Content uploaded by Samir Tambe
Author content
All content in this area was uploaded by Samir Tambe on Oct 09, 2015
Content may be subject to copyright.
ILASS Americas, 21st Annual Conference on Liquid Atomization and Spray Systems, Orlando, Florida, May 18-21 2008
A Study of Liquid Jets Injected Transversely into a Swirling Crossflow
S. B. Tambe*, and S. -M. Jeng
Department of Aerospace Engineering and Engineering Mechanics
University of Cincinnati
Cincinnati, OH 45221-0071 USA
Abstract
An experimental study has been conducted to study the effect of a swirling crossflow on transversely injected liquid
jets. Three in-house designed axial swirlers with vane exit angles of 30°, 45° and 60° were used to generate the
swirling crossflow. Water jets were injected from a 0.5 mm dia orifice located on a cylindrical centerbody that pro-
truded through the hub of the swirler. The measurement technique used was multi-planar 2-D PIV and Mie-
Scattering. The crossflow generated by the swirlers was observed to be fully-developed, axisymmetric and had a
vortex structure similar to solid body rotation. Total crossflow velocities peaked at a radial location near r = 15 mm.
Liquid jets injected in this crossflow were observed to spin along with the crossflow. The radial penetration of the
jet continued to increase even at far downstream locations due to centrifugal forces and a reduction in crossflow
velocities at high r. The radial penetration of the jet increased with an increase in the swirl strength of the crossflow
and with the momentum flux ratio (q). The jet plume continued to expand as it moved downstream.
*Corresponding author
Introduction
The transversely injected liquid jet in crossflow has
numerous applications including fuel injection [1],
thrust vector control of rockets [2] and lubrication of
the bearing chamber [3]. The injection of a liquid fuel
jet into a crossflow is also one of several possible con-
cepts of a premix module for lean, premixed, pre-
vaporized (LPP) combustion for aviation gas turbines.
The demands for higher efficiency of power production
and smaller engines lead to an increase in the operating
temperatures and pressures. This leads to increase in the
production of effluents like oxides of nitrogen (NOx)
since NOx formation rates increase with temperature
[4]. With the increasingly stricter ICAO regulations on
engine emissions, there is a strong emphasis on the de-
velopment of low-emission combustion techniques.
One technique is the LPP combustion, where a lean
homogeneous fuel-air mixture is created just upstream
of the combustor inlet. The presence of excess air
throughout the primary zone ensures that the combus-
tion temperature is low enough to suppress NOx forma-
tion. The LPP model requires a premix duct where fuel
and air are mixed together. To achieve a good homoge-
neous mixture and to avoid coking, fine atomization
and careful fuel placement are needed. The liquid-jet-
in-crossflow has characteristics of rapid atomization
and controllable penetration [1], which make it a good
choice for LPP fuel injection.
Jet-in-crossflow is a fundamental flow field and
has been the subject of numerous experimental as well
as computational studies. Tambe [5] and Elshamy [6]
have conducted detailed literature reviews of the work
done in this area. However, most of these studies fea-
ture uniform crossflows, i.e. where the crossflow veloc-
ity does not change in the transverse direction. Only a
few studies have reported crossflows with non-uniform
velocity profiles. Becker and Hassa [7, 8] injected liq-
uid jets into a counter-swirling double-annular cross-
flow and studied the effect of momentum flux ratio and
the air pressure on the jet behavior. Gong et al [9] pub-
lished a preliminary report on studies conducted for the
Lean Direct Wall Injection (LDWI) concept, where
they injected liquid jets into a swirling flow at different
injection angles. The authors previously conducted a
study where jets were injected transversely into a cross-
flow laden with a shear layer [10, 11]. The shear layer
was generated in the form of a slip plane between two
co-flowing airstreams and produced quasi-linear veloc-
ity gradients across the height of the test chamber, with
the jet being injected from a nozzle flush with the bot-
tom wall. The shear layer strength and the sense of the
crossflow velocity gradient had significant impact on
the jet penetration and the post-breakup spray
The objective of the present work is to study the
behavior of a liquid jet injected transversely into a
swirling crossflow. The effect of swirl strength on the
jet penetration and droplet velocity distribution is inves-
tigated. Three different axial swirlers were used to vary
the swirl strength of the crossflow. The measurement
technique used was 2-D Particle Image Velocimetry
(PIV). The study was divided into two parts. The first
part focused on characterizing the crossflow produced
by the different swirlers used. In the second part, the
impact of these crossflows on the jet was investigated.
