air premixed flame

air premixed flame

Combustion and Flame 188 (2018) 199–211 Contents lists available at ScienceDirect Combustion and Flame journal homepage: www.elsevier.com/locate/com...

4MB Sizes 4 Downloads 57 Views

Combustion and Flame 188 (2018) 199–211

Contents lists available at ScienceDirect

Combustion and Flame journal homepage: www.elsevier.com/locate/combustflame

Effects of internal flue gas recirculation rate on the NOx emission in a methane/air premixed flame Baolu Shi a, Jie Hu b,∗, Hongwei Peng c, Satoru Ishizuka d a

School of Aerospace Engineering, Beijing Institute of Technology, No.5 ZhongGuanCun South Street, Haidian, Beijing, 100081, PR China Department of Mechanical Engineering, Kitami Institute of Technology, 165 Koen-cho, Kitami, Hokkaido, 090-8507, Japan c MIURA Co., Ltd, 7 Horie, Matsuyama, Ehime, 799-2696, Japan d Department of Mechanical Science and Engineering, Graduate School of Engineering, Hiroshima University, 1-4-1 Kagamiyama, Higashihiroshima, 739-8527 Japan b

a r t i c l e

i n f o

Article history: Received 21 April 2017 Revised 22 May 2017 Accepted 28 September 2017

Keywords: NOx emission Internal flue gas recirculation Recirculation rate Tubular flame

a b s t r a c t This study fundamentally investigates the effects of internal flue gas recirculation (IFGR) on the combustion characteristics of methane/air mixtures, particularly on the NOx formation. To induce IFGR, a prototype tubular flame burner with internal recirculation has been made, in which 12 recirculation paths are designed. The IFGR rate is controlled by varying the number (n) of opening recirculation paths (n = 0-12). To quantify IFGR rate and furthermore evaluate its influence on the NOx formation, the recirculated gas volume is determined by measurements of velocity distribution with a Particle Image Velocimetry (PIV) system. The results illustrate that by increasing n, the flame length increases while flame luminosity weakens, steady tubular flame range in equivalence ratio shrinks, and the flame temperature decreases. Temperature of recirculated burned gas drops significantly before diluting the fresh reactants. Detailed observations show that the measured IFGR rate is about 13% (convert to STP) when all the recirculation paths are opened (n = 12), in which the NOx concentration significantly decreases to almost a half. Owing to IFGR, the local oxygen mole fraction and equivalence ratio in the reaction zone decrease, which directly reduce the flame temperature and oxygen atom concentration, resulting in a significant decrease of thermal type NO formation. Variations of species concentrations with IFGR rate have been examined, and the obtained relations provide a useful guide for the operation of small boilers adopting IFGR, which can reduce NOx emission without great cost and space. © 2017 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction Nitrogen oxide (NOx ) is recognized as a critical hazard to environment and human health [1,2]. Recently legislative regulations for NOx emission are increasingly more stringent. Therefore, great efforts have been made to reduce NOx emission. Many effective methods have been proposed, including: (1) low excess air combustion [3,4], (2) staged combustion [5,6], (3) flue gas recirculation (FGR) [7,8], (4) low NOx burners with use of fuel staging, air staging or fiber-matrix, to reduce peak temperature [8,9], (5) oxy-fuel combustion [8,9], (6) reburning to convert NO to N2 [8], (7) selective catalytic and non-catalytic reduction [6,8], (8) high temperature air combustion [10], and (9) MILD combustion [11]. Among them, FGR is one of the most widely used methods of reducing NOx emission from combustion [8]. FGR usually includes two types, external and internal recirculation. In the external recir-



Corresponding author. E-mail address: [email protected] (J. Hu).

culation (EFGR), flue gas is recirculated outside of the furnace and added to the combustion air [8,12]; hence, the equipment may become larger and more expensive. Common EFGR applications include gasoline [12] and diesel engines [13]. It has been pointed out that EFRG can reduce fuel consumption, decrease NOx emission levels and inhibit knock in gasoline engines [12]. In the internal flue gas recirculation (IFGR), combustion products are recirculated inside the furnace back to the burner and added into the flame; hence, the equipment can be compact and costs a little. Thus, IFGR becomes an ideal method to reduce the NOx emission, particularly for small boilers which have been widely adopted in industries owing to higher boiler efficiency and economically optimal operation according to need. In a lab-scale boiler, Shinomori et al. [14] utilized peripherally arrayed air injection around a liquid fuel-spray nozzle to induce internal recirculation, in which the NOx emission decreased to a half. With a swirl, flue gas was recirculated into the flame zone to reduce the temperature gradient, improve reactants pre-heating and lean the mixture, positively influencing NOx emission [15]. Recently IFGR techniques are also adopted in gas turbine engines typically in the form of staged combustion [16,17]. In

https://doi.org/10.1016/j.combustflame.2017.09.043 0010-2180/© 2017 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

200

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

Nomenclature

β β d D Df

t Φ ΦU k1 Lf n

B R U QAir QB QM QR QU P0, P Ru T, Ti T0 Vxi X Y

internal flue gas recirculation rate (converted to STP condition), IFGR rate local oxygen mole fraction in the unburned gas diameter of the recirculation path diameter of the inner wall flame diameter residence time equivalence ratio in the fresh reactant equivalence ratio in the unburned gas (including fresh reactant and recirculated gas) reaction rate constant flame length number of opening recirculation path volume concentration in the burned gas volume concentration in the recirculated gas volume concentration in the unburned gas air flow rate in the fresh reactant flow rate of burned gas flow rate of fresh methane/air mixture flow rate of recirculated gas flow rate of unburned gas (including fresh reactant and recirculated gas) pressure universal constant temperature initial temperature axial velocity at a specific vertical position around open path inlet (i = 1,2,…,k) axial position vertical position

addition, in waste incineration, FGR is utilized to improve efficiency and reduce emissions [18]; in the oxy-fuel natural gas combustion, the reduction ratio of NO emission is approximately 85% with 40% FGR ratio, in which the effects of CO2 ratio and FGR rate on the NO emission are analyzed [19]. It is seen that FGR can significantly reduce the NOx emission in various combustion systems. Extensive studies have illustrated that FGR directly influences the flame temperature, oxygen concentration in the mixture and concentrations of NO-generating radicals associated with the flame front, which lead to variation of NOx emission [8]. These effects are closely correlated with the main composition of recirculated gas, which generally includes N2 , CO2 , O2 and H2 O. Hence the effects of dilution on NOx formation have been fundamental investigated in methane/air [20,21], ethylene/air [22] and syngas/air mixtures [23]. At atmospheric pressure, with constant temperature and residence time, in methane/air flames NOx emission (expressed in terms of CH4 emissions index) increased with dilution owing to an enhancement of prompt NOx pathway resulting from an increase in the CH radical concentrations; dilution with N2 yielded a higher NOx emission than dilution with CO2 [24]. While at a constant adiabatic flame temperature, a significant reduction in NOx emission was observed by H2 O and N2 dilution, and reducing effect of H2 O dilution was a factor of 2 larger than that of N2 dilution [25]. In ethylene/air flame, N2 dilution was found to decrease NO at any equivalence ratio [22]. With an opposed-jet burner, the combined effects of N2 , CO2 and H2 O dilution on NOx emission and flame reactivity were assessed at constant flame temperatures [21], in which FGR tends to increase the relative weight of prompt NO formation and reduce the contribution from thermal NO.

