## Abstract

The characterization of the two-phase kerosene/air flow near the nozzle of an aero-engine combustor is important in order to understand the combustion characteristics of the burner. Typically, particle image velocimetry (PIV) or laser Doppler velocimetry is used to measure velocities inside aero-engine combustors. However, these measurement techniques rely on tracer particles to visualize the flow field and are usually only able to measure the velocity field of one phase at a time. In the case of PIV measurements, both the flow tracers and the kerosene droplets scatter the laser light, and thus appear on the PIV recordings. Depending on droplet size and flow velocity, these kerosene droplets do not necessarily follow the airflow leading to errors in the derived velocity field. This work presents a method on how to separate kerosene droplets from flow tracers depending on their optical characteristics in the PIV recording. This phase separation enables the independent measurement of the flow fields of both the gaseous and liquid phase at the same time as well as the instantaneous slip velocity between droplets and gaseous flow using a standard PIV setup. The method is demonstrated on a laboratory scale aero-engine combustor operated at atmospheric conditions. The obtained results show that in the setup under investigation, gaseous and liquid phase can have significantly different flow fields with kerosene droplets moving in the opposite direction of the recirculating airflow.

## Motivation and Introduction

Particle image velocimetry (PIV) is a powerful tool to investigate flow fields in combustion systems and is commonly used to study gas turbine combustors. The measurement technique works by seeding the flow with tracer particles, which scatter the light of a laser sheet directed through the measurement plane. By recording the scattered light at different times, the particle displacement can be determined, which is assumed to be equal to the flow displacement. In gas turbine combustors with liquid fuel injection, as it is typical for aero-engine application, also the fuel droplets scatter the incident laser light. This can lead to several errors in the analysis, since these fuel droplets do not necessarily follow the gaseous flow and at the same time scatter more light than the tracer particles.

Generally, the ability of a droplet to follow the airflow is described by the Stokes Number St, which is defined as the ratio of the droplet relaxation time to the characteristic time scale of the flow. For values of $St<0.1$, it is generally assumed that the particle follows the airflow closely; however, for typical aero-engine application, only the smallest droplets will fulfill this criteria, and most droplets will not follow the airflow exactly [1].

In order to measure the velocity of the phases of a two-phase flow separately with PIV, the signals from each phase need to be separated and evaluated independently. For this signal or phase separation, there exist two general concepts.

The first concept is to separate the signals optically and to detect the signal of each phase with a different detector. Typically, this is achieved by substituting the particle tracers of one phase by fluorescing substances like fluorescein or rhodamine. In addition to scattering the laser light, these tracers also emit light at a different wavelength as the laser. By using two cameras and appropriate optical filters, the PIV signals of the phases can be separated. Application examples include the investigation of particle-laden turbulence [2] and air/water mixing [3]. There are also more complex techniques using multiple lasers [4] and multiple fluorescing tracers [5]. The drawback of all of these methods is the increased experimental complexity and limited applicability to combustion experiments, because of the low thermal stability of the fluorescing tracers.

The second concept is to record PIV data conventionally and to separate the phases depending on how they appear optically on the PIV pictures. This can be used to split the original pictures into ones containing only signal of one phase and consequently evaluate each phase using standard PIV and particle tracking velocimetry (PTV) methods. This phase separation can be performed by looking at the intensity of the scattered light and applying a threshold filter [6], by filtering the signal of small flow tracers [7] and/or using edge detection filters to detect one phase [8,9] or by separating the particles by the size they appear on the PIV recordings [10,11]. These approaches can be improved by using a combination of size and intensity [12,13] or a combination of intensity threshold and edge detection filters [14]. In contrast to methods already used for PIV in aero-engine combustors, which try to mitigate the effects of light scattered by fuel droplets by masking out regions with high intensity, the phase separation allows the independent analysis of gas and liquid phase flow fields.

