Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

75

MULTIPHASE
FLOWS
Multiphase flows are encountered in a
large range of different applications and
therefore represent an important subject
in Fluid Dynamics. This domain has been
a subject of investigation at the VKI for
about 35 years. Gas-particle flows were an
initial research topic and gas-liquid flows
were added later. Spray dynamics is an
important topic with applications in the
safety of industrial processes and of internal
flows in solid propellant rockets. Gas
bubbles are also the subject of studies with
multiple applications in domains such as
the continuous casting of steel or flow
measurement systems. Slug flows have
also been investigated using advanced
experimental techniques such as PIV.
2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
76
The full-scale experiments have been performed in
collaboration with the Ecole des Mines dAls in
France [MP80].
Chlorine or carbon dioxide clouds are produced on an
open area where concentrations are measured at
ground level permitting an evaluation of the disper-
sion factor for different wind conditions and release
rates up to 20kg/min. The curtain presently tested is
2m to 3m high and 4.8m to 10m wide. It is composed
of 90 full-cone nozzles of 5mm in diameters with 0.4m
spacing. The total water flow rate is 300kg/min. A typ-
ical view of the curtain in downward operating mode
with a release of chlorine is shown in Figure 3. Local
mean chlorine concentration is measured by bubbling
a sample in an absorbing solution. Photo-ionisation
detectors measure local instantaneous concentra-
tions. The nozzles used have been systematically char-
acterised in the VKI-Water-Spray facility in terms of
droplet size and velocity. Typical lateral distribution of
the ground concentration measured after the curtain
during field tests are plotted in Figure 4.
Two-dimensional CFD simulation of field tests has also
been performed with the code Fluent [MP108]. Typical
vertical predicted concentration profiles upstream and
downstream of the spray curtain are plotted in Figure 5.
A substantial reduction of concentration - a factor of
five - is predicted at the ground level. The water cur-
tain efficiency for cloud dispersion is expressed in
MITIGATION OF MAJOR INDUSTRIAL
HAZARDS USING SPRAY CURTAINS
Industrial hazards and the mitigation of their conse-
quence is a great safety concern in the petro-chemical
and industrial gas fields. Nowadays, the water-curtain
is recognized as a promising technique to mitigate the
consequences of various types of industrial hazards.
Its simplicity of use, efficiency and adaptability to dif-
ferent types of risks such as gas dispersion, absorp-
tion, as well as fire-effect attenuation, makes it an at-
tractive tool.
The spray behaves as a direct-contact reactor ex-
changing heat, mass and momentum with the gas
phase. In case of toxic leakage, the water curtain acts
as a diluting, absorbing and heating device. In case of
a fire, it affords thermal shielding to maintain the in-
tegrity of nearby structures such as storage tanks.
Along with these applications, the water spray curtain
has become an important research topic at the VKI
where the methodology adopted involves laboratory
tests, full-scale experiments and multidimensional
CFD.
Laboratory tests of mechanical dispersion have been
conducted in the VKI Wind-Gallery (WG) facility with
simulating gases [MP71]. A close view of the test sec-
tion in Figure 1 shows the spray curtain acting as a
fence leading to the formation of a upstream gas re-
circulation bubble.
This effect may be modelled by the curtain-to-wind
momentum ratio, RM. Figure 2 shows that as RM
increases, the vertical concentration profile becomes
more uniform and the ground concentration reduces
as well.
Figure 2: Vertical concentration profile in WG
Figure 3: Field tests of water spray curtain
Figure 1: Cloud visualisation in the Wind-Gallery
2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
terms of the forced dispersion factor D
F
defined as the
ground concentration ratio without to with sprays op-
erating. The upwind recirculation zone due to wind-
source and curtain interaction as observed during
tests (see Figures 1 & 3) has been reproduced by the
numerical results.
This parameter is plotted versus the curtain-to-wind
momentum ratio in Figure 5 for the three considered
approaches. Despite the scatter due to different wind
conditions (intensity and direction) and the different
values of the curtain-to-cloud height ratios, the fair
agreement observed between the three approaches
denotes that the D
F
-RM is a pertinent representation
of the forced dispersion by the water-spray curtain
[AJ68].
The outcome of several years of such activities in this
field has been a general engineering code, CASIMIRE,
developed to design an efficient spray curtain accord-
ing to the accident scenario selected. The code has
been evaluated through several industrial exercises.
A typical scenario considers an accidental release
of chlorine characterised by a mass flow rate of
3.5kg/s. A cloud develops in atmosphere the stability
of in modelled by the Pasquil-Turner classes.
The height of the downward curtain is 3m and it is lo-
cated 15m from the release point. Figure 7 shows that
the increase of the wind speed deteriorates the cur-
tain efficiency. Above a wind speed of 4m/s no effect
of the water spray curtain is expected.
As a general rule of thumb it can be stated that if the
curtain-to-cloud height ratio is sufficiently large,
a RM-value of 10 should lead to D
F
-value of 10.
In the case of a reservoir fire, water sprays can pro-
vide thermal shielding to maintain the integrity of
neighbouring structures. Behaving as a filter, the wa-
ter droplets afford significant reduction of the incident
radiation that impinges on sensitive surfaces such as
petro-chemical or LNG storage tanks. The water cur-
tain may be located vertically downward in front of the
surface to be protected or oriented towards the tank
to form a continuous fog casing protecting the entire
tank wall. A physical modelling of the thermohydraulic
behaviour of liquid sprays including phase change and
radiative heat transfer has been developed and im-
plemented in the engineering code CASIMIRE [AJ69].
Thermal radiative shielding experiments by water
spray curtains have been conducted on the VKI Water-
Spray facility implemented with propane burners to
simulate a source of 14kW at 1000K and a IR concen-
trator to produce a uniform radiant flux of 11kW/m
2
on
a curtain characterized by nozzle density varying from
10 to 33 nozzles per meter.
Figure 6: Comparison of the different approaches
Figure 7: CASIMIRE
prediction
Figure 5: CFD simulation of full-scale experiments
Figure 4: Ground concentration after water curtain
77 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
78 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Figure 8 summarizes a comparative exercise between
numerical simulation and laboratory experiments. The
attenuation factor is plotted versus the water flow rate
per unit of curtain length. The increase in pressure,
which leads to a decrease of the droplet size and an
increase of the droplet concentration, improves the
shielding performance.
