Unsteady vortex breakdown in an atmospheric swirl stabilised combustor. Part 1: Chamber behaviour

Combustion and Flame 162 (2015) 388–407

Contents lists available at ScienceDirect

Combustion and Flame j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m b u s t fl a m e

Unsteady vortex breakdown in an atmospheric swirl stabilised combustor. Part 1: Chamber behaviour Adam Ruggles a,b,⇑, James Kelman b a b

Combustion Research Facility, Sandia National Laboratories, 7011 East Avenue, Livermore 94550, USA Cranfield University, College Road, Cranfield, MK43 0AL, UK

a r t i c l e

i n f o

Article history: Received 14 February 2014 Received in revised form 15 April 2014 Accepted 5 July 2014 Available online 19 August 2014 Keywords: Gas turbine combustor Combustion instabilities Vortex breakdown Swirling flows High speed diagnostics SPIV

a b s t r a c t This paper presents the behaviour of three very different and unique flame and flow structures within an atmospheric swirl-stabilised dump combustor supplied with a lean premixed mixture of methane and air. The reactant flow was artificially perturbed with frequencies of 100 Hz, 200 Hz, and 400 Hz. Phase average behaviour and temporal dynamics were characterised using phase locked high speed CH chemiluminescence and high speed stereo particle imaging. The interaction between the flame and flow field, in particular the internal recirculation zone of the vortex breakdown, was determined to be responsible for differences observed in behaviour at the three forcing frequencies. The 100 Hz perturbation frequency displayed simple oscillatory motion. Higher perturbation frequencies of 200 Hz and 400 Hz gave rise to a second toroidal vortex ring which formed within the internal recirculation zone adjacent to the inner shear layer. This caused additional out of phase modulation of the heat release rate and flame area. Twin counter rotating vorticity structures attached to the annulus were formed as a result of the chamber geometry. The oscillating inlet flow and oscillating reversed flow region of the inner recirculation zone caused oscillations in vorticity magnitude which were responsible for flame wrinkling and stretch effects upon the flame front. Vorticity within the shear layers was found to be the source of harmonic frequency generation of the imposed perturbation frequencies. The data is presented in detail to facilitate CFD model comparisons, particularly LES. Published by Elsevier Inc. on behalf of The Combustion Institute.

1. Introduction The social and legislative demand to reduce the quantities of pollutants (NOx and CO) and emissions of CO2 associated with hydrocarbon derived electricity has led to the dry low NOx approach that uses premixed fuel–air mixtures with lean equivalence ratios (u = 0.7–0.9 typically). Lean premixed combustion (LPC) implementation has been hampered by self-sustaining combustion instabilities that couple the unsteady heat release and chamber pressure coinciding with one or more acoustic modes of the combustor. The inability to predict the onset, type, and behaviour of instabilities during the combustor design stage highlights the limits of scientific understanding of the phenomena. Aerodynamic techniques over physical flame holders have become popular to stabilise high velocity combustion. Recirculation zones circulate radicals and hot burnt products back to ignite the fresh gas maintaining a constant reaction rate, improve mixing, ⇑ Corresponding author at: Combustion Research Facility, Sandia National Laboratories, 7011 East Avenue, Livermore 94550, USA. E-mail address: [email protected] (A. Ruggles). http://dx.doi.org/10.1016/j.combustflame.2014.07.016 0010-2180/Published by Elsevier Inc. on behalf of The Combustion Institute.

and increase residence times. A dump plane, or sudden expansion, creates a toroidal vortex structure often referred to as a corner recirculation zone (CRZ) between the dump plane and chamber wall. Using swirl introduces a tangential velocity component and axial vorticity to the inlet flow. With strong swirl, S > 0.6, (ratio of axial flux of angular momentum and axial thrust [1]) off axis vortex breakdown is observed due to the tangential velocity component creating the internal recirculation zone (IRZ) of reversed axial flowing fluid located around the chamber centreline. The specific arrangement of the recirculation zones is highly dependent upon the combustor geometry, combustion type (premixed or diffusion), fuel injection method, swirler type and design, exhaust profile, and Reynolds and swirl numbers. Comprehensive reviews can be found in the literature [1–3]. The vortex breakdown is dependent upon the swirl number and the subsequent axial pressure gradient. This gradient forms as the flow moves away from the axial direction. Above a critical swirl number the kinetic energy of the flow will be unable to overcome the swirl dependent axial pressure gradient and flow reversal occurs [4]. Combustion chamber geometry also contributes to the strength of the pressure gradient and corresponding decrease of

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

mean axial kinetic energy, such features include a bluff body which creates a low pressure focus at its tip [5], a sudden expansion – redirecting axial flow, and any flow restriction at the exhaust [6]. Once vortex breakdown has formed the behaviour is controlled by the ratio of axial and tangential velocity. In total, seven forms of vortex breakdown have been observed [7] based upon three types: the spiral, double helix, and bubble (axisymmetric) [8,9]. The spiral/helical types are characterised by the vortex core/cores adopting a spiral shape that persist for several revolutions before breaking up into turbulence [10]. The vortex core is wrapped around the boundary of reversed flow and can rotate (precess) around the centre axis, distorting the reverse flow region at acoustic mode frequencies [11]. The bubble type is described by an expansion that envelopes the reversed flow region and can develop from the spiral type or directly from swelling of the vortex core. The nature of the internal structure of the bubble is not fully understood. Bubbles with a single toroidal vortex ring [4,7,12] or two counter rotating toroidal vortex rings have both been observed with no clear explanation to account for the difference. Both the spiral and bubble type of vortex breakdown have been observed in computational fluid dynamics (CFD) simulations and measurements of combustors [3,13–20]. For instabilities to become self-sustaining the unsteady heat release rate and pressure must be in phase, satisfying the Rayleigh criterion [21]. The consensus of the literature available is that there is no single fluid mechanical mechanism responsible for relating these two scalars. Experimental and CFD investigations have shown shear layer instabilities shed off the inlet lead to the periodic rippling of the flame front, thereby causing an unsteady heat release rate [15,16,18,22–24]. Coherent vortex roll up – the formation of a toroidal ring off the inlet – can cause large scale modifications to the flame structure or cause local extinction resulting in the undesired transport of fresh gas downstream to be consumed out of phase, both resulting in an unsteady heat release rate [25,26]. Recent attention has focused upon the precessing vortex core (PVC), the spiral shaped structure associated with strong swirl flame stabilisation. The vortex core rotates around the combustor centreline causing wrinkling of the flame front similar to the shear layer instability mechanism [16,27]. A thorough review of precessing vortex cores and vortex breakdown can be found in the literature [1,2,4,5,8,9,28]. In addition to fluid mechanical mechanisms, variations in equivalence ratio can also cause instabilities [29,30]. With respect to the vortex breakdown phenomena, only the PVC has been identified with a role in combustion instabilities. The range of potential mechanisms, dependency on combustor geometry and operating conditions, the coupling of thermo-acoustic processes, and interactions among these variables need to be better understood for instabilities to be mitigated and LPC to reach full potential. Flame transfer functions (FTF) have become a popular choice to determine the response of a specific combustor when exposed to acoustic forcing, thereby identifying frequencies and excitation values which cause dramatic instabilities [31–36]. This allows combustor designers some indication of which frequency ranges could be susceptible to instabilities provided the geometries are similar. These investigations require some measure of the unsteady heat release rate and a measurement of the inlet velocity. Such measurements are typically performed using chemiluminescence (occasionally a PLIF technique has been used [37,38]) imaging and a calibrated pressure measurement. One dimensional acoustic analysis methods are also being developed to attempt to predict the FTF without experimentation and construction of the combustor. It has been shown that similar atmospheric combustors experiencing the same instability mechanism possess similar FTFs [39] but that this cannot be extended to instabilities experiencing different mechanisms [40]. The FTF itself however does not yield

