PHASE-LOCKED PIV AND OH- CHEMILUMMINECENCE VISUALIZATION ON A SWIRL STABILIZED GAS TURBINE BURNER.

hammercoupleMécanique

22 févr. 2014 (il y a 3 années et 6 mois)

242 vue(s)

12
TH

INTERNATIONAL SYMPOS
IUM ON FLOW VISUALIZ
ATION

September 10
-
14, 2006, German Aerospace Center (DLR), Göttingen, Germany

1



ABSTRACT

In this work the heat release and flow field of a gas turbine burner has been studied using phase
-
locked
PIV and OH
-
chemiluminescence. The aim of these experiments was to obtain a better understanding of
thermoacoustic interactions in the comb
ustion process. Therefore, the interaction between heat release
fluctuations, acoustic fluctuations and vorticity fluctuations in the combustion process has been investigated..
In order to analyze these unsteady phenomena a novel phase locking method has b
een developed that enables
to obtain phase
-
resolved data in a post processing step. Using this method synchronous phase
-
locked PIV and
OH
-
Chemiluminence visualizations have been obtained. The velocity field obtained in this way have been
decomposed in thei
r rotational and irrotational parts. The unsteady rotational part represents the fluctuating
vorticity field, the irrotational part represents the (thermo
-
) acoustic motion. This, together with the
chemiluminescence data (which correlates strongly with the

heat release rate), enabled to study
thermoacoustic interactions on a full
-
size swirl
-
stabilized gas turbine burner at various operating conditions.

1 INTRODUCTION

Modern design of low emission combustors is characterized by swirling air in the combustor
's
dome coupled with distributed fuel injection to maximize mixing. This design results in efficient
combustion with extremely low emissions. The ALSTOM EV
-
type burners have the unique
property of flame stabilization in free space near the burner outlet
utilizing the sudden breakdown of
a swirling flow, called vortex breakdown. The swirler is of exceptionally simple design, consisting
of two halves of a cone, which are shifted to form two air slots of constant width [1]. Gaseous fuels
are injected into th
e combustion air by means of fuel distribution tubes comprises of two rows of
small holes perpendicular to the inlet ports of the swirler. Complete mixing of fuel and air is
obtained shortly after injection.


The characteristics of combustion stabilizatio
n by vortex breakdown are controlled by the flow
dynamics associated with this particular flow phenomenon. Vortex breakdown is defined as a flow
instability that is characterized by the formation of an internal stagnation point on the vortex axis,
follow
ed by reversed flow.

Upstream of the vortex breakdown location, the velocity profile is highly jet
-
like with a peak
velocity almost three times greater than the mean velocity. Very close to the downstream of the
breakdown, the flow in the core may complet
ely stagnate and then change to a wake
-
like flow.
Downstream of the breakdown turbulence increases, axial velocities are substantially lower and
reverse flow is possible.

PHASE
-
LOCKED PIV AND OH
-
CHEMILUMMINECENCE VI
SUALIZATION ON A SWI
RL
STABILIZED GAS TURBI
NE BURNER.


Bruno Schuermans*, Felix Guethe*, Adrienn Scivos*,

Melanie Voges**, Chris Willert**

*ALSTOM, 5401 Baden, Switzerland

**
German Aerospace Centre, 51147 Cologne, Germany

bruno.schuermans@power.alstom.com




Keywords
: keywords
(no more than 5)


Bruno Schuermans, Felix Guethe, Adrienn Scivos, Melanie Voges, Chris Wil
lert

2



In this work flow and flame visualization techniques have been used simultaneousl
y to study
unsteady phenomena in such combustion processes. The fluctuations of the velocity field consists of
a superposition of acoustic and vortical or turbulent contributions. Moreover, fluctuations of heat
release in the flame can cause temperature fl
uctuations that result in velocity fluctuations as well. To
study these different effects, special phase
-
locking techniques have been used in conjunction with a
post processing technique that separates acoustic from vortical contributions.