Moreover, in the frame of the common European proj-
ect ASTRRE, campaigns of full-scale experiments have
also been conducted to evaluate the water curtains in
real industrial conditions. Trials have been performed
successively on a Gaz de France site located at
Saint Etienne de Montluc near Nantes (France) and on
the GESIP site, near Vernon (France). Fires of few tens
of kW/m
2
were produced in a LNG pool, 2m wide -
25 m long (Figure 8), and in an unleaded petrol basin
of 1000 litres.
A typical radiometer signal recorded during a LNG
field test is plotted in Figure 10. The shielding effect of-
fered by the water spray curtain appears clearly:
a mean attenuation of about 70% may be expected for
flow rates ranging from 60 to 120kg/min. The compari-
son between the field test data and the CASIMIRE sim-
ulation reported in Table 1 shows that the code is able to
predict thermal shielding in an industrial situation.
After deeply investigating the performance of thermal
shielding by vertically downward water curtains, the
VKI studies have been oriented towards more prom-
ising configuration based on liquid sprays impinging
directly on the structure to be protected. Potential im-
provement is then anticipated because the thermal
shielding is now due to both the impinging droplet
phase and the water film blanketing the wall.
To assess the performance of such a system, labora-
tory tests have been also conducted in the VKI Water-
Spray facility adapted for such thermal shielding ex-
periments (Figure 11) [AJ69]. The effects of different
design parameters such as the nozzle-to-wall distance,
the spacing between the nozzles and the liquid mass
flux have been studied. For configuration of sprays,
characterised by a liquid mass flux ranging from 0.11
and 0.18 l/s/m

, the global attenuation factor exceeds


easily 85%.
A theoretical development relying on the modelling of
radiative transfer in vertical sprays, thermohydraulic
and thermal absorption in a turbulent falling film has
been carried out [MP70]. It allows the prediction of the
downward evolution of the attenuation factor, Very
good agreement between prediction and laboratory
tests is found.
Figure 8: Attenuation factor
of shielding by vertical spray curtain
Figure 9: LNG fire and thermal
shielding by spray curtain
Pressure [MPa] 0.44 0.7
Field Tests 35 [%] 57[%]
1D model 45 [%] 60[%]
Table 1: Attenuation factor: comparison of
field test-and CASIMIRE prediction
Figure 10: Fluxmeter signal during field tests
Multiphase Flow
79 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
The model has been harnessed to
perform the dimensioning of
thermal shielding for the mostly
sensible LNG terminal facilities.
Figure 12 shows the mapping of
the water mass flux and the film
temperature on an 11m high com-
pression station building exposed
to a radiative flux of 45kW/m
2
.
Attenuation of 95% is achieved
and liquid temperature does not
exceed 100C.
DYNAMICS OF TWO-PHASE
FLASHING FLUIDS
Two-phase flows hold an interest in many areas of sci-
ence and engineering. In the safety field, one such top-
ic is the accidental release of flammable and toxic pres-
sure-liquefied gases; the failure of a vessel or pipe in the
form of a small hole results in the formation of a two-
phase jet containing a mixture of liquid droplets and va-
por. These droplets are generated by flash boiling (boil-
ing violently because the droplet temperature is above
the boiling point in ambient pressure) and aerodynam-
ic fragmentation. Of specific interest in this area are the
mathematical models and predictive computer codes,
which may be applied to such releases. It is hoped that
these models will embody an adequate understanding
of the processes involved, and may be used in design-
ing and assessing techniques to improve transport and
storage design, site location, layout and other safety
features. Experimental data on the size and velocity dis-
tribution of the droplets are required to make an accu-
rate prediction of the likely consequences of any given
two-phase release from known initial storage condi-
tions. This study focuses on the understanding of the
source processes, with emphasis on flashing release of
flammable pressurized liquids.
Thermal non-equilibrium is necessarily present in flash-
ing situations, i.e. when changes in the system pressure
result in superheating of the liquid and production of va-
por. As an example of the absence of thermal equilibri-
um, let us consider the rapid depressurization of a liq-
uid system; in this case the liquid may drop well below
the saturation value corresponding to the temperature
of the liquid. It gives time for the bubbles to nucleate
and grow by drawing heat from the surrounding liquid.
Under adiabatic conditions, the vapor formed can ob-
tain its latent heat of vaporization only at the expense of
the sensible heat of the remaining liquid. Because of the
relatively large vapor pressure of the material, a com-
bination of fluid dynamic and thermodynamic insta-
bilities will then lead to break-up into small droplets
(Figure 1).
Equilibrium will be reached when the fraction of the liq-
uid converted to vapour has extracted enough energy
from the residual liquid. For the axial temperature evo-
lution of this jet with air entrainment, using thermocou-
ples and different nozzle exit diameters, it is seen that
the temperature goes below the boiling temperature
which is T
boiling
=- 26.7C at 1atm and reaches a plateau
at -60C (Figure 2). Ambient air entrainment leads to ice
formation on the thermocouple and further down-
stream, it provokes an increase of the temperature.
High quality data sets to study the flashing expansion
in detail are being provided by employing advanced
measurement techniques such as high speed visuali-
zation devices, laser techniques (Particle Image
Velocimetry (PIV), Global Rainbow Thermometry
(GRT), Phase Doppler Anemometry (PDA)) as well as
conventional techniques. Preliminary studies were
performed to develop an understanding of droplet
size, velocity distributions and temperature evolutions
using laser-based and intrusive techniques. From the
velocity evolution obtained from a simple preliminary
set-up using PIV and PDA along a two-phase flashing
jet where break-up into small droplets has already oc-
curred, it is observed that the velocity increases down-
stream of the nozzle due to flashing evaporation and
both techniques show similar trends ([MP22,MP23,
AJ55], Figure 3).
A new experimental set-up, which allows the pres-
surisation of liquid R134A under different pressures
and testing of different orifice openings without al-
lowing flashing before the nozzle exit, has been in-
stalled.
Figure 11:
Thermal shielding by
impinging sprays
Figure 12: Thermal shielding
by impinging sprays
Mass flux [kg/s.m2] Temperature [C]
Multiphase Flow
The non-equilibrium nature of
the near field regions creates
difficulties in obtaining accurate
data measurement with intru-
sive techniques like thermocou-
ples. Such probes introduce se-
rious difficulties in measuring the liquid jet temperature
because intrusiveness induces flashing break-up and
changes the flow characteristics at the contact point.
Therefore, a non-intrusive technique like Infrared
Thermography is also applied to the R134A jet, which
exits the orifice as a liquid core and breaks up down-
stream. Figure 4 is a typical thermogramme where the
liquid core and a thermocouple placed on the centreline
of the jet can be identified [MP74, MP87].
