Paper # 070IC-0307 Topic: Internal Combustion and Gas Turbine Engines

8th U. S. National Combustion Meeting

Organized by the Western States Section of the Combustion Institute

and hosted by the University of Utah

May 19-22, 2013

Combustion Instability Interaction with Fuel-Air Mixing in a

Partially-Premixed Gas Turbine Model Combustor

Patton M. Allison David A. Rosenberg James F. Driscoll

Department of Aerospace Engineering,

University of Michigan, Ann Arbor, MI 48109, USAA

Fluid mechanical coupling between injected reactants and the flame-base is observed in a

gas turbine model combustor. A natural combustion instability introduces oscillations in fuel

and air mass flow rates which affects mixing, stabilization, instability amplitude, and

anchoring. This interaction has been studied for varying fueling conditions in methane and

propane flames. Redistributions of air flow rates between the individual swirlers strongly

affects instability strength and shear layer mixing. Laser Doppler velocimetry (LDV)

measurements of radial velocities in the various flame show velocity fluctuations which

correspond to flame motion oscillations captured by high speed chemiluminescence imaging.

Acetone has been introduced as a fuel tracer, for use in planar laser-induced fluorescence

(PLIF) measurements. Acetone PLIF of the fuel and high speed chemiluminescence imaging

reveals that coupling between the flame position and shape occurs. Flame shape has been

correlated with instability strength and presence. Rayleigh indices have been compared to

fuel-air distributions to understand how mixing affects instability strength.

1. Introduction Flame stabilization and anchoring, in gas turbine burners, is achieved with swirling flow which leads to the

creation of a precessing vortex core (PVC) and enhanced mixing. The addition of swirl to a combustor has been

shown to allow stable operation over a range of conditions and regimes[1]. However, they are also subject to

thermoacoustic oscillations that can lead to acoustic instabilities. These instabilities arise when there is proper

coupling between the pressure field and heat release field to satisfy the Rayleigh criterion[2], such that they are in

phase. These pressure fluctuations may also induce fluctuations in the oxidizer and fuel flow rates such that

equivalence ratio oscillations occur[3]. These equivalence ratio oscillations continue to feed the flame causing

cyclical variations in heat release and pressure. Figure 1 demonstrates some of the mechanisms involved in

inducing instabilities in swirl flames. This paper focuses on a dual-swirl burner that undergoes a natural

combustion instability at a frequency in the range between 250 to 480 Hz, dependent on fueling properties. The naturally occurring combustion instability has previously been shown to be the product a Helmholtz and a

convective-acoustic mechanism which also displays a flame speed dependence[4]. This instability introduces

oscillations in fuel and air mass flow rates which affects mixing, stabilization, instability strength, flame shape,

and anchoring. These combustion parameters have been investigated by studying the interaction between the

flame and fuel-air mixing for varying fueling conditions in methane and propane flames.

The swirl burner, known as the gas turbine model combustor (GTMC), was developed by Meier and colleagues at

DLR Stuttgart. It has the advantage that it is of canonical axisymmetric swirler-design yet it exhibits the

fundamental physics associated with gas turbine flames; it contains two swirling air streams which surround an

annular fuel stream. Comprehensive measurements have been conducted in the burner-facility by Meier and

coworkers[5,6] to investigate flame-structure, flame-dynamics, and flow-field structures for a few selected

2

methane-fueled conditions. The goal of this paper is to investigate the effect of each of the components in this

feedback system, which feed into the flame at the flame base, as depicted by the larger black box in Fig. 1.

These components will be evaluated by studying fuel-air distributions via PLIF measurements, high speed

chemiluminescence videos of flame motion and shape, and LDV velocity measurements of swirler flow rate

distributions. LDV measurements of radial velocities in the various flames show velocity fluctuations which correspond to flame motion oscillations captured by high speed chemiluminescence imaging. Fuel-air mixing has

been investigated through PLIF measurements, where acetone has been introduced as a fuel tracer. For each fuel

type, the spatial distribution of fuel and air in the GTMC has been shown to significantly differ. Flame liftoff

height oscillations can be induced by variations in the mixing as well as air flow rate distribution. Flame shape has

been correlated with instability strength and presence. Rayleigh indices have been compared to fuel-air

distributions to understand how mixing affects instability strength. When air flow is redistributed to the outer

swirler, larger radial velocities are recorded. Acetone PLIF images depict low-lying regions of fuel extending

towards the outer edge of the burner. This mixing pattern encourages a flat flame shape which is unanchored and

oscillates up and down. Rayleigh indices of flat flames indicate stronger acoustic coupling and amplification as

compared to cases where the flame may be V-shaped and/or anchored. The latter flames correspond to flow

conditions where a lower radial air velocity is observed and the flame is non-resonant.