The impacts of FGR on NOx emission are recognized to be influenced by pressure. Li et al. [26] found that at a constant flame temperature FGR lead to an increase of NOx formation at pressures below 5 atm, while EGR reduced NOx emissions at pressures above 5 atm. In the study of ElKady et al. [27], with 35% FGR in a lean premixed CH4 flame at 10 atm, NOx emission decreased more than 50% at a constant flame temperature. In addition, in combustion systems with FGR, NOx formation is also found to be influenced by the degree of turbulence and residence time [22,28,29], however, their impacts are correlated with other parameters in a specific combustor configuration. Another important influence introduced by FGR is reaction chemistry. Vitiated air condition and moderate increase of CO2 concentration will influence the combustion reactivity and NOx emission. To identify the chemical structure and mixture composition on the NOx formation, quantitative measurements with high spatial resolutions have been conducted for both alkane and alcohols [29] as well as syngas and biogas blends [23]. Many chemical models have been examined and compared with measurements [29–32], which provide useful data for further development of mechanisms with NOx prediction for different fuels. In most of the aforementioned studies, a mixture or single gas of N2 , CO2 and H2 O is supplied to mimic the FGR oxidizer composition in the combustor. In this study, rather than adopting simulated FGR, a specially designed tubular flame burner is employed to assess the IFGR rate on NOx reduction, in which part of the burned gas is recirculated through internal recirculation paths to dilute the fresh reactants and participate the reaction. As a fundamental type of flame, during the past decades both the swirl type [33,34] and non-swirl type tubular flames [35] have been extensively studied experimentally and numerically. Particularly, owing to overwhelming thermal stability and aerodynamic stability as well as wide flammability limit from lean to rich, the swirl type tubular flame has been widely adopted in industrial fields, for example, burning low-heat-value byproducts from steel making [36], and obtaining a uniform and large-area laminar flame to heat iron slab or to reduce steel sheet surface [37,38]. Main progresses in tubular combustion have been summarized in literatures [39,40]. Recently a prototype, self-recirculation tubular flame configuration has been proposed to reduce NOx emission. In a 1.25-inch diameter burner, 8 recirculation paths were designed to induce IFGR in premixed methane/air combustion [41]. In the following study, to avoid hazards of flame flash back, the rapidly mixed type tubular flame was tested, in which methane and air were individually injected into the combustor [42]. Then in the kerosene combustion using 8 and 12-inch burners, the tangential ejector effect on NOx emission was addressed [43]. These burners provide a heat output between 8 and 100 kW, and the NOx emission significantly reduces by using either natural gas or kerosene as the fuel. However, the IFGR rate has not yet been quantitatively analyzed, and the essential mechanism for NOx reduction is neither examined. In the present study, a 1.25-inch diameter IFGR burner, in which the number of recirculation path is increased and tangential ejector has been optimized, is adopted to systematically investigate the flame characteristics. The IFGR rate can be conveniently adjusted by varying the number (n) of opening recirculation path. Flame appearances, steady combustion range, temperature distributions and concentrations of major concerned species such as NOx and CO are investigated. Particularly, based on flow velocity measurements with using a Particle Image Velocimetry (PIV) system, IFGR rate is quantitatively analyzed in the test. Then IFGR rate is gradually increased to discuss the variations of main species concentrations in the reaction zone, recirculation zone and burner exit; gas temperatures in the main reaction zone and recirculation path are also measured. Based on theoretical analysis and numerical simulation,

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

201

Fig. 1. Schematics of internal recirculation type tubular flame burner.

NOx formation with IFGR rate is examined. In addition, variations of main species concentrations with IFGR rates are investigated for the lean to stoichiometric methane/air combustion, which provides a useful guide to the optimal operation conditions of small boilers and radiant tubes utilizing IFGR. 2. Experimental and numerical methods The schematics of the burner are shown in Fig. 1. The burner is made of stainless steel. The methane/air pre-mixture (indicated by green arrows) is injected inwards at high speeds from 12 circular slits (Fig. 1(a)), which are distributed in two stages (Section B-B and C–C). On each stage the injection slits are provided in symmetrical positions on the circumference (Fig. 1(b) and (c)). The central line of the slit at the first stage (B-B) is 8 mm away from the closed end (Fig. 1(a)); while the distance between the central lines of slits at the first and second stages is 15 mm. The premixed gas is injected from a tube of 2-mm inner diameter towards a hole of 6mm inner diameter tangentially drilled at the inner surface of the burner wall (Fig. 1(d)). Between the 2- and 6-mm holes, an internal recirculation path of 12-mm inner diameter is situated normal to the direction of the injection. The injected velocity through the 2-mm tube is so high (∼50 m/s) that the static pressure is reduced and a hot burned gas (indicated by red arrows in Fig. 1(a)) is induced through the recirculation path into the tangentially injected combustible mixture. By diluting the combustible mixture with the recirculated burned gas, the NOx concentration at the exit may be reduced because of reduction of O atom concentration and temperature. In this burner, the diameter of the inner wall is 35 mm and its length is 150 mm. At the end of the inner wall, a copper made inner pipe is attached. The total length of the inner pipe is 60 mm, and the length outside the inner wall is 40 mm. A combustion quartz tube with inner diameter of 76 mm and length of 300 mm is attached at the outlet of the burner to photograph the flame appearances in the side view. In temperature measurement, a stainless-steel pipe is used instead of the quartz tube. Figure 2 shows the burner configuration (inner pipe is not shown here) in the side view (Fig. 2(a)) and front view (Fig. 2(b)). The dotted curve in the side view illustrates the closed end. The

Fig. 2. Burner configuration and measurements in the recirculation path (a: side view, b: front view).

positions of tangential injector centers at the first stage (first dashdot line from left) and second stage (second dash-dot line) are also illustrated. To measure the main species concentrations and temperatures in the recirculation path, a hole is drilled 15 mm downstream the second stage injector (red point in Fig. 2(b)). A sampling probe or thermocouple is inserted from the outer wall. In this study, the number of recirculation path is varied stepwise such as n = 0 (without recirculation), 6 and 12, and the variations of flame appearance, temperature and NOx emission are analyzed. Flame appearances were recorded by two conventional digital cameras from the side view and front view. The burned gas temperature along the burner axis was measured with a Pt/Pt-

202

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

Fig. 3. Schematic of experimental setup for PIV measurement.

13%Rh thermocouple (0.2 mm in wire diameter). The recirculated gas temperature in the open path was also measured (Fig. 2(b)). The thermocouple is coated with SiO2 , for which catalytic heating can be eliminated [44]. Thermometric error due to radiation is corrected through the method proposed by Kaskan [45]. In the measurements, the temperature values are recorded by a chart recorder (Yokogawa LR8100E, ±0.05% of reading +1 °C). The total uncertainty in the temperature measurement is approximately ±2.2%. On the other hand, the combustion gas was sampled with a quartz probe (5 mm in inner diameter) and the concentrations of NOx and other species (CO and CO2 ) were determined 80 mm upstream the burner exit. NOx was determined with a chemiluminescence formula NOx meter (Shimadzu, NOA-70 0 0) and the concentrations of CO and CO2 were determined with an infrared gas analyzer CO/CO2 meter (Shimadzu, CGT-70 0 0). For the NOx analyzer, it has a precision of ±0.5% of the full scale with a full-scale range of 10 0 0 ppm. Before each measurement, NOx analyzer is calibrated with a standard mixture which contains pure nitrogen and 0.1% volume percentage of NO (the stated accuracy of ±1.0%). For the CO/CO2 analyzer, it has a precision of ±1.0% of the full scale (10 0 0 ppm). To minimize error of each measurement, the analyzer is also calibrated with a standard mixture (pure nitrogen and 0.1% volume percentage of pure carbon monoxide, stated accuracy of ±1.0%). The experimental uncertainty of NOx measurement is about ±3.5%, and that of CO measurement is ±3.9%. To quantitatively analyze the recirculation volume, a PIV system (TSI) was adopted to measure the axial velocity around the inlet of the recirculation path, the setup of which is shown in Fig. 3. The PIV system consists of a high-resolution CCD camera (TSI, 1280 × 1024 pixels, 10-bit grayscale), a double-pulsed Nd:YAG laser (120 mJ/pulse, 15 Hz), a host computer, and a synchronizer.