2 EXPERIMENTAL

SET
-
UP

The optical setup was mounted on a facility
-
decoupled support to minimize vibration on the
camera system and the light sheet optics. A double
-
cavity 120 mJ Nd:YAG laser with 532 nm
wavelength was used. The laser beam was guided to the light sheet o
ptics via an articulated laser
arm. To form the laser light sheet (LLS) with a thickness of 1.5
-

2 mm a set of three lenses was used
(fsph.1 =
-
50 mm, fsph.2 = 100 mm, fzyl.3 =
-
50 mm). With the help of a mirror the LLS was
delivered from the bottom into
the combustion chamber and was adjusted in the burner's centre in a
plane perpendicular to the burner exit. The mirror as well as the lens setup was purged with
compressed air during burner operation for cooling and to avoid any contamination of the sensit
ive
optical parts.

For flow observation a thermo
-
electrically cooled PIV CCD camera with a spatial resolution of
1600´1200 px at a frame rate of 15 Hz was used. The benefit of the high frame rate is that laser and
camera can run at the same frequency which

reduces measurement time. Additionally it is possible to
run the laser at its design frequency, where the optimum beam profile is achieved. The camera was
mounted on a Scheimpflug adapter to optimize alignment of the camera optics with the LLS. The
focal
length of the chosen camera lens was f = 55 mm. A precise calibration target was used to align
the LLS plane with the PIV camera object plane. Due to the high flame luminosity a band pass filter
was used with the camera optics to enable the desired laser s
tray light from the seeding particles and
at the same time suppress the background luminosity.

As the outer recirculation zones of the combustion chamber were of major interest, the cameras
viewing direction was adjusted to the upper half of the burner exi
t flow. To optimize the field of
view the camera was therefore positioned under a certain angle related to the LLS. The correct
mapping in the image plane of the camera optics was achieved using a Scheimpflug adapter.

Due to the high mass flow rates reache
d during combustor operation two particle generators
were installed in parallel. Using dry, compressed Nitrogen a seeding mass flow of about 5 g/s (1% of
total mass flow) was established and injected to the main flow of the burner.


Camera and laser were o
perated at 15 Hz repetition rate with a pulse separation of 7 µs and an
exposure time of 10 µs. Due to the increasing window contamination (seeding particle deposits) with
the number of PIV measurement sequences the burner needed to be run down app. every
10th
sequence for cleaning. Each sequence consisted of about 200 images.



3


PHASE
-
LOCKED PIV AND OH
-
CHEMILUMMINECENCE VI
SUAL
IZATION ON A SWIRL
STABILIZED GAS TURBI
NE BURNER.

Side window
for optical
access of
camera
Glass slit for
optical access
of laser
Swirl
-
stabilised
premix burner
Flow
Flame front
Outer
recirculatio
n zone
Inner
recirculation
zone
Region
captured by
PIV



Fig
.

1
: [Left] Picture of test rig with windows. [Right] Sketch of typical flow field and flame,
together with the region captured by PIV.



Two water
-
cooled microphones (one u
pstream and the other downstream of the burner) have
been used to record the acoustic pulsation levels in the combustor. For the frequency range of
interests, the acoustic wavelength is at least one order of magnitude larger than the axial extend of
the op
tical access, so two microphones are sufficient to characterise the acoustic field.


The exhaust of the test facility has a variable geometry, which allows to change the acoustic
reflection coefficient (R) from nearly fully reflecting ( R >0.9 ) to almos
t fully absorbing ( R<0.20).
A large reflection coefficient was chosen for all experiments, in order to be able to study unsteady
combustion phenomena.



The microphone signals are digitised and recorded on a PC with data
-
acquisition board.
Simultaneously,

the trigger signals form the laser, the PIV camera and the chemiluminescence
camera were recorded. Sample
-
and
-
hold hardware and a sufficiently high sampling rate ensured that
all five signals, could be recorded with high phase accuracy.


