This paper describes experimental measurements of forced ignition of prevaporized liquid fuels in a well-controlled facility that incorporates nonuniform flow conditions similar to those of gas turbine engine combustors. The goal here is to elucidate the processes by which the initially unfueled kernel evolves into a self-sustained flame. Three fuels are examined: a conventional Jet-A and two synthesized fuels that are used to explore fuel composition effects. A commercial, high-energy recessed cavity discharge igniter located at the test section wall ejects kernels at 15 Hz into a preheated, striated crossflow. Next to the igniter wall is an unfueled air flow; above this is a premixed, prevaporized, fuel–air flow, with a matched velocity and an equivalence ratio near 0.75. The fuels are prevaporized in order to isolate chemical effects. Differences in early ignition kernel development are explored using three synchronized, high-speed imaging diagnostics: schlieren, emission/chemiluminescence, and OH planar laser-induced fluorescence (PLIF). The schlieren images reveal rapid entrainment of crossflow fluid into the kernel. The PLIF and emission images suggest chemical reactions between the hot kernel and the entrained fuel–air mixture start within tens of microseconds after the kernel begins entraining fuel, with some heat release possibly occurring. Initially, dilution cooling of the kernel appears to outweigh whatever heat release occurs; so whether the kernel leads to successful ignition or not, the reaction rate and the spatial extent of the reacting region decrease significantly with time. During a successful ignition event, small regions of the reacting kernel survive this dilution and are able to transition into a self-sustained flame after ∼1–2 ms. The low-aromatic/low-cetane-number fuel, which also has the lowest ignition probability, takes much longer for the reaction zone to grow after the initial decay. The high-aromatic, more easily ignited fuel, shows the largest reaction region at early times.
With interest in future energy sources , attention has increased in examining alternative jet fuels for aircraft turbine engines . There are a number of operational constraints that an effective jet fuel must meet; one of these is reliable forced ignition. In aircraft engines, the ability of the spark igniter to induce ignition is critical for both ground engine starts and emergency high altitude relight . A number of previous ignition studies have examined the minimum energy required to induce ignition [4–8]. Though minimum ignition energy is a practical and useful quantity, it can over-simplify the issues associated with successful forced ignition in a jet engine combustor.
One step in successful ignition is the spark discharge itself and specifically the processes that determine the energy deposition into the spark kernel . After the discharge, however, the spark kernel needs to transition into a self-sustained flame . In that study, the authors demonstrated the dramatic decrease in ignition probability if the spark kernel had to transit through a thin layer of unfueled air before encountering a fueled region. Finally, for successful lighting of a combustor, the self-sustained flame must propagate or convect to some location in the combustor where a stable flame can be held . In this study, we focus on the second process, which is the transition of the spark kernel into a self-sustained flame.
A number of factors can control whether a spark successfully transitions to a self-sustained flame. Physical properties of liquid fuels will affect atomization [3,11,12] and vaporization [2,11], thereby influencing local fuel–air ratios. Chemical components of various jet fuels blends have also been shown to have statistically significant effect on ignition probability [6,13]. For example, a study of forced ignition in a striated flow with a sunken fire igniter  compared various conventional and synthesized fuels. The fuels were prevaporized to remove the influence of physical properties and isolate chemical composition effects. The study demonstrated significant differences in ignition probability for the different fuels, with aromatic content showing a reasonable correlation with ignition probability. The two fuels in that study that showed the largest difference in ignition behavior from a standard Jet-A fuel were the ones designated C-1 and C-5. However, no detailed diagnostics were applied in that study to examine the chemical evolution of the kernel. High-speed imaging has been used successfully to study forced ignition in spark-ignition engines  and gas turbine model combustors [15,16].
Therefore, the primary objective of this work is to elucidate the processes by which a forced ignition kernel evolves into a self-sustained flame in the early times after a spark discharge using high-speed imaging diagnostics. Additionally, this work examines the effect of fuel composition on forced ignition success by comparing the three fuels identified in the earlier study .
