7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
Analytical Investigation of Vortex formation in NPP with ANSYS CFX 12.0
F Blimeling and P. Pandazis and A. Schaffrath
TUV NORD SysTec GmbH & Co. KG, Department of Reactor Core and Thermofluiddynamics (ETR)
Grol3e Bahnstragle 31, D22525 Hamburg, Germany
fbloemeling @tuevnord.de
Keywords: vortices, pumps, water covering, emergency cooling in nuclear power plants, CFD
Abstract
Emergency core cooling systems require a clean and reliable water source for maintaining longterm recirculation
following a Loss of Coolant Accident. The heat removal pumps of Pressurized Water Reactors for instance take their
water from flooding tanks or the containment sump whereas Boiling Water Reactors use pump suction intakes located
in the suppression pool or the wet well. The pump suction intakes have to fulfill requirements which were developed
to ensure an undisturbed operation of the heat removal pumps. Particularly these requirements demand a vortexfree
inflow and an adequate submerge of the intakes to avoid surface vortices. Surface vortices quickly develop air cones
which lead to air entrainment into pumps. Air entrainment and inhomogenous inflow have strong negative influence
on the pump performance and should be avoided in general. Therefore the water level above the intake should exceed
a value based on design recommendations or suitable tests. Today TUV NORD SysTec is testing a third approach for
the determination of the appearance of surface vortices and the shape of their air cones.
In the first step the focus is put on the onset of air cones at the water surface since air entrainment due to vortices is
surely avoided if they do not appear. For this purpose two different experiments were analysed. In the first experiment
the necessary submerge for a scaled model of a flooding tank was calculated. In this case the inflow turns out to be
free of circulation. The necessary water levels and the shape of the surface computed by CFX are in good agreement
with the experimental data. The influce of circulation is examined in the second experiment. In this experiment an
assembly of vanes in a small cylindrical vessel induces a circulation on the flow field approaching a vertical intake.
Although the measured shape of the surface vortices for rotating flow could not be reproduced exactly the principle
appearance of air cones is satisfactorily predicted.
Introduction and motivation
After a Loss of Coolant Accident (LOCA) a longterm
recirculation of coolant is required in order to remove
the decay heat from the core. Therefore coolant is in
jected from a clean and reliable source to refill the reac
tor pressure vessel and to replace coolant which is lost
through the break. For this purpose the heat removal
pumps in Pressurized Water Reactors (PWR) take water
from flooding tanks or the containment sump, while in
Boiling Water Reactors (BWR) the water supply is based
on pump suction intakes located in the suppression pool
or the wet well. Thus, recirculation pump performance
under post LOCA conditions must be evaluated for both
types of light water reactors (LWR), cf. NRC (2003).
A reliable operation of the pumps requires in general a
homogenous inflow that is free of circulation. In the past
several problems occur at pumps during their comission
ing or even later under operating conditions. Often it is
difficult to identify the source of the disturbances. How
ever, in most of these cases the inflow conditions are
involved (Weinerth (2003); Knauss (1987)). Vortices for
instance which develop at bounding walls or the water
surface may cause a circulation in the flow approaching
the pump. Moreover the surface vortices might develop
air cones which could grow downflow and may lead to
air ingression in the pump. The consequences of un
favorable inflow conditions are, e.g., vibrations, noise,
decreasing flow rates or high power consumption. At
long sight these conditions can lead to (severe) mechan
ical defects or even the complete failure of the pumps,
see Weinerth (2003).
To avoid problems associated with air ingression re
quirements of pump suction intakes for emergency core
coolant pumps were developed by the German Reac
tor Safety Comission (RSK). The RSK demands, e.g.,
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
an adherence of an adequate intake submerge, so that
air suction by hollow vortex formation can be excluded
(RSK (2005)). Moreover the RSK proposes that for
vortexfree suction the water level above the intake must
exceed the value resulting from suitable tests or if such
tests are not available from the socalled ANSI for
mula, see ANSI (2000). Furthermore it is possible to
avoid harmful inflow conditions through constructive
optimization of the pump suction intake.
Guidelines like ANSI are developed to cover a large
variety of pump intakes. So they do not consider spe
cific geometric properties such as inclination of the in
take or walls. Therefore they lead to conservative values
for the submerge and distances to walls or boundaries.