Magnesium oxide (MgO) particles with a few microns in diameter are seeded into the mixture. Measurements are made under both cold flow and combustion conditions with different numbers of open paths. The laser sheet is directed to the burner, paralleling the burner axis, in addition, passing through both the center axis of a recirculation path and the burner axis; the CCD camera is set to perpendicular to the laser sheet. The fuel used is methane, which is supplied by a high pressure cylinder; and air is supplied by a compressor. Their flow rates are measured individually by the volumetric flow rate controllers (accuracy of ±2% of the reading), and then uniformly mixed and supplied to the burner. The volumetric flow rate meters are calibrated with a wet gas meter (Shinagawa W-NK-10, accuracy of ±0.15% of the reading), yielding an uncertainty of ±2.1% in each flow rate. In most tests, the air flow rate QAir is maintained at a constant value of 9.53 m3 /h. This yields a heat output of 10 kW at the equivalent ratio (Φ ) of 1.0. According to the operation condition of boilers, combustion tests are mainly conducted from lean to the stoichiometric condition in this study. To analyze observed variation of NOx formation with IFGR rate, the flame structure was simulated with the corrected temperature profile using the 1-D premixed laminar burner-stabilized flame model [46,47] in the Chemkin-PRO software. With this model, Knyazkov et al. [48] analyzed the formation and destruction of nitric oxide in methane flames at atmospheric pressure. In recent numerical simulations, various natural gas thermochemical models and their associated sub-models for NOx formation have been used to analyze NOx emission in various burner systems [21,30,32,48,49]. Among currently available thermochemical mechanisms, GDF-Kin®3.0 mechanism [50] gives the best predictions for NO formation in the lean and stoichiometric CH4 /air flames at atmospheric pressure [23,30]. In this study, detailed GDF-Kin®3.0 mechanism is used. Primary NOx chemical pathways included in the sub-model are thermal, prompt based on NCN, N2 O, and NNH. Multicomponent diffusion and thermal diffusion options are taken into account. Adaptive mesh parameters are set to grad = 0.06 and curv = 0.08, and the numerical results are ensured to be gridindependent. 3. Results At first, flame appearances were analyzed by varying n. Figure 4 illustrates the flame appearances in the side view (left) and front view (right) for the mixture of Φ = 0.8 (a) and 1.0 (b). In the case of Φ = 0.8, when all the recirculation paths are closed (n = 0, upper row in Fig. 4(a)), the flame in the side view cannot be observed (only the regions downstream of the inner pipe can be seen through the quartz tube), however, its length can be roughly estimated (as indicated by the cyan cylinder). Since the flame is a typical tubular flame when n = 0, and the flame length can be roughly estimated through the relation QM = π Df Lf Su [51], in which QM is the flow rate of fresh reactants, Su is the laminar burning velocity, Df and Lf are flame diameter and length, respectively. Df can be given through the flame image in the front view, in which a luminous circular zone is observed. Su can be obtained from literatures [52,53]. When n = 6 (second row), owning to increased gas volume induced by the recirculated gas and a possible decrease of Su , at the exit of inner pipe, a short, blue flame is observed; in the front view, the flame luminosity weakens and its diameter decreases. When n is increased to 12, in the side view, an expanded blue flame zone is observed, which illustrates a recirculation trend to the open path from upper edge of the flame. In the front view a thin, luminous circular zone is observed. In the case of Φ = 1.0 (Fig. 4(b)), the inner pipe is heated to radiate. In the side view, a yellow flame zone is observed at

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

203

Fig. 5. Steady tubular flame region under various air flow rates.

in a reduced rich limit [55]. By increasing n, the steady flame range in equivalence ratio gradually shrinks, however, it does not change with air flow rate. Then, the emissions were analyzed. Figure 6 shows the NOx and CO concentrations measured around the exit of the burner, in which n is selected as a parameter. The concentrations are determined for a wide range of equivalence ratios between Φ = 0.57 and 1.0 under a constant air flow rate of QAir = 9.53 m3 /h. The measured NOx values are converted to the values at 0% O2 concentration on the basis of O2 concentration determined in the exhaust gas through the following relation,

[NOx ](0% O2 ) =

Fig. 4. Flame appearances with recirculation path closed (n = 0, upper row), 6 paths opened (n = 6, middle row) and 12 paths opened (n = 12, bottom row ) for mixtures at Φ = 0.8 (a), and Φ = 1.0 (b), in which QAir = 9.53 m3 /h.

the exit of the inner pipe; by increasing n, the yellow flame zone gradually expands. In the front view, inside the circular blue zone, a yellow flame zone is formed in the center for the combustion of n = 0 and 6; when n is increased to 12, the yellow flame zone diminishes, indicating less soot formation. It is seen that the IFGR rate determined by n can greatly influence the flame structures. Next, the steady tubular flame range was determined for various air flow rates. Figure 5 shows the variation of equivalence ratio at extinction, in which QAir is increased from 4.77 to 11.44 m3 /h for n = 0, 6, and 12. The dotted lines correspond to the lean and rich flammability limits in Ref. [54]. In the case of n = 0, a steady flame can be established over a wide range between Φ = 0.5 and 1.5. The rich limit is a bit smaller than the standard value of Φ = 1.68 due to the so called Lewis number effects [55]. Since for methane mixture close to rich limit the effective Lewis number is usually larger than unity, and hence the flame diameter cannot become small. Thus, heat loss to the burner surface is increased resulting

0.2095 [N Ox ]m 0.2095 − [O2 ]m

(1)

in which [O2 ]m and [NOx ]m refer to measured O2 and NOx volume concentrations in the combustor exit, respectively. For all the cases of n = 0, 6 and 12, the NOx concentration increases as Φ is raised to 1.0, however, it is very important to note that the NOx emission significantly decreases as n is increased. For example, at the stoichiometry (Φ = 1.0), the measured value is 76 ppm (0% O2 ) when n = 0. This value decreases to 65 ppm for n = 6, and further drops to 46 ppm for n = 12. At the condition of Φ = 0.76, measured NOx is 35 ppm (0%O2 ) for n = 0; it decreases to 28 and 18 ppm by increasing n to 6 and 12, respectively. Comparing the two cases of n = 0 (without recirculation,♦) and n = 12 (all the recirculation paths are opened, ), the NOx emission reduces to almost a half. On the other hand, within the range of Φ = 0.65–0.9, the CO concentration around the burner exit is closed to 0 ppm for all the cases of n = 0, 6 and 12; hence, the combustion seems almost complete. In the cases of Φ = 1.0, the CO concentration increases abruptly, however, the value gradually decreases from 70 to 40 ppm when n is increased from 0 to12. It is seen that between Φ = 0.6 and 0.9, both NOx and CO emissions are significantly low with n = 12, which will be profitable in smaller boilers. Finally, the temperatures (T) were measured by SiO2 coated thermocouples. Figure 7 shows the temperature distributions along the burner axis for three cases of n = 0, 6 and 12 at Φ = 0.8 (corrected T due to radiation is shown only for n = 0, broken line). The positions of the tangential slits centers and the exit of the inner pipe are indicated by the dash-dot lines. It is seen that T (uncorrected) gradually increases from the closed end; thereafter, it rapidly decreases close to the exit of the inner pipe; and then gradually decreases approaching the exit of the burner. By increasing n from 0 to 12, around the closed end, T decreases from 1430

204

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

Fig. 6. Variations of NOx (a) and CO (b) concentrations with n (QAir = 9.53 m3 /h).

