numerical investigation of the role of

2 downloads 0 Views 8MB Size Report
indicated that increasing the swirl number broadened the stability limits; however, locally unburnt fluid samples had been shown to ap- pear in swirl flames but ...
Fuel 236 (2019) 1226–1242

Contents lists available at ScienceDirect

Fuel journal homepage: www.elsevier.com/locate/fuel

Full Length Article

An experimental/numerical investigation of the role of the quarl in enhancing the blowout limits of swirl-stabilized turbulent non-premixed flames

T



A.M. Elbaza,b, , S. Yuc, X. Liuc, X.S. Baic, I. Kheshoa, W.L. Robertsa a

Clean Combustion Research Center, King Abdullah University of Science and Technology, KAUST, Thuwal 23955-6900, Saudi Arabia Faculty of Engineering-Mattaria, Helwan University, Cairo, Egypt c Division of Fluid Mechanics, Lund University, 22100, Sweden b

A R T I C LE I N FO

A B S T R A C T

Keywords: Quarl swirl stabilized Non-premixed swirl flames PIV/OH-PLIF LES Transported PDF

The blowout limits of methane/air non-premixed swirl-stabilized flames were measured with and without quarl. The addition of a quarl significantly enhances the flame blowout limits. The transition from attached flame to blowout was mapped. To explore the role of the quarl, a series of OH-PLIF/PIV experiments, coupled with large eddy simulations (LES) using a transported probability density function (PDF) model, were carried out on flames with and without quarl over a wide range of fuel jet velocity, Uf. The results show that the mean flow field is characterized by two recirculation zones. The existence of the quarl enhances this flow field by triggering a larger scale of reversal flow, penetrating deeply upstream into the quarl. This results in much earlier fuel, extending down into the air tube, where a diffusion flame is stabilized around the stoichiometric mixture contour and locally low scalar dissipation rates. The relative delay in fuel/air mixing in non-quarl flames results in a locally strong scalar dissipation rate layer overlapping the stoichiometric mixture contour, and thus, the flame is highly sensitive to local extinction with increasing fuel jet velocity. At high Uf, in the liftoff flame region, the existence of the quarl enhances the jet spreading and a weak recirculation zone around the highly strained jet is observed. Together with fuel jet spreading, partial oxidization of the mixture upstream the lifted flame base creates a wider range of burnable mixture along the axis in the quarl flames. On the contrary, the high scalar dissipation rate and the absence of a recirculation region in the proximity of the fuel nozzle in the non-quarl flame give rise to an earlier blowout.

1. Introduction Due to the complexity of the flow fields coupled with combustion and heat release, non-premixed swirl stabilized flames continue to challenge both the experimentalists and modelers. However, with the recent development and advances in the laser diagnostics and computational power, these research challenges are now being revisited and investigated. There is a real need to explore the interlinks between the complex swirl flow structure, the mixing field and flame stability, hence, enhancing and optimizing the combustor designs. Non-premixed swirl stabilized flames are used extensively in practical combustion systems, particularly gas turbines, furnaces and boiler [1]. Non-premixed swirl flames offer the possibility to control the flow and mixing, thus to improve the stability limits and control NOx emission by manipulating the aerodynamics of the inlet conditions [2]. With a burner configuration consisting of a central fuel jet surrounded by swirling air, ⁎

different aspects of non-premixed swirl stabilized flames have been studied [2–6]. Introducing the fuel jet along the centerline of the formed toroidal vortex, the fuel-air mixing rate has been shown to be enhanced by a factor of five times greater than that of the jet flame [4]. Due to swirl, the fuel jet-vortex interaction was shown to reduce the fuel jet velocity and thus strongly enhance the lifted flame stability and improve the rich flame blowout limit [5]. The blow-off limit in non-premixed swirling flames was first investigated by Feikema et al. [6], who compared the flame blow-off limits with and without swirl, and concluded that the reason behind the enhancement of the lean stability limit was the reduced local strain rate. In studies of the Sydney swirl non-premixed flames [7–12], a parametric investigation on the stability map and blow-off limits of non-premixed swirling flames was carried out [7]. Raman and Rayleigh point measurements in these non-premixed swirling flames were conducted. The main findings of the compositional flame structure

Corresponding author at: Clean Combustion Research Center, King Abdullah University of Science and Technology, KAUST, Thuwal 23955-6900, Saudi Arabia. E-mail address: [email protected] (A.M. Elbaz).

https://doi.org/10.1016/j.fuel.2018.09.064 Received 13 July 2018; Received in revised form 7 September 2018; Accepted 12 September 2018 0016-2361/ © 2018 Elsevier Ltd. All rights reserved.

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Fig. 1. (a) Swirl stabilized non-premixed flame burner and, (b) OH-PLIF/PIV experimental setup.

study had three specific purposes: (1) to gain deeper understanding of the effect of the quarl on the blowout limits of non-premixed swirl stabilized flames, (2) to characterize the development of the flame structure and flow field at the different regimes from the blowout with and without quarl, and (3) to delineate the effect of the quarl on the mixing/flow fields and the interlinks between the mixing/flow fields and the flame liftoff mechanism. The blowout limits for quarl and nonquarl methane/air swirling flames, showing the different regimes from attached, liftoff, to blowout, are measured first. An integrated study combining OH-PLIF/PIV measurements and LES are then used to interpret the flame stabilization mechanism and structure. The recirculating flow, the instability of the flow field, and precession are considered the main challenges in the simulating of these swirlstabilized flames. Large Eddy Simulations (LES) is considered a valuable tool capable of computing the complex transient features, where it has shown an accurate predictor of the flow instabilities associated with a bluff body stabilized flames [20,21]. Hybrid methodologies that involve LES, as well as either conditional moment closure (CMC) or PDF methods to represent combustion at the sub-grid levels, are under development. However, this development is in need of experimental data with well-defined boundary conditions. In swirling reactive flows, LES has great potential for predicting the detailed features of the flow field [22,23]. The combustion mode of the present flames is complex, thus, a transported probability density function (PDF) model is used in the LES, which is known to be able to handle different combustion modes [24,25].

indicated that increasing the swirl number broadened the stability limits; however, locally unburnt fluid samples had been shown to appear in swirl flames but not in non-swirling bluff-body stabilized flames before flame blowout [9,11]. In many studies of non-premixed swirling jet flames [3–6], a divergent nozzle (quarl) was attached to the burner exit. It was found that the existence of the quarl at the exit of the burner enhances the flame’s stability at higher jet Reynolds number. Examination of the flow field revealed that the quarl causes the formation of a recirculation zone early in the flame region, which anchors the flame to the burner tip [13,14]. A parametric investigation of the quarl geometry’s influence on swirling flame stabilization has been studied [15]; however, there is insufficient data on either the vector or scalar fields included in their work to adequately explain their observations. The non-reacting swirling flow issuing from a straight exit burner [16] was compared with existing data of a diverging quarl burner [3], where the quarl confinement was shown to act as a flame holder. An experimental study of the flame structure stabilized in the quarl of an oxy-fuel swirl burner has been conducted [17]. With flow field and OH-PLIF information, they noticed a complex flame structure inside the quarl, and speculatively suggested that flames around the inner recirculation zone were in multiple combustion regimes and became partially premixed flames. Recently, in a burner similar to that in refs. [3–6] with a central fuel jet surrounded by a swirling air flow, the temperature/gas species inside the quarl and the downstream flow field were investigated [18]. They show that a significant change of both flame structure and emissions with the change of the quarl geometry, indicating a change in the combustion mode. However, the reason behind this change was unclear. Consequently, in this same burner, a quartz quarl was used and optical measurements were carried out close to the flame base [19]. The results of a large eddy simulation (LES) and OH-PLIF/PIV measurements showed that the combustion mode was changing from a diffusion flame stabilized upstream of the diverging quarl to a partially premixed flame stabilized downstream of the straight quarl. The studies of [19] was focused on a low speed condition where the flames were far from blow off. Thus, it is unclear how the flame structures and flow field would change near blow off. The influence of the quarl at the burner exit on both the flow and mixing fields, flame structure near blowout and hence flame stability, is still uncertain. It is evident that the present knowledge of the effect of quarl on flame stability, flow and mixing fields, and flame structure near blowout is incomplete. The present work is aimed at providing a detailed understanding of the quarl effect on the non-premixed swirl stabilized flame structure, flow and mixing fields under conditions near blowout. The present