.

2. Methods

A gas turbine model combustor (GTMC) as seen in Fig. 2[5] is used to study acoustic behavior and performance. The injector consists of a central air nozzle, an annular fuel nozzle, and a co-annular air nozzle. Both air nozzles supply

swirling dry air at atmospheric pressure and temperature from a common plenum. The inner air nozzle has an outer

diameter of 15 mm and the annular nozzle has an inner diameter of 17 mm and an outer diameter of 25 mm. The

measured swirl number is approximately 0.55. Non-swirling fuel is provided through three exterior ports fed through the

annular nozzle which is subdivided into 72 0.5 mm x 0.5 mm channels. The width of the fuel annulus is less than 0.5

mm. The exit plane of the central air nozzle and fuel nozzle lies 4.5 mm below the exit plane of the outer air annulus.

The exit plane of the outer air annulus will be referred to as the burner face. The combustion chamber has a square cross

section of 85mm in width and 110 mm in height. The exit of the burner has a tapered lid which leads to an exhaust

chimney with a diameter of 40 mm and a height of 50 mm. The burner is operated with 4 fused silica windows, with a

thickness of 1.5 mm, for flame visualization. The burner was fired using methane and propane fuels. An external

cylindrical chamber was used for fuel mixing and fuel line separation. This chamber allowed for the equal division of the fuel flow into three separate lines which lead to the three fuel ports on the burner.

Mass flow rates for the air and individual fuel lines leading to the burner were controlled by sonically choked

orifices. Table 1 lists the operating parameters investigated with regards to the LDV and chemiluminescence

measurements. Of note is the fueling case primarily studied by Weigand et al. [6] identified as Case B. This methane

case has an air mass flow rate of 282 g/min and an equivalence ratio of 0.75. The fueling conditions for Case B have

been the standard for which these measurements have been based on. The air mass flow rate and equivalence ratio for

Figure 1. Proposed system of thermoacoustic relationships.

3

various fuels have been varied in a range around the Case B operating condition. A PCB piezoelectric transducer was

mounted in the plenum, which has a diameter of 80 mm and a length of 72 mm, and in the chamber, although not

simultaneously. For a given operating condition, pressure measurements are taken at a set burner reference temperature.

The frequency and amplitude in the plenum was previously shown to be not significantly different from that in the

combustion chamber, with the exception of a phase shift[7, 8]. The reported frequencies are accurate to ±0.5%. Pressure

measurements were collected at 25 kHz. High speed videos of the total line of sight flame chemiluminescence were taken using a Phantom V7.1 high speed camera. Videos were taken at a frame rate of 2500 fps with an exposure of 150μs

at 768 x 768 pixels. The resolution of each of the images is 114 μm/pixel.

Velocity measurements were taken using a TSI LDV system using a Ar-ion laser at 514.5 nm with data rates well

above 1 kHz. The radial velocity of the flow was measured at a single point located a distance of 5mm above the burner

face and 20mm from the axial centerline. A single measured location was deemed sufficient to investigate the presence

of a swirler switching effect. Duan et al. [9] previously showed that the radial velocity at this location oscillates over a

range of 6 m/s over the course of a single acoustic cycle through phase resolved measurements. The axial velocity at this

location undergoes a much smaller range of fluctuation.

Figure 2. GTMC schematic drawing depicting air and fuel flow paths, as well as swirl vane locations. The red dotted

box indicates the field of view of the acetone PLIF measurements. [5]

Table 1. Fueling parameters investigated in LDV and chemiluminescence studies

Fuel ma, g/min ΦGlobal

Methane 140-282 0.5 – 1.2

Propane 140-282 0.55 – 1.4

The acetone PLIF measurements were taken as part of a separate ongoing study to investigate flame index in partially premixed flames[10, 11]. This study involved the use of tracking fuel and air flow with tracers, specifically acetone for

the fuel and NO2 for the oxidizer. This investigation uses the acetone PLIF results from that study. Three fueling cases

for methane and propane were studied. The methane case, M-1, matches the fueling conditions of Weigand’s Case B.

The propane cases, P-1 and P-2, from the flame index study are given in Table 2. P-1 attempts to match similar fueling

conditions to Case B, except with propane rather than methane. The acetone seeding in the propane was achieved by

bubbling the propane through an acetone bath. The bath had a bypass line, through which the flow of propane could be

controlled. Heat tape, connected to a PID temperature controller, kept the acetone in the bath at a constant temperature

±2oC to more precisely control the acetone concentration.

