Aerospace Science and Technology 43 (2015) 388–394

Contents lists available at ScienceDirect

Aerospace Science and Technology

www.elsevier.com/locate/aescte

Combustion noise characteristics of an experimental 2D trapped

vortex combustor
P.K. Ezhil Kumar, D.P. Mishra ∗
Combustion Lab, Department of Aerospace Engineering, Indian Institute of Technology, Kanpur, 208016, India

a r t i c l e i n f o a b s t r a c t

Article history: The objective of present investigation is to understand the effects of various operating parameters on
Received 13 November 2014 the combustion noise characteristics of a 2D trapped vortex combustor (TVC). Spectral analysis of the
Received in revised form 14 March 2015 microphone signal indicates that the power spectral density (PSD) magnitude increases with primary air
Accepted 18 March 2015
velocity (V p ) for both mainstream Reynolds number (Rems ) 20 000 and 40 000 cases considered in this
Available online 24 March 2015
study. Combustion noise measurements reveal that the spectral amplitude of the combustor increases
Keywords: with cavity equivalence ratio (φc ). Moreover, for a particular V p case, the PSD magnitude is observed
Trapped vortex to be higher for Rems = 20 000 as compared to that of Rems = 40 000 case due to the quenching of
Combustion noise mainstream flame in the latter case. For a particular mainstream Reynolds number, the noise level
Momentum flux ratio increases with the momentum flux ratio between cavity and mainstream flow. Analysis of results reveals
that the noise level varies between 80 dB for non-merged flame and 100 dB for merged flame cases.
An attempt has also been made to correlate the overall sound pressure level (OASPL) with the non-
dimensional operating variables.
© 2015 Elsevier Masson SAS. All rights reserved.

1. Introduction TVC brought out the difference in frequency spectrum between the
reacting and non-reacting flow noise. In this study, the mainstream
Aviation and power generation industries often use swirl sta- flow was not premixed. In order to vary the cavity equivalence
bilized combustors where, the pilot flame is directly exposed to ratio, they fixed the cavity fuel flow rate while varying the pri-
the mainstream flow [1] and as a result, their operating range mary air flow rate. For non-reacting flow cases, they observed a
is quite narrow [1,2]. In order to extend the operating range of broad band spectrum in the frequency domain. Similarly, a broad
the combustor, the pilot flame has to be well shielded from the band spectrum was observed under lean operating conditions as
mainstream flow. Cavity can be utilized to shield the pilot flame well. On the other hand, a distinct peak in spectral amplitude was
from the mainstream flow and the combustor using this concept is observed in the frequency spectrum for rich operating condition.
termed as trapped vortex combustor (TVC) [2,3]. In TVC, fuel rich They argued that the cause for this peak in spectral amplitude
mixtures are usually maintained at the cavity zone to establish a might be due to the coupling between the heat release and the
pilot flame and this flame is responsible for sustaining combustion flow. However, they did not provide any support for this argument.
at the mainstream region [4,5]. It can be noted that the nature of In another study, Gutmark and co-workers [10] investigated the ef-
flow and flame prevailing in the TVC cavity is highly sensitive to
fects of two injection strategies namely the axial and tangential
the cavity aspect ratio [3,4,6], momentum flux ratio between the
air injection cases on the noise level of a can type external cav-
cavity and the mainstream flow [4,7]. It was demonstrated that the
ity TVC. Their study revealed that the noise level for tangential air
secondary air jet plays a vital role in altering the cavity flow struc-
injection was considerably higher than that of the axial injection
ture under reacting flow conditions [6,7].
case. Besides this, they also observed a weak coupling between
Combustion noise is one of the main sources of gas turbine
the OH∗ chemiluminescence and pressure signals for tangential in-
engine noise [8,9,1]. In the context of trapped vortex combustor,
jection case. On the other hand, axial injection case did not exhibit
limited studies are reported for combustion noise; for example,
any coupling between the heat release and acoustic signal [10].
Sturgess and Hsu [9] through their experiments in an axisymmetric
Through that study, they also brought out the role of overall equiv-
alence ratio on the noise level. Apart from these two studies, no
* Corresponding author. Fax: +91 512 2597561. work is available for combustion noise study in a trapped vortex
E-mail address: mishra@iitk.ac.in (D.P. Mishra). combustor.

http://dx.doi.org/10.1016/j.ast.2015.03.014
1270-9638/© 2015 Elsevier Masson SAS. All rights reserved.
 P.K. Ezhil Kumar, D.P. Mishra / Aerospace Science and Technology 43 (2015) 388–394 389

Fig. 2. (a) Experimental setup showing the combustor inside the semi-anechoic
chamber for combustion noise measurement and (b) schematic showing various an-
gular positions and y locations for microphone data acquisition.

