ISIJ International,

ISIJ International, Vol. 57 (2017), No. 3 Vol. 57 (2017), No. 3, pp. 468–477

Experimental Investigations of Rotary Electromagnetic Mould

Stirring in Continuous Casting Using a Cold Liquid Metal Model

Bernd WILLERS,1) Martin BARNA,2) Jürgen REITER3) and Sven ECKERT1)*

1) Helmholtz-Zentrum Dresden-Rossendorf (HZDR), Institute of Fluid Dynamics, Bautzner Landstraße 400, 01328 Dresden,
Germany. 2) Johannes Kepler University (JKU) Linz, Institute of Fluid Mechanics and Heat Transfer (ISW),
Altenbergerstraße 69, 4040 Linz, Austria. 3) Voestalpine Stahl Donawitz GmbH, Kerpelystraße 199, 8700 Leoben,
Austria.
(Received on September 9, 2016; accepted on November 9, 2016; J-STAGE Advance published date:
February 10, 2017)

This paper presents the experimental study of an electromagnetically stirred mould flow using a 1:3
scale acrylic glass model of the round bloom caster from voestalpine Stahl Donawitz GmbH. An electro-
magnetic stirrer was installed at the strand producing a rotating magnetic field (RMF). Flow measurements
were performed in the eutectic alloy GaInSn at room temperature by means of the ultrasound Doppler
velocimetry (UDV). Up to 10 ultrasonic transducers were employed simultaneously in order to obtain a
two-dimensional reconstruction of the flow structure. The experiments contribute to a better understand-
ing of electromagnetically stirred mould flows and provide an extensive and valuable data base for the
validation of numerical methods. The flow measurements reveal a distinct influence of the secondary flow
on the distribution of the angular velocity in various regions of the mould. The submerged jet intensifies
this secondary motion in the upper part of the mould and thus causes a strong deformation of the free
surface of the melt. The jet is deflected, bent and rotates around the strand axis.

KEY WORDS: continuous round bloom casting; eutectic alloy GaInSn; experiments; electromagnetic stir-
ring; rotating magnetic field; ultrasound Doppler flow measurements.

embedding into the solidified shell by flow velocities of

1. Introduction
about 0.3 m/s. Using an experimental model in which mol-
Electromagnetic stirring (EMS) by means of AC mag- ten steel was stirred by an RMF during solidification Miki
netic fields is a valid method for contactless flow control in et al.8) observed an enrichment of agglomerate inclusions at
billet/bloom casters for about 40 years,1–4) where the rotary the axis of rotation in the resulting ingots.
stirring of the mould flow by means of a rotating magnetic Moreover, it is expected that the forcing of melt flow in
field (RMF) can be considered as an established application. close vicinity to the solidification region promotes a transi-
The stirring is supposed to enhance the homogeneity of the tion from columnar to equiaxed solidification. An intense
molten steel and to improve the quality of the solidified flow along the solidification front causes a fragmentation of
steel strand therefore significantly decreasing the number of the dendrites and hence a multiplication of nuclei. Su et al.9)
casting defects. In particular, it is assumed that non-metallic suggest a dendrite fragmentation criterion for low carbon
particles are prevented from becoming entrapped in the and high carbon steels casted under the influence of EMS.
solidified shell by the so-called washing effect.5) Particles Industrial experience has shown that the particular stir-
attached to the growing columnar dendrites are prevented rer design, position and operating conditions have a strong
from engulfment by the mushy zone, washed out by the influence on the metallurgical quality. First of all, a high
forced flow and transported back into the bulk liquid. In stirring intensity should be guaranteed by a suitable coil
addition, the stirring shall stimulate the agglomeration and design and appropriate magnetic field parameters. A stirrer
the formation of larger particles which rise up more easily operation at low magnetic field frequencies (typically 2 to
towards the free surface. Yamada et al.6) analyzed the size 8 Hz) is required because at higher frequencies the thick
of alumina inclusions in solidified steel. They suggested copper walls of the mould impede a deep penetration by the
that the maximum diameter of the clusters remaining in the AC magnetic field. This restriction to low frequencies limits
product corresponds to the EMS- induced velocity of molten the driving electromagnetic Lorentz force. Furthermore, an
steel. On the basis of these data Matsumiya7) concluded that intensive swirling flow at the free surface poses the risk of
inclusions with a size of 100 μm can be prevented from slag or mould powder entrapment due to the depression of
the surface around the nozzle. The absorption of impurities
* Corresponding author: E-mail: [Link]@[Link] is known to lower the quality of steel products significantly.
DOI: [Link] A specific electromagnetic stirring system of two stirrers

