ISIJ International, Vol. 40 (2000), No. 7, pp.

685–691

Model Study on the Entrapment of Mold Powder into Molten

Steel

Manabu IGUCHI, Jin YOSHIDA,1) Tomoyuki SHIMIZU,1) and Yoshiteru MIZUNO1)

Division of Materials Science and Engineering, Graduate School of Engineering, Hokkaido University, Kita-13, Nishi-8, Kita-ku,
Sapporo, 060-8628 Japan. 1) Graduate Student, Graduate School of Engineering, Hokkaido University, Kita-13, Nishi-8,
Kita-ku, Sapporo, 060-8628 Japan.
(Received on January 5, 2000; accepted in final form on February 16, 2000 )

In continuous casting molds of steel, mold powder is placed on the meniscus of molten steel in order to
prevent heat loss from the meniscus and to use it as a lubricant between the solidifying steel and the mold.
The mold powder is sometimes entrapped into the molten steel. Such mold powder entrapment affects the
quality of the steel product significantly. In this study we focused on the shear flow instability between the
molten steel flow and mold powder as one of causes for the mold powder entrapment and investigated the
effect of the kinematic viscosity of the mold powder on the onset of the entrapment. Model experiments
were carried out using salt water and some kinds of silicone oils. The velocity of flow around an interface
between the silicone oil and salt water was measured with a particle imaging velocimetry (PIV) based on
the cross-correlation method. The kinematic viscosity of silicone oil hardly affected the critical flow velocity
for the onset of the mold powder entrapment caused by Kelvin–Helmholtz instability, but it affected the
wave length of the instability.
KEY WORDS: continuous casting; slab caster; mold powder entrapment; Kelvin–Helmholtz instability; parti-
cle imaging velocimetry.

model experiments is relatively easy.2,5,7) In this research we

1. Introduction
also focused on the possibility of the second entrapment
Figure 1 shows a schematic of the current continuous phenomenon. Previous researchers treated a steady flow in
casting mold. The entrapment of mold powder into molten the mold and carried out model experiments using water
steel strongly influences the quality of the steel product. and some kind of silicone oils. The flow field in the mold
Investigations on the mechanism of the entrapment there- can be schematically shown in Fig. 2(a). The silicone oil is
fore have been extensively made by many researchers.1–8) pushed away from the narrow face of the mold by reversing
Also, much effort has been devoted to prevent the mold flow reflected from the narrow face.2) Silicone oil droplets
powder entrapment.9–11) The following three types of en- are pulled from the silicone oil layer into the water layer in
trapment phenomena have been considered to occur in the the presence of high shear stress. Such a situation may not
mold3): be realistic in the actual processes, because when the mold
(1) Karman’s vortex streets formed behind the immersion powder is pushed away from the narrow face, it can not be
nozzle entrap the mold powder placed on the meniscus supplied between the solidifying steel and the oscillating
of the mold.2–4,6,8) mold. Mold powder is known to be supplied continuously
(2) High shear stress between the reversing molten steel between them even if mold powder entrapment takes place.
flow reflected from the narrow face and mold powder
induces a shear flow instability to cause the entrap-
ment of the mold powder.2,5,7,8)
(3) Argon gas injected into the immersion nozzle for pre-
venting the clogging of alumina on to the inner wall of
the immersion nozzle attacks the molten steel/mold
powder interface to cause the mold powder entrap-
ment.
Unfortunately, a clear explanation of a critical occur-
rence condition for each entrapment in actual processes is
not given at the present stage. Among the aforementioned
three kinds of entrapment phenomena the second has been
most actively investigated, probably because carrying out Fig. 1. Schematic of mold powder entrapment.

