air

Combustion and Flame 212 (2020) 403–414

Contents lists available at ScienceDirect

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

Effects of non-thermal plasma on the lean blowout limits and CO/NOx emissions in swirl-stabilized turbulent lean-premixed flames of methane/air Gyeong Taek Kim a, Chun Sang Yoo a,∗, Suk Ho Chung b, Jeong Park c,∗ a b c

Department of Mechanical Engineering, Ulsan National Institute of Science and Technology, Ulsan 44919, Republic of Korea Clean Combustion Research Center, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia Department of Mechanical Engineering, Pukyong National University, Busan 48547, Republic Korea

a r t i c l e

i n f o

Article history: Received 22 June 2019 Revised 11 August 2019 Accepted 14 November 2019

Keywords: Swirl-stabilized lean-premixed methane/air flames Non-thermal plasma (NTP) Dielectric barrier discharge (DBD) Ozone Lean blowout (LBO) limit CO/NOx emissions

a b s t r a c t This study investigates experimentally the effects of non-thermal plasma (NTP) induced by a dielectric barrier discharge (DBD) reactor on the characteristics of swirl-stabilized turbulent lean-premixed methane/air flames in a laboratory scale combustor by systematically varying the applied AC voltage, VAC , and frequency, fAC . Especially, it is elucidated how the NTP influences the lean blowout (LBO) limits and the characteristics of CO/NOx emissions depending on flame configuration. Without applying the NTP as the mixture equivalence ratio, φ , decreases from the stoichiometry to an LBO limit, the flame configuration changes from an M-flame (Regime I) to a conical flame (Regime II) and to a columnar flame (Regime III) for the whole range of the mixture nozzle exit velocity, U0 , (4–10 m/s). With the NTP, however, it exhibits only Regimes I and II at relatively-low U0 range (4–6 m/s), while all three regimes at relativelyhigh U0 range (7–10 m/s). For both velocity ranges, the LBO limits are significantly extended by the NTP enhancing the flame stability. Under the relatively-low U0 range, streamers induced by the DBD reactor play a critical role in stabilizing the flames such that the degree of extension of the LBO limit depends linearly on VAC and fAC . Under the relatively-high U0 range, however, ozone generated by the DBD reactor in Regime III is found to be a major reason in extending the LBO limit, which is substantiated by another flame regime diagram with ozone addition only, and hence, the extension of LBO limit minimally depends on fAC . Simultaneously, the NTP considerably reduces CO emission, while slightly increases NOx emission near the LBO limits due to the enhanced combustion by ozone. © 2019 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction In gas turbine combustors and industrial burners, lean premixed combustion has been widely adopted to comply with strict NOx emission regulations because thermal NOx can be significantly reduced at low flame temperatures [1–3]. Since progressively stringent emission regulations are expected in the foreseeable future, it is required to develop a novel premixed combustion system operating under extremely-lean conditions. For this purpose, numerous experimental and numerical studies of low-NOx gas combustors have been performed [4–9] and have successfully demonstrated the effectiveness of lean-premixed gas combustors and their emission control.

∗

Corresponding authors. E-mail addresses: [email protected] (C.S. Yoo), [email protected] (J. Park).

At such extremely-lean conditions, however, premixed flames become vulnerable to combustion instability, especially to flame blowout. When the flow speed is high enough or the mixture equivalence ratio is low enough in a premixed gas combustor, the premixed flames cannot be stabilized at a desired position and subsequently propagates downstream, eventually leading to flame blowout [10,11]. Therefore, it is needed to develop a new technology to effectively stabilize flames within a combustor together with a swirler or a bluff body, while further reducing NOx /CO emissions from the current levels without employing expensive facilities for exhaust gas clean-up. During the past two decades, plasma has emerged as one of the promising techniques to enhance combustion, reduce pollutant emissions, and control combustion instability [12–14]. Many studies have successfully demonstrated that NOx and/or particulate matter (PM) emissions can be significantly reduced by utilizing various types of plasmas such as gliding arcs [15], pulsed corona discharge [16–18], and dielectric barrier discharge (DBD) [19,20].

https://doi.org/10.1016/j.combustflame.2019.11.024 0010-2180/© 2019 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

404

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

Especially, the applicability of non-thermal plasma (NTP) such as microwave discharge and DBD to combustion systems has been investigated because it has relatively-low power consumption compared to thermal plasmas [21–24]. More specifically, Cha et al. [23] applied DBD to a flame region and demonstrated that the NTP applied through a DBD can suppress soot generation in co-flow jet diffusion flames by showing significant reduction of yellow luminosity radiated from soot particles. Cathey et al. [22] found that transient NTP can enhance ignition delay, peak pressure, and heat release rate by inducing streamers, as compared with conventional spark ignition systems. They suggested that the transient NTP has great potential for improving lean combustion and controlling pollutant emissions. Song et al. [24] characterized the reduction of PM, hydrocarbons (HC), and NOx emissions by applying a DBD to diesel exhaust gas. They reported that PM and HC emissions are decreased with increasing the peak voltage of the DBD, of which removal efficiency ultimately levels off at high voltages. However, the removal efficiency of NOx is reduced with increasing the peak voltage at high voltages. NTPs have also been applied to laboratory-scale dump combustors to enhance combustion stability [25–30]. The effects of nano-second pulsed discharges (NSPD) on the dynamics of swirlstabilized flames were investigated in several studies [25–28]. Especially, Lacoste et al. [25] demonstrated that an unstable flame can be effectively stabilized by applying NSPD because it can reduce velocity fluctuation amplitude by an order of magnitude due to its thermal effect. Versailles et al. [29] elucidated the effects of DBD on flame flashback in a dump combustor. An ionic wind induced by a DBD reactor increases flow velocity in the boundary layer inside the fuel/air nozzle. Consequently, relatively-high velocity in the boundary layer can prevent flame flashback at high equivalence ratios, leading to the extension of the stable operation regime of the combustor. DBD reactors generate streamer discharges or in short, streamers which are a transient electric discharge, featuring the release and transmission of electricity with applied electric field through a gas [31]. The streamers are generated through synergistic interaction between electrons and electric field, and thus, they play a critical role in stabilizing flames by enhancing chemical reactions [32,33]. Another example of the DBD reactors is for large-scale ozone generation in the industry. Ozone can be generated by applying high AC voltage to a DBD reactor for which the discharge space is filled with oxygen or air. Chemical reaction pathways for ozone generation in a DBD reactor starts with the dissociation of oxygen molecules to oxygen atoms by the electron impact in an electric field, and subsequently, oxygen atoms react with oxygen molecules to yield ozone. Therefore, the energy efficiency of ozone generation is strongly related to the production efficiency of oxygen atoms in the DBD reactor [34–36]. In general, ozone initiates and accelerates the chain branching reactions of hydrocarbon oxidation, and hence, even small amount of ozone addition to flames can improve the combustion efficiency [37–39]. These previous studies have successfully demonstrated that the NTP can control flame stability and reduce pollutant emissions by generating streamers and/or ozone. However, it has not been fully elucidated how the NTP can enhance lean premixed combustion according to flame configurations and change the characteristics of pollutant emission in a premixed gas combustor. Therefore, the objective of the present study is to experimentally investigate the effects of NTP applied through a DBD reactor on turbulent lean-premixed methane/air flames in a laboratory scale swirlstabilized dump combustor. Especially, the influences of NTP on the extension of lean blowout (LBO) limit of the premixed flames and the characteristics of CO/NOx emissions are elucidated. For the present study, we adopt the DBD technique to generate NTP because it can be applied to conventional gas combustors with