4

Table 2. Fueling parameters investigated in acetone PLIF measurements

As suggested by prior studies [12–14] and a previous flame modeling study [11], acetone makes an excellent fuel tracer,

which fluoresces from 400 nm to 500 nm when excited by a 266 nm laser. A diagram of the layout of the lasers,

cameras, and burner for the simultaneous acetone and NO2 PLIF system can be seen in Fig. 3. The acetone PLIF was

achieved with a frequency quadrupled Nd:YAG laser (Spectra Physics GCR-130), Laser #2, at a wavelength of 266 nm.

The 266 nm laser was run at 30 mJ/pulse. The laser had a pulse length of 6 ns, and pulsed at a rate of 10 Hz.

The laser beams were formed into sheets and focused using one or more cylindrical lenses, and then passed through

10% pick-off mirrors to a dye cell with an optically thick Rhodamine 6G solution to correct for the non-uniformity of the

laser sheet. The distances were set up such that the laser sheets would be focused inside the dye cell. The remaining

90% of the laser sheets were then combined using a dichroic beam splitter (CVI Melles Griot BSR-25-2025), and passed through the burner. At the test section, the 266 nm laser sheet had a height of 10 mm, and a 1/e thickness of

250 µm at its focal point over the back edge (farther from the lasers) of the burner’s inner tube. A knife edge was used

to create a sharp edge in the observed PLIF signals and the fluorescence observed in the dye cell, so that the two images

could later be aligned, as suggested by Clemens [15].

Figure 3. Diagram of laser and camera setup for simultaneous PLIF. BD - beam dump, BS – beam splitter, CL -

cylindrical lens, K - knife edge, M - mirror, ND - neutral density filter, PM - partial mirror. The acetone fluorescence was also observed by an ICCD camera (Andor iStar DH734-25F-03), ICCD #2, with an

interference filter (Omega Optical 500ASP) to allow light at wavelengths of 400 nm to 500 nm to pass. The camera

was fitted with a 105 mm f/2.8D Micro-Nikkor lens. A CCD (Sony XCD-X710) imaged the dye cell, and was fitted

with a 50 mm f/2.8 Nikkor lens. A neutral density (ND) filter, with an optical density of 2, was placed to cover only the

half of the dye cell that the 532 nm laser sheet hit. A second ND filter, with an optical density of 1, was placed to

cover the entire image. The acetone camera was gated to turn on 50 ns before the arrival of the 266 nm laser pulse.

The gate width was set to 100 ns. A total of 450 images was taken or each condition.

5

The ICCD had a pixel array of 1024 × 1024 pixels. The camera was re-focused each day to ensure the best possible

images. The resolution of the camera was 21.7 ± 3.3 µm/pixel. After each day of runs, a series of background images

was taken.

3. Results and Discussion It has been observed in the GTMC that air flow redistribution can occur between the two independent swirlers,

which are fed from a common plenum. Uneven partitioning of air flow between the individual swirlers strongly

affects instability strength and shear layer mixing. Dynamic swirler flow rate variations can result in large redistributions of heat release which can affect flame shape and the pressure field. LDV measurements of radial

velocities in the various flames show velocity fluctuations which correspond to flame motion oscillations captured

by high speed chemiluminescence imaging.

a. LDV Results:

Figure 4 depicts the variation in the radial velocity at the LDV measurement location. The cold flow case represents non-

reacting measurement with only air flowing. The radial velocity increases linearly with increasing flow rate, as expected.

For this non-reacting measurement, a set distribution of air flow rate between the inner and outer swirler has been established and remains unchanging. The reacting methane case shows interesting behavior. The first data at 160 g/min

represents a flame that is non-resonating. The flame is lifted and has a V-shape. At higher flow rates, the methane flame

liftoff height is reduced, the shape is flatter, and the flame begins to resonate at 308 Hz. For increasing air flow, the fluid

velocity remains almost unchanged. For all the propane cases studied, the flames are resonant and flat in shape. The

velocity appears to be independent of airflow rate. Of note is the increase in radial velocity observed between the non-

reacting and reacting flows. This is suggestive of a redistribution of flow to the outer swirler, which imparts a greater

radial flow component.

Figure 4. Measured radial velocities with varying air mass flow rate. Reacting cases measured at an equivalence

ratio of 0.75.

