Application of diffuse background illumination to lean direct injection combustion undergoing oscillations

Abstract

Diffuse background illumination (DBI) diagnostics is applied to study spray instabilities in lean direct injection combustion systems for gas turbines. Experiments were performed in a reacting kerosene spray at atmospheric pressure using DBI, Mie Scattering, and OH* chemiluminescence (OH*-CL) imaging to delineate instability dynamics. Comparison of DBI and Mie scattering results shows that the Mie scattering is effective in illustrating the planar structure of the dispersed spray, but the line-of-sight DBI provided an improved visualization of the off-axis features of the spray to aid in understanding the spray dynamics. Measurements were postprocessed into phase-reconstructed data to illustrate the dynamic relationship between spray and OH*-CL oscillations and to demonstrate the effectiveness of DBI imaging for illustrating the influence of spray structure on the flame topology. Results show that DBI provides a clear illustration of how spray oscillations govern the switching between premixed (lean or rich) and diffusion modes of combustion over the course of the oscillation cycle.

Keywords

DBI fuel sprays LDI combustion low emissions flow oscillations

Introduction

A robust understanding of the spray structure in liquid-fueled combustion systems is of critical importance because fuel introduction into the reaction zone governs the nature of the resulting flame, and thereby, the pollutants generated in the process. The methods to study fuel sprays vary widely depending on the application of interest, the nature of the spray, and the objectives of the study. In the diesel engine community, nonreacting and reacting high-pressure jet sprays are commonly studied to describe macroscopic spray parameters such as the penetration length and the divergence angle.^1–4 Liquid spray length is of interest to offer insight into the underlying physics of fuel evaporation, fuel-air mixing, and combustion and to avoid unwanted spray-wall interactions.⁵ Additionally, the jet divergence angle offers insight into air entrainment, which affects fuel evaporation and fuel-air mixing before combustion.⁶ In these relatively dense sprays, the fuel dispersal and fuel-air mixing in the near field are limited by the intermittent and extremely short injection times (on the order of a few ms) of high-pressure, high-velocity fuel jets.

In contrast, aviation gas turbines typically employ twin-fluid atomizers to form continuous, highly dispersed sprays in the near field which enhances fuel evaporation and fuel-air mixing. Particularly, lean direct injection (LDI) combustion seeks to pursue rapid fuel-air mixing to approach lean premixed (LPM) combustion to greatly reduce pollutant emissions.⁷ Lean direct injection combustion systems are prone to hydrodynamic and thermoacoustic instabilities, which motivated the present study to probe continuous, disperse sprays that can couple with flow instabilities or acoustic modes of a combustor leading to worsened emissions⁸ and/or catastrophic failure.⁹ An ability to appropriately diagnose spray oscillations is crucial to mitigate this behavior.

Several optical diagnostics are available to measure the spray characteristics. Phase-Doppler particle analysis is a point measurement technique to obtain droplet size and velocity distributions in sprays, and it is commonly used to study sprays in diesel engines,¹⁰ industrial burners,¹¹ and gas turbines.¹² Mie scattering is an imaging technique, wherein a laser sheet illuminates a cross-section of the spray, and light is scattered toward a camera sensor depending upon the laser sheet intensity and droplet diameter. Mie scattering has the advantage of assessing the spray cross-section in a snapshot and it has been used extensively to quantify spray length and divergence angle in diesel applications,¹³ and to characterize spray structure and spray dynamics in gas turbine applications.^9,14 However, in dispersed and unsteady sprays, Mie scattering can be challenging to apply because of the high sensitivity of the measurement to the illumination source.¹⁵

Diffuse background imaging (DBI) is a line-of-sight diagnostic, wherein the spray is back-illuminated by a diffuse light source, and liquid droplets in the spray attenuate the light along with a camera's line-of-sight according to the Beer–Lambert law.¹⁵ Diffuse background illumination technique was developed to acquire benchmark data in relatively dense diesel sprays.¹⁶ In recent years, DBI has become the recommended technique to study diesel spray structure within the Engine Combustion Network¹⁷ because of its ability to reliably measure the liquid volume fraction along a line-of-sight in relatively dense diesel sprays.¹⁸ However, the beam-steering effects within the vapor phase of the spray must be eliminated by carefully selecting an optical diffuser with a sufficiently large diffuser angle.¹⁹

Diffuse background illumination has been used mainly to study intermittent, dense diesel sprays,^20,21 but in recent years, it has also been employed to study dispersed sprays relevant to aviation gas turbines. In highly dispersed evaporating sprays, multiphase effects render it impossible to accurately apply the Beer–Lambert law to quantify the liquid volume along with the line-of-sight in the spray. In such cases, the self-calibrating nature of the images still renders DBI as an attractive diagnostic to evaluate the global spray behavior including penetration length, divergence angle, and temporal evolution of spray structure.²² Recently, Cardona et al. employed microscopic DBI to study the droplet diameter and velocity distributions in a pressure-swirl atomizer for various fuels and air coflow rates.²³ Microscopic DBI enhances the ability to track individual droplets in sufficient resolution by direct imaging but at the expense of resolving the global spray structure.²⁴ Additionally, Li et al. employed DBI to study combustion within a rarefied coal particle field.²⁵

As noted earlier, LDI combustion increases the propensity for combustion instability²⁶ which has numerous potential triggers, and their interactions are highly nonlinear.²⁷ In a classic combustion instability, one or more natural acoustic modes of a combustor become excited, causing bulk oscillations in the flow of reactants feeding the flame.²⁸ Additional mechanisms include the presence of hydrodynamic instabilities which induce dynamic behavior in a flame,²⁹ harmonic excitation of vortex-shedding structures at the combustor dump plane,³⁰ and oscillations inherent to the fuel injection process.^31,32 In liquid-fueled systems, the heat release rate (HRR) oscillations could also be related to the oscillations in the spray field, either resulting from flow rate oscillations and/or localized variations in the atomization and mixing processes.³³

