You are on page 1of 16

Experimental and numerical investigations of jet active

control for combustion applications


Vincent Faivre and Thierry Poinsot


Institut de Mecanique des Fluides de Toulouse, Allee du Pr. Camille Soula, 31400
Toulouse, FRANCE
Abstract. Controlling the mixing of a gas (usually fuel) issuing from a tube into surrounding
air is a basic problem in multiple combustion systems. The purpose of the present work is
to develop an actuator device to control the mixing enhancement of an axisymmetric nonreactive jet. The actuators consist of four small jets feeding the primary jet flow. These four
jets are oriented to add an azimutal component to the velocity field. The influence of jets
deflection and position along the main jet duct is discussed. Schlieren photographs and Planar
Laser Induced Fluorescence measurements are used to compare the efficiency of the three
configurations of interest. The effect of control-to-main mass flow rates ratio is quantified
through hot wire anemometry results. Large Eddy Simulations (LES) of both forced and
unforced configurations are also performed. The objectives of the numerical part of this work
are to understand the actuator effect and to validate LES as a tool to study active control.

Submitted to: Journal of Turbulence

To whom correspondence should be adressed (thierry.poinsot@imft.fr)

Active jet control for combustion applications

1. Introduction
Combustion instabilities may occur in closed combustion chambers, resulting from the
coupling between acoustics and combustion [1, 2, 3, 4, 5, 6, 7]. They can appear in many
combustors such as gas turbines or industrial furnaces. Those instabilities are responsible
for noise, vibrations and sometimes complete device failure [8]. Being able to control such
phenomena is an important research path in the combustion community. There are two ways
to control a flow. Passive control consists in modifying the geometry of the burner [9] and/or
the combustion chamber ; on the other hand active control consists in injecting external
energy through actuators [10, 11]. The quality of the control depends directly on the design
of the actuators. Some of them are specific to combustion applications but most actuation
techniques are encountered in both reactive and non reactive applications : loudspeakers
[10, 12], synthetic jets [13], flaps [14]. The purpose of the present study is to quantify,
experimentally and numerically, the effects of forcing on the aerodynamic field in a model
configuration : a non-reactive jet of air.
The actuators are designed to modify the flow in two ways (figure 1) :
a radial fluid injection into the main jet enhances its mixing with the ambient air [15, 16],


a swirl addition drastically changes the aerodynamic pattern of the flow and can be used
to stabilize the flame [17].

To obtain these effects simultaneously the actuators of the present work consist of four
small jets feeding the primary jet flow. These four jets are oriented to add an azimutal
component to the velocity field. To visualize the effect of the actuators on the main flow,
hot-wire anemometry, Schlieren photographs and Planar Laser Induced Fluorescence (PLIF)
measurements are used.
Introducing swirl to increase mixing or to stabilize flames is not new [17, 18, 19, 20, 21,
22, 23]. The effect of swirl jets has been studied in details by many authors and the objective
of the present work is not to repeat these analysis. Here, the objective is to design a control
device (the four jets) which can be easily tuned to adjust swirl and turbulence levels without
any geometry modification. Another objective is to minimize pressure losses which are often
encountered in swirling devices using vanes for example.
The interaction of the actuators jets with the main jet can also be simulated using Large
Eddy Simulations. Such LES provide a detailed insight into the physical mechanisms which
is difficult to obtain experimentally. At this stage, LES cannot be used for all regimes to
investigate the most efficient actuation mode : for such a task, experiments are more efficient.
However LES is more powerful to investigate one regime in details so that a combination
of experiments and LES is used in this work where the strategy is the following : first,
experiments are used to establish which actuator configuration is the most efficient in terms of
mixing enhancement ; LES of this configuration only are then performed to understand how
the actuator actually affects the main jet.

Active jet control for combustion applications

Figure 1. Generic scheme of the SW90, A90 and A45 configurations.

Nozzle
A90
A45
SW90

( )
90
45
90


h (mm)
30
30
8

Table 1. Control parameters for each configuration of actuator.