If it is desired to specify more precise design options
for a given geometry precise, complex and normally ex
pensive experiments in a scale larger than 1:20 (NRC
(2003); ANSI (2000)) are necessary.
TUV NORD investigates and evaluates a third ap
proach to determine the occurence of vortices and to
specify the amount of air entrainment in the pump by
using CFD codes. Since surface vortices first produce a
dimple at the surface before they develop air cones air
entrainment can be avoided if no dimples occur on the
water surface. Therefore the aim of this paper is to in
vestigate in a first step the occurence of dimples on the
water surface during pump operation and the determina
tion of a necessary submergence of the intake by using
the commercial CFD code ANSYS CFX 12. For com
parison two different experiments with different initial
flow fields are evaluated.
The first experiment uses a 1:4 scaled model of a
typical flooding tank of a German PWR. The water is
pumped down with constant flow rates such that the wa
ter level in the vessel is continuously lowered. In the
experiment the focus is targeted on the onset of vortices
and the transport of air by those vortices. Furthermore
it turned out that an initial rotation of the flow field has
no significant influence on the onset of surface dimples
under the initial and boundary conditions which are ad
justed in the tests. In the second experiment we focus
on the influence of a forced circulation on the flow field.
For this purpose water is pumped out of a small cylin
drical vessel in vertical direction. But in contrast to the
first experiment water enters the vessel from the sides
and is given a variable tangential direction. Hence, a
specific circulation can be prescribed and its influence
on the vortices is analyzed.
The paper is structured as follows: In section 1 we de
scribe models which are suitable to predict a necessary
level for the submergence of the intake under certain as
sumptions. After a short classification of surface vor
tices models for the determination of the submergence
for nonrotational and rotational inflows of the pump are
f Coherent surface swirl
2
S Surface Dimple
Coherent swirl at surface
Dye core to intake
coherent swirl throughout
water column
4
0
0
a
Vortex pulling floating
trash but not air
Vortex pulling air
Sbubles to intake
o
SFull air core
to intake
Figure 1: Classification of surface vortices
given in 1.1 and 1.2 respectively. In part 2 we turn to
wards the CFD analyses. First we give a very short
overview over the modeling equations in 2.1 followed
by the flooding tank experiment in 2.2. The results con
cerning an influence of circulation are given in 2.3.
1 Models to determine a sufficient intake
submergence to avoid air entrainment
As already mentioned air entrainment into pumps can be
avoided by a sufficient submergence of the pump suction
intake. If it is too small first dimples occur at the water
surface in the beginning which might grow and develop
an air cone. This air cone might be stretched in the di
rection of the pump which causes air entrainment in the
end. The occurence of dimples depends on the state of
the flow field near the intake. It turns out that the height
of the necessary submergence increases with the vortic
ity of the flow field and constant intake mass flow. But
even if there is no circulation in the approaching flow
dimples and air cones can be observed. In this special
we talk about Bernoulli cones. The necessary submer
gence for Bernoulli cones is smaller than for swirling
vortices. Thus, in practice a vortex free flow should be
enforced by adding some internals. Otherwise the more
restrictive criteria for swirling vortices must be used.
As a criterion for the classification of air entrainment
the vortex type can be used. Hecker (Weinerth (2003))
classified vortices by six types which are shown in figure
1. Depending on the pump types one and two might be
tolerated, whereas types five and six must be avoided
since they cause a continuous entrainment of air into the
pump.
For practical considerations air entrainment can be ex
cluded if no or only very slight dimples are observable
at the surface. That means only vortices of type one and
two are allowed to occur. Due to the unsteady and tran
sient behaviour of the vortices optical observations of
the surface are not sufficient to classify the category of
air entrainment.
In the following we present some useful correlations
which are suitable to determine a minimal submergence
of pump suction intakes. First we assume only nonro
tational flows approaching the intake and in the second
subsection also rotational flows are considered.
.1 Nonrotationa
For nonrational flows KSB proposed an extension of
the Bernoulli equation (Wagner (2004)) to determine a
sufficient submergence of the intake to avoid hollow vor
tex formation. They added an offset of 0.1 m to the
Bernoulli equation to address also nonuniform inflows.
