03 MTB Meng CON1D Reprint Post

Heat-Transfer and Solidification Model of Continuous
Slab Casting: CON1D

YA MENG and BRIAN G. THOMAS
A simple, but comprehensive model of heat transfer and solidification of the continuous casting of
steel slabs is described, including phenomena in the mold and spray regions. The model includes a
one-dimensional (1-D) transient finite-difference calculation of heat conduction within the solidifying
steel shell coupled with two-dimensional (2-D) steady-state heat conduction within the mold wall.
The model features a detailed treatment of the interfacial gap between the shell and mold, including
mass and momentum balances on the solid and liquid interfacial slag layers, and the effect of oscillation
marks. The model predicts the shell thickness, temperature distributions in the mold and shell, thickness
of the resolidified and liquid powder layers, heat-flux profiles down the wide and narrow faces, mold
water temperature rise, ideal taper of the mold walls, and other related phenomena. The important
effect of the nonuniform distribution of superheat is incorporated using the results from previous three-
dimensional (3-D) turbulent fluid-flow calculations within the liquid pool. The FORTRAN program
CONID has a user-friendly interface and executes in less than 1 minute on a personal computer.
Calibration of the model with several different experimental measurements on operating slab casters
is presented along with several example applications. In particular, the model demonstrates that the
increase in heat flux throughout the mold at higher casting speeds is caused by two combined effects:
a thinner interfacial gap near the top of the mold and a thinner shell toward the bottom. This modeling
tool can be applied to a wide range of practical problems in continuous casters.
I. INTRODUCTION move downward, causing slag to be consumed at a rate

balanced by the replenishment of bags of solid powder to
HEAT transfer in the continuous slab-casting mold is the top surface. Still more slag is captured by the oscillation
governed by many complex phenomena. Figure 1 shows a marks and other imperfections of the shell surface and carried
schematic of some of these. Liquid metal flows into the mold downward at the casting speed.
cavity through a submerged entry nozzle and is directed by These layers of mold slag comprise a large resistance
the angle and geometry of the nozzle ports.[1] The direction to heat removal, although they provide uniformity relative to
of the steel jet controls turbulent fluid flow in the liquid the alternative of an intermittent vapor gap found with oil cast-
cavity, which affects delivery of superheat to the solid/liquid ing of billets. Heat conduction across the slag depends on the
interface of the growing shell. The liquid steel solidifies thickness and conductivity of its layers, which, in turn, depends
against the four walls of the water-cooled copper mold, while on their velocity profile, crystallization temperature,[6] viscosity,
it is continuously withdrawn downward at the casting speed. and state (glassy, crystalline, or liquid). The latter can be
Mold powder added to the free surface of the liquid steel determined by the time-temperature-transformation (TTT)
melts and flows between the steel shell and the mold diagram measured for the slag, knowing the local cooling
wall to act as a lubricant,[2] so long as it remains liquid. The rate.[7,8,9] Slag conductivity depends mainly on the crystallinity
resolidified mold powder, or “slag,” adjacent to the mold of the slag layer and on the internal evolution of its dissolved
wall cools and greatly increases in viscosity, thus acting like gas to form bubbles.
a solid. It is thicker near and just above the meniscus, where Shrinkage of the steel shell away from the mold walls
it is called the “slag rim.” The slag cools rapidly against the may generate contact resistances or air gaps, which act
mold wall, forming a thin solid glassy layer, which can as a further resistance to heat flow, especially after the slag
devitrify to form a crystalline layer if its residence time in is completely solid and unable to flow into the gaps. The
the mold is very long.[3] This relatively solid slag layer often surface roughness depends on the tendency of the steel shell
remains stuck to the mold wall, although it is sometimes to “ripple” during solidification at the meniscus to form an
dragged intermittently downward at an average speed less uneven surface with deep oscillation marks. This depends
than the casting speed.[4] Depending on its cooling rate, on the oscillation practice, the slag-rim shape and properties,
this slag layer may have a structure that is glassy, crystalline, and the strength of the steel grade relative to the ferrostatic
or a combination of both.[5] So long as the steel shell remains pressure, mold taper, and mold distortion. These interfacial
above its crystallization temperature, a liquid slag layer will resistances predominantly control the rate of heat flow in
the process.
Finally, the flow of cooling water through vertical slots
in the copper mold withdraws the heat and controls the
YA MENG, Graduate Student, Department of Materials Science and temperature of the copper mold walls. If the “cold face” of
Engineering, and BRIAN G. THOMAS, Professor, Department of Mechan-
ical and Industrial Engineering, are with the University of Illinois at the mold walls becomes too hot, boiling may occur, which
Urbana–Champaign, Urbana, IL 61801. Contact e-mail: [email protected] causes variability in heat extraction and accompanying
Manuscript submitted August 27, 2002. defects. Impurities in the water sometimes form scale deposits
METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—685