In this study, interactions between spray and flame oscillations will be explored using DBI, Mie scattering, and OH* chemiluminescence (OH*-CL) techniques. Measurements will be acquired simultaneously in pairs of DBI and OH*-CL and Mie scattering and OH*-CL. A proper orthogonal decomposition (POD) technique developed by the authors³⁴ will be employed to phase-reconstruct data at a resolution of 1°. Results will be analyzed to accomplish two objectives: First, different postprocessing methods for DBI and Mie scattering will be compared to draw contrast between the two techniques. Subsequently, simultaneous DBI and OH*-CL measurements will be presented in detail to illustrate the ability of DBI in describing the dynamic relationship between spray and heat-release processes. The following sections include details of the experimental set-up, data processing techniques, results and discussion, and finally conclusion of the study.

Experimental procedure

This section provides details of the test facility and operating conditions, and optical layout of different diagnostics.

Test facility

Figure 1 shows an LDI combustion setup using a swirl-venturi injector scheme developed by NASA for aviation gas turbines.³⁵ The atmospheric pressure combustor is an 80 mm OD, 400 mm long, naturally cooled quartz tube to confine the flame. Compressed air from a compressor is dehumidified and regulated to a desired supply pressure, and the airflow rate is measured by a laminar flow element (CME 10-5-300G). Flow then passes through a large box of glass beads to homogenize the airflow. The air temperature can be regulated by a preheater, but this capability is not employed in the study. Atomizing air is supplied to the injector from a compressed air bottle manifold after regulating it to a safe operating pressure. Flow control is provided by an Alicat PCT-500 pressure controller upstream of a sonic nozzle that is used to measure the atomizing airflow rate. Kerosene is supplied from a fuel tank pressurized by air. Fuel flow is metered by a needle valve and measured with an Emerson LF4M-series Coriolis flow meter. Methane regulated by a Sierra SmartTrak C100 mass flow controller is employed to initiate the flame in the combustor. Once the methane flame is established, the kerosene flow is enabled gradually to achieve the desired fuel flow rate. Concurrently, the methane flow rate is reduced until stable operating conditions are obtained with liquid fuel only.

Figure 1.

Thermoacoustic test cell schematic.

A commercial air-blast atomizer (Delevan SNA-type 30609-2) is used to create a dispersed spray typical of aviation gas turbines. Figure 2(a) illustrates the integration of the injector and swirl-venturi nozzle with the plenum and Figure 2(b) presents internal details of the swirl-venturi nozzle. As shown in Figure 2(a), atomizing air (bounded by dark blue surfaces) enters from the bottom and liquid fuel (red) enters from the side of the nozzle. The atomizing air flows through channels in the injector body until it passes through the atomizing air swirler located near the injector tip with orifice diameter of 0.76 mm. The liquid fuel and atomizing air initially interact near the injector tip, where the swirling atomizing air induces primary atomization. Secondary atomization occurs when the partially atomized fuel spray interacts with the counter-rotating swirling coflow (light blue) of primary (combustion) air, and fuel atomization and fuel-air mixing are further enhanced as the flow accelerates through the venturi into the combustor.

Figure 2.

(a) Plenum cross section; (b) swirl-venturi cross section; (c) injector assembly. (1) – Injector housing; (2) – Fuel pin; (3) – Atomizing air swirler; (4) – Injector tip; (5) – Primary air swirler; (6) – Venturi; (7) – Spray.

The primary swirler has six vanes at 28° to the horizontal and a swirl number of 1.5. The swirler feeds directly to the venturi. The swirl-venturi constricts to throat ID of 12.5 mm before expanding to 25 mm ID at the exit coinciding with the combustor dump plane. Ordinarily, the fuel atomizes and evaporates within the swirl-venturi such that the spray is barely noticeable downstream of the dump plane. Some conditions, however, enable substantial flow oscillations, and under these conditions, large-scale spray structures can penetrate cyclically into the field of view (FOV). This study will focus on strong spray oscillations produced at an overall equivalence ratio of 1.0, atomizing air-to-liquid mass ratio of 2.0, and HRR of 9 kW.

Optical setup

Figure 3 shows a schematic of the optical setup for DBI, Mie scattering, and OH*-CL. The DBI setup includes a series of optical components to create a spatially uniform background of high-intensity illumination. The light source is a single-pulsed broadband LED controlled by a driver from Lightspeed Technologies (HPLS-36DD7500). The LED is placed at the focal point of a Fresnel lens with a diameter of 100 mm and a focal length of 71 mm (Edmund Optics, 32-683) to generate collimated light. An engineered diffuser (RPC Photonics, EDC-15-15132-A) with a divergence angle of 16° is employed to obtain a uniform DBI with minimum beam steering. The DBI images of the spray were acquired by Photron FASTCAM Nova S12 1000 K camera with a Sigma 105 mm F2.8 Macro lens (L1 in Figure 3) providing a FOV of 30 mm × 30 mm, with the bottom of the image at the dump plane and centered on the combustor axis and spatial resolution of 30 μm per pixel. Images were acquired at a framing rate of 12 kHz and a camera exposure time of 1 μs which reduced motion blur to subpixel level for the LED pulse duration of 500 ns. However, individual droplets were not visualized because of the dispersed nature of the spray.

Figure 3.

Optical setup schematic.

