AISTech 2019 — Proceedings of the Iron & Steel Technology Conference

6–9 May 2019, Pittsburgh, Pa., USA

DOI 10.1000.377.135

Effect of Bubble Behavior for a Stopper Rod With Multiple Side-Channel Injection on Bubble
Size Distributions in Nozzle and Mold During Continuous Casting of Steel
Seong-Mook Cho and Brian G. Thomas

Department of Mechanical Engineering, Colorado School of Mines,

Brown Hall W370-B, 1610 Illinois Street, Golden, Colorado 80401, USA.
Phone: 303-273-3309
E-mail: [email protected] (Seong-Mook Cho) and [email protected] (Brian G. Thomas)

Keywords: continuous casting; argon; bubbles; size distributions; breakup; coalescence; accumulation; water model

INTRODUCTION
Argon gas is often injected to prevent nozzle clogging during continuous casting of steel [1]. In addition, it is well known that
the argon gas bubbles help to remove non-metallic inclusions in the mold by capturing the inclusions onto the gas bubble
outer-surfaces and floating them upward towards the top slag layer [2,3]. However, unoptimized argon gas injection may be
detrimental to final steel product quality. Abnormally high gas flow rate compared to the molten steel flow rate can produce
annular or slug flow inside the Submerged Entry Nozzle (SEN) [4,5]. This results in asymmetric jet flow [4,5], shallower jet
angle [6], and more turbulence [6], causing excessive surface velocity and severe surface level fluctuations in the mold. In
addition, very large bubbles float directly upwards to the surface near the SEN, which may cause slag entrainment into the
molten steel pool [7,8,9]. Finally, bubbles can be entrapped by the solidifying steel shell, especially if those are smaller than
the primary-dendrite arm spacing [10,11] or if deep hooks form in the meniscus [12,13]. This can lead to bubble-related
defects, including blisters [5,14], segregation [15], and slivers from the associated inclusions covering the bubble surface
[14]. To decrease these bubble-related defects, it is important to better understand bubble behavior and size distributions in
the nozzle and mold.
In this work, the behavior and size distributions of bubbles, injected through a stopper rod with multiple downward-inclined
channels in its head, are investigated using 1/3-scale water-model measurements, analytical model calculations, and
computational model predictions. Bubble formation at the gas channel exits, bubble breakup, coalescence, and accumulation
in the nozzle, and bubble size distributions in the nozzle, port, and mold regions are quantified by visualizing the phenomena
using a high-speed video camera, measuring the gas pressure and the bubble size distributions, and analyzing those
phenomena, including calculations of gas pressure for bubble formation, initial bubble size, bubble terminal descending
velocity, residence time, and changes in bubble size distribution due to bubble accumulation. This paper builds on previous
work [16] as velocity and turbulence of the fluid flow in the nozzle are calculated with Computational Fluid Dynamics
(CFD). Finally, the initial bubble size model validated with the water model measurements, is extrapolated to estimate argon
bubble size and bubble frequency in the molten steel caster.

METHODOLOGY

Water Model Experiments

Water-air system experiments were conducted using a 1/3 scale water model to investigate argon bubble behavior and size
distributions in the nozzle and mold during continuous steel-slab casting. The model consists of the tundish, stopper rod,
SEN, and mold as shown in Figure 1. Vertical movement of the stopper rod controls the water flow rate from the tundish
through the SEN into the mold, by changing the size of the annular gap between the stopper-rod head and the bottom of the
tundish where it curves into the SEN. Water exits holes in the bottom of the mold into a holding water bath and is pumped
continuously back up to the tundish. Dimensions and process conditions for both the real caster and the water model are
given in Table I. The casting speed was chosen based on maintaining a constant Froude number. Air is injected through
multiple gas channels on the side near the bottom of the stopper rod, as shown in Figure 1(c). Geometric details of the stopper

© 2019 by the Association for Iron & Steel Technology. 1317

rod and the gas channels are given in Table II. Air flow rate in the water model was chosen to maintain a constant gas volume
fraction as follows:

qW q R( 1873 K) λq R( 273 K)
  (1)
QW  qW Q R  q R( 1873 K) Q R  λq R( 273 K)

where QW is water volume flow rate, QR is molten-steel volume flow rate, qW is air volume flow rate at 298 K, q R ( 1873 K) is
argon volume flow rate at 1873 K, q R ( 273 K) is argon gas volume flow rate at standard temperature and pressure conditions (1
atm, 273K), and  is gas expansion factor. Because argon gas injected into the molten steel pool is greatly affected by the
high temperature of the molten steel and pressure at the branch channel exit,  is calculated as follows:

q R( 1873K) ρ R( 273 K) P   1873K   P273 K   1873K 

λ    273 K         (2)
  273K   Ps,tundish_level  ρ s g H tundish  H bc    273K 
q R( 273 K) ρ R( 1873K) 
 P1873K

where, ρ R( 273 K) is argon gas density at standard conditions (1atm, 273K), ρ R( 1873 K) is argon gas density at 1873K, P273K is
1 atm, and P1873 K is molten steel pressure at the branch gas channel exits. Ps, tundish_level is pressure at the top surface of the
tundish (1atm), ρs is molten steel density, and Htundish – Hbc is hydrostatic pressure head distance from the tundish top surface
to the branch-channel exit near the tundish bottom. In this work,  is 4.1, and fluid properties for both molten steel-argon gas
and water-air systems are given in Table III.
Once the water-air flow reached steady state in the stopper rod, nozzle and mold, bubble behaviors in the nozzle (Regions 1-
7) and mold (Region 8) were captured using a high-speed video camera at 1200 frames/s. The recorded image snapshots were
analyzed using the image-analysis program, Image J [17], to quantify the size distributions of the bubbles in each analysis
window (ie. region). In addition, during gas injection, instantaneous gas pressure at the main channel inlet was measured
every 0.01 s for 10 s using a pressure gauge with 0.05 kPa pressure resolution and a data logger as shown in Figure 1(b).