only a simple modification [30]. Note that the plasma effects on the LBO and CO/NOx emissions of swirl-stabilized and bluff-body stabilized methane/air flames have been investigated in previous studies [30,40]. In the present study, however, we focus more on the NTP effects on the LBO limit extension, which appear in different forms depending on the mixture velocity or the flame configuration, especially the effect of ozone addition. As such, the degree of LBO limit extension can be characterized by the applied voltage and frequency of NTP and the mixture velocity. 2. Experiment The experimental apparatus consists of a swirl-stabilized dump combustor, a gas supply system, an NTP-generating system, a gas analyzer, an ozone generator and analyzer, and a visualization system, as schematically shown in Fig. 1. The burner has a center body made of stainless steel with the diameter 12 mm through which the fuel (methane, 99.95%) is supplied. Near the exit, it has a tapered region (23.6◦ ). Compressed dry air is supplied through a concentric region with the diameter of 18 mm. The exit region is also tapered and the exit diameter is 11 mm. A single axial swirler with the vane angle of 45◦ is installed just before the tapered region. The fuel is injected into the concentric region via ten holes with 0.1 mm diameter on the center body at 23 mm below the swirler and at the exit of the swirler (see the red lines in Fig. 1a). The fuel and air are expected to be reasonably well mixed before entering the combustor section. A ceramic nose cone with 3 mm diameter, which protrudes by 8 mm from the mixture exit, is installed to effectively generate streamers near the flame edge. To visualize flames and streamers, the combustion chamber is surrounded by a quartz tube (inner and outer diameters are 33.5 and 40 mm, respectively). For the present study, flow rates of fuel and synthetic air (21% O2 and 79% N2 by volume) are determined according to the mean velocity at the mixture nozzle, U0 , and the corresponding mixture equivalence ratio, φ , is evaluated based on them. The flow rates are controlled by mass flow controllers. In the present study, φ varies from 1.0 to a value at which the flame is blown out for a given U0 , which changes from 3 to 10 m/s. Therefore, the mixture Reynolds number, ReDh ( ≡ Dh U0 /ν ), varies from 1911 to 6370, where Dh and ν are the hydraulic diameter and the kinematic viscosity at the mixture exit, respectively. Note that measurement error in φ is approximately 1% for a given U0 . For U0 , its measure error is less than 0.5%. The swirl intensity is characterized by the swirl number defined by the ratio of the axial flux of the tangential momentum to the product of the axial momentum flux and the characteristic radius [41,42]. The specific swirl number needs to be determined based on the injector configuration and flow profiles. For a typical singleelement injector with a flat vane swirler, the swirl number, S, can be expressed as [41,43]:

2[1 − (ri /ro ) ] 3

S=

3[1 − (ri /ro ) ] 2

tanθ ,

(1)

where ri and ro are the inner and outer radii of the swirler with 6 and 9 mm, respectively, and θ is the angle of the swirler vane of 45◦ . As such, S is estimated to be 0.84, which is large enough to generate vortex breakdown and develop a central toroidal recirculation zone downstream of the nozzle [41,44]. The recirculation zone extends as the swirl number increases [45]. Such flow structure could provide a dominant flame stabilization mechanism. The NTP generator is composed of a DBD reactor and a power supply system (Trek, 10/10B-HS). The DBD reactor is composed of the stainless cone section (applying high voltage) and a cylindrical woven stainless-steel mesh electrode with 20 mm in height and 1 mm thickness surrounding the outer quartz tube. The stainless

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

Fig. 1. Schematics of (a) the swirl combustor with non-thermal plasma generator and (b) the experimental setup.

405

406

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

steel center body serves as a central high voltage electrode while the mesh acts as the ground electrode. The installation location of the mesh is optimized above the ceramic mixing chamber to maximize the generation of streamers. As mentioned previously, the ceramic sharp nose cone is installed on top of the center body. The reason being that is it can increase the circumferential distance at the top of the stainless center body in generating more distributed streamers and increasing optimal ranges of applied root mean voltage, VAC , and applied frequency, fAC , as compared with a pointing center body and our previous tip configuration with 8 pointed tips on top of a center body with flush end [30]. For instance, streamers are generated for 0.2 ≤ fAC ≤ 2.2 kHz at 7 kV (rms) with previous sharp tips because there exists a limitation of frequency over which the power supply is shut down at a fixed voltage. However, it can be operated for 1 ≤ fAC ≤ 4 kHz at 7 kV (rms) with the installation of the present ceramic cone, which significantly intensifies streamers. Note that when fAC increases gradually for a given VAC , the induced streamers become stronger because higher electric power can be achieved as in [46,47]. In this study, streamers are generated only when flames exist. This result implies that streamers and flames have synergistic effects on each other [32,33], which will be discussed later. The dielectric barrier such as the quartz tube and the ceramic nose cone enables electrons to be accumulated dynamically on the electrodes, consequently allowing them to generate an opposing electric field to the applied field in its every half cycle [48]. This opposing field is crucial in preventing glow-to-arc transition at atmospheric pressure. The electric power is supplied to the fuel nozzle by applying AC voltages and frequencies in the range of 5–7 kV and 1–4 kHz, respectively. In a comparative study with ozone addition, a small fraction of the O2 stream is by-passed to the ozone generator (Ozonetech, LAB-II) as shown in Fig. 1b to achieve high yields of ozone without generating undesirable plasma byproducts such as NOx [49]. The ozone concentration in the synthetic air is then measured with an ozone analyzer (Ozonetech, OM-1500B), of which measurement error is found to be less than 1%. As such, we can accurately measure absolute ozone concentration supplied to the fuel/air mixture. Finally, the synthetic air stream containing ozone is mixed with the fuel at the mixing region. Therefore, the equivalence ratios for both flames with plasma and with ozone addition are identical. To measure the relative chemiluminescence intensity of ozone in flames with/without plasma and the ozone from the ozone generator only, a bandpass filter (254 ± 10 nm) is used to improve the fidelity of the measurement by blocking ambient light. The wavelength of 253.7 nm is used based on a peak absorption cross section area of ozone of 1.137 × 10−17 cm2 at 300 K [50,51]. To measure CO and NOx emissions, exhaust gas is sampled at the exhaust end of the combustor marked as (ii) in Fig. 1b and analyzed using a gas analyzer (Maxilyzer NG). A digital camera is used for direct flame visualization. It has been reported that spontaneous light emission from streamers in air primarily occurs by N2 (C3 u − B3 g ) transition at 337.1 nm [52]. Thus, the streamers can be visualized with an intensified charge-coupled device (ICCD) camera (Princeton Instrument, PI-MAX4:2048f) through a narrow bandpass filter (337 ± 10 nm) to minimize the interference from the chemiluminescence of flame. In addition, the streamer field is measured with a high-speed camera (Olympus, iSpeed3) with 300 fps. 3. NTP characteristics in flames From the DBD reactor, non-thermal plasma is induced in the form of electric discharges or streamers that are highly affected by flames [23,30,53]. For instance, it was found from a previous study [30] that the existence of flame significantly enhances the plasma