For Mie scattering, a Quantronix Hawk-Duo Nd-YAG laser is employed to produce a laser sheet using a turning mirror and a sequence of spherical (F = 500 mm) and cylindrical (F = −25 mm) lenses to create a laser sheet for illumination. The laser sheet passes through the midplane of the combustor, where fuel droplets in the flow scatter light toward the camera, depending upon the sheet intensity and droplet diameter. The laser pulse duration of 5 ns essentially eliminated any motion blur. A Photron Nova camera is employed in the same manner as for DBI, with the addition of a bandpass filter centered at 532 nm with a bandwidth of 2 nm.

The OH*-CL system employs a Photron FASTCAM SA5 camera with a 50 mm lens (L3), in conjunction with a UVi 1850-10 ultraviolet intensifier and attached 105 mm UV lens (L2). A bandpass filter assembly centered at 315 nm with a 15 nm bandwidth was used to detect OH*-CL emissions mainly at 308 nm. For synchronous operation with spray imaging at 12 kHz, the camera produces an 80 mm × 80 mm FOV at spatial resolution of 104 μm/pixel. The intensifier gain is 70% of the full scale, and intensifier gate time is 20 μs, delayed 10 μs after the end of the DBI image exposure to eliminate interference from the LED. Exposure time for OH*-CL was optimized based on the FOV, pixel resolution, framing rate, and intensifier gain. In this time interval, the fastest ballistic droplets (at about 30 m/s)³⁶ will travel about 0.6 mm or less than 1% of the OH*-CL image size. The timings and synchronization for the LED, laser, and cameras are enabled by Berkeley Nucleonics Model 575 digital delay generator. In this study, synchronous measurements were acquired in pairs: OH*-CL and DBI and OH*-CL and Mie scattering. For all diagnostics, a total of 18,000 images were recorded for a test duration of about 1.5 s.

Data processing

Data processing consists of three sequential steps. In the first step, instantaneous DBI and Mie scattering spray images are created for direct comparison by eliminating the background effects. In the second step, a POD technique developed by the authors³⁴ is employed to phase reconstruct OH*-CL data and synchronously acquired spray data either by DBI or Mie scattering. The third step involves averaging of multiple images at a phase angle window of 1°, producing a sample of approximately 50 snapshots per bin to ensure accurate statistical analysis. For DBI and Mie-scattering, different averaging algorithms are employed to illustrate the spray structure. In the case of OH*-CL, Abel inversion is used to obtain the planar data from the averaged image at each phase angle.

Instantaneous spray images

Data processing to create instantaneous DBI or Mie scattering images is similar, with a few minor differences. In the case of DBI, a set of background images was acquired before the experiment, and the arithmetic mean of brightness at each pixel location was taken as the background brightness in the image. During the experiment, the spray attenuates or extinguishes light to produce DBI images with reduced brightness. Thus, the light attenuation at each pixel location in the spray is determined by subtracting DBI image brightness from the background image brightness. This procedure results in positive values, approximately in proportion to the liquid droplets present along with a given pixel's line of sight. Next, the light attenuation fraction is obtained by normalizing the local light attenuation by the local background brightness to characterize the relative features of the spray throughout the combustor.

In the case of Mie scattering, images of the laser sheet interactions with the combustor are acquired prior to testing, and the average brightness of these images is taken as the background brightness. Mie scattering directs light toward the camera to increase brightness in areas subjected to fuel droplets. Thus, the light enhancement at each pixel location in the spray is determined by subtracting background image brightness from the Mie scattering image brightness. The difference in brightness of these images is normalized by the camera's saturation intensity. Pixel saturation is not a concern because of the focus on overall spray structure rather than individual droplets. In this way, DBI and Mie scattering images can be compared with each other.

Figure 4 presents sequential snapshots of DBI (a) and Mie scattering (b) images at approximately the same time, although a precise temporal match is not possible. In the first (top left) DBI image of Figure 4(a), the spray appears as a narrow central jet, penetrating up to Z = 15 mm. A wide droplet cloud ranging from R = −10 mm to R = 10 mm is also visible in the near field. The corresponding Mie scattering image (top right) in Figure 4(b) illustrates a droplet cluster along with the centerline at Z = 10 mm. Both diagnostics illustrate the width of the central spray column as ranging from R = −5 mm to R = 5 mm. The DBI captures the diverging droplet cloud on the spray periphery, a feature unrecognized by Mie scattering because few droplets reside in this region within the imaging plane.

Figure 4.

Sequence of instantaneous images of normalized DBI light attenuation (a), normalized Mie-scattered light (b).

The DBI image in the second row illustrates an elongated central spray column extending to Z = 18 mm before dispersing both radially and axially in subsequent snapshots. Note that DBI can detect regions of the spray with fewer droplets because of the line-of-sight nature of the diagnostics. Diffuse background illumination images in rows 3–5 illustrate dispersed spray regions with relatively few droplets. In contrast, Mie scattering images are effective at illustrating axial progression of spray structures near the centerline, but outer spray regions with fewer droplets are not easily detectable because fewer droplets reside in the imaging plane. In these outer regions, very few droplets will exist in the imaging plane of Mie scattering. Thus, DBI images tend to depict wider and taller disperse spray structures in comparison to Mie scattering images.