The effect of different initial pressures, temperatures
and orifice diameters on the break-up pattern, droplet
size and velocity distributions along the radial and ax-
ial directions is studied in the new set up by means of
high-speed imaging and Phase Doppler Anemometry
(PDA). It is observed that the droplet sizes and
velocities are larger in the center and
they decrease toward the sides. The
droplet sizes and velocity mean val-
ues decrease going farther from the
nozzle due to evaporation and the jet
becomes wider. For the same back-
pressure and superheat, the high-
speed camera images display that the
jet exiting from a larger nozzle pres-
ents a more violent breakup than the
smaller one. Additionally, the disinte-
gration of the larger jet is closer to the
nozzle exit. The high-speed images
presented in this study show also that
the behavior of the jet with superheat
increment up to 6.4C changes from
slow expansion to complete shatter-
ing of the jet with a spray-like behav-
ior (Figure 5).
PDA measurements are performed at
the axial distances of x/D = 110 and
x/D = 220. The D
10
and D
32
values are
computed on the total cross-section at
these axial locations. Their evolution
is plotted in Figure 6 with respect to
Jacob number, Ja, that is the ratio
between latent and superheat (Ja = (C
pL
/ T
12
) / h
LG2
).
A clear decrease in droplet diameters is observed with
the increase of Ja, whatever is the pressure, nozzle di-
ameter or axial location. With the superheat effect, the
D
32
and D
10
values approach each other significantly.
In other words, with very high superheat the distribu-
tion converges to a mono disperse one. The increase
of the storage pressure results in different flow pat-
terns with different flow temperatures. High liquid
pressure provides a less sharp decrease in mean drop
size with increasing superheat than low liquid pres-
sure. Nozzle diameter has almost no effect on the drop
sizes compared to superheat effect. These conclusions
are valid for the nozzle diameters investigated.
Increasing the length-to-diameter ratios of the nozzle
result in smaller droplet diameters compared to the
sharp edge orifice. The non-dimensional velocity
profiles for the droplets and gas phase show a self-
similar pattern and can be represented with the
Figure 1: General view from two-phase flashing jet
Figure 2: Thermocouple tem-
perature measurements along
the jet axis using different ori-
fice diameters
Figure 3: Centerline velocity
evolutions along the jet axis
Figure 4: Infrared thermography
image for the flashing two-
phase jet under the pressurisa-
tion of P = 14 bars
Figure 5: High speed images for three different su-
perheat values for a jet exiting a 1mm nozzle
80 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
81
correlations in the literature for single-phase turbulent
jets and two-phase jets [MP122,MP123,MP145,MP146,
MP151,MP230,MP234,MP237,TH21].
The direct application of this research is the operation
and design of safe industrial plants dealing with flam-
mable and/or toxic gases due to improved prediction
capabilities and thus reduces the possible conse-
quences both for human lives and investment, if an
accident occurs.
BUBBLE DYNAMICS AND SLUG FLOWS
During continuous steel casting, argon is injected at
the submerged entry nozzle (SEN) to prevent clogging
in the upper part of the SEN (see section on continu-
ous casting), to promote the floatation of solid inclu-
sion particles and to avoid re-oxidation of the steel.
Nevertheless, buoyancy of the gas bubbles changes
the flow pattern significantly in the upper zone of the
mould. By studying the flow properties, the mecha-
nisms of entrapment of gas bubbles and mould slag
can be determined.
In the course of research on continuous steel casting,
a water model of a mould has been built in Plexiglas
at a scale 1:2. A Froude-based similarity criterion is ap-
plied to transpose laboratory results to the actual
process. The LIF/PIV technique has been used in the
upper part of the mould to investigate the recircula-
tion zones and estimate the maximum velocities un-
der the free surface. Figure 1 shows a photograph of
the bubbly flow obtained using backlighting. A spe-
cific pre-processing of the images has been developed
to measure the liquid phase velocities in the areas
with high void fraction [MP33]. In these areas, bubbles
reflect the laser light in all directions and excite out-
of-plane fluorescent particles. Therefore, the fluores-
cent emission does not comes not only from the plane
but from the entire volume. Bubbles will then act as
small lenses and appear in the diffraction images. To
limit their influence in the cross-correlation algorithm,
a mathematical operator that is sensitive to the gradi-
ents of intensity was applied (high-pass filtering)..
The high-pass digital filter has been applied to 300 im-
ages recorded from the PIV/LIF technique. Mean ve-
locity vectors and magnitude of the flow of liquid are
reported in Figure 2.
The institute started
with a new partner
(Centre Technique des
Industries Mcaniques,
CETIM) to investigate
the measurement of
the flow rate and pres-
sure losses in two
phase flow (gas-liquid)
at ambient tempera-
ture (thermo dynami-
cal equilibrium) or close to the triple point (high evap-
oration). This research is particularly important to
design correctly the control valve and the safety valve
protecting the cooling network of pipes of the pres-
surised two-phase circuit. The present research focus-
es on the measurement techniques that are not intru-
sive because it is specially important to avoid
supplementary pressure losses increasing the gas
phase volume fraction. For this purpose, electromag-
netic flow-meters are used. If this apparatus produces
an accurate measurement in single phase flow, the
use of it in two-phase flow is challenging.
Figure 6: The evolution of global D
10
and D
32
values
obtained from total cross-section with Jakob
number, Ja.
2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Figure 1: Photograph
of the bubbly flow
Figure 2: Velocity of the liquid phase
Multiphase Flow
82
The principle of the measurement is based on the
perturbation of the voltage recorded at an electrode
when bubbles are inducing perturbations of the con-
ductivity field associated with a constant or pulsed
magnetic field.
The development of this two-phase technique is sup-
posed to handle any gas quality (i.e. ratio of gas mass
flow to total mass flow) and therefore be used for any
regimes; from bubbly flow at low quality (Figure 3), to
bubbly flow at higher quality (Figure 4) but also to an-
nular flow (Figure 5) or slug flow (Figure 6). For all
these regimes, a measurement of the two-phase pres-
sure losses is done on a dedicated set-up. The regular
losses (horizontal and vertical) and singular losses
(bends, sudden restrictions and contractions) are in-
vestigated and compared to classical correlations.
After this stage, the pressure losses due to different
transparent but industrial valves (gate, butterfly,
globe,) are investigated.
