You are on page 1of 9

reviewed paper

from R&D

Sujoy Pandit Patil and Dipak Mazumdar:

Prediction of strand superheat in continuous casting:


modelling and industrial scale measurements in steelmaking
tundish systems
Mathematical modelling of thermal fields in two different, industrial scale, continuous casting tundish systems including a rectangular shaped 28 t and a delta shaped 10 t tundish, respectively, was carried out employing the commercial CFD software
Fluent. Here, temperature distribution within the tundish was predicted from a given ladle opening temperature embodying
various boundary heat flux values reported in the literature. Parallel to numerical computation, experimental measurements
were also carried out in which immersion thermocouples together with available expert systems were used to measure and record temperatures at predetermined locations in the two tundish systems. In general, the predicted temperature was observed
to be somewhat higher than those measured on the shop floor. While the general temperature prediction framework appears
to be reasonably sound, it is argued that a closer fit between experimental and predicted temperature is required from the view
point of superheat prediction and control in the continuous casting mold.
Thus the subject of heat loss through various tundish surfaces together with the reported heat flux values were carefully examined and analysed. Based on this, the heat flux through the tundish free surface was modified and tuned empirically to fit the
experimentally measured data derived from the 28 t tundish. It was subsequently demonstrated that similar free surface heat
flux values are operational in the 10-t tundish, despite its markedly different operating conditions viz., size, different surface
area to volume ratio, mass flow rates etc. The two tundishes held melt covered by a thin slag (i.e., rice husk), similar free board
heights and employed a top refractory cover as well as identical refractory lining material and thickness and were, therefore,
expected to experience similar heat loss rates through various surfaces. Industrial measurements together with computational
results presented in this work have confirmed that even in physically covered tundish systems with an overlying slag phase,
heat loss through the free surface is dominant, with an area average contribution that is at least an order of magnitude higher
than those through refractory lined solid walls.

In continuous casting, melt super heat (i.e., the temperature of steel melt over and above liquidus temperature) is a
critical process parameter as this profoundly influences the
structure and properties of the continuously cast products. It
is rather well known that casting with low superheat eliminates centre-line and axial porosity and increases the percentage of equiaxed grains in continuously cast products
[1], thereby improving mechanical properties of steel.
Hence, prediction and control of superheat have justifiably
been the subject matter of considerable interest to steelmakers and researchers alike. In such a context, knowledge
of the molten steels temperature is a pre-requisite and this
necessitates continuous tracking of thermal history of the
melt from BOF/EAF to mold, either experimentally or
computationally. Rigorous and routine mapping of temperature of molten steel during various stages of steelmaking is cumbersome and difficult due to the high operating
temperatures and a relatively large size of steelmaking fur-

naces/transfer vessels. Furthermore, a point measurement


technique such as an immersion thermocouple often fails to
provide a representative estimate of the overall thermal
state of the system, particularly in larger vessels, which are
prone to thermal stratifications. Alternative to this, appropriately validated mathematical models can be applied to
predict the evolution of temperature during various stages
(from tapping through to mold filling) of steelmaking and
thereby, provide estimates of the associated superheat. The
present study is primarily concerned with this latter approach and involves modelling and mapping of temperature
fields in steelmaking tundish systems.
Recently, the present authors were engaged in a detailed
process engineering study of the hollow jet nozzle (HJN), a
device used for super heat extraction between the tundish
and the continuous casting mold. In this, as illustrated in
fig. 1, the typical SEN design is altered to such an extent as

Fig. 1: Schematic of tundish - SEN-mold assembly; a) conventional casting and b) casting with a hollow jet nozzle