Proper orthogonal decomposition–based phase reconstruction

The OH*-CL images were processed by normalizing the emission intensity by the maximum light intensity recorded in the dataset. These normalized images were used to perform POD to facilitate the phase reconstruction of OH*-CL emissions and the synchronized spray measurement, DBI, or Mie scattering. Proper orthogonal decomposition allows for the decomposition of a turbulent flow field into deterministic functions, such that portions of the total fluctuating kinetic energy can be defined.³⁷ A turbulent flow field, i’, which is the measure of fluctuations in the field, can be represented according to equation (1), as the infinite summation of the products Φ_k, the deterministic spatial functions (modes), and a_k, the associated time coefficients. These Φ_k define correlation in space, and the respective time coefficient function, a_k, for each mode defines how that structure oscillates in time:

i_{(x, t)}^{'} = \sum_{k = 1}^{\infty} a_{k} (t) φ_{k} (x)

(1)

The numerical basis for this process is described in detail throughout the literature; Weiss provides an outstanding introduction to the associated numerical methods and their interpretation.³⁸ Previous work by the authors³⁴ determined that the time coefficients of POD decomposition can be used to identify the cyclical behavior of the flame dynamics. A 100 Hz passband centered at the frequency of flame oscillations is employed to filter out the background noise at other frequencies. The two most significant POD modes oscillate at the same peak frequency and can be used to evaluate the phase angle of each snapshot in a representative cycle using the relation described in equation (2). Here, each time coefficient is normalized by the eigenvalue of the associated mode, and a scatter plot of a_1j vs a_2j provides approximately a unit circle for sinusoidally oscillating time coefficients. The cycle phase angle (θ) of each snapshot then describes where that snapshots’ point lies azimuthally on the resulting unit circle:

\arctan (θ_{j}) = \frac{a_{2 j} \sqrt{2 λ_{1}}}{a_{1 j} \sqrt{2 λ_{2}}}

(2)

Phase averaging

Phase reconstruction organizes OH*-CL, DBI, and Mie scattering images into 1° bins, each containing an average of 50 of the 18,000 total images. In the case of OH*-CL, a simple averaging was performed in each bin, and then, an Abel inversion algorithm, described in detail by Dasch³⁹ and Kohle and Agrawal,⁴⁰ was used to obtain local data in an axisymmetric field. This is analogous to a planar measurement of flame activity such as OH-PLIF imaging. Although POD phase reconstruction was applied to OH*-CL images only, the simultaneously acquired DBI or Mie scattering images could also be phase averaged similarly. For each phase reconstruction, the starting value of the phase angle is arbitrary, and so for each set of reconstructions, the phase angle θ = 0° was defined as the bin with maximum liquid penetration into the combustor.

Once the spray images are binned, the average spray structure in each bin or phase angle can be described in various ways. Ensemble averaging described by Sieber⁴¹ is the simplest method, wherein the DBI or Mie scattering signal at each pixel location is averaged across all images in the phase angle bin. Although simple, this analysis technique introduces errors because the presence or absence of droplets in individual images is not taken into consideration to calculate the averages. Alternatively, as shown by Wanstall,⁴² the spray can be described probabilistically, and so within an instantaneous snapshot, pixels exceeding a threshold signal intensity are said to contain a droplet, and the droplet probability at a pixel can be equated to the percentage of snapshots containing a droplet at a given pixel location.

Results and discussions

In the first part of this section, DBI and Mie scattering results will be compared to highlight the differences between the two techniques. Both ensemble and probabilistic methods will be employed for phase averaging, and to display the unique features of DBI diagnostics for investigating unsteady disperse sprays. In the second part, synchronous DBI and OH*-CL measurements will be presented to explore the dynamic relationship between spray and flame oscillations.

Comparison of spray structures from DBI and Mie scattering

Figure 5 compares averaged results for the phase angle of 1°, showing DBI in the left column and Mie scattering in the right column. The ensemble-averaged results are shown in the top row, while the bottom row displays probabilistically averaged results. Figure 5 reveals that all four plots are different, indicating that both diagnostics and analysis techniques affect the visualization of the spray structure. The DBI plot in the top row shows a central column of fuel penetrating along with the centerline up to Z = 18 mm. The droplet cloud in the near field of up to Z = 10 mm has a short divergence angle, and it ranges from R = −10 mm to R = 10 mm. Interestingly, fine droplets observed at farther radial and downstream locations observed in Figure 4 are washed out by the averaging procedure in Figure 5(a). Note that dispersed, fine droplets reside at different spatial locations in each snapshot, and hence, the averaging process diminishes the signal because these droplets rarely occupy the same line-of-sight in multiple DBI snapshots. Overall, DBI is shown to provide an excellent visualization and quantification of the spray structure, that is, a dense droplet cluster near the venturi exit that decreases gradually in both radial and axial directions.

Figure 5.

Spray penetration at 1° phase angle illustrated by (a) ensemble-averaged DBI, (b) ensemble-averaged Mie-scattering, (c) probabilistic averaged DBI with 6% binarization threshold, and (d) probabilistic averaged Mie scattering with 6% binarization threshold.

Ensemble-averaged Mie scattering image illustrated in Figure 5(b) displays similar spray structures, that is, a dense droplet cluster near the venturi exit, decreasing gradually in both radial and axial directions. The Mie scattering image reveals (1) farther spray penetration to Z = 25 mm, and (2) a wider droplet cloud near the venturi exit extending to R = ±12 mm. In Mie scattering, fine dispersed droplets are present at nominally the same planar location. Thus, the averaging procedure accentuates signal at a given pixel location, which is opposite to the result of ensemble-averaged DBI images.

The bottom row of Figure 5 shows probabilistic averaged results using an arbitrary binarization threshold of 6% to maximize droplets detected without introducing significant sensor noise and beam steering effects. In practice, the binarization threshold should be chosen judiciously since a lower threshold would capture more droplets but is more likely to suffer from beam steering. In the case of DBI, any pixel in an instantaneous image with background illumination attenuated by 6% or more is assumed to contain at least one droplet along its respective line-of-sight. This approach has the effect of accentuating the average signal strength or probability in regions with a sufficiently large number of droplets. As such, a dense region of spray near the venturi exit in Figure 5(c) displays a 100% probability of droplet presence along the line-of-sight; note that the probability value varies somewhat relative to the chosen threshold. Results show that the probabilistic averaging for DBI also tends to wash out rarefied droplets, similar to the ensemble-averaging approach.