© 2017 ISIJ 468

 ISIJ International, Vol. 57 (2017), No. 3

was considered to influence the meniscus stability.10) Two

2. Experimental Setup
independent rotating magnetic fields superimposed upon
each other are applied to achieve a flexible control of the A – The flow model
stirring motion in the meniscus zone regardless of whether The experiments were conducted at the mini-LIMMCAST
an intensive stirring is generated deeper in the mould. In this facility at HZDR. A detailed description of this experimental
way even an opposite stirring direction near meniscus can equipment can be found in previous publications.23,24) About
be chosen with respect to the main stirrer. 12 litres of the ternary eutectic alloy Ga68In20Sn12, which is
A considerable amount of previous research work was liquid at room temperature, were used as a model fluid.25) A
dedicated to the prediction of the mould flow and related schematic view of the experimental facility can be seen in
heat and mass transfer under the action of EMS, mainly Fig. 1. The tundish is a cylindrical vessel made of stainless
by numerical simulations (see for example).11–17) Numeri- steel with a circular outlet into the submerged entry nozzle
cal calculations can provide a better understanding of the (SEN). A stopper rod controls the mass flow rate through
complex flow behaviour, but experimental data are indis- the SEN. The SEN is an ordinary circular pipe made of
pensable with respect to validation of these CFD models. acrylic glass with an inner diameter of 10 mm and a length
Experimental investigations in an electromagnetically stirred of 300 mm. A SEN immersion depth of 35 mm below the
Woods metal model were performed by Partinen et al.12,18) surface level of the liquid metal is chosen for the experi-
The authors measured the deformation of the free surface of ments presented here. The mould is also made of acrylic
the melt and determined the surface velocity using alumina glass and has a circular cross section with an inner diameter
particles on the surface. The motion of these particles was of 80 mm and a length of 800 mm. The melt discharges
recorded by means of a high-speed video camera. Fluid flow from the SEN into the mould as a submerged jet. To close
measurements within the liquid metal were carried out using the experimental test loop, after the mould the liquid metal
an electromagnetic Vives probe. The comparison between flows over a dam into a storage vessel. The vertical position
the measurements and the results obtained by the numerical of the spillway controls the free surface level in the mould.
model shows fairly good agreement, but the measurements The experiments presented here were performed in a con-
of the bulk flow by the Vives probe turned out to be rather tinuous mode. A magnetic pump conveys the melt from the
challenging. The authors reported severe problems arising vessel into the tundish at a maximum rate of about 7.5 l/min.
from the size of the invasive probe and the interference Owing to the eddy currents the operation of the pump causes
of the magnetic field driving the flow and the field of the a temperature rise in the metallic melt of approx. 0.1 K/min.
permanent magnet incorporated in the probe. In particular, The liquid flow rate was controlled by lifting the stopper
it became apparent that the immersion of the probe into the rod into a defined position. The liquid levels in both the
melt caused distinct changes of the surface deformation. tundish and the storage vessel were monitored using a laser
Since this pioneering work performed 20 years ago new distance sensor. The tundish level was controlled and kept
ultrasonic measuring techniques were adapted to be applied constant during the continuous operation by readjusting the
successfully in liquid metal flows.19–21) The used ultrasonic pumping rate. Maximum level fluctuations of ± 10 mm were
Doppler velocimetry (UDV) is a non-intrusive method which observed in a period of about two hours. The flow measure-
instantaneously delivers the liquid velocity profile along the ments were started after achieving fully developed steady
ultrasonic beam. New developments of ultrasonic sensor flow conditions. Flow rate measurements were conducted
arrays allow for a multidimensional flow mapping.22) Such
imaging techniques become more and more important for
detailed explorations of three-dimensional turbulent flows,
in particular with respect to the generation of a suitable
experimental data base for an efficient validation of numeri-
cal simulations.
The motivation of our study is to provide a rather generic
experiment equipped with advanced ultrasonic flow measur-
ing technique for detailed investigations of the liquid metal
flow in a long cylindrical column driven by an RMF. Espe-
cially, the consequences of variations of the field strength
and the frequency on the stirring intensity and the behavior
of the free surface are considered here. Furthermore, we
compare the case of a pure RMF-driven flow with the more
realistic configuration where the swirling flow in the mould
is superimposed by an emergent jet from a vertically aligned
submerged entry nozzle. The flow measurements were con-
ducted in a 1:3 scale acrylic glass model of the round bloom
strand caster from voestalpine Stahl Donawitz GmbH. A
description of the model experiment, the magnetic field
system and the measuring technique will be given in the
subsequent chapter. The measured results are presented and
discussed in the chapters 3 and 4, respectively. Concluding
remarks can be found in chapter 5. Fig. 1. Schematic drawing of the experimental setup.