models are very sophisticated (even requiring supercomputers
to run), so are infeasible for use in an operating environment.
The earliest solidification models used one-dimensional
(1-D) finite-difference methods to calculate the temperature
field and growth profile of the continuously cast steel
shell.[16,17] Many industrial models followed.[18,19] These
models first found application in the successful prediction
of metallurgical length, which is also easily done by solv-
ing the following simple empirical relationship for dis-
tance (z) with the shell thickness (S) set to half the section
thickness:
S 5 K Ö z /Vc [1]
where K is found from evaluation of breakout shells and
computations. Such models found further application in
troubleshooting the location down the caster of hot-tear
cracks initiating near the solidification front[20] and in the
optimization of cooling practice below the mold to avoid
Fig. 1—Schematic of continuous casting process showing slag layers (not subsurface longitudinal cracks due to surface reheating.[21]
to scale). Since then, many advanced models have been developed
to simulate further phenomena such as thermal stress and crack-
related defects[12,22,23] or turbulent fluid flow[24–28] coupled
on the mold cold face, which can significantly increase the together with solidification. For example, a two-dimensional
mold temperature, especially near the meniscus where the (2-D) transient, stepwise coupled elasto-viscoplastic finite-
mold is already hot. After exiting the mold, the steel shell element model tracks the behavior of a transverse slice through
moves between successive sets of alternating support rolls a continuously cast rectangular strand as it moves down through
and spray nozzles in the spray zones. The accompanying the mold at casting speed.[12] This model is suited for simulating
heat extraction causes surface-temperature variations while longitudinal phenomena such as taper design,[29] longitudinal
the shell continues to solidify. cracks,[30] and surface depressions.[31] Other casters have
It is clear that many diverse phenomena simultaneously been modeled using three-dimensional (3-D) coupled fluid
control the complex sequence of events which govern heat flow–solidification models[27] based on control-volume or finite-
transfer in the continuous casting process. The present work difference approaches, at the expense of greater computation
was undertaken to develop a fast, simple, and flexible model time and memory.
to investigate these heat-transfer phenomena. In particular, To study the temperature distribution and/or distortion of
the model features a detailed treatment of the interfacial gap the mold, 3-D finite-element thermal-stress models have been
in the mold, which is the most important thermal resistance. applied.[10,32] These models have been crucial in determining
The model includes heat, mass, momentum and force the axial heat-flux profile based on measured temperatures
balances on the slag layers in the interfacial gap. in the mold walls.[20,32,33] This procedure is sometimes auto-
This model is part of a larger comprehensive system of mated with inverse heat-conduction models.[20]
models of fluid flow, heat transfer, and mechanical behavior, One of the greatest resistance to heat transfer from the liquid
which is being developed and applied to study the formation steel to the mold cooling water is the interface between the
of defects in the continuous casting process. These other mold and shell. Heat transfer across this interface is controlled
models are used to incorporate the effects of mold distor- by the thickness and thermal properties of the materials that
tion,[10] the influence of fluid flow in the liquid pool on fill the gap. Despite its known importance, most previous
solidification of the shell,[11] and coupled thermal-stress mathematical models characterize the interface as a boundary
analysis of the shell to find the reduction of heat transfer condition for a model of either the shell or the mold alone.
across the interface due to air-gap formation.[12] Even models of both usually use a simplified treatment of
This article first describes the formulation of this model, the gap.[34,35,36]
which has been implemented into a user-friendly FORTRAN A few models have considered more-detailed treatment
program, CONID, on personal computers and UNIX work- of the resolidified powder layers in the gap and can calculate
stations. Then, validation of the model with analytical slag-layer thicknesses,[37,38] the slag velocity profile along
solutions and calibration with example plant measurements the film thickness,[38,39] and interface friction.[37,38,39] Common
are presented. Finally, the effect of casting speed on mold oversimplifications include neglecting the solid slag layer,[40]
heat transfer is investigated as one example of the many assuming a constant slag-layer thickness,[39] or assuming
applications of this useful modeling tool. a constant slag viscosity. [41,42] The highly temperature-
dependent slag viscosity has been modeled with a simple
II. PREVIOUS WORK inverse function of temperature [38] or with an Arrhenius
equation,[37,39,43] by fitting the low viscosities (usually less
Many mathematical models have been developed of the than 10 Pa?s) measured at high temperatures and then extra-
continuous casting process, which are partly summarized in polating to lower temperatures. Even the best interface
previous literature reviews.[13,14,15] Many continuous-casting models generally oversimplify the shell and/or the mold.
686—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

Thus, there is a need for a comprehensive model of the shell,
mold, and gap, which is fast and easy to run, for use in both
research and steel plant environments.
III. MODEL FORMULATION