389

any information regarding the flame or flow field structure, nor the fluid mechanical mechanisms present. Provided the heat release was determined from an imaging method an analysis of the flame structure can be attempted although limited by the nature of data acquisition (phase locked or not). However behaviour of the flow field and fluid mechanical mechanisms will remain unknown. If a FTF is going to represent a particular combustors’ response it is important to understand the physical mechanisms that underlie the FTF. This means investigations of the combustor physics subject to imposed perturbations are needed to understand why a particular FTF is what it is. Additionally changes to the combustor geometry (or pressure regime [41] or fuel composition [42]) will cause changes to the combustor physics, particularly the vortex breakdown and swirling dynamics, which will require the FTF to be determined again. These changes of physics will be captured by the newly measured FTF but unless the physics are understood interpretation of the FTF is restricted as changes to the flow field physics will not be known. Understanding how vortex breakdown responds to geometry changes and how it behaves under unstable conditions are essential to allow the prediction of an FTF that is independent of combustor geometry. This knowledge will also improve FTF interpretation with respect to the fluid mechanics and therefore improve the FTFs applicability to real sized gas turbines. Many of the investigations cited here were performed using large eddy simulation (LES). LES has become the most suitable CFD method to study instabilities and to develop design tools for gas turbine manufactures. This is primarily due to the resolving of temporal and spatial scales larger than an appropriate filter size, yielding instantaneous realisations of the flow. Still in its development stage, LES requires high quality validation data to gain confidence in it as a predictive development tool that requires little or no a priori experimental data. Such experimental investigations have been performed in atmospheric or low pressure combustors. Utilising optical access, these combustors have been probed nonintrusively, using precise optical diagnostics and sensory instruments, when experiencing instabilities (see experimental references above) [43,33,44–47]. The collected data are used for physical insight, typically of a single operating condition, and as LES validation data. Parallel studies of experimentation and LES have used the experimental data to gain confidence in the simulation, which was then used for an advanced analysis that could not be performed with the limited diagnostics [11,15,16,48,49]. Currently, the most comprehensive validation data set has been acquired by the German national laboratory, DLR [17,27,50–54]. This publication builds upon the work previously presented [55] where an atmospheric model gas turbine combustor designed for instability research and LES validation was introduced. A lean premixed methane/air mixture with an equivalence ratio of u = 0.8 was artificially perturbed at 100 Hz, 200 Hz and 400 Hz, and the resulting flame dynamics were characterised with high speed CH chemiluminescence and stereo particle imaging velocimetry (SPIV). A full uncertainty analysis of the applied diagnostics was documented, and the ensemble average results were presented. This second publication details the phase locked averaged results and presents a global description of the instabilities. Rather than determine the FTF of the combustor this work seeks to investigate mechanisms present within the chamber during these perturbations through a more detailed investigation. Of importance is the observation of additional toroidal vortex rings forming adjacent to the flow reversal region upstream of their expected position during the instabilities. These additional vortex rings caused additional out of phase flame area modulation. To our knowledge this phenomenon has not previously been observed experimentally and is a newly documented fluid mechanical structure that can impact combustion instabilities. The actual mechanisms that affected heat

390

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

release were changes of flame surface density (FSD) and flame stretch by strain inducing structures, depicted by the vorticity of two counter rotating vorticity structures attached to the annulus. In parallel to the physical insight the data are presented in a detailed manner to be suitable for CFD model validation case studies and are accompanied with boundary conditions, heat release characterisation, flame and flow field characterisation, and acoustic analysis. The three cases presented here reveal unique flame and flow field structures and behaviours. 2. Experimental set up and data processing A lean premixed methane/air mixture with an equivalence ratio of u = 0.8, density 1.13 kg m3, and mass flow rate of 20 g s1 was supplied to an atmospheric swirl-stabilised dump combustor previously described in detail [55]. The swirl number was S = 0.865 imparted by an axial vane swirler [1]. The inlet annulus inner and outer diameters were 20 mm and 28 mm, respectively. The chamber was made from optical grade fused silica with inner and outer diameters of 130 mm and 142.7 mm. Both the dump plane and top of the bluffbody possessed a ceramic inlay to make the chamber as adiabatic as possible. Sensory measurements of temperature and pressure prior to chamber entry were possible using machined mounting points. These were located 135 mm and 116 mm upstream of the annulus entry, respectively. Temperature measurements within the ceramic dump plane were also made possible with a thermocouple embedded within the ceramic at a radial position of 35 mm from the centreline and at a depth of 5 mm below the dump plane (ceramic inlay was 10 mm deep). A siren device upstream of the optical chamber and swirler was used to impose artificial perturbations of desired frequencies and intensity. The design was such to create a near sinusoidal profile of aperture. The combustor was operated prior to data collection to achieve thermal equilibrium. For this investigation the mixture flow rate was perturbed at frequencies of 100 Hz, 200 Hz, and 400 Hz. An acoustic analysis of the chamber under these operating conditions revealed the first longitudinal mode at 840 Hz with radial and tangential modes higher in frequency [56]. The perturbation frequencies investigated were below all the acoustic modes and no evidence of any natural acoustic mode being active was detected. Phase locked high speed CH chemiluminescence and SPIV were used to investigate combustor behaviour. For both diagnostics a synchronisation signal was generated by the siren as each cycle of perturbation was instigated. A signal generator used this signal to create the individual pulses used to trigger the respective diagnostic at each phase position. CH chemiluminescence was used to determine the unsteady heat release rate and assess the phase averaged flame structure after undergoing an Abel three point deconvolution procedure. The spatial resolution was 0.4 mm/pixel. Image acquisition was triggered to achieve 8 images per cycle with 200 images per phase recorded (total 1600 images). High speed SPIV was used to acquire instantaneous 3 component velocity fields to characterise the chamber flow field in its entirety. The cameras were arranged in the forward scatter configuration with each camera 45° to the image plane. Three distinct imaging regions were interrogated to characterise the chamber. These were termed the corner recirculation zone, the flame zone, and upper recirculation zone and are referred to as the CRZ, FZ, and URZ. Using the processing method previously described [55] the vector field resolution was 0.75 mm/vector. The inlet flow was characterised with the imaging region termed, Annulus 3D, which had an improved spatial resolution, 0.59 mm/vector. Zirconia stabilised with yttria was chosen as the seeding medium due to its superiority as a refractory material compared to other seed materials. This extended the life of the fused silica chamber. The SPIV system was operated to acquire 10 images per cycle, corresponding to a

little over 100 vector fields per phase (total approx. 1000) completely filling the camera memory. The difference in temporal resolution between CH and SPIV measurements resulted from unanticipated hardware limitations. A full analysis of the geometrical, parameter, and gradient uncertainty has previously been presented [55]. The geometrical errors were 0.058 ms1 and 0.06 ms1 for the in plane and out of plane directions giving an error ratio of 1.04 for the three chamber imaging regions. For the annulus region these were 0.036 ms1 and 0.037 ms1 with an error ratio of 1.06. These agree with the literature predicted value of unity for a forward scatter arrangement with both cameras at 45° to the imaging plane. The total parameter uncertainty for the chamber regions was an rms value of 0.772 ms1 and 0.816 ms1 for the annulus region. A conservative range of bias induced by gradients was 1 ms1 for shear and ±3 ms1 for strain, occurring in the regions of strongest gradients, the shear layers. The vector fields were calculated using LaVision software using the approach previously described [55]. To improve all statistical calculations the Jackknife resampling technique was used to determine statistical uncertainties and to yield a more converged statistics for all measurements [57]. The Jackknife method removes the first data sample from a set of size n and calculates the desired statistic. The omitted sample is re-introduced and the next sample is removed and the statistic re-calculated. This process is repeated leading to n calculations of the desired statistic. The mean of these calculations yields a more converged statistic and the uncertainty is the 2r value of the calculations. Temporal spectra of the pressure, CH chemiluminescence, and SPIV measurements within this paper were processed using the following approach. From the continuous data set in question a truncated, half-length set was created, retaining the first data point to the middle data point, and the power spectrum was calculated. A second truncated data set was then created by incrementing the start and end data points by one position and the power spectrum calculated. This process was continued until the truncated data set was formed from the second half of the original data set. This process yielded hundreds of individual power spectra from which the mean was calculated. This reduces the dynamic range of the power spectra by a factor of 2 while reducing the uncertainty by p a factor of N. 3. Boundary conditions The boundary conditions both upstream and downstream of the optical chamber were determined using a combination of temperature, pressure, and SPIV measurements. The upstream mixture gas temperature was a steady 296 K for the three experiments. The upstream pressure perturbations and corresponding frequency spectra are shown in Fig. 2 along with the exhaust pressure. The temperatures within the ceramic dump plane after achieving thermal equilibrium were 418 K, 492 K, and 510 K, showing that temperature at this location increased with perturbation frequency. The acoustic impedance at the inlet was estimated using the upstream rms pressure and the rms integrated velocity magnitude across the inlet determined using SPIV (from the Annulus 3D region). This was multiplied by the cross sectional area to determine the volumetric rms flow. The impedances were 42 kPa s m3, 59 kPa s m3, and 112 kPa s m3, respectively. The increases in impedance reflect the increasing amplitude of pressure oscillation. The strength of perturbation was defined by using the rms integrated velocity magnitude across the inlet divided by the mean velocity across the inlet. This yielded perturbation intensities of 0.14, 0.12, and 0.15. The upstream pressure information revealed the presence of multiple weaker integer harmonic frequencies for all three cases.

