Ignition of a sequential combustor: Evidence of flame propagation in the autoignitable mixture

ARTICLE IN PRESS

JID: PROCI

[m;July 28, 2018;18:50]

Available online at www.sciencedirect.com

Proceedings of the Combustion Institute 000 (2018) 1–8 www.elsevier.com/locate/proci

Ignition of a sequential combustor: Evidence of flame propagation in the autoignitable mixture D. Ebi a,∗, U. Doll a, O. Schulz b, Y. Xiong b, N. Noiray b a Laboratory b Department

for Thermal Processes and Combustion, Paul Scherrer Institute, Villigen 5232, Switzerland of Mechanical and Process Engineering, CAPS Laboratory, ETH Zurich, Zurich 8092, Switzerland Received 30 November 2017; accepted 7 June 2018 Available online xxx

Abstract Ignition of the second stage in a lab-scale sequential combustor is investigated experimentally. A fuel mixing section between jet-in-cross-flow injection and the second stage chamber allows the fuel and vitiated, hot cross-flow to partially mix upstream of the main heat release zone. The focus of the present work is on the transient ignition process leading to a stable flame in the second stage. High-speed OH-PLIF as well as OH chemiluminescence imaging is applied to obtain complementary planar and line-of-sight integrated information on the ignition. We find experimental evidence for the co-existence of two regimes dominating the chamber ignition, i.e. autoignition and flame propagation. As the mass flow of the dilution air injected downstream of the first stage is increased (i.e. mixing temperatures in the fuel mixing section are decreased), we transition from an autoignition to a flame propagation dominated regime. Hysteresis in the ignition behavior is observed indicating that the first stage in a sequential combustor may be operated at leaner conditions than required for ignition of the second stage. The time traces of integral heat release obtained simultaneously with a photomultiplier tube show distinct features depending on the dominating regime, which is important for high-pressure testing with limited optical access. © 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved. Keywords: Autoignition; Turbulent flame propagation; Sequential combustor; Jet-in-cross-flow; High-speed PLIF

1. Introduction Sequential or staged combustion systems are developed to address the demand for increased operational and fuel flexibility while at the same time maintaining high efficiency and low emissions [1,2].

∗

Corresponding author. E-mail addresses: [email protected] (D. Ebi), [email protected] (N. Noiray).

The second stage combustor is operated by feeding fuel into the vitiated, hot gas stream supplied by a conventional first stage burner. The fuel jet may enter the vitiated hot gas stream in a cross-flow or co-flow configuration. The main focus in previous studies has been on conditions where the flame is anchored in the immediate vicinity of the fuel injection [3–11]. The current work instead focuses on the transient ignition process of the second stage in a practical sequential combustor geometry. Ignition

https://doi.org/10.1016/j.proci.2018.06.068 1540-7489 © 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Please cite this article as: D. Ebi et al., Ignition of a sequential combustor: Evidence of flame propagation in the autoignitable mixture, Proceedings of the Combustion Institute (2018), https://doi.org/10.1016/j.proci.2018.06.068

JID: PROCI

2

ARTICLE IN PRESS

[m;July 28, 2018;18:50]

D. Ebi et al. / Proceedings of the Combustion Institute 000 (2018) 1–8

The distance from the fuel injection to the inlet of the 2nd stage combustion chamber (cross-section: 88 × 88 mm2 ) was 220 mm. An orifice plate was installed at the combustor outlet (orifice diameter: 35 mm), which was sufficient to prevent thermoacoustic instabilities upon ignition of the second stage. 2.2. Diagnostic setup

Fig. 1. Sketch of the sequential combustion system with an enlarged view of the 2nd stage combustor. The black box marks the camera field-of-view. Dashed lines indicate the area illuminated by the UV light sheet, dash-dotted lines outline the flow channel walls.