The model in this work computes 1-D transient heat flow
through the solidifying steel shell, coupled with 2-D steady-
state heat conduction within the mold wall. Superheat from
the liquid steel was incorporated as a heat source at the steel
solid/liquid interface. The model features a detailed treatment
of the interfacial gap, including mass and momentum
balances on the liquid and solid slag layers, friction between
the slag and mold, and slag-layer fracture. The model
simulates axial (z) behavior down a chosen position on the
mold perimeter. Wide-face, narrow-face, and even corner
simulations can, thus, be conducted separately.
A. Superheat Delivery
Before it can solidify, the steel must first cool from its
initial pour temperature to the liquidus temperature. Due
to turbulent convection in the liquid pool, this “superheat” Fig. 2—Model of solidifying steel shell domain showing typical isotherms
contained in the liquid is not distributed uniformly. A small and heat flux conditions.
database of results from a 3-D fluid-flow model[11] is used to
determine the heat flux (qsh) delivered to the solid/liquid inter-
face due to the superheat dissipation, as a function of distance included in the effective specific heat (Cp*steel ), explained in
below the meniscus. The initial condition of the liquid steel
Section IV–C.
at the meniscus is then simply the liquidus temperature.
This equation assumes that axial (z) heat conduction is
Previous work[11] found that this “superheat flux” varies
negligible in the steel, which is reasonable past the top
linearly with the superheat temperature difference and also
10 mm, due to the large advection component as indicated
is almost directly proportional to casting speed. The super-
by the large Péclet number: Pe 5 Vc Z mold rsteel Cpsteel/ksteel 5
heat-flux (qsh) function in the closest database case is adjusted
0.0167 3 0.81 3 7400 3 670/30 5 2236. The simulation
to correspond with the current superheat temperature
domain for this portion of the model is a slice through the
difference (DTsup ) and casting speed (Vc) as follows:
liquid steel and solid shell, which moves downward at the
DTsup Vc casting speed, as pictured in Figures 2 and A-1, together with
qsh 5 q0sh [2] typical interface conditions. At the internal solid/liquid steel
DT 0sup V 0c interface, the superheat flux delivered from the turbulent
where q 0sh is the superheat-flux profile from the database case, liquid pool (Section III–A) is imposed as a source term. From
with conditions of a superheat temperature difference of DT 0sup the external surface of the shell, interfacial heat flux (qint) is
and a casting speed of V 0c. Further adjustments are made to lost to the gap, which depends on the mold and slag-layer
translate the heat-flux peak to account for differences in nozzle computations, described in the following two sections.
configuration between the current conditions and the database. Appendix A provides the explicit finite-difference solution
Examples of the superheat-flux function are included in of Eq. [3], including both of these boundary conditions.
Figure 2, which represents results for a typical bifurcated,
downward-directed nozzle. The influence of this function is
insignificant to shell growth over most of the wide face, where C. Heat Transfer Across the Interfacial Gap
the superheat flux is small and contact with the mold is good. Heat transfer across the interfacial gap governs the heat
flux leaving the steel, qint , to enter the mold. To calculate
this at every position down the mold, the model evaluates
B. Heat Conduction in the Solidifying Steel Shell
an effective heat-transfer coefficient (hgap ) between the
The temperature in the thin solidifying steel shell is surface temperature of the steel shell (Ts) and the hot face
governed by the 1-D transient heat-conduction equation, of the mold wall (Tmold ):
which becomes the following on applying the chain rule to
the temperature-dependent conductivity: qint 5 hgap ( Ts 2 Tmold ) [4]
T 2 T k steel T 2 dair dsolid

rsteel Cp*steel 5 k steel 1 [3] hgap 5 1 rcontact 1 1
t x 2
T x kair ksolid
[5]
Temperature-dependent properties are given in Appen- dliquid deff
11 1 1 1 hrad
dix D.[44,45] Both the sensible and latent heat of steel are kliquid keff