2. Experimental apparatus The burner used in this work is similar to that of refs. [18,19], where more details are provided. A schematic of the burner is shown in Fig. 1a. The swirling flow around a central fuel tube (df = 4.4 mm) is generated via four tangential air inlets to the air tube (with an inner diameter, dA, of 27 mm). The tangential air streams merge with the axial air upstream of the burner. The exit of fuel (methane) tube terminates with the air tube exit, at x = 0 mm. Flames with quarl and without quarl are investigated. The quarl of 15° half cone angle and a 40 mm length is fitted to the air tube exit. Characteristics of the swirling jet are usually defined in terms of the swirl number, S. The swirl number S as given by Syred and Beer [26] is defined as the ratio of the flux of the angular momentum at the throat of the burner to the axial flux momentum times the burner throat radius. Thus, dA /2 dA /2 S = ∫d /2 (ρrUa Wa)2πdr )/[ ∫d /2 ρ (Ua2−Wa2/2)2πdr (dA/2) ], where ρ is f

f

the air density, r is the radial distance, Ua is the average axial air 1227

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

measurements. A dual cavity, diode-pumped, solid-state Nd: YLF laser (LDY 300) and two high-speed CMOS cameras were used for PIV setup. The laser was capable of producing 35 W per head at 10 kHz with a nine ns pulse width. The double-pulsed beams were formed into a sheet along the flame central plan using three cylindrical lenses. Two high-speed CMOS cameras were located at ± 15° to the sides of the OHPLIF ICCD camera to acquire the stereoscopic particle images (528 × 692 pixels resolution). With 105 mm/f4 objective lens (Nikon UV Micro-Nikkor) and fitted with Scheimpflug adaptors equipped with a 527 nm bandpass filter, the Mie scattered flame images were acquired by the PIV cameras. Perspective distortion was corrected using a dual plane, three-dimensional imaging target (LaVision Type 11). Using the self-calibration function, the coordinate system and the camera calibration of the Stereoscopic Particle Image Velocimetry, SPIV, setup were adjusted so that the Y = 0 mm plane was adjusted to be exactly in the middle of the laser light sheet. The swirling air and fuel jet were seeded with titanium dioxide particles (0.5 μm nominal diameter). An adaptive multi-pass vector evaluation technique was used (Davis 8.1 software) to compute the vectors fields, with interrogation boxes ranging from 128 pixels to 16 pixels. With a spatial resolution of 1.2 × 1.2 × 1 mm, 2500 images were recorded for each flame. Spurious vectors were detected and removed using a function of Q-Factor equal to 1.3. With a median filter size of 3 × 3 vectors, the resulting vectors were filtered. The vectors were eliminated if the difference between them and their neighbors’ was more than four standard deviations from the mean of the surrounding vectors. By using the cross-correlation peak finding module in the Davis 8.1 algorithm with ± 0.1-pixel uncertainty, the random uncertainty of the PIV measurements is estimated to be ± 0.35 m/s. A frequency-doubled high-speed dye laser (Sirah, Cerdo-Dye) pumped at 532 nm by a frequency-doubled, diode-pumped solid-state INNOSLAB laser (Edgewave IS16II-E) was used to generate the necessary UV laser light for OH-PLIF imaging. The dye laser produced a fundamental laser beam at 566 nm with Rhodamine 6G dye. This fundamental beam was then frequency doubled to produce 283 nm laser beam at 10 kHz repetition rate by using a BBO crystal. At the exit, the average laser power was 2.8 W at 283 nm. A collimated UV sheet of approximately 60 mm height was created and expanded by using two cylindrical lenses. With a third cylindrical lens of a focal length of 600 mm, the sheet was focused to approximately 130 μm width at the center of the burner. The OH signal was acquired at 90° angle to the laser sheet via CMOS ICCD camera (LaVision IRO), equipped with Nikkor UV lenses (f/4.5, f = 105 mm) and appropriate band-pass interference filter of > 80% at 310 nm. The Q1(6) transition in the band of the A2Σ ← X2Π(1, 0) OH system was pumped. Each OH-PLIF image was processed to remove the laser sheet inhomogeneity, background noise, and corrected for both shot-to-shot laser fluctuations and laser sheet profile. The shot-to-shot variation was corrected by monitoring the energy of individual laser pulses, by using energy monitor equipped with a fast photodiode, simultaneously with the acquisition of ICCD images. The laser sheet’s profile correction was determined using an ensemble average of 10,000 individual images of the laser sheet passing through a cuvette filled with acetone placed in the center of the field of view.

Fig. 2. Stability diagram of quarl and non-quarl swirl stabilized non-premixed flames.

velocity at the burner throat location, defined in Fig. 1a, and Wa is the azimuthal air velocity components. The aforementioned formula eliminates the need to measure static pressure and allows the use of stereo PIV [27] data to calculate the swirl number. However, it was found to be more convenient to track S by monitoring the tangential and axial mass flow rates of air by determining the geometrical swirl number, Sg. The geometrical swirl number is defined as [5] Sg = (π ro dA/2At) (mθ/(mθ + mA))2, where mθ and mA are the tangential and axial air flow rates respectively, At is the area of the four tangential air inlets, and ro is the air tube radius. The current burner configuration is relatively simple geometry, but replicates many of the fundamental swirling flame behaviors of real gas turbine combustors. To study the flame stability map, we kept the geometrical swirl number constant at 12 while changing the axial air bulk velocity. However, the actual air swirl number based on the axial and tangent air streams only, without fuel injection, was estimated based on the stereo-PIV measurement at the burner exit. The air flow field measurements at a bulk air velocity of 4.9 m/s show a significant reduction in the swirl intensity to 0.32 of the geometrical swirl number, Sg. PIV/OH-PLIF measurements at 5 KHz repetition rate were conducted for the region just above the quarl exit. The OH-PLIF measurements were extended to cover the entire quarl region, see Fig. 1a for more illustration. However, for non-quarl flames, the PIV/OH-PLIF measurements covered the region downstream of the burner exit. Details of the stereoscopic PIV/OH-PLIF setup are presented elsewhere [19] and is schematically shown in Fig. 1b. Two counter-propagating laser sheets crossed the vertical plane of the burner for the PIV/OH-PLIF Table 1 The main parameters and designations of the investigated flames. Sg, Ua (m/s)

Sg = 12,

Uf(m/s)

4.3 8.7 33 55 100

Non-quarl flames PIV √ √ √

OH-PLIF √ √ √ √

Flames with quarl LES √ √ √ √ √

S 0.982 0.766 0.611 0. 485 0.308

Ua = 4.9 m/s

1228

Flame Fno4.3 Fno8.7 Fno33 Fno55 Fno100

PIV √ √ √ √

OH-PLIF √ √ √ √ √

LES √ √ √ √ √

S 0.986 0.820 0.654 0.543 0.393

Flame Fq4.3 Fq8.7 Fq33 Fq55 Fq100

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Fig. 3. Instantaneous OH-PLIF shots with increasing Uf of non-quarl flames, Fno4.3, Fno8.7, Fno33, and Fno55.