generation compared to nonreacting flow without flame. To identify the effect of flame on the generation of NTP in the present swirl combustor, the characteristics of streamers induced by the DBD reactor are investigated first by varying the mixture equivalence ratio, φ . Figure 2 shows the temporal evolutions of voltage and current between the electrodes measured by the oscilloscope in the presence of flame with three different φ . In addition, those for a nonreacting flow is also shown in the figure. When AC with VAC = 7kV and fAC = 4 kHz is applied, the voltage exhibits a sinusoidal waveform and its peak value is approximately 10.6 kV. It is readily observed from Fig. 2a that several spikes of electric current appear during a period of the voltage variation. Such electric current spikes represent the occurrence of streamers induced by the DBD reactor in the flame, featuring filamentary microdischarges with a duration of a few nanoseconds [54]. The streamers tend to be randomly distributed in the flame front region as well as in time (see Supplementary Material-I (SM-I)). Note that in the presence of flame, the streamers are less frequently generated and their intensities become weaker as φ decreases from 0.9 to 0.52, while there exists no streamers for the nonreacting flow. This implies that the characteristics of streamers are highly affected by the flame strength. In other words, the enhancement of streamer generation is primarily attributed to high gas temperature and abundant ions in flames [54,55]. In addition, the current spikes appear only when the central electrode becomes anode or during the positive-half cycle of the voltage variation because the streamers are usually induced through typical positive corona [56]. More details of steamer generation in flames were reported previously [23,30,53]. To further characterize the NTP, we measure the power consumption of NTP based on a method proposed in [40,57,58]. As shown in Fig. 3, the power consumption of NTP increases with increasing VAC and fAC while it does not change much with U0 and φ . The maximum power consumption is found to be approximately 20 W at 7 kV and 4 kHz, corresponding to 1–2% of thermal loading from combustion near the LBO limit similar to [40]. This result implies that it is reasonable to use NTP for combustion enhancement because only a small fraction of thermal load is needed to generate the NTP. 4. Overall flame characteristics 4.1. Flame regimes without NTP Figure 4 shows the direct images of the lean premixed flames stabilized in the burner. Similar to previous studies [59,60], typical flames in the present study can be classified into three different modes according to their configurations when the equivalence ratio varies at a specified mixture velocity: (1) M-flame (Regime I), (2) conical flame (Regime II), and (3) columnar flame (Regime III) as shown in Fig. 4a. For U0 = 8 m/s (Fig. 4b), for example, the flame shape changes from an M flame (φ = 0.9) to a conical flame (φ = 0.57) to a columnar flame (φ = 0.54) with decreasing φ , and it is ultimately blown out beyond a LBO limit. The regime diagram of swirl-stabilized turbulent lean-premixed methane/air flames without NTP in the U0 − φ space is shown in Fig. 4a. Note that the experimental procedure is that the boundaries of all the flame regimes are identified by decreasing φ from the stoichiometry to φ at the LBO limit for a fixed U0 . In Regime I, a stable M-shaped flame attached near ceramic nose cone is observed as shown in Fig. 4b-i. As φ decreases for a given U0 , however, an unstable flame mode develops at the transition from Regime I to Regime II, marked as the blue symbols. At such equivalence ratio, the flame shape starts to alternate between an M-flame and a conical flame for the whole velocity range, which is often called the outer recirculation zone flame flickering

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

407

Fig. 2. Temporal evolutions of voltage-current waveforms at VAC = 7 kV and fAC = 4 kHz in the presence of flame with (a) φ = 0.9, (b) 0.54, and (c) 0.52, and for (d) nonreacting flow with φ = 0.9 at U0 = 8 m/s.

Fig. 3. Power consumption of NTP as a function of VAC for various fAC for premixed flame with φ = 0.5 at (a) U0 = 5 m/s and (b) U0 = 8 m/s.

Fig. 4. (a) Flame regime diagram of swirl-stabilized turbulent lean-premixed methane/air flames without applying non-thermal plasma in terms of mixture velocity, U0 , and equivalence ratio, φ , and (b) representative flame images of (i) Mflame (Regime I; φ = 0.90), (ii) conical flame (Regime II; φ = 0.57), and (iii) columnar flame (Regime III; φ = 0.54) with U0 = 8 m/s.

[59]. As φ is further decreased, another flame shape change occurs from a conical flame (Regime II) to a columnar flame (Regime III) as shown in Figs. 4b-ii and iii. Finally, the flame is lifted off from the ceramic nose cone and is then blown out beyond the LBO limit.

It is also readily observed increases with decreasing φ . previous experimental studies [30,59,60]. Moreover, under

from Fig. 4b that the flame length These results are consistent with of swirl-stabilized premixed flames relatively-high velocity conditions

408

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

Fig. 5. (a) Flame regime diagram with/without non-thermal plasma in terms of mixture velocity, U0 , and equivalence ratio, φ , at VAC = 7 kV and fAC = 4 kHz, and (b) representative flame images with/without non-thermal plasma at (i) φ = 0.90, (ii) 0.54, and (iii) 0.52 with U0 = 8 m/s.

(7 – 10 m/s), Regimes II and III exist in a broader area in the U0 − φ space than those under relatively-low velocity conditions (4 – 6 m/s). In general, the flame strength becomes weaker with decreasing φ and increasing U0 . Thus, the regime transition from Regime I to Regime II to Regime III occurs at larger φ with increasing U0 as shown in Fig. 4a [61]. As the columnar flame in Regime III approaches its LBO limits, it is lifted off and elongated along the inner recirculation zone. For large U0 , however, the inner recirculation zone becomes stronger and holds the lifted flame more effectively with increasing U0 [62]. Therefore, the lifted flame is blown out at smaller φ with increasing U0 under relatively-large U0 conditions (7–10 m/s), while the inner recirculation zone is not fully developed at relatively-small U0 (4–6 m/s), and consequently, the columnar flame comes to be blown out at larger φ with increasing U0 . 4.2. Flame regimes with NTP Figure 5 shows the regime diagram with NTP at VAC = 7 kV and fAC = 4 kHz in the U0 − φ space. For comparison, the regime diagram without applying NTP (see Fig. 4) is marked as the dotted lines together with their representative snap shot images with/without NTP at φ = 0.9, 0.54, and 0.52. Since the regime diagram is influenced with varying VAC and fAC , a case most significantly affected by the NTP is shown in Fig. 5. Three points are noted. First, the NTP can significantly extend the LBO limits for the whole velocity range as compared with those without applying NTP, especially at relatively-large U0 conditions. More specifically, Regimes II and III are extended appreciably in terms of φ range with the NTP. However, Regime I is minimally influenced by the NTP except for U0 = 3 m/s. Moreover, the flame flickering does not occur for the flames with NTP at the transition from Regime I to Regime II. These results imply that the enhancement of combustion by the NTP can effectively stabilize the flames in Regimes