Application of these techniques to combustion systems is not common. So far, there are investigations of the particle-laden flow in model coal combustors at nonreactive conditions [15] but no application for gas turbine combustion with liquid fuels. The application of these two-phase PIV methods to combustion systems may suffer from distortions by broadband flame luminosity, changes in seeding density due to temperature gradients, seeding depositions onto the combustor windows, and locally very high spray density.

The scope of this work is to apply the two-phase PIV phase separation methods and adapt them for the application in kerosene spray flames, to test their applicability and to study the two-phase flow characteristics close to the fuel nozzle in a laboratory scale aero-engine combustor.

## Experimental Setup

Experiments were carried out on an atmospheric, single sector combustion test rig, which is designed to be close to an aero-engine rich-quench-lean (RQL) combustor. The test rig is shown in Fig. 1. The test rig features realistic pressure drop and inlet temperatures as well as overall dimensions. The combustion chamber has optical access through quartz glass windows from all sides and only the area of the secondary air injection is accessible via two sides. The secondary air injection consists of five holes per side in two rows in a staggered configuration. There is no window purging or external cooling in order to mitigate effects of cold walls.

The fuel nozzle consists of a pilot fuel atomizer, injecting kerosene fuel (Jet A-1) directly into the combustion chamber, and six main fuel atomizers, which inject the fuel directly after the main air swirler, allowing for some evaporation and mixing of the fuel before entering the combustion chamber. Main and pilot air have counter-rotating swirl with swirl numbers of 0.7 and 1.34, respectively. The combustion air is preheated to 600 K using electrical heaters, and the pressure drop over the nozzle and secondary air injection is set to 3%. The air fuel equivalence ratio in the primary zone is set to $Φ=1.43$. Figure 1 also shows a typical flow field of the combustor measured via PIV [16]. The swirling flow creates a strong inner and outer recirculation zone and an unsymmetrical flow field due to the staggered mixing hole configuration.

The injection of pilot fuel directly into the combustor creates a spray that is strongly visible in the PIV recordings. The velocity fields of the setup under consideration were measured with PIV for several operating conditions [16], where the general sensitivity of the flame position to the pilot fuel spray injection was revealed. The penetration of the pilot spray was assessed by comparing recordings of the spray to PIV results recorded without pilot fuel injection, and depending on the operating conditions, fuel droplets of the pilot atomizer have been observed traveling against the flow direction of the inner recirculation zone. However, switching off the pilot fuel injector changes the overall velocity field as well as the location of the flame, so this might not be a valid comparison. Consequently, this study focusses on the region surrounding the pilot fuel injection and its interaction with the combustion air and the recirculated exhaust gases.

For certain operating conditions, two different flames types, an attached and a lifted flame, can be stabilized for the same boundary conditions. Figure 2 shows OH* chemiluminescence line-of-sight recordings of the corresponding flame types. In order to address resulting differences in the mean flow field far from the nozzle, both flame types were examined in this study.

## Particle Image Velocimetry

The experimental setup consists of a dual cavity high-speed Nd:YLF laser (Litron LDY304, Warwickshire, UK) and a high-speed camera (Phantom High Speed Star, 1024 × 1024 px resolution) with a recording frequency of 1.5 kHz double frames. Time separation between laser pulses was 10 μs, and 1024 double frame pictures were taken per recording. The flow was seeded with aluminum oxide seeding using fluidized bed seeders. In this study, the field of view was adjusted directly over the nozzle exit, and only the primary airflow was seeded. Figure 3 shows the field of view of the PIV measurements.

During operation, there is fast seeding deposition onto the combustor windows, which need to be cleaned after every recording. The fuel-rich primary zone leads to strong broadband light emissions from the soot particles, which can interfere with the PIV recordings. The bigger fuel droplets scatter the laser light considerably stronger than the tracer particles, so that careful adjustment of the laser power was necessary in order to avoid camera saturation and to ensure, at the same time, the visibility of the tracer particles.

Phase separation was done using a self-developed matlab program, while vector calculation and postprocessing were performed using a commercial piv software (davis 8, lavision). Details on the vector calculation parameters will be given in the Phase Separation section.