Fig. 7. Variations QAir = 9.53 m3 /h).

of

axial

temperature

distributions

with

n



=

0.8,

to 1100 °C; while the maximum T also decreases from 1550 to 1260 °C. At the exit of the inner pipe, for n = 0, T is 1300 °C; by increasing n to 6 and 12, T decreases to1200 and 1150 °C, respectively. Downstream of the inner pipe, the burned gas temperature also decreases by increasing n. For n = 12, T even drops to 815 °C. In addition, temperature of the recirculated gas in an open path (see Fig. 2(b)) was also measured. Figure 8 illustrates the temperature distribution at Φ = 0.8. In the cases of n = 6, T increases from 200 to 400 °C from the outer brim to the inner brim (close to inner flame zone). The temperature of the recirculated burned gas is much lower than its initial temperature at the exit of the inner pipe owing to heat transfer to the wall of recirculation path. By increasing n to 12, T decreases at least 50 °C comparing with that of n = 6. Owning to special burner configuration, a remarkable drop in temperature (around 10 0 0 K) appears in the recirculated burned gas before mixing the fresh reactant. It is different from the high temperature gas recirculation observed in the IFGR combustion systems conducted by Sorrentino et al. [28] and Sung et al. [56]. It is seen that by increasing n, the axial temperature in the burner as well as T in the recirculated path decreases. This is one possible reason for the decrease of NOx emission.

Fig. 8. Temperature distributions of the recirculated gas in the path (n = 6 and 12, Φ = 0.8, QAir = 9.53 m3 /h).

4. Discussion To quantitatively investigate IFGR rate and furthermore evaluate its influences on the combustion characteristics, such as reactant concentrations, local equivalence ratios and NOx formation, a systematic study is conducted through velocity as well as concentration measurements, theoretical analysis and numerical simulation.

4.1. IFGR rate To quantify IFGR rate, the axial velocity is determined by adopting a PIV system (Fig. 3) for both the cold flow and combustion conditions with different numbers of n. In tests the interval time between double pulses from the PIV laser was set to 50 μs. Figure 9 illustrates the axial velocity distribution around the inlet of the recirculation path, which is an averaged value of 50 continuous velocity measurements. Though optical distortion caused by cylindrical quartz tube occurs, it only influences the radial component of velocity vector [57], whose errors are less than 0.1% in the concerned domain of calculating recirculated gas flow rate (QR ). In this study, QR is mainly determined by the axial component, and the errors caused by optical distortion can be negligible.

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

205

Fig. 10. Mean velocity profiles in a plane containing tube axis under the combustion condition of Φ = 0.85. The result is obtained by averaging 50 measurements (n = 12, QAir = 9.53 m3 /h).

To quantitatively analyze the IFGR rates for different operating conditions, at first, the flow rate of recirculated gas is calculated based on measured velocity around the inlet of the open path. As indicated by Fig. 11, the axial velocity determined at the plane of X = 2 mm (bold blue line in the left-end), i.e., 2 mm offset the inlet of open path, is selected to represent the initial velocity of the recirculated gas at the inlet of one path. The velocities between the upper brim (Point A, X = 2 mm, Y = 34 mm) and the lower brim (Point B, X = 2 mm, Y = 22 mm) are measured to calculate the flow rate entered into the open path. Based on the grid size in the PIV measurement, the axial velocity distribution between upper and lower brims, i.e., a distance of path diameter (indicated as d = 12 mm), is averagely divided into k parts, denoted as Vx1 to Vxk , as illustrated in the right end of Fig. 11. In each part, a velocity vector is given through PIV measurement. In addition, the distance for each part is d/(k-1). Then the flow rate can be estimated by assuming that recirculated gas passing through a circular ring, whose outer and inner radii can be given by d/(k-1)∗ [( k-1)/2-i + 1] and d/(k-1)∗ [( k-1)/2-i], respectively. And the velocity is calculated as the averaged values of four velocities situated on the circular ring, i.e., (Vxi + Vxi+1 + Vxk-i + Vxk-i+1 )/4, as indicated by the shadow circular ring. Then the flow rate recirculated into n paths (QR ) can be calculated as k−1

QR = n

2  Vxi +Vxi+1 +Vxk−i + Vxk−i+1

4

i=1

Fig. 9. Mean velocity profiles in a plane containing tube axis. The result is obtained by averaging 50 measurements (QAir = 9.53 m3 /h).

Here a coordinate is defined, whose origin is located at the burner axis, parallel the inlet of recirculation path. As shown in Fig. 9(a), the recirculation paths are closed (n = 0), however, an obvious recirculation zone is observed outside the inner pipe; by increasing n to 6 (Fig. 9(b)), part of the recirculated gas outside the inner pipe enters into the recirculation path; in the case of n = 12 (Fig. 9(c)), the volume of recirculated gas increases obviously based on measured velocity vectors at the inlet of the open path. The measurements are also conducted in the combustion condition. Here an example at Φ = 0.85 is analyzed, as illustrated in Fig. 10. All the recirculation paths are opened, i.e., n = 12. It is seen that the velocity magnitude of the recirculated gas entering into the open path increases much comparing with the cold flow condition in Fig. 9(c).



−π

d k−1



k−1 −i 2

2 

   2 d k−1 π −i + 1 k−1

2

(2)

Based on Eq. (2), the values of QR for cold flow and combustion conditions are calculated for n = 6 and 12. The results are plotted in Fig. 12. For the cold flow at the condition of n = 12, QR is calculated as 1.3 m3 /h; for the combustion of Φ = 0.85, the value is about 8.2 m3 /h, which is almost close to the fresh air flow rate (9.53 m3 /h). By converting the hot burned gas flow rate to STP condition, QR becomes 1.4 m3 /h, which is close to that of the cold flow. In the case of n = 6, QR becomes almost a half of n = 12 (STP). Here a parameter is defined to quantitatively analyze the IFGR rate (β ), as

β = QR /QB × 100%

(3)

in which QB is the flow rate of burned gas, and QB and QR are converted to STP condition. In the cases of n = 6 and 12, the values

206

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

Fig. 11. Calculation method of the recirculated gas volume with using the results of velocity vector profiles obtained by a PIV system.

Fig. 12. Variation of internal recirculated gas flow rate with n.

of β are 6% and 13% respectively in the combustion. The result of n = 12 is close to that of 10% (STP) in literature [14]. 4.2. Concentration variations of main components with IFGR rate As β varies, the concentrations of main components in reaction zone (inner tubular flame zone) will change accordingly. Here the flow rates and volume concentrations of main reactants are theoretically calculated and compared with the measured values. The global reaction CH4 + 2O2 →CO2 + 2H2 O is adopted for simplification. Hence, five main components CH4 , O2 , CO2 , H2 O and N2 are analyzed. In the tests the mixtures from lean to stoichiometry are used, in which the fuel is assumed to be completely consumed. The flow rates and volume concentrations of the five components in the unburned gas (including fresh reactants and recirculated gas), burned gas and recirculated gas are calculated and expressed by Φ and β . The results are listed in Table 1. Detailed calculations can be found in Appendix. Then the variation of volume concentration in the unburned gas (U ) can be calculated. Figure 13 illustrates the results for different mixtures of Φ = 0.6, 0.8 and 1.0, in which β is varied from 0 to 100%. For each Φ , the value of U-N2 is maintained constant; that of CH4 gradually reduces to 0; U-O2 reduces to 0 for Φ = 1.0, while it does not become zero for Φ = 0.8 and 0.6, indicating that U-O2 decreases much rapidly