II and III such that stable combustion can be achieved in a wider range of φ , leading to the extension of the LBO limits. It is of interest to note that in Regime I, very strong streamers are generated everywhere in the flame zone while in Regime II, streamers are mostly generated near the flame fronts (see Fig. 5b and SMI). As such, streamers in Regime I may not play an important role in stabilizing the flame nor affect Regime I boundary extension; however, those in Regime II can help to stabilize the flame, consequently extending the boundary of Regime II. Second, the LBO limit with NTP is extended with increasing U0 in the relatively-high U0 range, which is opposite to the trend without NTP. At relatively-high U0 , the LBO limit without NTP is extended due to recirculation zone enhanced by high velocity as discussed above. In the presence of NTP, however, the effect of NTP overwhelms that of enhanced recirculation zone, which renders turbulent premixed flame with NTP to be weak with increasing U0 . Third, the LBO limit extension is achieved by Regime II extension at relatively-low U0 conditions (4–6 m/s) and by Regime III extension at relatively-high U0 conditions (7–10 m/s), which implies that the combustion enhancement mechanism by the NTP may differ for the two different regimes or for the high- and lowU0 conditions. The effects of NTP on the flames in different regimes are further elucidated by comparing their combustion characteristics with/without NTP at φ = 0.9 and 0.54 (see Figs. 5b-i and ii). For the flame with NTP at φ = 0.9, the streamers propagate radially towards the outer electrode of woven mesh once they are generated from the central electrode that coincides with the flame front. As such, they are mostly observed at the flame front region near the mixture nozzle as shown in Fig. 5b-i. Since abundant positive ions exist in high temperature flame region, the coincidence of streamer and flame locations indicates that the flame can serve as an extended electrode by reducing the effective distance between the electrodes [53]. This is also caused by synergistic interaction

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

409

Fig. 6. Snap shots of streamers taken by ICCD camera through 337-nm filter with 100-ms exposure at various fAC for premixed flames near LBO limits with (a) U0 = 5 m/s, VAC = 7 kV, and φ = 0.55 in Regime II, and (b) U0 = 8 m/s, VAC = 7 kV, and φ = 0.50 in Regime III.

between streamers and flames. In general, abundant positive ions such as CHO+ , C2 H3 O+ , H3 O+ , and CH3 + in flame enhance the electric current in a unipolar electric field that effectively increases the streamer generation, which strengthens the flame [53,63,64]. Thus, the flames with NTP in Regime I and II strengthen themselves by enhancing the generation of streamers. Due to such synergistic interaction, the intensity and distribution of streamers are stronger and wider in Regime I (φ = 0.9) than those in Regime II (φ = 0.54) as shown in Figs. 5b-i and ii, which can be also identified from Figs. 2a and b. Moreover, a columnar flame without NTP in Regime III is drastically changed to a conical flame in Regime II by the NTP as shown in Fig. 5b-ii (also see SM-II). Finally, when AC is applied to the electrodes at relatively-high U0 , a conical flame (Regime II) changes to a columnar flame (Regime III) with decreasing φ , consequently leading to the extension of the LBO limits. However, the intensity and distribution of streamers become significantly weak and narrow, respectively, in Regime III, compared to those in Regime II as shown in Figs. 5b-ii and iii. Although the streamers exist near the fuel nozzle, the LBO limits are significantly extended under relatively-high U0 conditions (7–10 m/s). As such, it is reasonable to expect that there could be another factor affecting the LBO limits other than streamer generation. Based on the observation of stronger streamers occurring in Regime II than Regime III and previous experimental studies of ozone effect on combustion enhancement [37,39,65], we can conjecture that the streamers induced by the DBD reactor may extend Regime II at relatively-low U0 (4 – 6 m/s) while the ozone produced from the NTP enlarges Regime III at relatively-high U0 (7 – 10 m/s). This point is further discussed below. 5. Effects of NTP on lean blowout limit 5.1. Effect of streamers at low U0 As discussed above, the LBO limit extension at relatively-low U0 is attained mainly through Regime II, which is conjectured to occur by the synergistic interaction between flames and streamers (see Fig. 5). In general, NTP, which is induced in the form of streamers, can enhance reactions in flames by increasing temperature, producing a pool of ions/electrons, excited species, and radicals, and/or inducing ionic winds and fuel fragmentation [32,33]. Since such thermal, kinetic, and transport effects cannot be differentiated from one another in the present study, henceforth we collectively refer to them as the “streamer effect”. To further elucidate the generation of streamers for flames near the LBO limits of Regimes II and III, streamer images were taken by the ICCD camera through 337 nm bandpass filter with short time exposure of 100 ms as shown in Fig. 6, where bright regions rep-

resent the accumulated paths of streamers. The streamer intensity is increased with increasing fAC for a specified VAC at U0 = 5 m/s (Regime II), while the intensity becomes significantly weaker at U0 = 8 m/s (Regime III). Although not shown here, the streamer intensity also increases with increasing VAC at specified U0 = 5 m/s. These results indicate that streamer generation is appreciably influenced by both the applied VAC and fAC in Regime II, whereas it is much suppressed in Regime III as identified from the direct flame images in Fig. 5. To characterize quantitatively the streamer generation near the LBO limits in Regimes II and III, the averaged number of streamer generation per cycle, Nst , (refer to Fig. 2) and the averaged maximum intensity of streamers from the current, Ist,max , are measured from the oscilloscope signals by varying the applied VAC and fAC as shown in Fig. 7. Note that Nst and Ist,max are averaged over 100 cycles. It is readily observed that for U0 = 5 m/s (Regime II), Nst increases with increasing fAC and Ist,max does not change much with fAC while both Nst and Ist,max are monotonically increased with increasing VAC . However, for U0 = 8 m/s (Regime III), Nst is appreciably smaller and Ist,max is significantly weaker, as compared to those for U0 = 5 m/s. These results support that streamers with high intensity are more frequently generated by the synergistic interaction with flames in Regime II than in Regime III, which in turn leads to the LBO limit extension of Regime II as shown in Fig. 5. These streamer characteristics indicate that the premixed flames in Regime II become stronger with increasing Nst and Ist,max that monotonically increase with VAC and fAC . As such, it can be expected that the LBO limits can be more extended by streamers with increasing VAC and fAC [63]. Moreover, premixed flames generally become weaker with increasing U0 , leading to the LBO limit reduction. Therefore, one may correlate the degree of extension of φ at the LBO limit by streamers with VAC · fAC · U0−1 for various VAC , fAC , and U0 as shown in Fig. 8. Note that the degree of extension of φ , Y, is defined by the difference between φ at the LBO limit with and without NTP, |φLBO,AC − φLBO,0 |, normalized by φ at the LBO limit without NTP for a specified U0 , φ LBO,0 . The result shows that the degree of extension of φ at the LBO limits is well correlated with X = VAC · fAC · U0−1 and the best linear fit is found as Y = 0.01X with the correlation coefficient of R = 0.86. This correlation substantiates that the LBO limits are extended by streamers generated from the DBD reactor at relativelylow U0 and the degree of its extension is linearly proportional to the applied AC voltage and frequency. For instance, the degree of extension of φ is approximately 8.4% at VAC = 7 kV and fAC = 4 kHz, and U0 = 4 m/s while it is only 2% at VAC = 6 kV and fAC = 2 kHz, and U0 = 6 m/s. It is of interest to note that for VAC < 6 kV and fAC < 2 kHz, the NTP has no effect on the extension of the LBO limits because streamers are not generated below the threshold values of AC voltage and frequency [54].

410

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

Fig. 7. Variations of the averaged number of streamer generation per cycle, Nst , (a,c) and the averaged maximum intensity of streamers, Ist,max , (b,d) as a function of fAC (top) and VAC (bottom) for U0 = 5 m/s (Regime II) and 8 m/s (Regime III).