## Phase Separation

The phase separation method in this work is adapted from previous work from Khalitov and Longmire [12] and Becker et al. [15] with differences in the particle detection to cope with varying background light intensities and seeding densities in the reacting environment.

The method starts with conventionally recorded, background-subtracted PIV pictures. Figure 4(a) shows a representative section of a two-phase PIV picture. Some of the particles can clearly be identified as big fuel droplets due to their size and high peak intensities. Other smaller particles with varying size are also present and can be either fuel droplets or tracer particles. In a first step, the pictures are filtered using a two-dimensional Wiener filter. This adaptive noise-removal filter is used to suppress most of the signal from the tracer particles, as they have similar properties as random image noise. This is done in order to suppress unwanted results in the following edge detection steps. The size and intensity of the filter is set so that all clearly visible particles remain, and the result of the filtering is depicted in Fig. 4(b).

Following an edge detection method is used, based on maximum gradients using the Sobel approximation. The edge detection should be tuned so that most particles, which are still visible after the noise reduction filtering, are detected. This may be fuel droplets and tracer particles. Figure 4(c) shows the results of the edge detection. Using an edge detection filter instead of an intensity threshold offers the advantage that detection works better with uneven background illumination and varying seeding densities. This is especially useful in reacting conditions, as broadband light emission from the flame and strong laser scattering at the fuel droplets lead to increased background intensities, which cannot be easily removed. In the current setup, additional seeding deposition onto the combustor windows leads to different background light intensities throughout one measurement.

After the edge detection, the resulting image mask is dilated, and the pixels inside one masked region are filled. Every set of connected pixels is labeled as one particle, and the size and average intensity of every detected particle are calculated. Depending on the size and average intensity, each particle is then classified as liquid phase or tracer particle, and a mask is created containing all liquid phase particles. The exact values of these cut off parameters depend on many different factors and need to be chosen for every dataset individually. This is done by applying the particle detection method to PIV pictures containing only tracer particles and comparing the size and intensity of the detected particles to results derived from two-phase pictures. In Fig. 4(d), a plot of particle size over average intensity for a PIV picture containing only tracer particles and for a two-phase flow PIV picture is shown. The lower plot shows the evaluation for the tracer-only picture, and apart from a few stray points, most tracer particles are smaller than 25 px and have an average intensity lower than 120. The plot on top shows the same analysis for a picture containing both tracers and fuel droplets, and many particles with higher average intensity and size are now present. The particles with a higher average intensity of 120 or with a size bigger than 25 px are classified as duel droplets.

After assigning all particles to a phase, separated PIV pictures can be build. Figures 4(e) and 4(f) show the resulting picture containing only tracers and fuel droplets. For the tracer picture, the signal of the detected fuel droplets is removed, and the rest of the picture is unchanged.

For vector calculation of the gas phase velocities, the generated mask is used to mask out all liquid phase particles. A sequential cross correlation is used, with window sizes starting at 96 × 96 px and ending at 48 × 48 px with 50% overlap. The chosen window size resulted in a spatial resolution of 1.5 mm. Once a correlation window consists of more than 30% masked out pixels, no vector is calculated. This is important as cross correlation at the edge of detected liquid phase particles can lead to increased amount of false vectors. The results were postprocessed by deleting vectors with a peak ratio smaller than 1.4 and then median filtered using the “universal outlier detection”-method integrated into the piv software. For the calculation of the liquid phase velocities, the particle mask is dilated because the calculation is very sensitive to partially cut out particles. Single tracer particles might be unmasked by this step, but the effect of this is negligible because of their small numbers compared to the bigger particles and the bias of the cross-correlation method to more intense signals. Velocities are calculated by tracking single particles using the PTV algorithms integrated into the piv software.