Figure 1. One-third-scale water model showing (a) photos and (b) schematic including region numbers and (c)
details of the stopper rod with multi-channels for gas injection.

1318 © 2019 by the Association for Iron & Steel Technology.

 Table I. Caster dimensions and process conditions
Real caster (R) 1/3 scale water model (W)
Dimensions
Nozzle bore diameter: inner / outer 75 / 138 mm 25 / 46 mm
Nozzle bottom well depth 18 mm 6 mm
Nozzle port size: width / height 69.9 / 80.1 mm 23.3 / 26.7 mm
Nozzle port angle -35° (down) angle
Tundish level, Htundish 1020 mm 340 mm
Distance from nozzle inlet to nozzle port 1449 mm 483 mm
Mold size: thickness / width 225 / 1500 mm 75 / 500 mm
Process conditions
Steel flow rate QR: 545.6 LPM QW: 35 LPM
Casting speed UCasting,R: 1.61 m/min UCasting,W: 0.93 m/min
Gas flow rate* Cold (at 273 K): qR (273K): 0.8, 1.5, 2.3, 3.1, At 298 K: qW: 0.2, 0.4, 0.6, 0.8,
3.8, 4.6, 5.4, 6.1 SLPM 1.0, 1.2, 1.4, 1.6 LPM
Hot (at 1873 K): qR (1873K): 3.1, 6.2, 9.4,12.5,
15.6, 18.7, 21.8, 24.9 LPM
Gas volume fraction (hot)* 0.6, 1.1, 1.7, 2.2, 2.8, 3.3, 3.8, 4.4 %
Submerged depth of nozzle 165 mm 55 mm
* The 3 cases in bold were investigated in detail in the nozzle and mold.

Table II. Dimensions of the stopper rod with the gas channels
Dimension Value
Diameter of stopper rod: Dst 42 mm
Diameter of main gas channel: Dmc 10 mm
Diameter of branch gas channel: Dbc 1 mm
Length of main gas channel: Lmc 517.3 mm
Length of branch gas channel: Lbc 15.7 mm
Gap size between stopper and tundish well: Lg 2 mm
Vertical-angle of branch channel: θ v 15o (downward angle)
Horizontal-angle of branch channel: θh 60o
Height from tundish bottom to branch channel exit : Hbc 10 mm (for water flow rate 35 LPM)

Table III. Comparison of physical properties between molten steel-hot argon and water-air systems
Molten steel-hot argon system Water-air system
Liquid density: ρl 7000 kg/m 3
998.2 kg/m3
Cold (at 273 K): 1.623 kg/m3
Gas density: ρ g At 298 K: 1.225 kg/m3
Hot (at 1873 K): 0.399 kg/m3
Liquid dynamic viscosity: μl 0.007 kg/mꞏs 0.001 kg/mꞏs
Gas dynamic viscosity: μ g -5
2.1 × 10 kg/mꞏs 1.8 × 10-5 kg/mꞏs
Interfacial tension coefficient: σ 1.192 N/m 0.073 N/m

Analytical Modeling

Gas pressure for bubble initiation

Video observations show how the initial gas bubbles form at the branch channel exits in the stopper rod. Bubbles form at the
branch channel exits in four stages of initiation, expansion, elongation, and detachment stages, as shown in Figure 2. To
initiate the bubble formation (Figure 2(a)), gas pressure at the branch channel exit around the liquid pool in the tundish
bottom, Pbc-out must overcome the liquid hydrostatic pressure, Phyd and a threshold pressure based on the maximum surface
tension force, σκ max , as given in Eq. (3) following the Young-Laplace equation. Alternatively, Pbc-out can be given by
Bernoulli’s equation considering the pressure drop from the main channel inlet to the branch channel exit, as follows:

© 2019 by the Association for Iron & Steel Technology. 1319

 Pbc-out  Phyd  σκ max  Pin 
1
2
 2 2

ρ g u mc  u bc  ρ g g H mc  H bc   ΔPchannels (3)

where Pin is inlet gas pressure at the top of the main channel, ρ g is gas density, umc is gas velocity in the main channel, ubc is
gas velocity in the branch channel, g is acceleration of gravity, Hmc is height from the tundish bottom to the main channel
inlet, Hbc is height from the tundish bottom to the branch channel exit, and ΔPchannels is total pressure drop including three
pressure drops: across the main channel ( ΔPmc ), across a branch channel ( ΔPbc ) and due to sudden contraction of the cross-
sectional area from the main channel to the branch channel ( ΔPcont ).