469 © 2017 ISIJ

 ISIJ International, Vol. 57 (2017), No. 3

at the end of each measuring campaign when the pump was the propagation direction of the ultrasonic beam.19) The
switched off. While emptying the tundish the overall liquid measurements in the present study were performed using the
flow rate can be derived from the descent of the surface level DOP2000 velocimeter (model 2125, Signal Processing SA,
in the tundish. The velocity inside the SEN could be esti- 1073 Savigny, Switzerland). This instrument is equipped
mated in this indirect way to be in the range of 1.2–1.5 m/s. with an internal multiplexer allowing for sequential data
B – The magnetic field system acquisition from up to ten sensors.
The electromagnetic system has a bore diameter of The outer cylindrical wall of the mould was partly
200 mm and a height of 310 mm. An arrangement of 12 coils machined to a flat surface in order to enable an exact posi-
generates a rotating magnetic field (RMF) with a maximum tioning of the US sensors at different distances to the strand
magnetic flux density of B = 20 mT. The coils are installed axis and heights. The transducers were attached at the outer
around the mould with the top edge of the magnetic stirrer wall using a technical vaseline. Acrylic glass and GaInSn
located 60 mm below the free surface of the liquid metal. are a well-suited material combination for non-invasive
Systematic flow measurements were conducted for different ultrasound measurements through the mould wall because
values of the magnetic field strength (peak value) varying these two materials have almost the same sound velocity of
in the range between 4.1 mT and 18.3 mT. Magnetic field
frequencies between 2.5 Hz and 50 Hz were chosen. Every
change of the magnetic field parameters in the course of
a measuring campaign was accompanied by an idle time
of at least 2 min prior to the next measurement. This was
necessary in order to guarantee a complete adaptation of the
flow pattern to the new magnetic field parameters. A Gauss
meter (LakeShore model 460, sensor type MZZ-2512-UH)
was applied to determine the field distribution. The spatial
distribution of the magnetic flux density measured for a
feeding current I = 10 A and a field frequency f = 10 Hz is
presented in Fig. 2(a). Figure 2(b) shows the vertical profile
of the magnetic field measured along the axis for three differ-
ent combinations of field strength and stirring frequency. A
domain of uniform field strength occurs in the inner zone of
the stirrer whereas the magnetic induction starts to decrease
at a vertical distance of about 50 mm from the edges of the
stirrer. The magnetic system generates a remarkable stray
field, which extends to mould regions far below the stirrer.
C – Flow measurement technique
The fluid velocity in the mould was measured by means
of the ultrasound Doppler velocimetry (UDV). This method Fig. 3. The schematic view showing the off-center measuring
positions of the US Sensors with US-beam lines at R = 15
is based on the pulse-echo technique and delivers instan-
mm and R = 30 mm and the horizontal velocity vectors of
taneous profiles of the velocity component projected onto tangential and radial flow component.

Fig. 2. (a) Magnetic flux density inside the RMF; the mesh represents a volume of Ø80 × 410 mm 2; I = 10 A, f = 10 Hz;
B = 9,9 mT in the center of the RMF. (b) Magnetic flux densities at the centerline inside and outside of the stirrer
region.

© 2017 ISIJ 470

 ISIJ International, Vol. 57 (2017), No. 3

approximately cs = 2 750 m/s.26) Hence, velocity measure-

ments through the wall are possible as there is no significant
diffraction of sound at the solid-liquid phase boundary.
A vertical array of ten 4 MHz transducers (TR0405AS,
acoustic active diameter 5 mm) was assembled with a
distance of 10 mm between two adjacent transducers.
Horizontal velocities were measured along chords parallel
to the diameter of the strand situated at specific distances
from the axis of the cylindrical mould, namely at the posi-
tions R = 15 mm and R = 30 mm. The locations of the
measuring lines in the mould cross section are displayed in
Fig. 3. The exact values of the tangential velocity can be
detected on both radial measuring positions at a distance of
a = 43 mm from the sensor where the propagation line of
the ultrasonic beam is parallel to the tangential direction of
the flow velocity.
The height of the model facilitates a vertical measuring
range of 660 mm. The centre of the electromagnetic stirrer
was taken as the vertical zero point of the coordinate system.
The vertical distance of 10 mm between neighbouring sen-
sors leads to 2 × 67 fixed acoustic beam lines for acquiring
linear velocity profiles. The data recording for the multiple
sensor arrangement was conducted by multiplexing using an
overall scan rate between 3 Hz and 10 Hz for all 10 ultra-
sonic sensors. The measuring volume of a particular sensor
comprises a sequence of separate cylindrical disks lined up
concentrically along the propagating ultrasound beam. In the
present study, the size of the individual disks in the axial
direction was about 1 mm. Owing to the divergence of the
ultrasonic beam the lateral size of the disks increases with
increasing distance from the transducer. The divergence
angle of the 4 MHz transducers used in this study is φ =
2.3°.26) Hence the lateral resolution varies from 5 mm at
the sensor to approximately 9 mm at a distance of 100 mm
from the sensor. The velocity resolution is about 0.4% of the
velocity range, i.e. a resolution of 4.4 mm/s can be achieved
for a velocity range of 1.1 m/s.