In several projects that led to two PhD theses [TH22],
a deep investigation of the slug flow has been per-
formed. This flow is characterised by long Taylor
bubbles rising and almost filling a pipes cross-section.
Liquid moves around the bubbles and fills the space
between two successive gas slugs. While the nose of
the gas slugs is very stable, the area between it is
greatly agitated. In industry, slug flow may appear in
nearly any application employing two-phase flow in
pipes. The objective of the present research is to mod-
el unsteady slug flow, namely the coalescence be-
tween two Taylor bubbles, which is known to be de-
pendent on the flow in the wake of the Taylor bubble
rising ahead. However, to get a better understanding
of the mechanism of coalescence, the approach in-
volves characterising the hydrodynamics associated
with a single Taylor bubble for both stagnant and mov-
ing liquid in the pipe.
A facility consisting of a vertical column of Plexiglas
has been designed and built at VKI. A technique of PIV
using fluorescent particles was developed and used
to determine the velocity fields around one single
Taylor bubble [IB7]. A synchronisation method that al-
lows the acquisition of PIV images, always at the same
position, for different bubbles, has been implemented.
A spurious optical effect, called the mirage effect, oc-
curs at the bubble/liquid interface when a laser sheet
illuminates it and mirages of PIV particles appear on
the bubble interface, therefore it is not possible to de-
termine the gas-liquid interface from the PIV images.
This phenomenon has been extensively studied
and the results showed that the virtual images of par-
ticles appear due to reflection and refraction on the
gas/liquid interface. This mirage effect hinders the
determination of the bubbles shape from the PIV im-
age and the measurements of the shape of the bub-
bles have been performed with a shadowgraphy tech-
nique.
A recent technique for performing simultaneous
Particle Image Velocimetry and Shadography has been
applied for the first time to slug flow [MP27, MP38]. As
seen on Figure 7, the unsteadiness of slug flow creates
the need of recording the shadow of the Taylor bub-
bles at the precise instant and position of the PIV
measurements. Therefore, this experimental technique
is used in order to characterise simultaneously the
flow in the liquid and the shape of a single gas slug
(Taylor bubble) rising through a vertical column of
stagnant liquid using only one CCD camera [AJ35].
Figure 3: Dispersed bubbly flow (quality = 0.00046%)
Figure 4: Dense bubbly flow (quality = 0.0093%)
Figure 5: Annular flow (quality = 0.04%)
2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
The PIV images are recorded by seeding the liquid
with fluorescent particles and illuminating it with a
double cavity pulsed Nd: Yag laser. The shadowgraph
images are obtained by illuminating the flow from the
back with a double-pulsed light emitting diodes (LEDs)
array. The LEDs emit light at 650nm and the Taylor
bubbles produce a shadow, at that wavelength, that
passes through the optical filter and reaches the cam-
era. A composite image is therefore recorded that con-
tains the PIV and the shadow information. The contour
of the bubble slug is determined by digital filtering of
the image and the flow field by standard application
of the PIV processing algorithm WIDIM. An example
of such a composite image is shown on Figure 8. The
profiles of optical density are shown on this figure.
Figure 6: Chronologic pictures of a rising slug
(quality = 0.002%)
Figure 7: Unsteadiness of the slug
Figure 8: Composite image with optical density
profiles
83 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
84 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
During the post-processing the shadowgraph results
are combined with the velocity field acquired with PIV
solving some of the basic processing errors, which ap-
pear at the interfaces.
This study has been extended to non-Newtonian flu-
ids and experiments with aqueous solutions of car-
boxy-methyl-cellulose (CMC) have been performed
[AJ43]. The flow fields around the nose and the rear of
a gas slug rising through a vertical column with
non-Newtonian fluid are presented in Figure 9 and
Figure 11. These are the first quantitative measure-
ments for slug flow in non-Newtonian fluids and the
results are of great interest. The rear end of the slug
moving in a 1% CMC solution is shown on Figure 10.
The non-Newtonian nature of the fluid causes the slug
end to take the shape of the rear of a boat.
The results of this
work, developed dur-
ing these last years,
are essential to vali-
date existing models
and to understand the
mechanisms of slug
flow for Newtonian
and non-Newtonian
fluids.
SPRAY COOLING
To achieve specified metallurgical and mechanical
properties, stainless steel strips have to be annealed
and pickled, after the hot rolling process, which has
given the desired thickness to the sheet. This process
involves heating the strip to a temperature as high as
1150C and then to cool it, at high rate down to room
temperature. Nowadays the cooling technology leads
one to envisage complete water spray cooling.
A study has been undertaken at the VKI to develop an
engineering model allowing the design and optimisa-
tion of water spray cooling systems. The methodolo-
gy adopted combined both numerical modelling of the
strip heat transfer and experimental characterisation
of spray nozzles to be used on the industrial lines.
A generic schematic of a spray cooling system is
shown in Figure 1. It indicates also the main input pa-
rameters of the model.
Figure 9: Shape of the nose of the Taylor bubble
and flow field around it (solution of CMC 1%)
Figure 10: Boat Shape of the
rear of the Taylor bubble
(solution of CMC 1%)
Figure 11: Velocity field and streamlines in the
wake of a Taylor bubble (solution of CMC 1%)
Figure 1: Schematic of spray cooling of moving
strip
Multiphase Flow
85 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
The heat balance in the metal sheet is expressed ac-
cording to fin theory taking into account gas convec-
tive heat transfer, thermal radiation and water spray
cooling, respectively. The heat removal due to water
sprays is modelled from a close examination of the
different boiling regimes as a function of the local wall
temperature and local liquid mass flux. A general cor-
relation, of the local heat flux normalized by the criti-
cal heat flux is established.
The water mass flux distribution at the impact of the
spray is determined on the VKI Water-Spray facility al-
lowing tests with flow rate of 1 liter/s under a pressure
of 800kPa. Single nozzle or arrangements of nozzles
are tested. The facility is equipped with a phase
Doppler anemometer for the real time measurement
of velocity and size of droplets and with an array of
calibrated tubes to determine the mapping of the liq-
uid mass flux.
The final model results in a 2D non-linear heat equa-
tion, which is solved numerically using a finite differ-
ence method with application of the boundary condi-
tions appropriate to the industrial problem to be
simulated. The outcome is an engineering code which
can iterate on the water supply pressure to achieve the
prescribed thermal target.
Different simulations of industrial quenching have
been conducted. Figure 2 illustrates a typical output
of the engineering code. Figure 3a shows the water
mapping of the mass flux distribution in the case of
quench partitioning. The plot presented in Figure 3b
shows that such a configuration may lead to a smooth
decrease of the strip temperature [MP233].