685 © 2000 ISIJ

 ISIJ International, Vol. 40 (2000), No. 7
Fig. 2. Entrapment of mold powder due to reversing flow.
(a) Steady entrapment, (b) Unsteady entrapment
We therefore consider that the mold powder entrapment
occurs under unsteady conditions, as shown in Fig. 2(b).
According to previous studies on the slab casting mold, the
steel flow in the mold is essentially unsteady due to uneven
flows discharging out of the ports of the immersion noz-
zle.12–14) The mold powder entrapment under unsteady flow
conditions seems to be closely associated with the
Kelvin–Helmholtz instability, abbreviated as KHI.15,16)
Some researchers also suggest that the KHI is one of causes
for the onset of mold powder entrapment and that a critical
velocity difference can be predicted by the original KHI
theory for an infinitely extending interface.8,17) However, it
is not clear whether the original KHI theory is applicable to
a finite interface between mold powder and molten steel.
The original KHI theory15,16) implies that the critical ve-
locity difference giving the onset of mold powder entrap-
ment is influenced by the interfacial tension and the densi-
ties of molten steel and mold powder. In addition to these
parameters, the kinematic viscosity of mold powder must
be taken into consideration. As a first step, particular atten-
tion is paid, in this study, to the effect of the kinematic vis-
cosity of the mold powder on the entrapment, and model
experiments are carried out using salt water and some kinds
of silicone oils.
In order to determine the critical velocity difference for
the initiation of KHI, it is necessary to measure the veloci-
ties of salt water flow and silicone oil flow near the silicone
oil/salt water interface. A particle imaging velocimetry
(PIV) is chosen, as the PIV is the most adequate for this
purpose.
2. Experimental Apparatus and Procedure
The experimental apparatus is schematically shown in
Fig. 3. The vessel made of transparent acrylic resin had an
inner length of 1 000 mm, an inner width of 100 mm, and an
inner height of 70 mm. These dimensions were decided by
referring to the current continuous slab casting molds of
© 2000 ISIJ
Fig. 3. Schematic of apparatus for mold powder entrapment.
Table 1. Physical properties of salt water and silicone oil at
298 K.
steel. That is, this model is an approximately 1/2 model of
the slab casting molds. Only the geometrical similitude for
the inner length and the inner width is taken into considera-
tion in this model because the effects of the physical prop-
erties of upper and lower liquids and the depths of the two
layers on the onset of the KHI can be evaluated from exist-
ing equations. Salt water and some kinds of silicone oils
with different physical properties were used. Both the tem-
peratures of the two liquids in the vessel were 298 K. The
physical properties of the model liquids at this temperature
are listed in Table 1. The actual mold powder layer is not
melted completely, but only the lower part is in liquid
phase, and hence, we assume that the upper part behaves
like a solid wall. The upper end of the silicone oil layer
therefore was covered with a flat solid wall in order to
model the non-melted mold powder. The depth ratio of the
salt water layer to the silicone oil layer was changed from
1 : 1 to 5 : 1.
Solid particles having diameters ranging from 75 m m to
150 m m were dispersed in the salt water layer as tracers for
PIV measurements. They had a density of 1 013 kg/m3, and
accordingly, the density of the salt water was adjusted to
become the same as that of the tracer particles. Also, parti-
cles having the same density as that of silicone oil 10 were
used as tracers for the measurement of the flow velocity of
the silicone oil 10. This 10 means that the kinematic viscos-
ity of silicone oil is 10 mm2/s. Adequate tracer particles are
available for the silicone oil 100, too. For the remaining sil-
icone oils there exists no adequate tracer at present.
A laser sheet having a width of 10 mm was used for the
illumination of the flow field to make the movements of
686
 ISIJ International, Vol. 40 (2000), No. 7

Fig. 4. Symbols used in this study.