In order to reduce errors of the two-phase PIV evaluation, the following points are important: the seeding quality needs to be good and the density sufficiently high. Since some fuel droplets scatter light similarly to the tracer particles, they cannot reliably be discriminated from tracer particles. As the fuel droplets are smaller in number compared to the tracer particles, a high seeding density helps to mitigate this as long as the correlation windows are big enough to contain several tracer particles. Regions with high fuel droplet density are generally harder to evaluate. If droplet density is too high, no gaseous flow vector can be calculated at that position, since there are not enough tracer particles left for a reliable calculation. The corresponding regions should be masked out completely, since the PIV cross correlation might calculate false vectors at these positions. A camera with higher resolution and a higher dynamic range is generally better in order to record both tracer particles and fuel droplets with sufficient accuracy.

## Calibration and Validation

The particle separation method requires many parameters to be tuned manually by the person performing the data evaluation, and no universal set of parameters could be identified so far. This means that the phase separation needs to be calibrated and validated for every application case. A reliable way to do so is to use artificial two-phase PIV pictures with the same optical properties as the examined case [12,15].

In this work, this is done by superimposing pictures of the liquid phase without tracer particles onto pictures, which were recorded when pilot fuel injection was switched off and only tracer particles were visible. These artificial PIV pictures do not represent any physical flow but do have similar optical properties to the two-phase pictures in terms of tracer particle and liquid droplet light scattering and also in terms of background disturbances and image noise since all pictures were recorded in reacting flows. The idea behind these artificial two-phase pictures is that the original single-phase pictures can be evaluated using conventional PIV/PTV methods, serving as a baseline reference to evaluate the correct set of parameters for the phase separation and to quantify the errors introduced by the two-phase evaluation.

Figure 5 shows the results of the validation process. In Figs. 5(a) and 5(b), the PIV and PTV results of the original single-phase recordings are shown. Figures 5(c) and 5(d) show the results after the single-phase pictures have been combined to one artificial two-phase picture, separated again with the method described in the Phase Separation section, and evaluated using PIV and PTV. The overall flow structure was successfully recovered. In the case of the gaseous flow, differences in the streamlines are visible close to the pilot inlet, which is expected since spray density is very high in this region.

In the original single-phase recordings, 67% of the calculated vectors were validated in the postprocessing, and for 1% of the positions, no vector could be calculated. Alternative vectors derived from other local correlation maxima replaced the remaining positions. In the case of the two-phase analysis, only 61% of the vectors were validated directly, and in 8% of the positions, no vector could be calculated at all. In Fig. 6, the number of vectors used for the calculation of the mean gas-phase velocity vector is shown for the original single-phase results (Fig. 6(a)) and after the two-phase evaluation (Fig. 6(b)). The number of vectors drops significantly in the region of the spray.

In the case of the liquid fuel spray, the data rate in the two-phase pictures is also lower compared to the single-phase pictures, since most of the weaker signals (from either smaller droplets or droplets only partly inside the laser sheet) are outshined by the tracer particles. This results in some wrinkled streamlines and lower spread compared to the original solution. Both gaseous and liquid phase average vectors were only calculated if 10% of all snapshots at a given position contained a vector.

In Figs. 5(e) and 5(f), the difference in velocity magnitude between the original results and the results after two-phase separation is shown. For the gaseous phase, increased errors occur in regions with high spray density, as would be expected. Errors can reach up to 10 m/s but are mostly below 5 m/s (10% of maximum velocity). For the spray velocities, errors are also around 5 m/s except on the edges of the spray where errors can be significantly higher. Raising the threshold of 10% for mean vector calculations can be used to mask out areas with higher errors but leads to larger holes in the derived mean velocity fields.

Overall, the reconstructed velocity fields represent the original mean flow structure well and the technique, and the selected evaluation parameters can be applied to the real two-phase flow in the combustion chamber.

## Results

Results for the reacting two-phase flow in the RQL combustor were obtained at 600 K air inlet temperature and an equivalence ratio in the primary zone of 1.43. Both main and pilot fuel injection were active. As reported before, this setup can stabilize both a lifted and an attached flame for this operating condition, and the results of both flame types are presented in Fig. 7. Both cases were evaluated using the same parameters.