Experimental Setup
Test Section
In-house developed axial swirlers were used to pro-
duce the swirling crossflow. The swirlers used had a
hub diameter of 2.22 cm (0.875”) and a tip diameter of
7.62 cm (3”). The test chamber had a square cross sec-
tion with an internal dimension of 7.62 cm (3”) and was
30.48 cm (12”) long. The walls of the test chamber
were constructed of 3.175 mm (1/8”) thick acrylic ma-
terial for optical access. The walls were mounted on
aluminum corner struts, which also provided a chamfer
to suppress corner recirculation zones (CRZ).
Jet injection was conducted through a centerbody,
which protruded from the swirler hub and extended
through the length of the test chamber. The centerbody
was a stainless steel tube of outer diameter 1.905 cm
(3/4”) and a wall thickness of 1.651 mm (0.065”). The
jet was injected from a 0.5 mm orifice located 2.54 cm
(1”) downstream of the swirler.
Figure 1a shows a Solidworks model of the test
section, and the co-ordinate frame of reference used for
the experiments. Figure 1b shows the actual test cham-
ber. The origin of the coordinate frame was located on
the axis of the centerbody at the streamwise location of
the center of the jet orifice. X-axis is located in the
X
Y
ZX
Y
Z
Figure 1. Test Chamber, a) Solidworks model, b) Ac-
tual chamber
a)
b)
ILASS Americas, 21st Annual Conference on Liquid Atomization and Spray Systems, Orlando, Florida, May 18-21 2008
Table 1. Swirler Properties
Vane Exit Angle No of vanes Mean Solidity Mean Aspect Ratio Swirl Number
30° 16 3.53 0.95 0.41
45° 12 2.83 0.85 0.71
60° 12 3.2 0.75 1.23
streamwise direction; Y-axis points vertically upwards
and the Z-axis points to the left as viewed from down-
stream. The centerbody was aligned so that the jet ori-
fice was centered about the Y-axis.
The test chamber assembly was mounted onto the
Horizontal Rig, which acts as a settling chamber for the
air flow. The Horizontal Rig is a long pipe of diameter
15.24 cm (6”), equipped with an inline 72 kW heater.
Air is provided from a Kaesar variable speed rotary
compressor capable of flow rates up to 0.907 kg/s (2
lb/s) at pressures of up to 13 bars (175 psig), connected
to a dryer and tank. A network of 10.16 cm (4”) and
5.08 cm (2”) dia pipes connects the air tank to the Hori-
zontal Rig. Air flow rate is metered by a Micro Motion
CMF300 coriolis flow meter and controlled by a stan-
dard globe valve.
The injected liquid was water, and was housed in a
water tank. A Nitrogen cylinder was used to pressurize
the water tank, in order to drive the flow as well as
maintain a constant flow rate. The water flow rate was
measured by a Micro Motion CMF010 coriolis flow
meter. A Parker metering valve was used for precise
control of the water flow rate. Figure 2 shows a sche-
matic of the experimental setup.
Swirlers
The axial swirlers used for the study were designed
in-house and were fabricated by the Rapid Prototyping
SLA technology. Three different swirlers were used,
with vane exit angles of 30°, 45° and 60°. Since the
focus of the experiments was on jet behavior, a smooth,
non-circulating, helical flow was desired.
A swirler configuration with straight, radial vanes
was chosen. The swirler streamwise length was taken to
be 2.54 cm (1”). The number of vanes for the 45° and
60° swirlers was chosen to be 12. To achieve reason-
able solidity, this number was increased to 16 for the
30° swirler. Then the only design parameter is the
variation of vane angle, i.e. the angle formed between
the vane and the corresponding radial plane, with
streamwise distance (x). Figure 3 shows 3 different
vane angle configurations for a swirler with a vane exit
angle of 45°. The simplest vane configuration is a heli-
cal vane, where the vane angle remains constant with x.
However the drawback of helical vanes is the sudden
change in direction of the air flow at swirler entry,
which might add to the flow turbulence. This change in
direction can be avoided by using a quadratic design,
where the vane angle changes quadratically with x, in-
creasing from 0° at inlet to the required angle at exit.
From figure 3, we observe that the quadratic vane pro-
duced a circumferential shift of 22.5°. A good rule of
thumb to achieve circumferential flow uniformity is to
ensure that the circumferential shift produced by the
swirler is close to 360°/(no. of vanes), which equals 30°
for the 45° swirler. To achieve the required shift, a mix-
ture of quadratic and helical vane designs was imple-
mented. For the 45° swirler, a quadratic design up to x
= 1.778 cm (0.7”) and a helical design henceforth was
implemented, which produced a circumferential shift of
Figure 2. Schematic of the setup
Figure 3 Swirler vane angle design
Figure 4 45° Swirler, a) Cutout view of the design, b)
Actual swirler
a
)
b)
28.2°. Figure 4 shows a cutout of the final swirler de-
sign and the actual swirler. Table 1 lists the detailed
parameters of the swirlers.