Hence, the submergence KSB is given by
) 2
SKSB 0.1 m + 
2.q
where v refers to the fluid velocity in the intake. The
validity range of equation (1) is given in Wagner (2004):
The velocity v should be less than 3 m and the diameter
of the vessel should be larger than the diameter of the
intake by a factor of 1.5. Furthermore minimal distances
of the intake to the bounding walls must be taken into
account.
.2 Rotationa.
In this case the Hydraulic Institute Standard proposed
a correlation which was also taken over in the ANSI
Standard (American National Standards for pump intake
design), see ANSI (2000). The correlation states that the
submergence of the intake should be larger than
ANSI d 1 + 2.3 (2)
where d refers to the intake diameter. It can be applied
in analogy to the KSB correlation if the intake velocity
does not exceed 3 m and the correlation is valid for all
angles of intake lines.
2 CFD analyses with CFX
Before we describe the experiments and compare the re
sults from CFX with the experimental data we start with
a short description of the basic partial differential equa
tions. To determine the appearance of dimples on a wa
ter surface both fluids the water phase and the gaseous
phase above the water surface (in this case air) have to
be modelled. Fortunately it can be assumed that the
gaseous phase is approximately at rest. Therefore its in
fluence on the liquid phase is neglectable and the equa
tions simplify. More details on this topic are given in the
following section.
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
2.1 Modelling equations
The fluid field is modelled by a pure Eulerian ap
proach. The basic idea is that both fluids can be present
as a mixture at a given position and time. Therefore it
is necessary to average the basic equations. Let for in
stance X(k) (X, t) be the indicator function of phase k,
i.e.
(k 1, if phase k is present at (x, t)
)() 0, otherwise.
If the indicator function is averaged over an appropriate
small time interval At we obtain the local void fraction
of phase k, i.e.,
(k) t+At
() () X(k) t) dt. (3)
Furthermore it follows directly from equation (3) that
a1) + a(2) 1. (4)
The mixture density p is obtained by
2
p ~( (k)
k=1
if p(k) describes the density of the single phase k. How
ever, to determine a suitable mixture velocity i the Favre
average
PU (6)
P
has to be used.
Since the influence of the air on the water is ne
glectable it can be assumed that the phases are mechani
cally balanced and share one mixture velocity field. This
approach is known as the homogenous model of two
phase flow, cf. Ishii and Hibiki (2006). So one finally
obtains the mass conservation equations for the mixture
S+ V(p)= 0 (7)
at
and for every single phase k
p a(kA )
P t + p (k) (. V) a(k) = Dk, (8)
where kl, I / k, describes the mass exchange from
phase k to 1. Thus, in our case this exchange term simply
vanishes.
Since we deal with turbulent flow the turbulence has
to be modelled additionally. For this purpose the fluid
dynamic quantities are splitted in main and fluctuating
parts which gives
U = i, + ut'
average fluctuations
and pu' 0
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
for the velocity.
Substituting equation (9) into the momentum equa
tion and averaging yields the Reynolds Averaged Navier
Stokes (RANS) equation for the mixture of air and water
(Ishii and Hibiki (2006); Wilcox (1994)):
pi + P( *V) u
att
 Vp + J
+ V (GS + at).
The terms on the right side of equation (10) are the
acceleration due to pressure, gravitation and the viscous
and turbulent stresses. Note, that due to our assumptions
no momentum exchange between the phases has to be
modelled in the momentum equation. For instance there
is no drag force between air and water. Furthermore the
influence of surface tension has been dropped. Due to
Knauss (1987) surface tension asur is unimportant for the
development of hollow vortices for Weber numbers
We pv2d
We
Usur
between 121 and 33856 (remind that d is the diameter of
the intake and v is the absolute value of the fluid velocity
in the intake) which is fulfilled for all experiments.
The turbulent stresses in the momentum equation (10)
represented by the Reynolds stress tensor
(at)j pa'.,, i, j 1,..., 3 (11)
are unknown in advance and have to be modelled by a
turbulence model. In the case of mainly isotropic turbu
lence twoequation models like the SSTModel (Menter
(1994)) are widely used. These models define an eddy
viscosity (t by two more turbulent quantities, e.g. the
turbulent kinetic energy k and the turbulent dissipation
frequency w,
k
PLt = p
LO
such that
(Ut)y = it + 
Therefore the number of transport equations raises by
two additional equations for k and w. The SST model is
used for the analysis of the flooding tank experiment in
section 2.2.