refers to the entire ignition process starting (in this case) with autoignition kernels and ending with a stabilized flame in the second stage chamber [12]. The combustor design is such that the fuel injection site is spatially separated from the main heat release zone, i.e. fuel and vitiated cross-flow are allowed to at least partially mix prior to the main reaction zone. Therefore, the second stage combustion process cannot be attributed to a single regime but instead premixed or partially premixed propagation, non-premixed combustion and autoignition may contribute to the ignition process and flame stabilization [13]. The objective of the present work is to investigate experimentally, with the support of reactor simulations, the combustion regimes dominating the transient ignition process for a range of operating conditions. 2. Experimental setup 2.1. Test rig A schematic view of the atmospheric pressure, sequential two stage combustion system is depicted in Fig. 1. The first stage consists of a swirl burner with a single axial swirler and a cylindrical center body [14,15]. It was operated with natural gas (NG) in a fully premixed mode for the present investigation. Downstream of the first stage, dilution air was injected into the vitiated cross-flow through a port in the top wall. The fuel for the second stage (again NG) was supplied with a jet-in-cross-flow at the top wall immediately upstream of two vortex generators of tetrahedral shape, which are located inside the fuel mixing section. The cross-sectional area of the fuel mixing section was 40 × 40 mm2 .

The reacting flow was characterized by the successive acquisition of the line-of-sight integrated OH∗ chemiluminescence signal as well as planar laser induced fluorescence (PLIF) of the OH radical in the vertical center plane of the 2nd stage combustor. The detection system consisted of a highspeed CMOS camera (LaVison HSS X) and a lens coupled high-speed intensifier (LaVision HS-IRO), equipped with a 100 mm f/2.8 UV lens (Cerco) and a bandpass interference filter (Chroma, transmission > 70 % at 310 nm, FWHM 10 nm). PLIF images were acquired at a repetition rate of 2 kHz with intensifier gate width and gain set to 150 ns and 60 to 70%, respectively. For OH∗ imaging, the repetition rate was set to 5 kHz with an extended intensifier gate width of 100 μs in order to reach sufficient signal strengths. The PLIF laser system was composed of a DPSS pump source (Edgewave IS400-2-L) followed by a dye laser (Radiant Dyes NarrowScan HighRep) and an external frequency-doubling unit. The frequency-doubled dye laser output was tuned to excite the Q1 (9) and Q2 (8) lines of the A-X (ν  = 1, ν  = 0) band near 284 nm. At 2 kHz repetition rate, UV energies of 0.5 ± 0.05 mJ per pulse were achieved. The laser beam was expanded into a divergent sheet and entered the combustor through the exhaust duct. A mirror protected by an air-cooled window was placed in the hot exhaust gas stream for that purpose. The sheet thickness was approximately 0.5 mm within the field-of-view of the camera. While the lower half of the 2nd stage combustor was almost completely covered by the UV light sheet, laser light in the upper half was partly blocked due to geometrical restrictions. Absorption of laser energy along the exhaust gas duct was sufficiently low such that flame structures could be visualized along the entire width of the field-ofview (Fig. 4 c). The OH-PLIF images were corrected for the mean laser intensity profile. However, beam steering effects due to refractive index variations induced by the highly turbulent hot exhaust gas remained clearly visible in the PLIF images but were acceptable for qualitative imaging. All PLIF images have been filtered with a sliding median filter (3 × 3 pixel neighborhood). In addition, the integral heat release of the second stage was monitored at 50 kHz repetition rate by means of a Hamamatsu H10721-110 photomultiplier tube (PMT) equipped with a bandpass

Please cite this article as: D. Ebi et al., Ignition of a sequential combustor: Evidence of flame propagation in the autoignitable mixture, Proceedings of the Combustion Institute (2018), https://doi.org/10.1016/j.proci.2018.06.068

ARTICLE IN PRESS

JID: PROCI

[m;July 28, 2018;18:50]

D. Ebi et al. / Proceedings of the Combustion Institute 000 (2018) 1–8

3

Table 1 Summary of operating conditions: 1st stage conditions and 2st stage fuel mass flow were constant for all OPs. The OPs are labeled by amount of dilution air (DA), e.g. DA5 ˙ DA = 5 g/s. corresponds to m 1st stage ˙ NG,1 = 0.6 g/s, m ˙ MA = 15.3 g/s, φ = 0.67 m 2nd stage OP

˙ NG,2 m

˙ DA m

φ

TTC

DA5 DA7 DA8.5 DA10

0.6 g/s 0.6 g/s 0.6 g/s 0.6 g/s

5.0 g/s 7.0 g/s 8.5 g/s 10.0 g/s

0.98 0.81 0.72 0.65

1278 K 1223 K 1172 K 1129 K

interference filter (bk Interferenzoptik, maximum transmission ∼ 40 % at 313 nm, FWHM 7 nm). The fuel mass flow to the 2nd stage combustor was recorded simultaneously. Acoustic signals were acquired with four microphones (Kistler Type 211B5) at various axial locations and monitored to ensure quiet operating conditions for the current investigation. A shielded K-type thermocouple was installed immediately upstream of the 2nd stage fuel injection port to estimate the hot gas temperature in the mixing section.