391

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

Both the 100 Hz and 400 Hz profiles are of a general sinusoidal form whereas the 200 Hz profile is clearly not. The spectral plots reveal that in the 200 Hz case the first harmonic was almost equal in strength with the imposed frequency. The presence of such a strong first harmonic explains the unique shape of the pressure profile. This was not observed for the 100 Hz or 400 Hz case where the first harmonic is at least one order of magnitude weaker than the imposed frequency. This dominance suggests that the chamber dynamics were controlled by the imposed frequency. For the 200 Hz case the strength of the first harmonic implies increased complexity of the chamber dynamics.

chemiluminescence technique. It is used to observe flame area trends rather than an absolute measurement of flame area. Figure 3 shows the heat release and flame area behaviour. Both the 100 Hz and 400 Hz cases exhibit simple oscillating heat release behaviour where as the 200 Hz case demonstrates a more complex, multi peak profile with much reduced amplitude. The phase averaged flame brush area for the 100 Hz case is in phase with the heat release and has a large amplitude of oscillation. With increasing perturbation frequency it can be seen that a phase lag develops, a single phase position for the 200 Hz case, and 2 phase positions for the 400 Hz case. Additionally as perturbation frequency increases the cycle average flame area increases and the amplitude of oscillation decreases. At the same it must be noted that the flame area profiles are of the same form as the heat release rate profiles, matching single and twin peak occurrences. Flame Surface Density (FSD), representing flame–turbulent interactions, and flame stretch under a premixed flamelet assumption, are both proportional to the progress variable reaction rate [56] which can be represented by the measured heat release rate. During unsteady inlet flows it is reasonable to assume that the local turbulent conditions upon the flame front change in space and time resulting in a varying FSD. These differing turbulent structures will also to a lesser degree impart varying flame stretch effects. As perturbation frequency increases, the flame area trends described above

4. Unsteady heat release rate The corrected instantaneous CH chemiluminescence images were integrated and normalised by their respective ensemble average before being phase averaged to characterise the unsteady heat release for each case. The rms value at each phase position was calculated and represents the variation due to turbulent fluctuations. Additionally, the phase averaged flame brush area was also calculated (only on the phase averages after deconvolution). Not to be confused with flame surface density, this was calculated by simply summing the number of pixels that were non-zero after processing and represents the flame brush area as measured by a Fuel Injector Assembly Siren Assembly

Electric Motor Assembly

Combustion Chamber

Swirler Assembly

Exhaust

Exhaust pressure measurement

Upstream pressure and temperature meeasurement 58mm 95mm

241mm

400mm

163mm

249mm

480mm

Global imaging areas

Inlet imaging area

Centreline

Centreline

54mm

120mm

URZ FZ

81mm

42mm

40mm

33mm

CRZ 3mm Centreline

3mm Centreline

Fig. 1. Overview of combustion chamber and SPIV interrogation regions CRZ, FZ, URZ, and Annulus 3D.

392

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

indicate the flame itself is being additionally distorted by the fluid mechanics present. This is particularly evident in the 400 Hz case which experiences a large heat release rate oscillation but small flame area oscillations of a large flame. As flame size variation is decreased it can be hypothesised that to account for the very large heat release rate oscillation, FSD and flame stretch variations become larger with perturbation frequency. Physically this can be interpreted as different vortex breakdown behaviour increasing the amount of flame–turbulence interaction (wrinkling) and causing flame stretch above what would normally be expected by general turbulence, such as in the 100 Hz case. This hypothesis could be verified with the use of a OH-PLIF investigation. The behaviour of the vortex breakdown in these cases will be presented in the following sections. Examination of the heat release rate spectra revealed the same integer harmonics present with the same strength relative to the

imposed frequencies in each case as those found from the upstream pressure spectra. A single exception was the 1.2 kHz harmonic observed in the pressure spectrum for the 400 Hz case was absent in the heat release rate spectrum. The spectra of both the 100 Hz and 400 Hz cases reveal the dominance of the imposed frequency over the harmonics whereas the spectra of the 200 Hz case reveals a first harmonic nearly equal in strength to the imposed frequency. 5. Inlet velocity behaviour To characterise the inlet velocity behaviour the SPIV measurements from the Annulus 3D region (Fig. 1) were integrated across the inlet. The values determined from the instantaneous vector fields were phase averaged for each velocity component and the velocity magnitude. Additionally, the spectra of the components and magnitude were determined. These are presented for each case in Figs. 4–6.

Pressure (Bar g)

100Hz

200Hz

400Hz

0.045

0.06

0.08

0.03

0.04

0.04

0.015

0.02

0

0

0

0.01

0.02

0.03

0

0.04

0

0.005

0.01

Time (s)

0.015

Inlet

Power (Bar2s)

−0.04

0.02

10-4

10-6

10

-6

10-6

-8

10-8

10

10

500

750

10

1000

0

500

Frequency (Hz)

0.0075

0.01

1000

1500

3000

4000

-8

10-10

-10

250

0.005

Exhaust

10-4

0

0.0025

Time (s)

10-4

10-10

0

Time (s)

2000

0

2000

1000

Frequency (Hz)

Frequency (Hz)

Fig. 2. Mean upstream and downstream pressure profiles with corresponding upstream frequency spectra.

200Hz

400Hz

×10

Heat release (au)

1.4

1.4 1.2 1.2 1.0

1.0

0.8

Heat release rate Flame area 0.8

0.6 2

4

6

8

2

4

6

2

8

4

6

Phase

Power (au)

4

10

14

10

12

10

10

10

8

10

6

Flame area (mm2)

100Hz 1.6

8

2

4

6

8

2

12

10

10

10

8

10

6

10

4

4

10

200

300

Frequency (Hz)

400

10

0

200

400

8

2

4

6

8

Phase

14

100

6

Phase 10

0

4

600

Frequency (Hz)

800

10

14

10

12

10

10

10

8

10

6

10

4

0

400

800

1200

1600

Frequency (Hz)

Fig. 3. Mean integrated heat release normalised by the cycle mean, and mean flame area plots with corresponding frequency spectra. Error bars are the rms values at each phase.

393

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407 80

Power (m/s)2s

103

Velocity (ms-1)

60

103

100Hz - vx

102

102

101

101

100

100

10-1 40

10-1 0

100

Power (m/s)2s

0 2

4

6

8

10

2

4

6

8

vy (axial)

300

400

0

500

100

200

300

400

500

100Hz - vy

400

500

100Hz - vmag

102

102

101

101

100

100

10 10-1

10-1

Phase vx (radial)

200

103

103 20

100Hz - vz

0

vz (tangential)

100

200

300

400

500

0

100

200

Frequency (Hz)

300

Frequency (Hz)

Fig. 4. Integrated phase averaged velocities and spectra for the 100 Hz case.

80

Power (m/s)2s

103 102

102

101

101

100

100

10-1

40

200Hz - vz

10-1

0

200

400

600

800

0

1000

103

200

Power (m/s)2s

20

0 2

4

6

8

10

2

4

6

8

vy (axial)

800

1000

800

1000

200Hz - vmag

102

102

101

101

100

100

10-1

10-1 0

vmag

vz (tangential)

600

10

Phase

vx (radial)

400

103

200Hz - vy

200

400

600

800

1000

0

200

400

Frequency (Hz)

600

Frequency (Hz)

Fig. 5. Integrated phase averaged velocities and spectra for the 200 Hz case.

80

103

103

400Hz - vz

Power (m/s)2s

400Hz - vx

60

102

102

101

101

100

100

10-1

10-1

40

0

400

103

Power (m/s)2s

Velocity (ms-1)

Velocity (ms-1)

60

103

200Hz - vx

20

0 2

4

6

8

10

2

4

6

8

vy (axial)

1200

1600

2000

102

102

101

101

100

100

vmag

400

800

1200

1600

2000

1600

2000

400Hz - vmag

10-1

10-1

vz (tangential)

0 103

400Hz - vy

10

Phase

vx (radial)

800

0

400

800

1200

1600

2000

0

Frequency (Hz)

Fig. 6. Integrated phase averaged velocities and spectra for the 400 Hz case.