Fig. 4. Instantaneous OH-PLIF shots with increasing Uf of quarl flames, Fq4.3, Fq8.7, Fq33, Fq55, and Fq100.

premixed flames, with possible local extinction and re-ignition. To model such complex flame modes, LES with a comprehensive turbulence/chemistry interaction model based on transported probability density function (TPDF) in the Eulerian Stochastic Field (ESF)

3. Numerical approach From the experimental measurements, it was understood that the present flame was made up of multiple modes, e.g., diffusion and 1229

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

eddy diffusivity that is determined from the Smagorinsky model. The last term on the r.h.s of (1) represents the effect of micro-scale mixing on the joint PDF, which is modeled here using the interaction by exchanging with the mean (IEM) model [29]. The last term on the l.h.s of (1) represents the effect of chemical reactions on the joint PDF, which is in closed form and requires no further modeling. Equation is solved using the ESF method proposed by Valino [30]. In the ESF method, the reactive scalar fields are represented by N stochastic fields for each of the Ns scalars (species mass fractions and enthalpy or temperature), denoted here as ξαn (X , t ) , with 1 ≤ n ≤ N , 1 ≤ α ≤ Ns . ∼ The joint PDF, Psgs ( ψ; x , t ) , is computed by an ensemble of the stochastic fields,

1 ∼ Psgs ( ψ; x , t ) = N

N

Ns

∑ ∏ δ (ψα−ξαn) n=1 α=1

(2)

where δ is the Dirac delta-function. The mean and the moments of the reactive scalar variables can be approximated from the ensemble of N stochastic fields. For example, the mean is,

1 ∼ ϕα = N

N

∑ ξαn

(3)

n=1

The governing equation for the nth stochastic field is: n

n

∂ξ ∂ξ ρ¯ n ∼ ∂ 1 ¯ αn = −ρu ¯ ∼j α dt + (Γt α ) dt + ρω ¯ ̇α ( ξ n ) dt − Cϕ ρdξ (ξ −ϕ ) dt ∂x j ∂x i ∂x i 2 τsgs α α + ρ¯ 2

n Γt ∂ξα dWin, ρ¯ ∂x i

(4)

n where ξ n = [ξ1n, ξ2n, ⋯, ξNs ] is the composition/enthalpy vector for the nth stochastic field. dW n represents a vector Wiener process that is spatially uniform but different for each field. Cϕ is the model constant for micro-mixing, taking a value of 2 [31–33]. The sub-grid mixing time scale τsgs is calculated as [34],

τsgs =

framework was used. The composition LES-ESF method was employed. In this approach, spatially filtered continuity equation and NavierStokes equations are used to describe the velocity field on the resolved scales, whereas the composition and temperature fields are determined from joint PDFs. The sub-grid scale (SGS) terms in the spatially filtered Navier-Stokes equations, known as the SGS stresses, are modeled using the Smagorinsky model [28]. The one-point, one-time joint PDF ∼ Psgs ( ψ; x , t ) of the composition and temperature vector ψ that describes the random motion in the SGS is modeled using the following transported PDF equation [24,25],

∼ Ns ∂Psgs ( ψ) ∂ ∼ ¯ ∼j + ∑ [ρω ¯ ̇ ( ψ) Psgs ( ψ)] + ρu ∂x j ∂t α = 1 ∂ψα ∼ Ns ∂ ⎡ ∂Psgs ( ψ) ⎤ Cϕ ∂ ∼ ∼ = Γ [ρ¯ (ψα −ϕα ) Psgs ( ψ)] ∑ − t ∂x i ⎢ ∂x i ⎥ τsgs α = 1 ∂ψα ⎣ ⎦

(5)

where Δ is the length of the spatial filter. The LES-TPDF approach has the advantage of being combustion mode-free, since it is based on the direct integration of finite-rate chemistry with PDF, in which the sub-grid turbulence effect on the chemical reaction rate source terms is closed directly in the model. LESESF approach has been used successfully in numerical simulations of multiple modes combustion processes, e.g., refs. [33,35–38], where the model was able to capture the multiple combustion modes (diffusion flame and partially premixed flame in gas turbine engines, ignition in compression ignition engines, and lean blow-out of turbulent premixed flame). Following the works of Jones et al. [35,36] and Liu et al. [19], we have used eight stochastic fields in the present calculations, which has been shown to give an acceptable accuracy with a reasonable computational cost for several different applications. Following our previous work [19], the chemical kinetic mechanism of Smooke and Giovangigli [39,40] is used, which is made up of 16 species and 35 reactions, and is known to give similar laminar flame speeds for premixed flames and quenching scalar dissipation rate for non-premixed flams as those from more comprehensive mechanisms, e.g. GRI3.0 [41]. The computational cost required with their mechanism [41] is relatively low; thus, it is preferred here since the PDF calculation is computationally demanding. The open source CFD code, OpenFOAM [42], is adopted for the numerical integration of the governing equations. A second-order filtered linear implicit scheme is used for the spatial discretization, and the second-order backward Euler scheme is used for the temporal integration. The pressure implicit with the splitting of operator (PISO) algorithm is employed for the pressure-velocity coupling. The four air

Fig. 5. Instantaneous OH mass fraction from LES for the flames with and without quarl at Uf = 4.3, 8.7, 55 and 100 m/s; upper row: non-quarl, lower row: with quarl.

ρ¯

ρ¯Δ2 μ + μsgs

∼ ∂Psgs ( ψ)

(1)

where a quantity with overbar is the spatially filtered; quantity with over-tilde denotes density weighted spatially filtered. ρ is the density of the local mixture; uj is the velocity component along Cartesian coordinate xj direction. Γt is the sum of molecular diffusivity and SGS 1230

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Fig. 6. Mean streamlines for non-quarl flames (a-c), quarl flames (d-g) at different Uf, red dotted line shows reversed flow boundaries. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

flame blowout occurs. Note that the actual swirl number of the fuel/air mixture at the exit plane of the fuel nozzle decreases with increasing Uf. Thus, the flow structure approaches that of the non-swirl jet flow at high Uf. For non-quarl flames with increasing Uf, the attached flames start to liftoff before blowout; here chemiluminescence from OH* was used to monitor the flame liftoff using a camera exposure time of 40 μs, and the corresponding Uf was recorded. In Fig. 2, at the liftoff limit shown by the dashed line in the plot, a distinction is made by the vertical arrow showing two different of transitioning from an attached flame to a lifted flame. To the left of the arrow, the flames appear to be similar to nonswirl jet flames. However, to the right of the arrow, compact flames start to elongate with increasing Uf (at the same axial air velocity, Ua), forming a long flame tail stabilized on the conical base flame. Once the flame base is extinguished, the flame lifts off. Implementing a quarl to the burner exit dramatically extends both the liftoff and blowout limits; this enhancement of the liftoff limit is more significant at high Ua. This is also true when Ua is zero (no coaxial air) where the blowout fuel jet velocity increases from 54 m/s of nonquarl flame to 62.3 m/s of quarl flames. The liftoff mechanism at low Ua (left of the black arrow) in the quarl flame is similar to the non-quarl flame. However, at high Ua, the base of the non-quarl flame is extinguished whereas the base of the quarl flame is not extinguished (details of the flame structures will be discussed in Section 4.2). This shows that the addition of the quarl significantly improves flame

jets, the air tube, and the fuel tube are included in the computation; the topology of the computational domain is identical to that of ref. [19]. Grid sensitivity studies are carried out and it is found that much finer grid is needed for the flame cases with high fuel jet speed. Two different grids are therefore used in the results presented here: for the low-speed cases, the minimum grid size is 0.1 mm in the quarl and the reaction zone, and the total cell number is about 1.8 million. In the high fuel jet velocity cases, the minimum grid size around the quarl and the reaction zone is 0.05 mm; the grid size is 0.02 mm close to the wall of the fuel tube, with the total number of grid about 6 million. These grids are shown to give essentially grid-independent mean flame structures and mean flow fields. 4. Results and discussion 4.1. Flame stability and selected flames Fig. 2 shows the measured stability diagram of quarl and non-quarl swirl stabilized non-premixed flames. By keeping the ratio of the tangential to axial air flow rates constant, the geometrical swirl number of the air flow was kept constant at Sg = 12 as the axial air velocity, Ua, (based on the bulk air flow rate at the air tube exit) was increased. On the vertical axis of the stability diagram, the fuel jet velocity, Uf, at the exit of the fuel tube, is presented against the axial air velocity, Ua, on the horizontal axis. The fuel jet velocity is increased gradually until 1231