2. Experimental facility
Figure 1 shows a generic scheme of the nozzle equipped with the actuator. The exit diameter
of the main jet (D) is 10 mm while the exit diameter of each small secondary jet is 2mm.
Previous investigations have shown that the exit diameter of the small jet (d) is one of the
main parameters having an influence on the control efficiency [24]. Here, two other parameters
are tested : the orientation of the four small jets relative to the main one () and the distance
between the actuator and the main jet exit (h). To evaluate the importance of these parameters,
three configurations of actuators have been tested (table 1).
For all configurations, the control jets are tangential to the main one to add an azimutal
component to the velocity field.
The nozzle is connected to a cubic box so that the jet flow is confined. The characteristic
length of the box is 10 diameters of the main jet. The outlet of the box ends directly into
free atmosphere. The effects of the box size were investigated numerically : even though the
uncontrolled flow was slightly affected by the box length, the effects of control on this flow
were independent of it.
The main air flow is delivered by an hot air generator. This device allows to reach mass
flow rates up to 30 g/s at 400 C. Temperature regulation is performed using PID regulation.
The control flow rate is measured by a DANTEC S2140 Mass Flow Transducer which has
been specially modified for the present work to measure unsteady flow rates up to 1 g/s at a
frequency up to 500Hz. The ratio between the mass flow rate of the control flow (m act ) and


Active jet control for combustion applications

the mass flow rate of the main jet (m jet ) is :


q


m act
m jet

(1)

It is adjusted with a MOOG servovalve which works in both continuous and pulsated regime.
The jet spreading is characterized by the mean and rms velocity fields measured with a single
hot-wire probe. The flow is also visualized through Schlieren photographs to measure its
spreading angle and through PLIF snapshots to characterize the flow structure.
2.1. Velocity measurements
Velocity profiles are obtained using a DANTEC Streamline constant temperature anemometer
coupled to a CHARLYROBOT displacement table which allows micrometric probe shifting.
The single hot-wire calibration is performed on DANTEC 54H10 calibrator. Each part of
the anemometry rig is linked to a computer equipped with a KEITHLEY DAS 1800ST/HR
acquisition card. Velocity signals are recorded at a frequency of 25000 Hz during 1 second.
2.2. Optical diagnostics
Two optical diagnostics have been used for the present study : Schlieren technique and Planar
Laser Induced Fluorescence (PLIF). For the Schlieren visualization the main jet is preheated
(70 C). Since the present experiments are carried out using non fluorescent gases, PLIF
measurements require the flow to be seeded with acetone. Acetone fluorescence is induced
by a laser sheet (Nd YAG l=266nm in UV, I0 =40mJ, sheet thickness=500nm). An intensified
CCD camera sensitive to the acetone fluorescence wave length is used to collect PLIF images.


3. Numerical setup
The principle of LES is to resolve the larger scales of turbulence while modelling the smaller
ones [7]. LES is therefore a good tool to predict the effect of control on the flow because large
structures are certainly essential in this mechanism.
LES of the experimental configuration are performed using the parallel CFD code
AVBP developed at CERFACS in Toulouse (www.cerfacs.fr) and at IFP in Paris, France.
AVBP solves the full compressible Navier-Stokes equations on 2D or 3D meshes. Meshes
can be structured, unstructured or hybrid. The numerical approach is based on finite-volume
schemes using the cell-vertex method. The scheme provides third-order accuracy both in
space and time. It has been tested on multiple simple (isotropic turbulence, vortices, acoustic
waves, channel, etc...) and complex cases [25, 26, 27, 28, 29, 30].
The boundary conditions treatment is based on the NSCBC approach [31, 32] : acoustic
waves are identified and treated independently on boundaries to satisfy the boundary
conditions. At the inlet of the main jet and of the four actuator jets, velocity profiles are
imposed. At the box outlet, a non reflecting condition is fixed [31, 7]. On all walls, no slip
conditions are imposed (figure 2).

Active jet control for combustion applications

Figure 2. Computational domain. 1 : actuators. 2 : main flow. 3 : jet blowing in the box.

The subgrid-scale model is the WALE (Wall Adapting Local Eddy viscosity) model
[25, 33]. This model is based on the square of the velocity gradient tensor and degenerates
correctly near the walls.
All simulations are three-dimensional, the mesh is hybrid and contains 1 million cells for
600.000 points. Figure 2 shows the computational domain. Note that, even for the unforced
case, the actuators are included in the mesh.
4. Experimental results : determination of the optimal configuration
To quantify the actuators effect, a criterion based on the jet spreading angle measured on
Schlieren views is used (figure 3). The estimated value of the jet spreading angle includes
the large structures of the shear layer. This diagnostic is sufficient to compare solutions and
identify the most efficient actuators.