MICROFLUIDICS
Microfluidics is the study of fluid flow in and around
submillimeter-geometries. For several years, it has re-
ceived increased attention at VKI; it concerns the trans-
port of ions, bubbles and heat in electrochemical coat-
ing processes, but also the stability of liquid jetting
through orifices with diameters less than 50m. For
these dimensions, optical diagnostics are necessary
but only possible if the required resolution is larger
than the wavelength of the employed light.
Particle image velocimetry (PIV) at micrometer scale is
called micro-PIV, which is available at VKI. Images are
taken through a microscope, magnifying the flow up to
64x. Laser-light reflections from the walls are separated
by means of an epifluorescent filter from the fluorescent
light scattered by submicron seeding particles.
Experimental results are compared to CFD simula-
tions, whether it concerns the formation of droplets
by an inkjet printer or bubbles during an electroplat-
ing process.
Figure 2: Boiling curve in water spray cooling
a)
Figure 3: Output of the water spray cooling code
b)
Figure 1: VKI micro-PIV equipment
Multiphase Flow
When governing forces and timescales are well esti-
mated, the upscaling of the microscale phenomena is
possible. This has lead to the Silicon-Jet facility, where
oil is jetted through a nozzle of several millimeters in
water in order to be able to neglect buoyancy forces.
Visualization techniques and standard measuring de-
vices can help to perform parametric studies on
geometries and fluid properties that are not possible
at micro-scale.
Upscaled measurements are performed using the
Phantom V7.1 high-speed camera (150000 frames/s).
The camera records images of droplet shadows or of
seeding particles; subsequently, in-house PIV algo-
rithms compute velocity vector fields. In the future,
heat transfer will be measured using infrared ther-
mography, and ion transport will be analyzed using
absorption spectroscopy.
SLAG ACCUMULATION
IN SOLID ROCKET MOTORS
Large solid rocket motors (SRM), such as the Ariane 5
boosters, are composed of a submerged nozzle and
segmented propellant grains separated by inhibitors.
During propellant combustion, a cavity appears
around the submerged nozzle and liquefied alumina
droplets are generated from chemical reaction of the
aluminium composing the solid propellant. The alu-
mina droplets being carried away by the hot burnt gas
are flowing towards the nozzle. Meanwhile the
droplets interact with the vortices formed by the in-
ternal flow and thus may modify their properties. On
the other hand, some of these droplets are entrapped
in the cavity, instead of being exhausted through the
throat, which generates an alumina puddle, also called
slag (see Figure 1). This slag, which accounts for about
five tons for the two Ariane 5 boosters, reduces the
performance of the solid propellant motor.
The goal of the VKI research, performed for the
European Space Agency, is to characterise experi-
mentally and numerically the driving parameters of
the slag accumulation in a stagnant area modelling
the nozzle cavity.
Figure 3: SiliconJet Facility for investigation of
droplet formation at scale 300:1
Figure 4: Sequence of shadow images
Figure 5: PIV image inside a droplet
Figure 2: CFD of droplet formation at microscale
Figure 1: Interaction of the droplets
with the nozzle
86 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
The two-dimensional transparent test section, connect-
ed to the stagnation chamber of a standard low-speed
wind tunnel, is designed in a way that allows the use of
optical measurement techniques to characterise the flow
field (Particle Image and Particle Tracking Velocimetry)
and to capture the liquid accumulation (free surface lev-
el detection technique LeDaR). This experimental mod-
el (Figure 2) simulates the main characteristic features
of an Ariane 5 booster: the vortices are shed from an in-
hibitor and the nozzle is also present together with its
neighbouring cavity. However, the flow is injected axi-
ally and the sonic condition at the nozzle throat is not re-
spected. The models of alumina droplets are produced
via a water spray installed in the stagnation chamber.
Different geometrical and flow parameters can be var-
ied: inhibitor size (h) and position (L
i
), nozzle throat open-
ing (o), cavity width (w), flow velocity (U
0
) and spray
characteristics.
The project started with a single-phase flow analysis
performed both experimentally and numerically
[MP208]. The PIV measurements are acquired in the
mid-span plane of the test-section. The region of in-
terest covers the whole middle plane of the test-sec-
tion from the inhibitor until the nozzle throat, includ-
ing the cavity. Numerically, the flow field is solved with
the commercial solver CFD-ACE+, with Large Eddy
Simulation (LES) of turbulence on a hybrid 3D mesh
[MP198]. The inlet conditions are set to respect the ex-
perimental conditions.
The comparison of the experimental and numerical re-
sults shows a good agreement (Figure 3). Furthermore,
the numerical simulation reveals complex instanta-
neous velocity fields and flow structures. In particular,
Qvisualizations are used to isolate the part of the flow
which is dominated by vorticity. It appears that the in-
hibitor is generating most of the coherent structures
through the break down of the shear layer (Figure 4).
The two-phase flow experiments are carried out using
a water spray to simulate the alumina droplets. First,
the Level Detection and Recording (LeDaR) technique
is applied to detect the free surface of the accumulat-
ed water (simulating the slag) in the cavity and to
study its mean shape along with the rate of increase
of its volume. A parametric investigation allows the
determination of the most important geometrical and
flow properties that influence the slag accumulation
process. An example of liquid accumulation and in-
stantaneous shape is provided in Figure 5. Among all
the parameters, it has been observed that the most
important one is the distance between the tip of the
inhibitor and the tip of the nozzle along the transver-
sal axis [MP242].
Then, the two-phase flow
field is characterized.
However, it is not possi-
ble to perform these
measurements with the
classical PIV and PTV
techniques since the trac-
er particles and the water
droplets cannot be distin-
guished from each other
due to fundamental opti-
cal and image formation
problems.
Figure 3: Single-phase flow transversal velocity
fluctuation (top: PIV; bottom: CFD)
Figure 2: The 2D test section
Figure 4: Q visualizations (color map: velocity
magnitude)
Figure 5: Liquid surface
captured by LeDaR and
liquid accumulation in
the cavity
87 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
88
Therefore, a two-coloured technique has been devel-
oped, where different fluorescent dyes are dissolved
in the tracer particles of the air-phase and in the water
droplets. Furthermore, the illuminating laser sheet
contains two different wavelengths (532nm and
266nm). Each of the applied dyes is absorbing light
only at one of the wavelengths and therefore they pro-
duce fluorescent emissions at two distinct regions of
the visible spectrum. Thus, using two PIV cameras
equipped with the proper optical filters, separate images
of the two phases can be recorded simultaneously.