Fig. 2. Autoignition times τ AI as a function of mixture fraction Z for four mixture compositions representative of the investigated operating conditions. Squares indicate the stoichiometric mixture fraction Zst . The equivalence ratios corresponding to the most reactive  mixture fractions ZMR (minimum τ AI ) are φ = ( ) 0.06, (◦) 0.04, (♦) 0.03, (∗) 0.02.

and the cold fuel jet. The composition of the fuel stream was defined as YCH4 =0.865, YC2 H6 =0.074, YC3 H8 =0.017, YCO2 =0.030, YN2 =0.014, which represents the natural gas used for the current investigation (higher hydrocarbons with mass fractions Y < 0.4% were neglected). The corresponding stoichiometric fuel-to-oxidizer ratio, which was used to compute the global φ listed in Table 1, is 0.252.

3. Results 3.1. Operating conditions

3.2. Stationary operation: Modes of flame stabilization

The ignition of the 2nd stage combustor has been investigated at the four operating conditions (OPs) listed in Table 1. The first stage was operated with ˙ NG,1 and a fixed main a fixed fuel mass flow rate m ˙ MA corresponding to an equivalence ratio air flow m of φ = 0.67 and a thermal power of approximately 28 kW. The second stage was operated with a fixed ˙ NG,2 but with a range of dilution fuel mass flow m air (DA) mass flow rates. An increase in DA corresponds to a decrease in global equivalence ratio and a decrease in gas temperature entering the fuel mixing section. The measured thermocouple temperatures TTC given in Table 1 provide an estimate for the hot gas temperature prior to injection of the 2nd stage fuel stream. A radiation correction was applied, which suggests that these readings underestimate the gas temperature by ∼ 50 K [16]. Farther downstream, at the exit of the fuel mixing section, the bulk flow velocities based on the total mass flows and the estimated mean temperatures range from about 43 m/s to 46 m/s for the 5 g/s and 10 g/s dilution air case, respectively. Autoignition times τ AI for the four operating conditions were obtained from 0-D reactor simulations with Cantera using the UCSD chemical kinetic mechanism [17]. The results are shown in Fig. 2. Z describes the mixing between the vitiated hot gases (Tin ) entering the mixing section

Before discussing the transient ignition process, the flame shapes and modes of flame stabilization associated with a stationary operation of the combustor are presented. The mean flame shapes (based on 2000 images) are shown in Fig. 3 for each of the four operating conditions. The top row shows time-averaged OH∗ chemiluminescence images in the field-of-view indicated in Fig. 1. The height of the field-of-view for the PLIF imaging (bottom row) was slightly smaller due to constraints on the laser sheet width. The mean OH∗ image for a dilution air mass flow of 5 g/s (case DA5) shows that the majority of the heat release occurs along the core of the combustion chamber. In particular, heat release occurs along the full length of the field-of-view, which is also seen in the corresponding instantaneous OH∗ image in Fig. 4 a. The OH-PLIF imaging reveals that this heat release distribution results from a continuous stream of autoignition kernels, which emanate from the mixing section. A representative single shot OH-PLIF image is shown in Fig. 4 c. Individual autoignition kernels are not seen in the OH∗ images (Fig. 4 a) at this operating condition, since discrete kernels at various depth locations overlap in the line-of-sight integrated OH∗ images. The autoignition kernels emanating from the mixing section, which themselves are on the order

Please cite this article as: D. Ebi et al., Ignition of a sequential combustor: Evidence of flame propagation in the autoignitable mixture, Proceedings of the Combustion Institute (2018), https://doi.org/10.1016/j.proci.2018.06.068

JID: PROCI

4

ARTICLE IN PRESS

[m;July 28, 2018;18:50]

D. Ebi et al. / Proceedings of the Combustion Institute 000 (2018) 1–8

Fig. 3. Time-averaged OH∗ images (line-of-sight integrated, top row) and OH-PLIF images (center plane, bottom row) during stationary operation (flow from left to right). Grey solid lines indicate the exit of the mixing section duct at x = 0 and the outline of the 2nd stage combustor walls.