Figure 5 represents radial velocity histograms for an interesting air flow set point unique to the GTMC. As discussed

previously, the distribution to each swirler is naturally set by the system and cannot be individually controlled. The cold

flow case is Fig 4, generally represents a 60/40 split with the outer swirler receiving most of the flow. With the

combustion chamber removed from the system, such that the burner injector is no longer encased, two unique flow

patterns can be establish for this open air, non-reacting set-up. If a slight back-pressure is applied to the burner face with

a bluff object, such as one’s hand, the flow distribution can be set such that the majority of the air flow is directed to the

outer swirler.

Interestingly enough, when the blockage or bluff object is removed, the flow pattern through the outer swirler will

remain. This results in the open air radial mode as depicted by the red line histogram in Fig 5, with a mean velocity of

150 200 250 300 350 4001

2

3

4

5

6

7

Air Mass Flow Rate [g/min]

Velo

city [

m/s

]

Cold Flow

Methane

Propane

6

3.1 m/s. However, if a back pressure is reapplied, the flow pattern may switch such that the majority of the flow is

redirected to the inner swirler. Again, this flow pattern is maintained when the back pressure is removed. The inner

swirler has no contoured shape and thus the air flow is generally directed down the axial direction. This results in the

open air axial mode as depicted by the blue line histogram in Fig 5, with a mean velocity of 2.05 m/s.

This unique set of flow patterns is interesting because it establishes the precedent that higher radial velocities are correlated with flow to the outer swirler and that changes in back pressure to the burner face can cause this distribution

of flow to “switch” between the self described radial and axial modes. The presence of a reacting flow also tends to force

more flow to the outer swirler as observed with the higher velocities seen in Fig 4. It is hypothesized that the flame and

recirculation zone created by the swirling flow can affect the back pressure on the injector face which will feedback to

the airflow distribution and alter the flame and recirculation bubble. Thus conditions correlated with onset of resonating

flames are established with the initiation of a particular flow distribution through the swirlers.

Figure 5. Open air radial mode (red). Open air axial mode (blue). Air mass flow rate = 282 g/min, non-reacting.

Figure 6 depicts the variation of radial velocity with varying equivalence ratio in methane flames. The behavior shows a

non-monotonic trend with the measured velocity peaking at an equivalence ratio of 0.75. A similar trend is observed in

previous studies [4] of acoustic trends versus equivalence ratio. The frequency behavior of resonating methane flames is

also non-monotonic with increased fueling which suggests a correlation of higher measured frequencies with higher

measured radial velocity.

Figure 6. Methane, reacting case, radial velocity variation with equivalence ratio, with air mass flow at 282 g/min.

0 1 2 3 4 5 60

0.5

1

1.5

2

2.5

3

3.5

Velocity [m/s]

Perc

ent

of

Da

ta

0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.33.5

4

4.5

5

5.5

6

6.5

7

Velo

city [

m/s

]

Equivalence Ratio

7

A separate case study was observed for methane flames near lean blowout (LBO). These flames were operated at the

Case B air flow rate, but the equivalence ratio was reduced to 0.48, leading to the onset of LBO. At this condition the

flame is weakly resonating with a low acoustic amplitude. However the flame is very sensitive to disturbances and will

fluctuate between a lifted and anchored state. When anchored and resonating, the flame is flat and displays a higher

radial velocity near 4.5 m/s, as seen in Figure 7. When the flame is highly lifted, the shape changes to be more V-like

and the radial velocity drastically drops to 2.5 m/s. This suggests that the air flow is being redistributed is a similar to fashion as observed in Fig 6, dependent on flame position. These drops in velocity correspond to moments when the

flame is about to blowoff. Finally at the end of the series, after 60 sec, the flame is finally blown off. It can be reasoned

that high radial velocities are conducive to flame stabilization and good anchoring.

Figure 7. Methane flame operated near LBO, air mass flow rate = 282 g/min, phi = 0.48.

A secondary case study was conducted for an interesting behavior observed in rich propane flames. It was seen that the

flame shape would transition between a flat mode and a lifted, V-shape mode, similar to the previous methane LBO case.

Figure 8 depicts the measured velocities corresponding to this flame motion. This behavior has been displayed as 3

segments in Fig 8. Initially, the flame is resonant and flat with a high radial velocity. The flame then transitions some time later to the lifted, V-shape mode, marked by low radial velocities. In the final segment, the flame is seen to

dynamically transition back and forth in a “flapping” motion. This flapping motion is naturally occurring and unforced.

It is marked by the large fluctuations in velocity as the flame shape changes.

Figure 8. Rich propane flames operated at phi = 1.4 for a reduced air flow rate of 160 g/min. “Flapping” captured

exaggerated velocity fluctuations.