In Figs. 7(a) and 7(b), the gas-phase velocity fields are plotted. The axial velocity inside the inner recirculation zone is higher for the lifted flame than for the attached flame. This was previously also observed in the flow field further away from the nozzle and is linked to the opening of the flow cone due to the change in flame position, which, in turn, lowers the recirculation velocity. For both cases, the inner recirculation zone extends all the way upstream up to the pilot fuel injection.

The velocity field of the gaseous phase of the lifted flame is very close to the single-phase result from Fig. 5(a). This was expected, since the single-phase pictures were recorded by turning off the pilot fuel injection and thus causing the flame to lift off in a very similar way than the lifted flame with pilot fuel injection. The velocity field of the attached flame, on the other hand, is different to the single-phase result, with lower axial velocities in the inner recirculation zone. Another notable difference is in the streamlines and velocity magnitude of the pilot air visible in the bottom of the graph. In case of the attached flame, higher velocities and further penetration of the air into the region of the fuel spray are observed. This is favorable for the ignition of the fuel spray. Figures 7(c) and 7(d) show the corresponding fuel spray velocities. Streamlines are similar for both cases, and velocity magnitude is generally higher for the attached flame. In both cases, the spray extends into the inner recirculation zone. The faster deceleration of the spray in the lifted flame can be attributed to the interaction with the gaseous phase and the higher negative axial velocities in the inner recirculation zone. The penetration of the fuel spray into the recirculation zone is an important observation, since in this RQL setup, the fuel rich primary zone leads to very low oxygen levels in the recirculated exhaust gases, and fuel injected here has a different effect on the combustion characteristics of the setup as fuel injected into the areas containing fresh air.

The slip velocity between droplet and gaseous phase has a big influence on the heat and mass transfer between the phases and is important for the correct modeling of the spray vaporization. One unique property of the two-phase PIV technique is that this slip velocity can be calculated for every instantaneous snapshot. In Figs. 7(e) and 7(f), the average slip velocities between fuel droplets and gas phase are shown. Since the average slip velocities are not necessarily the same as the difference between the average velocities of the separated phases [15], the slip was calculated using the instantaneous snapshots. This was done by subtracting the velocity of the fluid phase particles from the closest corresponding gas-phase vector. If no gas-phase vector was calculated (e.g., in regions of very dense spray), no slip velocity was calculated. The results show that there is considerable slip between the two phases with slip velocities of up to 50 m/s. The slip velocities are higher for the attached flame. The increase in spray velocity for the attached flame has a bigger effect on the slip velocity than the decrease in axial velocity in the inner recirculation zone.

In Fig. 8(a), time series of PIV snapshots is shown. Background color and streamlines represent the gas phase, and the white arrows represent the velocity vectors of the fuel spray. The time series starts at a snapshot with a strong recirculation-event with high axial velocities in the inner recirculation zone. Over the series, the flame transitions from the lifted to the attached flame, which results in slower axial velocities in the inner recirculation zone, as observed in the time averaged results. The snapshot at 0 ms, with the high velocities inside the inner recirculation zone, is very typical for the lifted flame, where this behavior is observed in many instances, whereas this is not the case anymore after the flame anchors at the burner with the snapshot at 4.67 ms being representative for the attached flame.

In Fig. 9, the instantaneous velocity magnitude of the fluid phase at the position x = 0.3r0 and y = 0.5r0 is shown. The time zero marks the onset of the flame change and is the same as in Fig. 8. With the flame change, the velocity in the spray increases, as was also observed in the averaged results. Additionally, the spray is not fluctuating as strongly as before. It seems as the strong recirculation-events in the lifted flame drive the velocity oscillations in the spray.

