5th International Symposium on Particle Image Velocimety
Busan, Korea, September 22–24, 2003
PIV’03 Paper 3207
General design and operating rules for seeding atomisers
C.J. Kähler
Abstract In a previous investigation it was shown that high concentrations of monodisperse tracer particles with a
mean diameter below
m can be produced by means of multi-hole nozzles only when the liquid volume inside
the atomiser is properly adjusted with respect to the flow rate delivered by the nozzles, Kähler et al. (2002). This
experimental result was explained by assuming that the nozzle has basically three functions. Firstly, it generates
the particles at the nozzle outlet. Secondly, it changes the fluid mechanical state of the fluid into a two phase liquid
which promotes the generation of smaller particle size distributions. Thirdly, the remaining kinetic energy which
is not consumed for the generation of the particles or the transition of the fluid mechanical state is transfered into
the turbulent motion of the liquid and act as an active impactor. In this paper four additional questions will be
examined. First of all, the assumption that the liquid volume inside the seeding generator has a major effect on the
particle size distribution will be proven. Beside the academic interest, this examination is important for all tracer
based measurement techniques because a varying liquid-level, due to consumption or contamination of the nozzle
outlets, increases the slip of the tracer particles. Secondly, the effect of the seeding temperature on the particle size
distribution will be examined. This is important for applications in powerful wind-tunnels without a cooling system
or when an experiment is performed at low temperature. Thirdly, the development of the particle size distribution
with time will be examined. This is important for investigations in closed circuit wind-tunnels where the tracer
particles remain for many cycles and interact with each other and with the surrounding atmosphere. Finally, the
mass delivered by the atomiser was measured for various pressure states. This investigation allows to calculate
the time to generate an appropriate seeding density in a wind-tunnel as a function of the nozzle-outlet number and
nozzle-pressure.
1
Atomiser
The generation of the particles was performed with a 300 mm height metal atomiser with an internal diameter of
200 mm, see left image of figure 1. The cover, that was made of perplex for monitoring of the liquid-level and
the operation of the nozzles, is equipped with three orifices. One 45 mm threaded hole in the centre for the nozzle
mount and the pressurised air adapter, a 5 mm hole with seal ring for the temperature measurement with a PT500
and a 1/2 inch hole with quick-release fastener for the delivery of the aerosol, see figure 1. The large dimension
of the outlet is important because high velocities in a small diameter pipes pushes liquid inside the test facility
and adulterates the measurements. A 220 mm long brass nozzle with 5 mm internal diameter and four crosswise
arranged outlets, each 1 mm in diameter and 20 mm ahead from the closed end of the tube, was built. To avoid free
surface effects during the mixing process and a bias due to the presence of the solid boundary, the nozzle-outlets
are located sufficiently below the liquid level. To adjust the liquid volume without changing the liquid load and
thus the liquid level above the nozzle outlets, three pipe-insets with a diameter of 50, 100 and 150 mm were built
which can be mounted inside the generator, see right image of figure 1. For the suppression of any secondary flow
between the flow regions separated by the pipe-inlets and the generator bottom, the insets where placed nuts, 5
mm in depth, which where carved on the inner side of base plate. The determination of the volumetric particle size
distribution was achieved with a laser diffraction technique which can resolve particle diameters down to
m
by analysing the three dimensional diffraction pattern of a particle ensemble according to ISO 13320-1.
Institut für Strömungsmechanik, TU Braunschweig, Bienroder Weg 3, 38106 Braunschweig, Germany, E-mail: [email protected]
1
Figure 1: Heatable atomiser with adjustable volume.
2
Liquid volume effects
In the first series of experiments the dependence of the particle size distribution on the liquid volume was examined.
Therefore, the atomiser was filled with 6 litre DEHS and a nozzle with four outlets was installed which was
operated at 2 bar overpressure. Figure 2 shows the measured particle size distribution and the cumulative sum for
three liquid volumes. Clearly visible is the decreasing number of large particles with decreasing liquid volume as
predicted by Kähler et al. (2002).
100
Cumulative percentage [%]
Frequency distribution
2
1
0
1
Particle size
[µm]
10
50
0
3
V = 6.28 dm
3
V = 3.53 dm
3
V = 1.57 dm
0
1
2
3
4
5
6
7
Particle size
[µm]
Figure 2: Volumetric particle size distribution for various liquid volumes measured at 8
9
10
bar.
Based on that result it can be stated that the ratio between the the volume flux through the nozzle outlets per second
and the liquid volume inside the seeding generator should be around
s . This is realized when a nozzle with
four mm drills is operated in 1.5 litre liquid at 1 bar pressure for example. In this case the jet velocity becomes
"!#
m/s at the orifice and the volume flux of the air is around 1 litre per second. As the liquid volume can be
controlled easily care must be taken with the contamination of the nozzle outlets. However, it is also important
that the nozzle outlets are sufficiently below the liquid surface in order to avoid an interaction between the air jet
and the liquid surface. This would lead to a strongly increased number of large particles as the particle generation
mechanism becomes different.