0 10 20 30 40 50 60 702

2.5

3

3.5

4

4.5

5

5.5

6

Velo

city [

m/s

]

Time [sec]

LDV Signal

0 20 40 60 80 100 120 1401

2

3

4

5

6

7

8

Velo

city [

m/s

]

Time [sec]

LDV Signal

8

This investigation has shown that flame shape is controlled by the distribution of air through the two swirlers. Two

further studies were conducted where the swirlers were independently blocked. When the inner swirler is blocked, air

only flows out of the outer annulus, and a flat flame shape is established. When the outer swirler is blocked, air is only

permitted to flow through the inner opening. This results in a highly lifted flame. Acoustics are highly amplified in the

flat flame case whereas they are weak, compared to background noise, when only the inner swirler is used. This finding

implies that there is a coupling between reflected pressure waves, flame shape, and the distribution of air through the swirlers. As a flat shape, the flame acts as larger blockage within the combustion chamber. Reflected pressure waves

may act on the incoming airstreams influencing flame shape. With a flat flame, the heat release is focused along a

smaller location in the burner, which allows for stronger amplification of the existing instability.

b. Pressure and Chemiluminescence Results:

Pressure measurements were taken simultaneously with high speed videos of the line of sight chemiluminescence. These

measurements showed detailed flame motions occurring over a given acoustic cycle. For the chemiluminescence images,

information regarding the average flame shape, liftoff height, average liftoff fluctuation distance, and frequency of heat release could be investigated. When combined with pressure data, Rayleigh indices [16] can be calculated which

determine the degree of coupling between the pressure field and heat release.

Figure 9 represent flame motions over the course of an acoustic cycle for a propane flame operated at Case B conditions.

The flame lifts off the face of the burner and then returns to a flat shape. The two red circles marked in Scene 2 of Fig 9

represent the location immediately above the fuel injector, where the propane is entering. These regions will be further

analyzed with acetone PLIF measurements. As the flame moves back upstream to the injector, This region remains

unburned. In Scene 3, there is still some unburned space in this region until the flame fully propagates back to the burner

face in Scene 4. The cycle repeats with the flame then lifting off again. Similar behavior is observed in methane flame.

Figure 9. Propane flame motion for P-1, conditions over a single acoustic cycle. The red circles represent the

regions investigated with acetone PLIF.

This liftoff and flashback cycling occurs at the same frequency as the pressure oscillations as given in the top images for

Figure 10 a and b. For methane flames, the average liftoff height was 6mm compared to 8mm for propane flames. Propagation speeds for the flame motion were also calculated for the upstream and downstream motions. For M-1 and P-

1, the propagation speeds in each direction were the similar. The upstream and downstream speeds for methane were

both 1.5 m/s. For propane the upstream speed was 7.4 m/s and the downstream speed was 7 m/s. The average

propagation fluctuation for a methane flame was 0.6±1.6mm and for propane, the average propagation distance was

2.9±8mm. These results indicate a high degree of motion and rapidly changing flame sizes and positions. The pressure

oscillations for both resonating cases, M-1 and P-1, occur around 300 Hz. The total chemiluminescence for each flame

can be seen to also fluctuate are near the same frequencies as given in Fig 10 a and b.

9

a) b)

Figure 10. a) Pressure and chemiluminescence power spectral densities methane flames operated at Case B.

b) Pressure and chemiluminescence power spectral densities propane flames operated at Case B.

For these Rayleigh indices, the pressure was recorded at a port located in the combustor wall, 1 cm above the injector

face. Figure 11 depicts the Rayleigh index field of each respective flame. Rayleigh indices greater than zero represent

thermo-acoustically amplified flames where the pressure field and heat release are in phase, thus satisfying the Rayleigh

criterion. Indices less than zero represent dampening. Figures 11a and 11b show the flame profile for methane and

propane flames. The index field is populated by values that are close to zero or slightly negative which is indicative of a

field with little coupling between the pressure and heat release. Figures 11a and 11b display the same characteristics for a resonating flame and there exists a strong region of amplification surrounded by dampened regions, similar to that shown

by Kang et al. [17]. There are noticeable differences in the flame shapes between the resonating flames and the non-

resonating flames. A non-resonating flame shows better anchoring and has an axially extended V-shape profile. While

for a resonating flame, its profile is wider at its base than the non-resonating flame, which implies a higher degree of

premixing. Flat flames are thermally compact and concentrate heat release in a smaller portion of the burner than an

axially extended V-flame. Thus, V-flames are less effective at applying thermal energy toward acoustic amplification

and satisfaction of the Rayleigh criterion.

a) b)

Figure 11. a) Rayleigh index for methane flame, M-1. b) Rayleigh index for propane flame, P-1.