Rearranging the above equations gives the minimum pressure at the main channel inlet, Pin, that is needed to exceed the
threshold pressure at the exit of at least one branch channel and initiate gas bubble formation:

Pin  Phyd  σκ max 

1
2
 2 2

ρ g u mc  ubc  ρ g g H mc  H bc   ΔPchannels (4)

Phyd is calculated as follows:

Phyd  ρl g H tundish  H bc  (5)

where ρl is liquid density and H tundish is height of the top-surface water level above the tundish bottom.

Surface tension force is maximum with the largest bubble-surface curvature when the hemisphere bubble diameter equals the
branch channel diameter.

2σ (6)
σκ max 
rbc

where σ is surface tension coefficient and rbc is radius of the branch channel exit. ΔPchannels is calculated as follows:

128 μ g  Lmc q g,mc Lbc q g,bc


 1    D 2
  ρ g  0.421  bc

 u bc 2 (7)
ΔPchannels  ΔPmc  ΔPbc  ΔPcont  
π  Dmc 4 Dbc 4  2 
 
  D 2
 mc



where μ g is gas dynamic viscosity, Dmc is main channel diameter, D bc is branch channel diameter, Lmc is main channel
length, Lbc is branch channel length, q g,mc is total gas volume flow rate in the main channel, and q g,bc is gas volume flow
rate in each branch channel (qg,mc/6).

Initial bubble size

After initiating bubble formation, the gas bubble starts to protrude from the channel exit, Figure 2(a), and then expands and
elongates as shown in Figure 2(b) and (c). Then, the bubble is detached from the stopper-rod wall due to the shearing drag
forces of the fast liquid flow through the gap between the nozzle inlet and the stopper-rod head region as shown in Figure
2(d). To calculate this initial bubble size, a two-step semi-analytical model of bubble formation in downstream flow by Bai
and Thomas [18] was applied, using the process parameters for the current stopper-rod gas-injection system.
During the expansion stage (model step 1), Figure 2 (b), the bubble sits on the tip of the gas channel exit and is assumed to
attain a spherical shape, instantaneously obeying the following force balance including drag force Fd (left term), buoyancy /
gravity force Fb (right first term), and surface tension force Fsur (right second term),

1
2
 
Cd ρl ul π rex   π rex  ρl  ρg gcosθv  πrexσfθ ul
2 2 4
3
3 1
2
  (8)

1320 © 2019 by the Association for Iron & Steel Technology.

where C d is drag coefficient according to Reynolds number [19], u l is steady average liquid velocity across expanding
bubble, rex is expanded radius of the gas bubble to be solved, θ v is vertical angle of the branch channel, and f θ is a contact
angle function that depends on the static contact angle and difference between contact angles (advancing contact angle, θa
and receding contact angle, θr) above and below the bubbles [18].
In the elongation stage (model step 2), Figure 2(c), the bubble elongates along the stopper-rod surface when the drag force
starts to exceed the surface tension and buoyancy force. Thus, the bubble has an ellipsoid shape at the stopper-rod head
surface, due to this shearing effect of the flowing liquid. The equation [18] to predict final bubble size, d g  2rel,s e just  
before it detaches from the stopper-rod head is given as follows:

rel,s
πU tun  ar 22 / 7 
5.2692   r ar  b
15 / 7 3/ 2
 ar  b1/ 2 dr  2rel,s e3 / 2  Dbc  rex (9)
qg,bc Dtun 
1/ 7
rex
2  2

where qg,bc is average gas volume flow rate into each active branch channel (qg,mc/nact), nact is number of activated branch
channels, Utun is average liquid velocity near the branch channel exit, Dtun is hydraulic diameter in the tundish bottom region,
a and b are constants related with rex, the expansion radius, rel,s , the horizontal radius of the elongated bubble to be solved,
and e, the elongation factor of the bubble [18].

Figure 2. Bubble formation mechanism (left: photo and right: schematic) showing: (a) initiation, (b) expansion,
(c) elongation, and (d) detachment.

Bubble terminal descending-velocity

To calculate the residence times of bubbles in the nozzle, the bubble terminal descending-velocity, u g, terminal, i is calculated
from a force balance between the drag force (left term) and the buoyancy / gravity force (right term) on each gas bubble size
as follows:

ρl ul  u g,terminal,i  Cd Ai  d g,i  ρl  ρg g

1 2 1 3
(10)
2 6

where dg,i is gas bubble diameter i, ul is liquid velocity in the nozzle, Ai is cross sectional area, which is calculated as
  2
π d g,i / 4 , and Cd is drag coefficient which varies with relative Reynolds number and Weber number for the tap water-air
system [20].

© 2019 by the Association for Iron & Steel Technology. 1321

Rearranging Eq. (10) gives u g, terminal, i as follows:

Ql 4 (ρl  ρg ) d g,i (11)

u g,terminal,i   g
Anozzle 3 ρl Cd

where Ql is liquid flow rate in the nozzle and Anozzle is cross-sectional area of the nozzle.