Heat conduction depends on the thermal resistances of The oscillation marks are assumed to be filled with either
four different layers of materials contained in the gap: the slag, air, or a mixture, depending on the local shell-surface
oscillation marks, liquid slag, solid slag, and a possible air temperature. This governs the value of kmark.
gap. These depend on the time-averaged thickness profiles Except for perhaps a microscopically thin glassy surface
down the mold of the different layers and their corresponding layer, experience has shown that the cold slag layer adjacent
thermal conductivities. The model for gap heat conduction to the mold wall is usually crystalline[48,49] and, thus, opaque.
is illustrated in Figures 3 and 6. The most important resis- Thus, radiation occurs only across the semitransparent hot
tances are usually the slag layers, whose thicknesses are glassy or liquid layer above Tfsol,[50,51] according to Eq. [9],
calculated as described in the next section. The latent heat as shown in Figure 3:
evolved by liquid-slag solidification is less than 3 pct of the
heat transferred across the gap, so it is neglected in this m2 s T sK
2 2
1 TfsolK TsK 1 TfsolK
model. h rad 5 (Tmold , Tfsol ) [9]
The equivalent air gap (dair) is specified as input data and 1 1
0.75a d liquid 1 d eff 1 1 21
includes contact resistances[46] at the slag/shell and slag/mold «slag «steel
interfaces. It may also include a gap due to shrinkage of the where m is the slag refractive index; TsK and TfsolK are Ts
steel shell, which can be calculated using a separate thermal- and Tfsol expressed in Kelvin, respectively; a is the average
stress model.[12] The shrinkage gap is affected by the mold absorption coefficient of the slag, assuming graybody
taper and also by mold distortion, which can be calculated radiation ( «slag 5 0.9). If the liquid slag runs out, so
by another model.[10] This gap is important when simulating that Ts , Tfsol , then h rad 5 0. In the unlikely event that
down positions near the corner. Tmold $ Tfsol , «slag would be replaced by «mold , and Tfsol by
All nonuniformities in the flatness of the shell surface Tmold in Eq. [9]. Jenkins showed that this simple equation
are incorporated into the model through the prescribed to characterize radiation with absorption across a gap
oscillation-mark depth (dmark) and width (L mark), as pictured (Eq. [9]) is accurate to within 10 pct relative to a full
in Figure 4. Assuming shallow triangle-shaped marks or multiview factor analysis including radiation-conduction.[52]
depressions, d osc is the volume-averaged oscillation-mark This is sufficiently accurate, because the radiation compo-
depth: nent itself usually contributes only on the order of 10 pct
of the gap heat transfer.
0.5 Lmark dmark
dosc 5 [6]
Lpitch
D. Mass and Momentum Balance on Powder
where Lpitch is the ratio of the casting speed to the oscillation Slag Layers
frequency. Slag is assumed to flow down the gap as two distinct layers:
The oscillation marks affect heat transfer in two different solid and liquid. The solid layer is assumed to move at a
ways. First, the oscillation marks consume mold slag, so they time-averaged velocity (Vsolid ) which is always between zero
affect the slag-layer thicknesses, as described in Section III–F. and the casting speed, according to the input solid-slag speed
Second, they reduce heat conduction by effectively providing ratio ( fv).
an extra gap. This extra gap is represented by d eff, calculated
based on a heat balance[47] which produces the same total heat
flux as found by combining the heat fluxes across the two par- Vsolid 5 fv Vc [10]
allel conduction paths (at or adjacent to the oscillation mark),
averaged spatially (z direction) using an appropriate weighted
average.
0.5 L mark d mark

d eff 5
d mark kgap [7]
(L pitch 2L mark ) 1 1 0.5 1 L mark
d liquid 1 d solid k mark
d solid k liquid
kgap 5 dliquid 1d solid 11 1 h rad_liquid [8]
k solid d liquid
Fig. 3—Thermal resistances used in the interface model. Fig. 4—Model treatment of oscillation marks.

The downward-velocity profile across the liquid slag combining Eqs. [5] and [15]. The transport of slag by the
layer is governed by the simplified Navier–Stokes equation, oscillation marks depends on the lubrication state, dis-
assuming laminar Couette flow: cussed next.
Three different regions are distinguished down the mold,
Vz according to the lubrication condition. Close to the meniscus,
m 5 (r steel 2 r slag) g [11]
x x a solid slag rim exists against the mold wall. Its thickness
profile must be specified, as it depends on transient phenomena
A small body force opposing flow down the wide face not yet in the model.
gap is created by the difference between the ferrostatic The second region, shown in Figure 6, allows the solid
pressure from the liquid steel (rsteel g), transmitted through slag layer to move at the slow time-averaged velocity Vsolid .
the solid steel shell, and the average weight of the slag It always also includes oscillation marks filled with molten
(rslag g) The time-averaged velocity of the liquid slag (Vz), slag and a continuous liquid slag layer, which remains
described by Eq. [11], is subjected to boundary conditions
constraining it to the casting speed on its hot side and to the
solid-slag velocity (Vsolid ) on its cold side.
The viscosity of the molten slag (m(T)) is assumed to vary
exponentially with temperature:
n
To 2 Tfsol
m 5 mo [12]
T 2 Tfsol
where the parameters Tfsol and n are chosen empirically to fit
measured data and mo is the viscosity measured at the
reference temperature, To, usually chosen to be 1300 °C.
A typical curve obtained with this function is shown in
Figure 5 together with the measured viscosities by Lanyi that
it was fit to match.[53] Mold slags in service absorb some
Al2O3 from the steel, which changes their properties, includ-
ing decreasing the solidification temperature.[53,54] The second
curve in Figure 5 was constructed for a reported solidification
temperature of 1045 °C and viscosity at 1300 °C of 1.1 Poise
and was used later in model calibration.
By approximating the temperature across the gap to vary
Fig. 5—Comparison of model mold slag viscosity curves and measure-
linearly, Eqs. [10] through [12] can be solved for the time- ments.[53]
averaged velocity distribution across the slag layers, which
is illustrated in Figure 6. Integrating across the liquid region
yields an average velocity for the liquid layer (Vliquid ) of
(r slag 2 r steel ) gd2liquid Vc 1 Vsolid (n 1 1)