Measurement Technique
2-D Particle Image Velocimetry (PIV) was used for
flow measurements. A The PIV system used is a LaVi-
sion commercial PIV system. The Laser is a Double
Pulsed Nd:YAG laser, NewWave Solo PIV, with a
pulse energy of up to 120 mJ/pulse at a wavelength of
532 nm. The camera used is the LaVision Image In-
tense camera, which is a double frame - double expo-
sure CCD camera, capable of taking either two frame
buffers, or two images in rapid succession. A bandpass
filter at 532 ± 3 nm is used to restrict the light absorbed
to the laser wavelength. PIV records two images in
rapid succession with a known time difference. The
images are divided into interrogation windows, and
cross correlation is carried out on the two images to
determine velocity vectors for each interrogation win-
dow. For more details on the PIV system, please refer
to Elshamy [6].
For the current experiments, the PIV laser sheet
was aligned parallel to the X-Y planes. Multiple planes
were measured for each test condition and the results
were compiled together to create a 3-D map of the flow
field. The PIV laser and camera were mounted on a
three-axis Lintech traverse for precise control of loca-
tion and alignment in the different planes. A Velmex
VP9000 controller interfaced with a computer was used
to control the traverse movement. A schematic of the
PIV setup is shown in figure 5.
The results obtained from PIV were the Mie-
Scattering images and the PIV vector plots. The Mie-
Scattering images are sensitive to the droplet density at
any location and provide information about the jet
plume location. The PIV vector fields indicate the dis-
tribution of droplet velocities within the jet plume. For
each measurement plane, 200 images were recorded
and the Mie-Scattering images and PIV vector plots
were averaged over the 200 images.
The swirlers used in this study produce a clockwise
swirl as viewed from downstream. Since the jet orifice
is centered on the Y-axis, the measurement domain of
interest is the top, right quarter plane (y > 0, z < 0).
From figures 1 and 5, we can see that the optical access
to the chamber is mainly restricted by the corner strut,
which limits the measurement domain to y < 32 mm
and z > -32 mm. The measurement domain employed
for PIV is -5 mm ≤ x ≤ 42 mm, -5 mm ≤ y ≤ 31 mm, 0
mm ≤ z ≤ -30 mm.
The laser sheet, being parallel to the X-Y plane,
was incident upon the centerbody for a portion of the
measurement domain (z = 0:-9.55 mm). Due to this, a
part of the reflected light is incident upon the camera.
To avoid saturating the CCD and to reduce the noise
created by laser reflections, the centerbody was coated
with Fluorescent Paint (figure 1b). The paint converts a
portion of the reflected green laser light into red fluo-
rescent light, which gets blocked by the bandpass filter
on the camera. This alleviated the problems due to laser
reflection. However, there were certain locations where
the reflection intensity was too intense to be able to take
good data, and these locations had to be skipped.
X
Y
010 20 30 40
0
10
20
30
X
Y
010 20 30 40
0
10
20
30
Figure 7. PIV planar velocity distributions for case 2
(45°, We = 105.9), a) z = 0 mm, b) z = -25 mm
Figure 5. Schematic of the PIV setup
Z
Y
0-30
-5
30
Z
Y
0-30
-5
30
Z
Y
0-30
-5
30
Z
Y
0-30
-5
30
Figure 6. Measurement domain, a) Crossflow studies,
b) Jet injection studies
a
)
b)
a
)
b)
ILASS Americas, 21st Annual Conference on Liquid Atomization and Spray Systems, Orlando, Florida, May 18-21 2008
Table 2. Test Conditions for Crossflow Studies
Case No Swirler ma (kg/s) Vx (m/s) Vcf (m/s) We
1 30° 0.404 98.57 113.82 105.7
2 45° 0.347 81.15 114.76 105.9
3 45° 0.245 56.63 80.08 52.7
4 60° 0.245 56.73 113.45 105.5
Table 3. Test Conditions for Jet Studies
Case No Swirler ma (kg/s) Vx (m/s) Vcf (m/s) ml
(kg/min)
Vl (m/s) We q
5 30° 0.404 97.65 112.75 0.141 12.03 104.7 9.6
6 45° 0.349 80.44 113.16 0.141 12.03 106.7 9.4
7 45° 0.245 56.41 79.78 0.1 8.53 52.5 9.6
8 45° 0.349 80.44 113.76 0.199 17.01 106.7 18.8
9 60° 0.245 57.07 114.13 0.141 12.01 106.1 9.4
Test conditions
The 45° swirler was considered as the base swirler.
A base Weber number of 100 for air and base momen-
tum flux ratio of 10 for the jet was chosen. For cross-
flow studies, in addition to testing a We = 100 for all
swirlers, and additional We = 50 was tested for the base
swirler. Table 2 lists the test conditions for the cross-
flow studies. For jet injection, the case of We = 100, q =
10 was tested for all swirlers. The effect of a change in
We and q was also studied for the 45° swirler. Table 3
lists the test cases for the jet studies. Both We and q
have been defined based upon the total air velocities.