5 (2007) No. 2

119

from R&D

reviewed paper
1

to convert the latter into an effective heat exchanger [2] . It


is readily apparent that for a desired superheat/melt temperature in the mold, the intensity of cooling (i.e., the corresponding water flow rate) in the HJN can be explicitly
prescribed provided temperature of the incoming steel from
the tundish is known. A key to better process control and
efficient operation of the HJN system is an accurate, a priori knowledge of melt temperature distribution in the
steelmaking tundish system.
In the liquid steel processing circuitry, the tundish is the
last reactor or vessel in which molten steel undergoes both
physical and chemical changes. Thus, a precise knowledge
of temperature distribution within the tundish is required
for correct inference of superheat prior to casting. Mathematical modelling of thermal fields and prediction of temperature distribution in the tundish (or, as a matter of fact,
any industrial furnaces/reactors) embody empiricism as
these rely on measured heat flux/temperature at tundish
walls and free surfaces. Such empirical inputs largely determine the accuracy with which temperature fields in such
processing units can be predicted. To this end, a search in
relevant literature indicates that uncertainties still exist as

Fig. 2: Schematic of the 28-t slab casting tundish

Consequently, the primary objective of the present study


has been to assess the adequacy and appropriateness of the
reported surface boundary condition on tundish systems
and thereby, develop a mathematical model for prediction
and control of melt superheat during continuous casting operation. To demonstrate the present approach, measured melt
Table 1: Surface heat flux values as applied by two different groups of investigators for
temperatures derived from a 28-t
modelling of thermal fields in steelmaking tundish systems
slab casting and a 10-t bloom
Researcher
surface heat flux, kW/m
casting tundish system, operated
bottom
longitudinal
transverse
top surface
at 3.5 t/min and 0.54 t/min, rewall
vertical wall
wall
spectively, were applied and diChakraborty and Sahai [3] 1.4
3.2
3.8
15
rectly assessed against matheJoo and coworkers [4]
2.6
2.6
2.6
75 (with 30 mm slag layer)
matical model predictions.
Table 2: Physical dimensions and operating parameters in
the 28-t single-strand slab casting tundish
Parameters
tundish length (L): at base and top

numerical values
4284.1 - 4562.6 mm

tundish width (W): at base and top

840 - 1132.6 mm

melt depth (H)

1042 mm

inlet stream velocity

1.37m/s

inlet nozzle diameter

88mm

outlet nozzle diameter

154mm

working fluid

steel

inlet stream temperature

1848 - 1873 K

mass flow rate

3.5 t/min

Present work
Industrial tundish systems and experimental measurements. Industrial measurements of melt temperatures
in tundishes were undertaken in two different steel plants,
varying widely in their capacity and steelmaking technology. In one of these, a 210-t ladle, a 28-t tundish and a thin
slab caster while in the other, a 44-t ladle, a 10-t tundish
and a two strand bloom caster were used to continuously
transform molten steel into thin slabs and blooms, respectively.
Fig. 2 is a schematic representation of the slab casting
tundish. There, as can be clearly seen, the tundish has a

far as thermal boundary conditions on various tundish surfaces are concerned [3; 4], see table 1. There [3; 4], markedly different heat fluxes through the top melt surface were
considered, while practically similar heat flux through various refractory walls have been predetermined. Furthermore,
despite many studies in the area [5...7], the reliability of
such boundary heat flux is not known with any certainty
since predicted temperature fields have rarely been validated against corresponding industrial scale measurements.
The present study seeks to fill the above mentioned gaps in
the literature through mathematical modelling and high
temperature measurement of thermal fields in industrial
scale tundish systems.
1