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Table 1. As illustrated in Table 1, among the experimentally investigated cases, only flames at Uf = 4.3, 8.7, 55 and 100 m/s were selected for LES investigations. The flames are designated as FqUf for quarl flames and FnoUf for non-quarl flames. These values of the fuel jet velocity were chosen to investigate the changes in the flame structure and flow/mixing fields while the flames are approaching blowout conditions with the transition from attached flames to lifted flames. The fuel jet velocity influences the actual swirl number, S, at air tube exit. To quantify this effect, the simulated flow field at the burner tube exit where the fuel jet momentum is included, is used to calculate the actual swirl number for both quarl and non-quarl flames. The actual swirl intensity including the fuel jet momentum was defined as the ratio of the axial flux of the tangential momentum and the axial flux of the axial momentum [43]. The estimated actual swirl numbers are illustrated in Table 1, which clearly show a significant decrease in the swirl number with increasing fuel jet velocity for the quarl and non-quarl flames. The existence of the quarl at the burner exit leads to a slight increase in the swirl number. This indicates that the vertical axis of the stability plot (shown in Fig. 2) represents different swirl numbers at the same bulk air velocity where the higher the fuel jet velocity, the lower the swirl number. 4.2. Flame structure To study the flame structure and the transition from attached to lifted flames, OH-PLIF images were acquired. The OH-PLIF image extends from one side of the flame to the other, spanning approximately 65 mm. The laser sheet covers a region from the burner exit plane to approximately 60 mm downstream. In these OH-PLIF images, the dark region without OH at the central flame region shows gas at low to medium temperature, fresh fuel and/or fuel/air mixture and possibly some recirculated burned gas. The brightest OH regions indicate the region of super-equilibrium OH concentrations typically found at the flame front on the air side. The strong temperature gradients at the regions of the super-equilibrium OH concentrations results in strong gradients in the OH-PLIF images, thus, the steep gradient from dark inflow region to the maximum fluorescence is a good marker for the flame front [44]. OH-PLIF has strong signals; however, because it is a long-living intermediate at elevated temperatures, it appears in regions of high temperature in addition to areas of high reaction rates. The CH radical is short-lived intermediate, existing mostly in a small range of stoichiometry on the slightly rich side of hydrocarbon flames, and is a very specific tracer of the flame front. On the other hand, the PLIF of

Fig. 7. Radial axial velocity profiles normalized with jet fuel velocity, U/Uf, at different axial locations PIV/LES results for non-quarl flames (a-c) and corresponding quarl flames (b-d). PIV results are shown in solid black squares, while the solid line shows the LES results.

stabilization and allows the burner to provide higher heating power relative to the non-quarl burner. Based on the stability map, flames with and without quarl, at the same Sg = 12, and at Uf of 4.3, 8.7, 33, 55 m/s were selected for detailed PIV/OH-PLIF measurements. In addition, the flame with quarl at Uf = 100 m/s was also studied. For all investigated flames, Ua was kept constant at 4.9 m/s while Uf was varied. The main flow parameters of these selected flames and the corresponding designations are listed in

Fig. 8. The tangential velocity contours overlaid by the streamlines plot of flames Fno4.3 in (a) and Fq4.3 in (b), while the radial profiles of the tangential velocity component, W, at axial distance x = 47 mm of flames Fno4.3 and Fq4.3 are shown in (c). 1232

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Fig. 9. Mean velocity vectors superimposed on the instantaneous OH-PLIF images: (a) flames Fno4.3, (b) flame Fno33. The average maximum principle strain rates of the region bounded by the blue rectangular in (a) and (b) are extracted in (c) for flame Fno4.3 and in (d) for flame Fno33. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 10. Mean velocity vectors superimposed on the instantaneous OH-PLIF images: (a) flames Fq4.3, (b) flame Fq33. The average maximum principle strain rates of the region bounded by the blue rectangular in (a) and (b) are extracted in (c) for flame Fq4.3 and in (d) for flame Fq33. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

bounded by the OH layer, the oxygen concentration is relatively very low (the detailed mixing field will be discussed in Section 4.3). This inner region is dominated by the fuel jet spreading. As indicated in Fig. 5a, a fairly similar OH distribution can be found from the LES and the OH-PLIF imaging presented in Fig. 3a. As the fuel starts to penetrate through the recirculation zone with increasing Uf, shown in flame Fno8.7, Figs. 3b and 5b, the flame becomes elongated, and a long tail of OH layer is observed, forming a neck on the top of the conical OH base. Flame Fno33, as shown in the OH-PLIF image in Fig. 3c, displays a shrinking conical base, with frequent local extinction and re-ignition at the flame neck zone, and a highly contoured structure is observed downstream. With further increase of Uf, liftoff occurs when the conical flame base is extinguished, as shown in Fig. 3d, flame Fno55, where flame kernels at the base of the lifted flame can be seen.

OH is a good indicator of the actual position of the transition from unburnt to burnt regions of flames, such as in lifted flames [45], or to investigate the flame front-flow field interactions [46]. The instantaneous OH-PLIF images for four non-quarl flames (Fno4.3, Fno8.7, Fno33, and Fno55) are shown in Fig. 3 and the corresponding quarl flames as well as flame Fq100 are presented in Fig. 4. The OH mass fraction field from LES for the flames with and without quarl is shown in Fig. 5 for comparison. Flame Fno4.3, shown in Figs. 3a and 5a, displays the structure of the typical compact swirl flame with a thin conical OH layer laying on the boundaries of the recirculated flow (cf. Fig. 6a for information about the flow field) and is wrinkled and becomes comparatively thicker downstream. This thin conical OH layer confines a dark, relatively cold, central region. In this flame configuration, it was observed that within this dark inner region 1233

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Fig. 11. Instantaneous distribution of the mixture fraction (left), scalar dissipation rate (middle), and the temperature (right) from LES for Uf = 4.3 m/s. The stoichiometric mixture fraction contour is imposed on the images and is denoted by the white contour line. The upper row shows non-quarl flame while the bottom row for quarl flame.

Fig. 12. Instantaneous distribution of the mixture fraction (left), scalar dissipation rate (middle), and the temperature (right) from LES for Uf = 100 m/s. The stoichiometric mixture fraction contour is imposed on the images and is denoted by the white contour line. The upper row shows non-quarl flame while the bottom row for quarl flame.

The LES results for Fno55, given in Fig. 5c, show local extinction in the middle height of the flame, while the base flame in the air flow tube is still active. With further increase of the fuel jet flow velocity, e.g., to Uf = 100 m/s, the base flame is quenched, cf. Fno100, Fig. 5d. The flame structure with quarl is very different from the non-quarl flames, as shown by the instantaneous OH-PLIF images of the quarl flames in Fig. 4a–e, and the corresponding LES results in Fig. 5e–h, especially at high Uf. These flames show a radially wider conical OH layer (relative to the initial conical cone in the case of non-quarl flames) along the quarl wall that becomes corrugated and thicker downstream. This is due to the change in the flow structure resulting from the quarl. Downstream of the quarl rim, the OH layer is pushed towards the burner centerline at the low fuel jet velocity, i.e. flame Fq4.3, Figs. 4a and 5e. As will be indicated by the mean flow streamlines in Fig. 6d, the existence of the reversed flow at the quarl corner corrugates the flame sheet, leading to enlarged flame area, and strong convective mixing between burned and unburned gas. With increasing Uf, flame Fq8.7, Figs. 4b and 5f, the reaction zone confining the swirling flow inside the quarl continues outside the quarl, but resides primarily in the shear layer along the interface between the incoming unburned gas and the high temperature recirculating products. Further downstream, the OH

radicals start to be pushed away from the central flame region. Increasing the fuel jet velocity sustains the flow in the forward direction within the central flame region, and therefore there is less penetration of the back flow into the quarl exit. A further increase of Uf to 33 m/s shows that the reaction zone becomes corrugated near the location of vortex breakdown, where the vortex rolls up the flame sheet and the reaction zones become more corrugated, as indicated by OH-PLIF in Fig. 4c. Local extinction and flame fragmentation occurs at much higher Uf relative to the non-quarl flames. Flame Fq55 in Figs. 4d and 5g, shows intermittence of local extinction and re-ignition at the neck formed between the conical OH layer and the downstream jet-like OH layers. The downstream OH structure shows a highly contoured and fragmented flame structure, and local extinction appears within the conical zone inside the quarl. In all the flames investigated, it is clear that OH radicals close to the burner exit do not start from the tip of the fuel tube, but rather, they start radially shifted from the fuel tube nozzle. This radial shift