Fig. 4. Instantaneous OH∗ (top row) and OH-PLIF images (bottom row, not imaged simultaneously) for cases DA5 and DA10.

of the mixing channel height in terms of size, generally feature small-scale structures with sharp fronts, i.e. strong gradients in the OH-PLIF signal, which is characteristic of established flame fronts following autoignition events [8,18]. Depending on the trajectory and radial spread of autoignition kernels entering the 2nd stage chamber, the kernels may come in contact with unburnt gas near the walls and ignite it as seen along the bottom wall in Fig. 4 c. At this operating condition, Reh ≈ 8000 based on the step height h. The length of the recirculation zone (reattachment length xR ) downstream of the backward-facing step scales as xR /h ∼ 6, which corresponds to xR ≈ 150 mm [19]. The gas near the bottom wall in Fig. 4 c, which ignites upon arrival of autoignition kernels, may then be attributed to being inside the recirculation zone downstream of the backward-facing step in agreement with LES results in the same combustor at a similar operating condition [20]. No reaction is observed upstream of the location where kernels first come in contact with the recirculation zone. In particular, no OH-PLIF sig-

nal is observed in the recirculation zones near the backward-facing step at the chamber inlet at this operating condition. The envelope of the mean PLIF signal intensity for the DA5 case (Fig. 3) supports this finding as the mean PLIF signal does not bend upstream near the wall. Furthermore, based on the full movie sequences, a temporal gap in the supply of kernels from the fuel mixing section correlates with a downstream movement of the flame front in the 2nd stage. Autoignition in the recirculation zones (independent of the arrival of kernels emanating from the mixing section) therefore does not dominate the stabilization of the 2nd stage flame at this operating condition, albeit a certain contribution cannot be excluded. As the dilution air is increased to 7 g/s (case DA7), autoignition kernels still form regularly but less frequently in the mixing section, which leads to a decrease in mean OH∗ signal intensity in the core flow near the chamber inlet (Fig. 3). A further increase in dilution air to 8.5 g/s shifts the heat release to the second half of the field-of-view. At 10 g/s, kernel formation in the mixing section is a rare event, which will be discussed further in Section 3.3. Instead, the OH-PLIF image shows an inverse V-shaped flame, which is anchored near the chamber walls. Regions with non-zero OH-PLIF signal near the wall are frequently observed farther upstream compared to the main heat release region indicated by the OH∗ images similar to the findings by Sidey et al. [8] in a different burner configuration. The V-shape associated with the DA10 case flame is not seen in the chemiluminescence images during stationary operation (Fig. 4 b), which can be explained by the wedge-shaped part of the flame not just forming along the bottom and top wall but also at the front and back wall of the combustion chamber (line-of-sight imaging). In contrast, the Vshaped flame front and absence of autoignition kernels is clearly seen in instantaneous OH-PLIF images, e.g. in Fig. 4 d.

Please cite this article as: D. Ebi et al., Ignition of a sequential combustor: Evidence of flame propagation in the autoignitable mixture, Proceedings of the Combustion Institute (2018), https://doi.org/10.1016/j.proci.2018.06.068

JID: PROCI

ARTICLE IN PRESS D. Ebi et al. / Proceedings of the Combustion Institute 000 (2018) 1–8

Fig. 5. Time traces of 2nd stage fuel supply and overall heat release (based on PMT signal) of five ignition attempts for cases DA5 (top) and DA10 (bottom). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

3.3. Ignition sequences For the investigation of the transient sequential stage ignition process, the 2nd stage fuel valve was opened by directly controlling the valve opening percentage with a ramp rate that was held constant for all operating conditions. The final valve opening corresponded to the specified fuel mass ˙ NG,2 = 0.6 g/s, which was reached within flow of m ∼ 500 ms. For each fuel ramp, the 2nd stage fuel mass flow rate as well as the PMT signal as a measure of integrated heat release in the 2nd stage combustor were monitored. Imaging data was obtained for a total of 40 ignition sequences split among the four operating points. For brevity, only the DA5 and DA10 cases are discussed in this section. 3.3.1. Combustor ignition delay Figure 5 shows time traces of fuel supply (red) and PMT signal (blue; filtered signal shown in black) for five ignition attempts (cycles) for cases DA5 and DA10. The fuel mass flow was modulated such that during each cycle fuel was supplied for 10 s followed by a period of 12 s without fuel injection. The PMT signal quantifies if ignition was successful and when it occurred with respect to the opening of the fuel valve. An enlarged view of the first cycle is shown in the yellow box for each case. For case DA5, heat release was detected shortly (milliseconds) after the fuel valve began to open, i.e. already with only a small amount of fuel released into the vitiated cross-flow. In the MILD combustion regime, two competing effects exist, which cause autoignition not to occur near the stoichiometric mixture fraction but for very lean conditions (Fig. 2) [18]: When a cold fuel stream is released into a hot oxidizer stream, the fuel concentration increases locally (Z increases)