For case M-1, there is a region of amplification at the base of the flame surrounded by regions of dampening. A similar

profile is given for P-1, except that the region of amplification is much stronger and larger. Despite showing similar oscillation frequencies, the amplitude of the instability for P-1 is much stronger than M-1 [4]. Both flames exhibit

behavior symmetric around the center line, with weaker amplification occurring at the centerline. The strongest

amplification occurs just about the fuel annulus, displaced a few millimeters to the left and right of centerline.

200 250 300 350 400 450 5000

1

2

3

4x 10

-5PSD Pressure

Frequency (Hz)Pow

er

Sp

ectr

al

Den

sit

y (P

si2/H

z)

200 250 300 350 400 450 5000

1

2

3x 10

10PSD Chemiluminescence

Frequency (Hz)Pow

er

Sp

ectr

al

Den

sit

y (P

si2/H

z)

200 250 300 350 400 450 5000

0.5

1

1.5

2x 10

-5PSD Pressure

Frequency (Hz)Pow

er

Sp

ectr

al

Den

sit

y (P

si2/H

z)

200 250 300 350 400 450 5000

1

2

3x 10

11PSD Chemiluminescence

Frequency (Hz)Pow

er

Sp

ectr

al

Den

sit

y (P

si2/H

z)

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

10

c. Acetone PLIF Results:

Methane and propane were bubbled through an acetone bath to seed the flow for fuel marking. The location of the

acetone signal marks regions of unburned fuel. Acetone PLIF measurements were taken over a field of view located

above the fuel injector to the right of centerline. Figure 12 shows the average acetone signal for case M-1 over 450

images. Two distinct lobes of fuel exist in the burner. Lobe A is located closer to the injector at a radial distance of 12-14

mm. Lobe B is lifted and located at a radial distance of 8-10mm from centerline, which is almost directly above the fuel

annulus.

Figure 12. Average acetone PLIF signal marked with ROIs of interest for M-1, with GTMC injector schematic.

These two regions have been investigated for dynamic behavior. It has been observed that the unburned fuel location

switches between Lobe A and B over the course of the data set. In order to develop a numeric metric, the entire signal

within the regions of interest (ROI) marked on Figure 12 are respectively considered. Each region of interest is equally

sized. For each frame in the 450 image set, the total signal in each ROI is summed. A ratio of the signal in region A versus region B is calculated. When this ratio is larger than 1, there is more fuel located near location A. For ratios close

to 1, there is an equal amount of fuel located in both regions A and B. For ratios much smaller than 1, the signal at

location B is much larger than that at location A. Figure 13 depicts sample frames for various calculated ratios of 57, 1,

and 0.08.

This metric can be used to track the movement of unburned fuel between these two regions as it switches back and forth.

These measurements were not time resolved, such that a time sequence of the motion can be captured. However,

individual statistics based on each frame can be used to evaluate this motion. Figure 14 displays a histogram of the

logarithm of these ratios, which allows for more balanced scaling. Values near zero correspond to an equal distribution

of unburned fuel. Positive values correspond to concentrations favoring ROI A, and negative values correspond to

concentrations favoring ROI B. Based on the histogram, 60% of the cases have more acetone in ROI A. There are also several cases where the signal in A is up to 100x lager than in region B. However from the average acetone image, the

total signal in each ROI is roughly the same, which indicates that even when the fuel is concentrated to one ROI or the

other, the signal strength is about the same. This back and forth motion can be related to the large scale fluctuations in

radial velocity which occurs over an acoustic period.

6 8 10 12 14 16 18 20 22 24-6

-4

-2

0

2

4

6

8

r [mm]

y [

mm

]

A A

B

11

Figure 13. Sample acetone PLIF images for various ROI ratios in case M-1.

Figure 14 . Histogram of M-1 ROI ratios. Positive ratios indicate fuel concentrated mostly in region A and

negative values indicate fuel mostly concentrated in region B.

Similar PLIF measurements were taken in propane flames as given in cases P-1 and P-2. Figure 15 depicts the average

acetone signal for the lean and rich propane cases. These images show similar dual lobe behavior as compared to the

average acetone methane signal. However, for lean P-1, the unburned fuel is concentrated to ROI B, whereas for P-2, the unburned fuel is concentrated to region A. The same ROI ratio was calculated for cases P-1 and P-2 as given in Figure

16. The ROIs are spatially located in the same locations relative to the injector as in the methane study. As well, each

ROI corresponds to equally sized areas as the methane study.