in Φ = 1.0; the values of U-H2 O and U-CO2 gradually increase with β , and they are largest in the case of Φ = 1.0. The calculated values for β = 13%, which correspond to the IFGR rate of n = 12 in tests, are indicated by the red dash lines. By inducing the recirculated gas, as illustrated in Fig. 13, in the reaction zone the oxygen mole fraction and equivalence ratio will change. Here the local oxygen mole fraction is defined as β  = U-O2 /(1-U-CH4 ), while the local equivalence ratio in unburned gas is defined as Φ U = (U-CH4 /U-O2 )/ (CH4 /O2 )stoic , in which stoic means stoichiometry. The values of β  and Φ U corresponding to Fig. 13 are presented in Fig. 14. It is seen that β  decreases with an increase of β ; in addition, the decrease of β  becomes more rapidly by increasing Φ in the premixed methane/air. As for Φ U , its value is maintained at 1.0 for Φ = 1.0, i.e., Φ U = Φ = 1.0. However, as indicated in Fig. 14(a), the value of β  decreases. In the cases of Φ = 0.6 and 0.8, Φ U gradually decreases with an increase of β . Particularly, when β exceeds 50%, Φ U rapidly decreases. Since a majority of burned gas is recirculated back to increase overall oxygen flow rate in the unburned mixture, in which methane flow rate is maintained constant, leading to a rapid decrease of Φ U. It is seen that in the test of n = 12, owing to decreases of both β  and Φ U (indicated by the dash lines), NOx formation (Fig. 6) as well as burned gas temperature (Fig. 7) decreases. It should be noted that in the practical operation, the initial value of Φ in the fresh reactant should be a bit higher than the lean extinction limit to avoid flame extinction owing to decreases of Φ U and β  when β is non-zero. In the measurements, the minimum value of Φ is 0.64 for n = 12, as illustrated in Fig. 5. To give a more detailed instruction for the operation of the methane/air combustion with different fuel compositions, the values of β  and Φ U in the unburned gas are examined by varying initial Φ in the fresh reactant, which is summarized in the supplementary material (Fig. S1). Based on Table 1, the volume concentrations of the five components in the recirculation path (R ) can also be calculated for various values of Φ . Here the results of R-O2 and R-CO2 are calculated, both of which are only determined by Φ , i.e., independent with β or n (last two rows in Table 1). The results are plotted in Fig. 15 (solid line). By increasing Φ , R-O2 gradually decreases while R-CO2 increases. In the tests, R-O2 and R-CO2 at the center axis of the recirculation path are measured. The results for n = 6 () and 12 () are plotted in Fig. 15, whose values are consistent, indicating that R-O2 and R-CO2 are independent with n. The measurements for Φ = 0.7−0.9 show a good agreement with calculation. In the cases of Φ = 0.6 and 1.0, R-O2 becomes larger than calculated value while R-CO2 becomes smaller. Since CO is formed in these two cases (Fig. 6). On the whole, the calculated values in Table 1 can be used to well examine the species concentrations

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

207

Table 1 Flow rate and volume concentration of main components in unburned, burned and recirculated gases. Type

Flow rate

Unburned gases

QU-CH4 QU-O2 QU-N2 QU-CO2 QU-H2 O

1 (2/Φ −2)/(1-β ) + 2 (2∗ 79)/[21Φ (1-β )] β /(1-β ) 2β /(1-β )

Volume concentration

U-CH4 U-O2 U-N2 U-CO2 U-H2 O

21Φ (1-β )/(200 + 21Φ ) 42(1-Φβ )/(200 + 21Φ ) 2∗ 79/(200 + 21Φ ) 21Φβ /(200 + 21Φ ) 42Φβ /(200 + 21Φ )

Burned gases

QB-CH4 QB-O2 QB-N2 QB-CO2 QB-H2 O

0 (2/Φ −2)/(1-β ) (2∗ 79)/[21Φ (1-β )] 1/(1-β ) 2/(1-β )

B-CH4 B-O2 B-N2 B-CO2 B-H2 O

0 42(1-Φ )/(200 + 21Φ ) 2∗ 79/(200 + 21Φ ) 21Φ /(200 + 21Φ ) 42Φ /(200 + 21Φ )

Recirculated gases

QR-CH4 QR-O2 QR-N2 QR-CO2 QR-H2 O

0

R-CH4 R-O2 R-N2 R-CO2 R-H2 O

0 42(1-Φ )/(200 + 21Φ ) 2∗ 79/(200 + 21Φ ) 21Φ /(200 + 21Φ ) 42Φ /(200 + 21Φ )

β (2/Φ −2)/(1-β ) (2∗ 79β )/[21Φ (1-β )] β /(1-β ) 2β /(1-β )

and local equivalence ratios, particularly for the mixtures between Φ = 0.7−0.9, which are usually adopted in the boilers.



In methane/air combustion the major routes to NO are the thermal (Zeldovich) route and the prompt (Fenimore) route [8]. In the former route, the NO production rate can be estimated by

d[NO] = 2k1 [O][N2 ] dt

(4)

in which k1 is the reaction rate of O + N2 →NO + N, and here k1 = 1.8 × 1014 exp(−318kJ•mol−1 /(RT)) cm3 /(mol•s) [6,9]; [NO], [O] and [N2 ] are the molar concentrations. Here O2 and O are assumed to be in equilibrium. Hence [O] can be estimated based on [O2 ] through the following relation [9],

[O ] =

Ru T

[O 2 ]

1 2

(5)

in which Kp is equilibrium constant, Ru is universal constant (8314.51 J/(kmolK)), P0 the pressure (101,325 Pa) and T the temperature. According to Table 1, the volume concentrations of oxygen and nitrogen in the unburned gas can be calculated as U-O2 = 42(1Φβ )/(200 + 21Φ ) and U-N2 = 158/(200 + 21Φ ). By convert the volume concentration to molar concentration with unit of kmol/m3 , the following relation is obtained,



[O2 ] = U−O2 RPu T = [N 2 ] = 

P U−N2 Ru T

=

42−42Φβ P 200+21Φ Ru T 158 P 200+21Φ Ru T

(kmol/m3 ) (kmol/m3 )

(6)

in which P is the pressure (Pa). And hence the thermal route NO can be estimated through the following relation [9], d[NO] = 2k1 [O][N2 ] dt



= 2k1

= 2k1 e

e

L T0 4(QAir + QC H4 )/π D2 T

(8)

in which D is diameter of the inner burner (35 mm), T0 = 273.5 K. Thus, Eq. (7) can be calculated as

4.3. NOx formation

K P 12 p 0

t =

66713T −2.54×108 Ru T

Ru T 66713T −2.54×108 2Ru T

P0

12 

42(1 − Φβ ) P 200 + 21Φ Ru T

 12 

158 P 200 + 21Φ Ru T

P 2  42(1 − Φβ )  12 158 Ru T 200 + 21Φ 200 + 21Φ



(7)

in which P0 = P = 101,325 Pa. In the reaction zone from the closed end to the exit of the inner pipe, i.e., Z = 0-190 mm, the residence time t in an axial distance of L = 10 mm can be calculated through

[NO] =

t

d[NO] dt dt

0



P 2  42 − 42βΦ  12 = 2k1 e Ru Ti 200 + 21Φ 1  

158 L T0 × 200 + 21Φ 4(Qair + QC H4 )/π D2 Ti i 

66713T −2.54×108 2Ru Ti

(9)

in which Ti is the measured temperature along the burner axis. Then the molar concentration of NO is converted to volume percentage (ppm) through the following relation,

R T

[NO] ppm = [NO] u  P

P  42 − 42βΦ  12 = 2k1 e Ru Ti 200 + 21Φ 1  

158 L T0 × 200 + 21Φ 4(QAir + QC H4 )/π d2 Ti i 

66713T −2.54×108 2Ru Ti

(10)

With Eq. (10), the thermal type NO in this burner can be estimated. For example, in the case of ΦΦ = 0.8 (QAir = 9.53 and QCH4 = 0.8 m3 /h), with the measured axial temperatures (corrected values) at β = 0 (n = 0), the thermal NO can be calculated as [NO]ppm = 86.3 ppm; in the cases of β = 6% (n = 6),  [NO]ppm becomes 13.7 ppm, and for that of β = 13% (n = 12), [NO]ppm reduces to 1.3 ppm. By converting the values to those at 0% O2 , [NO]ppm becomes 102.5 ppm for n = 0; for n = 6, [NO]ppm = 16.8 ppm; and that of n = 12 is 1.6 ppm. It is seen that by increasing n, which directly reduces the flame temperature and oxygen atom concentration, thermal NO decreases significantly. While in the measurements, for n = 0, [NO]ppm = 40 ppm, which is much lower than the calculated value (102.5 ppm). Since in Eq. (10), the central high temperature is selected to calculate the thermal NO, which may result in large deviations. In addition, in Eqs. (8) and (10) the residence time is also roughly estimated, which may greatly influence NO formation. In the case of n = 6, the measured value (30 ppm) becomes almost double of the estimated value; in the case of n = 12, the measured value is 20 ppm, much larger than the estimated value of 1.6 ppm. For n = 6 and 12, the relative large NO concentration in measurements may be attributed to prompt type NO formation, as reported in Refs. [21,24],

208

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

Fig. 14. Variations of local oxygen mole fraction β  (a) and equivalence ratio Φ U (b) corresponding to Fig.13.