For water, a density of 996 kg/m3 and a surface tension
of 0.072 N/m have been assumed.
Results and Discussion
Crossflow Measurements
Initial tests were conducted to characterize the
crossflow. The crossflow was seeded with olive oil
droplets for these tests. 2-D PIV measurements were
carried out at a plane separation of 5 mm, from z = 0 to
-30 mm, as shown in figure 6a. The red line in figure 6a
represents the plane at z = -5 mm, which had to be
skipped due to intense reflections, so that 6 planes were
measured.
Figure 7 shows typical PIV results for the z = 0 mm and
-25 mm planes for case 2 (45°, We = 105.9), which is
the base crossflow case. For the z = 0 mm plane, the
centerbody exists up to y = 9.55 mm, explaining the
regions with no velocities. Even at z = -25 mm, the
presence of the centerbody in the background and sec-
ondary reflections were responsible for blank patches in
the velocity field. From figure 7, it can be observed that
there is no significant change in both the x- and y- ve-
locity components with x. Similar observations for
other cases led to the conclusion that the flow is fully
developed in the measurement domain. As a result, it
was possible to derive the velocities in the test chamber
cross-section by averaging over x. Figure 8 plots the
contours of Vx and Vy in the cross-section for case 2.
The negative values for Vy are a result of the clockwise
sense of the swirl, as viewed from downstream. The
absence of valid velocity measurements in the region
where the centerbody was present in the background
can be clearly observed as the areas with zero veloci-
ties. The trends in the velocities in figure 8 can be un-
derstood clearly if the cross-section is considered as a
polar frame of reference. Then Vx is observed to reduce
as we move radially outwards, while the magnitudes of
Vy increase as we move circumferentially away from
the Y-axis.
Figure 8. Cross Sectional Velocity distributions for case
2 (45°, We = 105.9), a) Vx, b) Vy
a
)
b)
The absence of the z-components of the velocity
hinders complete understanding of the flow. However
the flow could be assumed to be axisymmetric since it
originates from an axial swirler. Figure 9 tests the valid-
ity of this assumption by plotting Vx along two direc-
tions: along the lines z = 0 (Y-axis) and y = 0 (negative
Z-axis). In spite of the difference in the number of
points measured, the velocities compare very well, con-
firming the validity of the assumption of axisymmetry.
It was previously seen in figure 8 that the velocities
could be understood better from a polar frame of refer-
ence. Since the crossflow is axisymmetric as well, the
cross-section was converted into polar coordinates, as
shown in figure 10, with θ increasing clockwise from
the Y-axis. Then Vy along the Y- and Z-axes can be
considered as the radial (Vr) and tangential (Vt) veloci-
ties respectively. Figure 11 plots the radial variation of
the velocity components. Vx initially increases with r,
peaking around r = 12 mm and then gradually reduces
with r. Vt was observed to increase with r, with a n-
parameter from the vortex equation (equation 1) of
around 0.8. The n-parameter for free vortex and solid
body rotation are -1 and 1 respectively, indicating that
the crossflow is very close to a solid body rotation. The
magnitudes of Vr are very small compared to Vx and Vt
and can be neglected.
n
t
Vconstr
=
⋅ (1)[12]
An axisymmetric model was created, with the axial
and tangential velocities modeled by 4th and 2nd degree
polynomials respectively. Figure 12 plots the computed
as well as experimental Vx and Vt values for cases 1, 2
and 4, which represent the We = 105 cases for the three
swirlers. It can be seen that there is good correlation
between the modeled and experimental data. The Vx and
Vt standard errors were of the order of 4.5 and 3 m/s
respectively, which are about 5 % of the peak veloci-
ties. Thus the axisymmetric velocity model is a good
representation of the crossflow.
The Vx and Vy velocities were used to compute the
total crossflow velocities (Vcf), which have been plotted
in figure 13. For all swirlers, Vcf values peak at radial
locations near r = 15 mm and decrease on either sides.