400

800

1200

Frequency (Hz)

394

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

The integrated velocity components and magnitude in the 100 Hz case are all oscillating at 100 Hz. The majority of combustion harmonics detected by the chemiluminescence and pressure spectra are present in all component and magnitude spectra. Of note is the radial and tangential components are in phase with each other but out of phase with the axial component by three phase positions. By comparison the 200 Hz case shows that the velocity magnitude is a product of near equal strength 200 Hz and 400 Hz contributions, as revealed by the pressure and chemiluminescence measurements. The strongest frequency in the tangential spectrum is actually the 400 Hz harmonic rather than the 200 Hz perturbation frequency. However, both axial and radial velocities are dominated by the 200 Hz frequency with only the axial component exhibiting the 400 Hz harmonic. Inspection of the integrated phase averaged profiles shows that the velocity magnitude possesses such a strong 400 Hz harmonic due to the phase relationship amongst the components. It is clear that the axial and radial components are 180 degrees out of phase. The 400 Hz case shows more similarity with the 100 Hz case in that all components and the magnitude are controlled by the perturbation frequency. The harmonics detected in the magnitude spectra are dominated by the axial component. Once again the radial and tangential components are out of phase with the axial component, but the respective oscillation amplitudes are insufficient to create a harmonic as strong as the perturbation frequency in the magnitude spectrum. Common to all three cases is that the amplitude of oscillation is greatest for the axial component, followed by the radial, and then the tangential component. Additionally, the largest valued component was consistently the axial component, followed by the tangential and then radial. This information, together with the reported boundary conditions describes the inlet conditions to the chamber. It must be acknowledged that the dynamics of combustion and vortex breakdown will also have an effect upon the inlet flow as it enters the chamber.

6. Chamber behaviour To show the flame and flow field behaviour within the chamber phase averaged images of the chemiluminescence imaging and SPIV (using representative streamlines derived from the phase averaged velocity fields – note these are not the mathematical definition of streamlines but are plotted as such) results are presented. These reveal the structural changes to the flame and flow field for each case. The images are split into two figures showing the first and second half of the chamber behaviour. For the 100 Hz case these are presented in Figs. 7 and 8, Figs. 9 and 10 for the 200 Hz case, and Figs. 11 and 12 for the 400 Hz case. For the chemiluminescence eight phase positions were used and 10 phase positions were used for the SPIV. The discussion of each case is considered with respect to the chemiluminescence cycle. As supplementary material phase average velocity profiles taken at 30 mm, 50 mm, and 70 mm downstream of the annulus of all three velocity components are also presented and are intended to be used, in conjunction with the images, for quantitative CFD comparison. This information is presented in Figs. S1–S9. Additionally, avi files of both the phase averaged chemiluminescence and SPIV results for each case are provided. The flame and flow field sequence for the 100 Hz case is shown in Figs. 7 and 8. The first three phases of CH chemiluminescesnce reveal a flame at an angle of 45° that retreats to the dump plane and in doing so becomes parallel to the dump plane. This retreat agrees well with the flow field images, which show the inlet structure bent over to become almost horizontal, compressing the corner vortex ring. The inner recirculation zone grows, gaining a near

horizontal aspect to its upstream boundary to match the new orientation of the inlet flow structure. This position corresponds to the minimum in heat release and flame area plots of Fig. 3. In phases 4 and 5 of the CH image sequence (Fig. 7) the flame begins to grow as the inlet flow increases. The flame is pushed downstream and curves backwards towards the dump plane over the corner vortex ring. The inlet flow structure begins to straighten and increases in strength, pinching the upstream part of the inner recirculation zone. CH phases 6 and 7 correspond to the peaks of heat release rate and flame area plots. The flame tips are angled downstream and the inner recirculation zone is clearly being compressed by the increased inlet flow and expansion from increased combustion. The increased penetration of the inlet flow structure stretches the corner vortex ring downstream, causing its elongation. CH phase 8 marks the start of the flame’s retreat. This agrees well with the final flow field image, which shows a weakening of the inlet flow structure and a corresponding radial expansion of the inner recirculation zone. Inspection of the 200 Hz sequence in Figs. 9 and 10 reveal a very different flame and flow structure which is controlled by the formation of a second, internal vortex ring within the internal recirculation zone. Phases 1–3 of the chemiluminescence sequence reveal a very different flame in retreat. The flame possesses an upstream vertical aspect before bending to a 45° outward angle. Rather than a marked axial movement, the flame appears to withdraw into itself while keeping the same overall form through these phases. This is due to the previously mentioned second vortex ring that forms close to the annulus and inline with the corner vortex ring. This new structure and the corner vortex ring causes the inlet flow, and hence the flame to straighten. Phase 4 shows a dramatic change in flame structure to two opposed ‘C’ shapes. This is attributed to the increased influence of the new vortex ring, which is able to bend the flame around itself and then push the flame structures radially outward in phase 5. In phases 6 and 7 the flame structure moves downstream and resembles a form similar to that seen in the 100 Hz case. By this point the more downstream vortex ring has disappeared and the remaining vortex ring has been pushed downstream to a more typical axial position, albeit with an overall triangular shape. The final chemiluminescence phase shows the flame starting to retreat as the peak intensity decreases. The formation of the second upstream vortex ring close to the inner shear layer is evident. Additionally, the chemiluminescence image shows combustion occurring on the inlet flow structure upstream of the main flame. This is also detectable on the final flow field image where wrinkling of the inlet flow is seen. The peak of heat release and flame area occurs at phase 6 on the plot of Fig. 3. This phase position, as previously described, is similar to that of the 100 Hz case and possesses only a single vortex ring within the internal recirculation zone. The formation and growth of the upstream vortex ring during the flame retreat phases halts the expected drop in heat release rate and flame area (Fig. 3), and is responsible for the second spike in both plots. Its presence causes additional flame stretch in the inner shear layer in particular. This increases local combustion and flame area. The 400 Hz case (Figs. 11 and 12) has simple sinusoidal type upstream pressure and chamber heat release profiles, as did the 100 Hz case. However, the flame structure is markedly different from the 100 Hz case, with more similarities to the structures of the 200 Hz case. Phase 1 (Fig. 11) corresponds to the peak of heat release rate (Fig. 3). The flame is pushed to the chamber walls and substantial CH signal is present within the internal recirculation zone, which possesses a single vortex ring. The flame retreats during CH phases 2–6. The inlet flow structure straightens and experiences a sharp radial turn. A second vortex ring within the internal recirculation zone is observed on the inner shear layer. Starting with a near vertical upstream orientation, the flame

P2 of 10

P3 of 10

P4 of 10

P5 of 10

110 100 90

Axial Length (mm)

80 80 70 60 50 40 40 30 20 10 0

0

10

20

30

40

50

0

Chamber Radius (mm)

Axial Length (mm)

120

10

20

30

50

40

Chamber Radius (mm)

10

0

30

40

50

20

10

0

Chamber Radius (mm)

P2 of 8

P1 of 8

20

30

40

50

20

30

40

50

Chamber Radius (mm)

Chamber Radius (mm)

P3 of 8

10

0

P4 of 8

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

Fig. 7. First half of the 100 Hz case. Progressing from bottom to top. Abel deconvolved phase averages of CH chemiluminescence and phase averages of SPIV.

120 P1 of 10

100 80 60 40 20 0 −60 −40 −20

0

20

40

60

Chamber Radius (mm)

−60 −40 −20

0

20

40

60

Chamber Radius (mm)

−60 −40 −20

0

20

40

60

Chamber Radius (mm)

−60 −40 −20

0

20

40

60

Chamber Radius (mm) 395

396

P6 of 10

P7 of 10

P8 of 10

P9 of 10

P10 of 10

110 100 90

Axial Length (mm)

80 80 70 60 50 40 40 30 20 10 0

0

10

20

30

40

50

0

Chamber Radius (mm)

Axial Length (mm)

120

10

20

30

50

40

Chamber Radius (mm)

P5 of 8

0

10

20

30

40

50

Chamber Radius (mm)

P6 of 8

10

0

20

30

40

50

Chamber Radius (mm)

P7 of 8

0

10

20

30

40

Chamber Radius (mm)

P8 of 8

100 80 60 40 20 0 −60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

50

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

Fig. 8. Second half of the 100 Hz case. Progressing from bottom to top. Abel deconvolved phase averages of CH chemiluminescence and phase averages of SPIV.