3 POST PROCESSING

The post processing of the raw PIV images to obtain the velocity vectors will not be discussed
in this paper, the reader is referred to [7, 8]. However, two techniques have been developed in the
course of this work will be highlighted here: A phase
-
locked

post
-
processing technique and a method
for separating acoustic from vortical motion.

3.1 Off
-
line Phase
-
locking

In order to visualize the acoustic interactions between heat release, acoustics and vorticity, a
special phase
-
locking procedure has been devel
oped. More conventional methods of phase
-
locking
typically use a band
-
pass filter, pulse generator and a delay line. The band pass filter is set to a
frequency range that contains some dominant pulsation peak of interest. The filtered microphone
signal the
n triggers a pulse from the pulse generator. The delay line can be adjusted to achieve the
Stagnation

point

Bruno Schuermans, Felix Guethe, Adrienn Scivos, Melanie Voges, Chris Wil
lert

4

desired phase shift of the pulse, which is then send to the camera/laser. In this way the images have a
fixed phase with respect to the microphone signal (or other r
eference signal), averaging a sequence
of images yields a phase
-
locked image. Repeating this for a range of phase angles yields a phase
-
locked sequence that gives a visual representation of the periodic motion at the specific frequency.
This method yield g
ood results and has been reported in numerous papers. However, for PIV
applications it is important that the method is very fast. Because of window contamination by the
seeding particles, as much data as possible should be recorded within the given time. M
oreover, the
online band pass filtering operation will inevitability introduce a phase distortion in the pass
-
band,
which impairs the quality of the averaging procedure.


The newly developed method makes use of the recorded time traces of the acoustic sign
al and
the camera (or laser) trigger signal. To do so, a band
-
passed Hilbert transform of the microphone
signal is calculated. The phase of this complex
-
valued time trace represents the instantaneous phase
of the acoustic signal within the frequency band o
f interest. Because the filtering procedure is done
off
-
line, zero
-
phase distortion can be achieved for the band pass filtering operation. The time instants
at which each image is recorded is known because the trigger signal of the camera has been
recorded
, thus since the instantaneous phase is also known a s a function of time, each image can be
assigned an instantaneous phase for one (or more) frequency bands. Then the range of phases is
divided in a number of equally spaced intervals, all images correspo
nding to a certain range of phase
angles are a then averaged. Finally, the averaged images for each phase interval yields the phase
locked sequence. Because the phase
-
locking is done off
-
line, the method is not restricted to only
calculate the mean values
oat each phase angle. For example, results will be shown where the
standard deviation (RMS) is shown at each phase locked sequence. An additional advantage of this
method is that the dynamic behaviour at several frequencies can be obtained from one experim
ent.

3.2 Decomposition into acoustic and vortical fields

The unsteady velocity field recorded by PIV is generally a superposition of an acoustic motion,
turbulence and vortical coherent structures. The energy transfer from the acoustic field to the
vortici
ty field is appreciated as a strong damping mechanism for the acoustic field. Coherent vortical
structures can also be a source of sound, especially if interaction with the heat release process takes
place. One could argue that an appropriate phase lock
ing technique could remove the contributions
of turbulence and vorticity form the acoustic motion, because the coupling between these fields is
only of second order. However as discussed in [3], this is only true in domains far away of solid
boundaries, wh
ich is clearly not the case here. In order to separate the different contributions, the
velocity field is decomposed in a rotational part and an irrotational part. The irrotational part is
representative of the acoustic field, whereas the rotational part i
s representative for the vorticitcal
field (turbulence + coherent structures).