Fig. 8. Correlation of the degree of extension of φ at LBO limit by the influence of streamers with VAC · f AC · U0−1 under relatively-low velocity conditions (4 – 6 m/s)..

5.2. Effect of ozone at high U0 As discussed above, the LBO limits are found to be extended by streamers at relatively-low U0 in Regime II (see Fig. 7 and SM-III). However, only weak streamer or corona is observed at relativelyhigh U0 in Regime III (see Fig. 7 and SM-IV). Since the DBD reactor can generate ozone without strong streamers [66,67], we conjecture that ozone generated by the DBD reactor could affect chemical reactions near flame region, enhancing the combustion in Regime III and ultimately extending the LBO limits. Henceforth, we refer to it as the “ozone effect”, which is actually related to the kinetic effect of NTP to generate radicals such as O, O3 , and NO [32,33]. To test the effect of ozone generation on the extension of the LBO limit, we first measure the relative chemiluminescence intensity of ozone induced by the NTP for various VAC and fAC using the ICCD camera with a narrow bandpass filter (254 nm ± 10) as explained in Section 2. Since we need relative ozone amounts at different conditions, we measure ozone intensity with

the ICCD camera rather than its concentration with extensive laser diagnostics. Figure 9 shows the radial distributions of ozone chemiluminescence intensity, IO3 , for various VAC and fAC , which is normalized by the maximum ozone intensity, IO3 ,max , at VAC = 7 kV and fAC = 4 kHz for φ = 0.50 and U0 = 8 m/s. Note that the axial location for measuring the ozone intensity is 5 mm above the mixture exit nozzle, where the flamebase is located. The radial direction is normalized by the flame radius, rf , that is approximately 5 mm at the flamebases of the columnar flames. Also note that the asymmetric measured ozone intensity is primarily attributed to the limitation of ozone measurement through the obstruction of the woven mesh in the DBD reactor. It is readily observed from Fig. 9 that the ozone chemiluminescence intensity is increased with increasing VAC when VAC is greater than 5.0 kV for fAC = 4 kHz. Theoretically, an oxygen molecule is split into two single atoms by NTP, which then recombine in triplets to form ozone [35,49]. As such, when the electric energy applied to the DBD reactor is not high enough, oxygen molecules would not be decomposed, leading to low level of ozone concentration similar to that with VAC = 5.0 kV in Fig. 9a. It is also found from Fig. 9b that the effect of fAC on the ozone intensity is negligible, similar to [57]. These results imply that the LBO limit extension by ozone is affected mainly by VAC . Although the ozone concentration in the lean premixed flames was only qualitatively measured, we could effectively estimate the relative amount of ozone generation for different VAC and fAC by measuring its chemiluminescence intensity with the ICCD camera. In this time, we supply ozone generated from the ozone generator directly to flame without NTP and measure its ozone chemiluminescence intensity at 5 mm above the mixture exit nozzle. By comparing the ozone intensities for cases with ozone addition only to those for cases with/without NTP, we can quantitatively estimate the ozone amount generated by NTP as shown in Fig. 10. As discussed in Section 2, ozone supplied directly to the flame is generated from the oxygen flow by the ozone generator and its concentration is accurately measured by the ozone analyzer (see Fig. 1b). The ozone chemiluminescence intensity is measured

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

411

Fig. 11. Flame regime diagram of swirl-stabilized turbulent lean-premixed methane/air flames with three different levels of ozone addition only such as 10 0 0, 40 0 0, and 10 0 0 0 ppm in terms of mixture velocity, U0 , and equivalence ratio, φ .

Fig. 9. Normalized radial distributions of ozone intensities divided by the maximum intensity with VAC = 7 kV and fAC = 4 kHz along the fuel nozzle electrode blow ceramic nose cone at (a) various voltages and (b) various frequencies with φ = 0.50 and U0 = 8 m/s.

Fig. 10. Radial profiles of ozone intensity at various conditions: without plasma and ozone addition for φ = 0.55, with plasma at VAC = 7 kV and fAC = 2 kHz for φ = 0.50, and with 40 0 0 and 10 0 0 0 ppm ozone additions for φ = 0.50.

under various conditions: without plasma and ozone addition at φ = 0.55, with plasma at VAC = 7 kV and fAC = 2 kHz at φ = 0.50, and with 40 0 0 and 10 0 0 0 ppm ozone addition at φ = 0.50. We adopted two different equivalence ratios for the ozone intensity measurement because the LBO limits are different from one another: for instance, it is φ = 0.522 without plasma and ozone addition, φ = 0.467 with plasma, and φ = 0.495 with the addition of ozone of 40 0 0 ppm, respectively. Note that ozone addition without plasma can also extend the LBO limits. This is because O3 molecules supplied to the mixture nozzle split into O radicals in the preheat zone, which initiate and accelerate the chain-branching reactions of hydrocarbon oxidation, ultimately enhancing combustion [37–39].

Several points are noted from Fig. 10. First, there are no ozone traces for the case without plasma and without ozone addition as can be expected. Second, the intensity and distribution of ozone become stronger and wider as the amount of ozone addition increases from 40 0 0 to 10 0 0 0 ppm. Moreover, the maximum ozone intensity with 10 0 0 0 ppm ozone addition becomes comparable to that with NTP, which substantiates that the LBO limit is also extended by the ozone addition. Third, ozone generated by NPT is concentrated near the central electrode where corona is observed (see SM-IV); however, the supplied ozone through the ozone generator is relatively broader than that with NTP, which is primarily attributed to the ozone supply method through the mixture nozzle. This result implies that ozone can be generated by O atoms induced by NTP reacting with O2 between the central electrode and the lifted flame even though O atoms typically react with fuel molecules much faster than they recombine with other O2 molecules in flame. These results substantiate that the ozone generated by NTP can enhance the chain-branching reactions of the premixed flames, effectively leading to the extension of the LBO limits at relatively-high U0 (Regime III). To directly estimate the effect of ozone addition on the LBO limit extension, another flame regime diagram for premixed flames with different amounts of ozone addition is shown in Fig. 11. For comparison purposes, the flame regime diagrams with/without NTP are also shown in the figure. As the amount of ozone addition is increased from 10 0 0 ppm to 10 0 0 0 ppm, the LBO limit is significantly extended in the relatively-high U0 range (7–10 m/s); especially, the LBO limit with 10 0 0 0 ppm ozone addition becomes comparable to that by NTP. In the relatively-low U0 range (4– 6 m/s), however, the LBO limit remains nearly identical to that without NTP regardless of the amount of ozone addition. These results substantiate that the LBO limit extension in the relativelyhigh U0 range is primarily attributed to ozone generation by NTP. Moreover, we can also conjecture from the results that a conical flame in Regime III can survive at lower φ with the help of ozone, only when the inner recirculation zone becomes strong enough to stabilize it. As a columnar flame approaches its LBO limit in the relatively-high U0 range, it is lifted off by high U0 and is stabilized by the relatively-strong inner recirculation zone, which subsequently provides enough space between the flamebase and the mixture nozzle exit for ozone generated by NTP or ozone supplied from the ozone generator to enhance reactions at the flamebase, leading to the LBO limit extension. In the relatively-low U0 range, however, a columnar flame without NTP cannot survive at low φ regardless of the amount of ozone addition because its inner