## Limitations

The method presented in this work relies on differences in the scattered light of the investigated phases. As such, small droplets with similar scattering properties as the tracer particles cannot be detected. This means that in regions of the spray with very small droplets, no liquid phase velocity can be calculated, and as no size measurements using the PIV setup are possible, the smallest measurable droplet size cannot be readily determined. This might also limit the application to test cases with very fine spray (e.g., high-pressure applications).

We have found that the successful gas-phase vector calculation typically requires larger inter-rogation windows as for comparable single-phase PIV, leading to a loss in spatial resolution. The error in the phase separation depends strongly on the amount of large particles (in this work, kerosene droplets) and the quality of the seeding of tracer particles. Preliminary tests in nonreacting environments and carefully adjusted density of large particles showed that much smaller errors as presented in this work are possible. However, in practical application, these parameters cannot be set freely and so the resulting errors cannot be easily influenced.

Finally, the presented validation procedure to determine the errors of the method works well for the average values, but there is currently no way of determining the errors of the single snapshots. Comparison of the artificial two-phase PIV snapshots to the original single-phase snapshots show that some are reconstructed almost perfectly, whereas others suffer from increased errors, and so far there is no way of identifying these snapshots in the actual two-phase recordings.

## Conclusions

In this work, we presented the application of two-phase PIV to a laboratory scale jet-engine combustion chamber and showed that the method can successfully be applied to recover both the airflow and fuel droplet velocity field using a standard PIV system with no additional hardware. The method was validated using artificial two-phase PIV pictures. The method was able to recover the original flow structures very well, and errors in velocity magnitude were quantified. Even though the conditions for applying this method are more adverse compared to previous studies (reacting flow with strong soot broadband emissions, rapid seeding decomposition onto combustor walls, and reduced spatial and dynamic resolution of the high-speed setup), the method is still working satisfactorily. Application to the real two-phase flow in the combustion chamber revealed that the inner recirculation zone extents all the way upstream up to the pilot fuel injector for both investigated flame types and that fuel droplets of the pilot fuel injector travel against the flow of the inner recirculation zone. The change in flame type affected both the velocity field of the gaseous phase as well as the pilot spray. The derived air velocity fields differed from the results using single-phase images, showing the necessity to investigate the two-phase flow directly. The measurement technique allowed the calculation of slip velocities between liquid and gaseous phase, and considerable slip between the phases was detected.

## Funding Data

• The authors gratefully acknowledge funding from the European Union within the SOPRANO H2020 project under Grant Agreement No. 690724 (Funder ID: 10.13039/501100007601).