Heat from the flowing melt is removed by circulating water along


the periphery of the central cylindrical portion of the HJN

120

Fig. 3: Schematic of the 10-t bloom casting tundish

5 (2007) No. 2

reviewed paper

from R&D

rectangular cross-section with sloping sides and is equipped


with a turbo stop as well as a small dam at the far end. The
side walls have a taper of about 7o while the turbostop and
dam are provided to aid surface directed flows for better inclusion float out and removal. The corresponding operating
parameters are summarised in table 2. During casting, a
submerged stream of liquid steel enters from the ladle to the
tundish via a shroud, submerged approximately 617 mm
below the free surface to avoid any entrainment of air into
the melt.
Similarly, a schematic of the 10 t bloom casting tundish is
shown in fig. 3 while corresponding physical dimensions
and operating parameters are summarised in table 3. As
shown, the given tunduish has a delta shape and is provided
with three strands. Of these, one of the strands is currently
not available and the tundish is operated in a two-strand
mode, as has been illustrated in fig. 3. The given tundish,
like its other typical industrial counterpart, has sloping side
walls, is equipped with a turbo-stop and is operated with a
shrouded and submerged liquid steel stream.
Table 3: Physical dimensions and operating parameters in
the two-strand 10-t bloom casting tundish
Parameters
tundish front length (L): at base and top

numerical values
2000 - 2168.44 mm

maximum tundish width (W): at base and top 1580- 1740 mm


melt depth (H)

700 mm

inlet stream velocity

0.6548 m/s

inlet nozzle diameter

50 mm

outlet nozzle diameter

30 mm

working fluid

steel

inlet stream temperature

1837 - 1865 K

mass flow rate

540 kg/min

In both the steel plants, melt temperatures in the tundish


during casting are routinely measured via an immersion
thermocouple. This is coupled to an expert system which
converts EMF output of the thermocouple into a representative temperature following well established procedures.
Most of the temperature measurements considered and reported in this study were from periods during which the
tundishes were operated under a practically steady state
condition as was confirmed by a near constant melt level.
During a typical experimental campaign, two sets of time
and temperature were recorded and these included the time
of placement of the ladle on the turret and the corresponding melt temperature (termed as ladle opening temperature)
as well as the instant of measurement and the corresponding melt temperature. These allowed us to infer melt temperature at the shroud at the instant of temperature measurement by incorporating appropriate corrections due to
thermal losses from the ladle (see below). The immersion
thermocouples were always dipped into the melt in a specified region as was determined by the local operator. Successive measurements in one of the two tundishes indicated
that variations in the measuring locations to the tune of 30
cm or so, does not have much influence on the recorded
temperature. In addition to these, in a few heats, two successive measurements were carried out and these in general
indicated good reproducibility of measured data. Experimental measurements of temperature were carried out dur5 (2007) No. 2

ing many sequence casting operations, involving various


grades of steel and different ladle opening temperatures.
Mathematical modelling. Assumptions in modelling. A
mathematical model for thermal energy transport in industrial scale tundish systems was derived in conformity with
the actual industrial practices, embodying the following assumptions and idealization:
the flow is three-dimensional, turbulent and macroscopically steady i.e., the phenomena involved during
filling and emptying of the tundish were not considered
in the present study;
the melt surface is assumed to be flat and mobile. The
formation of waves or the presence of slag at the free
surface were ignored as far as fluid flow simulation was
concerned [8];
entrainment of air or gas via the pouring stream was neglected, i.e., the ladle stream, which was the industrial
case, was assumed to be shrouded and submerged;
the flow is Newtonian and incompressible;
the system is non-isothermal. One of the reasons for
non-isothermality is the different rate of heat losses
through the top surface and the tundish walls. As far as
thermal energy transport in the tundish is concerned,
forced convection effects were assumed to dominate
over free convection ( Gr / Re 2  1 ) and accordingly,
governing equations of flow and temperature were
considered to be coupled one way only. Thus, Boussenisques approximation was not invoked and the corresponding free convection effect not included in the
vertical direction momentum balance equation;
the temperature of the incoming liquid steel stream into
the tundish is assumed to be known from the corresponding ladle opening temperature, measured on the
turret and an imposed rate of heat loss. Due to the heat
loss, temperatures at the shroud were assumed to change
at the rate of -0.5 C/min [9], finally;
in the presence of a steady flow field, a thermal energy
balance equation was solved in an unsteady state mode
primarily to accommodate a time dependent temperature at the shroud. This required a time dependent
boundary condition in the ladle shroud region.
The implications of some of the assumptions are discussed in more detail in a subsequent section.
The governing equations and boundary conditions.
Within the framework of the assumptions mentioned above,
the governing equations of continuity, motion, turbulence
and the thermal energy balance can be conveniently cast in
the Cartesian co-ordinate system and represented in compact tensorial notations according to:
equation of continuity:

( ) =0

u j
x j

(1)

equation of motion:

ui u j
x j

) = p +
xi

u u j

eff i +
+ g i , (2)
x j
x j x i

121

from R&D

reviewed paper
Table 4: List of symbols

conservation of turbulence kinetic energy:

eff k

u j k
= G ,
x j
k x j

(3)

(4)

conservation of thermal energy:

( T )
t

u iT
K eff T
+
=

x j
x j C x j ;

(5)

for explanation of symbols, see table 4. The following


auxiliary relationships/expressions were embodied in eqs.
(2) through (4):

eff = 1 + t ,

(6)

t = C D k 2 /

(7)

universal dissipation rate constant

volumetric rate of generation of turbulence kinetic

u u u j
G = t i i +
.
x j x j xi

accelaration due to gravity

depth of liquid in the tundish

turbulence kinetic energy

thermal conductivity

length of the tundish

pressure

heat flux

time

temperature

horizontal velocity component

vertical velocity component

transverse velocity components

width of the tundish

spatial coordinate





viscosity
turbulence kinetic energy dissipation rate
density

suffix

and
(8)

The tundish systems, as already illustrated, are physically


bounded by two different types of surfaces viz., solid walls
(the bottom and the four side walls) and a free surface. At
these surfaces, the following boundary conditions were applied to the governing equation of flow, turbulence and heat
transfer i.e.,
at the bottom wall,

u = 0 , v = 0 , w = 0 , k = 0 , = 0 , qwall = q1 ;
at the free surface,

v = 0,

specific heat

C

energy

conservation of dissipation rate of turbulence kinetic energy:

eff C 1G C 2

u j
=
k
x j
x j

u
w
k
= 0,
= 0,
= 0,
z
z
z

i, j

tensorial notations

eff

effective

laminar

turbulent

wall

tundish refractory wall

were computed first on the basis of applicable mass flow


rate, and applied subsequently to predict temperature fields
in the tundish to infer the associated superheat in the mold
during continuous casting. Thermal calculations were carried out considering turbulent Prandtl number, t equal to
unity.
Numerical simulation. Numerical simulation was initiated
by creating the tundish geometry and laying the numerical
grid within. These were accomplished in Gambit [11] (the
geometry and mesh building interactive tool), the preprocessor to Fluent [10]. The volume meshing produced
3,52,968 and 4,06,035 cells in the 10-t and 28-t tundish, respectively. During meshing, due efforts were made to en-

= 0 , qfreesurf = qrad ;
z
at the solid side walls,

u = 0 , v = 0 , w = 0 , k = 0 , = 0 , qwall = q2 ;
at the front and rear walls,

u = 0 , v = 0 , w = 0 , k = 0 , = 0 , qwall = q3 .
In the present study, q1, q2, q3 and qrad, the heat fluxes
through the various tundish walls and free surfaces were assigned to their values shown in table 1 in a straightforward
fashion. Eqs. (1) through (8), together with their associated
boundary conditions were solved numerically via Fluent
[10], a state of the art CFD (computational fluid dynamics)
software. In this, the steady state flow and turbulence fields
122

Fig. 4: Flow diagram of the Fluent based computational


model

5 (2007) No. 2

reviewed paper
sure as large a proportion of hexahedral volume elements in
the domain as possible. Subsequent to this, solver version
(Fluent 6.1) was selected and the continuum as well as the
various surfaces were specified. Finally, the grid was
checked for internal consistency and exported to Fluent as
a.msh file, where solver settings, models, material properties, necessary initial and boundary conditions etc. were
specified to set up a case file as per the geometry, governing equations and boundary conditions presented earlier. A
typical computation was initiated by iterating the case file
and terminating the same once the scaled residual for flow,
turbulence and temperature fell below 10-3 and 10-8, respectively. The salient features of the Fluent based computational scheme, fig. 4, are listed below:
solver characteristics: 3D, segregated, implicit and
steady;
viscous model: k- model [12] with standard wall functions;
energy equation: transient with prescribed heat fluxes at
all the bounding surfaces;
material: steel; see, for example, table 5.