The decomposition is made by solving the following system of four equations for the four
unknown velocity components of
r
u


and
i
u




u
u
u
u
u













i
r
i
r
0
0



5


PHASE
-
LOCKED PIV AND OH
-
CHEMILUMMINECENCE VI
SUAL
IZATION ON A SWIRL
STABILIZED GAS TURBI
NE BURNER.

where
u


is the measured velocity in the plane (it consists of an axial and radial component) the
subscripts ( r ) and ( i ) refer to the rotational and irrotational components of the velocity,
respectively. An ap
proximate solution can be found by solving the spatial Fourier transform of the
system of equations, and than back
-
transforming the results to space domain using the inverse
Fourier transform in order to obtain
r
u


and
i
u

. Because a Fourier transform is used, this
approximation works very well for periodic signals. However since the flow fields are generally not
periodic, they are made spatially periodic in an artificial way, by copying and mirroring the flow
field f
our times to created a four times larger symmetric flow field.

4 RESULTS

In order to investigate the interaction between combustion and the velocity field, the heat release was
visualized by taking OH*
-
chemiluminescence (CL) from the inherent flame emissio
n of
electronically excited OH radicals. The OH* CL images are recorded using a UV filter where the
light is detected. The CL images contain light from all over the combustor integrated along the line
of sight. Thus, in contrast to the PIV data that only m
easures a plane within the combustor, in CL
analysis a three
-
dimensional volume is considered to produce the images. Single shot images of PIV
velocity vectors superimposed onto CL intensity data are shown in Figure XX and XX for two
random samples.

The C
L
-
camera was synchronized with the PIV system. Similar to the PIV data, all CL pictures are
assigned a phase angle for the considered frequency bandwidth. Then the CL data are phase
-
locked
to yield a similar eight
-
picture result.

A s
ingle shot images of PI
V velocity vectors superimposed onto CL intensity data are shown in
Figure 2 for a random time instant.



Fig. 2: Synchronously detected PIV and CL single shot plotted overlaid.

Bruno Schuermans, Felix Guethe, Adrienn Scivos, Melanie Voges, Chris Wil
lert

6

The post processed phase locked results in Fig. 3 show the fluctuation of th
e velocity field with
respect to phase angle for one frequency. The phase increases in clockwise direction, the most right
picture corresponds to the zero phase position. The arrows represent the velocity fields, the colors
indicate the axial component of

the velocity. The white colored region indicates an axial velocity of
zero, the blue a negative axial velocity and the red a positive axial velocity. Thus, the white areas in
Fig. 3 correspond to lines that can be considered as contours of the recirculati
on zones. The inner
recirculation zone is clearly visible; the outer zone is barely in the region captured by PIV. The
stagnation point of the flow will be defined here as the axial location of this contour on the
centerline. As will be discussed later, it

of interest to note that the position of the stagnation point
moves periodically with phase angle.









Figure 3: Phase averaged velocity field. Colors indicate
axial velocity (positive: red, zero: white, negative:
blue), phase increa
ses clockwise with 45°.


Figure 4: Phase averaged unsteady velocity field (mean
removed).

Apart from the periodically moving stagnation point, no strong phase dependence can be seen in Fig.
3. The reason is of course that acoustic perturbations are gene
rally small with respect to the mean
flow (rarely as high as 20%). To be able to study better the phase dependence, the mean of the flow
fields of all phases is subtracted from the field for each phase. The velocity fields obtained in this
way are referred

to as the “unsteady velocities” and are plotted in Fig. 4. Now a clear phase
-
dependence is observed: images that are 180° of phase angle apart, are each other’s negative.


7


PHASE
-
LOCKED PIV AND OH
-
CHEMILUMMINECENCE VI
SUAL
IZATION ON A SWIRL
STABILIZED GAS TURBI
NE BURNER.



Figure 5: Irrotational component of the velocity field:
representing (thermo
-
) aco
ustic motion.


Figure 6: Rotational component of the velocity field:
representing motion of vortices and turbulence.