In strongly swirling flows the isotropy assumption is
not valid. Therefore we take advantage of a different
turbulence model in section 2.3 where the influence of
forced circulation on the onset of air cones is examined.
The SSG Reynolds stress model (Speziale et al. (1990))
does not use the eddy viscosity concept. Instead it solves
Figure 2: flooding tank model (Siempelkamp (2006))
six transport equations for the independent entries of the
Reynolds stress tensor. This model is also suitable for
flows with anisotropic turbulence.
Since all calculations are isothermal with constant
material properties equations (7), (10), (8) and (4) yield
a closed system of fluid dynamic equations which is
solved by ANSYS CFX 12.
2.2 Flooding tank experiment
As already explained in the introduction it is im
portant for nuclear reactor safety to exclude the oc
curence of hollow vortices during operation of the heat
removal pumps in case of an emergency. Therefore it
was checked for every pump intake of German NPP if a
sufficient submergence is available for every estimated
operating condition. For this purpose the existing sub
mergence was compared with the values obtained by the
ANSI formula (see 1.2). In those cases where the exist
ing submergence was below the calculated one exper
imental proofs were necessary. The experiments in a
1:4 scaled model of a PWR flooding tank (Siempelkamp
(2006)) were performed in this context.
A sketch of the test vessel is shown in figure 2. The
experimental setup consists of the tank model which is
connected to a water supply via a pump. The tank is
2.75 m high and 1.1 m in diameter. It has two horizontal
connections. One connection is located near its top to
fill the vessel with water. The pump suction intake is at
the bottom part and connects the vessel with the pump
via a siphon. The whole model is made from purspex.
Therefore it is possible to observe the water surface and
air entrainment with a video camera.
During the experiments the test vessel was emptied
with different but constant mass flow rates. To consider
flow rates which would be relevant for the original flood
ing tanks they have to be scaled by means of the Froude
5
/
1 suction pipe
2 pump
3 water supply
4 feed pump
5 inlet nozzle
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
number
Fr =
A flow rate of 4.84 kg/s would approximately fit the
maximal possible flow rate in a flooding tank.
In every test run Siempelkamp documented the corre
sponding water levels above the intake for three different
phenomena:
1. onset of dimples,
2. the air cones reach the intake and
3. the air cones reach the pump.
In the following we investigate only the first phenom
ena (onset of dimples) since these values give the neces
sary submergence of the intake to avoid vortices of types
three or higher.
To investigate the influence of an initial rotation of the
water in the test vessel Siempelkamp initiated a recircu
lation in several runs. In these cases water was pumped
into the vessel via the upper connection and out of the
vessel through the lower one with a mass flow rate of
1.03 kg/s. Then the recirculation was stopped and the
water level in the flooding tank model was decreased.
Hence, only at the pump suction intake a higher mass
flow with a prescribed flow rate was retained. Since the
experiments started with an initial water level of 2.68 m
it took about six minutes to lower the water surface to
the level at which dimples were observable. In the test
report it was mentioned that the fraction forces damp the
circular movement of the water at the surface during this
time period. As a consequence the initial rotation did
not influence the measured results.
Therefore the analyses focus on two issues. The first
issue is to determine the intensity of the rotation and its
influence at the beginning and during the experiment.
These analyses were used to explain the observation of
the experimenters that the initial rotation has no signifi
cant influence on the submergence because it is damped
after a short time.
For the investigation of the first issue the whole vessel
was modelled by a structured grid with 1.2 million hex
aeder cells. At the locations of the initial water surface
and the nozzles the mesh is refined. Further refinements
are located at the walls to address the turbulent boundary
layer.
The initial conditions for this first analysis are deter
mined by a steady state single phase calculation. The
recirculation of water was induced by injection of wa
ter through the upper nozzle just like in the experiments.
The water was pumped out again via the lower nozzle
with the same mass flow rate of 1.03 kg/s. The result
ing flow field is shown in figure 3. By using this ini
tial flow field a transient two phase calculation is set up
for which the water surface above the upper nozzle is
inlet nozzle
outlet nozzle
Figure 3: initial rotation in the flooding tank (mass flow:
1.03 kg/s)
refined locations
Figure 4: mesh of the small flooding tank model
given as initial condition. The water injection is stopped
and the mass flow rate at the lower intake is increased to
4.84 kg/s. After approximately two minutes the circu
lar movement of the water in the test vessel was damped
to a very low level which is in good agreement with the
observations during the experiment.