The angle made by the total velocity vector (ψ) to
the Vx vector is representative of the direction of the
streamline at that location. These angles were calcu-
lated using equation 2 and have been plotted in figure
14. We observe that ψ values undergo a minimum cor-
responding to the location of maximum Vx (r = 12-15
mm) as expected. Average ψ values for r > 12 mm are
10 15 20 25 30 35
40
50
60
70
80
90
y, -z (mm)
V
x
(m/s)
z = 0
y = 0
Figure 9. Variation of Vx for case 2 (45°, We = 105.9)
10 15 20 25 30 35
-20
0
20
40
60
80
100
r (mm)
V (m/s)
V
x
V
t
V
r
Figure 11. Variation of Vx, Vr and Vt with r for case 2
(45°, We = 105.9)
θ
R
Y
-Z
Figure 10. Transformation from Cartesian to polar co-
ordinates
10 15 20 25 30 35
-20
0
20
40
60
80
100
r (mm)
V
x
(m/s)
30° model
30° exp
45° model
45° exp
60° model
60° exp
10 15 20 25 30
30
40
50
60
70
80
r (mm)
V
t
(m/s)
30° model
30° exp
45° model
45° exp
60° model
60° exp
Figure 12. Computed velocities for We = 105. a) Vx, b)
Vt
a)
b)
30.4°, 41.2° amd 45.4° for the 30°, 45° and the 60°
swirlers respectively. We observe that the 30° swirler
(ψ = 30.4°) is able to achieve flow angles close to the
vane exit angle, while the 45° and 60° swirlers (ψ =
41.2° and 45.4°) do not provide enough turn, with the
highest deficit being for the 60° swirler.
1
tan t
x
V
V
ψ
−⎛⎞
=⎜⎟
⎝⎠
(2)
Since the test chamber used had a square cross-
section, small corner recirculation zones (CRZ) will
form at the intersection of the walls, though the pres-
ence of chamfers at the corners is expected to reduce
their extent. Due to the presence of the corner struts, no
measurements could be taken near the corner of the
cross-section. As a result, for cases 1-4, no CRZs were
detected in the crossflow except for the 60° swirler
(case 4) in the measurement domain. For case 4, the
presence of CRZ could be detected as a sudden increase
in the negative Vy values for y > 27.38 mm in the z = -
30 mm plane.
Jet Measurements
The crossflow studies were followed by jet injec-
tion studies, where a water jet was injected from the
orifice on the centerbody. For jets, the separation be-
tween measurement planes was reduced to 2.5 mm for
better flow field resolution. Measurements were carried
out from z = 0 to -30 mm, as shown in figure 6b. The
plane at z = -7.5 mm (marked red in figure 6b) had to
be skipped due to excessive reflections, resulting in 12
measured planes.
The base case for the jet was case 6, i.e. We =
106.7, q = 9.4 for the 45° swirler. Figure 15 shows a 3-
D view of the jet for the base case obtained from collat-
ing the Mie-Scattering images from all planes. Figure
16 shows 3 views of the same jet as viewed along the
Cartesian axes. From these figures, we observe that the
jet emerges vertically upwards, and then starts spinning
along with the crossflow. As the jet moves downstream,
10 15 20 25 30
20
30
40
50
60
70
80
90
100
r (mm)
Vcf (m/s)
30°
45°
60°
Figure 13. Variations of Vcf with r
10 15 20 25 30
20
30
40
50
60
70
80
90
r, mm
ψ
(deg)
30°
45°
60°
Figure 14. Variations of flow angle (ψ) with r
Figure 15. Base jet case, case 6 (45°, We = 106.7, q =
9.4)
Figure 16. Views of case 6 (45°, We = 106.7, q = 9.4),
a) View from top, b) View from left, c) View from
downstream
a)
b)
c)
even though the penetration in the y-sense seems to
decrease, the radial penetration keeps increasing. This
is seen more clearly in figure 16c, where the cross-
section of the centerbody has been included for refer-
ence. Then we can clearly see that the radial penetration
kept increasing even far downstream. In fact, for all
cases tested, the jets hit the wall of the test chamber
well before completing half a revolution around the
centerbody.
In a typical jet in crossflow flow field, for low-to-
moderate q, the rate of increase of jet penetration be-
comes very small at large downstream distances. How-
ever, for the swirling crossflow, the jet penetration kept
increasing even far downstream. One of the explana-
tions for this phenomenon is the presence of centrifugal
forces acting on the jet due to the rotating nature of the
crossflow. These centrifugal forces aid the jet in pene-
trating further. Another effect is similar to what the
authors observed in their previous study on jets injected
in a shear laden crossflow [11]. In cases where the
crossflow had a negative velocity profile (Vcf decreased
with y) the reduction in the penetration slope decreased
as the jet penetrated further, allowing higher incre-
mental penetration, as compared to that for a uniform
crossflow. Since crossflow velocities decrease, the
crossflow aerodynamic forces also decreases, allowing
higher penetration. For the current study, the modeled
total crossflow velocities were observed to reduce with
r for r > 15, so a similar effect can be attributed towards
the continually increasing jet penetration.