The dashed square box highlights the test section where ignition events were observed. The height, width, and length of the test section are 54.0, 85.7, and 215.9 mm, respectively. The walls of the test section are equipped with quartz windows, allowing full optical access to the test section. Inflow air was preheated to 478 K and regulated to target a 12 m/s mean flow velocity in the test section. A perforated plate after the inlet is used to produce a uniform flow velocity. Downstream of the perforated plate, a 0.635-mm-thick steel plate separates the heated air flow into an upper main flow and a lower kernel flow. Three fuel injection rods deliver prevaporized fuel into the main flow 30 cm upstream of the test section, allowing for uniform distribution of fuel and air in the main flow. The height of the splitter plate is adjustable between 6.35 and 12.7 mm above the floor of the test section. In this study, the splitter plate was set at 6.35 mm.
An industrial gas turbine engine igniter, utilized in previous studies [9,13], produced sparks with a nominal deposition energy of 1.25 J/spark at a frequency of 15 Hz. After each discharge, the resulting high pressure in the cavity ejects a high-temperature plasma kernel into the crossflow. The time between discharges is sufficient that all remnants of the high temperature kernels or ignited flames are swept from the test section before the next discharge (and there is no flameholding in the test section). In this study, the top surface of the igniter was raised 3.18 mm above the test section floor to improve ignition probability.
Fuel Delivery and Prevaporization System.
To isolate the influence of the fuels' chemical composition on forced ignition (and remove the influence of physical properties like heat of vaporization), the liquid jet fuels were prevaporized before introduction to the facility. Liquid fuel was supplied from a stainless steel cylinder, pressurized to 200 psi by nitrogen gas. The fuel was bottom-fed from the tank through a flow regulating needle-valve to flow metering and vaporization equipment.
The liquid fuel was sprayed and mixed into a carrier air flow prior to entering a Bronkhorst heater set at 473 K, enhancing vaporization. Heat trace and fiberglass insulation were wrapped around the delivery lines to prevent fuel condensation. Fuel vapor temperature was monitored immediately before introduction to the test section to ensure both adequate vaporization and limit pyrolysis. Through controlling the air and fuel flow rates, the equivalence ratio in the fuel/air mixing region was set at ∼0.75. Further details of the delivery and metering system have been described previously . Complete vaporization at the operating conditions was verified using a continuous-wave HeNe laser (632.8 nm) to detect any liquid droplet scattering in the test section.
Fuel Distribution Uniformity.
Nonuniform mixing of fuel and air can lead to inconsistent ignition behavior. Three fuel injection rods with approximately twenty evenly spaced holes span the region pass by the main flow upstream of the test section. As observed during the experiments, the vaporized fuel A-2 fluoresces when subject to the OH planar laser-induced fluorescence (PLIF) laser (discussed below in the Imaging Diagnostics section). This fuel fluorescence allows verification of proper mixing of fuel and air in the main flow prior to entering the test section. As the laser sheet passes through a plane centered on the ignitor, the fuel fluorescence provides the fuel distribution in the plane of greatest interest. The high-temperature kernel ejected from the ignitor traverses through this plane. An image of A-2 fuel PLIF is presented in Fig. 2. The horizontal coarse variation in PLIF signal is due to the spatial profile of the laser sheet.
Three test fuels were studied to observe the effects of chemical composition on ignition performance. Fuel A-2 corresponds to a conventional Jet-A fuel. Fuels C-1 and C-5 are test fuels, designed to have specific chemical and physical properties. Fuel C-1 has a low C/H ratio, aromatic content, and cetane number, while C-5 has a high aromatic content. Selected chemical properties and derived cetane number are listed in Table 1. Additional physical properties of the fuels can be found in the fuel properties section of .
The evolution of the ignition kernels was characterized by high-speed images acquired nearly simultaneously from three imaging systems: a schlieren setup, a planar laser-induced fluorescence system tuned to excite and detect the OH radical, and a ultraviolet (UV) emission system designed to capture chemiluminescence from electronically excited OH, denoted as OH*. A small time delay between the OH PLIF and chemiluminescence systems was used to prevent interferences from the PLIF signal on the chemiluminescence image. The light sources used for the OH and schlieren setups are described first, followed by the imaging systems used to collect the three signals.