Figure 3. Schlieren photographs of the unforced flow (a) and the forced one (b). Evaluation
of the jet spreading angle.

Active jet control for combustion applications

Different values of control-to-main mass flow rate ratio (q) have been investigated and
figure 4 shows the jet spreading angle evolution as a function of q. For every actuator the jet
spreading angle increases with q but not in the same way. Note that the jet angle significantly
varies with control ranging from 19 (unforced) to 70 (forced cases).


Figure 4. Jet spreading angle vs. control-to-main mass flow rate ratio (q) for A45, A90 and
SW90 configurations.

4.1. Effect of actuators deflection ()


The efficiencies of A90 (=90 ) and A45 (=45 ) actuators are compared to investigate the
influence of (see figure 1). The A90 configuration is the most efficient : except for very low
values of q, the jet controlled by the A90 actuator spreads more than the one controlled by
the A45 actuator at the same control mass flow rate. The orientation of the four small jets has
therefore an effect on the control efficiency.
One possible explanation is that, for equal control flow rates, the azimutal component
added to the velocity field is larger for the A90 than for the A45 configuration. The A90
actuator only adds an azimutal component to the velocity field while the A45 actuator adds
both azimutal and axial components. Therefore the main flow sweeps the actuator effect along
the jet axis more easily.


4.2. Effect of the distance from actuators to main jet exit (h)
The efficiencies of SW90 (h=8mm) and A90 (h=30mm) actuators are now compared to
investigate the influence of h (see figure 1) on the jet angle. Figure 4 shows that the distance
h between the actuator and main jet exit influences the control efficiency. For each value
of q, the jet spreading is higher when the flow is controlled with the SW90 actuator which
has the smallest h value. This result shows that the control effect vanishes along the nozzle
presumably because of wall friction and confinement. As the SW90 configuration has been
identified as the most efficient it was retained for all further studies.

Active jet control for combustion applications

4.3. Effect of control flow rate on the velocity field


Figure 5 shows the effect of increasing q on the radial profiles of mean axial and rms velocities
at x/D=6 where D is the main jet exit diameter. Those profiles are obtained by using a single
hot wire probe. Each profile is normalized by the mean centerline velocity at the jet nozzle
exit of the free jet configuration (Uc0 ). Figure 5 shows that higher q values lead to larger
mixing enhancement : the increase of q leads to a decrease of the mean centerline velocity
and to an expansion of both mean and rms velocity profiles. Velocity profiles on figure 5 also
show that the entrainment of the ambient air is favored when increasing the control-to-main
mass flow rate ratio.

Figure 5. Mean radial profiles of axial velocity (up) and rms axial velocity (down) at x/D=5
for different control flow rates.

Active jet control for combustion applications

Figure 6. Mean radial profiles of axial velocity. Comparison between LES and experiments.
Unforced case (q=0).

5. Numerical results
In addition to experimental characterization, the actuators effect was also studied using
simulations to :
understand the phenomena induced by control and identify the mechanisms responsible
for mixing enhancement,


validate LES as a tool to study active control by comparing LES and experimental results.

5.1. LES validation for the unforced case


LES is validated first on the unforced case. Figure 6 shows radial profiles of mean and rms
axial velocities at different distances from the jet nozzle exit (x/D=3 and 7) for both numerical
and experimental tests. LES seems to be able to reproduce the experimental results : the
mean profiles extracted from the simulations are in good agreement with the experimental
ones. The amplitude of the velocity fluctuation is slightly overestimated at x/D=3, especially
in the mixing zone. Nevertheless, figure 7 shows that LES predicts the entrainment rate,
in agreement with experiments and literature on free round jets [9]. Both experimental and
numerical streamwise mass fluxes are calculated by integrating the mean velocity profile,
assuming that the flow is axisymmetric.

Active jet control for combustion applications

Figure 7. Normalized streamwise mass flux. Comparison between LES, experiments and
Gutmark and Grinstein data [9].