V liquid 5 2
1 [13]
m s (n 1 2) (n 1 3) (n 1 2)
where ms is the slag viscosity at the liquid layer/steel shell

interface. A mass balance was imposed to match the
measured (known) powder consumption (Q slag (kg/m2 ))
with the total molten slag flow rate past every location
down the interfacial gap, neglecting the carbon-content
component, which burns off. This consumption rate is
expressed as mass of slag per unit area of the strand surface,
which can be found from the consumption per mass of
product (Mslag (kg/ton)):
W3N
Q slag (kg/m2) 5 M slag ( kg /ton) 3 rsteel 3 [14]
2 (W 1 N)
where W is the slab width and N is the slab thickness. Slag

can be carried downward by the solid layer, the liquid layer,
and in the oscillations marks:
Qslag 3 Vc
5 Vsolid d solid 1 Vliquid d liquid 1 Vc dosc [15]
rslag
The liquid- and solid-layer thicknesses are obtained

by solving a fourth-order polynomial equation, found by Fig. 6—Velocity and temperature profiles assumed across interfacial gap.

present so long as the outer-surface temperature of the steel coefficient between the mold cold face and the cooling water
(Ts9) exceeds the slag solidification temperature, Tfsol : (hfin) incorporates heat flow to both the root and sides of the
water channels, the latter treated as heat-transfer fins.
deff
Ts¢ 5 Ts 2 q int 3 [16]
k mark h w wch Ö 2h w k mold (L c h 2wch ) 2
2h w d ch
h fin 5 1 tanh [20]
L ch L ch km old (L ch 2 wch)
Slag in the oscillation marks remains liquid longer, due
to the higher local shell temperature at their roots (Ts). Once where the mold-geometry parameters Lch, wch, and dch are
the oscillation-mark roots cool below the slag solidification shown in Figure 7(b). The heat-transfer coefficient between
temperature, however, the slag entrapped in them solidifies. the water and the sides of the water channel (hw) is calculated
This defines the third region, which consists of totally solid assuming turbulent flow through an equivalent-diameter pipe
slag, moving downward at the uniform speed. The oscillation using the empirical correlation of Sleicher and Reusse,[57]
marks no longer transport slag and, so, become filled with which is reported[58] to be more accurate than other relations
air. The transition between the second and third regions is such as that of Dittus and Boelter:[59]
gradual.
kwater
It is important to emphasize that this model represents hw 5 (5 1 0.015 Re cwaterf
1 c2
Pr waterw ) [21]
steady, time-averaged behavior only. To investigate tran- D
sient phenomena, a transient version of this model is being where D 5 2wch d ch /(wch 1 d ch ) is the equivalent diameter
developed to calculate stress inside the slag layer based of the water channel, and c1 5 0.88 2 0.24/(4 1 Pr waterw ) and
on a force balance with friction, which is described c2 5 0.333 1 0.5e20.6 Pr are empirical constants.
waterw
elsewhere.[55]
E. Heat Conduction in the Mold

A 2-D, steady-state temperature within a rectangular
vertical section through the upper portion of the mold is
calculated assuming constant conductivity:
2T 2T
1 50 [17]
x2 z2
This equation is solved using a standard Fourier-series
product solution,[56] applying a fixed heat flux of q int 5
2kmold T/x, and convection, hwater, and Twater as boundary
conditions, as shown in Figure 7(a). This copper domain is
generally chosen to extend from the top of the mold to
100 mm below the meniscus. Below this meniscus region,
heat flow is 1-D through the thickness. The temperature at
the copper hot face (Thotc ) is then
1 d mold
Thotc 5 Twater 1 q int 1 [18]
h water k mold
where d mold is the copper mold thickness calculated in
Appendix B. Coating layers are incorporated as needed to
find the mold hot-face temperature (Tmold ) by adding extra
dcoat/kcoat resistances to Eq. [18] as needed. In addition to the
heat flux across the interface (qint ), this calculation requires
the initial cooling-water temperature (Twater) input as a
boundary condition, and the effective water heat-transfer
coefficient (hwater), discussed as follows.
F. Convection to the Cooling Water

The effective heat-transfer coefficient between the cooling
water and the cold face (“water side”) of the mold is
calculated to include a possible resistance due to scale
deposits on the surface of the cooling-water channels:
d scale 1
h water 5 1 1 [19]
k scale h fin
To account for the complex nature of heat flow in the Fig. 7—Simulation domain in mold (a) vertical section along casting
undiscretized width direction of the mold, the heat-transfer direction and (b) horizontal section through mold.