Bubble size distribution based on accumulation

Changes in the bubble size distribution due to gas accumulation in the nozzle were simply estimated based on terminal
descending velocity of each bubble size. Inside the nozzle between Regions 3 and 6, the total number of bubbles of each
diameter, n3-6,i was predicted based on the bubble size distribution measured in Region 3, n3,i as follows:

n36 , i  n3 , i f t residence, i (12)

where i is bubble diameter, f is bubble frequency, tresidence,i is residence time of a bubble between Regions 3 and 6, calculated
as follows:

L36
t residence,i  (13)
u g,terminal,i

where L3-6 is vertical distance from Region 3 to 6. From the total number of bubbles of each diameter calculated to reside
within Regions 3-6, the number of bubbles in Region 6, n6,i was estimated as follows:

L6
n6 , i  n36 , i (14)
L36

where L6 is the vertical height of the analysis window in Region 6. Finally, the population proportion of each bubble diameter
in Region 6, P6,i is calculated as follows:

n3 , i
n6 , i u g, terminal, i (15)
P6 , i   100   100
n n
n3 , i
n
i 1
6,i 
i 1 u g, terminal, i

Computational Modeling
A three-dimensional finite-volume CFD model was applied to quantify the fluid flow velocity and turbulence in the stopper-
rod nozzle of the 1/3 scale water model. The steady-state Reynolds-Averaged Navier-Stokes (RANS) equations with the
standard k-ɛ model have been solved with a commercial CFD program, ANSYS FLUENT [21] to quantify the time-averaged
single-phase (water) turbulent flow in the nozzle. The model used a half domain (adopting 1-fold symmetry) and included the
stopper-rod head and nozzle. The domain consists of 0.24-million hexahedral cells. Constant velocity (1.06 m/s) was fixed as
the inlet condition at the nozzle inlet which is between the nozzle inner wall and stopper-rod head outer-surface, along with
10-5 m2/s2 for Turbulent Kinetic Energy (TKE) and 10-5 m2/s3 for TKE dissipation rate. The velocity was calculated according
to the flow rate in the water model and the surface area of the inlet. At the nozzle port, pressure-outlet boundary conditions of
~650 Pa of gauge pressure considering the hydrostatic pressure due to the head of water in the mold, 10-5 m2/s2 for backflow
TKE and 10-5 m2/s3 for backflow TKE dissipation rate were applied. The inner walls of the nozzle were given by a stationary
wall with the no-slip boundary condition. Convergence of solving the equations was defined when all scaled residuals were
stably reduced below 10-4. 

1322 © 2019 by the Association for Iron & Steel Technology.

 RESULTS AND DISCUSSIONS

Bubble Formation Mechanism and Initial Size

The number of activated branch channels for bubble formation was measured from the video recordings for various gas flow
rates. At lower gas flow rates, all of the branch channels are not activated simultaneously. The number of activated channels
generally increases with higher gas flow rate. All six gas-channels are activated when the gas flow rate exceeds a critical
value, which is between 1.4 and 1.6 LPM for the conditions in this work. It is likely that slightly non-uniform pressure drops
across the branch channels or slightly asymmetric flow near the channel exit in the tundish causes this phenomenon. When
the gas flow rate barely exceeds the minimum threshold, only the single branch channel with the largest diameter is able to
generate bubbles, which inhibits the system pressure from building up. A significant increase in flow rate is required to
increase the system pressure enough to overcome the pressure threshold in the next largest branch channel, and so on.
Gas pressure at the main gas channel inlet was measured for 10 s of steady bubble formation for various gas flow rates [16].
The measured pressure at the main channel inlet are compared with the predicted pressure threshold for bubble formation
(calculated using Eqns.4-7) at the same location for several gas flow rates, as shown in Figure 3(a). Both the measurements
and the predictions show that higher gas flow rate increases gas pressure at the channel inlet. The predicted pressure
threshold increases with flow rate due to the larger pressure drop. The measured average pressure is slightly greater than the
predicted pressure threshold as expected, in order to enable flow. The difference generally increases with increasing flow
rate, except for jumps when flow through a new branch channel is activated.
After passing through the channels, gas can form bubbles only if the gas pressure can overcome the hydrostatic pressure of
the liquid and surface tension force at the branch channel exit. The total bubble frequency increases with increasing gas flow
rate, as shown in Figure 3(b). However, as more channels become activated, the frequency per branch channel is always ~100
/s, which means that the formation of one bubble takes ~0.01 s. With increasing gas flow rate, this frequency decreases
slightly, which is another cause for the bubble size to increase slightly. Based on the measured total bubble frequency, f, the
initial volume-averaged bubble diameter at the branch channel exit, d avg,g can be calculated as follows:

1
 6 q g,mc 3 (16)
d avg, g   
 πf 

These average initial bubble sizes measured in the water model at several gas flow rates are compared in Figure 3(c) with the
analytical model predictions of d g using Eqs. (8) and (9) and with several empirical models suggested by other researchers
[22-24]. As shown in Figure 3(c), the empirical models for stagnant liquid flow systems overpredict the initial bubble
diameter at the branch channel exit in the stopper rod. This is expected because in the current work, the shearing effect from
the high velocity liquid flowing into the gap between the tundish bottom and the stopper rod shortens the time of bubble
formation and results in smaller bubbles with higher bubble frequency. In contrast, the two-step analytical model [18]
predictions show much better agreement with the measurements. This is because the analytical model was formulated for gas
injection into downward flowing liquid, such as encountered in the current flow system. Thus, the analytical model is
validated, and can be applied to molten steel-argon gas system in the real plant.

Figure 3. (a) Predicted gas pressure threshold compared with measured gas pressure at the main channel inlet, (b) bubble
frequency and (c) initial bubble size at the branch channel exit in the water-air model.