120

P2 of 10

P3 of 10

P4 of 10

P5 of 10

110 100 90

Axial Length (mm)

80 80 70 60 50 40 40 30 20 10 0

0

10

20

30

40

50

0

Chamber Radius (mm)

Axial Length (mm)

120

10

20

30

50

40

P1 of 8

10

0

Chamber Radius (mm)

20

30

40

50

0

Chamber Radius (mm)

P2 of 8

10

20

30

40

50

0

Chamber Radius (mm)

P3 of 8

10

20

30

40

50

Chamber Radius (mm)

P4 of 8

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

Fig. 9. First half of the 200 Hz case. Progressing from bottom to top. Abel deconvolved phase averages of CH chemiluminescence and phase averages of SPIV.

120 P1 of 10

100 80 60 40 20 0 −60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

60

Chamber Radius (mm) 397

398

P7 of 10

P8 of 10

P9 of 10

P10 of 10

110 100 90

Axial Length (mm)

80 80 70 60 50 40 40 30 20 10 0

0

10

20

30

40

0

50

Axial Length (mm)

120

10

20

30

50

40

0

Chamber Radius (mm)

Chamber Radius (mm) P5 of 8

10

20

30

40

50

0

Chamber Radius (mm)

P6 of 8

10

20

30

40

50

Chamber Radius (mm)

P7 of 8

0

10

20

30

40

P8 of 8

100 80 60 40 20 0 −60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

50

Chamber Radius (mm)

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

Fig. 10. Second half of the 200 Hz case. Progressing from bottom to top. Abel deconvolved phase averages of CH chemiluminescence and phase averages of SPIV.

120 P6 of 10

P2 of 10

P3 of 10

P4 of 10

P5 of 10

110 100 90

Axial Length (mm)

80 80 70 60 50 40 40 30 20 10 0

0

10

20

30

40

50

0

Chamber Radius (mm)

Axial Length (mm)

120

10

20

30

50

40

P1 of 8

10

0

Chamber Radius (mm)

20

30

40

50

0

Chamber Radius (mm)

P2 of 8

10

20

30

40

50

0

Chamber Radius (mm)

P3 of 8

10

20

30

40

50

Chamber Radius (mm)

P4 of 8

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

Fig. 11. First half of the 400 Hz case. Progressing from bottom to top. Abel deconvolved phase averages of CH chemiluminescence and phase averages of SPIV.

120 P1 of 10

100 80 60 40 20 0 −60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

60

Chamber Radius (mm) 399

400

P7 of 10

P8 of 10

P9 of 10

P10 of 10

110 100 90

Axial Length (mm)

80 80 70 60 50 40 40 30 20 10 0

0

10

20

30

40

50

0

Chamber Radius (mm)

Axial Length (mm)

120

10

20

30

50

40

0

Chamber Radius (mm)

P5 of 8

10

20

30

40

50

0

Chamber Radius (mm)

P6 of 8

10

20

30

40

50

Chamber Radius (mm)

P7 of 8

0

10

20

30

40

P8 of 8

100 80 60 40 20 0 −60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

50

Chamber Radius (mm)

−60 −40 −20

0

20

40

Chamber Radius (mm)

60

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

Fig. 12. Second half of the 400 Hz case. Progressing from bottom to top. Abel deconvolved phase averages of CH chemiluminescence and phase averages of SPIV.

120 P6 of 10

401

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

mechanical mechanisms with a significant role in the documented instabilities. From the literature [1–11] a clear hierarchy of vortex breakdown behaviour with respect to swirl has been established. Once flow reversal has occurred, increases in swirl cause the formation of a vortex bubble with a recovered vortex tail at its end. Further increases in swirl cause multiple bubbles to form sequentially, eventually leading to the upstream movement of the vortex tail of the final bubble, destroying adjacent bubbles and causing the swelling of the first bubble in the series. As stated in the introduction the internal structure of a vortex bubble has been observed with one or two vortex rings with no theory to predict each scenario. For this work the ratio of tangential to axial velocity at the inlet was used to represent swirl and has been determined and phase averaged to show the effective swirl behaviour during the instabilities. In combination the instantaneous velocity ratios were used to determine the frequency spectra. Figure 13 shows the phase averaged velocity ratio for the three cases and surprisingly shows a single peak oscillation per cycle for all cases. Further similarities include the amplitudes of oscillation and cycle mean values are all approximately equal. This indicates that the values of swirl experienced for all three cases are approximately equal. The fact that only one case is dominated by the additional vortex ring shows that the oscillating swirl values themselves are not responsible for the additional vortex ring and that the 200 Hz case does not experience excessive swirl intensities that could be expected. It is also clear that there is no second swirl frequency present in the 200 Hz case. This is confirmed by the unexpected absence of a 400 Hz harmonic in the velocity ratio spectrum, which was present in the spectra of both velocity components (Fig. 5). However, the velocity ratio spectra for both the 100 Hz and 400 Hz cases have the same harmonics as those detected in the respective velocity component spectrums. Therefore, the inlet swirl conditions are suspected not to be directly responsible for the observed additional vortex ring. The behaviour of the reversed flow region itself was analysed using the phase averaged absolute axial (vy) velocity component taken on the centerline at 30 mm, 50 mm and 70 mm downstream axial positions for each of the three cases. The instantaneous data were used to determine the frequency spectra at these positions. These data are shown in Figs. 14 and 15. For all three cases there

becomes horizontal to match the flow structures and impinges upon the chamber wall, from which the flame structure extends both upstream and downstream. The flame withdraws along this form during these phases until phase 5, where the flame has retreated away from the chamber wall. Phase 6 shows an upstream bulge on the flame structure which corresponded to a wrinkling of the inlet flow structure. Phases 7–8 show the growth of a new flame structure similar to the ‘C’ structures of the 200 Hz case. However, the upstream vortex ring is absent. The bend of the inlet flow structure and flame structure is due to the strength of the single inner toroidal vortex. The phase lag between the heat release rate and flame area profiles shown in Fig. 3 is two positions, with maxima occurring at phases 1 and 3, respectively. The same lag is observed regarding the minima. Of the three different flame/flow behaviours characterised here the 100 Hz case is regarded as simple oscillatory motion. The 200 Hz case is complicated by a second vortex ring forming during part of the instability. This caused additional heat release out of phase with the imposed perturbation. Simple behaviour could have been expected for the 400 Hz case after examination of the pressure and heat release rate profiles but this was not observed. A second vortex ring also formed, albeit much smaller in size and influence. It did not appear to cause any out of phase additional heat release. The capability of the combustion/flow field interaction to respond to the frequency of perturbation manifested itself in the differences of flame and flow field structures. This response was only revealed through the phase locked measurements. To the authors’ knowledge, the presence of an additional vortex ring that exists for only a part of the instability has not previously been reported as a mechanism influencing combustion instabilities. Fundamental vortex breakdown studies have observed steady state bubbles with two rings [4,7,12]. The present novel results highlight the advantage of phase locked measurements and the need for validated LES over simple RANS modelling. The ensemble average results of these data previously presented also do not reveal the behaviour and structures revealed here [55]. 7. Vortex breakdown behaviour The combustor behaviour presented identified the internal recirculation zone and the additional vortex ring as important fluid

vz/vy ratio

100 Hz

200 Hz 0.9

0.9

0.8

0.8

0.8

0.7

0.7

0.7

0.6

0.6

0.6

0.5

0.5

0.5

0.4

2

4

6

8

10

2

4

6

8

10

0.4 2

4

6

Phase

Power (au)

400 Hz

0.9

8

10

2

4

6

8

0.4

10-1

10-1

10-2

10-2

10-2

10-3

10-3

10-3

10-4

10-4

10-4

10-5

10-5

10-5

0

100

200

300

Frequency (Hz)

4

6

400

500

10-6

0

200

400

600

Frequency (Hz)

8

10

2

4

6

8 10

Phase

10-1

10-6

2

10

Phase

800

1000

10-6

0

400

800

1200

1600

Frequency (Hz)

Fig. 13. Phase averaged velocity ratio vz/vy, at the chamber inlet and corresponding frequency spectra for all three cases.

2000

402

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

vy velocity (ms-1)

200Hz

100Hz

45

45

35

35

35

25

25

25

15

15

15

45

50mm

30mm 5

70mm 5

5 2

4

6

8

10 2

4

6

400Hz

8 10

2

4

Phase

6

8

10 2

4

6

8 10

2

4

6

8

10 2

4

6

8 10

Phase

Phase

Fig. 14. Phase averaged |vy| component taken on the chamber centerline at 30 mm, 50 mm and 70 mm axial positions for all three cases.