OH PLIF Laser.
An Edgewave Nd:YAG laser operating at 10 kHz and emitting approximately 50 W at 532 nm was used to pump a tunable wavelength Sirah Credo dye laser, which was then frequency doubled, resulting in an output of 2 W of ∼283 nm emission with a pulse duration of roughly 10 ns. The laser was tuned to excite the Q1(7) transition in the Α2Σ(v′ = 1)←X2Π (v″ = 0) band (also designated laser as the (1,0) band) of the OH molecule. The excitation wavelength was optimized for OH fluorescence by tuning the dye laser and observing LIF from a propane-air Bunsen flame. Fused silica lenses were used to form and direct a 1-mm-thick laser sheet approximately 30 mm × 70 mm into the test section, illuminating the region of expected early kernel development.
The hot ignition kernel was imaged with a point-blocked schlieren system to detect the density gradient between the hot kernel and lower temperature surrounding flow, as utilized in a previous work . The system consisted of a point light source, two off-axis parabolic condensing mirrors, and an opaque point spatial filter to block unrefracted light as depicted in Fig. 3.
Three eight-bit high-speed cameras captured the individual PLIF, chemiluminescence, and schlieren signals. As the 10 kHz detection systems could not be synchronized to the 15 Hz exciter that controlled the igniter discharge, a fourth camera operating at a higher framing rate was used to determine the delay between the high-speed images and the discharge pulse. Table 2 provides a summary of the imaging components. A top view of the relative position of the four cameras with respect to the test section is shown in Fig. 4. A 10 kHz transistor-transistor logic (TTL) signal from a gate and delay generator was used to trigger the three image system cameras and the Edgewave Nd:YAG pump laser. Thus, the pump laser, PLIF camera, chemiluminescence, and schlieren cameras were synchronized.
A Photron SA5 camera (C4) equipped with a Lambert HiCATT intensifier (I2) was used to capture fluorescence from OH radicals. The PLIF camera operated in externally synchronized mode, with the intensifier synchronized to the camera frame rate. The gates of the intensifiers were set such that only the PLIF camera could capture emission from laser-induced fluorescence. The relations between the two intensifiers and the laser pulse are shown in the timing diagram of Fig. 5. A bandpass filter centered at 315 nm with 15 nm full-width-half-maximum was placed in front of the lens to pass red-shifted OH fluorescence in the (1,1) and (0,0) vibrational bands. During data acquisition, approximately 1 s of video was recorded at a resolution of 896 × 848 pixels.
A second Photron SA5 camera equipped with a LaVision high-speed IRO intensifier was used to capture emission from the kernel. The bandpass filter (centered at 320 nm) installed on the lens was used to detect primarily OH* chemiluminescence, though any UV emission in this wavelength range is also captured. The camera and the intensifier were synchronized to the same 10 kHz signal used by the Nd:YAG pump laser and the OH PLIF system. The size of the recorded images (in pixel dimensions) was identical to the PLIF camera.
A Photron SA1 camera, also externally synchronized to the 10 kHz triggering signal, was used to capture schlieren images, with a pixel resolution of 704 × 704. Finally, the fourth camera (Photron SA3), without spectral filtering, was used to image the emission from the plasma discharge. The resolution of this camera was 512 × 104 pixels, focusing on a small field of view directly above the igniter. It was operated at 60 kHz but synchronized to the 10 kHz external TTL signal and thus also synchronized to all the other camera systems. With its higher framing rate, the uncertainty in the time delay between the diagnostic images and the discharge event was reduced from 0.1 ms to 0.016 ms.
The trajectory path and velocity of the hot air kernel were obtained from the high-speed schlieren images. The images were first background subtracted based on an image taken before the kernel was created (i.e., before the discharge), as illustrated in Fig. 6. A Sobel edge tracking algorithm was then used to track the edge of the kernel's boundary. A constant threshold value was used to ensure consistency in edge tracking. The coordinates of the kernel edge closest to the x-axis were obtained and transformed into distances from the igniter based on a calibration plate image taken prior to testing. Velocities were calculated from the differences in distances between frames.