Figure 17 shows the 3-D and the top view for case
9 (60°, We = 106.1, q = 9.4), which had operating con-
ditions identical to case 6, but for the 60° swirler. Com-
paring figures 15-17, two significant differences can be
deduced from the jet behavior in these two cases. The
jet for the 60° swirler hits the domain boundary at an
earlier streamwise location as compared to the jet for
the 45° swirler. This is expected, because of the higher
turn angles generated by the 60° swirler. From figure
16a, the streamwise location where the jet centerline for
case 6 hits z = -30 mm is approximately 37 mm, which
corresponds to a mean jet flow path angle of 39°. For
case 9, from figure 17b, the streamwise location for z =
-30 mm was around 28 mm, yielding a mean jet flow
path angle of approximately 47°. The mean flow angles
for the corresponding crossflows were 41.2° and 45.4°
respectively (table 4). Thus the jet can be seen to try to
follow the crossflow. However, care should be taken to
note that this was just a 1-D analogy of the flow. The
jet flow is highly three-dimensional in nature, and re-
quires a more detailed analysis.
The second difference between cases 6 and 9 is in
the jet penetration. The radial penetration for the 60°
swirler case is significantly higher than that for the base
swirler. This can again be explained in terns of the cen-
trifugal forces and the crossflow velocity gradients.
Since the crossflow spins more sharply for the 60°
swirler, the centrifugal forces experienced by the jet can
be assumed to be higher than for the case of the 45°
swirler. Also, the total crossflow velocity drops more
sharply for the 60° swirler, leading to even lesser oppo-
sition to jet penetrations at large r.
It was observed above that it makes more sense to
describe the jet penetration in a radial sense. Hence the
measurement domain was transformed into a cylindrical
frame of reference. The cross-section of the cylindrical
domain was identical to the polar domain used for the
crossflow. For the analysis of the crossflow described
previously, it was enough to convert the coordinates of
the measured points from Cartesian to polar due to the
assumption of axisymmetry. However, for the jets,
there is a need to track the jet boundaries in the cylin-
drical reference frame for jet penetration analysis.
Hence, a true cylindrical domain was desired. This was
achieved by first creating a very dense 3-D Cartesian
domain by interpolating the Mie-Scattering intensities
among the measured points. A cylindrical domain was
then defined and the intensity at each of these points
was obtained as an average of the points lying within
half a cell distance in each direction. To ensure good
results after averaging, the maximum cell dimension of
the intermediate dense grid needed to be at least as
small as the smallest cell dimension of the cylindrical
grid. The final cylindrical grid had a cell size of 0.25
mm in x, 0.5 mm in r and 5° in θ. The cylindrical do-
main used was -5 mm ≤ x ≤ 30 mm, 9.5 mm ≤ r ≤ 43
mm and 0° ≤ θ ≤ 100°.
Figure 18 shows representative contours of the jet
in the X-R (θ = 40°), R-Θ (x = 10 mm) and X-Θ (r = 20
mm) planes. It is to be noted that both the R-Θ and X-Θ
planes have been mapped onto a Cartesian grid for dis-
Figure 17. Views of jet case 9 (60°, We = 106.1, q =
9.4), a) 3-D View, b) View from top
play in figure 18. The R-Θ plane is actually a portion of
an annulus of inner radius 9.5 mm and outer radius 43
mm, while the X-Θ plane from figure 18c is the curved
surface of a cylinder of radius 20 mm. In each of these
planes, the jet can be observed to be oblong in shape
suggesting that none of them capture the true jet cross
section, which is true since the path of the jet centerline
is not aligned with any of the domain axes. This also
suggests that it is necessary to track the mean path of
the jet and predict the penetration and spread along this
path to be able to compare with existing jet-in-
crossflow data.
From the Mie-Scattering data in the cylindrical
domain, it is possible to track the radial as well as tan-
gential location of the edges of the jet periphery both in
the radial as well as circumferential directions, from
which we can determine the jet penetration and spread.
Figure 19 shows the x-variation of the radial and
circumferential locations of the upper radial edge of the
jet plume for cases 5, 6 and 9, which represent the ef-
fect of swirl angle with constant operating parameters
(We ≈ 105, q ≈ 9.5). From figure 19a, we observe that
the radial penetration increases with an increase in the
swirl angle, which has already been explained above.
Also, initially, for x < 4 mm, a slightly higher penetra-
tion for the 30° swirler is observed. However the radial
penetration up to this location is less than r = 18 mm,
which was radial location for the peak in the total cross-
flow velocity for the 30° swirler (Figure 13). As soon as
the jet enters this high crossflow velocity region, it
bends sharply and the radial penetration stays low
downstream of this location.
From figure 19b, we understand how the jet spins
with the flow. Figure 19b suffers from a lack of good
resolution in the circumferential direction, and hence a
lot of discrete jumps can be observed. For both the 30°
and 45° swirlers, the θ angle increases steadily with x.