3. Experimental Results
3.1. Rotating Flow without Submerged Jet
A first campaign of velocity measurements was carried
out for the situation of a solely rotating flow without the
influence of a submerged jet. Fig. 4. Magnetic stirred rotating flow without the influence of a
Vertical profiles of the mean tangential velocity Vt,r30 submerged jet. Vertical profiles of the time-averaged, tan-
gential velocity measured at a radial position of 30 mm (a).
obtained at R = 30 mm for six different magnetic field
Vertical profiles of the time-averaged, tangential velocity
intensities and stirring frequencies are shown in Fig. 4(a). measured at a radial position of 15 mm (b).
The time-averaged values are calculated from 100 to 300
flow profiles recorded over a period of ca. 30 s to 60 s. The field strength and frequency, for instance, the application of
magnetic stirrer was positioned in the domain highlighted a magnetic field of B = 18.3 mT at f = 2.5 Hz results in a
by the two horizontal black lines in the diagram. The flow maximum velocity of Vt,r = 30 = 186 mm/s. The application
is driven by the Lorentz force which is mainly induced of higher frequencies leads to higher fluid velocities, even
inside the stirrer region. Hence, the largest values of the at lower magnetic field intensities. However, because of
mean tangential velocity approximately coincide with the the copper mould’s shielding effect, it is quite inefficient to
geometric centre of the stirrer. Internal friction and second- apply field frequencies of 10 Hz and larger at a real caster.
ary flows are responsible for the rotating motion of the Figure 4(b) shows corresponding measurements of the
fluid above and below the magnetic stirrer. At a vertical mean tangential velocities recorded at the radial position R =
position of z = − 400 mm (245 mm beneath the stirrer’s 15 mm. Although the tangential velocity is smaller at the
lower edge) a reduction of the tangential velocity to approxi- lower radial position, the respective frequency of fluid rota-
mately 50…60% of the maximum value is observed. The tion is almost the same (fF,r = 15 = 1.1 s − 1 and fF,r = 30 = 1 s − 1
intensity of the swirling flow grows with increasing both for a magnetic field frequency f = 2.5 Hz). A remarkable

471 © 2017 ISIJ

 ISIJ International, Vol. 57 (2017), No. 3

difference with respect to the velocity profiles presented in

Fig. 4(a) is the fact that the vertical velocity profiles do not
show a pronounced maximum in the region of the magnetic
stirrer. Instead, the tangential velocity in Fig. 4(b) appears
to be almost uniform along the height of the strand showing
only minor variations. The velocity in the lower part of the
mould is even slightly exceeding the values measured within
the stirrer region of the magnetic field. Although no jet flow
was applied, the SEN was immersed into the melt during all
measurements carried out at f = 2.5 Hz. The presence of the
nozzle leads to a local braking of the flow which becomes
apparent by undershoot in the velocity profiles in the domain
150 mm < z < 200 mm. This drop in the tangential veloc-
ity can also be observed at an outer radial position (see Fig.
4(a)), however, the perturbation is not that pronounced here
as in the region close to the cylinder axis.
As a next step we want to compile a horizontal profile
of the tangential velocity along the ultrasonic beam line. In
general, the measured horizontal velocity has to be decom-
posed into a tangential and a radial component as shown in
Fig. 3 for the measuring line at R = 15 mm. Let us assume
an axisymmetric flow, which basically means that the abso-
lute value of all velocity components remains unchanged
along the circumference at a given radius r. In the case of
a radially inwards flow (converging flow) the radial and
tangential velocity components Vr and Vt, respectively, can
be determined from the values VUS measured by the UDV
system by solving the following equations:

VUS (a1 ) = Vt cosα + Vr sin α (a1 <43 mm,1. measuring range)

VUS (a 2 ) = Vt cosα − Vr sin α (a 2 >43 mm, 2. measuring range)
whereas cosα = R / r

From values of the tangential velocity component Vt, the

angular velocity ω can be easily derived as ω = Vt/r. Figure
5 shows radial profiles of the time-averaged angular veloc-
ity recorded at different heights z for different values of the
magnetic field strength and frequency. Figure 5(a) contains
the case of a magnetic field of B = 18.3 mT and a stirrer
frequency of 2.5 Hz. The diagram reveals flat profiles of the Fig. 5. Radial profiles of the angular velocity recorded along a
angular velocity near the core of the cylinder for positions chord at a distance of R = 15 mm from the axis of the
cylindrical fluid vessel in the first measuring range (a < 43
inside and in the vicinity of the stirrer. This indicates the
mm): (a) B = 18.3 mT, f = 2.5 Hz, (b) B = 9.2 mT, f = 10
occurrence of a solid body rotation in a wide domain of the Hz, (c) B = 4.1 mT, f = 50 Hz.
liquid metal column. In the lower part of the mould (z ≤
− 200 mm) the angular velocity increases with decreasing without the influence of a jet discharging from the SEN.
distance to the cylinder axis. That implies a remarkably The flow measurements in this section are concerned with
higher rotation rate of the fluid in the core domain as in the combination of the RMF-driven flow and the submerged
the outer zone. It is interesting to note that the core rota- jet. The behaviour of the tangential velocity was determined
tion rate below the stirrer is found to be larger than inside along the height of the cylindrical mould. Figure 6 displays
the stirrer where the fluid experiences the electromagnetic the time-averaged velocity as well as the related minimum
driving force directly. Corresponding results are presented and maximum values obtained for a magnetic flux density
in the Fig. 5(b) for a stirrer frequency of 10 Hz at B = 9.2 of 18.3 mT and a frequency of 2.5 Hz at radial positions of
mT and in Fig. 5(c) for a stirrer frequency of 50 Hz at B = 30 mm (Fig. 6(a)) and 15 mm (Fig. 6(b)), respectively. For
4,1 mT, respectively. The shape of the curves appears to be comparison the diagrams also contain the corresponding
similar as found for the low magnetic field frequency (see vertical profiles for the solely RMF-driven flow. The most
Fig. 5(a)), whereas the rotation rate inside the stirrer region striking difference with respect to the solely swirling flow
decreases slightly with increasing distance from the mould is the fact that the fluid rotation rate is rather high near the
axis at the higher magnetic field frequencies. free surface and reaches a distinct maximum just above the
stirrer region. Moreover, it has to be noted that significantly
3.2. Rotating Flow with Submerged Jet higher tangential velocities were found at r = 15 mm in
All results presented so far considered a rotating flow comparison to the measurements conducted at the position