from R&D

Fig. 5: Comparison between predicted and experimental


temperatures at location X ~ 225 mm, Y ~ 0 and Z ~ 729.4
mm in the 28-t tundish embodying the two different sets of
surface heat flux values summarised in table 1

stream of liquid steel into the tundish was taken to vary


with time. Based on such information and the measured ladle opening temperature (corresponding to an arbitrarily defined
Results and discustime, t = 0), an expression for the
Table 5: Thermophysical properties of steel used in
sion. The ratio Gr/Re2
variation of melt temperature at the
numerical computations
in the two tundish sysladle shroud as a function of time
Property
numerical value
tems estimated on the
was deduced and incorporated in
density, kg/m3
7000
basis of conditions
the heat transfer model.
specific heat, J/kgK
632
(velocity and temThus, embodying two sets of thermal
thermal conductivity, W/mK
41
perature) prevalent at
boundary conditions [2; 3] summarized
viscosity, kg/ms
0.007
the ladle shroud was
in table 1 in the mathematical model,
found to be significomputations were carried out for
cantly smaller than unity implying, essentially, that free many heats (viz., characterized by grades of steel and ladle
convection effects are not important in the study of fluid opening temperature) in the 28-t tundish. Results thus obflow and associated phenomena in the tundish systems. tained were subsequently compared with temperatures reFrom a computational standpoint, this further implies that corded with thermocouples immersed at X ~ 225 mm, Y ~ 0
flow and turbulence model equations can, in principle, be and Z ~ 729.4 mm, respectively. A direct comparison besolved in isolation of the thermal energy balance equation. tween experimental and predicted temperature is illustrated
Thus, despite a variable incoming liquid steel temperature, in fig. 5. On this basis, the following observations can be
it is legitimate to assume turbulent flow phenomena to be made:
steady since Gr / Re 2  1 and the melt maintained at a the boundary heat flux reported by Chakraborty and Sahai [3] in general produces relatively higher estimates of
constant level. As a consequence, a steady turbulent flow
temperature than those of Joo et al. [4]. This is to be exmodel was used for flow simulation and an unsteady therpected since the values reported by the former authors
mal model for temperature prediction.
are somewhat smaller than those of the latter;
Despite a slag cover in practice, the presence of a slag
layer at the melt surface was ignored for modelling flow predicted temperatures in general tend to be higher than
those measured experimentally. The largest deviation
phenomena. It has been assumed that the overlying slag
between experimental measurements and computed rephase does not influence flow phenomena in tundish syssults amounts up to 34 C indicating a maximum mistem to any significant extent as experimental RTD characmatch between the two of about 2 % or so;
teristics with and without a simulated slag were shown to
be largely similar [8], particularly in water model tundish however, instead of absolute temperature, if experimental measurements and computed results had been
systems. Such observations essentially justify bulk flow
visualized from the standpoint of melt superheat, more
velocities and turbulence parameters in tundish system not
pronounced differences between prediction and measbeing critical to the details of the free surface conditions.
urements than that mentioned above would have reThe temperature of liquid steel in the ladle is expected to
sulted (for example, about 40 % or so);
change as the ladle is being continuously emptied into a
tundish. This, in turn, will render thermal energy transport the discrepancy between experimental measurements
and numerical predictions, as illustrated in fig. 5, exan unsteady phenomena in the tundish, despite a steady
hibits a systematic trend (viz., the predicted values are
flow condition in the system. Available evidence in the litconsistently higher than the experimental ones). This
erature appear to indicate that temperature drop to the extends to point to the inadequacy of the calculation protent of 0.5 C/min is typical of ladles during holding pericedure rather than the measurement technique;
ods [9]. Accordingly, the temperature of the incoming
5 (2007) No. 2