10

2

10

2

Power (m/s)2s

30mm - 100 Hz

30mm - 400 Hz

1

101

10

1

10

0

100

10

0

10 10

-1

-1

10

-2

10

10

-2

10

-3

0

100

200

300

400

500

10

-3

0

200

Frequency (Hz)

400

600

800

1000

-1

-2

10

-3

0

50mm - 200 Hz

101

101

100

100

100

-1

10-1

10

10-2

10-2

10-2

0

100

200

300

400

500

10-3 0

200

Frequency (Hz) 10

102

70mm - 100 Hz

101

10

400

600

800

1000

10

0

10

70mm - 200 Hz

1

10-1

10-2

10-2

10-2

400

500

10-30

1600

2000

70mm - 400 Hz

10-1

300

1200

101

10-1

200

800

2

10

100

400

Frequency (Hz)

100

10-3 0

2000

-3

0

10

1600

-1

Frequency (Hz)

2

1200

50mm - 400 Hz

101

10-3

800

Frequency (Hz) 102

102

50mm - 100 Hz

10

400

Frequency (Hz)

102

Power (m/s)2s

2

10

10

Power (m/s)2s

10

30mm - 200 Hz

200

Frequency (Hz)

400

600

800

Frequency (Hz)

1000

0

10-3 0

400

800

1200

1600

2000

Frequency (Hz)

Fig. 15. Frequency spectra for the axial velocity component taken on the centerline at 30 mm, 50 mm, and 70 mm axial positions for all cases.

were no dominant frequencies identifiable in the spectra of the radial or tangential velocity components. For the 100 Hz case the amplitude of mean oscillation was approximately the same at the three positions, with a single phase lag (1/10 of a cycle) observed at the 70 mm position. Close to the bluff body, at 30 mm, the cycle average velocity is considerably less than other positions. The dominant frequency is 100 Hz at all positions in this case, with the second harmonic of 200 Hz detected at 30 mm and 50 mm. The 200 Hz case unexpectedly shows a simple, single peaked oscillation at the 30 mm position. At this position the amplitude and cycle average are largest. The 50 mm profile is reduced in amplitude and exhibits a now expected multi peak profile, not too dissimilar to the corresponding heat release profile in Fig. 3. The 70 mm profile is substantially further reduced in amplitude. This is also evident in the velocity spectrum in Fig. 15, where there was no dominant mode at that

position. 200 Hz was the dominant frequency at 30 mm and 50 mm. The only harmonic detected was a very strong 800 Hz frequency at the 50 mm position. The profiles for the 400 Hz case are the flattest of the three experiments. Only the profile at 30 mm exhibited any large oscillatory behaviour. Profiles at 50 and 70 mm are approximately flat. The spectrum at 30 mm reveals a signal at 400 Hz with a surprisingly strong harmonic at 1.6 kHz. This is not present in the heat release spectrum (Fig. 3) but is present in the pressure spectrum of Fig. 2. The 1.6 kHz harmonic is not present in spectra from the 50 mm or 70 mm positions. It is clear that in all experiments the IRZ is oscillating axially at the perturbation frequency. However the majority of the harmonics detected by both the heat release and pressure spectra are absent in the IRZ motion spectra. It is also important to note the axial penetration and magnitude of axial oscillations (at a given position) within the IRZ decrease with increasing perturbation frequency. This

403

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

implies the responses of the vortex breakdown/combustion processes are becoming insufficient to respond simply to higher frequencies. It has been shown that neither the inlet swirl nor IRZ behaviour are directly responsible for the formation of the additional vortex rings, although undoubtedly they are involved. Swirl itself affects the axial pressure gradient. Other factors affecting this include geometry, but as this is constant will not be considered in the comparison of the three cases. Other influencing factors include the radial redistribution of flow away from the axial direction and the response of the combustion process itself. It is suspected that one or a combination of these two factors could trigger the formation

of additional vortex rings. The radial and tangential components in each case were in phase with each other while both were out phase with the axial component. This phase difference was exactly half a perturbation period in the 200 Hz case and is the reason the velocity magnitude profile of Fig. 7 shows a twin peak profile. As both the 100 Hz and 400 Hz profiles showed radial components out of phase with the axial, it is not possible to be certain this was the reason for the second vortex ring, although it can be regarded as a contributing factor as it will strengthen the axial pressure gradient. It is worth noting that many fundamental vortex breakdown studies investigate steady state swirling flows and not forcibly perturbed flows [1–11].

60ms−1

45

Axial Length (mm)

40 35 30 25 20 15 10 5 0

10

20

30

40

50

Chamber Radius (mm)

Γ1

45

Γ2

45

Identified vorticies

45

1 0.8

Axial Length (mm)

0.6 35

35

35

25

25

25

0.4 0.2 0 -0.2

15

15

15

5

5

5

-0.6

0

10

20

30

40

0

50

Chamber Radius (mm)

10

20

30

45

40

-0.8

50

0

Chamber Radius (mm)

Fig. 16. Instantaneous vector field from imaging region CRZ from the 200 Hz case with only Bottom right: identified vorticies.

Axial Length (mm)

-0.4

45

P1

1=

4

10

20

30

40

50

-1

Chamber Radius (mm)

of the vectors displayed. Bottom left: C1 scalar, Bottom centre: C2 scalar.

45

P3

0.6

P5

0.4 35

35

35

0.2 25

25

0

25

-0.2 15

15

15

-0.4 -0.6

5

5

0

10

20

30

40

50

5

0

Chamber Radius (mm)

Axial Length (mm)

45

10

20

30

40

50

0

45

P1

45

P3

35

35

25

25

25

15

15

15

5

5 10

20

30

40

Chamber Radius (mm)

50

20

30

40

50

Chamber Radius (mm)

35

0

10

Chamber Radius (mm)

P5

5 0

10

20

30

40

Chamber Radius (mm)

50

0

10

20

30

40

50

Chamber Radius (mm)

Fig. 17. Phase average C2 images and vortex location maps of the imaging region CRZ for the 200 Hz case for phases 1, 3, and 5. ‘+’ anticlockwise ‘o’ clockwise.

404

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

8. Vortex shedding The phase averaged streamline plots of Figs. 7–12 show no evidence of vortex shedding with no vortex formation attached to the annulus, either on the inner or outer diameter, of the type that could be described as coherent vortex shedding [25,26] or shear layer instabilities [15,16,22]. In order to verify this, a vortex identification study was performed. Many different algorithms are available such as the Q criterion [59] or lambda 2 criterion [60]. Common to all is the comparison between rotation and strain of the velocity gradient tensor. The specific approach used was that proposed by Graftieaux et al. [61]. Two scalars are derived. The first, C1, is used to locate the individual vortex centre. The second, C2, is used to compare local principle strain rate and local rotation. An example of an instantaneous vector field with the calculated scalars and identified vortices are shown in Fig. 16. This approach was applied to all instantaneous vector fields from all three imaging regions for all three cases presented. Results from the 200 Hz case from the CRZ imaging area for three phases are shown in Fig. 17. The mean C2 images clearly show that the scalar is strongest on the shear layers attached to the annulus, very much akin to a pure vorticity image (see Section 9) and proving that these regions are dominated by vorticity rather than principle strain. There is clearly no coherent vortex rollup attached to the annulus. The vortices identified are clearly restricted to the shear layers (anticlockwise on the inner layer, clockwise on the outer layer) and the vortex rings identified in Figs. 7–12, coinciding with the CRZ vortex ring and vortex breakdown vortex rings. There is no evidence of a cluster of vortices forming on the annulus and moving downstream with phase that would be expected with coherent vortex shedding or shear layer instabilities. This was true for all three cases. It is also worth noting that an investigating of open swirling jet flames undergoing perturbations yielded an inverse relationship between the forcing amplitude and frequency of perturbation [26]. Plotted as a function of Strouhal number the three cases here correspond to Strouhal numbers of 0.043, 0.084, and 0.164 for the 100 Hz, 200 Hz and 400 Hz cases. With perturbation intensities of 0.14, 0.12, and 0.15 these cases fall below the required intensities where coherent vortex roll up and shedding can be expected to occur. 9. Vorticity structures The geometry of the chamber and use of vortex breakdown as a fluid mechanic stabilising mechanism not only creates the flow reversal and the recirculation zones described, but also creates vorticity and strain structures. These structures will impart normal and tangential strains (with respect to the flame normal) upon the flame, termed flame stretch, and alter the local FSD. While experiencing an instability, not only do the recirculation zones change but the shape and strength of the strain inducing structures do also, thereby altering the local flame–turbulence interaction and flame stretch, particularly when the new toroidal vortex rings form close to the inlet. It is this mechanism that physically causes the varying heat release rate and flame area. The changes to these structures are controlled by the varying inlet flow velocity and the axial oscillation of the flow reversal region. The SPIV used in this investigation yielded all three velocity components in a single plane but only allows the derivation of six components of the stress–strain (deformation) tensor. Tomogrpahic PIV would be required to complete the tensor [58]. As an example, a single phase from each of the three experiments has been converted to vorticity, as shown in Fig. 18. These images immediately show the presence of two oppositely rotating vorticity structures originating at the annulus. The first, an anticlockwise rotating structure (negative vorticity), is created by the inlet flow and flow reversal region and is attached at the edge of