The flow field shown in Fig. 3 has been decomposed into its rotational and
irrotational velocity components prior to performing the pha
se averaging. The unsteady irrotational
component of axial velocity is shown in Fig 5, and the unsteady rotational picture is shown in Figure
6. The unsteady irrotational part is representative for the acoustic motion of the fluid. So, it is not
surprising

to see a predominantly axial motion in Fig 5, since the frequency of oscillation was well
below the acoustic cut
-
on frequency of the combustion chamber. However, the gradient of the
acoustic velocity is much larger than what would be expected based on the

acoustic wavelength for
this frequency of oscillation. Previous detailed acoustic investigations have shown that this
frequency corresponds to the quarter wave mode of the test facility; the corresponding wavelength is
in the order of 5
-
10 meters. Clearly

the length of the diagnostic window is only a fraction of this
length. The motion displayed in Fig. 6 has a wavelength which is in the order of the length of the
diagnostic window. So, this motion does not correspond to a purely acoustic field associated
with the
phase
-
locking frequency. This motion can be explained by analyzing the Chemiluminesence data: it
is found that the unsteady heat release and entropy waves are responsible for the behavior. The
unsteady heat release acts a source term to the acoust
ic field. Compared to the acoustic wave length,
the flame can be seen as discontinuity or ‘jump’ in density and thus as a ‘jump’ in (acoustic)
velocity. This is of course caused by the (unsteady) heat release. This heat release is typically related
to the
acoustic velocity (REF) with a time delay and is considered as the driving mechanism for
combustion instabilities. Thus the strong velocity gradients in Fig 5, are caused by periodically
changing density due to the heat release fluctuations.

Further invest
igation of Fig. 5 shows that the irrotational velocity behind the flame has a convective
motion, i.e., there is not only a strong gradient, but the area of high velocity seems to travel from left
to right with phase angle. This motion cannot be explained b
y purely acoustic motion (because the
wavelength is to short, and because under these reflective conditions predominantly standing waves
Bruno Schuermans, Felix Guethe, Adrienn Scivos, Melanie Voges, Chris Wil
lert

8

are expected, not traveling waves), this convective motion cannot be explained by the heat release
process only.

A pla
usible explanation for the observed phenomenon would be the presence of entropy waves: the
heat release process in the flame causes temperature fluctuations that are convected downstream with
the mean flow. So, ‘pockets’ of hotter gas a periodically releas
ed by the flame and travel
downstream with mean flow speed. The associated density difference then affects the velocity field
in a traveling wave like manner. It should be noted that in order to see this effect the temperature
fluctuations should be non
-
is
entropic, i.e, they should be caused by entropy waves. This only occurs
if heat release fluctuations are caused by fuel to air ratio fluctuations or if incomplete combustion
takes place (periodical extinction of part of the flame) [4]. This can easily be u
nderstood by
considering that for a given mixture temperature and gas composition the temperature of the
reactants is defined by the equivalence ratio. Thus if these do not fluctuate, the temperature will not
fluctuate. Incomplete combustion did not take p
lace in this case (increased levels of unburned
hydrocarbons and CO would have indicated this), so, such a observation is an indirect way to
demonstrate that equivalence ratio fluctuations were responsible for driving the acoustic field in this
experiment.


If the same experiment would have been repeated, but then with fuel and air entirely mixed before
entering the burner, then case the wave
-
like velocity behavior after the flame would not be observed.
In [5] a model has been proposed for thermoacoustic in
teraction that depends not only on
equivalence ratio fluctuations, but also on a periodic variation of flame area or burning velocity due
to periodically changing vortical velocity. The heat released would then periodically be enhanced by
periodically chan
ging turbulence intensity (small fluctuations), which would cause a periodically
changing burning velocity, and hence a modulated heat release. Or, in a similar way, large scale
vortical structure could be responsible for periodical changing of flame area
or periodically enhanced
mixing of fresh gasses with combustion gases. In either way, the coupling would take place via the
rotational part of the velocity. The model presented in [5] fitted remarkably well with experimental
transfer function data. If this

model would be correct, then this would imply that that the variance of
the rotational velocity would be proportional to the acoustic velocity, and that the heat release would
be proportional to this variance. The upper part of Fig. 7 does indeed shows th
at the standard
deviation of the velocity for every phase angle has a periodic motion.