412

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

Fig. 12. Correlation of the degree of extension of φ at LBO limit by the influence of ozone with VAC · U0−1 under relatively-high velocity conditions (7–10 m/s).

recirculation zone is not fully developed and is too weak to stabilize the flame. Consequently, the columnar flame in the relativelylow U0 range is readily blown out despite the help of ozone immediately after a flame mode change occurs from a conical flame to a columnar flame. This result also supports that in the relativelylow U0 range, the extension of the LBO limit cannot be attributed mainly to the effect of ozone generation. The ozone effect on the chain-branching reactions implies that the premixed flames in Regime III become stronger with increasing ozone concentration that is monotonically proportional to VAC for a specified U0 . As such, it is expected that the LBO limits can be more extended by the amount of ozone generated by the NTP with increasing VAC . Therefore, one can correlate the degree of extension of φ at the LBO limits by ozone with VAC · U0−1 for relatively-high U0 (7 – 10 m/s) with VAC = 6–7 kV and fAC = 2–4 kHz as shown in Fig. 12. As readily observed from the figure, the degree of extension of φ at the LBO limits is well correlated with X = VAC · U0−1 and the best correlation is found as Y = 0.40X − 0.23 with the correlation coefficient of R = 0.93, which also verifies that the LBO limits are extended by ozone generated by NTP at relatively-high U0 and the degree of its extension is linearly proportional to the applied AC voltage. The effect of fAC on the LBO limit extension is marginal at relatively-high U0 conditions even though very weak electric discharge occurs near the bottom of flame. In addition, the effect of NTP on the LBO limit extension at relatively-high U0 is more significant than that at relatively-low U0 . For instance, the maximum degree of extension of φ is 18.7% at relatively-high U0 (7–10 m/s), which is much larger than that of 8.4% at relativelylow U0 (4–6 m/s). It is of importance to note that the effects of NTP on the LBO limit extension highly depends on the flame configuration of swirlstabilized turbulent lean-premixed flames. When a conical flame (Regime II) exists near the LBO limit, the NTP enhances combustion through streamer discharges. However, the NTP improves the flame strength in the form of ozone when the premixed flame exhibits a columnar flame configuration (Regime III) near the LBO limits. As shown in Fig. 5b, the conical flame in Regime II covers the whole combustor, and hence, serves as an extended electrode by reducing the effective distance between the electrodes [53], leading to the improvement of streamer generation. However, once the flame configuration changes from a conical flame (Regime II) to a columnar flame (Regime III) due to high velocity and/or low equivalence ratio, the flame can cover only narrow region near the central fuel nozzle (see Fig. 5b), which comes to significantly reduce the streamer generation.

Fig. 13. Variations of the emissions of (a) NOx and (b) CO as a function of the equivalence ratio without plasma, with ozone addition only (40 0 0 ppm), and with the NTP (VAC = 7 kV and fAC = 4 kHz) for U0 = 8 m/s. The vertical dotted line represents φ at the LBO without NTP for U0 = 8 m/s, φ LBO,0 .

5.3. Effects of NTP on CO/NOx emissions In the previous sections, we discussed how the non-thermal plasma extends the LBO limits of swirl-stabilized turbulent leanpremixed methane/air flames. This LBO limit extension is usually achieved by enhancing combustion with streamers and ozone depending on the flame regimes. However, the combustion enhancement by NTP is usually accompanied by the reduction of CO emission and the increase of NOx emission. To quantitatively understand the effects of NTP on CO and NOx emissions, we measure CO and NOx emissions of the premixed flames with/without NTP and with ozone addition at U0 = 8 m/s as shown in Fig. 13. The amounts of CO and NOx emissions are measured using the exhaust gas analyzer and corrected on a dry basis of 15% oxygen. We cannot measure more than 10 0 0 ppm emission including both CO and NOx due to the limitation of the gas analyzer used for the present study, which is denoted by ‘measurement limitation’ in the figure. Several points are noted from Fig. 13. First, it is readily observed that NOx emission from the flame with NTP is decreased as the flame approaches the LBO limit. Its amount, however, is relatively larger than those without NTP and with ozone addition only. For the flame with NTP in Regime II, combustion enhancement by intense streamers increases local flame temperatures, ultimately leading to relatively-high level of NOx emission ( ≈ O(10) ppm) [32,33] compared to those without NTP and ozone addition only. In Regime III near the LBO limit, however, overall flame temperature is relatively low due to low φ and local flame temperature increase by ozone is marginal compared to that by streamers, resulting in a considerable decrease of NOx emission compared to that in Regime II. Second, CO emissions are significantly increased with decreasing φ for all flames, which is primarily attributed to the occurrence

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

of incomplete combustion as the flames move towards their LBO limits. However, CO emission from the premixed flame with NTP is relatively lower than those without NTP and with ozone addition only for a specified φ near the LBO limit. In general, ozone supplied to flames can enhance complete combustion, and hence, it can significantly reduce CO emission and somewhat increase NOx emission [37,68]. Note that NOx can be produced by the plasma, which may add to NOx emission. Typically, combustion enhancement by NOx occurs only when its amount is large enough (e.g. 10 0 0 – 10 0 0 0 ppm) [69]. However, the amount of NOx emission in the present experiments is at most 20 ppm as shown in Fig. 13 although it is not measured within the flames. The residence time for NOx formation is much longer (so its effective formation occurs in the post-flame zone of a premixed flame) than that for the oxidation reactions in a flame zone. As such, the formation of NOx is expected not to enhance combustion. In this context, the LBO limit extension in the relatively-high U0 range can be attributed to ozone generation. It will be interesting to test the combustion enhancement effect by NOx produced by a plasma in a future study.

413

Declaration of Competing Interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Acknowledgments This work was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (NRF-2018R1A2A2A05018901). SHC was supported by KAUST. Supplementary material Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.combustflame.2019.11. 024 References

6. Conclusions The effects of non-thermal plasma on the extension of the lean blowout limit and the characteristics of CO/NOx emissions of turbulent lean-premixed methane/air flames were investigated experimentally in the laboratory-scale swirl-stabilized dump combustor with the dielectric barrier discharge reactor by varying the applied AC voltage, VAC , and frequency, fAC . The followings are the results found from this study. 1. Without the NTP, the flames are categorized into three regimes according to their configurations: (1) M-flame in Regime I, (2) conical flame in Regime II, and (3) columnar flame in Regime III. When the NTP is applied to the flames, their configurations are drastically changed and the LBO limits are significantly extended. The LBO limit extension by the NTP occurs by the extension of Regime II at relatively-low U0 while it does by the extension of Regime III at relatively-high U0 . This is because the NTP affects the flames in the form of streamers/ozone at relatively-low/relatively-high U0 range, respectively. 2. In Regime II at relatively-low U0 conditions (4–6 m/s), the intensity and the generation frequency of streamers are found to be proportional to applied VAC and fAC , and hence, the degree of extension of φ at the LBO limits is well correlated with VAC · fAC · U0−1 . This result indicate that streamers induced by the DBD reactor play a critical role in stabilizing the flames by their synergistic interaction with flames. 3. In Regime III at relatively-high U0 (7–10 m/s), the ozone concentration is found to be proportional to only applied VAC . As such, the degree of φ extension at the LBO limits is found to be well correlated with VAC · U0−1 . This implies that the ozone induced by the DBD reactor in Regime III extends the LBO limit by enhancing complete combustion in the combustor. Another flame regime diagram with ozone addition only substantiates that the LBO limit extension in the relatively-high U0 range is primarily attributed to ozone generated by the NTP. 4. It is also found that the NTP can considerably reduce CO emission compared to those without NTP and with ozone addition only. However, the NTP produces O(10) ppm NOx emission in Regime II, which is significantly reduced in Regime III near the LBO limit. All of the results are attributed to the enhanced combustion by streamers or ozone produced by the DBD reactor.