5.2. LES validation for the forced case


Figure 8 shows radial profiles of mean and rms axial velocities at different distances from the
jet nozzle exit (x/D=3 and 7) for both numerical and experimental tests. The configuration
corresponds to a control mass flow rate of 15% of the main mass flow rate (q=0.15). As for
the unforced case, figure 8 demonstrates that LES reproduces both mean and rms velocity
when control is on. The agreement between numerical and experimental results is better for
the forced than for the unforced case. It is probably due to the fact that the large structures
involved by the control are correctly captured by LES.
5.3. Effects of control
Figure 9 presents the mean axial velocity field in a transverse plane at a distance of 0.5 mm
upstream from the nozzle exit. Two-dimensional vectors have been superimposed to the field :
each of those is the projection of the local three-dimensional velocity vector on the plane. As
expected, figure 9 shows a global rotation movement of the flow inside the nozzle. LES
reveals that each of the four actuation jets induces a longitudinal vortex. This mechanism is
also found for classical jets in cross flow (JICF) where the interaction between jet and cross
flow leads to a contra-rotating vortex pair (CVP) [34]. In the present case, one vortex only is
observed because the other one is constrained by the wall and has no place to develop.
The jet spreading after the jet nozzle exit can also be visualized by computing the velocity
in a cylindrical coordinate system. Figure 10 shows the orthoradial velocity (u ) field at a
distance of x/D=1 from the jet nozzle exit for both forced and unforced configurations. The
level of u reached in the forced configuration is very high compared the unforced case (scales
are different in both cases). It shows that the control device has the expected effect : it adds
swirl to the flow. This swirl is responsible for the mixing enhancement with the ambient air.

Active jet control for combustion applications

10

Figure 8. Mean radial profiles of axial velocity. Comparison between LES and experiments.
Forced case (q=0.15).

Figure 9. Actuator effect on the flow inside the nozzle : mean axial velocity field and 2D
velocity vectors.

Active jet control for combustion applications

11

Figure 10. Flow field of orthoradial velocity component at x/D=1 for both forced and unforced
cases.

Figure 11. Effect of control on the flow. Vortex detection using Q criterion. Control is
activated at t=T0 . Animation (1.57Mo).

Figure 11 illustrates the important topological differences between unforced and forced
cases by displaying isosurfaces of Q criterion used for vortex detection [35]. Without control
(at T0 ), the Q isosurfaces are organized in rings which destabilize at a distance of three to four
nozzle diameters downstream. When control is activated these rings destabilize much faster
and the jet becomes turbulent right after the nozzle (T0 5 10 4 s to T0 1 5 10 3 s).


Active jet control for combustion applications

12

Figure 12. Jet spreading enhancement. PLIF visualizations. Influence of the control flow
density.

6. Scaling parameters controlling actuation efficiency


The previous results have shown that the control-to-main mass flow rate ratio q controls the
efficiency of the actuators. All those results were obtained by injecting the same gas (air) in
the main jet and in the four actuators jets. Since it is well known for JICF that density ratio
between jet and crossflow plays an important role, both experimental and numerical tests are
repeated injecting a lighter (air-helium mix) or a heavier gas (CO2 ) in the actuators.
Figure 12 presents different instantaneous concentration fields (obtained using PLIF) of
the flow for different values of q and different species injected through the actuators. At a
constant value of q, lighter gases produce a higher jet spreading angle : for each of the three
values of q the jet spreading is higher when using the air-helium mix and lower when using
CO2 than when using air. In those configurations, the value of q is not representative of the
control efficiency. The same trend is observed for each value of q between 0 and 0.5, as shown
on figure 13. On the other hand, when plotting the jet spreading angle vs. the impulsions ratio,
defined by :
J


2
act Uact
2
jet U jet

(2)

Active jet control for combustion applications

13

Figure 13. Jet spreading vs. q (left) and J (right). Effect of the control flow density.

the three curves corresponding to each species collapse and the jet spreading angle grows
linearly with J.
Simulations highlight the phenomena responsible for efficiency differences that occurs
when the control gas density changes. Figure 14 shows concentration isocontours of the
species injected through the actuators at various distances from the control flow injection.
The three configurations use the same flow rate ratio (q=0.15) but different gases in the control
device.
Figure 14 confirms the experimental results : q is not representative of the control effect.
LES reveal that the concentration fields exhibit two main modifications, concerning the global
rotation movement and the interaction between the secondary jets and the main one. The
global rotation movement can be visualized through the rotation of the vortical axial structures
around the main jet : the injection of light gas enhances the swirl intensity ; the vortical
structures turn faster around the main jet axis and so does the main flow. The flow topology
is also affected by the control gas density through the interaction between the control jets and
the main one. For heavy gas (J 1) the control jets are entrained by the main flow and do not
penetrate it. On the other hand, for light gas (J 1) the penetration is higher.