However, for the 60° swirler, after x = 15 mm, the slope
of the θ-curve drops, indicating that the jet may no
longer be spinning with the crossflow and may have
gained enough momentum to penetrate straight on. It
can be noted that at x = 15 mm, the jet for the 60°
swirler has already penetrated 28.5 mm in the radial
direction, where the crossflow total velocity (Vcf) is less
than 70 m/s.
Figure 20 plots the circumferential width of the
spray for the cases from figure 19. We observe that the
spray for the 30° swirler starts narrowing after x = 8
mm, while the spray periphery for the 45° swirler is the
widest. However, a higher resolution in the circumfer-
ential direction is needed for a detailed analysis.
Figure 21 plots the radial jet penetrations for the
cases 6 (base), 7 (We = 52.5) and 8 (q = 18.8), and
demonstrates the effect of We and q on the jet penetra-
x (mm)
r (mm)
Jet In tensity in X-R pla ne
-5 0 5 10 15 20 25 30
10
15
20
25
30
35
40
0
100
200
300
400
500
r (mm)
θ
Jet Intensity in R-
Θ
plane
10 15 20 25 30 35 40
0
20
40
60
80
100
0
100
200
300
400
500
x (mm)
θ
Jet In tensity in X -
Θ
plane
-5 0 5 10 15 20 25 30
0
20
40
60
80
100
0
100
200
300
400
500
Figure 18. Typical contours of case 6 (45°, We = 106.7,
q = 9.4) in the cylindrical frame, a) X-R plane (θ =
40°), b) R-Θ plane (x = 10 mm), c) X-Θ plane (r = 20
mm)
-5 0 5 10 15 20 25 30
5
10
15
20
25
30
35
40
x (mm)
r (mm)
30°
45°
60°
-5 0 5 10 15 20 25 30
0
10
20
30
40
50
60
70
x (mm)
θ
30°
45°
60°
Figure 19. Location of the upper periphery for cases 5,
6 & 9 (We = 105, q = 9.5) a) Radial location, b)
Circumferential location
a
)
b)
c
)
a)
b)
tion for the 45° swirler. Reducing We (case 7) leads to a
minor increase in the radial jet penetration. Increasing q
(case 8) leads to a significant increase in the jet penetra-
tion. In the far downstream region, a sudden drop in
penetration was experienced for case 8, the reason for
which is not yet known. The effect of We and q are
identical to what is observed in traditional jet-in-
crossflow studies [1].
Figure 22 shows the droplet velocities in 4 of the
measured planes for the base case, case 6 (45°, We =
106.7, q = 9.4). Here, the absence of the knowledge of
the z-components does hamper the understanding of the
3-D flow since both radial and tangential velocities are
important for the jet flow. At the z = 0 mm plane (Fig-
ure 22a), the droplet velocities are mainly directed up-
wards. At z = -15 mm (Figure 22b), the jet has moved a
bit downstream. Also a large portion of droplets have
negative Vy velocities, especially near the Z-axis (y =
0). Additionally, the droplets near the upper periphery
still have positive Vy velocities, indicating that the jet is
expanding. Similar observations can be deduced from
the z = -20 and -25 mm planes. This indicates that the
jet periphery is still spreading even at z = -25 mm. Also
the spray coverage in the velocity plots is observed to
be significantly larger than that observed in the Mie-
Scattering images. This occurs because the Mie-
Scattering intensities are roughly proportional to the
droplet densities, so that regions with small number of
droplets are not captured in the Mie-Scattering images.
The trends observed in figure 22 were similar to
those observed for other cases. The largest droplet ve-
locities were observed away from the centerbody (z ≤ -
15 mm) and these were mostly located near y = 0 mm
with negative Vy velocities. The jet spray was observed
to expand both in the vertically upward and downward
directions, though the velocities at the upper peripheries
were significantly smaller than those at the lower pe-
riphery.
No CRZs were observed in the measurement do-
main for the crossflows, except for the 60° swirler,
where presence of CRZ was detected for r > 40.6 mm
(z = -30 mm, y > 27.38 mm). Since the jet radial pene-
trations were lower than r = 40.6 mm for all cases, the
CRZs are believed to have no effect on the jet behavior.
Conclusions
An experimental study has been conducted to study
the effect of a swirling crossflow on the behavior of a
water jet injected transversely into it. Axial swirlers
were used to generate the swirling crossflow, and jets
were injected from a cylindrical centerbody protruding
through the hub of the swirler.
The crossflow was observed to be fully-developed
and axisymmetric, with a vortex motion similar to solid
body rotation. An axisymmetric velocity model was
created, which was observed to have very good agree-
ment with the experimental data. The model predicted
that the total crossflow velocities exhibited a peak near
r = 15 mm. The flow angles predicted by the model
reveal that the flow turn achieved by the swirlers lagged
the vane exit angle except for the 30° swirler.