[m;July 28, 2018;18:50]

5

Fig. 6. Examples of transient time traces after fuel valve opening showing the acoustic pressure in the sequential chamber and the PMT signals for operating conditions a) DA5 and b) DA10. The power spectral densities of signals recorded under stationary condition after ignition are shown in the insets.

but the temperature decreases compared to the initial temperature of the hot oxidizer stream. The shortest autoignition times occur when the fuel concentration has increased sufficiently but the temperature is still high. Favorable conditions for autoignition to occur with a small τ AI at this operating point therefore already exist before the fuel valve is fully open, which allows a smooth ignition. In contrast, for case DA10, a seemingly random delay, typically on the order of several seconds, was found between opening the fuel valve and heat release. During the second cycle of the data set presented in Fig. 5, ignition did not occur at all during the 10 s window of fuel injection. If ignition occurred, it was in general harsh compared to the DA5 case, i.e. associated with a spike in heat release and chamber pressure as shown in Fig. 6. The pressure spikes reached amplitudes of ∼ 8 kPa, which was more then tenfold the r.m.s. acoustic pressure level associated with the stable flame after ignition. In contrast, for case DA5 for example, the acoustic level rises smoothly. For both conditions, after ignition, one can see relatively broad peaks in the power spectral densities, which correspond to the thermoacoustic modes of the sequential combustor. As mentioned earlier, these modes are linearly stable thanks to the use of an orifice plate at the exit of the combustor. 3.3.2. Autoignition dominated ignition regime The ignition process for the DA5 operating condition can be divided into three phases. The initial phase is characterized by the formation of

Please cite this article as: D. Ebi et al., Ignition of a sequential combustor: Evidence of flame propagation in the autoignitable mixture, Proceedings of the Combustion Institute (2018), https://doi.org/10.1016/j.proci.2018.06.068

JID: PROCI

6

ARTICLE IN PRESS

[m;July 28, 2018;18:50]

D. Ebi et al. / Proceedings of the Combustion Institute 000 (2018) 1–8 ∗ ratio τAI /τcomb. ≈ 0.4 < 1 then suggests that pockets characterized by ZMR form well within the mixing section. The autoignition time simulation therefore agrees with the visual observation that the majority of autoignition kernels form approximately half way between fuel injection and inlet to the 2nd stage chamber. The 0-D reactor simulation further suggests that even pockets characterized by a mixture fraction near Zst autoignite before reaching the 2nd stage combustion chamber outlet in agreement with the OH-PLIF images (Fig. 4 c). The second phase of the ignition process is characterized by the formation of a continuous stream of autoignition kernels as depicted at time instants 132.5 ms through 133.5 ms in Fig. 7 (left column). During this phase, the size of the ignition kernels remains small compared to the chamber height. In a final phase, the kernels entering the 2nd stage chamber are larger in size ( ∼ mixing channel height). In addition, the kernels spread rapidly as they enter the combustion chamber thus causing igniting across the entire chamber height as shown at t = 239.5 ms. An image representative of a fully ignited second stage was already shown in Fig. 4 c. The rapid spreading and larger size of the kernels may be attributed to having reached a global equivalence ratio at which flame propagation with a significant flame displacement speed is possible.