Fig. 1. (a) 3D schematic of the TVC considered in the present investigation and
(b) injection strategy adopted in the present work. can be noted that in the present investigation, fuel (compressed
natural gas, CNG with 96% methane content) and air are injected
into the cavity only from the aft wall. Mainstream air is supplied
From the reviewed literature on the noise characteristics of TVC,
from a high pressure blower and the flow rate is measured using
it is brought out that limited studies have been carried out by only
an orifice flow meter (C d = 0.65).
two groups. In both works the noise level for various operating
Air for cavity (primary) injections is fetched from a compres-
conditions are not quantified. It is very important to separate the
sor. The air flow pressure is regulated using a pressure regulator
hydrodynamic noise from the combustion noise, which is not being
and the flow rate is measured using a calibrated mass flow me-
considered in the previous studies. Besides this, it will be interest-
ing if a scaling law could be brought out for the noise level data, ter. It can be noted that atmospheric pressure is maintained at all
which has not been reported for TVC to the best of our knowl- inlets. In order to visualize the flow and flame prevailing inside
edge. Interestingly, Shivashankara and Crouch [8], long time ago in the cavity, the front side of the combustor was equipped with a
1977 brought out correlations between sound power and operat- quartz window. The injection scheme used in the present work is
ing variables for a gas turbine combustor. Similar correlations are depicted in Fig. 1b. In the diagram, ‘L’ and ‘D’ refer to the cav-
not available for trapped vortex combustors and such correlations ity length and depth respectively. In the present investigation, the
may be helpful to relate the influence of various parameters on the aspect ratio (AR), L / D is maintained at 1.2.
sound pressure level. In the previous paragraph, it is mentioned
that the primary air velocity is instrumental in the ignition of 2.2. Test facility
mainstream fuel–air mixture. Hence, in the present investigation,
the effects of primary air velocity, mainstream Reynolds number In order to avoid the interference of ambient noise sources, the
and cavity equivalence ratio on the combustion noise character- combustor is kept in an enclosure, known as semi-anechoic cham-
istics of the 2D trapped vortex combustor (TVC) are brought out ber, made of sound absorbing materials as shown in Fig. 2a. The
under non-reacting and reacting flow conditions. outer structure of the enclosure is made up of hardboard material
with 3.6 cm sound absorbing material (polyurethane foam) lining.
2. Experimental details The inner structure is also made up of polyurethane foam with air
gap between the inner structure and the outer enclosure. It can be
2.1. The trapped vortex combustor noted that the polyurethane foam in the inner enclosure is corru-
gated shaped blocks of 0.3 × 0.3 × 0.3 m3 .
Fig. 1a shows the 3D schematic view of the trapped vortex At the top of the enclosure, an opening of around 0.5 × 0.5 m2
combustor with the fuel and air injection locations. The combustor is made for the hot gases to escape out of the enclosure (not
consists of three sections namely, upstream section, cavity section shown in Fig. 2a). Besides this the side portion (floor) of the en-
and the downstream section. The cavity is bounded by fore-wall, closure is also covered with sound absorbing material. In order to
aft-wall and side-wall as shown in Fig. 1a. Provisions are made to avoid reflections from the metal surfaces, all the metal surfaces
adjust the cavity size by sliding the bottom wall of the cavity. It inside the enclosure are covered with polyurethane foam. GRAS
390 P.K. Ezhil Kumar, D.P. Mishra / Aerospace Science and Technology 43 (2015) 388–394

microphone (Model 46 BP, sensitivity: 1.52 mV/Pa) along with a

GRAS preamplifier (Model 26TC) is used for noise measurements.
It can be noted that the microphone is having a flat response in
the frequency band 250 Hz to 20 kHz and the measurement uncer-
tainty is ±0.06 dB. The signals from the microphone are acquired
at a frequency of 10 000 Hz with the help of a 24 bit resolution NI
4462 DAQ card with a built-in anti-aliasing filter. A total number
of 65 536 (216 ) samples is acquired for 32.76 s and the frequency
resolution in this case is calculated to be around 0.03 Hz. Vari-
ous microphone positions are shown in Fig. 2b and the reason for
acquiring acoustic data from these locations are discussed in the
following section.