© 2019 by the Association for Iron & Steel Technology. 1323

Bubble Breakup, Coalescence, and Accumulation in Nozzle
Bubbles formed at the gas branch-channel exit enter into the highly turbulent flow field in the ~ 2 mm gap between the
stopper-rod head and the tundish bottom. As shown in Figure 4(a), bubble size drastically decreases, from an average initial
size of ~4.5 mm to <1 mm diameter after passing through the gap into Region 2. The narrow gap forces the bubbles to
elongate and the accompanying high velocity gradients produce high shear forces which tear the bubbles apart. In addition,
the high turbulent kinetic energy of eddies smaller than the bubble size contributes further to the breakup. Specifically,
velocity in the gap exceeds 3 m/s and TKE dissipation rate is ~30 m2/s3. This high turbulence allows a stable maximum
bubble size of only ~0.7 mm diameter according to the breakup criterion [25]. This is why the bubbles in the SEN are much
smaller than the initial bubbles formed at the branch channel exits.
In the nozzle, multiple bubbles are observed to collide and coalesce in complex manners during their flow downward. Some
examples of this bubble coalescence behavior are shown in Figure 4(b) for a ~3.4 ms time period. The solid red and dashed
blue circles identify two kinds of coalescence phenomena, which are commonly observed in the upper SEN for bubbles
smaller than 1.5 mm diameter. The solid red circles show three bubbles touching, causing two of the bubbles to coalesce
while the third bubble bounces away. On the other hand, the dashed blue circles show three different bubbles all coalescing
into one big bubble. This shows that bubble coalescence is an important mechanism to increase the average bubble size in the
nozzle. The big bubbles have lower descending velocity due to higher buoyancy during their downward flow in the nozzle.
This also contributes to the bubble size distribution evolving towards larger sizes. After reaching the nozzle well bottom and
the ports, the bubbles face high turbulence produced by the jet swirl which shows alternating flow directions [26,27]
(clockwise and counter-clockwise directions). Bubbles here experience breakup, as observed in Figure 4(c) near the nozzle
bottom. On the other hand, in the upper region of the oversized nozzle ports, where flow is stagnant due to the backflow,
bubbles accumulate and coalesce, as observed in Figure 4(c) near the port top.

Figure 4. Bubble evolution down the nozzle showing (a) bubble formation (upper), velocity and turbulence near the
stopper-rod tip, and resulting breakup into small bubbles (lower), (b) bubble coalescence between Region 3 and Region 4,
150-160 mm below the stopper-rod head tip (1.7 ms between snapshots), and (c) bubble breakup and coalescence in
Region 7.

Bubble Size Distributions in Nozzle and Mold

Bubble behavior including formation, breakup, coalescence, and accumulation greatly affects bubble size distributions in the
nozzle and mold. Bubbles moving down through four regions in the nozzle (Regions 3-6) were recorded using a high speed
video. Example snapshots are shown in Figure 5. To quantify the bubble size distributions in the nozzle and mold, Sauter
mean diameter d32 is calculated in each analysis window as follows:

 d 
3
d 
avg,V
3 i
(17)
d 32   i 1

d   d 
avg,A
2 n

i
2

i 1

1324 © 2019 by the Association for Iron & Steel Technology.

where di is measured bubble diameter, and n is the total number of bubbles in the analysis window. This equation uses the
n n
π
6 d i 3  π d 
2
i
i 1 6
3
volume-average diameter d , and area-average diameter d avg,A  i 1 . In addition, the standard
avg,V 
nπ nπ
deviation of the Sauter diameter, σb is calculated as follows:

 d  d 32 
2
i
σb  i 1 (18)
n

Bubble size distributions in the analysis window of each nozzle region, with air gas flow rates of 0.8 LPM, are given in
Figure 5. Going down through the nozzle regions, both average size and standard deviations increase, which agrees
qualitatively with previous work in a nozzle below a slide gate [28]. This is due to both bubble coalescence and the
accumulation of large bubbles in the descending nozzle flow. Between Regions 3 and 4, the number of bubbles smaller than
1.5 mm decreased greatly (by ~53%), likely due to the coalescence mechanism shown in Figure 4(b). Correspondingly,
bubbles larger than 1.5 mm diameter increased in number by 3  between Regions 3 and 4. In Region 6, bubbles larger than 2
mm diameter are slightly more frequent than those in Regions 4 and 5. This suggests that small bubble coalescence occurs
mainly between Regions 3 and 4, while larger bubbles gradually accumulate with distance down the nozzle.

Figure 5. Bubble size distributions in the nozzle with 35.0 LPM (water) and 0.8 LPM (air):
(a) Region 3, (b) Region 4, (c) Region 5, and (d) Region 6.

Figure 6(a) shows how terminal descending velocity of bubbles decreases with increasing bubble diameter up to 2.3 mm, as
calculated using Eq. (11). Bubbles larger than 2.3 mm diameter have ~0.96 m/s terminal descending velocity, ~20 % lower
than the mean liquid flow velocity (~1.2 m/s). Thus, bubbles larger than 2.3 mm diameter have ~ 20% longer residence time
in the nozzle than 0.5 mm diameter bubbles. Thus, both the bubble coalescence phenomenon, and the accumulation of larger
bubbles due to their longer residence time in the nozzle, are expected to cause the bubble size distribution inside the nozzle to
increase with distance down the SEN.