tracer particles visible. After confirming no motions of the

particles dispersed in the still salt water layer and silicone
oil layer, the vessel was declined with 10 deg. or 20 deg. to
the horizon at a rotation speed of 4.01 deg./s or 18.7 deg./s
using an air cylinder. A CCD camera mounted on the frame
of the vessel was used for recording the movements of the
particles at 30 frames/s. A velocity difference for the onset
of shear flow instability was measured with PIV. In the pro-
Fig. 5. Models for Kelvin–Helmholtz instability.
cessing of the video images, the cross-correlation method
was used.
cosity of mold powder on the onset of mold powder entrap-
The velocity of silicone oil flow near the interface, V1, is
ment in addition to the two parameters, s 12 and D r .
not always measured as mentioned above, and accordingly,
Existing KHI theories for fluids contained in a vessel of
it was calculated from the equation of continuity expressed
rectangular cross-section will be briefly reviewed (see Fig.
by
5). A critical velocity difference for the onset of KHI in a
V152(H2 /H1)V2 ............................(1) flow field illustrated in Fig. 5(a) was given by Milne–
Thompson,18) being expressed by
where H1 is the depth of the silicone oil layer, H2 is the
depth of the salt water layer, and V2 is the velocity of salt (V1cr2V2cr)25(r 22r 1)g(H1/r 11H2/r 2) ..........(3)
water flow near the interface, as can be seen in Fig. 4.
where H1 and H2 are the silicon oil layer depth and salt
water layer depth, respectively. In deriving Eq. (3), we as-
3. Some Aspects of Kelvin–Helmholtz Instability sumed that the effect of surface tension is negligible and
that KH1, ,1 and KH2, ,1, where K (52p /l ) is the wave
3.1. Critical Velocity Difference for the Onset of
number.
Kelvin–Helmholtz Instability
Kordyban and Ranov considered a flow field shown in
The original theory on the Kelvin–Helmholtz instability Fig. 5(b) and derived the following equation for the critical
is concerned with a shear flow instability occurring at an velocity difference.19)
interface between two horizontal liquid layers extending in-
finitely. A critical velocity difference for the onset of the (V1cr 2V2cr )2
KHI is expressed by15,16) ρ2 g 1
5 ⋅ ⋅
(V1cr2V2cr)454s 12 g(r 22r 1)(r 11r 2)2/(r 12r 22) .....(2) ρ1 K coth( KH1 20.9)10.45 coth 2 ( KH1 20.9)
where r 1 is the density of the upper liquid, r 2 is the density .........................(4)
of the lower liquid, s 12 is the interfacial tension, g is the ac-
where l is the wavelength. In addition, John et al.20) treated
celeration due to gravity, and the subscript cr denotes a crit-
a flow field shown in Fig. 5(c) and proposed
ical value. This equation cannot be applied directly to the
flow filed mentioned here, because the flow field in the ac- ρ1 1 ρ2  g 
tual mold is confined with the vertical walls of the mold (V1cr 2V2cr )2 5
ρ1ρ2  ( ρ2 2 ρ1 ) K 1σ 12 K  ...(5)
 
and non-melted mold powder. However, Eq. (2) affords us a
milestone for understanding of influential parameters for
although this flow field is somewhat different from that
the entrapment of mold powder.
mentioned here, Fig. 2(b).
According to Eq. (2), the interfacial tension, s 12, and
density difference, D r (5r 22r 1) appear to play an impor- 3.2. Wave length and amplitude of instability wave
tant role for the entrapment of mold powder. Although the The following equation,17,20) Eq. (6), is compared with
kinematic viscosity of mold powder is not included in Eq. measured values of the wavelength l .
(2), it is varied over a wide range in the actual processes
depending on the chemical compositions of molten steel l 52p /K ...........................................(6)
and casting speed, and thus mold powder having a higher 21 1/2
K5[(r 22r 1)gs 12 ] ........................(7)
kinematic viscosity is expected to suppress the KHI.
Accordingly, we focused on the effect of the kinematic vis- where K is the wave number. The amplitude of instability

687 © 2000 ISIJ

 ISIJ International, Vol. 40 (2000), No. 7

Fig. 7. Entrapment of silicone oil droplet observed from top of

Fig. 6. Entrapment of silicone oil droplet observed from side of the vessel [silicone oil 2, depth ratio (H2 : H1): 1 : 1, de-
the vessel [silicone oil 2, depth ratio (H2 : H1): 2 : 1, de- cline angle: 20 deg., rotation speed: 18.7 deg./s].
cline angle: 20 deg., rotation speed: 4.01 deg./s].

waves, Am, is given by21)

Am51/K ..................................(8)