Mie scattering plot in Figure 5(d) reveals relatively low droplet probability throughout the spray, peaking only at 20% near the venturi exit. The spray penetration to Z = 25 mm is consistent with the corresponding ensemble-averaged plot. However, the droplet cloud in the near field of the spray proved difficult to detect—the binarization threshold could not identify small droplets in this region and thus, the droplets were effectively washed out during the averaging process. In contrast, DBI with probabilistic averaging performed the best. In subsequent sections, probabilistic averaged DBI will be employed to study dynamic interactions between spray and flame structures.

Oscillation characterization

In this section, the instantaneous nature of spray and flame processes is examined using time series data along with the combustor centerline at Z = 15 mm to gain insight into the nature of the spray and flame oscillations. For both diagnostics, data in an 8 × 8 pixel region centered at Z = 15 mm are averaged for each snapshot to reduce noise. Figure 6(a) illustrates the transient response of spray and flame processes over a 25 ms time window. Dashed black lines represent snapshots within the bin closest to the start of the cycle. Typically, the peaks of DBI light attenuation fraction and OH*-CL emission intensity occur synchronously; notice the phase lag between peaks of spray light attenuation and OH*-CL emissions, and a longer duration of high OH*-CL levels. These results suggest that spray penetration into the combustor is relatively short; however, the fuel supplied by the spray evaporates and mixes with air to form a reaction zone that lasts relatively longer.

Figure 6.

(a) Instantaneous DBI light attenuation fraction and OH*-CL emission, fast Fourier transform of (b) DBI light attenuation fraction and (c) OH*-CL emissions, spectrogram of (d) DBI light attenuation fraction and (e) OH*-CL emissions.

Figure 6(b) illustrates the fast Fourier transform (FFT) of DBI light attenuation fraction signal and demonstrates a highly cyclic behavior with a fundamental oscillation frequency at 538 Hz and four harmonics of diminishing strength. Figure 6(c) illustrates the FFT of OH*-CL emissions signal indicating strong peaks in flame oscillation at the fundamental spray frequency of 538 Hz and its second harmonic. The shared fundamental frequency of these oscillations illustrates the strong degree of coupling between the spray and flame oscillations.

Figure 6(d) presents the spectrogram of DBI light attenuation fraction to identify any variations in the spectral behavior of the spray. The spectrogram is constructed with a window of 512 snapshots and an overlap of 256 snapshots. While the peak amplitude varies slightly, the peak frequency is nearly the same throughout, suggesting that very little deviation from this oscillation pattern occurs, implying the presence of a fully developed hydrodynamic or thermoacoustic instability in agreement with previous work.³⁶ Finally, Figure 6(e) illustrates the spectrogram for OH*-CL emissions, indicating no frequency drift over the test duration and strong correlation between spray and flame oscillations.

Synchronous DBI and OH*-CL measurements at selected phase angles

Figure 7 shows plots of phase-averaged droplet probability and Abel-inverted OH*-CL during an oscillation cycle to illustrate the influence of the spray structure on local flame topology. Abel inversion of the spray structure is physically unrealistic because light attenuation by droplets along with the line of sight is a nonlinear phenomenon. However, Figure 7 provides excellent insight into spray-flame interactions because the spray is relatively compact compared to the flame. Each spray snapshot represents probabilistic averaging over 1° phase angle.

Figure 7.

Phase averaged profiles of droplet probability and OH*-CL emissions.

Snapshots are shown initially at 60° intervals when little to no spray is present in the FOV, and at 30° intervals once the spray enters the FOV to better illustrate the temporal evolution of the spray structure. Note that the reference phase angle of 0° corresponds to the maximum spray penetration into the combustor. Ideally, fuel atomization, fuel vaporization, and fuel-air mixing should all occur upstream of the FOV, that is, inside the swirl-venturi, which will produce an approximately LPM flame inside the combustor. However, an acoustic instability in the combustor can temporarily elevate pressure and reduce/interrupt supplies of atomizing air and combustor air. These nonideal upstream processes introduce a visible spray structure in the FOV accompanied with flame structures signifying departure from LPM combustion.

In the first image at θ = −240° (alternatively θ = +120°), the spray in the FOV is largely absent, while the OH*CL plot displays the typical features of a swirl-stabilized LPM flame, anchoring within inner and outer shear layers formed by the diverging annular jets exiting from the swirl-venturi. At this phase angle, atomization occurs within the swirl-venturi, only a small quantity of unvaporized fuel reaches the central core in the FOV, and the majority of the fuel vaporizes and mixes with the combustion air to form a nearly LPM flame. Droplets that enter the central core can be correlated with the small, high-intensity OH*-CL (red color) zones near the swirl-venturi exit, where reactants are unlikely to burn in the LPM combustion mode.

At θ = −180°, virtually no droplets appear in the FOV, and at the same time, the reaction zone is confined to the near field (Z < 30 mm) and within R = ±15 mm, that is, the flame is localized near the swirl-venturi exit. Further examination would indicate that premixed reactants enter at the center of the combustor to form a lifted flame, some of premixed reactants also enter through the annular jet to combust outside R = ±15 mm and Z = 15 mm, and most importantly, the fuel supply to the combustor has been constrained severely. Reduction in fuel supply could be attributed to either increased acoustic pressure within the combustor or a local hydrodynamic instability that decreases combustion air, atomizing air, or a combination of both flows. In either case, the liquid fuel flow rate is virtually unaffected.⁴³ Thus, swirling airflow has weakened, and liquid fuel is confined mainly upstream in the central region, resulting in localized zones of elevated OH*-CL intensity.