Fig. 7. Left column: OH-PLIF image sequence of an ignition process dominated by autoignition (operating condition DA5). Same field-of-view as in Fig. 4 c. Right column: OH∗ image sequence of an ignition process dominated by propagation (operating condition DA10). Same field-of-view as in Fig. 4 a. Time instants are referenced to onset of fuel valve opening.

individual, small (∼half height of mixing section channel) autoignition kernels in the mixing section, which are convected through the 2nd stage combustion chamber without significant radial spread. The first three successive images shown in Fig. 7 (left column) are representative of this initial phase. The time-stamp is referenced to the start of the valve opening. ∗ The autoignition time τAI = τAI (Z = ZMR ) is about 2 ms at this operating condition (Fig. 2). The core flow velocity in the mixing section may be estimated from the speed at which kernels emanate from it. The kernel speed is evaluated by crosscorrelating successive PLIF images near the inlet of the combustion chamber such as the first three images in Fig. 7. Processing several thousand images yields a median kernel velocity of ∼ 48 m/s, which suggests a core flow velocity of similar magnitude. The relevant combustor length scale is the distance between fuel injection port and combustion chamber inlet (220 mm), which yields a combustor timescale of τ comb. ≈ 4.5 ms. The timescale

3.3.3. Propagation dominated ignition regime ∗ As the dilution air mass flow is increased, τAI in∗ ˙ DA = 7 g/s, the ratio τAI creases. At m /τcomb. ≈ 0.9 is in agreement with the time-averaged OH-LIF and OH∗ images in Fig. 3, which show that most of the kernels still form inside the mixing section. At ˙ DA = 8.5 g/s, the reactor calculation predicts that m ∗ τAI /τcomb. ≈ 1.5, which is supported by the experimental finding that fewer kernels form in the fuel mixing section and the mean flame shape transitions to an inverse V-flame (Fig. 3). ˙ DA = 10 g/s, the ignition process difFinally, at m fers fundamentally from the autoignition stabilized flame at 5 g/s. Comparing case DA10 to case DA5, ∗ τAI is larger by a factor of about 10 despite a mere decrease of about 10 % in inlet temperature owing to the exponential dependence of τ AI on Tin [12]. The autoignition time obtained from the perfectly stirred reactor calculation exceeds the combustor ∗ time scale, i.e. τAI /τcomb. ≈ 4.5 > 1. However, in reality, we do not have perfectly stirred conditions. Hot spots due to inhomogeneous mixing between 1st stage hot gases and cold dilution air likely exist. Conditions for the formation of autoignition kernels are favorable near ZMR and where scalar dissipation rates are low [18,21]. If such conditions coincide with hot spots and high residence times, e.g. in the wake of the vortex generators, formation of autoignition kernels may still occur at rare instants in time. For the DA10 case, such events are observed every few seconds. The combination of favorable

Please cite this article as: D. Ebi et al., Ignition of a sequential combustor: Evidence of flame propagation in the autoignitable mixture, Proceedings of the Combustion Institute (2018), https://doi.org/10.1016/j.proci.2018.06.068

JID: PROCI

ARTICLE IN PRESS D. Ebi et al. / Proceedings of the Combustion Institute 000 (2018) 1–8

conditions required explains the seemingly random delay between opening of the fuel valve and ignition as discussed previously (Fig. 5). In all ignition data sets obtained for operating condition DA10, a single small kernel such as the one shown in Fig. 7 (right column) at t = 2654.8 ms through 2656.0 ms was observed prior to chamber ignition (OH∗ image sequence is shown here since the single kernel responsible for ignition was reliably captured only with line-of-sight imaging). Owing to the time delay between fuel injection and formation of a single autoignition kernel, the chamber has already been filled with a combustible mixture at the global equivalence ratio of φ = 0.65. Once a kernel arrives, ignition occurs and a flame front spreads radially (t = 2657.6 ms). Subsequently, the flame propagates along the top and bottom wall as shown at time instants t = 2661.6 ms through t = 2662.8 ms. Therefore, flame stabilization, which is the final step of a successful combustor ignition, is achieved by flame propagation. Following the upstream propagation, an inverse V-shaped flame is established as discussed in Section 3.2. The flame shape and propagation shows similarities to channel boundary layer (BL) flashback of hydrogen-air flames at moderate inlet temperatures ( ∼ 300 to 600 K) [22,23]. Typically, 1st stage burners operated on natural gas with bulk flow velocities of ∼ 50 m/s are not prone to BL flashback. However, flames in the 2nd stage chamber are subject to high inlet temperatures. The laminar flame speed for the DA10 conditions predicted with Cantera is about 3 m/s, which is larger than that of a stoichiometric H2 -air flame at 300 K inlet temperature. In addition, critical gradients, below which BL flashback occurs, are particularly sensitive to the wall temperature [24]. High wall temperatures, present in a 2nd stage chamber even prior to ignition, decrease the quenching distance and thus promote flashback. We are hence in a regime where near-wall upstream flame propagation of natural gas flames is likely to occur in agreement with the experimental observations. The sudden ignition of the previously accumulated combustible mixture explains the spikes in heat release observed in the PMT time traces (Fig. 5) and the harsh ignition compared to the DA5 operating condition. 3.4. Ignition hysteresis The ignition of the 2nd stage flame shows hysteresis as a function of the dilution air mass flow rate. As indicated by the black dots (path 1) in ˙ DA ≥ 10.5 g/s. Fig. 8, ignition failed at all times for m Ignition was mostly successful at 10 g/s and always succesful for ≤ 8.5 g/s as discussed in Section 3.3.1. In contrast, if instead we start with a stable flame ˙ DA , i.e. decrease the mixing section and increase m gas temperature, the stable flame is supported be˙ DA = 12.5 g/s. This yond the ignition limit up to m