4. Experimental Results and Discussion

4.1. Visualized Flow Field and Velocity Vectors
Figures 6 and 7 show visualized images taken by a CCD
camera from the side and top of the experimental apparatus.
The instability did not appear near the end walls of the ves-
sel because of the end effect. In Fig. 6 a silicone oil droplet
can be seen after the rupture of instability waves. It is evi-
dent in Fig. 7 that the instability initiates over the entire,
horizontal cross section. Figure 8 shows velocity vectors of Fig. 8. Velocity vectors measured with PIV [silicone oil 2, depth
salt water flow around the silicone oil/salt water interface ratio (H2 : H1): 1 : 1, decline angle: 20 deg., rotation
just after the onset of KHI. Figure 9 shows some examples speed: 18.7 deg./s].
of the histories of the salt water flow velocity measured
with the PIV. The mean velocity ranges from 0 m/s to 0.32
m/s, and the acceleration ranges from 0 m/s2 to 0.20 m/s2.
We can conclude that PIV is useful for the measurement of
such an unsteady velocity field.
4.2. Critical Velocity Difference for the Onset of KHI
Figure 10 illustrates velocity distributions in the vertical
cross section for silicone oil 10. Both the distributions of
salt water flow and silicone oil flow are almost uniform at
every measurement time. On the other hand, in Fig. 11 the
velocity distributions for silicone oil 100 are not uniform
because of its very high kinematic viscosity.
Figures 12 and 13 show critical velocity differences de-
scribing the onset of KHI for the silicone oils 10 and 100,
respectively. Open circles were determined by using both Fig. 9. Histories of salt water velocity near interface [silicone oil
the measured values of V1cr and V2cr, while open triangles 2, depth ratio (H2 : H1): 1 : 1].

© 2000 ISIJ 688

 ISIJ International, Vol. 40 (2000), No. 7
Fig. 10. Change in velocity distribution [silicone oil 10, depth
ratio (H2 : H1): 1 : 1, decline angle: 10 deg., rotation
speed: 4.01 deg./s].
Fig. 11. Change in velocity distribution [silicone oil 100, depth
ratio (H2 : H1): 1 : 1, decline angle: 10 deg., rotation
speed: 4.01 deg./s].
Fig. 12. Critical velocity difference as a function of salt water
depth (silicone oil 10, decline angle: 20 deg., rotation
speed: 18.7 deg./s).
were determined by using the measured values of V2cr and
calculated value of V1cr based on the equation of continuity,
Eq. (1). For every salt water layer depth the two kinds of
symbols agree with each other in Fig. 12. As the kinematic
viscosity of silicone oil was raised from 10 mm2/s to
100 mm2/s, disagreement between the two symbols ap-
peared, as seen in Fig. 13, but the disagreement was at most
15%. Previous researchers2,5,7) also observed a critical ve-
locity difference of approximately 0.20 m/s for water-sili-
cone oil systems shown in Fig. 2(a).
689
Fig. 13. Critical velocity difference as a function of salt water
depth (silicone oil 100, decline angle: 20 deg., rotation
speed: 18.7 deg./s).
Table 2. Critical velocities of salt water layer, | V2cr | (silicone
oil 2).
Table 3. Critical velocities of salt water layer, | V2cr | (silicone
oil 2) (31022 m/s).
In the following, data on V2cr will be presented to com-
pare them with calculated values, because V1cr was obtained
only for silicone oils 10 and 100.
4.3. Comparison of Measured with Calculated Values
of Critical Salt Water Flow Velocity
For the silicone oil 2 the critical salt water velocity, V2cr,
was hardly dependent on the inclination angle and the rota-
tion speed of the vessel, as shown in Table 2. That is, the
acceleration of the salt water flow did not affect the critical
salt water velocity. Accordingly, the mean value of the mea-
sured critical salt water flow velocities is calculated and
shown in Table 3 and Fig. 14 for different depth ratios. In
calculating Eqs. (3) to (5), the following relationship was
assumed.
V1cr52V2cr ................................(9)
Also, in calculating Eqs. (4) and (5), K was evaluated by
substituting the measured values of l . In Fig. 14 an equa-
tion proposed by Asai22) is presented, too. This equation
was originally derived for the droplet formation illustrated
in Fig. 2(a), being expressed by
(V1cr2V2cr)4548s 12g(r 22r 1)/r 12 ...............(10)
The mean value of the critical salt water velocity is satisfac-
torily approximated by Eq. (3) for the salt water depth ratio
© 2000 ISIJ
 ISIJ International, Vol. 40 (2000), No. 7

Fig. 14. Critical velocities of salt water layer: silicone oil 2.