3. Results and discussion

In the present investigation, two mainstream Reynolds num-

bers (Rems ) are considered namely Rems = 20 000 and 40 000. For
each Rems case, the primary air velocity (V p ) is varied from 40
to 140 m/s while cavity equivalence ratio (Φc ) is varied from 1.0
to 1.4. Noise data from the combustor is acquired under both
non-reacting and reacting flow conditions. It can be noted that
location of microphone may affect the acquired noise data sig-
nificantly [11–14]. Hence, an attempt has been made to acquire
acoustic data at different angular locations and y locations nor-
mal to flow direction (x axis) as indicated in Fig. 2b (previous
section). Spectral analysis of the acquired data reveals that the an-
gular location is not affecting the signal significantly as shown by
the spectral plots in Fig. 3a. Hence, the microphone is located at
an angle of 90◦ from the combustor axis. This location is chosen
in order to prevent damage of microphone by hot exhaust gases.
Similarly, in order to identify the contamination of hydrodynamic
noise in the signal, noise data is acquired while moving the mi-
crophone away from the combustor axis as shown in Fig. 3b. From
this study, it is decided that to avoid the hydrodynamic noise, the
microphone has to be located at a distance 40 cm from the com-
bustor axis. In order to circumvent the hydrodynamic noise con-
tamination, Choy et al. [14] located the microphone 50 cm away
from the burner axis. Similarly, Prakash et al. [12] in their acous-
tic studies located the microphone 61 cm away from the burner
axis.

3.1. Effects of primary air velocity Fig. 3. (a) Power spectrum of the acoustic signal at various angular positions, and
(b) power spectrum of the acoustic signal at various y locations under non-reacting
flow condition.
The power spectral density plots of the noise signal for two
Rems cases 20 000 and 40 000 are shown in Fig. 4. The spectra
fuel–air mixture [15], and due to this, the heat release is likely
shown in Fig. 4a correspond to the noise generated within the
to get enhanced. This causes an increase in heat release as evident
combustion chamber. From this plot, it can be inferred that for a
from widening of CH∗ intensity profile at mainstream region (see
particular cavity equivalence ratio case, the energy content of the
noise signal increases with the primary air velocity (V p ). Increase Fig. 6). For Rems = 40 000; V p = 60 m/s case, weak non-merged
in the primary air velocity is likely to increase the heat release rate flame is observed in the mainstream region as evident from Fig. 5d.
(see Fig. 6) and as a result, the noise level also increases. Besides Due to the presence of week mainstream flame, noise level is rel-
this, with increase in the primary air velocity, a shift in the fre- atively low as shown in Fig. 4b. At higher V p cases (above 100
quency around 2000 Hz is evident as shown in Fig. 4a. This shift m/s), presence of merged main flame (see Fig. 5g) results in higher
in the frequency might be attributed to the enhanced interaction combustion noise as shown in Fig. 4b.
between the mainstream flame and the cavity flame, particularly As mentioned in the previous paragraphs, combustion noise
at higher primary air velocities. Power spectra for Rems = 40 000 level increases due to the increase in heat release rate. In order
case (see Fig. 4b) and its variation with V p is very similar to to bring out the influence of primary air velocity on heat release
Rems = 20 000 case. rate, CH∗ chemiluminescence intensity profile within the cavity for
In order to demonstrate the variation of flame appearance, various primary air velocity cases are plotted in Fig. 6. It can be
instantaneous flame images are compared for different primary noted that use of CH∗ signatures from flame as a marker for heat
air velocity cases in Fig. 5. It can be noted from Fig. 5a that release in natural gas–air flames is well established in literature
for Rems = 20 000; V p = 60 m/s case, both the cavity flames re- [16] and [17]. For Rems = 20 000, Φc = 0.3 case, at V p = 40 m/s,
main non-merged at locations downstream the cavity trailing edge. CH∗ intensity value is maximum at the cavity–mainstream inter-
However, for higher V p cases flame from both the cavities ap- face region (shear layer). However, at locations y = −30 and −5,
proach together and thus forming a merged flame as shown in CH∗ intensity value attains its minimum. As V p is increased to
Fig. 5b–c. This merged flame subsequently ignites the incoming 50 and 60 m/s, CH∗ intensity magnitude within the cavity region
 P.K. Ezhil Kumar, D.P. Mishra / Aerospace Science and Technology 43 (2015) 388–394 391