Spray and flame images at θ = −120° are consistent with the above explanations. Lack of spray in the FOV at θ = −120° indicates that the fuel prevaporizes within the swirl-venturi. Increase in the size of the reaction zone suggests that the fuel supply to the combustor has increased. Referring to OH*-CL intensity in the flame zone, one can infer that majority of the premixed reactants enter near the combustor center to produce a lifted flame. In addition, some of the premixed reactants also enter through the annular jet to react in regions radially away and detached from the combustor dump plane.

At θ = −60°, fuel atomization in the swirl-venturi has degraded causing liquid fuel to enter directly into the combustor. Poor fuel atomization increases the droplet diameter, and these high-momentum droplets follow a ballistic trajectory out of the injector to concentrate along with the center region.⁴³ In the flame image, OH*-CL concentration is highest in the center region where the fuel column burns closer to the dump plane in a flame mode deviating from LPM combustion desired for modern low-NOx systems. Note that some fuel still mixes with the primary air in the swirl-venturi, and thus, enters through the annular jet to react radially away from the central core region. The spray and flame structures at θ = −30° are similar to those at θ = −60° except that the droplet concentration has increased, especially in the near field of the dump plane coincident with the flame location. In this case, droplets are likely burning in diffusion mode because of the inadequate time for fuel prevaporization and fuel-air mixing. Diffuse background illumination images clearly illustrate the effects of the spray on radial shift in the reaction activity.

At θ = 0°, the FOV of the combustor displays the worst phase of fuel atomization in the swirl-venturi resulting from reduced atomizing airflow through the injector; note the time delay between injector exit and combustor inlet. At this phase angle, DBI reveals the central fuel jet to have reached its maximum downstream penetration, but the outer droplet shroud continues to develop, indicating an increased swirl within the injector. The reaction zone at this phase angle is distributed across the combustor, correlating with the radial dispersal of the fuel by the swirling primary flow. At θ = 30°, the spray structure is similar to that at θ = 0°, but the droplet probability has decreased significantly throughout the spray. These trends signify recovery of swirling airflows, which would improve atomization and fuel-air mixing within the swirl-venturi. The OH*-CL images show improved radial flame distribution across the combustor.

At θ = 60°, the spray structure has weakened, and fewer droplets remain beyond Z = 10 mm. At θ = 90°, almost the entirety of the spray has receded into the swirl-venturi. At these phase angles, localized flame zones in the center region have either diminished or disappeared while improved fuel atomization, fuel vaporization, and fuel-air mixing result in a flame approximating LPM combustion. Diffuse background illumination clearly illustrates the highly dispersed spray structure, correlating directly with reaction zones shifting radially outward in the flame images.

In summary, DBI data illustrate spray oscillations starting with an initially well-atomized mixture, changing to a poorly atomized central fuel jet as it enters into the combustor to produce a large flame before atomization recovers gradually to produce radially distributed spray structure with small droplets in agreement with other findings in the literature.⁴⁴ In this experiment, the combustor approximates LPM combustion only for a small portion of the cycle from about θ = 60° to 120°, while the rest of the cycle undergoes nonoptimal conditions causing combustion in diffusion and rich premixed modes. More importantly, results demonstrate that DBI diagnostics can be utilized to characterize the spray structures in a highly dynamic reacting flow system, and in conjunction with OH*-CL emissions measurements, it can provide a robust description of spray-flame interactions.

Synchronous DBI and OH*-CL measurements at selected axial locations

The discussion above focused on the dynamic relationship between spray and flame structures at selected phase angles. In this section, results will be presented for the entire phase space varying from −240° (or 120°) to 120° (or 480°) but at selected axial locations to gain further insight into dynamic relationship during the instability cycle. Figure 8 presents these results at near, mid, and far-field axial locations, that is, Z = 2 mm, 10 mm, and 20 mm. The left column presents droplet probability obtained by DBI and the right column shows the Abel-inverted OH*-CL emissions. The bottom row represents the axial location closest to the combustor dump plane (Z = 2 mm). In the following discussion, the oscillation cycle will be divided into three sequential phases: (1) when the spray structure can be detected in the FOV, (2) after phase (i) when the spray structure in the FOV is diminishing or has diminished, but the flame is distributed across the combustor, and (3) after phase (ii) or before phase (i) when the spray structure is absent in the FOV but the flame is localized within the combustor.

Figure 8.

Oscillation patterns in DBI line of sight droplet probability (left) and Abel-inverted OH*-CL emissions (right) at various axial locations.

At Z = 2 mm, the spray structure spans from about θ = −100° to 120° and it is most dense between θ = −60° and 60°. As discussed previously, this pulse of liquid fuel at the combustor inlet results from poor fuel atomization causing fuel droplets to follow a ballistic trajectory. Note that high acoustic pressure in the combustor during this cyclic period reduces atomizing and primary airflows. The OH*-CL plot at Z = 2 mm indicates that reactions occur mainly between θ = −100° and 120°, which coincides with the presence of the spray structure. Overlapping flame activity with spray structure near the dump plane suggests diffusion mode of combustion whereby fuel droplets evaporate within the stoichiometric reaction zones. Increased flame activity occurs between θ = −50° and 50° when the spray is most dense. After θ = 60° and until about θ = −210° (or 150°), small droplets are present in the FOV, and the flame is devoid of high-intensity reaction zones indicating effective fuel vaporization and mixing to approach LPM combustion. Between θ = −210° and −60°, the absence of both spray structure and reaction zone suggests that the incoming premixed reactants (if any) do not produce a stable flame near the dump plane.