2
3
Viscosity effects
In a second experiment the loaded atomiser was cooled below zero degree Celsius and afterwards heated up to
%$&
Celsius in intervals of 30 degrees. Figure 3 reveals the volumetric particle size distribution measured with a
)
&
four-outlet nozzle in DEHS at '( 0, 30, 60 and
Celsius and * 0.5, 1 and 2 bar.
o
o
+ C
T=0
+ C
T=30
2
2
∆p = 0.5 bar
∆p = 1.0 bar
∆p = 2.0 bar
Frequency distribution
Frequency distribution
∆p = 0.5 bar
∆p = 1.0 bar
∆p = 2.0 bar
1
0
1
1
0
10
o
T=60
+ C
C
T=120
,
2
∆p = 0.5 bar
∆p = 1.0 bar
∆p = 2.0 bar
Frequency distribution
∆p = 0.5 bar
∆p = 1.0 bar
∆p = 2.0 bar
Frequency distribution
10
o
2
1
0
1
1
Particle size
[µm]
10
1
0
1
Particle size
[µm]
10
Figure 3: Volumetric particle size distribution generated with a four-outlet nozzle for various DEHS temperatures
and pressure states. The *
bar graph is hidden behind the *
bar results in the lower right plot.
Clearly visible is the dependence of the particle size distribution from the viscosity of the liquid. When the
&
temperature is low ('-
C) a four-outlet nozzle is suited to generate a small-band particle size distribution with
a cut-off diameter at
m, even when the criteria given in the last section is violated. At higher temperatures this
is not anymore valid. However, in contrast to the investigation presented in the previous publication, which were
performed in olive oil at room temperature, no significant shoulder can be observed in the distribution between
.
/01
m. This indicates the different behaviour between both liquids. Figure 4 shows the same measurement
p but here the suction nozzle, pesented in Kähler et al. (2002), was applied instead of a four-outlet nozzle. The
motivation for this experiment is the statement in the cited publication that the particle size distribution generated
with this nozzle type might be similar to the multi-outlet nozzle results when the viscosity is changed. However,
it can be seen that the answer to this open question is negative in the viscosity range analysed here. For small
' a broad-band bimodal distribution can be observed which indicates that the particle generation is based on
two different physical mechanisms. When the temperature is increased, the second peak decreases and only a
3
o
o
T = 60 C
2
∆p = 0.5 bar
∆p = 1.0 bar
∆p = 2.0 bar
∆p = 3.0 bar
Frequency distribution
1
0
12
o
T = 560 C
Frequency distribution
2
8∆p = 0.5 bar
4
Particle size
[µm]
∆p = 3.0 bar
1
12
o
C
T = 120
7
2
∆p = 0.5 bar
∆p = 1.0 bar
∆p = 2.0 bar
∆p = 3.0 bar
12
8∆p = 2.0 bar
0
10
3
10
3
8
8∆p = 0.5 bar
8∆p = 1.0 bar
1
0
8
8∆p = 1.0 bar
Frequency distribution
Frequency distribution
2
T = 930 C
8∆p = 2.0 bar
∆p = 3.0 bar
1
0
10
3
12
4
Particle size
[µm]
10
3
Figure 4: Volumetric particle size distribution generated with a suction nozzle for various DEHS temperatures and
pressure states.
small shoulder can be observed, especially when the nozzle pressure is low. This analysis implies that strong
viscous fluids are less suited for the generation of small-band particle distribution when the atomiser is equipped
with a suction-nozzle but well suited when large particles are desired to compensate a low laser power or camera
sensitivity for example. In order to link the previous measurements with the viscosity of the seeding liquid the
temperature-viscosity dependence was measured. Figure 5 shows the dependentcy for DEHS and sunflower oil in
linear and semi-logarithmic representation. The measurement was performed with an air supported ThermoHaake
&
rheometer with probe cover. For temperatures larger :
Celsius the oil measurements are biased due to volatile
components and impurities and the DEHS results due to evaporation and gap-draining. Nevertheless, the viscosity
can be estimated with sufficient accuracy. Based on this result it can be seen that the viscosity was varied by more
)"#&
than one order of magnitude in the temperature range between zero and
Celsius.
4
200
1000
175
DEHS
Oil
100
125
η [mPa s]
η [mPa s]
150
DEHS
Oil
100
75
10
50
25
0
;
0
25
50
75
o
T [ C]
;
100
<
125
1
=
150
;
0
25
50
75
o
T [ C]
;
100
<
125
=
150
Figure 5: Dependence of the viscosity on the temperature for oil and DEHS.