123

from R&D
on the basis of the foregone discussion, it is therefore
apparent that if the ultimate objective of the computational exercise is to predict melt superheat in continuous
casting for engineering of HJN, cast microstructure etc.,
a much closer fit between experimental and computed
temperature fields is required.
To assess the general adequacy of the computational procedure developed, the influence of grid size, time step size,
materials property, continuity of mass inflow rate etc. on
predicted results were investigated numerically. It was observed that within the range of values investigated, numerical predictions are not critical to the precise choice of various numerical parameters and material property values. To
investigate this further, melt flow and the associated temperature fields in the 10-t tundish were computed at a
throughput rate of 540 kg/min for different shroud temperatures and melt chemistry, embodying the heat flux values summarized already in table 1. A comparison between
predicted and experimentally measured temperatures recorded at X ~ -260 mm, Y ~ -630 mm and Z ~ 600 mm (the
line on which shroud is located is used as the reference X =
0 and Y = 0) is illustrated in fig. 6. There, differences between experimental measurements and computed results
comparable to those in fig. 5, are readily evident. As can be
seen, the predicted melt temperatures are consistently
higher than those registered by the immersion thermocouples. Results and discussion presented so far therefore appear to suggest that currently available heat flux values are
inadequate and required to be improved, if melt superheat
in continuous casting is to be predicted with greater accuracy. However, before an effort is made in this direction, it
is worthwhile to re-visit the two previous studies [3; 4] and
carefully examine the general adequacy of heat flux values
presented in table 1.
Chakraborty and Sahai [3] were among the first to carry
out elaborate numerical computation of melt temperature
distribution in steelmaking tundish systems embodying heat
flux values reported in table 1. Their [3] wall heat flux values, as one would note, were derived by equating steady
state conductive heat losses through the walls with natural
convection at the external tundish surfaces embodying an
available temperature wall function correlation. It was further suggested that their estimates of wall heat fluxes agree
reasonably well with an equivalent estimate reported in literature. However, these authors provided no justification
for their choice of free surface heat flux value (15 kW/m2).
Subsequently, for a similar problem, Joo and coworkers [4]
also estimated the steady state heat fluxes through the walls
and a slag covered melt surface of a tundish through an
idealized thermal energy balance. Their [4] estimates of
wall heat fluxes, while they are very similar to those of
Chakraborty and Sahai [3], show that large differences
between respective estimates of free surface heat flux exist,
as reported by the two group of investigators (e.g., see table
1). Given that tundish lining material and thickness do not
vary appreciably from one practice to the other, similarity
in values in heat flux is expected and therefore values as
shown in table 1 can be taken to be reasonably representative and more or less universal for steelmaking tundish
systems. In contrast, heat flux through the top surface is
expected to be a function of specific industrial practice and
124

reviewed paper
therefore is likely to vary on a case by case basis. For example, a bare tundish as opposed to a slag covered or a slag
and physically covered tundish will exhibit a markedly different heat loss rate through the top surface. Similarly, the
thickness of the overlying slag phase is also likely to influence such estimates to some extent. Looked at from such
standpoints, heat flux acting through tundish top surface
can be expected to be plant specific and, therefore, somewhat different. Operating conditions (viz., tundish cover,
slag layer thickness etc.) for which the free surface heat
flux values in table 1 are tenable, have, however, not been
spelt out by the authors [3; 4] in any detail.
Rigorous determination of heat flux through the top surface under industrial conditions is likely to pose considerable difficulties since heat transfer through several composite layers is involved in the process. Typically, heat from
the melt surface to the ambient air, is transported through a
slag layer, an intermediate volume of air separating the tundish cover and the overlying slag and finally through the refractory shield/cover. The success with which reasonably