[1] U. Stopper, M. Aigner, W. Meier, R. Sadanandan, M. Stohr, I.S. Kim, Flow field and combustion characterization of premixed gas turbine flames by planar laser techniques, J. Eng. Gas Turb. Power 131 (2008) 021504. [2] T. Lieuwen, B.T. Zinn, The role of equivalence ratio oscillations in driving combustion instabilities in low NOx gas turbines, Proc. Combust. Inst. 28 (1998) 1809–1816. [3] R.C. Rawlins, SoLoNOx combustin system update, Turbomachinery Technical Seminar (2003). [4] J.J. Keller, Thermoacoustic oscillations in combustion chambers of gas turbines, AIAA J. 33 (1995) 2285–2287. [5] A.A. Peracchio, W.M. Proscia, Nonlinear heat-release/acoustic model for thermoacoustic instability in lean premixed combustors, J. Eng. Gas Turb. Power 121 (1999) 415–421. [6] T. Lieuwen, B.T. Zinn, Theoretical investigation of combustion instability mechanisms in lean, premixed gas turbines, AIAA Paper (1988) 98–0641. [7] J.M. Cohen, T.J. Anderson, Experimental investigation of near-blowout instabilities in a lean, premixed step combustor, AIAA Paper (1996) 96–0819. [8] D.W. Kendrick, T.J. Anderson, W.A. Sowa, T.S. Snyder, Acoustic sensitivities of lean-premixed fuel injector in a single nozzle rig, J. Eng. Gas Turb. Power 121 (1999) 429–436. [9] M.C. Janus, G.A. Richards, M.J. Yip, E.H. Robey, Effects of ambient conditions and fuel composition on combustion stability, ASME Paper (1997) 97–GT–266. [10] U. Stopper, M. Aigner, W. Meier, R. Sadanandan, M. Stohr, I.S. Kim, Combustion dynamics and control: Progress and challenges, Proc. Combust. Inst. 29 (2002) 1–28. [11] T. Lieuwen, K. McManus, Introduction: Combustion dynamics in lean-premixed prevaporized (LPP) gas turbines, J. Propul. Power 19 (2003) 721. [12] A. Starikovskuiy, N. Aleksandrov, Plasma-assisted ignition and combustion, Prog. Energy Combust. Sci. 39 (2013) 61–110. [13] G. Lou, A. Bao, M. Nishihara, S. Keshav, Y.G. Utkin, J.W. Rich, Ignition of premixed hydrocarbon-air flows by repetitively pulsed, nanosecond pulse duration plasma, Proc. Combust. Inst 31 (2007) 3327–3334. [14] W. Kim, M.G. Mungal, M.A. Cappelli, The role of in situ reforming in plasma enhanced ultra lean premixed methane/air flames, Combust. Flame 157 (2010) 374–383. [15] A. Czemichowski, Gliding arc. Applications to engineering and environment control, Pure Appl. Chem. 66 (1994) 1301–1310. [16] B.M. Penetrante, R.M. Brusasco, B.T. Merritt, W.J. Pitz, Plasma-assisted catalytic reduction of NOx, SAE Tech. Papers (1998) 982508. [17] H.H. Kim, K. Takashima, S. Katsura, A. Mizuno, Low-temperature NOx reduction processes using combined systems of pulsed corona discharge and catalysts, J. Phys. D.: Appl. Phys. 34 (2001) 604–613. [18] V. Puchkarev, M. Gundersen, Energy efficient plasma processing of gaseous emission using a short pulse discharge, Appl. Phys. Lett. 71 (1997) 3364–3366. [19] K. Urashima, J.S. Chang, T. Ito, Energy efficient plasma processing of gaseous emission using a short pulse discharge, IEEE Trans. Ind. Appl. 33 (1997) 879–886. [20] A. Khacef, J.M. Cormier, J.M. Pouvesle, NOx remediation in oxygen-rich exhaust gas using atmospheric pressure non-thermal plasma generated by a pulsed nanosecond dielectric barrier discharge, J. Phys. D.: Appl. Phys. 35 (2002) 1491–1498. [21] C.E. Haselfoot, P.J. Kirkby, The electrical effects produced by the explosion of hydrogen and oxygen, Lond. Edinb. Dubl. Phil. Mag. 8 (1904) 471–481. [22] C.D. Cathey, T. Tang, T. Shiraishi, T. Urushihara, A. Kuthi, M.A. Gundersen, Nanosecond plasma ignition for improved performance of an internal combustion engine, IEEE Trans. Plasma Sci. 35 (2007) 1664–1668.

414

G.T. Kim, C.S. Yoo and S.H. Chung et al. / Combustion and Flame 212 (2020) 403–414