For the acetone signal in Fig. 15a, the signal is on average twice as strong in region B as region A for lean P-1. From Fig

16, the range of ratios exhibited is larger than that seen in M-1. Fig 16a shows that for 40% of the time, the signal in ROI

B is at least double the signal in A. Likewise, 40% of the time, the signal in ROI A is at least double the signal in ROI B.

This suggests, in coordination with the average signal image, that when the unburned fuel is located in ROI B, the

concentration is large, compared to when the fuel is in ROI A, where the concentration appears to be weaker. However,

when the fuel is located in ROI A, the signal in ROI B must be very weak. For the acetone signal in Fig. 15b, the signal

is on average 15% stronger in region A than in region B for rich P-2. Fig 16b shows that for 23% of the time, the signal

5 10 15 20-6

-4

-2

0

2

4

6

8

A/B Ratio = 57

r [mm]

y [

mm

]

5 10 15 20-6

-4

-2

0

2

4

6

8

A/B Ratio = 1

r [mm]

y [

mm

]

5 10 15 20-6

-4

-2

0

2

4

6

8

A/B Ratio = 0.08

r [mm]

y [

mm

]

-1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 30

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1Probability of Relative Acetone Location

LOG ROI Ratio

Perc

ent

of

Sa

mp

le

12

in ROI B is at least double the signal in A. However, 47% of the time, the signal in ROI A is at least double the signal in

ROI B. This suggests, in coordination with the average signal image, that the fuel is typically found in ROI A on a frame

to frame basis. However, fuel can appear almost equally strong in signal for either ROI.

a) b)

Figure 15. a) Average acetone PLIF signal for lean P-1. B) Average acetone PLIF signal for rich P-2.

a) b)

Figure 16. Histogram of a) P-1 and b) P-2 ROI ratios. Positive ratios indicate fuel concentrated mostly in region A

and negative values indicate fuel mostly concentrated in region B.

For case M-1, a wider field of view encompassing the entire burner was investigated as seen in the average acetone

image in Figure 17. Unburned fuel is identified on either side of the fuel injector. Again a dual lobe pattern can be

observed on either side of the centerline. The average signal on the right side of the burner appears to be stronger than on

the left side. This may be due to uneven fuel distribution to each of the fuel ports or absorption of the laser sheet, since

the sheet enters from the right. Each of these sides was treated as ROI, with the left zone marked as ROI A and the right

zone marked as ROI B. The histograms in Figure 18 also display some degree of switching from left to right in average

signal strength. This may be due to precession of the recirculation zone which, in turn, causes precession of the flame

around the injector annulus, as evident in high speed chemiluminescence videos. As the flame precesses around, a portion of the injector has no flame present, allowing for the buildup of unburned fuel. This may explain the observed

switching from one side to the other. The magnitude of these ratios suggest that the concentration of unburned fuel does

not shift as dramatically as observed in the ROIs between two lobes on a given side of the injector.

4 6 8 10 12 14 16 18 20 22 24-6

-4

-2

0

2

4

6

8

r [mm]

y [

mm

]

4 6 8 10 12 14 16 18 20 22 24-6

-4

-2

0

2

4

6

8

r [mm]

y [

mm

]

-3 -2 -1 0 1 2 30

0.02

0.04

0.06

0.08

0.1

0.12Probability of Relative Acetone Location

LOG ROI Ratio

Perc

ent

of

Sa

mp

le

-3 -2 -1 0 1 2 30

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1Probability of Relative Acetone Location

LOG ROI Ratio

Perc

ent

of

Sa

mp

le

13

Figure 17. Average acetone PLIF signal for M-1, for entire combustor field of view.

Figure 18. Histogram of M-1 ROI ratios for entire field of view. Positive ratios indicate fuel concentrated mostly

to the left of center and negative values indicate fuel mostly concentrated to the right of center.

4. Conclusions

Fluid mechanical coupling between injected reactants and the flame-base is observed in a gas turbine model combustor

(GTMC). This instability introduces oscillations in fuel and air mass flow rates which affects mixing, stabilization,

instability strength, flame shape, and anchoring. It has been observed in the GTMC that air flow redistribution can occur

between the two independent swirlers, which are fed from a common plenum. Uneven partitioning of air flow between

the individual swirlers strongly affects shear layer mixing. Dynamic swirler flow rate variations can result in large

redistributions of heat release which can affect flame shape and liftoff height. LDV measurements of radial velocities in

the various flames show velocity fluctuations which correspond to flame motion oscillations captured by high speed

chemiluminescence imaging.

Rayleigh indices have been compared to fuel-air distributions to understand how mixing affects instability strength.