Numerically, the liquid phase is created by random in-
jection of water particles at the inlet plane, with a
droplet size distribution respecting the experimental
one. The Q visualizations together with the liquid
droplets show that the path of the particles tracked in
the free stream is perturbed by the coherent struc-
tures, modifying the trajectory of some of them
(Figure 6). This phenomenon may be responsible for
the entrapment of these particles into the cavity. In or-
der to quantify the amount of particles filling the cav-
ity, a dedicated user subroutine is coupled with the
above simulation. This tool is used to propose other
geometries minimizing the slag accumulation in the
stagnant areas [MP243].
FLUID DYNAMICS
OF CONTINUOUS CASTING
Continuous casting has become an increasingly im-
portant step in the manufacture of steel in the last
three decades. Concurrent with the increase in pro-
duction levels, the quality requirements of the final
steel product have also become quite stringent. A
high-quality steel should satisfy the customer specifi-
cations for composition and the content of non-metal-
lic inclusions. The tolerance level for overall inclusion
content and fluctuations in composition in the final
steel product has been steadily decreasing over the
years, and this trend is expected to continue.
A continuous casting set-up can be divided into three
parts (Figure 1): the ladle, the tundish, and the contin-
uous casting mold. The tundish is an intermediate ves-
sel between the ladle and the mold, serving as a steel
reservoir during ladle change operations.
Several years ago, the VKI initiated a research program
on the behaviour of the fluid flow in a continuous cast-
ing process with the aim of improving the under-
standing of the mechanisms generating the defects.
The methodology used at the VKI combines experi-
ments on water models, numerical simulations, theo-
retical modelling and industrial tests:
Water model experiments are performed on three dif-
ferent configurations, simulating the ladle, the tundish
and the mold.
- A 1:2 scale model of a mold, the VULCAIN facility has
been designed and assembled at VKI. The configura-
tion allows full visual access of the flow in any part of
the tundish/mold system including the three-plate
slide gate and the stopper rod. The different valves,
stopper rod and sliding gates are remotely controlled
and allow full simulation of an automated process. The
water model can simulate the following casting con-
ditions: thickness of the mold (240mm), width of the
mold (900-2000mm) and maximum casting velocity
(2.5m/min).
- A 1:4 scale model of a tundish (PLUTON facility) has
also been designed and assembled at VKI. It allows
for non-intrusive optical measurements such as
Particle Image Velocimetry (PIV) or Laser Doppler
Figure 6: Q visualizations of the coherent structures
and droplet positions (color map: droplet diameters)
- The experiments on water models are per-
formed for the study of flow unsteadiness, for the
investigation of slag entrapment, for the study of
two-phase (air-water) coupling and for the cali-
bration of numerical models (CFD codes). New
control techniques are also tested on water
models.
- Numerical simulations are applied to the study
of the heat transfer in the mold and in the
tundish. Simulations are also performed for the
tracking of inclusions inside the tundish and the
determination of the removal rates for different
diameters of inclusions. The commercial code
Fluent is used for these numerical simulations.
- Theoretical modelling is used to quantify heat
transport phenomena, and also to better under-
stand the creation of bubbles and slag entrap-
ment.
- Industrial tests are performed in different in-
dustrial plants for the implementation of the im-
provements that are recommended based on
this research and verification of their influence
on the quality of the final products.
2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
Velocimetry (LDV). The facility has been built to allow
investigations of any tundish geometry using these
optical techniques. For this purpose, a double shell
was installed to avoid dioptric effects that might in-
troduce optical deformations to the light that crosses
the tundish. The inner shell can be easily replaced to
study different tundishes. The outer shell is 2110mm
long, 690mm wide and 385mm high. The facility can
simulate steady state conditions and also a transient
configuration corresponding to a change in ladle in the
industrial process. Flow rates at the outlet of the la-
dle and at the outlets of the tundish are remotely con-
trolled from a PC to simulate an industrial sequence.
- A new 1:4 scale model of the ladle (CHARON facility)
has been designed and assembled at VKI. The experi-
mental approach for the ladle consists of reduced
scale model studies of isothermal and non-reacting
systems. The scale has been adjusted to 1:4 except for
the height, which has been limited to half this value
because of a space constraint. The ladle diameter has
been set to 850mm with a total height of 450mm.
Most of the parts of this model have been made in
Plexiglas to provide a good optical access. Moreover,
the ladle has also been inserted in a square box filled
with water to reduce optical aberrations.
Experiments on the water model simulating the mold
have been made for the analysis of the distribution of
air (argon) inside the Submerged Entry Nozzle (SEN)
and in the mold in order to have a better understand-
ing of the two-phase flow and the transition regime.
Figure 2 shows a more detailed view of the bubble
repartition at the outlet of the SEN for a dispersed bub-
bly flow regime inside the SEN.
The model is also fully instrumented for quantitative
velocity measurements:
- PIV is used for the measurement of water and air bub-
ble velocities. A typical example of the flow field of wa-
ter measured with PIV is represented in Figure 3. This
flow field corresponds to a single-phase configuration,
meaning that no air is injected inside the SEN.
- A special probe based on drag measurement using
strain gages has been designed to determine the lo-
cal surface velocity. This measurement is used to eval-
uate the shear at the slag metal interface.
A commercial optical probe for the investigation of the
two-phase structure inside the SEN has also been im-
plemented. This probe allows the determination of the
local void fraction as well as bubble size and velocity.
Profiles of void fraction inside the SEN have been
measured. Figure 4 shows the optical probe mount-
ed on a special support to control the displacement
and position of the fiber optics.
The work is now aimed at developing a more detailed
understanding of the hydrodynamics inside the mold,
using the different measurements. Also, the water
model is fully regulated and the influence of the reg-
ulation on the two-phase flow pattern inside the SEN
has being studied in greater detail. A new technique
has been designed to optically characterize the free
surface of the mold (Figure 5).
Figure 1: Continuous casting - schematic
Figure 2: Two-phase flow at the outlet of the SEN
Figure 3: Flow field measured with PIV
(single phase)
Figure 4: Optical probe with support
89 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
90 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
This technique, called Level Detection and Recording
(LeDaR), relies on a computer analysis of the light in-
tensity gradient of images produced by a camera and
provides temporal and spatial information in real time
processing. The technique is developed to reproduce
on the water model the behavior of the industrial sen-
sors used to monitor the position of the free surface
of the mold in the industrial plant. Their impact on the
flow pattern inside the mold when used in a regulation
loop is also tested. The aim is to optimize the stability
of the meniscus over the whole width of the mold.