[23] M.S. Cha, S.M. Lee, K.T. Kim, S.H. Chung, Soot suppression by nonthermal plasma in coflow jet diffusion flames using a dielectric barrier discharge, Combust. Flame 141 (2005) 438–447. [24] C.L. Song, F. Bin, Z.M. Tao, F.C. Li, Q.F. Huang, Simultaneous removals of NOx, HC and PM from diesel exhaust emissions by dielectric barrier discharges, J. Hazard. Mater. 166 (2009) 523–530. [25] D.A. Lacoste, J.P. Moeck, D. Durox, C.O. Laux, T. Schuller, Effect of nanosecond repetitively pulsed discharges on the dynamics of a swirl-stabilized lean premixed flame, J. Eng. Gas Turb. Power 135 (2013a) 101501. [26] D.A. Lacoste, D.A. Xu, J.P. Moeck, C.O. Laux, Dynamic response of a weakly turbulent lean-premixed flame to nanosecond repetitively pulsed discharges, Proc. Combust. Inst. 34 (2013b) 3259–3266. [27] W. Kim, J. Snyder, J. Cohen, Plasma assisted combustor dynamics control, Proc. Combust. Inst. 35 (2015) 3479–3486. [28] W. Kim, J. Cohen, Plasma-assisted combustor dynamics control at ambient and realistic gas turbine conditions, ASME Turbo Expo 2017: Turbomachinery Technical Conference and Exposition (2017). paper V04AT04A037. [29] P. Versailles, W.A. Chishty, H.D. Vo, Application of dielectric barrier discharge to improve the flashback limit of a lean premixed dump combustor, J. Eng. Gas Turb. Power 134 (2012) 031501. [30] G.T. Kim, B.H. Seo, W.J. Lee, J. Park, M.K. Kim, S.M. Lee, Effects of applying non-thermal plasma on combustion stability and emissions of NOx and CO in a model gas turbine combustor, Fuel 194 (2017) 321–328. [31] Y.P. Raizer, Gas discharge physics, Barcelona: Springer, 1991. [32] Y. Ju, W. Sun, Plasma assisted combustion: Progress, challenges, and opportunities, Combust. Flame 162 (2015a) 529–532. [33] Y. Ju, W. Sun, Plasma assisted combustion: Dynamics and chemistry, Prog. Energy Combust. Sci. 48 (2015b) 21–83. [34] W. Siemens, Ueber die elektrostatische induction und die verzȵgerung des stroms in flaschendrhten, Ann. Phys. Chem. 102 (1857) 66–122. [35] B. Eliasson, M. Hirth, U. Kogelschatz, Ozone synthesis from oxygen in dielectric barrier discharges, J. Phys. D: Appl. Phys 20 (1987) 1421–1431. [36] M.B. Chang, S.J. Wu, Experimental study on ozone synthesis via dielectric barrier discharges, Ozone Sci. Eng. 19 (1997) 241–254. [37] Z.H. Wang, B.L. L. Yang, Z.S. Li, Z.W. Sun, M. Alden, K.F. Cen, A.A. Konnov, Investigation of combustion enhancement by ozone additive in CH4 /air flames using direct laminar burning velocity measurements and kinetic simulations, Combust. Flame 159 (2012) 120–129. [38] F. Halter, P. Higelin, P. Dagaut, Experimental and detailed kinetic modeling study of the effect of ozone on the combustion of methane, Energy Fuels 25 (2011) 2909–2916. [39] X. Gao, Y. Zhang, S. Adusumilli, J. Seitzman, W. Sun, The effect of ozone addition on laminar flame speed, Combust. Flame 162 (2015) 3914–3924. [40] J. Choe, W. Sun, Blowoff hysteresis, flame morphology and the effect of plasma in a swirling flow, J. Phys. D Appl. Phys. 51 (2018) 365201. [41] J.M. Beer, N. Chigier, Combustion aerodynamics, Krieger and Malabar, 1983. [42] A.K. Gupta, D.G. Lilley, N. Syred, Swirl flows, Abacus Press, Tunbridge Wells, Kent, 1984. [43] Y. Huang, V. Yang, Dynamics and stability of lean-premixed swirl-stabilized combustion, Prog. Energy Combust. Sci. 35 (2009) 293–394. [44] X. Lu, S. Wang, H. Sung, S. Hsieh, V. Yang, Large-eddy simulations of turbulent swirling flows injected into a dump chamber, J. Fluid Mech. 527 (2005) 171–195. [45] S. Wang, S.Y. Hsieh, V. Yang, Unsteady flow evolution in swirl injector with radial entry. I. Stationary conditions, Phys. Fluids 17 (2005) 045106. [46] B. Dong, J.M. Bauchire, J.M. Pouvesle, P. Magnier, D. Hong, Experimental study of a DBD surface discharge for the active control of subsonic airflow, J. Phys D: Appl. Phys 41 (2008) 155201.

[47] N. Benard, E. Moreau, Electrical and mechanical characteristics of surface AC dielectric barrier discharge plasma actuators applied to airflow control, Exp. Fluids 55 (2014) 1846. [48] M.G. Kong, X.T. Deng, Electrically efficient production of a diffuse nonthermal atmospheric plasma, IEEE Tran. Plasma Sci. 31 (2003) 7–18. [49] T. Ombrello, S.H. Won, Y. Ju, S. Williams, Flame propagation enhancement by plasma excitation of oxygen. Part II: Effects of O2 (a1 g ), Combust. Flame 157 (2010) 1916–1928. [50] J. Malicet, D. Daumont, J. Charbonner, C. Parisse, A. Chakir, J. Brion, Ozone UV spectroscopy. II. Absorption cross-sections and temperature dependence, Atmos. Chem. 21 (1995) 263–273. [51] D.G. Imre, J.L. Kinsey, R.W. Fleld, Spectroscoplc characterization of repulsive potential energy surfaces: Fluorescence spectrum of ozone, J. Phys. Chem. 86 (1982) 2564–2566. [52] Y. Kim, S.H. Won, M.S. Cha, S.J. Kim, Measurements of electron energy by emission spectroscopy in pulsed corona and dielectric barrier discharges, J. Adv. Oxid. Technol. 6 (2003) 17–22. [53] J. Vinogradov, E. Sher, I. Rutkevich, M. Mond, Voltage-current characteristics of a flame-assisted unipolar corona, Combust. Flame 127 (2001) 2041–2050. [54] B. Eliasson, U. Kogelschatz, Nonequilibrium volume plasma chemical processing, IEEE Trans. Plasma Sci. 19 (1991) 1037–1077. [55] A.B. Fialkov, Investigations on ions in flames, Prog. Energy Combust. Sci. 23 (1997) 399–528. [56] J.S. Charng, P.A. Lawless, T. Yamamoto, Corona discharge processes, IEEE Trans. Plasma Sci. 19 (1991) 1152–1165. [57] T.C. Manley, The electric characteristics of the ozonator discharge, Trans. Electrochem. Soc. 84 (1943) 83–96. [58] R. Feng, G.S.P. Castle, S. Jayaram, Automated system for power measurement in the silent discharge, IEEE Trans. Ind. Appl. 34 (1998) 563–570. [59] S. Taamallah, Z.A. LaBry, S.J. Shanbhogue, A.F. Ghoniem, Thermo-acoustic instabilities in lean premixed swirl-stabilized combustion and their link to acoustically coupled and decoupled flame macrostructures, Proc. Combust. Inst. 35 (2015) 3273–3282. [60] S. Taamallah, S.J. Shanbhogue, A.F. Ghoniem, Turbulent flame stabilization modes in premixed swirl combustion: physical mechanism and karlovitz number-based criterion, Combust. Flame 166 (2016) 19–33. [61] F.H. Champagne, S. Kromat, Experiments on the formation of a recirvulation zone in swirling coaxial jets, Exp. Fluids 29 (20 0 0) 494–504. [62] A.F. Najafi, S.M. Mousavian, K. Amini, Numerical investigations on swirl intensity decay rate for turbulent swirling flow in a fixed pipe, Int. J. Mech. Sci. 53 (2011) 801–811. [63] J. Lawton, F. Weinberg, Electrical aspects of combustion, Oxford : Clarendon Press, 1969. [64] J.M. Goodings, D.K. Bohme, W.N. Chun, Detailed ion chemistry in methane-oxygen flames, Combust. Flame 35 (1979) 27–43. [65] T.M. Vu, S.H. Won, T. Ombrello, M.S. Cha, Stability enhancement of ozone-assisted laminar premixed Bunsen flames in nitrogen co-flow, Combust. Flame 161 (2014) 917–926. [66] X. Zhang, B.J. Lee, H.G. Im, M.S. Cha, Ozone production with dielectric barrier discharge: Effects of power source and humidity, IEEE Trans. Plasma Sci. 44 (2016) 2288–2296. [67] N. Merican-Bourdet, M.J. Kirkpatrick, F. Tuvache, D. Frochot, E. Odic, Effect of voltage waveform on dielectric barrier discharge ozone production efficiency, Eur. Phys. J. Appl. Phys. 57 (2012) 30801. [68] T. Ombrello, S.H. Won, Y. Ju, S. Williams, Flame propagation enhancement by plasma excitation of oxygen. Part I: Effects of O3 , Combust. Flame 157 (2010) 1906–1915. [69] T. Ombrello, Y. Ju, A. Fridman, Kinetic ignition enhancement of diffusion flames by nonequilibrium magnetic gliding arc plasma, AIAA J. 4 (2008) 2424–2433.