© 2017 ISIJ 472

 ISIJ International, Vol. 57 (2017), No. 3

r = 30 mm. The corresponding curves in Fig. 6(b) reveal a situation of a submerged jet without any stirring. No sig-
significant drop of the tangential velocity in the upper part nificant azimuthal flow can be found for the situation of the
of the stirrer. The lowest values for the axial profile of Vt submerged jet without electromagnetic stirring. The graphs
along the entire mould length can be observed at a vertical obtained for electromagnetic stirring and submerged jet
position of z = 50 mm. The spreading of the curves for the show the maximum tangential velocity in the domain above
minimum and maximum values increases dramatically for the stirrer and the minimum inside the stirrer, whereas the
the rotating flow with a superimposed jet, which indicates velocity magnitude grows gradually with increasing mag-
a much higher level of turbulent fluctuations than for the netic flux density. It is interesting to note that an inversion of
solely RMF-driven flow. the velocity appears for the lowest field strength of 5.8 mT.
Figure 7 shows the vertical profiles of the time-averaged The time average of the velocity reveals the occurrence of a
tangential velocity for the low stirring frequency of 2.5 Hz domain ( − 30 mm < z < 60 mm) with counter-rotating fluid
and different magnetic field intensities as well as for the (Fig. 7(a)). However, the existence of a flow pattern marked
by a really counter-rotating flow in some regions is highly
unlikely. It is rather likely that the interaction between the

Fig. 6. Vertical profiles of the time-averaged, tangential velocity Fig. 7. Vertical profiles of the time-averaged, tangential velocity
and both the corresponding minimum and maximum val- measured at radial positions of 30 mm (a) and 15 mm (b),
ues measured at B = 18.3 mT and f = 2.5 Hz at radial respectively (various values of the field strength, f = 2.5
positions of 30 mm (a) and 15 mm (b), respectively. Hz).

473 © 2017 ISIJ

 ISIJ International, Vol. 57 (2017), No. 3

jet and the rotating flow is responsible for the occurrence plots of the tangential velocity field obtained by a flow map-
of this phenomenon. ping using 10 ultrasonic transducers in a multiplex mode.
In the next step the temporal behaviour of the rotating flow The ultrasonic sensors were installed at R = 15 mm along a
will be considered here. Figure 8 presents two-dimensional vertical line at heights between z = 130 mm and z = 220 mm

Fig. 8. Sequence of snapshots of the tangential flow recorded at R = 15 mm within the domain 130 mm < z < 220 mm.