7. Conclusion
Experimental and numerical investigations of the control of a jet by four radial actuation
jets have shown that swirl injection (obtained by shifting the axis of actuators jets) is an
efficient way to control mixing and jet spreading over a very wide range. The effect of the
deflection and the distance between the control jets and the main jet axis has been studied. The
most efficient configuration in terms of mixing and jet spreading enhancement is the SW90
actuator device where actuation jets are located close to the nozzle and oriented to provide
maximum swirl. Large Eddy Simulations have been validated on both unforced and forced
cases. The simulations show which mechanisms are responsible for the control efficiency : the
control jets add swirl (which was expected from the design of the actuators) that enhances the
jet spreading. However, the LES reveals that the actuator choice (4 small jets) also induces

Active jet control for combustion applications

14

Figure 14. Mean concentration isocontours of the species injected through the actuators.
Effect of the control gas density. Controled flow with q=0.15

secondary vortices which may play a role for the reacting cases. Both experimental and
numerical results performed with lighter or heavier actuator gases confirm that the control
efficiency is governed by the impulsion ratio J (Eq. 2).
8. Acknowledgments
We gratefully acknowledge the support provided by Air Liquide and ADEME.

Active jet control for combustion applications

15

9. References
[1] Crocco, L. and Chang, S.I., Theory of combustion instability in liquid propellant rocket motors,
Butterworths Science Publications, London, 1956.
[2] Keller, J.O., Vaneveld, L., Korschelt, D., Hubbard, G., Ghoniem, A.F., Daily, J.W., Oppenheim, A.K.,
Mechanism of instabilities in turbulent combustion leading to flashback, AIAA Journal, 20, 254-262,
1981.
[3] Yang, V. and Culick, F. E. C., Analysis of low-frequency combustion instabilities in a laboratory ramjet
combustor , Combust. Sci. Tech., 45 1-25, 1986.
[4] Poinsot, T., Trouve A., Veynante D., Candel S. and Esposito E, Vortex driven acoustically coupled
combustion instabilities. Journal of Fluid Mechanics 177, 265-292, 1987.
[5] Candel, S., Combustion instabilities coupled by pressure waves and their active control. 24th Symp. (Int.)
on Combustion, The Combustion Institute, Pittsburgh, 1992.
[6] Lieuwen, T. and Zinn B. T. , The Role of Equivalence Ratio Oscillations In Driving Combustion Instabilities
In Low NOx Gas Turbines, Proc. of the Combustion Institute 27, 1809-1816, 1998.
[7] Poinsot, T. and Veynante, D. Theoretical and Numerical Combustion, Edwards, 2001.
[8] McManus, K.R., Poinsot T., Candel S.M., A review of active control of combustion instabilities, Prog.
Energy Combust. Sci., 19, p. 1-29, 1993.
[9] Gutmark, E.J., Grinstein, F.F., Flow control with non circular jets, Ann. Rev. Fluid Mech., 31, p. 239-272,
1999.
[10] Lang, W., Poinsot, T., Candel, S., Active control of combustion instability, Combustion and Flame, 70
281-289, 1987.
[11] Polifke, W., Poncet A. ,Paschereit C. O. and Doebbeling, K. Control of Thermoacoustic Instabilities in a
Premixed Combustor by Fuel Modulation. Journal of Sound and Vibration, 245(3): 483-510, 2001.
[12] Bloxsidge, G. J., Dowling, A. P., Hooper, N. and Langhorne, P. J., Active control of an acoustically driven
combustion instability, Journal of Theoretical and Applied Mechanics, 6, 1987.
[13] Davis, A. and Glezer, A., A mixing control of fuel jet technology : velocity field measurements, in
Proceedings of the 37th AIAA Aerospace Sciences Meeting and Exhibit, Reno, NV, AIAA 99-0447,
p. 1-15, 1999.
[14] Susuki, H., Kasagi, N., Susuki, Y. Active control of an axisymmetric jet with an intelligent nozzle, in
Proceedings of the 1st int. Symposium on Turbulent Shear Flow, 1999.
[15] Davis, M. R., Variable control of jet decay, AIAA Journal, 20 (5), 606-609, 1982.
[16] Delville, J., Collin, E., Lardeau, S., Lamballais, E., Barre, S., Bonnet, J.P., Control of jets by radial fluid
injection, ERCOFTAC Bulletin, Vol. 44, p. 57-67, 2000.
[17] Beer, J. M. and Chigier, N.A., Swirling Flows, in Combustion Aerodynamics, Krieger, Malabar, Florida,
p.100-142, 1972.
[18] Gupta, A. K., Lilley, D. G., Syred, N., Swirl Flows, Energy and Engineering Science Series, Abacus Press,
1984.
[19] Panda, J. and LcLaughin, D.K., Experiments on the instabilities of a swirling jet, Phys. Fluids, 6, 263-276,
1994.
[20] Chigier, N. A. and Chervinsky, A., Experimental investigation of swirl vortex motion in jets, Journal of
Applied Mechanics, 443-451, 1967.
[21] Lucca-Negro, O. and ODoherty, T., Vortex Breakdown : a review, Prog. Energy Combust. Sci., 27, 431481, 2001.
[22] Shee, H. J., Chen, W. J., Jeng, S.Y. and Huang, T. L., Correlation of swirl number for a radial-type swirl
generator, Experimental Thermal and Fluid Science, 12, 444-451, 1996.
[23] Mathur, M.L. and Maccalum, N. R. L., Swirling air jets issuing from vane swirlers. Part1 : free jets, Journal
of the Institute of Fuel, 214, 214-225, 1967.
[24] Faivre, V. and Poinsot, T., Experimental and numerical investigations of jet control for combustion
applications, in Proceedings of FEDSM02, Joint US ASME-European Fluid Engineering Division
Summer Meeting, Montreal, Canada, 2002.