1234

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Fno33, and Fno55), and their quarl counterparts, as well as Fq100 flames, are shown in Fig. 6. Flame Fno4.3, Fig. 6a, shows a round fuel jet penetrating a swirling flow, resulting in the formation of a coneshaped inflow jet confining a pair of large-scale recirculation zones, identified as RZA and RZB; the recirculation zone is identified as the region where the axial velocity is less than zero (U < 0). The sudden expansion of the swirling jet creates a toroidal vortex, RZA. An outer shear layer of high velocity fluctuation is formed between the recirculation zone, RZA, and the high velocity fuel jet. The interaction between RZA and RZB causes the fuel to be transported to the outer shear layer where it can penetrate upstream. With progressive increases in fuel jet velocity in the non-quarl flames, the recirculation zone RZB attenuates rapidly along with a gradual decrease of RZA size, see Fig. 6b and c for flames Fno33 and Fno55, respectively. Referring to the OH-PLIF distribution shown in Fig. 3, it is obvious that the conical OH layer tracks the zero velocity contour of the recirculation zone RZA and it becomes fragmented downstream at high fuel jet velocity where the fuel jet momentum dominates the flow field. The downstream quarl flow field of flame Fq4.3 (Fig. 6d) still shows the existence of both RZA and RZB, and they are extended longer downstream (taking into consideration that the PIV measurement is only conducted in the region downstream the quarl as indicated in Fig. 1a), indicating that long recirculation zones dominate the inner flow region of the quarl flame low into the quarl. The LES radial velocity profiles show that the recirculation zone reaches even down into the air tube. Thus, it reduces the effect of the fuel jet momentum on the swirling flow. Comparing the streamline plots of quarl and non-quarl flames with increasing Uf illustrates the stronger effect of the fuel jet in altering the flow pattern in non-quarl flames than quarl flames, specifically, fast and early attenuation of the recirculation zones in non-quarl flames. For example, the RZB zone at Uf = 33 m/s, Fig. 6b, is completely absent for non-quarl flames, while it is clearly visible in the corresponding quarl flame shown in Fig. 6e. Radial profiles of the mean normalized axial velocity from both PIV and LES at several axial locations (from inside the burner air tube to downstream of the exit plane for quarl flames) for two fuel jet cases of Uf = 4.3 and 100 m/s are shown in Fig. 7. Velocity profiles at three axial locations downstream the exit plane of the quarl and five axial locations downstream the burner tip for non-quarl flames are shown. The velocity profiles are normalized with the corresponding bulk fuel jet velocity, Uf. In general, the LES velocity profiles for the non-quarl flame Fno4.3, illustrated in Fig. 7a, compare very well with the experimental PIV results. For the velocity profiles of the quarl flame, Fq4.3 in Fig. 7b, the LES shows some over-predication of the velocity at the core flame region, while the overall flow features, peak values, and their locations from the PIV and LES are consistent with each other. The flame luminosity of the quarl and non-quarl flames indicates that the non-quarl flames are less sooty than their corresponding quarl flames. The LES is done without considering thermal radiation, which results in an over-prediction of the flame temperature, thus under-prediction of the density field, which may contribute to the discrepancies between the LES results and the experiments. The error attributed to the adiabatic flame assumption in the LES is expected to be larger in the quarl flames, since the quarl flames have more radiation heat loss due to the higher level of soot in the flame. As indicated in Fig. 7a and b in both the quarl and the non-quarl flames at Uf = 4.3 m/s, as the flow emerges from the air tube tip, it starts to spread radially outwards owing to the presence of the swirl motion. This radial spread can be tracked by monitoring the radial position of the peak axial velocity, which moves radially outwards downstream the air tube tip and it continues for a long downstream distance in the quarl flame. The flow fields show that in both cases, a recirculation zone exists in the air tube upstream the exit plane of the fuel nozzle. In both the quarl and non-quarl flames, the recirculation zone is seen to penetrate deeply into the air tube, where a deeper penetration is observed with the quarl flame. In addition, it is obvious that

Fig. 13. Instantaneous distribution of the mass fractions of HO2, H2O2, CH2O and the CO from LES for Uf = 100 m/s. The stoichiometric mixture fraction contour is imposed on the images and is denoted by the white contour line. The upper row shows non-quarl flame while the bottom row for quarl flame.

gradually decreases with increasing Uf. Figs. 4e and 5h, show the structures of flame Fq100, and illustrate flame liftoff at high Uf; the flame lifts off when the intermediate flame section and some portion of the conical flame base are extinguished, as showed in Fig. 4e. In the LES results, a small flame can still be seen slightly upstream of the fuel nozzle, whereas in the PLIF experiments a stronger base flame can be observed inside the quarl. LES successfully replicates the trend of local extinction at the middle height of the flame and subsequent extinction of the flame base with increasing Uf for both flames with and without the quarl. 4.3. Flow structures and mixing field In this section, the flow structure and the mixing field are discussed. Time-averaged PIV streamline plots of three non-quarl flames (Fno4.3, 1235

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Fig. 14. Radial profiles of the mean mixture fraction, the scalar dissipation rates, and the mean temperature at several axial locations for flame Fno4.3 (left column) and flame Fq4.3 (right column).

the existence of the quarl extends the recirculated flow region for a longer distance downstream of the fuel tube exit than the non-quarl case. This recirculated flow inside the air tube is the reason for the flame being anchored upstream of the fuel tube. This effect is more pronounced in the quarl flame. Clearly, the quarl shifts the peak axial velocity radially outward and thus it triggers a wider recirculation zone, which leads to faster attenuation in the peak axial velocity. For the high Uf flames, e.g. Uf = 100 m/s, Fig. 7c and d, where the flow field is dominated by a high central fuel jet velocity, the recirculation zone in the non-quarl flame is absent; however, in the quarl flame, there is still

a weak recirculation zone near the fuel nozzle exit. These results clearly indicate that the existence of a divergent nozzle at the burner exit affects the flow pattern of the swirling flow. The existence of the quarl at the burner exit influences the tangential velocity distribution as shown in Fig. 8a and b by the mean tangential velocity contours of flames Fno4.3 and Fq4.3. It is obvious that the peak tangential velocity downstream of the quarl is shifted radially outward. This indicates that the quarl conserves the swirling flow momentum for a longer distance downstream of the quarl. This reduces the radial pressure gradient and hence smaller adverse gradient

1236

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Fig. 15. Radial profiles of the mean mixture fraction, the scalar dissipation rates, and the mean temperature at several axial locations for flame Fno100 (left column) and flame Fq100 (right column).

along the axis of the quarl flames. This explains the observed radial shift of the recirculation zone; hence, the observed longer/wider recirculated flow region in the quarl flames. In Fig. 8c, the tangential velocity along the radial positions of downstream the fuel nozzle at x = 47 mm for the flames with and without the quarl at Uf = 4.3 m/s are presented. The radial profile of the tangential velocity in the non-quarl flames displays a peak tangential velocity at a radial distance of r ≈−13 mm. The tangential velocity increases linearly with r (for the region from r = 0 to −13 mm), reflecting the solid body rotation part of the vortex. However, in the quarl flames, the radial tangential velocity profile

indicates two rates at which the tangential velocity is increased with the radial distances, where it shows a steep increase in the tangent velocity at r ≈ –22 mm. This explains the stronger and closer RZA to the central flame region around the fuel jet in the case of the non-quarl flames. The peak tangential velocity in the quarl flame case is shifted to a larger radius. The diverging quarl shows the same flow features relative to the straight quarl in enlarging the downstream recirculation zone as observed by Elbaz and Roberts [18]. In order to better identify the effect of the quarl on the flow field and flame structure, the vector plots of the average flow fields together 1237

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Fig. 16. Probability density function, pdf, of the scalar dissipation rate conditions on the stoichiometric mixture fraction χ[Zst]: (a) for the flames at Uf = 4.3 m/s with and without quarl over the domain from the fuel nozzle exit to 40 mm downstream, (b) for flames at Uf = 100 m/s with and without quarl (b) over the domain from the fuel nozzle exit to 50 mm downstream.