Fig. 13. Variations of CH4 , O2 , N2 , CO2 and H2 O concentrations with IFGR rate β in the unburned gas (a: Φ = 1.0, b: Φ = 0.8, c: Φ = 0.6).

which can also be illustrated from the expended flame front observed in Fig. 4. Numerical simulations with using Chemkin-PRO are conducted to examine the NO formation route for the three cases of n = 0, 6 and 12 at Φ = 0.8. With detailed reaction mechanism GDFKin®3.0 [50], the relative contributions of four different chemical processes contributed to overall formation of NO are determined by removing the corresponding initiation reactions from the simulations [58]. Detailed initiation reactions for NO formation and method to calculate the contribution of an individual route can be found in literature [23]. In the simulations, initial mixture compositions for different values of n (or β ) are calculated based on Table 1 (unburned gas compositions); in addition, corresponding axial temperature profiles measured in experiments are adopted in the burner-stabilized flame model. It should be noted that the flame temperature decreases from n = 0 to 12 (Fig. 7), which is different from the studies that adopted constant flame temperature [21,24, 25]. Figure 16 shows the computed contributions of different chemical routes to the total NO concentration for different values of n. For all data points, NO formation is taken at a consistent residence time of τ = 10 ms downstream from the flame location, where the flame location is defined by the peak in CH mole

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

209

fraction [23]. To compare with experimental measurements as well as theoretical values, NO concentration is scaled to 0% oxygen concentration. The unscaled data is available as the supplementary material (Fig. S2). Figure 16 illustrates predicted NO concentrations from different formation routes (left) and relative weight of each route (right) under different values of n. The results indicate that with increasing n (or β ), the NO formation from each individual route decreases, leading to a significant decrease of total NO formation. Particularly, thermal NO formation decreases remarkably, which even drops to approach zero for n = 12. The variation trend of thermal NO is consistent with the theoretical analysis based on Eq. (10). While that of prompt NO differs from the results reported in Refs. [21,24, 25]. Because in the present study by increasing n = 0 to 12, the flame temperature decreases (by 10 0 ∼ 30 0 K); In addition, the same amount of fuel is burned in a large amount of inert gases, leading to reduced concentrations of O and OH radicals (see Fig. S3 of the supplementary material). On the other hand, the relative weight of prompt NO increases by increasing n, indicating an important contribution to overall NO formation. This conclusion agrees with that reported in Refs. [21,24]. The simulation results show some differences in total NO formation comparing with measurements, particularly those of n = 6 and 12. Since when n is increased to 6 and 12, the flame is expanded along the radial direction and stabilized in a recirculation zone at the exit of inner pipe (Figs. 4, 9, and 10), which may raise the in-flame residence time and hence increase the production of NO through the fast prompt/NNH routes [21,32]. In summary, the decrease of NOx formation in this study is mainly attributed to IFGR, which leads to a decrease of flame temperature as well as oxygen atom concentration, resulting in thermal NO reduction. Meanwhile, the relative weight of prompt NO contributed to total NO formation increases with IFGR rate. It should be noted that to reduce the NOx emission, β should be as high as possible, however, its maximum value is limited to (1) the designing of the combustor, and (2) the operation condition (for example, Φ ), in which extinction may occur owing to reduced local equivalence ratio and oxygen mole fraction, as indicated in Figs. 5 and 14. 5. Conclusion Fig. 15. Measured and calculated O2 (a) and CO2 (b) volume concentrations in the recirculation path for mixtures of different equivalence ratios.

Fig. 16. Predicted NO concentrations (scaled to 0% O2 ) from different formation routes (left) and relative weight of each route (right) for different IFGR rates. NO concentration is taken at a location τ = 10 ms downstream from the flame reaction zone.

In this study a prototype tubular flame burner with internal flue gas recirculation (IFGR) is designed to quantitatively examine the effects of IFGR rate (β ) on the combustion characteristics. With the unique configuration of current IFGR system, several new points have been attained in the present study, including: (1) combined dilution effects of internal recirculated N2 , CO2 , O2 and H2 O are considered; (2) burned gas IFGR rate can be conveniently adjusted by varying the number (n) of opening recirculation path inside the burner, which can be even applied to burn low-heat-value gas fuels by adopting a relatively small IFGR rate; and (3) in the unburned mixture containing fresh reactants and recirculated burned gas, variations of flow rate, local equivalence ratio and oxygen mole fraction with IFGR rate have been derived, which will be beneficial for optimal design and operation of tubular combustion with IFGR. By measuring velocity distribution with a PIV system, the IFGR rate β is quantitatively examined. It increases as n is raised. When n = 12, β reaches 13% (convert to STP), in which NOx emission significantly reduces to almost a half, and both NOx and CO emissions are much low for the mixtures of equivalence ratio between 0.65 and 0.9. By increasing β , the steady combustion region in equivalence ratio shrinks; flame luminosity decreases while its length increases; the burned gas temperature decreases, and particularly, a remarkable drop in temperature appears in the recirculated gas; based on calculated composition of the unburned gas, the concentration of oxygen atom also decreases, leading to a decrease of

210

B. Shi et al. / Combustion and Flame 188 (2018) 199–211

( 0 ≤ β ≤ 1.0 )

Fig. A1. Flowchart of methane combustion in the internal recirculation burner.

thermal type NO. The experimental data confirms that the present, simple device of internal recirculation type tubular flame burner can successfully reduce NOx emission without great cost and space. Acknowledgments This research was supported by GASR (Grants-in-Aid for Scientific Research) through JSPS. Part of the work is supported by National Natural Science Foundation of China (No. 51676016). Appendix The method to calculate the flow rates and volume concentrations of the main species in the unburned gas, burned gas as well as recirculated gas are analyzed. Here the global reaction CH4 + 2O2 →CO2 + 2H2 O is adopted. At the beginning, oxygen volume percentage in air is assumed as 0.21 and that of nitrogen is 0.79. The initial methane flow rate QCH4 is set as 1 m3 /h, and the initial equivalence ratio is Φ (in this study Φ ≤ 1.0). And hence the air flow rate can be expressed as 2/(0.21Φ ), while QO2 and QN2 can be calculated as 2/Φ and (2∗ 0.79)/(0.21Φ ), respectively. At first, the process of the internal gas recirculation in the combustion is analyzed by introducing the flowchart, as illustrated in Fig. A1. The internal flue gas recirculation (IFGR) rate is defined as