Image Registration and Alignment.
The PLIF, schlieren, and chemiluminescence cameras were positioned at different angles with respect to the test section. For the images in the three cameras to be properly compared, their different fields of view and orientation need to be corrected to match. In the beginning of each day's experiments, images of a transparent plate with equally spaced dots (2 mm spacing, 0.5 mm diameter) were acquired (Fig. 7). These dots were used to dewarp the images so that images in each camera were aligned with same field of view. Since the field of view of the Schlieren camera (C2, Fig. 4) was positioned perpendicular to the test section, calibration images from the schlieren camera were used as the reference image for the other two cameras.
The pixel positions of the dots on each camera were determined manually. The corresponding dots in the schlieren camera and the nonschlieren camera formed matching pairs. These pairs were then used to obtain transformation matrices (using standard matlab image processing functions), and the transformation matrices were applied to the images using a built-in dewarping function.
Contrast Enhancement and Gray Scale Inversion.
To enhance the ability to observe spark kernel development in the PLIF and schlieren images, the contrast of the gray scale images are linearly adjusted with constant multipliers. The constant multipliers are chosen through trial and error, such that the brightness of the kernel is enhanced without saturation. The contrast adjustment algorithm can be expressed as new image = constant multiplier × raw image. The constants used for PLIF and schlieren are 3.0 and 2.0, respectively.
In addition, because human perception is better at detecting dark details on white background than detecting bright details on dark background, the images presented in the Results section are inverted. The inversion is achieved through first creating a matrix with the same size as the aligned and registered images. The values in each pixel of this matrix are set to be 255, representing the maximum of the dynamic range. This matrix can be pictured as a pure white image. The original image is then subtracted from the pure white image and an inverted image is obtained.
Spark Kernel Characterization.
For the ignition studies presented here, the inflow velocity and air and fuel temperatures were controlled and held nearly constant during experimental runs, so as to isolate fuel composition effects. Of course, another important influence on successful ignition is the strength of the spark kernel generated by the igniter. As the number of igniter pulses characterized during the high-speed imaging experiments was limited to a statistically insignificant set, it is important to determine whether the shot-to-shot behavior of the igniter could be a controlling influence in the experimental results. To this end, the schlieren images were used to characterize the early spark kernel behavior.
During early times, before significant heat release, the behavior of the kernel should be controlled by the initial plasma discharge conditions that create the kernel. Specifically, the ejection velocity of the kernel from the cavity will be controlled by the pressure impulse created by the discharge, which in turn is a function of the energy deposited during the discharge. Thus, the vertical velocity of the kernel can be used as a good indicator of the spark kernel energy.
Figure 8 provides time-histories of the kernel velocities for A-2 runs, conditionally averaged over multiple igniter pulses, with successful events separated from failed cases. As expected, the kernels decelerate as they interact with the crossflow. More importantly, the velocity difference between the two cases is generally less than 10%. Furthermore, the difference is less than 2% at the first measurement time after the kernel discharge. The one time where there is a measured difference above 10%, at ∼0.1 ms, is likely due to limitations in the image processing algorithm's ability to reliably identify the kernel edge as it emerges from the mixing layer. Figure 9 provides similar results for successful ignition events for the three different fuels. Again, the initial velocities after the discharge are nearly the same for all three fuel cases. The velocities at later times are also similar, with no systematic differences observed.
Overall, we can conclude that the deposited spark energy is not varying enough to be the controlling factor for successful ignition. In addition, the deposited spark energies are similar for all three fuel tests. Therefore, any shot-to-shot variations in the ignition events may be attributable to changes in mixing between the kernel and crossflow.
Ignition Characterization—A-2 Fuel.