with the instantaneous OH-PLIF of flames at Uf of 4.3 and 33 m/s for non-quarl and quarl flames are illustrated in Figs. 9 and 10, respectively. In addition, in Figs. 9 and 10 the maximum component of the principle strain rate (2D) of the flame region bounded by the blue rectangular in the velocity field are extracted and plotted. The maximum principal strain rates are all positive, corresponding to extensive strain. Fig. 9a shows the OH-radicals to reside primarily in regions of low axial velocity for flame Fno4.3. Such regions tend to occur along the interface between high axial-velocity of the inflow jet (i.e unburned gas from the burner nozzle), and the lower or negative-magnitude axial velocity fluid of the recirculation zone. The flame is anchored by the early fuel-air mixing in the air tube. Unburned incoming gas is seen from the burner nozzle on one side and the high-temperature, combustion-radical laden burned gas transported from downstream by the relatively large RZ (located between the central fuel jet and the incoming swirling jet flow). Referring to Fig. 9a and the strain rate plotted in Fig. 9c, it is obvious that regions of high strain results from the high axial velocity, seen in the vicinity of the inflow jet shear layers as well as in the central jet shear layer. This indicates that the reversed flow region bounded by the boundaries of the recirculation zone has a local minima in strain rates. It is also seen that the regions of high strain rate of both swirl flow jet and central jet are merged downstream of the recirculation zone while at the same time the average peak strain rates decrease downstream. These results suggest that the flame is stabilized in regions of low strain rate and follow the passage of the high levels of strain around the outer periphery of the recirculation zone. It is remarkable that the flame sheet is oriented orthogonal to the axis of the principle strain rates at the shear layer of the swirling jet. The longer recirculation zone associated with the quarl extends downstream of the quarl exit and is close to the quarl edges, as illustrated in Fig. 10a–c for flame Fq4.3 and causes the OH layers to reside inside the central region of the low axial velocity and low strain rate, where the central jet is dissipated. With increasing fuel jet velocity, the flow field downstream of the burner nozzle or the quarl is altered along with the flame structure, as shown in Figs. 9b and 10b. This causes a reduction in the size of the RZ associated with flame Fno33, and therefore less penetration of the back flow into the burner air tube. Consequently, the early OH layer resides primarily in the shear layer along the swirling flow jet. Further downstream near the location of the vortex breakdown, the vortex is able to roll up the reaction zone towards the central jet where a higher strain rate is observed (Fig. 10d). This leads to a fragmentation of the OH layer and the appearance of the isolate OH-spots with frequent local extinction and spontaneous re-ignition at the base of the lifted flame, as indicated in the downstream location in Fig. 3c. At Uf = 33 m/s in the quarl flame the central jet had sufficient momentum to sustain flow in the downstream direction within the central flame region, and the level of strain

rates is seen to increase within the recirculation zone (see Fig. 10b–d). Thus, the OH layer confining the swirling flow inside the quarl continues outside the quarl, but lays mainly in vicinity of the swirling jet’s shear layer and it shows a more corrugated structure near the location of the vortex breakdown due to the higher turbulence. The difference in flow pattern between the quarl and non-quarl flames leads to significantly different mixing fields. The mixing field of two flames at Uf = 4.3 and 100 m/s are examined to illustrate this. The results from LES of mixture fraction, the scalar dissipation rate, and temperature field at Uf = 4.3 m/s and 100 m/s are shown in Figs. 11 and 12, respectively. At Uf = 4.3 m/s, Fig. 11, the mixture fraction distribution reveals the enhancement and early mixing close to the fuel nozzle in the quarl flame. This is attributed to the large reversed flow inside the quarl, convecting the fuel upstream into the air tube. The stoichiometric mixture fraction iso-contour is found to begin in the air tube at an upstream position relative to the non-quarl flame. It is also obvious from the mixture fraction distribution that the fuel jet has more spreading into the swirling air. This result explains the OH-PLIF /LES structures of the flames shown in Figs. 3–5. Comparatively, in the nonquarl flames, the axial velocity shows a slow decay, indicating less fuelair mixing. The weak effect of the recirculation zone close to the fuel jet exit leads to the confinement of the fuel to a narrow region around the burner axis, which affects the scalar dissipation rate distribution, χ. The scalar dissipation rate characterizes the local scalar gradients within the flow field and is defined as, χ = 2(DSGS + D) ∇Z∙∇Z , where D is the thermal diffusion coefficient of the local mixture, DSGS is the sub-grid scale eddy diffusion coefficient, and Z is the mixture fraction on the resolved scale. The scalar dissipation rate, χ, is the inverse of a characteristic diffusion time scale imposed by the mixing field [47]. From one-dimensional laminar flame calculation using the current chemical kinetic mechanism, and under counter-flow diffusion configurations, the quenching scalar dissipation rate, χcrt, is 35 s-1 [19]. As shown in both cases, with and without quarl, the scalar dissipation rate in the vicinity of the fuel nozzle is higher than the critical quenching scalar dissipation rate χcrt, which would prevent a flame at the fuel nozzle tip. In both cases, the flames are located in the air tube around the stoichiometric mixture fraction (where χ ≈ 2 s-1) rather than the shear layer of the fuel jet. Thus, the locally high scalar dissipation rate in the vicinity of the fuel jet has no significant impact on the flame. For flames at Uf = 100 m/s, the mixture fraction illustrated in Fig. 12 shows the fuel jet’s early radial spreading inside the quarl. In the non-quarl case, less spreading near the fuel tube exit is observed, consistent with the radial velocity profiles shown in Fig. 7c and d. The absence of a recirculation zone in the non-quarl flame coupled with a less spreading jet leads to a higher scalar dissipation rate, χ > χcrt, eliminating the possibility of a flame existing at the fuel nozzle, and 1238

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

Fig. 17. Scatter plots in temperature-mixture fraction space, and scalar dissipation rate-mixture fraction space for Uf = 4.3 and 100 m/s of quarl and non-quarl flames.

1239

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

presented in Fig. 15a–d for the quarl and non-quarl flames. They indicate that the non-quarl flame reaches a larger Zmean at x = 22 mm with a very steep decline at the fuel jet boundaries. This is explained by the high axial centerline velocity and the reduced jet spreading in the no-quarl jet close to the fuel nozzle. The shorter time for fuel/air mixing, especially with the disappearance of the recirculation zone, is shown in the streamlines plots, Fig. 6, and the radial velocity profiles, Fig. 7. This distribution leads to a high scalar dissipation rate associated with the stoichiometric mixture fraction contour as indicated in Fig. 15b. On the contrary, the existence of the quarl leads to more rapid spreading of the fuel jet and a gradual decrease of the mixture fraction at the jet boundaries, which reduces the scalar dissipation rate, Fig. 15d, at the iso-contour of the stoichiometric mixture fraction. This explains the observed OH radicals in the early part of the quarl in flame Fq100, Fig. 4 as well as the relatively high temperature noticed inside the quarl, Fig. 15e. Further downstream at x = 150 mm, flame Fq100 shows a higher temperature at r ≈ 22 mm than flame Fno100. More details of the comparison between the quarl and non-quarl flames at high and low fuel jet velocity are seen in the pdf plots of the scalar dissipation rates conditioned on the stoichiometric mixture fraction presented in Fig. 16. The pdf for the quarl flame at Uf = 4.3 m/ s, Fig. 16a, has a sharp peak at relatively low scalar dissipation rate which then steeply decreases to a moderate value with a plateau distribution towards the high scalar dissipation rate. This is due to the mixture fraction redistribution by the early formed recirculation zone. Then, this plateau gradual decreases at the higher χ. The peak value of the pdf is located at low χ showing a low probability in the non-quarl flame as compared with the quarl flame. For the high-velocity flames, Uf = 100 m/s, the quarl flame still shows a sharp pdf peak at low χ with a tail towards the high χ. Removing the quarl reduces the peak value of the pdf at low χ with a moderate decline in the direction of high χ. This clearly explains how the addition of the quarl at the exit of the swirl flames burner adds a stabilization effect to the flames. More information about the flame structure can be observed from Fig. 17, which shows the scatter plots of the flow points in temperature/ mixture fraction space, T-Z, and scalar dissipation rate/mixture fraction space, χ-Z, for the quarl and non-quarl flames at Uf = 4.3 and 100 m/s. The data point of each dot in Fig. 17 corresponds to the flow point in the three-dimensional flow field shown in Fig. 11 for flames at Uf = 4.3 m/s and in Fig. 12 for flames at Uf = 100 m/s. Features of turbulent diffusion flames are clearly evident from the T-Z plots for the low velocity flames with a relatively broader distribution of temperature in the rich-mixture of the quarl flame. However, for flame Fno100, Fig. 17 shows that the rich mixtures are at low temperatures (which corresponds to the chemically inert mixing in front of the liftoff position shown in Fig. 13). It is also seen that the scalar dissipation rate around Zst is higher than that of flame Fq100. The T-Z plot of flame Fq100 illustrates a relatively high temperature in the fuel-rich mixtures, and a further increase in temperature around Z = 0.15, which is attributed to the partial oxidization/ignition of the mixture upstream of the lifted flame shown in Figs. 12 and 13. The existence of the quarl increases the burnable range of the mixture fraction below the extinction scalar dissipation rate at high or low Uf, attributed to the wider jet spreading observed in both the flow field and mixture fraction distribution. This explains the enhancement of both the liftoff and blowout limits with the existence of the quarl.