the bluff body. The second, a clockwise rotating structure (positive vorticity), is formed by the inlet flow and the slower fluid close to the dump plane. This is attached to the outer diameter of the annulus. These two structures coincide with the shear layers either side of the inlet flow, and the flame front will exist at some location within each structure. This has been previously observed by LES [22]. Great effort was given to try and identify a vortex core wrapped around the reversed flow region. No evidence of the core or any precessing effects was identifiable using the instantaneous or phase locked data. As it has been shown that direct motion of the IRZ is not responsible for all of the combustion harmonics detected, the available vorticity data are considered. Spectra of vorticity from the imaging area, Annulus 3D, was determined at all vector positions. This imaging region was chosen to avoid the effect of flame generated vorticity, as it impossible to identify which vector positions are fresh or burnt gas without simultaneous OH PLIF as a flame front marker [17]. The spectra were obtained as previously described. For each spectrum a custom algorithm was used to identify the dominant frequency present, provided it was a factor of 2 greater than 3 times the rms plus the mean value of the spectrum. Once identified, the mean and rms values of the spectrum were re-calculated, omitting the value of the previously identified frequency, and the next strongest frequency was identified. This process was repeated to identify up to four frequencies, provided they all satisfied the minimum power criterion. This is shown in Fig. 19. This method was chosen over a proper orthogonal decomposition (POD) method [17,18] in this investigation for a number of reasons. Firstly, the chamber was interrogated in distinct imaging regions, none of which imaged the entire chamber, which yields discontinuities if instantaneous snapshots are merged. Additionally, the same mode number for each region will not align across the image boundary as images of statistical derived moments do. Secondly the number of vector fields, 1024, is arguably not sufficient to achieve convergence of energy over the derived modes. Thirdly, establishing the correlation between calculated modes and spectral frequencies is not possible; particularly as the modes themselves are spatial averages of the given field of view. For the 100 Hz case the dominant vorticity mode was 100 Hz for the inlet flow and the shear layers either side. The second mode however clearly shows 200 Hz, 300 Hz, and 400 Hz frequencies which were detected by the pressure transducer (and heat release spectrum). The 200 Hz frequency is present across the shear layers and inlet flow. However the 300 Hz and 400 Hz frequencies are restricted to the shear layers. Therefore, the shear layers are responsible for generating vorticity at these frequencies, which will alter the flame–turbulent interactions and flame stretch resulting in the subsequent heat release harmonics. For the 200 Hz and 400 Hz cases the dominant mode is mostly the imposed frequency. However, in both the inner and outer shear layers the dominant mode is in fact an integer harmonic, 400 Hz and 800 Hz, respectively. The frequencies of the second mode correspond to the detected integer harmonics of each case. This reinforces the importance of the shear layers in terms of instability harmonics and the above processes acting through vorticity which are the physical processes affecting the flame front. 10. Conclusions A premixed methane/air mixture with an equivalence ratio of

u = 0.8 was supplied to an atmospheric swirl stabilised dump combustor with optical access. The flow was perturbed prior to entry into the chamber at frequencies of 100 Hz, 200 Hz, and 400 Hz. Pressure and temperature measurements were made upstream of the chamber. Temperature within the ceramic inlay embedded in

405

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407 100 Hz - P2

400 Hz - P2

200 Hz - P2

120

120

120

110

110

110

100

100

100

90

90

90

80

80

Axial Length (mm)

0

10

20

30

40

80 10

0

50

20

30

40

50

0

80

80

80

70

70

70

60

60

60

50

50

50

40

40 0

10

20

30

40

10

20

30

40

50

40

40

40

30

30

30

20

20

20

10

10

10

0

0 0

10

20

30

40

50

20

30

40

50

0

10

20

30

40

50

0

10

20

30

40

40 0

50

10

0

10

Chamber Radius (mm)

20

30

40

0

50

Chamber Radius (mm)

50

Chamber Radius (mm)

Fig. 18. Counter rotating vorticity structures of SPIV phase 2 from each case. All images artificially saturated at 4000 s1 (blue) to 4000 s1 (red). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 100 Hz Vorticity - Mode 1

200 Hz Vorticity - Mode 1 30

20

10

0

10

20

30

20

10

0

40

30

Axial Length (mm)

Axial Length (mm)

Axial Length (mm)

30

0

0

10

Chamber Radius (mm)

30

0

40

10

20

30

40

20

10

0

100

200

300

Frequency (Hz)

400

30

40

0

10

20

30

40

20

10

0

0

10

Chamber Radius (mm)

500

20

30

Chamber Radius (mm)

0

10

400 Hz Vorticity - Mode 2

Axial Length (mm)

Axial Length (mm)

20

10

0

Chamber Radius (mm)

30

0

10

200 Hz Vorticity - Mode 2

30

Axial Length (mm)

20

20

Chamber Radius (mm)

100 Hz Vorticity - Mode 2

0

400 Hz Vorticity - Mode 1

0

200

400

600

800

20

30

40

Chamber Radius (mm)

1000

0

400

Frequency (Hz)

Fig. 19. Spatial location of the first two vorticity modes for all cases.

800

1200

Frequency (Hz)

1600

2000

406

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407

the dump plane was also recorded. Phase locked high speed CH chemiluminescence and high speed stereo PIV were used to characterise the chamber behaviour and inlet conditions. All three forcing cases had distinctly unique flame and flow structure behaviours and shapes. The 100 Hz case exhibited a simple oscillatory motion as revealed by the upstream pressure analysis, heat release rate analysis, and inspection of the phase averaged chemiluminescence and SPIV images. Typical flow features observed were the toroidal corner recirculation zone and the flow reversal region of the inner recirculation zone associated with the vortex breakdown with a single internal toroidal vortex ring. The 200 Hz case was greatly altered by the formation of a second toroidal vortex ring within the internal recirculation zone. This formed upstream, adjacent to the inner shear layer during part of the instability and was responsible for additional heat release and kept the flame area large due to the additional flame stretch. The internal recirculation zone of the 400 Hz case also experienced a second vortex ring, smaller than that observed in the 200 Hz case. To our knowledge, this is the first work to reveal additional vortex rings forming during an instability and to show that they are an additional fluid mechanical feature with an important role in combustion instabilities. The vortex breakdown was characterised by analysing the oscillations of local swirl values at the inlet and the oscillations of the IRZ. The local inlet swirl values in all three cases were found to be simply oscillating at the imposed perturbation frequency. There was no evidence of an unusually large swirl value or second swirling frequency that might be expected to cause the formation of the new vortex rings. It is hypothesised that either the ratio of radial to axial velocity or the influence of combustion could be factors affecting the axial pressure gradient and triggering the formation of the new vortex ring. Results of a vortex identification analysis revealed no evidence to suggest coherent vortex rollup or shear layer instabilities occurring to explain the formation of the new vortex rings. The IRZ in all three cases was oscillating solely in the axial direction at the imposed perturbation frequencies and a limited number of harmonics at specific locations. It was found that the axial penetration and magnitude of oscillation were both reduced with increased frequency. The flow reversal region of the vortex breakdown, inlet flow, and sudden expansion created two oppositely rotating vorticity structures (rotating about the tangential axis) attached to the annulus. These were used to identify regions where fluid mechanical strain and flame–turbulence interactions impact the flame front. Spectral analysis of vorticity revealed the integer harmonics detected by pressure and heat release rate measurements were generated at the shear layers. In parallel to the physical insights presented the chamber behaviour for the three forcing cases have been presented in great detail including boundary conditions, inlet characterisation, flame and flow field characterisation, quantitative velocity profiles, and vorticity characteristics in order to provide a comprehensive data set for CFD validation. Acknowledgments This work was funded by the EPSRC and conducted at Cranfield University, UK. Many insightful discussions were had with Sandia National Laboratory scientists, Robert Barlow in particular.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.combustflame. 2014.07.016.