β=

QR × 100% QB

(A1)

in which QR and QB are flow rates of recirculated gas and burned gas converted to STP condition. In the 1st period combustion, methane/air pre-mixture (QM ) is supplied to the burner; after complete reaction, part of the burned gas is recirculated through the open path (QR-1 ), which will be added to the 2nd combustion; the rest of the burned gas is then exhausted through exit (QE-1 ). In this process the oxygen flow rates in different stages are analyzed as an example. Before entering the reaction zone, the unburned gas flow rate QU-1 equals to QM , in which the oxygen flow rate QU-1 – O2 is (2/Φ −2) + 2; in the burned gas (QB-1 ), the oxygen flow rate QB-1 – O2 is calculated as 2/Φ −2; in the recirculated gas, QR-1 – O2 is given as β (2/Φ −2). In the 2nd combustion, methane/air mixture with the same flow rate (QM ) is provided, and before reaction, this pre-mixture is mixed with recirculated gas (QR-1 ), i.e., QU-2 = QM + QR-1 . The oxygen flow rate is given as QU-2 – O2 = 2/Φ + β (2/Φ −2) = (1 + β )(2/Φ −2) + 2. After complete burning, in burned gas QB-2 – O2 = (1 + β )(2/Φ −2); part of the burned gas is recirculated, for which QR-2 – O2 = β (1 + β )(2/Φ −2 )= (β + β 2 )(2/Φ −2); the rest burned gas is exhausted from the burner exit. The same process is repeated in the 3rd, 4th, …nth combustion. And in the 3rd combustion, the oxygen flow rates can be calculated as QU-3 – O2 = 2/Φ + (β + β 2 )(2/Φ −2) = (1 + β + β 2 )(2/Φ −2) + 2, QB-3 – O2 = (1 + β + β 2 )(2/Φ −2) and QR-3 – O2 = β (1 + β + β 2 )(2/Φ −2) = (β + β 2 + β 3 )(2/Φ −2). In the nth combustion, it can be obtained that QU-n-O2 = (1 + β + β 2 +…+ β n-1 )(2/Φ −2) + 2, QB-n-O2 = (1 + β + β 2 +…+ β n-1 )(2/Φ −2) and QR-n-O2 = (β + β 2 +…+ β n )(2/Φ −2). It should be noted that the residence time for the gas in the inner tubular flame zone can be estimated by Eq. (8), which is around 0.06 s when all recirculation paths are opened. While the residence time for the recirculated gas in the recirculation path is around 0.075 s. Then even within 1 s, the flue gas may be recirculated over ten times. Since 0 ≤ β < 1.0, after a short time (e.g., exceeds 10 s), in which the flue gas may be recirculated over hundreds of times, QU-n-O2 can be calculated as 2 + (2/Φ −2)/(1-β ); in addition, QB-n-O2 = (2/Φ −2)/(1-β ) and QR-n-O2 = β (2/Φ −2)/(1-β ). Hence as Φ and β are maintained constant, the flow rates such as QU-n-O2 , QB-n-O2 and QR-n-O2 can be assumed to not vary with time in the combustion tests. Then the flow rates of the other four components can be calculated with the same method. The results are illustrated in Table 1. The accuracy of the calculation is verified through discussion in Fig. 15. Supplementary materials Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.combustflame.2017.09. 043. References [1] S.C. Hill, L.D. Smoot, Modeling of nitrogen oxides formation and destruction in combustion systems, Prog. Energy Combust. Sci. 26 (20 0 0) 417–458. [2] K. Skalska, J.S. Miller, S. Ledakowicz, Trends in NOx abatement: a review, Sci. Total Environ. 408 (19) (2010) 3976–3989, doi:10.1016/j.scitotenv.2010.06.001. [3] A.F. Sarofim, R.C. Flagan, NOx control for stationary combustion sources, Prog. Energy Combust. Sci. 2 (1976) 1–25. [4] P. Basu, C. Kefa, L. Jestin, Boilers and burners, Springer-Verlag, New York, 20 0 0, pp. 258–266. [5] T. Kolb, P. Jansohn, W. Leuckel, Reduction of NO emission in turbulent combustion by fuel-staging/effects of mixing and stoichiometry in the reduction zone, Symp. (Int.) Combust. 22 (1988) 1193–1203. [6] J. Warnatz, U. Mass, R.W. Dibble, Combustion, Springer-Verlag, Berlin, 2001, pp. 237–256. 3 ed. [7] J.A. Wunning, J.G. Wunning, Flameless oxidation to reduce thermal NO-formation, Prog. Energy Combust. Sci. 23 (1997) 81–94.

B. Shi et al. / Combustion and Flame 188 (2018) 199–211 [8] C.E. Baukal Jr., Industrial combustion pollution and control, Marcel Dekker, New York, 2004, pp. 247–325. [9] S.R. Turns, An introduction to combustion, 2nd ed., McGraw Hill, Boston, 20 0 0, pp. 130–133, 168-171, 573-584. [10] H. Tsuji, M. Morita, A.K. Gupta, M. Katsuki, K. Kishimoto, T. Hasegawa, High temperature air combustion: From energy conversion to pollution reduction, CRC Press, Boca Raton, 2003. [11] G.G. Szego, B.B. Dally, G.J. Nathan, Scaling of NOx emissions from a laboratory-scale mild combustion furnace, Combust. Flame 154 (2008) 281–295. [12] H. Wei, T. Zhu, G. Shu, L. Tan, Y. Wang, Gasoline engine exhaust gas recirculation - a review, Appl. Energy 99 (2012) 534–544. [13] K. Al-Qurashi, A.D. Lueking, A.L. Boehman, The deconvolution of the thermal, dilution, and chemical effects of exhaust gas recirculation (EGR) on the reactivity of engine and flame soot, Combust. Flame 158 (2011) 1696–1704. [14] K. Shinomori, K. Katou, D. Shimokuri, S. Ishizuka, NOx emission characteristics and aerodynamic structure of a self-recirculation type burner for small boilers, Proc. Combust. Inst. 33 (2011) 2735–2742.. [15] A. Coghe, G. Solero, G. Scribano, Recirculation phenomena in a natural gas swirl combustor, Exp. Therm. Fluid Sci. 28 (2004) 709–714. [16] R. Salzmann, T. Nussbaumer, Fuel staging for NOx reduction in biomass combustion: experiments and modeling, Energy Fuels 15 (2001) 575–582. [17] M. Brower, E.L. Petersen, W. Metcalfe, H.J. Curran, M. Furi, G. Bourque, Ignition delay time and laminar flame speed calculations for natural gas/hydrogen blends at elevated pressures, ASME J. Eng. Gas Turb. Power 135 (2013) 021504-021501. [18] G. Liuzzo, N. Verdone, M. Bravi, The benefits of flue gas recirculation in waste incineration, Waste Manag. 27 (2007) 106–116. [19] H.K. Kim, Y. Kim, S.M. Lee, K.Y. Ahn, NO reduction in 0.03-0.2 MW oxy– fuel combustor using flue gas recirculation technology, Proc. Combust. Inst. 31 (2007) 3377–3384. [20] F.H.V. Coppens, J. Deruyck, A. Konnov, The effects of composition on burning velocity and nitric oxide formation in laminar premixed flames of CH4 +H2 +O2 + N2 , Combust. Flame 149 (2007) 409–417. [21] A.C.A. Lipardi, P. Versailles, G.M.G. Watson, G. Bourque, J.M. Bergthorson, Experimental and numerical study on NOx formation in CH4 -air mixtures diluted with exhaust gas components, Combust. Flame 179 (2017) 325–337. [22] A.A. Konnov, I.V. Dyakov, J. De Ruyck, The effects of composition on the burning velocity and NO formation in premixed flames of C2 H4 +O2 +N2 , Exp. Therm. Fluid Sci. 32 (2008) 1412–1420. [23] G.M.G. Watson, J.D. Munzar, J.M. Bergthorson, NO formation in model syngas and biogas blends, Fuel 124 (2014) 113–124. [24] K.B. Fackler, M.F. Karalus, I.V. Novosselov, J.C. Kramlich, P.C. Malte, Experimental and numerical study of NOx formation from the lean premixed combustion of CH4 mixed with CO2 and N2 , ASME J. Eng. Gas Turb. Power 133 (12) (2011) 121502–121507. [25] M.J. Landman, M.A.F. Derksen, J.B.W. Kok, Effect of combustion air dilution by water vapor or nitrogen on NOx emission in a premixed turbulent natural gas flame: an experimental study, Combust. Sci. Technol. 178 (2006) 623–634. [26] H. Li, A.M. Elkady, A.T. Evulet, Effect of exhaust gas recirculation on NOx formation in premixed combustion system, 47th AIAA Aerospace Sciences Meeting Including The New Horizons Forum and Aerospace Exposition, Orlando, Florida, AIAA (2009) 5-8 January2009-226. [27] A.M. ElKady, A. Evulet, A. Brand, T.P. Ursin, A. Lynghjem, Application of exhaust gas recirculation in a DLN F-class combustion system for postcombustion carbon capture, ASME J. Eng. Gas Turb. Power 131 (3) (2009) 034505. [28] G. Sorrentino, P. Sabia, P. Bozza, R. Ragucci, M. de Joannon, Impact of external operating parameters on the performance of a cyclonic burner with high level of internal recirculation under MILD combustion conditions, Energy (2017), doi:10.1016/j.energy.2017.05.135. [29] G.A. Chung, B. Akih-Kumgeh, G.M.G. Watson, J.M. Bergthorson, NOx formation and flame velocity profiles of iso- and n-isomers of butane and butanol, Proc. Combust. Inst. 34 (2013) 831–838. [30] G.M.G. Watson, J.D. Munzar, J.M. Bergthorson, Diagnostics and modeling of stagnation flames for the validation of thermochemical combustion models for NOx predictions, Energy Fuels 27 (2013) 7031–7043. [31] A.C.A. Lipardi, J.M. Bergthorson, G. Bourque, NOx emissions modeling and uncertainty from exhaust-gas-diluted flames, ASME J. Eng. Gas Turb. Power 138 (2015) 051506.