Water jets injected in the swirling crossflow were
observed to spin along with the crossflow. However the
radial penetration kept increasing even at large down-
stream distances. This is believed to be caused by the
presence of centrifugal forces acting on the jet and the
reduction in crossflow total velocities at high r. The
radial penetration increased with an increase in the
swirl strength.
For a detailed study of the jet penetration, the
measurement domain was transformed into a cylindrical
domain. Analysis in the cylindrical domain verified that
the jet radial penetration increased with an increasing in
the swirl strength and that the jet spins along with the
crossflow. However, as the jet penetrates higher into the
region of decreasing crossflow velocity, the rate of
circumferential movement of the jet decreases and it
seems to penetrate straight on without further spinning.
The jet for the 45° swirler was observed to have the
largest circumferential spread. Changing the Weber
number had little effect on the jet penetration; though
increasing the momentum flux ratio increased the radial
penetration considerably.
The droplet velocity distribution reveals that the jet
plume keeps expanding even as the jet travels far down-
-5 0 5 10 15 20 25 30
0
5
10
15
20
25
30
35
40
x (mm)
Spray width (deg)
30°
45°
60°
Figure 20. Variations of jet circumferential width with x
for cases 5, 6 & 9 (We = 105, q = 9.5)
-5 0 5 10 15 20 25 30
5
10
15
20
25
30
35
x (mm)
r (mm)
We = 52.5, q = 9.6
We = 106.7, q = 9.4
We = 106.7, q = 18.8
Figure 21. Effect of We, q on radial penetration for the
45° swirler
stream. Since the jet spins clockwise with the cross-
flow, high negative y-velocities were observed in the
bottom half of the plume. However, positive y-
velocities are observed at the top of the plume indicat-
ing that some droplets still keep penetrating higher. A
complete representation of the velocity field could not
be generated due to the missing z-velocity components,
since only 2-D measurements have been carried out.
Future Work
It was observed that the missing z-components of the
velocities hamper a full understanding of the 3-D flow
field, especially for the jet injection cases. Repeating
the tests with laser sheets aligned with Y-Z planes
could alleviate these problems. Also additional PDPA
testing could be carried out to investigate the droplet
distribution in the spray.
Nomenclature
m mass flow rate
q liquid-air momentum flux ratio
r radial coordinate, radius
V air, liquid velocity
We aerodynamic Weber number
x streamwise coordinate
y vertical coordinate
z lateral coordinate
θ circumferential coordinate, polar coordinate
ψ flow angle
Subscripts
a air
cf crossflow property (total velocity)
l liquid
r radial component
t tangential component
x x-component
y y-component
z z-component
References
1. Becker, J., and Hassa, C., Atomization and Sprays
11:49-67 (2002)
2. Chen, T. H., Smith, C. R., Schommer, D. G., and
Nejad, A. S., 31st Aerospace Sciences Meeting and
Exhibit, Reno, NV 1993
3. Birouk, M., Azzopardi, B. J., and Stäbler, T., Parti-
cle and Particle Systems Characterization 20:283-
289 (2003)
4. Lefebvre, A.W., Gas Turbine Combustion,
McGraw-Hill, 1983, Ch 9
5. Tambe, S. B., “Liquid Jets in Subsonic Crossflow,”
MS Thesis, Department of Aerospace Engineering
and Engineering Mechanics, University of Cincin-
nati, Cincinnati, OH (2004)
6. Elshamy, O.M., “Experimental Investigations of
Steady and Dynamic Behavior of Transverse Liquid
Jets,” PhD Dissertation, Department of Aerospace
Engineering and Engineering Mechanics, University
of Cincinnati, Cincinnati, OH (2007)
7. Becker, J., and Hassa, C., Journal of Engineering
for gas turbines and Power, 125:901-908 (2003)
8. Becker, J., Heitz, D., and Hassa, C., Atomization
and Sprays, 14:15-35 (2004)
Figure 22. Velocity planes for the base case, case 6 (45°, We = 106.7, q = 9.4), a) z = 0 mm, b) z = -15 mm, c) z = -
20 mm, d) z = -25 mm
a)
c)
b)
d)
9. Gong et. al., Journal of Propulsion and Power, 22,1:
209-210 (2006)
10. Tambe, S. B., Elshamy, O. M., and Jeng, S.-M, 45th
Aerospace Sciences Meeting and Exhibit, Reno, NV
Jan 2007
11. Tambe, S. B., Elshamy, O. M., and Jeng, S.-M., 43rd
AIAA/ASME/SAE/ASEE Joint Propulsion Confer-
ence & Exhibit, Cincinnati, OH, 2007
12. Gupta, A. K., Lilley, D. g., and Syred, N., Swirl
Flows, Abacus Press, 1984, Ch 1