© 2017 ISIJ 474

 ISIJ International, Vol. 57 (2017), No. 3

(10 mm distance between two neighbouring sensors). At a

depth of 43 mm the ultrasound beam intersects with the
radial position of r = 15 mm. The positions of the SEN and
the free surface are marked in the diagrams too. In general,
the measurements reveal that the assumption of an axisym-
metric flow cannot be maintained for situations with simul-
taneous occurrence of jet and electromagnetic stirring. For
that reason the tangential flows presented here were derived
from the measured values by taking the Doppler angle α as
Vp = VUS/cosα into account. Please note that the horizontal
velocities are plotted here without decomposition in radial
and tangential components. The radial component of the
velocity was neglected here. This procedure might implicate
a significant measuring error, however, the data is not taken
for quantitative analysis, but rather for a qualitative identifi-
cation of the dominating flow structures.
Measurements of the flow pattern were conducted for
the situations of a submerged jet without swirling flow and
a pure RMF-driven flow (not shown here). While no sig-
nificant horizontal flow is observed for a pure jet flow, the
velocity field generated by the RMF is clearly dominated
by the swirling flow indicated by the gradual increase of
the velocity towards larger radii. The measurements show
that large-scale temporal fluctuations of the flow structure
are almost negligible in both cases.
In contrast, significant changes of the flow pattern become
visible when the jet interacts with electromagnetic stirring.
Figure 8 contains a series of images recorded at various
arbitrary moments during the respective experiment. At no
time the typical flow structure of an axisymmetric rotating
flow can be observed here. In general, the RMF gener-
ates a swirling flow in counterclockwise direction which
is reflected by positive values (red colour) in the velocity
plots. A disappearance and even an inversion of the swirling
flow occurs in the nozzle region or along the bottom of the
measuring range. The strong modifications of the velocity
near the SEN can be attributed to the existence of a small
vortex in the vicinity of the nozzle which circulates around
the nozzle according to the direction of fluid rotation in the
strand. Such vortices have been observed at the free surface. Fig. 9. Photographs of the free surface: (a) only jet; (b) only RMF,
An example is shown in the circle of Fig. 9(c). The distinct B = 18.3 mT, f = 2.5 Hz; (c) RMF + jet , B = 18.3 mT, f =
perturbations of the flow pattern found in the lower part 2.5 Hz.
(see Figs. 8(c) and 8(d)) can only be explained by a distinct
deformation, namely a deflecting and bending, of the jet. 15 mm, Fig. 4(b)) reveal an almost uniform distribution of
These temporary occurring flow reversals are responsible the tangential velocity along the height of the strand. It is
for the strong reduction of the time-averaged tangential interesting to note that in the region below the stirrer the
velocities in the upper stirring zone as it becomes evident core rotates faster at higher angular velocities as compared
in the Figs. 6 and 7. to the stirrer zone where the melt is directly affected by the
electromagnetic driving force (see Fig. 4(b)). Moreover,
from a comparison between Fig. 4(a) it becomes obvious
4. Discussion
that at positions far below the stirrer even higher tangential
The electromagnetic stirrer installed at the cylindrical velocities are observed near the axis than at an outer radial
strand generates a Lorentz force inside the liquid metal position (e.g. Vt,r15 = 115 mm/s and Vt,r30 = 87 mm/s at a
driving a rotating motion in the strand which appears to be magnetic field at B = 18.3 mT, f = 2.5 Hz). In their clas-
almost axisymmetric for the case of a solely RMF-driven sical paper Davidson & Hunt27) developed an approximate
flow without jet. Hence, maximum values of the swirling analytical model which reveals that the secondary poloidal
flow are found – as supposed – in the region within the flow plays an important role in the overall dynamics. From
electromagnetic stirrer. On one hand, measurements of this model an effective depth of stirring of z/R = 2.3 was
the time-averaged tangential velocity near the side walls predicted corresponding to a depth of about 90 mm in our
(r = 30 mm, Fig. 4(a)) confirm this expectation. On the case. These analytical results are explained by the conser-
other hand, corresponding data recorded in the core (r = vation of angular momentum. Besides the primary rotating

475 © 2017 ISIJ

 ISIJ International, Vol. 57 (2017), No. 3

motion a secondary flow exists in the meridional plane of of about 22°. The jet position is found to vary in time, as
the strand which is directed from the sidewalls towards the the jet is advected by the rotating flow. Both the jet and the
cylinder axis in the regions outside the stirrer. That means manifestation of the small vortex appear in the images of
that the rotating fluid is transported inwards and increases its the tangential flow pattern with the same periodic time, but
angular velocity along this way. This mechanism is respon- shifted for half a rotation period. This indicates an inherent
sible for increasing tangential velocities in close vicinity to coupling between these two flow structures.