Fig. 5. Instantaneous direct flame images for different operating conditions namely
(a)–(c) Rems = 20 000; V p = 60, 80 and 100 m/s, (d)–(g) Rems = 40 000; V p = 60,
80, 100 and 130 m/s.

Fig. 4. Power spectral density of combustion noise signal for two mainstream
Reynolds number cases (a) Rems = 20 000, and (b) Rems = 40 000.

increases as shown in Fig. 6. CH∗ intensity value is expected to in-

crease further as V p is increased to higher values. The increase in
heat release for higher V p cases contribute to combustion noise as Fig. 6. CH∗ intensity variation along y location (at x = 42 mm) within the cavity for
discussed in the previous paragraph. Rems = 20 000 and for various primary air velocity cases.

3.2. Effects of mainstream Reynolds number mainstream flow interaction noises, which needs further investiga-
tions.

In order to understand the role of mainstream Reynolds num- 3.3. Effects of cavity equivalence ratio
ber on the sound emission level, frequency spectrum of two
Reynolds numbers at V p = 60 and 100 m/s are plotted in Fig. 7a Effects of cavity equivalence ratio on the power spectral den-
and b respectively. It is interesting to note from Fig. 7a that until sity for two mainstream Reynolds number cases namely (i) Rems =
1000 Hz, both the Reynolds number cases exhibit similar behav- 20 000, (ii) Rems = 40 000 are shown in Fig. 8a and b respectively.
ior. Beyond 1000 Hz, the nature is different, since Rems = 40 000 For Rems = 20 000 case, when the cavity equivalence ratio is in-
case exhibits distinct spikes at certain frequencies, which are ab- creased from 1.0 to 1.4, the magnitude of the power spectrum gets
sent for Rems = 20 000 case. Similarly for primary air velocity enhanced as indicated in Fig. 8a. It can be noted that with in-
of 100 m/s, until 1000 Hz, the frequency spectra is similar for crease in cavity equivalence ratio (Φc ), the heat release magnitude
both Reynolds number cases, however with a different magni- will increase, and hence, increase in the sound level at higher Φc
tude (see Fig. 7b). Beyond 1000 Hz, distinct peaks are observed is expected. Increase of heat release magnitude with equivalence
at certain frequencies; the frequency values are different for both ratio is discussed in the recent works of Jeong et al. [16] and Orain
Reynolds number cases. Occurrences of peak values in amplitude and Hardalupas [17]. Rems = 40 000 case also exhibits similar trend
at higher frequencies are likely to be caused due to the cavity– discussed for Rems = 20 000 case (see Fig. 8b).
392 P.K. Ezhil Kumar, D.P. Mishra / Aerospace Science and Technology 43 (2015) 388–394

level increases with the primary air velocity. Under non-merged

flame conditions, the overall sound pressure level is of the or-
der of 85 dB, and at merged flame conditions, the OASPL value
reaches as close as 100 dB. Besides this, the influence of momen-
tum flux ratio (MFR) on the OASPL value is shown in Fig. 9b. The
OASPL value increases proportionally with the MFR between the
cavity and the mainstream flow. Moreover, for a particular MFR
value, OASPL increases with mainstream Reynolds number (see
Fig. 9b).
From the above discussion, it is brought out that the overall
sound pressure level is influenced by the mainstream Reynolds
number, cavity power level and the relative momentum flux be-
tween the cavity and the mainstream flow. An increase in momen-
tum flux of the primary air augments the overall sound pressure
level as shown in Fig. 9b. Besides this, as discussed by Prakash et
al. [12], quenching effects of cavity flame can be characterized by
Damkohler number. Based on these understandings, a new corre-
lating parameter is devised and the variation of OASPL with this
parameter is shown in Fig. 10. Using this correlating parameter,
OASPL curves for different operating conditions are collapsed into
a single curve and the semi-empirical relationship is given as be-
low.
 