One way to characterize the ignition behavior of a spark kernel is to compare the evolution of a successful kernel with that of a kernel that fails to produce a sustained flame. As a growing flame will have a corresponding increase in heat release, and thus chemiluminescence, we would expect the spatially integrated signal captured by the emission camera to increase in time for successful ignition. To examine this, the signal from each emission image was background corrected and spatially integrated over the field of view, and then normalized by the number of pixels in the integration region. Figure 10 shows the resulting values as a function of time for two spark pulses from an A-2 experimental run: a successful ignition event and a failure.
At early times (below 0.5 ms), both events show a decreasing signal. At least in part, this is due to broadband emission from the decaying spark plasma that passes through the 320-nm bandpass filter on the camera. Thus, the captured emission signal at early times can be a combination of plasma emission and flame chemiluminescence. For the unsuccessful event, the emission signal reaches the camera background level between 0.3 and 0.4 ms. For the successful event, the signal decreases until ∼0.6 ms, at which point it rises continuously. Moreover, it never drops as low as the unsuccessful event. Thus, we can conclude that the emission signal after 0.5–0.6 ms is solely due to flame emission (primarily OH* based on the optical filter employed) and that “ignition” has occurred by this time. Furthermore, the time required to reach a minimum intensity can be used to characterize successful ignition. In this paper, this time will be (arbitrarily) defined as the ignition delay time.
What is unclear from this spatially integrated analysis is what happens before this time, as one might expect chemical reactions could occur as soon as the hot kernel reaches the fuel–air mixture in the region above the splitter plate. To examine the early time behavior of the kernel, we now turn to the simultaneous information from the three high-speed imaging systems: emission, PLIF and schlieren. Example images from a single event for a successful ignition event using the A-2 fuel are shown in Fig. 11. Similar images for a failed event from the same experimental run are shown in Fig. 12. As described in the Experimental Setup section, the images were registered and dewarped for the off-axis cameras (chemiluminescence and PLIF) to match the schlieren camera images.
As expected from the results of Fig. 10, the overall UV emission captured by the chemiluminescence camera decays during these early times after the kernel is created. In addition to the peak intensities dropping, the size of the emitting region of the kernel also shrinks. On the other hand, the schlieren images show an increasing kernel size, with the kernel resembling a pulsed jet. Since the schlieren demarks large density gradients (i.e., between the cold crossflow and hot kernel) while the plasma emission from the kernel should be a strong function of its temperature, the logical interpretation of these results is the kernel is mixing with the crossflow. Thus, the overall size of the hot gas region is increasing, but the kernels (average) temperature is decreasing. This is consistent with previous interpretations of the kernel's evolution in similar situations . Furthermore, the emission comes primarily from the upper portion of the kernel; the long trailing tails from the kernel seen in the schlieren images likely represent only “warm” gas.
The mixing of the hot kernel with the fuel air mixture has already occurred before the first image shown (0.133 ms), as evidenced in the schlieren image (top row, Fig. 11), where the top of the kernel is located above the splitter plate (seen as the rectangular dark shadow in the lower left edge of the image). Based on the velocity measurements presented above, the top edge of the kernel should reach the mixing layer within 60 μs after the spark discharge.
The PLIF images in Fig. 11 (middle row) display two notable features. First, the fluorescence is nearly uniform in distribution within the region illuminated by the sheet. This results from components of the A-2 fuel that fluoresce. The relative uniformity in the fluorescence in the upper regions indicates the evenness of the fuel seeding in the upper flow, while the lack of signal in the lower portion of the images verifies the absence of fuel in the flow beneath the splitter plate.
The second notable feature is the behavior of the PLIF signal in the region of the kernel (as demarked by the schlieren signal). Much of the kernel does not produce a PLIF signal, presumably because there is much less fuel within the kernel compared to the crossflow. There is, however, a region of quite strong signal within a narrow portion of the kernel. This is most evident in the PLIF image at 0.233 ms, coming from a highly convoluted structure within the kernel region marked by the schlieren, but nearer the kernel's edges. Moreover, the region containing PLIF signal overlaps a portion of the emission region. This increased fluorescence occurs in all the PLIF images of Fig. 11, but it peaks in the 0.233 ms image, with the region of bright fluorescence rapidly decreasing with time. Furthermore, this signal comes only from the region of the kernel that has passed into the upper (fuel–air containing) region. Thus, we can conclude that the source of the signal is associated with the high temperature kernel air interacting with entrained fuel (and air). By the 0.533 ms image, the kernel's PLIF signal is limited to a few small regions.