Parameters of the regulation will depend on the sensor
used and can be optimized using the water model.
Experiments on the water model simulating the
tundish are made for the analysis of the residence
time distribution of liquid steel when a new ladle is
feeding the tundish. Residence time distributions are
determined from the variation of conductivity be-
tween two different ladles (using salt as a conductor).
Minimum and mean residence time, as well as plug
volume, dead volume and mixing volume can be com-
puted from the residence time distribution. Different
operating conditions can therefore be compared us-
ing these parameters. They are correlated with inclu-
sion separation data obtained from industrial tests and
water model tests. The removal rate of inclusions in-
side the water model has been obtained first by using
the Phase Doppler Anemometer (PDA) technique and
by comparing the histogram of diameters of selected
particles at the inlet and outlet of the tundish. Particles
of appropriate size and density have been selected to
reproduce the behaviour of Alumina inclusions.
Two new optical techniques have been developed to
be used with mono-disperse particles. The first one is
used to measure the concentration of inclusions ver-
sus time by analysing the light scattered by the inclu-
sions when passing through a laser sheet. The second
one is used to measure the concentration of inclusions
versus time by measuring the extinction produced by
the clouds of inclusions when passing between a
diode laser and a photo-diode.
Colored dyes are also injected at different locations to
understand the flow structure and to verify results ob-
tained with the numerical simulations. The colorime-
try technique is also used for the analysis of unsteady
phenomena. A typical result is given in Figure 6, where
water exiting the ladle tube is colored in red (right side
of the figure), and is feeding a tundish full of trans-
parent water. Water will exit through the SEN (left side
Figure 5: Detection of the free surface of the
mold
Figure 6: Colorimetry test
Figure 7: Vortex formation in the ladle
Figure 8: Rotational velocity in the ladle versus
waiting time
Figure 9: CFD simulation
of a non-vortexing funnel
Multiphase Flow
91 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
of the figure). The model is fully instrumented for ve-
locity measurements, using either LDV or PIV.
Measurements of mean velocity and turbulence pa-
rameters have been made at different locations using
the PIV technique.
Experiments in the ladle have been performed to iden-
tify important phenomena during the teeming (emp-
tying) of steelmaking ladles. Vortex formation during
ladle teeming has been studied in considerable detail.
A typical result is shown in Figure 7. A graduated rule
is glued on the ladle and appears in the frame of the
camera to record simultaneously the vortex formation
and the water level.
PIV measurements have been performed to measure
the rotational velocity inside the ladle versus time
once the ladle has been filled with water. These meas-
urements have been used to validate the waiting time
used (300 seconds) before starting the experiments.
Figure 8 shows the evolution of the velocity at differ-
ent heights in the ladle versus waiting time.
The experimental results are also used to validate nu-
merical simulations. A typical velocity field coloured
by volume fraction is shown in Figure 9 (blue corre-
sponds to water). The beginning of the non-vortexing
funnel can be observed.
The investigation includes an assessment of the in-
fluence of several parameters such as water level,
slide gate opening, argon injection in the ladle tube,
argon injection around the outlet of the ladle, the bot-
tom geometry, immersion of the ladle tube, on vor-
tex formation, vortex behaviour and vortex diameter
versus time. In addition, different methods to sup-
press or delay vortex formation have been proposed
and have been tested.
LIQUID FILM COATING
Liquid film instabilities are encountered in many in-
dustrial processes such as strip painting; the produc-
tion of paper, photographic film, and magnetic tape;
packaging, etc. Usually, the liquid film should be thin,
continuous, uniform in thickness and smooth. For
many years the VKI has been deeply involved in re-
search on liquid film behaviour. Its expertise in this
field led the Institute to organise the 4
th
European
Coating Symposium [B1]. The methodology includes
mathematical modelling, and experimental and nu-
merical simulations. The studies are carried out in col-
laboration with several companies to examine the fun-
damental and applied aspects of these processes
pointing out their limitations.
One field of investi-
gation has been the
roll coating process
whereby liquid
flows into a narrow
gap between two
rotating cylinders,
the surfaces of
which move either
in the same direc-
tion (forward) or
opposite directions
(reverse). Some of
the liquid passes
through the gap
and splits down-
stream into two
thin films, each
coating one of the
rolls. This technique is extensively used in the paint-
ing, photographic and tape-recording industries for
covering a large surface area with one or several uni-
form layers. However, it suffers from severe lubrica-
tion and free-surface flow problems.
At high speed, a flow instability appears, in which the
coating becomes ribbed and uneven as visualised in
Figure 1. This is usually unacceptable in a coating
process, and so the study of this instability is of con-
siderable importance. A research programme involv-
ing experimental and theoretical aspects has been car-
ried out at the VKI to study the limitations of the
process.
Three experimental facilities of different complexity in
which the velocities of the cylinders and the gap sep-
arating them can be set to a high accuracy have been
developed to analyse the occurrence of ribbing and
the flow features in the nip. Two are two-roller set-ups
as shown in Figure 1: one is inclined, whereas in the
second the upper cylinder is vertically above the low-
er and is transparent. The third facility is a full-scale
prototype - PAINT4 - shown in Figure 2.
Figure 1: Formation of ribs
in roll coating. (This picture is
published in Ce que veulent les
fluides by E. Guyon , J.-P. Hulin,
L. Petit, 2005)
Figure 2: The VKI roll coating facility PAINT 4
Multiphase Flow
92
It consists of four rolls, two of which have rubber sur-
faces, and has been constructed to approach closely
the industrial process conditions. Forces due to the
elasto-hydrodynamic effect of rubber band rollers can
be measured by specially-designed strain gauges
mounted in the cylinders. A measurement technique
based on the laser triangulation method has been ap-
plied to gain precise knowledge of the local thickness
of the liquid film. A comparison between an elasto-hy-
drodynamic model and measurements on the PAINT4
facility is presented in Figure 3.
Different modes of roll coating including the gravure
coating have been tested on the PAINT4 facility. In par-
ticular it has been demonstrated that placing a taut
wire in the nip of the meniscus suppresses totally the
ribbing occurrence independent of the roll speed. A vi-
sualisation of such a phenomenon (shrewdness) is
proposed in Figure 4. Nowadays the PAINT4 facility is
currently associated with the facility LERO in the
Organic Coating Process Department of the Group
ARCELOR.