211

[32] G.M.G. Watson, P. Versailles, J.M. Bergthorson, NO formation in rich premixed flames of C1-C4 alkanes and alcohols, Proc. Combust. Inst. 36 (2017) 627–635. [33] S. Ishizuka, On the behavior of premixed flames in a rotating flow field: Establishment of tubular flames, Symp. (Int.) Combust. 20 (1985) 287–294. [34] R.J. Kee, A.M. Colclasure, H. Zhu, Y. Zhang, Modeling tangential injection into ideal tubular flames, Combust. Flame 152 (2008) 114–124. [35] P. Wang, J.A. Wehrmeye, R.W. Pitz, Stretch rate of tubular premixed flames, Combust. Flame 145 (2006) 401–414. [36] S. Ishizuka, D. Shimokuri, K. Ishii, K. Okada, K. Takashi, Y. Suzukawa, Development of practical combustors using tubular flames, J. Combust. Soc. Japan. 156 (2009) 104–113. [37] S. Ishizuka, R. Hagiwara, M. Suzuki, A. Nakamura, O. Hamaguchi, Combustion characteristics of a tubular flame burner, Jpn. Soc. Mech. Engrs. 65 (1999) 3845–3852. [38] R. Hagiwara, M. Okamoto, S. Ishizuka, H. Kobayashi, A. Nakamura, M. Suzuki, Combustion characteristics of a tubular flame burner for methane, Jpn. Soc. Mech. Engrs. 66 (20 0 0) 3226–3232. [39] S. Ishizuka, Characteristics of tubular flames, Prog. Energy Combust. Sci. 19 (1993) 187–226. [40] S. Ishizuka, D. Dunn-Rankin, R.W. Pitz, R.J. Kee, Y. Zhang, T. Takeno, D. Shimokuri, Tubular combustion, Momentum Press, New York, 2013. [41] H. Peng, B. Shi, D. Shimokuri, S. Ishizuka, An experimental study on a self-recirculation type low NOx tubular flame burner, 9th Asia-Pacific Conference on Combustion, Gyeongju, Korea (2013), pp. 19–22. May. [42] H. Peng, B. Shi, D. Shimokuri, S. Ishizuka, NOx emission characteristics of a self-recirculation type tubular flame burner, J. Combust. Soc. Jpn 56 (2014) 54–61 (in Japanese). [43] S. Ishizuka, D. Shimokuri, H. Peng, A study on the self-recirculation type low NOx burner with use of the ejector effect of tangentially injected gas streams, J. Jpn. Boiler Assoc. 382 (2013) 18–25. [44] W.E. Kaskan, The dependence of flame temperature on mass burning velocity, Symp. (Int.) Combust. 6 (1957) 134–143. [45] J.M. Madson, E.A. Theby, SiO2 coated thermocouples, Combust. Sci. Technol 36 (1984) 205–209. [46] R.J. Kee, J.F. Grcar, M.D., Smooke, J.A. Miller, A FORTRAN program for modeling steady laminar one-dimensional premixed flames. Sandia National Laboratories Report; 1985, SAND 85–8240. [47] R.J. Kee, G. Dixon-Lewis, J. Warnatz, M.E. Coltrin, J.A. Miller, A FORTRAN computer code package for the evaluation of gas-phase multicomponent transport properties, Sandia National Laboratories, 1986 Tech. Rep. SAND86-8246. [48] D.A. Knyazkov, A.G. Shmakov, I.V. Dyakov, O.P. Korobeinichev, J. De Ruyck, A.A. Konnov, Formation and destruction of nitric oxide in methane flames doped with NO at atmospheric pressure, Proc. Combust. Inst. 32 (2009) 327–334. [49] P. Gokulakrishnan, C.C. Fuller, R.G. Joklik, M.S. Klassen, Chemical kinetic modeling of ignition and emissions from natural gas and lng fueled gas turbines, ASME Turbo Expo 2012, GT2012-69902, Copenhagen, Denmark (2012), pp. 11–15. June. [50] A.EI Bakali, L. Pillier, P. Desgroux, B. Lefort, L. Gasnot, J.F. Pauwels, I. da Costa, NO prediction in natural gas flames using GDF-Kin®3.0 mechanism NCN and HCN contribution to prompt-NO formation, Fuel 85 (2006) 896–909. [51] S. Ishizuka, R. Hagiwara, M. Suzuki, A. Nakamura, O. Hamaguchi, Combustion characteristics of tubular flame burner, Jpn. Soc. Mech. Engrs 66 (1999) 3845–3851. [52] I. Yamaoka, H. Tsuji, Determination of burning velocity using counter-flow flames, Symp. (Int.) Combust. 20 (1985) 1883–1892. [53] B. Shi, D. Shimokuri, S. Ishizuka, Methane/oxygen combustion in a rapidly mixed tubular burner, Proc. Combust. Inst. 34 (2013) 3369–3377. [54] C.F. Coward, G.W. Jones, US Bureau of Mines, Bulletin 503 (1952). [55] S. Ishizuka, Determination of flammability limits using a tubular flame geometry, J. Loss Prev. Process Ind. 4 (1991) 185–193. [56] Y. Sung, S. Lee, S. Eom, C. Moon, S. Ahn, G. Choi, D. Kim, Optical non-intrusive measurements of internal recirculation zone of pulverized coal swirling flames with secondary swirl intensity, Energy 103 (2016) 61–74. [57] Y. Liu, P. Wang, J. Wang, Z. Du, Investigation of Taylor bubble wake structure in liquid nitrogen by PIV technique, Cryogenics 55-56 (2013) 22–29. [58] H. Guo, F. Liu, G. Smallwood, A numerical study on NOx formation in laminar counter-flow CH4 /air triple flames, Combust. Flame 143 (2005) 282–298.