll, W. J.; Takahashi, F.; Vangsness, M. D. Combustion and Heat Transfer Studies Utilizing Advanced Diagnostics; Report WL-TR-92-2110; University of Dayton: Dayton, OH, 1992. (12) Chterev, I.; Foti, D.; Seitzman, J.; Menon, S.; Lieuwen, T. Flow Field Characterization in a Premixed, Swirling Annular Flow. In L

DOI: 10.1021/acs.energyfuels.5b02921 Energy Fuels XXXX, XXX, XXX−XXX