MFR0.5 × ∅0overall
.35
OASPL = 80 + 0.88 (1)
Dac0.65

In Eq. (1), Damkohler number at the cavity zone (Dac ) is the

ratio between the cavity flow  scale (τres,c ) to the chemi-
 time
cal time scale (τchem ), Dac = τres,c /(τchem ), where, τchem = α / S 2L ,
τres,c = 2L / V p , α is the thermal diffusivity of mixture and S L is
the laminar burning velocity at the stoichiometric mixture. This
correlation is having an R 2 value of around 0.89 and is valid for
the following conditions Φc = 1.0 to 1.4; Rems = 20 000 to 40 000;
V p = 40 to 140 m/s.

4. Concluding remarks

Fig. 7. Power spectral density of combustion noise signal for two primary air velocity
cases (a) V p = 60 m/s, and (b) V p = 100 m/s.
In the present work, the effects of injection strategy and in-
jection momentum on the combustion noise characteristics of a
2D trapped vortex combustor (TVC) are studied experimentally.
3.4. Overall sound pressure level
Spectral analysis of the microphone signal at various angular lo-
cations indicates that change in noise level with angular location
Fig. 9a shows the variation of overall sound power level (OASPL) is insignificant. Besides this, in order to avoid the hydrodynamic
with the primary air velocity for the cavity equivalence ratio 1.2. noise contamination, the microphone is located at 40 cm from
From this figure, it can be observed that the overall sound pressure the combustor axis. Analysis of the microphone signal reveal that

Fig. 8. Effects of cavity equivalence ratio on the power spectral density for the mainstream Reynolds number cases (a) Rems = 20 000, and (b) Rems = 40 000.
 P.K. Ezhil Kumar, D.P. Mishra / Aerospace Science and Technology 43 (2015) 388–394 393

Fig. 9. Variation of overall sound pressure level (OASPL) with (a) primary air velocity and (b) momentum flux ratio (MFR).

Conflict of interest statement

The authors declare that there is no conflict of interest in pub-

lication of this work.

Acknowledgements

The authors gratefully acknowledge the financial support

(AE/DST/20100202) provided by the Department of Science and
Technology (DST), New Delhi, India, for carrying out the present
work. The authors also wish to thank the graduate students Ma-
hesh and Pravendar for their help in conducting experiments.