The rotational part of the velocity shown if figure 6 shows predominantly a large scale recirculation
motion as sketched in Fig. 1. The dependence of phase angle in Figu
re 4 shows that this structure
accelerates and decelerates periodically in clockwise direction. It is unclear how this could interact
with the heat release in the flame, however it seems likely that such periodic motion could affect the
flame front kinemat
ics in a similar manner as decribed in [] and therefore the heat release.

An other plausible interaction mechanism between heat release and the acoustic field, is that the
flame is anchored on the center line, due to a strong negative velocity gradient in
front of the
stagnation point of the flow. An acoustic modulation of the flow, and hence of the stagnation point,
would modulate the flame anchoring point, which would yield a periodic modulation of the flame

9


PHASE
-
LOCKED PIV AND OH
-
CHEMILUMMINECENCE VI
SUAL
IZATION ON A SWIRL
STABILIZED GAS TURBI
NE BURNER.

area due to flame front kinematics (via a simil
ar mechanism as in [6]). The variation of axial position
of the stagnation point is plotted versus the phase in figure 7. Clearly a periodic motion is observed,
the velocity of this motion is in the same order of magnitude as the acoustic modulation.



Fi
g. 7.

Standard deviation of the rotational velocity and location of the flow stagnation point as a
function of phase.

4 CONCLUSIONS

The flow field in full
-
scale swirl
-
stabilized gas turbine burner has been visualized using PIV. An
offline phase
-
locking

technique has been developed to investigate acoustic motion and vortical
motion of the flow field. This analysis showed a distinct phase
-
locked motion of the stagnation
point of the inner recirculation bubble in the flow field and turbulence intensity fl
uctuations that are
coherent with the acoustic motion These different contributions, their effect on the thermoacoustic
process in the flame and a qualitative comparison with models available in literature have been
discussed.


Bruno Schuermans, Felix Guethe, Adrienn Scivos, Melanie Voges, Chris Wil
lert

10







ACKNOWLE
DGEMENTS

This work was conducted in the framework of the FP5 EC
-
Project FuelChief.


REFERENCES


1.

Doebbeling, K., Eroglu, A., Joos, F. and Hellat, J., Novel technologies for natural gas combustion in
turbine systems,
Eurogas 99,
Ruhr University Bochum, Germa
ny, 1999

2.

Paschereit, C., Gutmark, E., and Weisenstein, W., Coherent structures in swirling flows and their role
in acoustic combustion control 1999, Physics Fluids, 11
-
9: 2667:2678, 1999.

3.

Chu, B.T., Kovásznay, S.G., Non
-
linear interactions in a viscous hea
t
-
conducting compressible gas,
Journal of Fuid Dynamics

3: 494
-
514, 1958.

4.

Chu, B.T., On the Generation of Pressure Waves at a Plane Flame Front,
Fourth Symposium
(International) on Combustion
, The Combustion Institute, pp. 603
-
612., 1953,

5.

Schuermans B., Be
llucci V., Guethe F., Meili F., Flohr P., Paschereit C. O
., A detailed analysis of
thermoacoustic interaction mechanisms in a turbulent premixed flame
, ASME Turbo Expo 2004,
GT2004
-
53831

6.

M. Fleifil, A.M. Annaswamy, Z. Ghoniem and A.F. Ghoniem, Response of
a laminar premixed flame
to flow oscillations: a kinematic model and thermoacoustic instability result,

Combustion and Flame
,
106:487
-
510, 1996.

7.

Raffel M, Willert C, Kompenhans J, Particle image velocimetry
-

A practical guide. Springer
-
Verlag:
Berlin, Hei
delberg, NewYork, ISBN 3
-
540
-
63683
-
8, 1998

8.

Willert C, Jarius M, Planar flow field measurements in atmospheric and pressurized combustion
chambers. Exp Fluids vol.33, pp.931
-
939, 2002