For the determination of the submergence when dim
ples occur it was therefore not necessary to model the
flow field of the entire test vessel. Actually it is suffi
cient to model only the lower part containing the pump
suction intake. Thus, a second model was built which
reaches only to a height of 200 mm above the intake.
The corresponding mesh consists of 0.5 million hex
aeder elements. It was locally refined in the area where
the occurence of dimples could be expected due to the
experiments (see figure 4).
Test runs with two different mass flow rates were con
sidered in the analyses. In the first test case the vessel
was emptied with 4.84 kg/s. For the second analysis the
mass flow rate was raised to 6.25 kg/s. In the CFX cal
culations dimples occur on the surface at water levels of
65.5 mm and 71.5 mm respectively. In fact the inter
phase between air and water is not sharply retained in
the homogenous model and smeared over a few element
Figure 5: air cone at the surface
Table 1:
ment
intake
velocity
[m/s]
0.97
1.25
Necessary submergence to avoid air entrain
mass
flow KSB
[kg/s] [mm]
4.84 148
6.25 179
submergence of the intake
Bernoulli experiment CFX
[mm] [mm] [mm]
48 65.5 67.5+5
79 71.5 76.55
layers. Therefore an uncertainty of the values of about
5 mm must be handled.
The experimental and analytical results are summa
rized in table 1. Taking into account the above men
tioned uncertainty of approximately 5 mm the CFD re
sults agree very well with the measured data. Moreover
in the experiments only non rotating vortices were ob
served which is also confirmed by videos of the experi
ments and CFX. The shapes of the water surface in the
CFD analyses agree well with the experimental obser
vations (see figure 5). Therefore we have compared the
CFD results with the KSB correlation for nonrotational
inflow. The values given by this correlation and the orig
inal Bernoulli equation (KSB minus 100 mm) are also
shown in table 1. The application of the Bernoulli equa
tion yields already a reasonably good approximation of
the necessary submergence in the present cases. Varia
tions from the experimental data are compensated by the
additional offset of 100 mm which is considered by the
KSB formula. For this reason the KSB correlation gives
larger values than the experiments or CFX for both in
take velocities.
2.3 Influence of a forced circulation
Up to this point it was possible to predict the onset of
surface vortices of type two (slight dimples on the sur
r = u ds.
,c
(12)
As a consequence of Stokes integral theorem the circu
lation can be connected to the vorticity of the fluid
r = / V x u dA,
vorticity
if A is the surface bounded by the contour C. In our case
the fluid enters the centre part of the vessel uniformely
through the vanes. Thus, if r, denotes the distance of the
vanes from the centerline of the vessel we can assume
a constant tangential velocity vt along the circle with
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
face) in case of vortex free flow approaching the intake.
Therefore the next step is to take a deeper look to the
influence of vorticity.
Jain et al. investigated the occurence of surface vor
tices and their shapes in an open cylindrical vessel Jain
et al. (1974). In the experiments the authors varied the
viscosity of the fluid, surface tension and in particular
the circulation in the fluid. However, subsequently we
will consider only variations of circulation.
The experiments were performed in two different ves
sels and we will focus on the smaller one which was
0.45 m high and 0.75 m in diameter. Figure 6 shows
the principle setup. The tank was divided into a cen
tre and a peripheral part by two screens. Both screens
are clearly visible in the upper view of figure 6. The
feeding water enters the tank through a rectangular cir
cuit around the perimeter of the tank which was fed by
two pipes. A uniform distribution of feeding water was
ensured by a sufficient number of small holes ontop of
the circuit. The peripheral part between the tank wall
and the outer screen was filled with gravel to destroy the
kinetic energy of the feeding water. After passing the
second screen the fluid flows through an adjustable vane
ring assembly which is placed concentrically around the
intake. It consisted of 64 vanes which could be fixed in
different angles with regard to the radial direction (0,
20, 450 and 600). Therefore it was possible to direct
the fluid entering the test section in the corresponding
direction.