The specific species that give rise to this early time PLIF signal is unclear. While the laser and camera bandpass filter were set to excite and detect OH fluorescence, a number of hydrocarbons are broadband absorbers in the UV and can produce fluorescence (as evidenced by the signal coming from the A-2 fuel components). For example, possible fuel pyrolysis includes benzene and toluene, both of which exhibit UV excited fluorescence that might be detected by the 320-nm bandpass filter employed here. So, the kernel's PLIF signal may be produced by species associated with either fuel pyrolysis or oxidation (and heat release).
At later times, ∼0.6–2.1 ms (Fig. 13), we see that the region of PLIF signal begins to grow, eventually becoming a self-sustained flame. At these times, the signal is likely due primarily to OH PLIF. It is also interesting to note that by 1.1 ms, it is difficult to identify a region without fuel in the upper portion of the flow. Presumably, the mixing of the fuel–air mixture with the kernel has preceded to such an extent that there is little high-temperature, unmixed “pure” kernel fluid left.
To help interpret the early time results from the successful ignition event, we again compared them to a case of failed ignition (Fig. 12). As before, both emission and PLIF signals occur after the spark discharge. In contrast to the successful ignition event, however, the signals decrease more quickly, essentially disappearing by ∼0.5 ms. If the PLIF signal from the kernel is associated with chemical reactions involving fuel, as suggested above, the similar temporal behavior for the PLIF and emission signals also suggests that at least part of the emission is due to chemical reactions, i.e., chemiluminescence, especially after the first 0.1–0.2 ms.
Thus, it is reasonable to conclude that at least partial oxidation of the fuel, with some heat release, is occurring within the first few hundred microseconds for both failed and successful ignition events. In addition, the rate and extent of these reactions is greatest in the first 0.1–0.2 ms after the igniter discharge (or shortly after the kernel begins entraining the fuel–air mixture). The decrease in reactions after this time may be due to the decrease in kernel temperature resulting from the mixing of the hot kernel with cold reactants. Of course, this would also suggest that any heat release from fuel oxidation is more than offset by cooling due to entrainment and mixing. Given the small spatial extent of the PLIF and chemiluminescence signals at 0.533 ms, it is clear that whether ignition is successful depends on the relative amount of heat release and dilution in a quite small region of the kernel. If the heat release is inadequate or the mixing too rapid, the temperature will rapidly decrease, extinguishing the reactions.
Ignition Characterization—Fuel Comparison.
Similar high-speed imaging results were also acquired for two other fuels, C-5 (Fig. 14) and C-1 (Fig. 15). The most noticeable difference between these results and the A-2 images is the lack of PLIF signal from the unreacted fuel. Neither of the C fuels contains cyclo-paraffins, while they are a significant component of A-2 (see Table 2). Thus, they are a possible source of the fuel PLIF seen in Figs. 11–13. On the other hand, both the C-5 and C-1 fuel PLIF images do exhibit the same kernel PLIF signals recorded for the A-2 fuel. Since the parent fuel produces no fluorescence, this provides further support that the source of the fluorescence is a species produced by chemical reactions between the high-temperature kernel and the entrained fuel–air mixture.
As with the A-2 results, the emission and PLIF signals decay after the first few hundred microseconds for both fuels. There is a notable difference between the C-5 and C-1 data; both the emission and PLIF are more pronounced and cover a greater portion of the kernel for the C-5 fuel compared to C-1. This may correlate with the higher probability of ignition for the C-5 fuel observed in an earlier study , though the number of successful kernel events recorded here is not sufficient to make this conclusion.