© 2019 by the Association for Iron & Steel Technology. 1325

To compare the relative importance of coalescence and accumulation, the change in bubble size distribution between regions
3 and 6 was predicted using Eq. (15), considering only accumulation, and compared with the measurements in Figure 6(b).
The predicted proportion of bubbles smaller than 0.5 mm diameter decreases in Region 6, owing to their short residence time.
This same trend is observed in the measurements. Correspondingly, bubbles larger than 1 mm accumulate due to their longer
residence times, so their proportions increase for both the prediction and measurements. The quantitative discrepancy
between the proportion predictions and the measurements is likely due to bubble coalescence inside the nozzle, which
appears to be more important than accumulation phenomena.

Figure 6. (a) Bubble terminal descending velocity and residence time in the nozzle, and (b) predicted bubble accumulation
from Region 3 to Region 6 inside the nozzle and measurements, with 35.0 LPM (water) and 0.8 LPM (air).

In the mold, smaller bubbles typically have longer residence times and accumulate in the mold more than large bubbles,
which exit quickly to the top surface [29,30]. Measurements of bubble size distributions were taken in the mold (Region 8),
in the yellow rectangular analysis window shown in Figure 7. Larger average and broader size-range (higher standard
deviation of bubble size) of bubble size distribution are produced in the mold, as shown in Figure 7. In addition, both average
diameter and standard deviation of the bubble size in the mold increase with higher gas flow rate. As the breakup and
accumulation mechanisms both tend to decrease the average bubble size in the mold, bubble coalescence in the upper port
(Region 7) is clearly the most important mechanism to explain these measured findings.

Figure 7. Bubble size distributions in the mold (Region 8) with 35.0 LPM (water) and (a) 0.2 LPM (air), (b) 0.8 LPM (air),
and (c) 1.6 LPM (air).

1326 © 2019 by the Association for Iron & Steel Technology.

Argon Bubble Size in Molten Steel for Real Caster
The initial bubble size model, validated with the water model measurements in Figure 3(c), was extended to predict the initial
size and frequency of argon bubbles in the real steel caster. The steel caster conditions, corresponding to the water model
based on Froude similarity and same gas volume fraction, are included in Table I. Figure 8 shows that the predicted initial
argon bubble size at the branch channel exit of the real stopper rod, d g,R is 10.5-13.5 mm, which is much larger than the 3-4
mm air-bubble diameter in the water-air system. Increasing the gas flow rate per branch gas channel makes bigger argon
bubbles, which matches the trend in the air-water system.
The average volume of the initial argon bubbles in the real caster is ~40  larger than the air bubble volume in the water
model. This is partly due to the ~15.6  higher argon gas volume flow rate that is needed to maintain the same gas volume
fraction in both systems. In addition, surface tension in molten steel-argon gas system is over 16  higher than water-air
system. This likely makes the argon bubble stay on the branch channel exit longer in the steel pool and produces larger
bubbles, even though molten steel flow has a larger shearing effect, due to its 1.7  higher velocity near the stopper rod. It is
interesting to note, however, that the scaled bubble diameter (dimension) in the steel caster roughly matches the expectation
of being 3  bigger, considering the 1/3 scale factor of the water model.
Rearranging Eq. (16) by replacing q g,mc and d avg, g with λq R( 273K ) and d g,R respectively, gives the total bubble frequency in the
real caster, fR as follows:

λq R( 273K ) (19)
fR 
π d g,R 
1 3

Compared to the measured total bubble frequency in the 1/3 scale water model (Figure 3(b)), the argon bubble frequency in
the real caster should be smaller. This is because the bubble formation time is longer in the molten steel-argon system than in
the water-air system and the bubble size is bigger in the real caster.
As investigated from the water-air model measurements, all 3 mechanisms of breakup, coalescence, and accumulation in the
nozzle are very significant to determination of the bubble size distributions in the nozzle and mold. Thus, CFD two-phase
flow models should include all of these complex bubble phenomena [31,32], if they aim to predict realistic argon bubble size
distributions in the nozzle and mold of the real steel caster.

Figure 8. Predicted initial bubble size and bubble frequency of argon gas in the real steel caster.

SUMMARY AND CONCLUSIONS

Bubble behavior and size distributions were investigated in the nozzle and mold of continuous steel-slab casting with a
stopper rod having multiple side-channels for gas injection. One-third scale water model experiments using a high-speed
video-camera and a pressure gauge, analytical model calculations, and CFD model predictions were applied to quantify
bubble formation, breakup, coalescence, and accumulation and their effects on bubble size distributions. The main findings
are as follows.