The presence of the water slots can either enhance or to 8.7 W/m2K everywhere. Larger values can be input for
diminish the heat transfer, relative to a tube mold with uniform h conv to reflect the stronger convection when there is air mist
thickness (dmold ) such as used in billet casting. Deep, closely in the cooling zone. Heat extraction into the rolls is calcu-
spaced slots augment the heat-transfer coefficient (hfin is lated based on the fraction of heat extraction to the rolls
larger than hw), while shallow, widely spaced slots inhibit ( froll ), which is calibrated for each spray zone:
heat transfer. In most molds, hfin and hw are very close.
Although it slightly underpredicts the mold temperature, f roll
h roll 5 [(hrad_spray 1 h conv 1 h spray )
Eq. [20] was shown, through comparison with many 3-D L roll contact (1 2 f roll )
computations for a variety of typical slab-casting mold L spray 1 (h rad_spray 1 h conv ) (L spray pitch 2 L spray 2 L roll contact )]
geometries and conditions, to match the temperature within [24]
1 pct at the water-slot root and from 0.1 to 6 pct at the hot
face.[47,60] For a typical hot-face temperature of 190 °C and A typical froll value of 0.05 produces local temperature drops
water temperature of 30 °C, it gives maximum errors of 2 °C beneath the rolls of about 100 °C. Beyond the spray zones,
and 10 °C. It is most accurate for molds with either deep, closely heat transfer simplifies to radiation and natural convection.
spaced slots[47] or very wide slots,[60] where the cold-face
temperature is most nearly constant, as assumed in Eq. [20]. H. Solution Methodology
The model requires simultaneous solution of three different
G. Spray zones below the mold systems of equations: 1-D transient heat conduction and solid-
ification of the steel shell, 2-D steady-state heat conduction
Below the mold, heat flux from the strand surface varies in the mold, and the equations balancing heat, mass, and
greatly between each pair of support rolls, according to the momentum in the gap. The simulation starts by setting the ini-
spray-nozzle cooling, based on water flux (hspray) ; radiation tial steel and mold temperatures to the pouring temperature
(h rad_spray ); natural convection (hconv ); and heat conduction and inlet cooling-water temperature, respectively. Phase-
to the rolls (h roll ), as shown in Figure 8. Incorporating these transformation temperatures and phase-fraction temperature
phenomena enables the model to simulate heat transfer during curves are then calculated, using one of the methods described
the entire continuous casting process. The heat extraction in the next section. Then, each time step begins by rearranging
due to water sprays is a function of water flow[61] of the and solving Eqs. [5] and 15 simultaneously for dliquid and dsolid ,
following form: based on heat and mass balance at the previous time step. The
heat flux qint is then calculated according to Eqs. [4] and [5],
h spray 5 A 3 Q cwater 3 (1 2 b 3 Tspray ) [22]
which is the boundary condition for both steel and mold
where Qwater (L/m2 s) is water flux in the spray zones, and domains. The heat-transfer coefficient hwater is calculated accord-
Tspray is the temperature of the cooling-water spray. In ing to cooling-channel conditions with Eqs. [19] through [21]
Nozaki’s empirical correlation,[62] A 5 0.3925, c 5 0.55, and is used to obtain mold temperatures. Applying the superheat
and b 5 0.0075, which has been used successfully by other flux-boundary condition (Eq. [2]) as an internal heat source
at the steel solid/liquid interface, the model uses an explicit,
modelers.[61,63]
central finite-difference algorithm originally developed by
Radiation is calculated by
Pehlke[64] to solve Eq. [3] for the shell temperature at each
2
h rad_spray 5 s 3 « steel (TsK 1 Tamb K ) (T sK 1 T amb K2 ) [23] time step (Appendix A). This limits the maximum time-step
size (Dt). When a node temperature drops below the liquidus
where Tamb is the ambient temperature and TsK and TambK are temperature, its solid fraction is calculated from the latent
Ts and Tamb expressed in Kelvin. Natural convection is heat evolved, and then the node temperature is adjusted[65]
treated as a constant input for every spray zone. For water (Eq. [A6]) according to the phase fraction–temperature curves,
cooling only, it is not very important, so it was simplified described in Section IV–A. The results are used as initial
conditions for the 2-D mold calculation, which solves Eq. [17]
analytically, relating the distance down the mold (z) to the time
in the shell through the casting speed. Subsequently, the entire
1-D shell-solidification model in the 2-D mold region is recom-
puted using the new 2-D mold temperatures as its boundary
condition. This stepwise coupling procedure alternates between
models until the 1-D mold temperatures converge to match
the 2-D results within 3 °C. This produces a self-consistent
prediction, which is stable for all coupled simulations inves-
tigated and usually converges in three to four iterations. Figure 9
gives a flow chart of the whole procedure.
The model has been incorporated into a user-friendly
FORTRAN program, CONID. [66] A 100-second-long
simulation with a 0.004-second time step and 100-node mesh
runs on a PENTIUM* III personal computer (using 3.1 Mbytes
*
PENTIUM is a trademark of the Intel Corporation, Santa Clara, CA.
Fig. 8—Schematic of spray zone region. of memory) in about 30 seconds.