[m;July 28, 2018;18:50]

7

Fig. 8. Hysteresis in combustor ignition for different dilution air flow rates. Path 1 (dots): Ignition is attempted starting at high dilution air flow rates. Path 2 (triangles): Dilution air is increased starting with a stable flame. Nonzero PMT signals indicate flame in 2nd stage.

finding provides further evidence that flame stabi˙ DA does not rely on autoignition lization at high m events in the mixing section. With regards to operating a real gas turbine, the finding suggests operating the 1st stage at leaner conditions below what is required for ignition of the second stage is possible. 4. Conclusion Ignition and modes of flame stabilization of the second stage flame in a sequential combustor were investigated experimentally in a lab-scale test rig. High-speed OH-PLIF and OH∗ imaging was applied to capture both line-of-sight and planar information on the transient phenomena. The findings provide experimental evidence for the co-existence of two regimes, i.e. autoignition and flame propagation, in a technical system within a narrow range of inlet temperatures. The temperature was varied by a change in dilution air flow rate. The ignition process differed fundamentally depending on which regime dominates. In the autoignition regime, where τ AI /τ comb. < 1, a continuous stream of large autoignition kernels emanating from the mixing section ignites and stabilizes the 2nd stage flame. Since kernels already form as the fuel valve is opening, the ignition is smooth and well-behaved, characteristics which are important for the operation of a gas turbine. In the propagation regime, a single small autoignition kernel forms at a random delay (typically within few seconds) relative to the opening of the fuel valve despite τ AI /τ comb. > 1. By that time, a combustible fuel-air mixture has already accumulated in the second stage, which is ignited rather far downstream of the chamber inlet once the kernel arrives. Chamber ignition, i.e. flame stabilization, is then enabled by flame propagation along the chamber walls. This latter mode of ignition is accompanied with a spike in heat release and chamber pressure, i.e. it is rather harsh and less desirable for the operation of a real combustor. The findings in this work suggest that the dominating regime may be identified based on a PMT

Please cite this article as: D. Ebi et al., Ignition of a sequential combustor: Evidence of flame propagation in the autoignitable mixture, Proceedings of the Combustion Institute (2018), https://doi.org/10.1016/j.proci.2018.06.068

ARTICLE IN PRESS

JID: PROCI

8

[m;July 28, 2018;18:50]

D. Ebi et al. / Proceedings of the Combustion Institute 000 (2018) 1–8

signal only, which is important for high-pressure testing with limited optical access.

Acknowledgments This research is supported by the Swiss National Science Foundation under Grant 160579. We further thank R. Bombach for the support with the laser system.