Fig. 16. Comparison between calculated V2cr from Milne-
Thomson equation and measured | V2cr |.

Fig. 15. Critical velocities of salt water layer: silicone oil 10.
Fig. 17. Wavelength l .
H2 /H up to around 0.75.
In Fig. 15 a good agreement can also be seen between Table 4. Wavelength (31022 m).
the measured values of critical salt water velocity and Eq.
(3) for silicone oil 10. Equations (2) and (4) are not suitable
for correlating the presently observed values. In summary,
Eq. (3) is the best for the prediction of the critical salt water
velocity, although KH1 and KH2 are orders of the magnitude
of unity, and hence, the conditions KH1, ,1 and KH2, ,1
are not satisfied.
Less agreement was observed between the measured
value and Eq. (3) for depth ratios of 0.8 and 0.833 in Fig.
14. In addition, for the silicone oils 50, 100 and 350 the salt Table 5. Amplitude (31022 m).
water attached to the top wall in the growing process of in-
stability for the same depth ratios. That is, when the ampli-
tude of the KHI is larger than H1, the interfacial instability
is disturbed by the top wall.
Figure 16 demonstrates that all the measured values of
the critical salt water velocity can be predicted by Eq. (3)
regardless of the kinematic viscosity of silicone oil.
Yamasaki et al.5) and Komai et al.17) also observed that the
critical velocity difference is a weak function of the kine-
matic viscosity of silicone oil. wavelength l . Yamasaki et al.5) observed that the diameter
of a liquid paraffin droplet became large with an increase in
4.4. Wavelength and Amplitude of KHI the kinematic viscosity of the liquid paraffin. This fact part-
Figure 17 shows that the wavelength decreases with an ly supports the presently obtained finding on the wave-
increase in the salt water layer depth. When the kinematic length.
viscosity of silicone oil is low, say 2 mm2/s, the measured The amplitude of KHI is a function of the depth ratio
values of l can be approximated by Eq. (6). As the kine- H2/H1 and the kinematic viscosity of silicone oil, as shown
matic viscosity of silicone oil increased, the wavelength l in Table 5. Equation (8) could not predict the amplitude, as
increased (see Table 4), and Eq. (6) overestimated the demonstrated in Fig. 18.