When air flow is redistributed to the outer swirler, larger radial velocities are recorded. This mixing pattern encourages a flat flame shape which is unanchored and oscillates up and down. Rayleigh indices of flat flames indicate stronger

acoustic coupling and amplification as compared to cases where the flame may be V-shaped and/or anchored. The latter

flames correspond to flow conditions where a lower radial air velocity is observed and the flame is non-resonant.

Resonating flames also show the pressure, chemiluminescence, and liftoff height all oscillate at the same frequency.

Acetone PLIF images depict low-lying regions of fuel extending towards the outer edge of the burner. They also depict

two distinct regions of localization in unburned fuel. The distribution of fuel between these two regions oscillates from a

central lifted position to a lower, radially extended position. Methane flames show even favoring between these two

regions, whereas propane flames show a position preference dependent on equivalence ratio. A similar switching from

side to side has also been observed between respective sides of the fuel injector.

Acknowledgements

This research was funded by ONR under grant N00014-10-10561 and by the DOE-UTSR program under grant

FE0007060. The authors thank Dr. Wolfgang Meier of DLR Stuttgart for permission to use the GTMC design. The

-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.60

5

10

15

20

25Probability of Relative Acetone Location

LOG ROI Ratio

Perc

ent

of

Sa

mp

le

14

acetone PLIF measurements were funded by the National Science Foundation award CBET 0852910 which was

monitored by Dr. Arvind Atreya and Dr. Ruey Hung Chen.

References

[1] Gupta, A.K., Lilley, D.G., and Syred, N., Swirl Flows, Abacus Press, Kent, 1984.

[2] J.W.S. Rayleigh and R. Lindsay, The Theory of Sound: Vol. 2, Dover Publications, 1945, p. 230.

[3] Lieuwen, T., Torres, H., Johnson, and C., Zinn, B.T., “A mechanism of combustion instability in lean premixed gas

turbine combustors”, J. Eng. Gas Turb. Power, 123, 2001, p. 182 – 189.

[4] Allison, P.M., Driscoll, J.F., and Ihme, M., Proc. Combust. Inst., 34 (2013) 3145–3153.

[5] Giezendanner, R., Weigand P., Duan X.R., Meier W., Meier, U., and Aigner M., “Laser-based investigations of

periodic combustion instabilities in a gas turbine model combustor”, J. Eng. Gas Turb. Power, 127, 2005, p. 492 – 496.

[6] Weigand P., Meier W., Duan X.R., Stricker W., and Aigner M., “Investigations of swirl flames in a gas turbine

model combustor I. Flow field, structures, temperature, and species distributions”, Combust. Flame, 144, 2006, p. 205-

224. [7] R. Sadanandan, M. Stöhr, and W. Meier, Combustion, Explosion, and Shock Waves 45 (2009) 518–529.

[8] A.M. Steinberg, I. Boxx, M. Stöhr, W. Meier, and C.D. Carter, AIAA J., 50 (4) (2012) 952 – 967.

[9] Duan X.R., Meier W., Weigand P., and Lehmann, B., “Phase-resolved laser Raman scattering and laser Doppler

velocimetry applied to periodic instabilities in a gas turbine model combustor”, Applied Physics B: Lasers and Optics,

80, 2005, p. 389-396.

[10] Rosenberg, D.A., Allison, P.M., and Driscoll, J.F., Method to measure flame index in a partially-premixed gas

turbine combustor. In 51st AIAA Aerospace Sciences Meeting, Grapevine, TX, 2013. AIAA 2013-1181.

[11] Rosenberg, D.A. and Driscoll, J.F., A method to image flame index in partially premixed flames. In 50st AIAA

Aerospace Sciences Meeting, Nashville, TN, 2012. AIAA 2012-0972.

[12] A. Lozano, B. Yip, and R. K. Hanson. Exp. Fluids, 13 (1992) 369–376.

[13] M. C. Thurber and Ronald K. Hanson. Appl. Phys. B, 69 (1999) 229–240. [14] Christof Schulz and Volker Sick. Prog. Energy Combust. Sci., 31 (2005) 75–121.

[15] Noel T. Clemens. Flow Imaging. In Encyclopedia of Imaging Science and Technology, Joseph P. Hornak,

editor,pages 390–419. John Wiley & Sons, Inc., New York, 2002. [16] W. Pun, S.L. Palm, and F.E.C. Culick, Combust. Sci. and Tech. 175 (2003) 499-521.

[17] D.M. Kang, F.E.C. Culick, and A. Ratner, Combust. Flame 151 (2007) 412–425.