hence the flame eventually lifts off. On the contrary, in the quarl flame, Fq100 shown in Figs. 4, 5, 7 and 12, there is still a weak recirculation zone in the air flow tube upstream of the fuel nozzle, which sustains a small base flame. Although this weak recirculation zone is unable to sustain a large flame as both the scalar dissipation rate and jet velocity are high, the low speed flow (with weak recirculation) inside the air tube allows for the mixture to partially oxidize, releasing heat and forming a large amount of combustion intermediates such as CO, HO2, H2O2, and CH2O in the region upstream the liftoff re-ignition location, cf. Fig. 13. In non-quarl cases, the flame is located around the stoichiometric mixture in the fuel jet shear layer downstream, where both the scalar dissipation rate and jet velocity are low. The temperature upstream the liftoff position in the non-quarl case is equal to the ambient fuel/air temperature; and no combustion intermediates could be observed in the mixture upstream the liftoff location, cf. Fig. 13. To yield more insight into the main features of the attached and lifted flames with and without quarl, the mean radial profiles at various axial locations from the fuel nozzle for the mixture fraction, Zmean, the mean scalar dissipation rate, χmean, and the mean flame temperature Tmean in flames at Uf = 4.3 m/s and 100 m/s are presented in Fig. 14and 15 , respectively. Also, Fig. 16 illustrates the probability density function, pdf, of the scalar dissipation rate conditioned on the stoichiometric mixture fraction, χ[Zst], in the same two flames of Uf = 4.3 and 100 m/s. The pdfs were calculated for the domain from the fuel nozzle exit to x = 40 mm downstream for the low fuel velocity flames (Uf = 4.3 m/s), and up to x = 50 mm for the high fuel jet velocity flames (Uf = 100 m/s). At x = 5 mm, the radial profiles of Zmean displayed in Fig. 14a and d of flames at Uf = 4.3 m/s for quarl and nonquarl, show similar profiles, where the Zmean peaks are found directly above the fuel nozzle exit, as expected. It is seen, however, that at x = 22 mm, the pattern of Zmean for quarl flame relative to non-quarl flame is much broad at the expense of mixture fraction at the central jet. This demonstrates the initial fast mixing resulting from the existence of an early wide recirculation zone with the quarl flame. On the contrary, further downstream at x = 44 mm, a relatively higher Zmean along the central region of the quarl flame than that of the non-quarl indicates an increase of mixing rates with the non-quarl flame, gaining from the strong recirculation zone formed on the central flame region downstream the fuel nozzle, as illustrated in Fig. 6 and the radial velocity profiles in Fig. 7. The features of the Zmean distribution are in agreement with the early OH-PLIF distribution (see Figs. 3 and 4), indicating that mixing and the main flame reactions are closely linked. It is further seen that in the near field of the nozzle, the mean mixture fraction distribution is considerably higher in the boundary of the wider recirculation zone with quarl, which enhances the stabilization effect of the RZ because they enable high temperature recirculated gases to mix with the inflow mixtures. The radial profiles of the mean scalar dissipation rate, χmean, indicate higher peak values at the fuel jet shear layer at x = 5 and 22 mm for the non-quarl case relative to the quarl flames and that close to the fuel nozzle, see Fig. 14b–e. On the other hand, the distribution of the mean temperatures displayed, in Fig. 14c–f, reflects the difference in mixing between the quarl and non-quarl flames. At x = 5 mm, the two flames exhibit a similar temperature distribution, where low-temperature regions are corresponding to the inlet of the fuel jet and it increases gradually at the jet shear layer, reaching the same peak temperature at r ≈ 8 mm. This demonstrates the dominant effect of the fuel jet close to the fuel nozzle. At x = 22 mm, the temperature profiles show a significant change in both flames, owing to the radial spreading of the mixture fraction in the quarl flame. The location of the peak flame temperature is gradually shifted in the outward direction (reaching a peak value at r = 21 mm, for x = 44 mm) with a low temperature in the central flame region. On the contrary, owing to the fast mixing at x = 44 mm in the no-quarl flame, a relatively uniform temperature is observed in the central flame region (r = 0 to 5 mm) with a gradual decay at the flame boundaries. The radial distribution of Zmean for flames at Uf = 100 m/s,

5. Conclusions The blowout limits of non-premixed swirl flames with and without quarl were measured experimentally. For the same geometric air flow swirl number (Sg = 12), gradually increasing the fuel jet velocity, Uf, (from Uf = 4.3 to 100 m/s) gives rise to the liftoff of the attached flames and then blowout at higher Uf. The quarl extends both the flame liftoff and blowout limits. To investigate the reasons for this, flame structure and flow fields for the quarl and non-quarl flames were studied from 1240

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al.

funding from King Abdullah University of Science and Technology (KAUST), Saudi Arabia. The work at Lund University is sponsored by Swedish Research Council (VR) and Swedish national center for combustion sciences and technologies (CeCOST). Senbin Yu and Xiao Liu were sponsored by China Scholarship Council (CSC). The computations were performed on resources provided by the Swedish National Infrastructure for Computing (SNIC) at HPC2N and PDC.

low to high Uf; using a combination of PIV/OH-PLIF measurements and large eddy simulations (LES) with a transported PDF combustion model. The main findings are:

• The primary features of the mean flow field are two recirculation









zones, the first zone located at the shear layer of the fuel jet, while the second zone is radially shifted to the swirling inflow jet. In both quarl and non-quarl flames, compact flames are observed at low Uf and turn into elongated tail flames with the gradual increase of Uf. The local extinction and re-ignition events start to occur with the increase of fuel jet velocity in the non-quarl flames at lower Uf compared to the quarl flames. At low Uf relative to the non-quarl flames, the quarl enhances the reversal flow reaching far upstream, promoting fuel-air mixing inside the air tube. This causes the stoichiometric mixture fraction contour to extend into the air tube upstream of the fuel tube exit, where both the scalar dissipation rate and the flow velocity are low. Thus leading to a stable diffusion flame anchored inside the air tube. Contrarily, in the non-quarl flame, the relatively weak recirculated flow delays the fuel-air mixing, hence increasing the scalar dissipation rate close to the stoichiometric mixture fraction where the flame is established. At high Uf, where both the quarl and non-quarl flames are lifted, the flow pattern is dominated by the central jet in the non-quarl flame; showing significant attenuation in the recirculated flow. This generates much higher scalar dissipation rates, and together with the higher flow velocity close to the fuel nozzle, it prevents the diffusion flames from being stabilized close to the fuel nozzle exit. Partial oxidation of the fuel/air mixture and the earlier fuel jet spreading provide more burnable mixture fraction at the base of the lifted quarl flames, resulting in wider liftoff and blowout limits. The Pdf plots of the scalar dissipation rates conditioned on the stoichiometric mixture fraction in the region close to the fuel nozzle showed a higher probability at low scalar dissipation rate for quarl flames relative to the non-quarl at Uf = 4.3 and 100 m/s. In addition at Uf = 4.3 m/s, due to the mixture fraction redistribution by the early formed recirculation zone, an attenuation in the pdf distribution is noticed towards the moderate and high scalar dissipation rate in the quarl flames. The scatter plots in temperature-mixture fraction space, T-Z, and scalar dissipation rate-mixture fraction space, χ-Z show significant differences between the low Uf (4.3 m/s) compact flames and the high Uf (100 m/s) lifted flames in both the quarl and non-quarl flames. The T-Z plots indicate a relatively wider distribution of the rich mixture in the high temperature regions at low Uf quarl flames. However, the non-quarl flame at Uf = 100 m/s indicates a rich mixture at low temperatures, representing the inert chemical mixture at the base of the lifted flame. On the contrary, in the quarl high Uf flame, a high temperature in the fuel-rich mixtures followed by a further increase in the temperature around Z = 0.15 are observed, indicating the partial oxidation/ignition of the mixture in the front of the lifted quarl flame.