© 2019 by the Association for Iron & Steel Technology. 1327

 • Gas injected through the multiple channels in this stopper rod forms bubbles in four stages: initiation, expansion,
elongation, and detachment.
• Bubble formation is initiated when the inlet gas pressure exceeds the threshold pressure that consists of the pressure
drops across the gas channels and due to the sudden area contraction between the gas channels, the hydrostatic
pressure from the liquid level in the tundish, and the surface tension force at the branch-channel exit. A higher
threshold is needed to produce flow through all 6 branch channels, which are activated in sequence as the gas flow
rate is increased.
• With higher gas flow rate, more channels becoming activated decreases the bubble frequency exiting each channel
slightly, so the average initial bubble size increases slightly.
• After detaching from the stopper-rod head, bubbles flow into the gap between the stopper-rod head and the nozzle
inlet, where they break up due to the high velocity gradients and high turbulence. This drastically decreases bubble
size in the region just below the nozzle inlet.
• While flowing down the nozzle, small bubbles coalesce in complex manners while larger bubbles accumulate due to
their longer residence time. Both phenomena lead to increasing bubble size with distance down the nozzle, although
coalescence is more important.
• In the nozzle well bottom and jet, bubbles breakup due to the high turbulence. In the stagnant upper region of the
nozzle port, bubbles coalesce. In the mold, small bubbles accumulate due to their longer residence time. Measured
bubble size distributions in the mold are larger and have a broader size range compared with those in the nozzle,
suggesting that coalescence is the dominant mechanism for the conditions of this work.
• In both the nozzle and mold regions, the average and standard deviation of the bubble size increases with increasing
gas flow rate.
• The semi-analytical model of bubble formation that considers expansion and elongation stages shows good
agreement with the measured air-bubble size at stopper-rod branch channel exit in the 1/3 scale water model.
• Applying the validated two-step bubble-formation model to molten steel-argon gas system in the real steel caster
predicts much larger (~3  ) bubble size and lower bubble frequency, relative to those in the water-air model, as
expected considering the 1/3 scale factor.
• The injection of argon gas through multiple branch channels in the sides of the stopper rod tip is expected to be
better than central single-channel injection of high gas flow rates through the stopper-rod tip, by producing more
stable bubbly flow in the nozzle, avoiding annular / slug flow, and leading to less transient flow variations in the
nozzle and mold.

ACKNOWLEDGEMENTS
The authors thank Seon-Hyo Kim and Dae-Woo Yoon, POSTECH for help with the 1/3 scale water model experiments, and
Shin-Eon Kang, POSCO Technical Research Laboratories for providing the water model. Support from the Continuous
Casting Center at Colorado School of Mines, the Continuous Casting Consortium at University of Illinois at Urbana-
Champaign, POSCO, South Korea (Grant No. 4.0002397), and the National Science Foundation GOALI grant (Grant No.
CMMI 18-08731) are gratefully acknowledged. Provision of FLUENT licenses through the ANSYS Inc. academic
partnership program is also much appreciated.

REFERENCES
1. K. Rackers and B. G. Thomas, "Clogging in Continuous Casting Tundish Nozzles", 78th Steelmaking Conference
Proceedings, Nashville, TN, April 2, 1995, Iron and Steel Society, Warrendale, PA, Vol. 78, 1995, pp. 723-734.
2. L. Zhang and B. G. Thomas, “State of the Art in Evaluation and Control of Steel Cleanliness,” ISIJ International, Vol.
43, No. 3, 2003, pp. 271-291.
3. L. Zhang, J. Aoki, and B. G. Thomas, “Inclusion Removal by Bubble Flotation in a Continuous Casting Mold”,
Metallurgical and Materials Transactions B, Vol. 37B, No. 3, 2006, pp. 361-379.
4. Zhang, L., S. Yang, X. Wang, K. Cai, J. Li, X. Wan, and B.G. Thomas, “Investigation of Fluid Flow and Steel
Cleanliness in the Continuous Casting Strand”, Metallurgical and Materials Transactions B, Vol. 38B, No. 1, 2007, pp.
63-83.
5. M.Burty, C.Pussé, M.Alvarez, P.Gaujé, and G.Grehan, “Fundamental Study of Gas-Liquid Flows in CC Machine”,
84th Steelmaking Conference Proceedings, Baltimore, MD., March 25, 2001, Iron and Steel Society, Warrendale, PA,,
Vol. 84 2001, pp. 89-98.

1328 © 2019 by the Association for Iron & Steel Technology.