© 2000 ISIJ 690

 ISIJ International, Vol. 40 (2000), No. 7

surements.
Nomenclature
Am : amplitude of Kelvin–Helmholtz instability
g : acceleration due to gravity
K : wave number
H5H11H2
H1 : depth of upper liquid layer
H2 : depth of lower liquid layer
V1 : velocity of upper liquid flow
V2 : velocity of lower liquid flow
V1cr : critical velocity of upper liquid flow
V2cr : critical velocity of lower liquid flow
Fig. 18. Amplitude Am. D r : density difference5r 22r 1
l : wavelength
n 1 : kinematic viscosity of upper liquid
4.5. Possibility of Mold Powder Entrapment in Actual n 2 : kinematic viscosity of lower liquid
Continuous Casting Mold Due to KHI s 12 : interfacial tension
We assume that the depth of the melted mold powder r 1 : density of upper liquid
layer is 0.05 m and that the depth of the molten steel layer is r 2 : density of lower liquid
equal to the thickness of the reversing molten steel flow, say
0.10 m. Substituting r 153 000 kg/m3, r 257 000 kg/m3, REFERENCES
H150.05 m, and H250.10 m into Eq. (3), we have
1) Jpn. Soc. Promotion Sci. (JSPS): Recent development in studies of
(V1cr2V2cr) of approximately 1 m/s. If H1 is changed from non-metallic inclusions in steel, JSPS, (1994), 87.
0.05 m to 0.02 m and H2 is changed from 0.1 m to 0.05 m, 2) N. Kasai, M. Kawasaki, K. Hanazaki and T. Sakashita: CAMP-ISIJ,
the critical velocity difference becomes approximately 3 (1990), 1114.
0.6 m/s. Such a steel flow velocity fluctuation is considered 3) H. Tanaka, H. Kuwatori and R Nishihara: Tetsu-to-Hagané, 78
(1992), 761.
to frequently appear in the actual continuous casting
4) Y. Yamaoka, M. Suzuki and K. Murakami: CAMP-ISIJ, 6 (1993),
mold.12–14) Accordingly, we can conclude that the possibili- 279.
ty of mold powder entrapment due to KHI is considerably 5) T. Yamasaki, H. Hasegawa, T. Huji and T. Imoto: CAMP-ISIJ, 10
higher, as long as we judge the onset of KHI based on Eq. (1997), 236.
(3). 6) K. Watanabe, M. Nakata, M. Suzuki and T. Shiomi: CAMP-ISIJ, 11
(1998), 860.
It should, however, be noted that the interfacial tension
7) T. Katai, Y. Otsuka, K. Shigematsu, T. Imoto and H. Iiboshi: CAMP-
between mold powder and molten steel is much higher than ISIJ, 11 (1998), 861.
that between silicone oil and salt water. Further investiga- 8) T. Miyake: Jpn. J. Multiphase Flow, 13, 2 (1999).
tion must be performed for getting clear explanation on the 9) S. Yokoya, R. Westhoff, Y. Asako, S. Hara and J. Szekely: Tetsu-to-
onset of KHI in the actual continuous casting molds. Hagané, 80 (1994), 759.
10) M. D. Santis and A. Ferretti: ISIJ Int., 36 (1996), 673.
11) S. Yokoya, S. Takagi, H. Souma, M. Iguchi, Y. Asako and S. Hara:
5. Conclusions ISIJ Int., 38 (1998), 1086.
12) N. Hatta, T. Fukui, Y. Nakajima and J. Ikeda: Tetsu-to-Hagané, 82
When the thickness of the silicone oil layer was larger (1996), 285.
than the amplitude of waves caused by shear flow instability 13) K. Tozawa, S. Idogawa and K. Sorimachi: CAMP-ISIJ, 8 (1995),
1114.
at the silicone oil/salt water interface, the critical salt water
14) D. Gupta, S. Chakraborty and A. K. Lahiri: ISIJ Int., 37 (1997), 654.
velocity was independent of the kinematic viscosity of sili- 15) L. D. Landau and E. M. Lifshitz: Fluid Mechanics, 2nd ed.,
cone oil, n 1, in a range of n 1 from 2 mm2/s to 350 mm2/s Translated by H. Takeuchi, Tokyo Tosho Co. Ltd., Tokyo, (1970),
and satisfactorily predicted by an analytical solution, Eq. 264.
(3), proposed by Milne-Thompson. 16) Jpn. Soc. Fluid Mech.: Handbook of Fluid Mech, 2nd ed., Maruzen,
Tokyo, (1998), 166.
The wavelength l increased with an increase in the kine-
17) T. Komai, Y. Saeki and N. Tsutsumi: JSPS, 19th Committee, 19-
matic viscosity of silicone oil. A previously derived analyti- 10801, Solidification-426, (1987).
cal equation for l , Eq. (6), agreed with the measured values 18) L. M. Milne-Thompson: Theoretical Hydrodynamics, 5th ed., The
for n 152 mm2/s, but overestimated l as n 1 became larger Macmillan Press Ltd., London, (1968).
than 10 mm2/s. The amplitude of the interfacial instability 19) E. S. Kordyban and T. Ranov: Trans. ASME, J. Basic Engng., 92
(1970), 857.
also was measured, but a clear explanation on it must be
20) J. S. Walker, G. Talmage, S. H. Brown and Neal A. Sondergaard:
left for a future study. Phys. Fluids, A5, 6 (1993), 1446.
21) K. Mishima and M. Ishii: J. Fluid Eng., 102 (1980), 441.
Acknowledgement
22) S. Asai: 100th and 101th Nishiyama Memorial Seminar, ISIJ, Tokyo,
The authors would like to express their thanks to Mr. N. (1984), 65.
Tsuda and T. Koseki (NSC) for suggestions on PIV mea-

691 © 2000 ISIJ