References

[1] A.H. Lefebvre, D.R. Ballal, Gas Turbine Combustion, 3rd ed., Taylor and Francis,
2010.
[2] W.M. Roquemore, Dale Shouse, Dave Burrus, Art Johnson, Clayton Cooper, T.
Fig. 10. Correlation for overall sound pressure level (OASPL). BeverlyDuncar, K.Y. Hsu, V.R. Katta, G.J. Sturgess, V. IllariVihinei, Trapped vortex
combustor concept for gas turbine engines, in: 39th AIAA Aerospace Sciences
Meeting & Exhibit, Reno, 8–11 Jan. 2001, 2001 (AIAA 2001-0483).
for a particular mainstream Reynolds number, an increase in pri- [3] K.Y. Hsu, L.P. Goss, W.M. Roquemore, Characteristics of a trapped vortex com-
mary air velocity (V p ), results in increase of power spectral density bustor, J. Propuls. Power 14 (1998) 1–12.
[4] D.P. Mishra, R. Sudharshan, Numerical analysis of fuel–air mixing in a two-
(PSD) magnitude. Results also indicate that the cavity equivalence dimensional trapped vortex combustor, Proc. Inst. Mech. Eng., G J. Aerosp. Eng.
ratio (Φc ) also influences the spectral amplitude. Moreover, the 224 (2009) 65–75.
PSD magnitude is observed to be higher for Rems = 20 000 as [5] R.C. Zhang, W.J. Fan, Q. Shi, W.L. Tan, Combustion and emission characteristics
compared to that of Rems = 40 000 case. For Rems = 40 000 case, of dual-channel double-vortex combustion for gas turbine engines, 130 (2014)
314–325.
the merging of cavity flames is inhibited at lower V p cases due
[6] Atul Singhal, R.V. Ravikrishna, Single cavity trapped vortex combustor dynam-
to the quenching of mainstream flame by the high momentum ics – part 2: simulations, Int. J. Spray Combust. Dyn. 3 (2011) 45–62.
mainstream flow. However, for a particular mainstream Reynolds [7] P.K. Ezhil Kumar, D.P. Mishra, Numerical simulation of cavity flow structure
number, the noise level increases with the momentum flux ratio in an axisymmetric trapped vortex combustor, Aerosp. Sci. Technol. 21 (2012)
16–23.
between cavity and mainstream flow. Analysis of results reveals
[8] B.N. Shivashankara, R.W. Crouch, Noise characteristics of a can type combustor,
that the noise level varies between 80 dB for non-merged flame J. Aircr. 14 (1977) 751–756.
and 100 dB for merged flame cases. An attempt has also been [9] G.J. Sturgess, K.-Y. Hsu, Combustion characteristics of a trapped-vortex com-
made to correlate the overall sound pressure level (OASPL) with the bustor, in: RTO AVT Symposium on Gas Turbine Combustion, Emissions and
Alternative Fuels, Portugal, 12–16 Oct. 1998, 1998.
non-dimensional operating variables. The proposed semi-empirical
[10] E.J. Gutmark, C.O. Paschereit, D. Guyot, A. Lacarelle, J.P. Moeck, S. Schimek, T.
correlation for OASPL in terms of MFR, Da and Φoverall is given be- Faustmann, M.R. Bothien, Combustion noise in a flameless trapped vortex re-
low heat burner, in: 13th AIAA/CEAS Aero-acoustics Conference, Italy, 21–23 May
  2007, 2007 (AIAA 2007-3697).
MFR0.5 × ∅0overall
.35 [11] M. Muthukrishnan, W.C. Strahle, D.H. Neale, Separation of hydrodynamic, en-
OASPL = 80 + 0.88 tropy and combustion noise in a gas turbine combustor, AIAA J. 16 (1977)
Dac0.65 320–327.
[12] S. Prakash, S. Nair, T.M. Muruganandam, Y. Neumeier, T.C. Lieuwen, J.M. Seitz-
Using this correlating parameter, OASPL curves for different op- man, B.T. Zinn, Acoustic sensing and mitigation of lean blowout in premixed
flames, in: 43rd Aerospace Sciences Meeting, Reno, Jan. 2005, 2005 (AIAA
erating conditions are collapsed into a single curve. This correla-
2005-1420).
tion is having an R 2 value of around 0.89 and is valid for Φc = 1.0 [13] R. Rajesh, T. Lieuwen, Acoustic radiation from turbulent premixed flames,
to 1.4; Rems = 20 000 to 40 000; V p = 40 to 140 m/s. J. Fluid Mech. 637 (2009) 357–385.
394 P.K. Ezhil Kumar, D.P. Mishra / Aerospace Science and Technology 43 (2015) 388–394

[14] Y.S. Choy, H.S. Zhen, C.W. Leung, H.B. Li, Pollutant emission and noise radia- [16] Y.K. Jeong, C.H. Jeon, Y.J. Chang, Evaluation of the equivalence ratio of the re-
tion from open and impinging inverse diffusion flames, Appl. Energy 91 (2012) acting mixture using intensity ratio of chemiluminescence in laminar partially
82–89. premixed CH4 air flames, Exp. Therm. Fluid Sci. 30 (2006) 663–673.
[15] P.K. Ezhil Kumar, Experimental and numerical studies of twin cavity 2D trapped [17] M. Orain, Y. Hardalupas, Effect of fuel type on equivalence ratio measurements
vortex combustor, PhD thesis, IIT Kanpur, 2014. using chemiluminescence in premixed flames, C. R., Méc. 338 (2010) 241–254.