Furthermore it was possible to insert pipe intakes with
different diameters (11.5 mm, 20.75 mm and 29.4 mm)
in the centre of the tank bottom. The intakes were bell
mouthed and the corresponding radius of the curvature
was equal to the radius of the pipes. The vertical intake
lines had a length of eight times their diameter and were
connected to a pump.
The authors vary the circulation of the fluid by chang
ing the angle of the vanes. Given a closed curve C the
circulation along C defined as the curve integral (Chorin
and Marsden (1992))
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
aNo depression on the
surface at large submergence
To o
rmtn f
bFormation of a dimple
1 Broncf fndr pip.
t. RActcngur .t (nnuit
*. GO rdo.il Pwrf
W h St.. DpOck
7. awlt at pip
I lorg vortxnk
10'y^ *;*<;* 
Figure 6: experimental setup (Jain et al. (1974))
radius r, and we obtain the circulation
r = 27rrTvt.
which is directly connected to the volume rate of the liq
uid Q. If we neglect the width of the vanes it holds that
tan(a)i (13)
r = Q) 0 (13)
h
tan(a) d2v.
h 4
In the above equations v characterises the velocity of the
liquid in the intake pipe, d its diameter, a is the angle of
the vanes and h the height of the water level. Note, that
the circulation which is induced by the vanes does not
depend on the radius r,.
The experiments were started with a large submer
gence of the intake. Then the liquid level was lowered
carefully by discharging some liquid from the loop in
order to observe the different types of surface vortices.
Very similar to the common definition of vortex types
the authors classified in Jain et al. (1974) basic steps of
vortex development (shown in figure 7). At the begin
ning there is a flat surface. After the liquid surface falls
below a certain limit a dimple will appear on the sur
face. This situation is followed by the development of
an air cone which grows until it reaches the intake. For
instance vortices of types one and two correspond to the
situations given in subfigures a) and b).
Two different liquids were used in the experiments,
water and a solution of isoamyl alcohol, to vary the sur
face tension under constant viscosity. The surface ten
sion of the isoamyl alcohol solution 0.044 N/m is lower
1 0
cVortex with air core as the dAir core is lengthened as
submergence is further decreased the %ortex becomes stronger
e Airentralning vortex
Figure 7: development of surface vortices (Jain et al.
(1974))
than the corresponding value of water which is given by
0.072 N/m. It turned out that both liquids show a compa
rable behaviour, see figure 8. Hence, surface tension has
no significant influence under these experimental condi
tions as we already mentioned in part 2.1.
Figure 8 shows the necessary submergences for three
angles of the vane assembly (200, 450 and 600). Fur
thermore two important liquid levels are given for se
lected intake velocities. The first one refers to the first
occurence of dimples and at the second level air enters
the intake.
In contrast to the flooding tank experiment the
Bernoulli equation does not lead to reasonable results.
The experiment shows that the necessary submergence
increases with the rotation of the inflow. Hence, the
Bernoulli equation which is derived for nonrotational in
flow should lead to values for the submergence which
are smaller than the measured data given in figure 8. But
that holds only for small intake velocities about 1 m/s.
Hence, the quadratic growth of the equation does not re
flect the real dependence of the submergence on the in
take velocity. Remind that the validity of the Bernoulli
equation requires not only nonrotational inflow but also
a water surface which is approximately at rest. Further
more it is originally derived for inviscous fluids.
For the CFD analysis of this experiment a model of
S.ectr Plto of Tan
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
IntMke velocly, V, f/s
& 6 S 10 1
075
I
os
04;5
4,0 4. 5.0
Figure 10: inflow conditions and opening
Figure 8: necessary submergence of the intake (Jain et
al. (1974))
Figure 9: mesh of the tank
the centre part of the tank was used. Because the rectan
gular inlet conduit and the vanes are not explicitly sim
ulated the initial rotation has to be initiated by boundary
conditions. The intake was modelled exactly like in the
experiments using a pipe diameter of 20.75 mm.
Figure 9 shows the mesh. It is refined at the locations
of the liquid surface, at the walls and the bellmouthed
intake. In vertical direction the finest element layer uses
0.5 mm at the surface. Altogether the structured mesh
is made from approximately one million hexaeder ele
ments (the exact number depends slightly on the water
level h).