At Z = 10 mm, droplet probability has reduced throughout the spray structure indicating noticeable fuel evaporation within the combustor between Z = 2 mm and 10 mm. The spray span has reduced to between θ = −70° and 100° because the finer droplets outside these bounds have evaporated completely to produce premixed reactants. At this axial location, the flame is spread across the combustor except for the intense reaction zones in the central region between θ = −90° and 60° and slight narrowing of the flame zone at around θ = −150°. Peak flame activity in the center region occurs between θ = −90° and 30° coinciding with the delivery of initially premixed (and likely fuel-rich) reactants, and later liquid droplets near the center of the combustor. After θ = 0°, atomization begins to improve and the spray widens to R = ±10 mm, and the reaction zone separates radially to follow these droplets. As fuel atomization improves later in the cycle (θ > 60°), the higher-intensity localized flame zones disappear and the flame spreads radially to approximate LPM combustion. The distributed flame persists from about θ = 60° to 180° (or −180°) and thereafter, high acoustic pressure and related phenomena take over as discussed previously.

Finally, at Z = 20 mm, additional fuel evaporation has decreased fuel droplet probability. The spray span has shrunk to R = ±5 mm and θ = 0° to 90°, with concurrent reduction in fuel droplet probability. In contrast, the flame activity shows large high-intensity reaction zones near the center of the combustor that last between θ = −180° and 60°. The reaction zone is initially separated radially from θ = −180° to −60°, illustrating the converging flow of premixed reactants as the radial distribution of reactants degrades, before diverging again from θ = 0° to 60° as the small amounts of surviving droplets pass. Because these high-intensity zones do not coincide with liquid presence, they represent reaction zones associated with (fuel-rich) premixed reactants formed upstream within the combustor. Localized high-intensity reaction zones are absent from θ = 60° to −180° (or 180°) when the flame intensity is uniform across the combustor signifying LPM combustion as discussed previously. Again, DBI diagnostics has proven useful to provide meaningful correlations with the flame structure in this highly complex reacting flow.

Conclusions

In this study, LDI combustion of liquid fuel supplied by a twin-fluid atomizer placed within a swirl-venturi nozzle located upstream of the combustor was investigated. Diffuse background illumination diagnostics was applied to study the spray dynamics in this highly disperse, unstable, gas turbine–relevant reacting spray. Mie scattering analysis was also performed to compare the two techniques and to demonstrate DBI's benefits at delineating off-axis, dispersed spray structures. Synchronous DBI and OH*-CL measurements at high framing rates were acquired to illustrate the temporal evolution of the spray structure and its relationship to the dynamic flame behavior.

Results show that spray and flame oscillations are highly coupled, indicating the presence of a hydrodynamic or thermoacoustic instability. Flow oscillations produce a spray structure cycling from a well-atomized fuel-air mixture to a central column of liquid fuel in a ballistic trajectory. Then, the fuel column disintegrates as atomization improves and droplets distribute radially outward. In conjunction with Abel-inverted OH*-CL emissions, this behavior is useful to predict oscillations in local burning behavior, as regions of the flame oscillate between premixed (lean and rich) and diffusion modes of combustion. Overall, the study shows the viability of DBI diagnostics to study complex reacting flows including spray-flame interactions.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by NASA Grant NNX13AN14A.

ORCID iDs

Mitchell Johnson

Ajay K. Agrawal

References

Boggavarapu

Ravikrishna

. A review on atomization and sprays of biofuels for IC engine applications. Int J Spray Combust 2013; 5: 85–121.

Gong

Jangi

Luccini

. Large eddy simulation of air entrainment and mixing in reacting and non-reacting diesel sprays. Flow Turbul Combust 2014; 93: 385–404.

Liu

Peng

, et al. A review of phenomenological spray penetration modelling for diesel engines with advanced injection strategy. Int J Spray Combust 2020; 12: 1–22.

Tian

Liu

, et al. Experimental study on spray characteristics of octanol biodiesel and modification of spray tip penetration model. Phys Fluids 2021; 33: 091902.

Pucilowski

Jangi

, et al. LES study of diesel flame/wall interaction and mixing mechanisms at different wall distances. Proc Combust Inst 2021; 38: 5597–5604.

Ghiji

Goldsworthy

Brander

, et al. Numerical and experimental investigation of early stage diesel sprays. Fuel 2016; 175: 274–286.

Asgari

Amani

. A multi-objective CFD optimization of liquid fuel spray injection in dry low-emission gas turbine combustors. Appl Energ 2017; 203: 696–710.

Alajmi

Adam

, et al. Fuel atomization in gas turbines: a review of novel technology. Int J Energ Res 2019; 43: 3166–3181.

Passarelli

Kazbekov

, et al. Experimental study of forced synchronization and cross-coupling in a liquid-fuelled gas turbine combustor at elevated pressure. Proc Combust Inst 2023; 39: 4751–4759.

10.

Jing

Zhang

, et al. Experimental investigation on the macroscopic and microscopic spray characteristics of diesel like fuel. Fuel 2017; 199: 478–487.

11.

Murugan

Kohle

. Experimental investigation into flow-blurring atomization. Exp Therm Fluid Sci 2021; 120: 110240.

12.

Aradhey

Meadows

. Coupled droplet dynamics: investigation of acoustic-spray interactions in a reacting flow field. Combust Flame 2023; 250: 112645.

13.

Anand

CTN

Deshmukh

Ravikrishna

. Laser-based spatio-temporal characterization of port fuel injection (PFI) sprays. Int J Spray Combust 2010; 2: 125–150.

14.

Renaud

Ducruix

Zimmer

. Experimental study of the precessing vortex core impact on the liquid fuel spray in a gas turbine model combustor. J Eng Gas Turb Power 2019; 141: 111022.

15.

Manin

Bardi

Pickett

LM.