## References

1.
,
R.
, and
Lenze
,
B.
,
2005
, “
Measurements of Droplets Characteristics in a Swirl-Stabilized Spray Flame
,”
Exp. Therm. Fluid Sci.
,
30
(
2
), pp.
117
130
.10.1016/j.expthermflusci.2005.05.002
2.
Poelma
,
C.
,
Westerweel
,
J.
, and
Ooms
,
G.
,
2006
, “
Turbulence Statistics From Optical Whole-Field Measurements in Particle-Laden Turbulence
,”
Exp. Fluids
,
40
(
3
), pp.
347
363
.10.1007/s00348-005-0072-y
3.
Zamankhan
,
P.
,
2010
, “
Large Eddy Simulation and Piv Experiments of Air–Water Mixing Tanks
,”
Commun. Nonlinear Sci. Numer. Simul.
,
15
(
6
), pp.
1511
1525
.10.1016/j.cnsns.2009.06.021
4.
Towers
,
D. P.
,
Towers
,
C. E.
,
Buckberry
,
C. H.
, and
Reeves
,
M.
,
1999
, “
A Colour PIV System Employing Fluorescent Particles for Two-Phase Flow Measurements
,”
Meas. Sci. Technol.
,
10
(
9
), pp.
824
830
.10.1088/0957-0233/10/9/309
5.
Tóth
,
B.
,
Anthoine
,
J.
, and
Riethmuller
,
M. L.
,
2009
, “
Two-Phase PIV Method Using Two Excitation and Two Emission Spectra
,”
Exp. Fluids
,
47
(
3
), pp.
475
487
.10.1007/s00348-009-0674-x
6.
Sakakibara
,
J.
,
Wicker
,
R.
, and
Eaton
,
J.
,
1996
, “
Measurements of the Particle-Fluid Velocity Correlation and the Extra Dissipation in a Round Jet
,”
Int. J. Multiphase Flow
,
22
(
5
), pp.
863
881
.10.1016/0301-9322(96)00014-6
7.
Kiger
,
K. T.
, and
Pan
,
C.
,
2000
, “
PIV Technique for the Simultaneous Measurement of Dilute Two-Phase Flows
,”
ASME J. Fluids Eng.
,
122
(
4
), pp.
811
818
.10.1115/1.1314864
8.
Shi
,
B.
,
Jia Wei
,
J.
, and
Zhang
,
Y.
,
2015
, “
Phase Discrimination and a High Accuracy Algorithm for PIV Image Processing of Particle–Fluid Two-Phase Flow Inside High-Speed Rotating Centrifugal Slurry Pump
,”
Flow Meas. Instrumen.
,
45
, pp.
93
104
.10.1016/j.flowmeasinst.2015.05.002
9.
Pang
,
M.
, and
Wei
,
J.
,
2013
, “
Experimental Investigation on the Turbulence Channel Flow Laden With Small Bubbles by PIV
,”
Chem. Eng. Sci.
,
94
, pp.
302
315
.10.1016/j.ces.2013.02.062
10.
Jakobsen
,
M. L.
,
Easson
,
W. J.
,
Greated
,
C. A.
, and
Glass
,
D. H.
,
1996
, “
Particle Image Velocimetry: Simultaneous Two-Phase Flow Measurements
,”
Meas. Sci. Technol.
,
7
(
9
), pp.
1270
1280
.10.1088/0957-0233/7/9/012
11.
Gui
,
L.
,
Lindken
,
R.
, and
Merzkirch
,
W.
,
Lain
,
S.
,
Bröder
,
D.
, and
Sommerfeld
,
M.
,
1997
, “
Phase-Separated PIV Measurements of the Flow Around Systems of Bubbles Rising in Water
,”
ASME
Paper No. FEDSM97-3103
12.
Khalitov
,
D. A.
, and
Longmire
,
E. K.
,
2002
, “
Simultaneous Two-Phase PIV by Two-Parameter Phase Discrimination
,”
Exp. Fluids
,
32
(
2
), pp.
252
268
.10.1007/s003480100356
13.
Cheng
,
Y.
,
Pothos
,
S.
, and
Diez
,
F. J.
,
2010
, “
Phase Discrimination Method for Simultaneous Two-Phase Separation in Time-Resolved Stereo PIV Measurements
,”
Exp. Fluids
,
49
(
6
), pp.
1375
1391
.10.1007/s00348-010-0878-0
14.
Muste
,
M.
,
Yu
,
K.
,
Fujita
,
I.
, and
Ettema
,
R.
,
2009
, “
Two-Phase Flow Insights Into Open-Channel Flows With Suspended Particles of Different Densities
,”
Environ. Fluid Mech.
,
9
(
2
), pp.
161
186
.10.1007/s10652-008-9102-7
15.
Becker
,
L. G.
,
von Langenthal
,
T.
,
Pielsticker
,
S.
,
Böhm
,
B.
,
Kneer
,
R.
, and
Dreizler
,
A.
,
2019
, “
Experimental Investigation of Particle-Laden Flows in an Oxy-Coal Combustion Chamber for Non-Reacting Conditions
,”
Fuel
,
235
, pp.
753
762
.10.1016/j.fuel.2018.08.076
16.
von Langenthal
,
T.
,
Zarzalis
,
N.
, and
Konle
,
M.
,
2019
, “
Experimental and Numerical Investigation of Different Flame Types Inside a Laboratory Scale RQL Combustion Chamber
,”
ASME
Paper No. GT2019-90127.10.1115/GT2019-90127