6. H. Bai and B. G. Thomas, “Turbulent Flow of Liquid Steel and Argon Bubbles in Slide-Gate Tundish Nozzles, Part II,
Effect of Operation Conditions and Nozzle Design,” Metallurgical and Materials Transactions B, Vol. 32B, No. 2, pp.
269-284, 2001.
7. L.C. Hibbeler and B. G. Thomas, “Mold Slag Entrainment Mechanisms in Continuous Casting Molds”, Iron and Steel
Technology, Vol. 10, No. 10, 2013, pp. 121-134.
8. T. Watanabe and M. Iguchi, “Water Model Experiments on the Effect of an Argon Bubble on the Meniscus Near the
Immersion Nozzle”, ISIJ International, Vol. 49, No. 2, 2009, pp. 182-188.
9. S. Yamashita and M. Iguchi, “Mechanism of Mold Powder Entrapment Caused by Large Argon Bubble in Continuous
Casting Mold”, ISIJ International, Vol. 41, No. 12, 2001, pp. 1529-1531.
10. B. G. Thomas, Q. Yuan, S. Mahmood, R. Liu, and R. Chaudhary, “Transport and Entrapment of Particles in Steel
Continuous Casting”, Metallurgical and Materials Transactions B, Vol. 45, No. 1, 2014, pp. 22-35.
11. K. Jin, B. G. Thomas, and X. Ruan, “Modeling and Measurements of Multiphase Flow and Bubble Entrapment in Steel
Continuous Casting”, Metallurgical and Materials Transactions B, Vol. 47B, No. 1, 2016, pp. 548-565.
12. J. Sengupta, H.-J. Shin, B. G. Thomas, S.-H. Kim, “Micrograph Evidence of Meniscus Solidification and Sub-Surface
Microstructure Evolution in Continuous-Cast Ultra-Low Carbon Steels”, Acta Materialia, Vol. 54, 2006, pp.1165-1173,
13. J. Sengupta, B. G. Thomas, H-J. Shin, G-G. Lee, S.-H. Kim, “A New Mechanism of Hook Formation during
Continuous Casting of Ultra-Low-Carbon Steel Slabs”, Metallurgical and Materials Transactions A, Vol. 37A, 2006,
pp. 1597-1611.
14. Y. Miki and S. Takeuchi, “Internal Defects of Continuous Casting Slabs Caused by Asymmetric Unbalanced Steel Flow
in Mold”, ISIJ International, Vol. 43, No. 10, 2003, pp. 1548-1555.
15. N. Kasai, H. Mizukami, and A. Mutou, “State of Segregation with Bubble in Continuously Cast Slab of Ultra Low
Carbon Steel”, Tetsu-to-Hagané, Vol. 89, No. 11, 2003, pp. 1120-1127.
16. S-M. Cho, B. G. Thomas and S-H. Kim, “Bubble Behavior and Size Distributions in Stopper-Rod Nozzle and Mold
during Continuous Casting of Steel Slabs”, ISIJ International, Vol. 58, No. 8, 2018, pp. 1443-1452.
17. C. A. Schneider, W. S Rasband, and K. W. Eliceiri, “NIH Image to ImageJ: 25 years of Image Analysis”, Nature
Methods, Vol. 9, No. 7, 2012, pp. 671-675.
18. H. Bai and B. G. Thomas, “Bubble Formation during Horizontal Gas Injection into Downward-Flowing Liquid”,
Metallurgical and Materials Transactions B, Vol. 32B, No. 6, 2001, pp. 1143-1159.
19. R. Clift, J. R. Grace, and M. E. Weber: Bubbles, Drops, and Particles., Academic Press Inc., 1978, pp. 111-113.
20. J. T. Kuo and G. B. Walis, “Flow of Bubbles through Nozzles”, Int. J. Multiphase Flow, Vol. 14, No. 5, 1988, pp. 547-
564.
21. ANSYS FLUENT 14.5-Theory Guide, ANSYS. Inc., Canonsburg, PA, USA, 2012.
22. I. Leibson, E. G. Holcomb, A. G. Cacoso, and J. J. Jamic: “Rate of Flow and Mechanics of Bubble Formation from
Single Submerged Orifices”, AIChE J., Vol. 2, 1956, pp. 296-306.
23. L. Davidson and E. H. Amick Jr., “Formation of Gas Bubbles at Horizontal Orifices”, AIChE J., Vol. 2, 1956, pp. 337-
342.
24. H. Ooyabu, A. Hiratsuka, R. Tsujino, and M. Iguchi, “Frequency of Bubble Formation from a Multi-Hole Nozzle
Attached to a Top Lance”, Mater. Trans., Vol. 50, No. 7, 2009, pp. 1812-1819.
25. G.M. Evans, G.J. Jameson, and B.W. Atkinson, “Prediction of the Bubble Size Generated by a Plunging Liquid Jet
Bubble Column”, Chemical Engineering Science. Vol. 47. No. 13/14, 1992, pp. 3265-3272.
26. D.E. Hershey, B.G. Thomas, and F.M. Najjar, “Turbulent Flow through Bifurcated Nozzles”, International Journal for
Numerical Methods in Fluids, Vol. 17, No. 1, 1993, pp. 23–47
27. S-M. Cho, B. G. Thomas, H-J. Lee, and S-H. Kim, “Effect of Nozzle Port Angle on Mold Surface Flow in Steel Slab
Casting”, Iron and Steel Technology, Vol. 14, 2017, pp. 76-84.
28. N. Kasai and M. Iguchi, “Water-model Experiments on Gas and Liquid Flow in the Continuous Casting Immersion
Nozzle”, Tetsu-to-Hagané, Vol. 91, No. 6, 2005, pp. 546-552.

© 2019 by the Association for Iron & Steel Technology. 1329

29. K. Jin, S. P. Vanka, and B. G. Thomas, “Large Eddy Simulations of Electromagnetic Braking Effects on Argon Bubble
Transport and Capture in a Steel Continuous Casting Mold”, Metallurgical and Materials Transactions B, Vol. 49B, No.
3, 2018, pp.1360-1377.
30. T. Zhang, Z. Luo, H. Zhou, B. Ni, and Z. Zou, “Analysis of Two-Phase Flow and Bubbles Behavior in a Continuous
Casting Mold Using a Mathematical Model Considering the Interaction of Bubbles”, ISIJ International, Vol. 56, No. 1,
2016, pp. 116-125.
31. H. Yang, S. P. Vanka, and B. G. Thomas, “A Hybrid Eulerian–Eulerian Discrete-Phase Model of Turbulent Bubbly
Flow”, Journal of Fluids Engineering, Vol. 140, 2018, pp. 101202.1-101202.12.
32. H. Yang, S. P. Vanka, and B. G. Thomas, “Modeling Argon Gas Behavior in Continuous Casting of Steel”, Volume
70, No. 10, 2018, pp. 2148–2156.

1330 © 2019 by the Association for Iron & Steel Technology.