Evaluation of the liquid length via diffused back-illumination imaging in vaporizing diesel sprays. In: Proceedings of international symposium on diagnostics and modeling of combustion in internal combustion engines, Fukuoka, Japan, 23-26 July 2012, SP2-4, pp. 665–673. The Japan Society of Mechanical Engineers.

16.

Ghandhi

Heim

. An optimized optical system for backlit imaging. Rev Sci Instrum 2009; 80: 056105.

17.

Engine combustion network. https://ecn.sandia.gov/diesel-spray-combustion , 2024 (accessed 08 May 2024).

18.

Weiss

Wensing

, et al. Development of limited-view tomography for measurement of spray G plume direction and liquid volume fraction. Exp Fluids 2020; 61: 1–17.

19.

Westlye

Penney

, et al. Diffuse back-illumination setup for high temporally resolved extinction imaging. Appl Opt 2017; 56: 5028–5038.

20.

Lazzaro

. Ray-tracing analysis of shadowgraph and diffuse back-illumination extinction imaging of a synthetic GDI spray. Exp Fluids 2021; 62: 30.

21.

Wang

, et al. Potential of kerosene-diesel blends as alternative fuels for diesel engines: perspectives from spray combustion characteristics. Fuel 2023; 335: 127112.

22.

Godbold

Poursadegh

, et al. Measurement of air-fuel mixing in a diesel spray at engine relevant conditions using UV-VIS DBI diagnostic. Proceedings of the ASME 2021 internal combustion engine division fall technical conference. Online, 13-15 October, 2021, V001T03A006. ASME.

23.

Cardona

Payri

, et al. Effects of varying the liquid fuel type and air co-flow conditions on the microscopic spray characteristics in an atmospheric co-flow spray burner. Fuel 2023; 335: 127018.

24.

Manin

Pickett

Crua

Microscopic observation of miscible mixing in sprays at elevated temperatures and pressures. In: ILASS Americas 27th annual conference of liquid atomization and spray systems, Raleigh, North Carolina, 15–17 May, 2015, paper no. 93. ILASS.

25.

Schiemann

, et al. Experimental investigations of single particle and particle group combustion in a laminar flow reactor using simultaneous volumetric OH-LIF imaging and diffuse backlight-illumination. Renew Sust Energ Rev 2021; 136: 110337.

26.

Lieuwen

. Unsteady combustor physics. 2nd ed. Cambridge: Cambridge University Press, 2021.

27.

Murusegan

Singaravelu

, et al. Onset of flame-intrinsic thermoacoustic instabilities in partially premixed turbulent combustors. Int J Spray Combust 2018; 10: 171–184.

28.

Huang

Yang

. Dynamics and stability of lean-premixed swirl-stabilized combustion. Prog Energy Combust Sci 2009; 35: 293–364.

29.

Gejji

Huang

Anderson

. Parametric investigation of combustion instabilities in a single-element lean direct injection combustor. Int J Spray Combust 2018; 11: 1–16.

30.

Raghunathan

George

, et al. Inhibiting the onset of thermoacoustic instability through targeted control of critical regions. Int J Spray Combust 2023; 15: 3–15.

31.

Kraus

Harth

Bockhorn

. Experimental investigation of combustion instabilities in swirl-stabilized partially premixed flames in single- and multiple-burner setup. Int J Spray Combust 2016; 8: 4–26.

32.

Kim

, et al. Self-pulsation characteristics of a gas-liquid swirl coaxial injector. Atomiz Sprays 2009; 19: 57–74.

33.

Hassa

Heinze

Dreizler

. Self-excited oscillation in a combustion chamber driven by phase change in the liquid fuel feed system. Int J Spray Combust 2011; 3: 273–284.

34.

Johnson

James

Agrawal

. Effect of insert porosity on combustion instability in a lean premixed combustor analyzed by a proper orthogonal decomposition-based phase reconstruction technique. J Eng Gas Turb Power 2024; 146: 081006.

35.

Tedder

Tacina

, et al. Fundamental study of a single point lean direct injector. Part I: effect of air swirler angle and injector tip location on spray characteristics. In: 50th AIAA/ASME/SAE/ASEE joint propulsion conference, Cleveland, Ohio, 28–30 July, 2014, AIAA 2014-3435. AIAA.

36.

Johnson

James

Agrawal

. Analysis of spray atomization effects on the performance of porous media as disruptors of thermoacoustic instability. In: Proceedings of the 13th U.S. national combustion meeting, College Station, Texas, 19–22 March, 2023, 180NCT-0376.

37.

Berkooz

Holmes

Lumley

. The proper orthogonal decomposition in the analysis of turbulent flows. Annu Rev Fluid Mech 1993; 25: 539–575.

38.

Weiss

A tutorial on the proper orthogonal decomposition. In AIAA Aviation 2019 forum, Dallas, Texas, 17–21 June, 2019, AIAA 2019-3333. AIAA.

39.

Dasch

. One-dimensional tomography: a comparison of Abel, onion-peeling, and filtered backprojection methods. Appl Opt 1992; 31: 1146–1152.

40.

Kolhe

Agrawal

. Abel inversion of deflectometric data: comparison of accuracy and noise propagation of existing techniques. Appl Opt 2009; 48: 3894–3902.

41.

Siebers

. Liquid-phase fuel penetration in diesel sprays. SAE Transact 1998; 107: 1205–1227.

42.

Wanstall

Juune

, et al. A robust statistical algorithm for boundary detection in liquid sprays. Atomiz Sprays 2019; 29: 1071–1085.

43.

Lefebvre

McDonell

. Atomization and sprays. 2nd ed. Boca Raton: CRC Press, 2017.

44.

Gajan

Strzelecki

, et al. Investigation of spray behavior downstream of an aeroengine injector with acoustic excitation. J Propuls Power 2007; 23: 390–397.