the cylinder axis and for transporting angular momentum Figure 10 compares the angular velocities for the jet
into regions outside the stirrer. and the rotating bulk flow measured at z = 130 mm which
The phenomenon of significant acceleration of the is the vertical position 55 mm below the nozzle port. It is
rotating fluid flow above the stirrer zone becomes more interesting to note that at both radial positions (r = 15 mm
pronounced for the situation of a jet emerging from the and r = 30 mm) the purely electromagnetically driven flow
submerged entry nozzle. In this case a striking feature of the rotates with less speed as compared to the combination of
flow field can be observed, which manifests in the occur- RMF-driven flow and submerged jet. Especially, extraor-
rence of large tangential velocities up to 300 mm/s for posi- dinary high angular frequencies are found at r = 15 mm
tions close to the free surface. These high swirling rates can which even exceed the frequency of the RMF (ω = 2πf =
be seen on the free surface where a strong vortex is formed 15.7 s − 1). The secondary flow is amplified by the jet lead-
around the SEN associated with a significant deflection of ing to an enhanced transfer of angular momentum from
the meniscus. Obviously, the vertical jet flow amplifies the the side walls towards the axis and causing therefore this
upper vortex of the secondary flow significantly because considerable increase of the circulation frequencies in the
rotating fluid is conveyed from the side walls towards the core region of the upper part of the mould. It is interesting
SEN. The radially inwards transport of angular momentum to note that a remarkable swirl enhancement can also be
is the reason for the distinct acceleration of the fluid near the observed at larger radii (r = 30 mm). Obviously, a part of
SEN. The formation of such intensive rotating vortices has the kinetic energy supplied by the jet is transferred to the
already been detected for fluid flows generated by a combi- swirling motion in the strand. A recurring appearance of the
nation of rotating and travelling magnetic field. This swirl jet has been detected by the measurements presented in Fig.
accumulation can even produce tornado-like flow structures 8. The temporary distinct perturbations of the flow pattern,
in a laboratory scale.28,29) namely the reversals of the flow direction (see Fig. 8(c))
The effect of swirl accumulation becomes visible by in the lower part of the measuring domain, are a distinct
strong deformations of the free surface. Fig. 9 contains a footprint of the jet. The jet is deflected radially outwards
comparison of photographs showing the free surface for the and moves with swirling flow. The lower curve in Fig. 10
cases of a submerged jet without RMF (Fig. 9(a)), a rotat- represents the angular frequencies of the jet which turn out
ing flow without jet (Fig. 9(b)) and the superposition of jet to be distinctly lower as those measured for the rotating
and RMF-driven rotating motion inside the mould. The free bulk flow. Likely, this difference can be explained by the
surface remains almost undisturbed as long as the RMF is high inertia of the inflowing jet which is not yet acceler-
idle. The swirling flow shows the typical parabolic shape of ated up to the mean rotation speed of the bulk flow here.
the melt surface, but the strongest depression of the free sur- The jet shows an irregular behaviour at low magnetic field
face is observed directly around the nozzle in Fig. 9(c). The strengths (B < 13 mT). No dominating angular frequency
secondary flow which exists in every rotating flow contained of the jet can be determined in this parameter region. The
in a finite vessel is significantly amplified by the impinging reason for this behaviour cannot be clarified so far. Obvi-
jet. The secondary flow driven by rotation and the jet flow ously, the interaction between the submerged jet and the
have the same direction, but, the intensity of the jet flow is swirling flow appears to be rather complex. For instance,
approximately two orders of magnitude stronger. Therefore, the flow measurements presented in this paper cannot fully
the acceleration of the rotation near the SEN and the deflec- clarify the reason for the flow reversals shown in Fig. 8.
tion of the free surface are much more pronounced with the
jet flow as for the case of the pure RMF-driven flow. Here,
the fluid above the stirrer rotates much faster than inside
the magnetically forced region, the angular velocity even
exceeds the magnetic field frequency of f = 2.5 Hz. Small
vortices may occur at the outer nozzle wall. In Fig. 9(c)
such a vortex is marked by a black circle. It was already
discussed above that this vortex revolving around the SEN
with the RMF-driven flow impacts the velocity measure-
ments as shown in Fig. 8. It appears intermittently as a
green-dominated (Figs. 8(b), 8(f)) spot at heights between
220 mm and 180 mm indicating a locally vanishing rotating
flow between the free surface and a height where the sub-
merged jet enters the mould. Moreover, the flow mapping
of the time-dependent tangential flow (Fig. 8) indicates a
radial deflection of the jet position. At a distance of about
40 mm below the nozzle port the jet is clearly detected at a Fig. 10. Comparison of the rotation rates of the jet and the rotat-
radial position of 15 mm corresponding to a deflection angle ing liquid in the strand.