The issue of our investigations was to reproduce the
results of the water experiments using a 450 angle of the
vanes. The intake velocity was prescribed as 3.5 m/s
which is implemented by a mass flow condition at the
outlet of the intake line. Furthermore the height of the
water surface h was set to desired values. The incom
ing water from the sides of the tank must have the same
volume rate than the outflow. Otherwise the water level
would change. So it would not be possible to analyse the
flow field by a stationary computation. Therefore the ra
dial velocity of the water at the tank sides is a result of
the intake velocity and the water level, i.e.
II 1 i'I 1I . [m ]
ir. =   .
h
Since the water enters the vessel in an angle of 450 the
tangential velocity component has to be defined by
Vt = t .Ii( !"')V = V7,
to ensure the correct fluid direction at the inlet boundary
which is occupied by water. The side parts above the
water surface are modelled as wall. At the opening on
top of the vessel air can enter or leave the tank at sur
rounding pressure of 1 bar. This setup is given in figure
10.
Three calculations were performed with different wa
ter levels (h 0.114 [m], 0.174 [m] and 0.234 [m]).
Two of them are smaller than the necessary submergence
given in figure 8. Thus, at least dimples should be visi
ble at the water surface in these cases. The third water
level is higher than the submergence determined in the
experiments. Hence, we should observe a flat water sur
face. The red circle in figure 11 marks the limiting water
level derived from the experiment.
In the CFX calculation a clear air cone at the wa
ter surface is detectable for the smallest submergence
h = 0.114 [m] (see figure 11). A water level height of
S0a N o Air Enlrainment SMALL TANK
I AirEntrain ment
32  01 072 N/m
IT. a 044 N//
28 lagOmyL Alcohol sokic io 
C0.044" N/m, lV Itr
___________,
10In t2t 2!35.0 3!
InLokl VIOCilty, V m/l
C'
I
IL 16
_l 2or
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
0
8 o, oo
1 0,001
I6
0,002
0
0S
i; 0,003
0,004
C
),12
h=0.114 m
h=0.174 m
h=0.234 m
0,06 0 0,06
x coordinate [m]
Figure 12: deviation of the surfaces (volume fraction of
water at the surface: 0.9)
I
Figure 11: water surfaces for different submergences
h = 0.174 [m] is still below the limit derived by the ex
periments and in fact a slight dimple is observable. In the
last case (h = 0.234 [m]) it is hard too see any variation
from a flat surface. Therefore we plotted the deviation
of the surfaces in figure 12. In all three cases dimples
are measurable. For smaller water levels the air cones
grow in depth which is in agreement with the experi
ments. Furthermore the vorticity of the surface vortices
also increases which can be seen in figure 13.
Thus, the appearance of air cones on the surface was
successfully predicted for three different water levels.
The agreement with the experimental results is satisfy
ing. However, the results do not completely agree with
the experimental data since we did not obtain fully de
veloped air cones that reach the intake for h = 0.114 [m]
and h 0.174 [m].
It is noteworthy that we also did runs with the SST
turbulence model although it is not really suitable for
swirling flow. In these cases no variations from a flat
surface were observable.
Conclusions
The topic of this work was to predict the onset of air
cones near pump suction intakes for nonrotational and
rotational inflow. In general complex experiments are
used to derive necessary submergences for intakes in nu
clear power plants to prevent the appearance of vortices
of types three or higher. If that is not possible the sub
mergence is estimated by conservative correlations like
the ANSI formula. TUV NORD investigates as a third
0
0,12
h=0.114 m
h=0.174 m
h=0.234 m
0,06 0 0,06 0,12
x coordinate [m]
Figure 13: vertical component of vorticity
approach the capabilities of CFX to determine the de
sired submergences.
For this purpose two experiments were analysed. The
first one uses a 1:4 scaled model of a flooding tank
which was emptied by different but constant mass flow
rates. At some point the water levels reaches a height at
which dimples were observable on the surface. The cor
responding values computed by CFX agreed very well
with the experimental data. An application of the mod
ified Bernoulli formula proposed by KSB led to larger
submergences than CFX or the experiment which re
confirms the conservative character of this correlation.
In some runs an initial rotation was initiated by a cir
culation at the beginning of the experiment. Actually it
turned out that this rotation did not influence the results.