IV. STEEL PROPERTIES a 0.044 pct C, 0.022 pct Mn, 0.006 pct S, 0.01 pct P, 0.009 pct
Si, and 0.049 pct Al plain-carbon steel, the equilibrium-phase-
The program includes several different choices for steel diagram model calculates Tliq 5 1528 °C and Tsol 5 1509 °C,
properties, including simple constants input by the user. By while with a 10 °C/s cooling rate, the segregation model
default, the liquidus temperature, solidus temperature, phase- gives Tliq 5 1532 °C and Tsol 5 1510 °C. Figure 10 shows the
fraction curve, thermal conductivity, specific heat, and thermal solid-fraction temperature curve in the mushy zone, obtained
linear expansion are all calculated as functions of composition from both models. Both models produce similar results. The
and temperature. The steel density (rsteel), latent heat (Lf), and surprising finding that the equilibrium model produces slightly
steel-surface emissivity («steel) are constants. For carbon steel, lower transformation temperatures shows that differences in
rsteel 5 7400 kg/m3, Lf 5 271 kJ/kg, and «steel 5 0.8. the coefficients which define the alloy-dependent equilibrium
lines are more important than the nonequilibrium effects due
A. Phase Fraction to segregation at the typical cooling rates, dendrite-arm spacing,
By default, equilibrium lever-rule calculations are performed and compositions considered.
on an Fe-C phase diagram, whose phase-field lines are specified
as simple linear functions of alloy content (including the B. Thermal Conductivity of Steel
influences of Si, Cr, Mn, Ni, Mo, Cu, Ti, P, S, Al, V, N, Nb, The thermal conductivity of carbon steel is calculated
and W) reported by Kagawa and Okamoto[67] in order to as a function of temperature, carbon content, and phase
calculate the steel liquidus, solidus, and peritectic temperature fraction, which was fitted from measured data compiled
and phase fractions. Alternatively, the user can choose a by K. Harste.[44] The specific functions are listed in Appen-
nonequilibrium microsegregation model to find these values, dix D. The stainless steel thermal conductivity is calculated
based on an analytical Clyne–Kurz-style equation developed according to a fitted equation based on measured data
by Won and Thomas,[68] which was extended in this work to compiled by Pehlke.[45] Figure 11 compares some typical
include the effects of 14 elements, given in Appendix C. For
Fig. 10—Phase fraction variation with temperature in mushy zone.
Fig. 9—Flow chart of CON1D program with 1-D steel solidification model
and 2-D mold heat conduction model. Fig. 11—Comparison of model thermal conductivities and measurements.[45]

plain-carbon steel, austenitic stainless steel, and ferritic calculations in order to enforce constant mass in the fixed-
stainless steel conductivities. The thermal conductivity of domain computation.
the liquid is not artificially increased, as is common in other Alternatively, the user may input a thermal linear-expansion
models, because the effect of liquid convection is accounted coefficient (a), so that
for in the superheat-flux function, which is calculated by
models which fully incorporate the effects of turbulent flow. TLE 5 a( T2Tsol ) [27]
This is done for stainless steel, where the value of a is
C. Effective Specific Heat of Steel taken from Pehlke.[45]
The specific heat is calculated as a function of temperature,
carbon content, phase fraction and steel grade. Appendix D V. MODEL VALIDATION
gives the specific-heat functions for carbon steel, found by
differentiating the enthalpy curve from K. Harste.[44] Refer The internal consistency and accuracy of the various com-
to Pehlke[45] for the specific-heat functions of stainless steel. ponents of this model have been verified through extensive
When the steel temperature is between the solidus and comparison with analytical solutions. The accuracy of the
liquidus temperatures, the latent heat is evolved using the 2-D mold heat-transfer model at the meniscus region was
liquid-phase-fraction curve found previously. The effective evaluated by comparison with full 3-D finite-element model
specific heat is then defined as computations on separate occasions using ABAQUS [69] and
with an in-house code.[70] In both cases, the CON1D model
dH dfs predictions matched within the uncertainties associated with
Cp* 5 5 Cp 2 L f [25]
dT dT mesh refinement of the 3-D model. The fin heat-transfer
equation was compared with 3-D model computations by
Figure 12 shows the specific-heat curve of AISI 1026 carbon Thomas et al.[71] and Langeneckert,[60] as already discussed.
steel using the microsegregation model, compared with Its accuracy is acceptable except near thermocouples located
measured data.[45] The curves for other alloys, such as those in a region of complex heat flow. Its accuracy there can be
used later, are similar except for within the mushy region. So improved by incorporating an “offset” distance, as discussed
long as it properly matches the total latent heat, its exact shape elsewhere.[26,60] Other obvious checks include ensuring that
has little effect on the shell growth or surface temperature. the temperature predictions match at the transition between
at 2-D and 1-D regions, which also indicates when heat
D. Thermal Linear Expansion of Steel flow is 1-D.
By default, the thermal linear expansion (TLE) needed The solidification model is verified here through com-
for shrinkage and ideal taper calculations is computed as a parison with an analytical solution for 1-D heat conduction
function of steel density: with phase change.[72] This solution assumes a constant shell-
surface temperature and constant steel properties. Table I
3 r0 lists the constants used in both the analytical solution and
TLE 5 21 [26] the CON1D validation case, which are chosen for typical
r(T)
conditions expected in practice. The difference between the
where r0 5 rsteel. The composition and temperature-dependent steel liquidus and solidus temperatures is only 0.1 °C to
steel-density function for carbon steel ( r(T)) is taken from approximate the single melting temperature assumed in
measurements tabulated by Harste[44] and is listed in Appen- analytical solution, which is set to the mean of Tliq and Tsol
dix D. A constant density ( rsteel) is adopted for the heat-flow used in CON1D. The pour temperature is set to the liquidus
because the superheat is neglected in the analytical solution.
For the CON1D model, the time-step size Dt is 0.004 seconds
and the node spacing is 0.5 mm.
Figure 13 compares results from the analytical solution
and CON1D model for (a) the temperature distribution
through the shell at different times and (b) the growth
of the shell thickness with time. The results show that the
predictions of the CON1D model are very accurate, so the
same time step and mesh size are used in the following cases.
Table I. Constants Used in Analytical Solution and