Fig. 6: Comparison between predicted and experimental


temperatures at location at X ~ -260 mm, Y ~ -630 mm and Z
~ 600 mm in the 10-t tundish embodying the two different
sets of surface heat flux values summarised in table 1

accurate heat flux values can be estimated under such condition is uncertain as this tends to depend on several interfacial contact resistances in the domain, which are either
unknown or at best approximate. As a much simpler alternative, experimentally measured temperatures reported in
this study can be applied to deduce a relatively more accurate top surface heat flux for the present industrial practice.
The heat flux values summarized in table 1 clearly indicate
that heat loss through tundish free surface is significantly
higher than those through refractory walls. Consequently, a
closer fit between experimental and predicted temperature
can be achieved by enhancing the free surface heat flux
value [4] to some extent. An attempt has therefore been
made to force fit predictions with experimentally measured
temperatures and thereby, deduce a more accurate heat flux
value operational in the slag and physically covered, slab
casting tundish system.
Thus, a large number of numerical computations were
carried out and temperature within the 28-t tundish predicted by enhancing the free surface heat flux value beyond
75 kW/m2. Successive comparisons between numerically
5 (2007) No. 2

reviewed paper

from R&D

Fig. 7: Comparison between predicted and experimental


temperatures at location X ~ 225 mm, Y ~ 0 and Z ~ 729.4
mm in the 28-t tundish embodying top surface heat flux value
of 240 kW/m2

Fig. 8: Comparison between predicted and experimental


temperatures at location X ~ -260 mm, Y ~ -630 mm and Z ~
600 mm in the 10-t tundish embodying a top surface heat
flux of 240 kW/m2 (included in the figure are also the data
points for 28-t tundish from the preceding figure)

predicted and experimentally measured temperatures indicated a best fit between the two for a free surface heat flux
value of about 240 kW/m2. This is illustrated in fig. 7. Embodying, such heat flux values, computations were subsequently carried out for the 10-t, delta shaped tundish system. The results thus obtained are shown in fig. 8. There,
substantially superior agreement between experimental
measurements and numerical predictions than those illustrated in fig. 6, is at once evident.
The heat flux from an uncovered melt surface is primarily radiative in nature and can be represented as:

q = s4 a4 .

(9)

Under steelmaking conditions, taking s ~ 1873 K and a Fig. 9: Comparison between predicted and experimental
~ 298 K (considering  ~0.8 and  = 5.65 10-8 W/m2 K4), temperatures at location X ~ -260 mm, Y ~ -630 mm and Z ~
the corresponding heat flux can be readily estimated and 600 mm in the 10-t tundish embodying top surface heat flux
2
found to be about 556 kW/m2. Consequently, the surface of 170 kW/m
heat flux value implicit in the results presented in figs. 7
and 8, as a first approximation, indicates that slag and tun- since the free surface area to volume ratio for the 10-t tundish covers combined could cut back radiation losses in in- dish is smaller than that of the 28-t slab casting tundish.
dustrial tundish systems to the extent of 57 % or so. Results Despite this, the present study appears to indicate that heat
presented in figs. 7 and 8, also as a definite possibility sug- fluxes operating through the top surface in industrial scale
tundish systems are
gest that much larger heat flux through the
considerably larger
top surface of tundishes, than those anticiTable 6: Predicted strand superheat in the 28-t
than those anticipated
pated earlier, are in fact operational during
tundish for various ladle opening temperatures
earlier [3; 4]. On the
continuous casting operations.
(grade of steel: low carbon (0.035 %C and liquibasis of the present
As a final point, a unique heat flux
dus temperature of 1808 K)
findings, it is therethrough free surface for different tundish
Heat
ladle opening predicted strand
fore suggested that
systems operated in two different steel
number
temperature, C
free surface heat flux
plants seem impractical since the surface
1
1864
1835
values in the range of
area to volume ratio, nature of slag cover2
1860
1831
ing, tundish free board height etc. varied
170 - 240 kW/m2 are
3
1861
1832
from one practice to the other. Indeed, adlikely to be represen4
1864
1835
ditional computations carried out to estabtative of slag and
5
1867
1838
lish a more accurate fitting between measphysically
covered
6
1855
1827
urement and prediction suggests a sometundish
systems,
7
1873
1844
what smaller free surface heat flux value
having a capacity in
8
1848
1820
the range of 10 to 28
(170 kW/m2) for the 10-t tundish system.
t. Based on the reThis is illustrated in fig. 9 in which, much
better correspondence between measurements and predic- vised surface heat flux values, temperature of liquid steel
tion than those illustrated in fig. 8, is readily evident. A leaving the 28-t tundish and the attendant superheat in the
smaller heat flux through the free surface is not unlikely mold were estimated via the mathematical model for differo