4
Coagulation and evaporation effects
Beside the investigations, which focus on the particle size distribution right behind the outlet of the seeding generator, the temporal evolution of the distribution was examined because this is important for flow investigations
in circulating wind-tunnels. For this experiment a large volume was filled with particles and the particle size distribution was measured in intervals shown in the legend displayed in figure 6. It can be seen that the maximum
100
Cumulative percentage
Frequency distribution
2
1
0
1
Particle size
[µm]
10
50
∆t = 0 min
∆t = 2 min
∆t = 5 min
∆t =10 min
∆t =20 min
0
0
1
2
3
4
5
6
7
Particle size
[µm]
8
9
10
Figure 6: Temporal evolution of the particle size distribution.
1
of the distribution moves gradually towards larger diameter with increasing time and particles larger
m arise.
1
However, after 10 minutes the distribution becomes stationary and the maximum is fixed at
m. This behaviour
can be explained by the coagulation of the particles. When only two species of particles are considered with >
and >%? being the radii, @
and @A? the coefficients of diffusion and B and BC? the number of particles per unit
volume, the coagulation rate D can be expressed by DFEHGIJ>
*K >)?)LMIN@ OK @A?%LPB B#? . This formula states that
the coagulation of equal size particles is small, relative to coagulation of particles which differ strongly in size,
because the coagulation rate becomes minimal when > ->%? . Qualitatively this is obvious because when > and >%?
are small it is unlikely that the particles meet each other and when both radii are large the coefficients of diffusion
SR
R
$UTV) ?XW
DZY becomes small due to the decreased movability -[\^] or mean square
@Q
' with %
@0` of the particles. The fact that no particles with a diameter larger
distance _ ? m appear in figure 6 can
5
be explained by the increased sedimentation of the large particles. The sedimentation velocity can be estimated
from the balance between the gravitational force ] g ba p c and the
law from Stokes ] w d:"GeC>f[ with
. friction
k0Tl%
Z
)
W
nm m/s or 1.6 mm/min
egihCB to [j
for
a
particle
with
m
and B p po
kg/m (DEHS), and
p
.
qrksTt% W
%
[A
m/s or 0.16 m/min for p m. The effect of evaporation should be mainly visible on the left
hand side of the distribution as the small particles evaporate mutch faster than the large particles becomes smaller.
However, no significant increase or decrease can be observed at all.
5
Mass flow analysis
To estimate the time to obtain a desired seeding density, the mass delivered by atomiser was determined for various
, 4 and 12), see figure 7. Based on the assumption that the
pressure states and number of nozzle orifices (uv
5
10
0.03
4
10
{
3
n=
{
60
n=
3
10
{
n=
{
n=
12
0.01
2
10
{
0.00
0
12
0.02
Volume [m ]
Mass flux [g/s]
n =1
n =4
n =12
MUB
NWB
LLF
4
1
n=
1
0
w
0.5
1
x
1.5
2
y
Pressure [bar]
2.5
3
z
10
1
x
10
Time [min]
100
w
Figure 7: Left: Efficency of the atomisers. Right: Loading times for wind-tunnels.
Z
$|Tl)
}
particles are spherical and
m in diameter, the number of particles delivered per second is
at *t
.
Assuming that a seeding density of 10 particles per cubic millimetre is sufficient for reliable PIV measurements
the loading time for a volume can be estimated, see figure 7. The horizontal lines indicate the volume of typical
low-speed wind-tunnels in Europe. The MUB is a closed-circled wind-tunnel at the TU-Braunschweig with a
W
volume of o m , the DNW-NWB is the 3 meter facility located at DLR Braunschweig with an estimated volume
kkf! W
$#
W
of
m and the DNW-LLF is the largest low-speed wind-tunnel in Europe with a volume of
m . The
oblique lines denote the capacity of atomisers with various nozzle-outlets operated at *g
bar. It can be seen
that an atomiser with a four-hole nozzle is sufficient to fill the MUB within 2 minutes at *
.
6
Conclusion
Four main conclusions of practical relevance can be deduced from the examination presented above. First, for
Z
the efficient generation of monodisperse tracer particles with a mean diameter below
m the ratio between the
volume flux provided by the multi-hole nozzles and the liquid volume inside the atomiser must be around
s .
Secondly, an increasing viscosity promotes the generation of small particles in case of the multi-hole and Laskin
nozzle. Thirdly, the maximum of the particle size distribution converge to a diameter of
m in time when the
seeding density is sufficiently heigh such that coagulation effects become significant. Finally, the capacity of
atomisers with different nozzle loadings was investigated to obtain a graphic that allows to estimate the loading
time for wind-tunnels.
6
7
References
Kähler CJ, Sammler B, Kompenhans J 2002 Generation and control of tracer particles for optical flow
investigations in air. Experiments in Fluids 33, 736–742
Acknowledgement
The research has been stimulated by work performed by the author within the EC funded EUROPIV 2 project. It
has partly been supported by the German Aerospace Center (DLR), Institute of Aerodynamics and Flow Technology, Göttingen.
7