© 2017 ISIJ 476

 ISIJ International, Vol. 57 (2017), No. 3

This feature can be caused by a bending of the jet against Experimental setups of cold liquid metal models are an
the direction of the rotating bulk flow or it can arise from important tool for the investigation of flow and transport
a strong intrinsic jet rotation. Further investigations will be processes in continuous casting. Moreover, these model
necessary to answer these open questions. experiments provide a valuable data base for benchmarking
or validation of numerical simulations.
5. Conclusions
Acknowledgements
This paper describes laboratory experiments in a 1:3 scale The research program of the Competence Center for
model for continuous casting of round blooms at voestal- Excellent Technologies in “Advanced Metallurgical and
pine Stahl Donawitz GmbH. The model mould has an inner Environmental Process Development” (K1-MET) has been
diameter of 80 mm and a length of 800 mm and was filled financially supported within the Austrian competence cen-
with the eutectic alloy GaInSn, which is liquid at room tem- tre programme COMET (Competence Center for Excellent
perature. A magnetic stirring system was placed just below Technologies) by the Federal Ministry of Economy, Family
the SEN outlet. A rotating magnetic field was applied to and Youth; by the Federal Ministry for Transport, Innova-
drive a swirling flow in the mould. Fluid velocities inside tion and Technology; by the provinces of Upper Austria,
the cylindrical mould were measured non-invasively by Styria and Tyrol, by the Styrian Business Promotion Agency
means of the ultrasound Doppler velocimetry (UDV). The and by the Tiroler Zukunftsstiftung. The authors from
measuring setup enables the detection of the horizontal flow HZDR acknowledge additional financial support from the
being a combination of the tangential and the radial velocity German Helmholtz Association in the framework of the
component. Helmholtz Alliance LIMTECH.
The flow measurements of time-averaged velocity pro-
files can be summarized by the following conclusions: REFERENCES
(1) The electromagnetically driven mould flow is com- 1) O. Ilegbusi and J. Szekely: ISIJ Int., 29 (1989), 1031.
posed of a primary, swirling flow and a meridional second- 2) S. Asai: Proc. 6th Int. Iron and Steel Cong., ISIJ, Tokyo, (1990), 370.
3) K.-H. Spitzer and O. Pesteanu: Proc. 3rd Int. Symp. on Electromag-
ary flow which consists of two toroidal vortices lying on netic Processing of Materials, ISIJ, Tokyo, (2000), 409.
top of each other. This secondary flow is responsible for a 4) S. Kunstreich: Rev. Metall., 4 (2003), 395, 1043.
redistribution of angular momentum within the strand. 5) S. Kittaka, K. Watanabe, T. Kanki and Y. Miura: Nippon Steel Tech.
Rep., 86 (2002), 68.
(2) Measurements of the tangential velocity for the situ- 6) W. Yamada, A. Kiyosa, J. Nakashima, J. Fukuda, K. Okazawa and
ation of a pure RMF-driven flow at r = 30 mm show a maxi- K. Miyazawa: CAMP-ISIJ, 12 (1999), 682.
7) T. Matsumiya: ISIJ Int., 46 (2006), 1800.
mum rotation speed in the stirrer region and a continuous 8) Y. Miki, H. Kitaoka, T. Sakuraya and T. Fujii: ISIJ Int., 32 (1992),
decrease of the tangential velocity with growing distance 142.
9) Z.-J. Su, J. Chen, K. Nakajima and J.-C. He: Steel Res. Int., 80
to the stirrer. On contrary, the tangential velocity does not (2009), 824.
vary significantly over the height at r = 15 mm. In fact, the 10) F.-C. Chang, J. R. Hull and L. Beitelman: Metall. Mater. Trans. B,
tangential speed outside the stirrer is even slightly higher as 35B (2004), 1129.
11) K.-H. Spitzer, M. Dubke and K. Schwerdtfeger: Metall. Trans. B,
in the stirrer region which can be attributed to the angular 17B (1986), 119.
momentum transfer by the secondary flow. 12) J. Partinen, N. Saluja, J. Szekely and J. Kirtley: ISIJ Int., 34 (1994),
707.
(3) Off-centre measurements at R = 15 mm with 13) T. T. Natarajan and N. El-Kaddah: Appl. Math. Modell., 28 (2004),
respect to the cylinder axis allow the reconstruction of the 47.
velocity distribution over the strand cross section for radii 14) K. Okazawa, T. Toh, J. Fukuda, T. Kawase and M. Toki: ISIJ Int.,
41 (2001), 851.
greater than r = 15 mm. 15) O. Pesteanu and K. Schwerdtfeger: ISIJ Int., 43 (2003), 1556.
(4) A discharging jet through the SEN intensifies the 16) M. Javurek, M. Barna, P. Gittler, K. Rockenschaub and M. Lechner:
Steel Res. Int., 79 (2008), 617.
meridional flow near the free surface. The superposition 17) M. Barna: Investigations of the Liquid Steel Flow in a Round Bloom
with the RMF-driven flow leads to a drastic acceleration of Caster with an Electromagnetic Mould-stirrer, JKU-Linz, Linz,
the rotating flow and a vortex formation near the SEN. This (2013).
18) J. Partinen, J. Szekely, C. Vives and L. Holappa: ISIJ Int., 35 (1995),
is accompanied by a strong depression of the meniscus at 292.
the nozzle. In the real casting process, such a flow phenom- 19) Y. Takeda: Nucl. Eng. Des., 126 (1991), 277.
20) D. Brito, H. C. Nataf, P. Cardin, J. Aubert and J. P. Masson: Exp.
enon is known to carry the risk of slag entrainment. Fluids, 31 (2001), 653.
(5) A rather complex interaction between the jet and the 21) S. Eckert, A. Cramer and G. Gerbeth: Magnetohydrodynamics-
Historical Evolution and Trends, eds. by S. Molokov, R. Moreau, H.
rotating flow in the strand can be observed. The jet revolves K. Moffatt, Springer, Berlin, (2007), 275.
at a distinctly lower rotation rate as the bulk flow. Further 22) S. Franke, L. Büttner, J. Czarske, D. Räbiger and S. Eckert: Flow
flow measurements (for instance a simultaneous detection Meas. Instrum., 21 (2010), 402.
23) K. Timmel, S. Eckert, G. Gerbeth, F. Stefani and T. Wondrak: ISIJ
of all three velocity components) are supposed to deliver Int., 50 (2010), 1134.
more information concerning the alignment and deforma- 24) K. Timmel, S. Eckert and G. Gerbeth: Metall. Mater. Trans. B, 42B
(2011), 68.
tion of the jet flow in the deeper zones of the mould and to 25) Y. Plevachuk, V. Sklyarchuk, S. Eckert, G. Gerbeth and R.
contribute to a better understanding of the behaviour of the Novakovic: J. Chem. Eng. Data, 59 (2014), 757.
jet. This is a work in progress. 26) Signal Processing (SA): [Link] (accessed
2016-09-01).
Further measurements are in progress in order to obtain 27) P. A. Davidson and J. C. R. Hunt: J. Fluid Mech., 185 (1987), 67.
more detailed information concerning the alignment and 28) I. Grants, C. Zhang, S. Eckert and G. Gerbeth: J. Fluid Mech., 616
(2008), 135.
behaviour of the jet flow in the deeper zones of the mould 29) T. Vogt, I. Grants, S. Eckert and G. Gerbeth: J. Fluid Mech., 736
in presence of electromagnetic stirring. (2013), 641.

477 © 2017 ISIJ