Study has also been addressed to the situation where
a slot gas jet impinges on a moving liquid film. This is
commonly called jet wiping. A fundamental research
program has been undertaken with the objective to
identify and model the actuators controlling the liquid
film behaviour. The study aims
at predicting the final coating
thickness and the eventual oc-
currence of liquid film insta-
bility.
Two facilities are available at
the VKI to study this mode of
coating process. The first set-
up called ESSOR is shown in
Figure 5. It includes a vertical
rubber strip, stretched be-
tween two rolls and set into
motion by the upper roll,
which is entrained by an elec-
tric motor. The strip velocity,
which can be adjusted precisely in the range 0.5m/s
to 4.5m/s, is measured by means of a tachometer. The
strip is dipped in a bath of water mixed with a surfac-
tant to ensure good wettability. The second setup is
presented in Figure 6. The liquid film is entrained by a
rotating cylinder in Plexiglas, allowing a complete op-
tical access. Using a diffuse light source inside the
cylinder, the liquid film patterns can be visualized pro-
vided that the working liquid is coloured. The pictures
are taken with a high-speed camera. Figure 7 illus-
trates typical visualisations of the liquid film experi-
encing surface instability after wiping. The film thick-
ness is measured by the light absorption technique
with an uncertainty of 7%. Digital image processing
yields complete mapping of the film thickness as
shown in Figure 8.
Figure 3: Effect of application force
of rolls on film thickness
Figure 4: Ribbing suppression
Figure 5 : Essor facility
Figure 6: New experimen-
tal setup:
(a) Sketch
(b) Picture of the setup
implemented for the light
absorption technique
( b)
( a )
2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
Another optical technique based on laser triangulation
is also used to measure the instantaneous local film
thickness with a spatial resolution less than 6%. It was
found that the instability amplitude is a function of the
nozzle-to-substrate standoff distance and substrate ve-
locity. The frequency analysis of the data proves that
several preferential wavelengths may coexist in most
conditions.
A linear theory based on the Orr-Sommerfeld equa-
tions of the liquid flow with the appropriate boundary
conditions has been developed. The solution is ob-
tained by the regular perturbation method.
The group velocity of the waves and their amplifica-
tion factor are computed. A dimensionless instability
criterion is derived in terms of the liquid film Weber
number. The comparison between experimental and
theoretical results is conducted on the basis of this in-
stability criterion. It is observed that the dominant
wavelengths found experimentally fall in the unstable
regime, indeed, as shown in Figure 9.
The instability features such as the mean amplitude of
the waves, their frequency content, as well as their ve-
locity, are computed. They are correlated to the wip-
ing actuators. Moreover, the importance of the wave
pattern type in the evolution of the instability ampli-
tude is put into evidence.
Attention has also been given to a more violent insta-
bility, called splashing, which occurs upstream of the
nozzle at high strip velocity [MP103,MP114]. Such an
instability, typical visualisations of which are shown
in Figure 10, leads to the formation of liquid droplets
that may plug the nozzle and degrade the wiping per-
formance. A theoretical and experimental investiga-
tion has been carried out on this topic.
The outcome is an engineering model, which allows
the prediction of the operating conditions leading to
droplet ejection. Good agreement has been found be-
tween the splashing criterion expressed in terms of
film Reynolds number and jet Weber number, and the
industrial line observations as depicted in Figure 11.
Figure 7: Visualizations of liquid film instability
Figure 8: Mapping of the local film
thickness in microns
Figure 10: Sequential visualizations of splashing
during jet wiping (high-speed camera)
Figure 9: Comparison of the observed
wavelengths with the instability criterion
Figure 11: Splashing criterion compared
to on-line observations
93 2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium
Multiphase Flow
94
All the findings of this research programme have been
compared with success to a LES-VOF simulation car-
ried out with the code Aquilon developed by TREFLE
laboratory of the University of Bordeaux, France
[AJ78]. An example is the validation of film thickness
predictions after wiping made in Figure 12.
VOF AND LES COUPLING
Requests for the numerical resolution of multiphase
problems involving a clear separation of the phases
through an interface are more and more common in
an industrial context. These systems require a correct
representation of turbulent motions for one or all the
phases. One may cite the dynamics of a liquid free sur-
face in a collection basin, or in industrial problems of
roll coating and jet wiping. In all those phenomena
where a typical density ratio of 1:1000 could be in-
volved, the gaseous phase is usually in a regime
where the flow can be considered definitely turbulent.
Multiphase simulations in such a circumstance are
definitely complicated by the necessity of resolving
correctly the turbulent motions in one phase and the
consequences in the other phase. This result can be
achieved when an industrial goal is fixed thanks to
the use of the Large Eddy Simulation technique (LES).
This last technique with its turbulence modeling of the
smallest scales (the Sub Grid Stresses -SGS- model-
ing), is more and more common in single phase sim-
ulations. Nevertheless, when a multiphase simulation
involves a free surface, the LES technique needs to be
appropriately modified to take into account the differ-
ent phases, the correct behavior of the SGS models at
the interface. The difficulty of coupling the multiphase
simulation technique (in this case the Volume Of Fluid
technique - VOF) and the LES one is clearly expressed
by its absence in most of the general purpose com-
mercial flow solvers, and by the rising interest it has
in the research environment. In this optic the EA
Department, thanks to its experience in the LES field,
has approached the problem using in-house rou-
tines for the SGS models to be used by two commer-
cial solvers, leaving to the latter the multiphase han-
dling. This explicit SGS modeling has been done
through, on one hand, User Subroutines in Fortran
(CFD-ACE+TM - ESI Group) and, in the other hand,
complex User Defined Functions in C (FluentTM). In
both circumstances, a proper developers environment
has been set up for advanced programming and par-
allel execution. Among the applications that are cur-
rently considered, an example is displayed in Figure 1,
where a jet at high speed is injected into a cavity con-
taining water. In Figure 2, an instantaneous flow field
is presented. Beside the use of commercial software,
the EA Department is also implementing in its paral-
lel LES solver (MIOMA) a multiphase capability in the
near future, in order to have an in-house parallel VOF-
LES solver.
Figure 12: Predictions of water film thickness
after wiping compared to measurements
Figure 1: Example Setup for VOF-LES Coupling
Figure 2: Air Water Interaction for a VOF-LES
simulation
2006, von Karman Institute for Fluid Dynamics, Rhode-St-Gense, Belgium

You might also like