References [1] J.M. Beer, N.A. Chiger, Combustion Aerodynamics, John Wiley and Sons, New York, US, 1972. [2] N. Syred, Prog. Energy Combust. Sci. 32 (2006) 93–161. [3] Y. Huang, V. Yang, Prog. Energy Combust. Sci. 35 (2009) 293–364. [4] O. Lucca-Negro, T. O’Doherty, Prog. Energy Combust. Sci. 27 (2001) 431–481. [5] W.L.H. Hallet, D.J. Towes, Exp. Fluids 5 (1987) 129–133. [6] M.G. Hall, Proc. Heat Trans. Fluid Mech. Inst. (1967) 319–340. [7] C. Brucker, W. Althaus, Exp. Fluids 13 (1992) 339–349. [8] T. Sarpkaya, J. Fluid Mech. 45 (3) (1971) 545–559. [9] T. Sarpkaya, AIAA J. 9 (9) (1971) 1792–1799. [10] T.B. Benjamin, J. Fluid Mech. 14 (1962) 593–629. [11] E. Serre, P. Bontoux, J. Fluid Mech. 459 (2002) 347–370. [12] M. Escudier, Prog. Aero. Sci. 25 (1988) 189–229. [13] P. Wang, X.S. Bai, Int. J. Num. Meth. Fluids 47 (2005) 99–120. [14] P. Wang, X.S. Bai, J. Klingman, Phys. Fluids 16 (9) (2004) 3306–3324. [15] C.E. Martin, L. Benoit, Y. Sommerer, AIAA J. (44) (2006) 741–750. [16] Y. Huang, H.G. Sung, S.Y. Hsieh, V. Yang, J. Propul. Power 19 (5) (2003) 782– 794. [17] I. Boxx, M. Stöhr, C. Carter, W. Meier, Combust. Flame 157 (2010) 1510–1525. [18] J. O’Conner, T. Lieuwen, Proc. ASME Turbo expo., Copenhagen, Denmark, June 11–15, 2012, GT2012-70053. [19] S. Wang, Phys. Fluids 17 (2005) 045107. [20] S. Roux, T. Lartigue, T. Poinsot, U. Meier, C. Berat, Combust. Flame 141 (2005) 40–54. [21] L. Rayleigh, Nature 8 (1878) 319–321. [22] F.F. Grinstein, C. Fureby, Proc. Combust. Inst. 30 (2005) 1791–1798. [23] J. O’Conner, T. Lieuwen, J. Eng. Gas Turbines Power 134 (2012) 011501. [24] J. O’Conner, T. Lieuwen, Phys. Fluids 24 (2012) 075107. [25] R. Balachandran, O.B. Ayoola, C.F. Kaminski, A.P. Dowling, E. Mastorakos, Combust. Flame 143 (2005) 37–55. [26] C. Kulsheimer, H. Buchner, Combust. Flame 131 (2002) 70–84. [27] A.M. Steinberg, I. Boxx, M. Stöhr, C. Carter, W. Meier, Combust. Flame 157 (2010) 2250–2266. [28] S. Leibovich, Annu. Rev. Fluid Mech. 10 (1978) 221–246. [29] J.G. Lee, K. Kim, D.A. Santavicca, Proc. Combust. Inst. 28 (2000) 415–421. [30] T. Lieuwen, V. McDonell, E. Perterson, D. Santavicca, Proc. ASME Turbo expo. Barcelona, Spain, May 6–11, vol. 1, 2006, pp. 601–615. [31] N.A. Bunce, B.D. Quay, D.A. Santavicca, Proc. ASME Turbo expo., San Antonio TX, June 3–7, 2013, 657–668, GT2013-95812. [32] K.T. Kim, D.A. Santavicca, Combust. Flame 160 (8) (2013) 1441–1457. [33] K.T. Kim, J.G. Lee, B.D. Quay, D.A. Santavicca, Combust. Sci. Technol. 183 (2) (2010) 122–137. [34] D. Durox, J.P. Moeck, J.-F. Bourgouin, P. Morenton, M. Viallon, T. Schuller, S. Candel, Combust. Flame 160 (2013) 1729–1742. [35] A. Chaparro, E. Landry, B.M. Cetegen, Combust. Flame 145 (2006) 290–299. [36] K.T. Kim, J.G. Lee, B.D. Quay, D.A. Santavicca, Combust. Flame 157 (2010) 1718–1730. [37] B.D. Bellows, M.K. Bobba, A. Forte, J.M. Seitzman, T. Lieuwen, Proc. Combust. Inst. 31 (2007) 3181–3188. [38] B.D. Bellows, M.K. Bobba, J.M. Seitzman, T. Lieuwen, Trans. ASME 129 (2007) 954–961. [39] D. Kim, J.G. Lee, B.D. Quay, D.A. Santavicca, K. Kim, S. Srinivasan, J. Eng. Gas Turbines Power 132 (2010) 021502. [40] K.T. Kim, D.A. Santavicca, Combust. Sci. Technol. 185 (7) (2013) 999–1015. [41] B. Schuermans, F. Guethe, D. Pennell, D. Guyot, C.O. Paschereit, J. Eng. Gas Turbines Power 132 (2010) 111503. [42] A.G. Hendricks, U. Vandsburger, Exp. Thermal Fluids Sci. 32 (2007) 126–132. [43] T. Sidwell, G. Richards, K. Castleton, D. Straub, D. Maloney, P. Strakey, D. Ferguson, S. Beer, S. Woodruff, AIAA J. 44 (2006) 434–443. [44] K.K. Venkataraman, L.H. Preston, D.W. Simons, B.J. Lee, J.G. Lee, D.A. Santavicca, J. Propul. Power 15 (6) (1999) 909–918. [45] Q. Zhang, D.R. noble, A. Meyers, K. Xu, T. Lieuwen, Proc. ASME Turbo expo., Reno NV, June 6–9, vol. 2, 2005, pp. 657–668. [46] P.M. Anacleto, E.C. Fernandes, M.V. Heitor, S.I. Shtork, Combust. Sci. Technol. 175 (2003) 1369–1388. [47] R.C. Orbay, K.J. Nogenmyr, J. Klingmann, X.S. Bai, Fuel 104 (2013) 133–146. [48] J.C. Broda, S. Seo, R.J. Santoro, G. Shirhattikar, V. Yang, Proc. Combust. Inst. 27 (1998) 1849–1856. [49] S.Y. Lee, S. Seo, J.C. Broda, S. Pal, R.J. Santoro, Proc. Combust. Inst. 28 (2000) 775–782. [50] W. Meier, I. Boxx, M. Stöhr, C. Carter, W. Meier, Exp. Fluids 49 (2010) 865–882. [51] W. Meier, P. Weigand, X.R. Duan, R. Giezendanner-Thoben, Combust. Flame 150 (2007) 2–26. [52] M. Stöhr, I. Boxx, C. Carter, W. Meier, Proc. Combust. Inst. 33 (2011) 2953– 2960. [53] M. Stöhr, R. Sadanandan, W. Meier, Proc. Combust. Inst. 32 (2009) 2925–2932. [54] P. Weigand, W. Meier, X.R. Duan, R. Giezendanner-Thoben, U. Meier, Flow Turbul. Combust. 75 (2005) 275–292. [55] A. Ruggles, J. Kelman, Combust. Sci. Technol. 186 (3) (2014) 313–331.

A. Ruggles, J. Kelman / Combustion and Flame 162 (2015) 388–407 [56] T. Poinsot, D. Veynante, Theoretical and Numerical Combustion, second ed., R.T, Edwards, Philadelphia, US, 2005. [57] L.H. Benedict, R.D. Gould, Exp. Fluids 22 (1996) 129–136. [58] B. Coriton, A.M. Steinberg, J.H. Frank, Exp. Fluids 55 (2014) 1743.

407

[59] J. Jeong, F. Hussain, J. Fluid Mech. 285 (1995) 69–94. [60] J.C.R. Hunt, A.A. Wray, P. Moin, CTR-S88, 1988. [61] L. Graftieaux, M. Michard, N. Grosjean, Meas. Sci. Tech. 12 (2001) 1422–1429.