The C-5 data also exhibit another feature not shown in the A-2 and C-1 images sequences, a bifurcated ignition kernel. This is most visible in the PLIF image sequence, where a structure appears to breaks from the main kernel and stays closer to the lower wall. Not only does this structure lack a vertical (upward) velocity, it also does not decay in size or intensity like the upward moving kernel. A corresponding structure of roughly the same size is evident in the chemiluminescence image. At later times, this second structure leads to a self-propagating flame that is independent of the flame produced by the upward moving kernel. The lack of decay observed in this structure suggests that it does not undergo the same amount of mixing with cold fluid compared to the main kernel. This behavior occurs in some sequences for all three fuels.
Finally, we can compare the ignition performance for the three fuels based on the ignition delay described in the discussion of Fig. 10. The spatially integrated emission/chemiluminescence data for the A-2, C-1, and C-5 fuels are shown in Fig. 16, averaged over three successful ignition events for each fuel. As expected from the images shown previously, the emission initially decays for all the fuels before eventually rising as the self-sustained flame grows.
Ignition delay times based on the minimum values in these profiles are listed in Table 3. The ignition delays for C-5 and A-2 are the same (∼0.5 ms), but about 0.3 ms faster than for C-1. Since C-5 was shown to have a higher ignition probability than A-2 in a previous study , it does not appear that this ignition delay is well correlated with ignition probability. In addition to the much longer delay, the C-1 fuel also has the slowest rate of increase in chemiluminescence/emission signal. This may correlate with the lack of aromatics in the C-1 fuel.
Simultaneous, high-speed (10 kHz) emission, OH PLIF, and schlieren images were acquired in a flow where a plasma kernel from a recessed cavity igniter is ejected vertically into a stratified crossflow. The portion of the crossflow in proximity to the igniter contained only air, while the upper flow contained a flammable, uniform mixture of prevaporized liquid fuel and air. Experiments were performed for three liquid fuels: A-2, a standard jet fuel (Jet-A), and two fuels with specific modifications. C-1 has a low C/H ratio, aromatic content, and cetane number, while C-5 has a high aromatic content. Data were acquired for all three fuels under the same operating conditions (air and fuel temperatures, flow velocities, and equivalence ratios).
Kernel velocities obtained from the schlieren images at early times (first few hundred microseconds) were essentially the same for all the fuels and for both successful and failed ignition events. This suggests that the energy deposited in the initial kernel by the discharge did not vary significantly. Comparison of the schlieren and emission images suggests rapid entrainment of crossflow fluid in the upwardly moving portion of the kernel. The PLIF images provide strong evidence of chemical reactions between the high temperature kernel air and the entrained fuel–air mixture that start within tens of microseconds after the kernel begins interacting with the fueled portion of the crossflow. While the source of the PLIF signal could be fuel decomposition products or OH, the spatial and temporal correlation between the PLIF and emission results suggests some chemiluminescence may be occurring, indicating the possibility of heat release at these early times. Whether the kernel leads to successful ignition or not, the reaction rate and the spatial extent of the reacting region decrease significantly after the first 0.1–0.2 ms. This can be explained by kernel cooling caused by dilution from crossflow fluid being rapidly entrained and mixed into the kernel. For successful ignition events, small regions of the reacting kernel survive the dilution process beyond 1 ms and are able to transition into a self-sustained flame.
All three fuels exhibit the same general behavior with respect to the ignition process. One major difference observed from the high-speed images is that the low-aromatic (C-1) fuel takes much longer for the reaction zone to grow after the initial decay. The ignition delay defined here is nearly the same for the A-2 and C-5 fuels, even though the C-5 fuel was shown to have a higher ignition probability in a previous study. The C-5 fuel did show a larger reaction region at early times compared to A-2. This may provide it a greater chance of surviving the dilution process and thus results in the higher ignition probability. The higher aromatic content of the C-5 fuel may be related to this behavior.
This research is sponsored by the FAA ASCENT project. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the FAA or other ASCENT Sponsors.
U.S. Federal Aviation Administration (FAA) Office of Environment and Energy as a part of ASCENT Project 13-C-AJFE-GIT-008 (FAA Award No. 27A).