5 (2007) No. 2

125

from R&D
ent ladle opening temperatures. For a typical low-carbon
grade (0.035 %C; liquidus temperature = 1808 K) this is
summarized in table 6. This suggests that the temperature
of steel entering the 28-t tundish must be constrained within
1850 - 1855 K to ensure a superheat of about ~20 K in the
mold.
Conclusions
A combined computational and experimental approach
has been adapted to investigate melt flow and associated
heat transfer in two different industrial scale tundish systems. Embodying the heat flux values published in literature, numerical simulation of thermal fileds were carried
out for a 10-t bloom casting tundish and a 28-t slab casting
tundish system. In general, the predicted temperatures were
found to be higher than those measured experimentally.
The reported heat flux values, particularly those working
through the free surface were critically analysed and fine
tuned with the aid of present industrial measurements. This
resulted in a much closer agreement between experimental
and predicted temperature fields. The present study has
demonstrated that free surface heat flux in the range of
170 - 240 kW/m2 together with the reported wall heat flux
values, produce reasonably accurate temperature fields for
the range of tundish operations considered in this study.
Industrial measurements coupled with computational results have further confirmed that heat fluxes operating
through the top surface in industrial scale tundish systems
are significantly larger than those anticipated earlier.
References
[1] J.P. Birat, M. Bobedilla, J.L. Jacquot, J.Rour, L.Baker and J.M.
Bastin: Proc. Continuous casting, Institute of Metals, London, 1985,
p. 18.1/18.9.

126

reviewed paper
[2] P. Naveau, S. Wilmotte and C. Albrecq: Proc. Steelmaking Conference, ISS, 1996, p. 2280/33.
[3] S. Chakraborty and Y. Sahai: Ironmak. Steelmak. 19 (1992) No. 6, p.
488/94.
[4] S. Joo, J.W. Han and R.I.L. Guthrie: Metall. Trans. 24B (1993), p.
767/77.
[5] R.D. Morales, J. De J. Barreto, S. Lopez-Ramirez, J. Palafox Ramos
and M. Diazcruz: Mater. Sci. Eng. 8 (2000), p. 781/801.
[6] H. Bebber and A. Kranz: steel res. 72 (2001) No. 11+12, p. 460/65.
[7] O. J. Ilegbusi and J. Szekely: steel res. 62 (1991) No. 5, p. 193/200.
[8] A. Kumar: Fluid flow and residence time distributions in a multistrand tundish system, Kanpur (India), 2005 (Ph.D diss.).
[9] A. Ghosh: Secondary Steelmaking, CRC Press, Boca Raton, USA,
2002.
[10] Fluent 6.1 Documentation CD, Fluent Inc, Lebenon, NH, 2003.
[11] Gambit 2.1 Documentation CD, Fluent Inc, Lebenon, NH, 2003.
[12] B.E. Launder and D.B. Spalding: Mechanic. & Engg. 3 (1974), p.
269/89.

Sujoy Pandit Patil


currently Graduate Engineer
Trainee
JSW Steel works, Torangallu
Bellary

Dipak Mazumdar
Department of Materials &
Metallurgical Engineering
Indian Institute of Technology
Kanpur
India

5 (2007) No. 2

You might also like