Active jet control for combustion applications

16

[25] Nicoud, F. and Ducros, F., Subgrid-scale modelling based on the square of the velocity gradient tensor,
Flow Turbulence and Combustion, 1999.
[26] Schonfeld, T. and M. Rudgyard, Steady and Unsteady Flows Simulations Using the Hybrid Flow Solver
AVBP, AIAA Journal, 37(11), p. 1378-1385, 1999.
[27] Angelberger, C., Egolfopoulos, F. and Veynante, D., Large Eddy Simulations of chemical and acoustic
effects on combustion instabilities, Flow Turb and Combustion 65(2), 205-22, 2000.
[28] Schluter, J. and Schonfeld, T. LES of jets in cross flow and its application to gas turbine burners, Flow
Turbulence and Combustion, 65(2), 177-203, 2000.
[29] Colin, O., Ducros F. , Veynante D. and Poinsot T., A thickened flame model for large eddy simulations of
turbulent premixed combustion. Physics of Fluids 12(7), 1843-1863, 2000.
[30] Colin, O. and Rudgyard M. Development of high-order Taylor-Galerkin schemes for unsteady calculations,
Journal of Computational Physics, 162(2): 338-371, 2000.
[31] Poinsot, T. and Lele S., Boundary conditions for direct simulations of compressible viscous flow, of
Computational Physics, 101, 104-129, 1992.
[32] Kaufmann, A., Nicoud, F. and Poinsot, T., Flow forcing techniques for numerical simulation of combustion
instabilities, Combustion and Flame, 131, 371-385, 2002.
[33] Priere, C., Gicquel L.Y.M., Kaufmann, P., Krebs, W. and Poinsot, T., Large eddy simulation predictions of
miking enhancement for jet in cross-flows, Journal of Turbulence, 5, 2004.
[34] Fric, T.F., Roshko, A., Vortical structure in the wake of a transverse jet, J. Fluid Mech., 279, 1-47
[35] Jeong, J., Hussain, F., On the identification of a vortex, J. FLuid Mech., 285, p. 69-94, 1995