Supplementary material Supplementary material associated with this article can be found, in the online version, at doi:10. 1016/j.proci.2018.06.068 References [1] D.A. Pennell, M.R. Bothien, A. Ciani, et al., Combust. Fuels Emiss.. 4B. V04BT04A043, doi:10.1115/ GT2017-64790. [2] H. Karim, J. Natarajan, V. Narra, et al., Combust. Fuels Emiss.. 4A. V04AT04A062, doi:10.1115/ GT2017-63998. [3] P. Domingo, L. Vervisch, D. Veynante, Combust. Flame 152 (3) (2008) 415–432, doi:10.1016/j. combustflame.2007.09.002. [4] W. Meier, I. Boxx, C.M. Arndt, M. Gamba, N.T. Clemens, J. Eng. Gas Turbines Power 133 (2) (2011). 021504, doi: 10.1115/1.4002014. http://link.aip.org/link/JETPEZ/v133/i2/p021504/ s1&Agg=doi. [5] J.M. Fleck, P. Griebel, A.M. Steinberg, C.M. Arndt, C. Naumann, M. Aigner, Proc. Combust. Inst. 34 (2) (2013) 3185–3192, doi:10.1016/j.proci.2012.05.039. [6] R. Sullivan, B. Wilde, D.R. Noble, J.M. Seitzman, T.C. Lieuwen, Combust. Flame 161 (7) (2014) 1792– 1803, doi:10.1016/j.combustflame.2013.12.022. [7] S. Lyra, B. Wilde, H. Kolla, J.M. Seitzman, T.C. Lieuwen, J.H. Chen, Combust. Flame 162 (4) (2015) 1234–1248, doi:10.1016/j.combustflame.2014. 10.014.

[8] J. Sidey, E. Mastorakos, 2015, Proc. Combust. Inst., 35, 3, 3537–3545, doi:10.1016/j.proci.2014.07.028. [9] D. Ahrens, M. Kolb, C. Hirsch, T. Sattelmayer, J. Eng. Gas Turbines Power 138 (8) (2016). 081506 doi: 10.1115/1.4032420 [10] P.P. Panda, M. Roa, C.D. Slabaugh, et al., Combust. Flame 163 (2016) 241–257, doi:10.1016/j. combustflame.2015.10.001. [11] J.A. Wagner, M.W. Renfro, B.M. Cetegen, Combust. Flame 176 (2017) 521–533, doi:10.1016/j. combustflame.2016.11.014. [12] T. Lieuwen, Unsteady Combustor Physics, Cambridge University Press, 2012. [13] O. Schulz, T. Jaravel, T. Poinsot, B. Cuenot, N. Noiray, 2017, Proc. Combust. Inst., 36, 2, 16371644, 10.1016/j.proci.2016.08.022. doi:10.1016/ j.proci.2016.08.022. [14] O. Schulz, U. Doll, D. Ebi, J. Droujko, C. Bourquard, N. Noiray, Proc. Combust. Inst., doi:10.1016/j.proci. 2018.07.089. [15] G. Bonciolini, D. Ebi, U. Doll, M. Weilenmann, N. Noiray, Proc. Combust. Inst., doi:10.1016/j.proci. 2018.06.229. [16] G.E. Glawe, R. Holanda, L. N. Krause, NASA Technical Paper 1099 NASA (1978). [17] http://combustion.ucsd.edu(2016-12-14). [18] E. Mastorakos, Prog. Energy Combust. Sci. 35 (1) (2009) 57–97, doi:10.1016/j.pecs.2008.07.002. [19] J. Eaton, J. Johnston, AIAA J. 19 (9) (1981) 1093–1100. [20] O. Schulz, N. Noiray, Combust. Flame 192 (2018) 86–100, doi:10.1016/j.combustflame.2018.01.046. http://www.sciencedirect.com/science/article/pii/ S0010218018300609. [21] C.M. Arndt, M.J. Papageorge, F. Fuest, J.A. Sutton, W. Meier, M. Aigner, Combustion and Flame 167 (2016) 60–71, doi:10.1016/j.combustflame.2016. 02.027. [22] A. Gruber, J.H. Chen, D. Valiev, C.K. Law, J. Fluid Mech. 709 (2012) 516–542, doi:10.1017/ jfm.2012.345. http://www.journals.cambridge.org/ abstract_S002211201200345X. [23] C. Eichler, T. Sattelmayer, Exp. Fluids 52 (2) (2011) 347–360, doi:10.1007/s00348- 011- 1226- 8. [24] A. Kalantari, V. McDonell, Prog. Energy Combust. Sci. 61 (2017) 249–292, doi:10.1016/j.pecs. 2017.03.001. http://linkinghub.elsevier.com/retrieve/ pii/S0360128516301253.

Please cite this article as: D. Ebi et al., Ignition of a sequential combustor: Evidence of flame propagation in the autoignitable mixture, Proceedings of the Combustion Institute (2018), https://doi.org/10.1016/j.proci.2018.06.068