References [1] Syred N, Chigier N, Beer J. Flame stabilization in recirculation zones of jets with swirl. Proc Combust Inst 1971;13(1):617–24. [2] Schmittel P, Günther B, Lenze B, Leuckel W, Bockhorn H. Turbulent swirling flames: Experimental investigation of the flow field and formation of nitrogen oxide. Proc Combust Inst 2000;28:303–9. [3] Tangirala V, Chen R, Driscoll JF. Effect of heat release and swirl on the recirculation within swirl-stabilized flames. Combust Sci Technol 1987;51:75–95. [4] Chen RH, Driscoll JF. The role of the recirculation vortex in improving fuel-air mixing within swirling flames. Proc Combust Inst 1989;22(1):531–40. [5] Feikema D, Chen RH, Driscoll JF. Enhancement of flame blowout limits by the use of swirl. Combust Flame 1990;80:183–95. [6] Feikema D, Chen RH, Driscoll JF. Blowout of non-premixed flames: maximum coaxial air velocities achievable, with and without swirl. Combust Flame 1991;86:347–58. [7] Al-Abdeli YM, Masri AR. Stability characteristics and flow fields of turbulent nonpremixed swirling flames. Combust Theor Model 2003;7:731–66. [8] Al-Abdeli YM, Masri AR. Recirculation and flow field regimes of unconfined nonreacting swirling flows. Exp Therm Fluid Sci 2003;27:655–65. [9] Masri AR, Kalt PA, Barlow RS. The compositional structure of swirl-stabilised turbulent nonpremixed flames. Combust Flame 2004;137:1–37. [10] Al-Abdeli YM, Masri AR, Marquez GR, Starner SH. Time-varying behaviour of turbulent swirling nonpremixed flames. Combust Flame 2006;146:200–14. [11] Kalt PA, Al-Abdell YM, Masri AR, Barlow RS. Swirling turbulent non-premixed flames of methane: flow field and compositional structure. Proc Combust Inst 2002;29:1913–9. [12] Masri AR, Kalt P, Al-Abdeli Y, Barlow R. Turbulence–chemistry interactions in nonpremixed swirling flames. Combust Theor Model 2007;11:653–73. [13] Elbaz AM. Early structure of LPG partially premixed conically stabilized flames. Exp Therm Fluid Sci 2013;44:583–91. [14] Mansour MS, Elbaz AM, Samy M. The stabilization mechanism of highly stabilized partially premixed flames in a concentric flow conical nozzle burner. Exp Therm Fluid Sci 2012;43:55–62. [15] Milosavljevic V, Taylor A, Whitelaw J. The influence of burner geometry and flow rates on the stability and symmetry of swirl-stabilized nonpremixed flames. Combust Flame 1990;80:196–208. [16] Shen D, Most JM, Joulain P, Bachman JS. The effect of initial conditions for swirl turbulent diffusion flame with a straight-exit burner. Combust Sci Technol 1994;100:203–24. [17] Becker LG, Kosaka H, Böhm B, Doost S, Knappstein R, Habermehl M, et al. Experimental investigation of flame stabilization inside the quarl of an oxyfuel swirl burner. Fuel 2017;201:124–35. [18] Elbaz AM, Roberts WL. Investigation of the effects of quarl and initial conditions on swirling non-premixed methane flames: flow field, temperature, and species distributions. Fuel 2016;169:120–34. [19] Liu X, Elbaz AM, Gong C, Bai XS, Zheng H, Roberts WL. Effect of burner geometry on swirl stabilized methane/air flames: a joint LES/OH-PLIF/PIV study. Fuel 2017;207:533–46. [20] Raman V, Pitsch H. Large-eddy simulation of a bluff-body-stabilized non-premixed flame using a recursive filter-refinement procedure. Combust Flame 2005;142:329–47. [21] Kempf A, Lindstedt R, Janicka J. Large-eddy simulation of a bluff-body stabilized nonpremixed flame. Combust Flame 2006;144:170–89. [22] Pitsch H. Large-eddy simulation of turbulent combustion. Annu Rev Fluid Mech 2006;38:453–82. [23] Gicquel LY, Staffelbach G, Poinsot T. Large eddy simulations of gaseous flames in gas turbine combustion chambers. Prog Energy Combust Sci 2012;38:782–817. [24] Pope SB. PDF methods for turbulent reactive flows. Prog Energy Combust Sci 1985;11:119–92. [25] Haworth D. Progress in probability density function methods for turbulent reacting flows. Prog Energy and Combust Sci 2010;36:168–259. [26] Syred N, Beer J. Combustion in swirling flows: a review. Combust Flame 1974;23:143–201. [27] Ribeiro M, Whitelaw J. Coaxial jets with and without swirl. J Fluid Mech 1980;96:769–95. [28] Smagorinsky J. General circulation experiments with the primitive equations: I. The basic equations. Mon Weather Rev 1963;91:99–164. [29] Dopazo C. Probability density function approach for a turbulent axisymmetric heated jet-Centerline evolution. Phys Fluids 1975;18:397–404. [30] Valiño L. A field Monte Carlo formulation for calculating the probability density function of a single scalar in a turbulent flow. Flow Turb Combust 1998;60(2):157–72. [31] Jones WP, Prasad VN. Large Eddy simulation of the sandia flame series (D–F) using the Eulerian stochastic field method. Combust Flame 2010;157:1621–36.

In conclusion, this integrated experimental and numerical work provides a better understanding of the flow fields, flame structures, and mixing fields of the swirl stabilized non-premixed jet flames. Future work will focus on further investigations inside and downstream the quarl using simultaneous OH-PLIF/PIV measurements emphasizing the local extinction and re-ignition phenomenon which occurs at the base of the lifted swirl stabilized jet flame. With this information, a clearer understanding of the turbulence-chemistry interaction in the vicinity of the lifted swirl-stabilized flame would be possible. Acknowledgments The experimental work was supported by competitive research 1241

Fuel 236 (2019) 1226–1242

A.M. Elbaz et al. [32] Jones W, Kakhi M. PDF modeling of finite-rate chemistry effects in turbulent nonpremixed jet flames. Combust Flame 1998;115:210–29. [33] Bulat G, Jones W, Marquis A. Large eddy simulation of an industrial gas-turbine combustion chamber using the sub-grid PDF method. Proc Combust Inst 2013;34:3155–64. [34] Mustata R, Valiño L, Jiménez C, Jones W, Bondi S. A probability density function Eulerian Monte Carlo field method for large eddy simulations: application to a turbulent piloted methane/air diffusion flame (Sandia D). Combust Flame 2006;145:88–104. [35] Jones WP, Navarro-Martinez S. Large eddy simulation of autoignition with a subgrid probability density function method. Combust Flame 2007;150:170–87. [36] Jones WP, Marquis AJ, Prasad VN. LES of a turbulent premixed swirl burner using the Eulerian stochastic field method. Combust Flame 2012;159:3079–95. [37] Jangi M, Lucchini T, Gong C. Bai XS. Effects of fuel cetane number on the structure of diesel spray combustion: an accelerated Eulerian stochastic fields method. Comb Theor Model 2015;19:549–67. [38] Hodzic E, Jangi M, Szasz RZ, Bai XS. Large eddy simulation of bluff body flames close to blow-off using an Eulerian stochastic field method. Combust Flame 2017;181:1–15.

[39] Smooke M, Giovangigli V. Reduced Kinetic Mechanisms and Asymptotic Approximations for Methane-air Flames. Berlin: Springer-Verlag; 1991. [40] Smooke M, Giovangigli V. Premixed and nonpremixed test problem results, in: Reduced kinetic mechanisms and asymptotic approximations for methane-air flames. Berlin:: Springer-Verlag; 1991. [41] Smith GP, et. al. GRI 3.0 mechanism web site: < http://www.me.berkeley.edu/grimech/ > ; 2017. [42] OpenFoam, Openfoam foundation web site: < http://openfoam.org/ > ; 2017. [43] Sadanandan R, Stöhr M, Meier W. Simultaneous OH-PLIF and PIV measurements in a gas turbine model combustor. Appl Phys B 2008;90:609–18. [44] Roman W, Jacques D. Combustion accelerated swirling flows in high confinements. Prog Energy Combust Sci 1992;18:349–67. [45] Hasselbrink Jr E, Mungal M. Observations on the stabilization region of lifted nonpremixed methane transverse jet flames. Proc Combust Inst 1998;27:1167–73. [46] Tacke M, Geyer D, Hassel E, Janicka J. A detailed investigation of the stabilization point of lifted turbulent diffusion flames. Proc Combust Inst 1998;27:1157–65. [47] Peters N. Laminar flamelets concepts in turbulent combustion. Proc Combust Inst 1986;21(1):1231–50.

1242