The rotation was damped by friction forces to low inten
sities before any dimples could appear.
Ikk
For a more detailed investigation of the influence of
the rotation on the necessary submergence post test cal
culations of the experiment of Jain et al. were per
formed. By using an assembly of concentric vanes in
a cylindrical vessel which could be adjusted in different
angles it was possible to initiate a circulation of the in
flow. The vertical intake was placed at the bottom of the
vessel. The shape of the water surface was determined
for a fixed intake velocity and three different water levels
by using CFX. In all cases surface vortices developed in
the center of the vessel. The air cones grow by decreas
ing the submergence. So far exact reproduction of the
shapes of the air cones like they were observed in the
experiment was not successful. Nevertheless, if the fo
cus is set on the onset of air cones the agreement with
the experimental data was satisfying.
In summary CFX was successfully used to predict the
principal appearance of surface vortices with air cones.
In case of nonrotational inflow the shape of the water
surface was computed correctly. For rotational inflow
the shape of the air cones obtained by CFX did not com
pletely agree with the measurements. So this topic re
quires more investigations. However, it seems that CFX
offers an alternative possibility to determine a submer
gence of intakes such that surface vortices of types three
or higher can be excluded, if the water level exceeds
the computed value. In other words air entrainment in
pumps can be avoided because this phenomenon does
not occur for vortices of types one or two.
Further studies concerning this topic are planned. The
analysis of the second experiment showed once again
that two equation turbulence models may not be able
to cover all aspects of swirling flows. Furthermore
the grids had large influence on the results and had to
be quite fine. Furthermore the exact determination of
shapes of air cones and the amount of air entrainment
will require the variation of the two phase model (e.g.,
two fluid model, interphase tracking methods).
Acknowledgements
We would like to thank E.ON and in particular Mr. Ko
ring for providing the data of the flooding tank experi
ment.
References
American National Standard for Pump Intake Design
(ANSI/HI9.81998), Hydraulic Institute, 2000, ISBN 1
880952262
A.J. Chorin and J.E. Marsden, A Mathematical Introduc
tion to Fluid Mechanics, Springer, 1992
7th International Conference on Multiphase Flow,
ICMF 2010, Tampa, FL, May 30 June 4, 2010
M. Ishii and T. Hibiki, ThermoFluid Dynamics of Two
Phase Flow, Springer, 2006
A.K. Jain, R. Raju and R.J. Garde, Vortex Formation
at Vertical Pipe Intakes, Journal of the Hydraulics Di
vision, Vol. 104, No. 10, 1978
J. Knauss, Swirling Flow Problems at Intakes, A.A.
Balkema, Rotterdam, 1987
F Menter, TwoEquation Eddy Viscosity Turbulence
Models for Engineering Applications, AIAA Journal,
Vol. 32, 1994
U.S. Nuclear Regulatory Commission (NRC), Regula
tory Guide 1.82, Water Sources for longterm recircula
tion cooling following a lossofcoolant accident, Revi
sion 3, November 2003
P. Pandazis, F Blameling and A. Schaffrath, Potential
von ANSYS CFX 12 zur Bestimmung der erforderlichen
Mindestilberdeckung von Ansaugstutzen von Pumpen
zur Vermeidung von Luftmitriss, Jahrestagung Kern
technik, Berlin, Mai 2010
Reactor Safety Commission (RSK), Statement "Control
of lossofcoolant accidents at PWRs under considera
tion of dead volumes in the containment", 2005
Siempelkamp Priif und GutachterGesellschaft, Unter
suchungen zum Ansaugverhalten der THPumpen aus
den Flutbehiltern, technical report, 2006
C.G. Speziale, S. Sarkar and T.B. Gatski, Modelling the
pressurestrain correlation of turbulence: an invariant
dynamical systems approach, J. Fluid Mech., Vol. 227,
pp. 245272, 1990
W. Wagner, Kreiselpumpen und Kreiselpumpenanlagen,
Vogel Buchverlag, 2004
J. Weinerth, Kennlinienverhalten und Rotorbelastungen
von axialen Kiihlwasserpumpen unter Betriebsbedin
gungen, Dissertation TU Kaiserslautern, 2003
D.C. Wilcox, Turbulence Modeling for CFD, DCW In
dustries, Inc., 1994