Validation Case
Conductivity, k steel 30 W/mK

Specific Heat, Cpsteel 670 J/kgK
Latent Heat, L f 271 kJ/kg
Density, rsteel 7400 kg/m3
Melting Temperature, Tmelt 1509.05 °C
Liquidus Temperature, Tliq 1509.1 °C
Solidus Temperature, Tsol 1509 °C
Shell Surface Temperature, Ts 1000 °C
Fig. 12—Comparison of model specific heat curve and measurements.[45]

Specifically, adjustments can be made to the velocity of
the solid slag layer, the value of the contact resistances
down the mold, and even the thermal properties of the mold
slag. Other influential input parameters include the average
powder consumption rate and the average oscillation-mark
depth and width.
In a slab caster with properly designed taper, there should
not be any air gap due to shrinkage down the center of the
wide face. This is because ferrostatic pressure pushes the long,
wide, weak shell against the mold to maintain as close a con-
tact as possible. This greatly simplifies model calibration when
simulating a slice through the wide face of the mold and shell.
The following sections report on the calibration, validation,
and results of simulations performed for two sets of condi-
tions given in Tables II and V. Input parameters for the
standard case, Table II, were calibrated to match the casting
conditions of the 0.225 3 1.78 m slabs of low-carbon steel
cast at LTV Steel (Cleveland, OH), where mold thermocouple
temperatures, cooling-water temperature-rise, and breakout-
shell measurements were available.[71,79] The steel compo-
sition is 0.044 pct C, 0.022 pct Mn, 0.006 pct S, 0.01 pct
P, 0.009 pct Si, and 0.049 pct Al.
A. Mold Cooling-Water Temperature Rise

The first step in model calibration is to match the total heat
extracted in the mold with the measured temperature increase
of the mold cooling water. The average rate of heat extracted
from the mold per unit surface area (Q) is found from
Vc
Q5 qint Dt [28]
Zmold mold
Fig. 13—Comparison of model results and analytical solution: (a) shell

This heat-transfer rate should equal the temperature
temperature distribution and (b) shell growth. increase of the mold cooling water (DTwater) flowing through
the “hot” channels, located adjacent to the slab width area:
qint Lch Vc D t
DTwater_hot channels 5 [29]
VI. MODEL CALIBRATION mold rwaterCpwater Vwaterwch d ch
Having shown the model to be internally consistent, it This equation assumes that the cooling-water slots have
cannot be used quantitatively until it is calibrated to match locally uniform rectangular dimensions, wch and d ch , and
measurements on the specific operating caster of interest. spacing, Lch. Heat entering the hot face (between two water
This step is necessary because so many of the inputs to the channels) is assumed to pass straight through the mold to
model are uncertain. heat the water flowing through the cooling channels.
To date, the model has been calibrated to match many To compare with the measured water-temperature increase,
different casters, including slabs at BHP LPD (Whyalla, the a forementioned prediction is modified as follows to
South Australia), LTV Steel (Cleveland, OH),[73] AK Steel account for missing slots due to bolts or water slots, or slots
(Mansfield, OH),[26] Allegheny Ludlum (Brackenridge, PA),[74] that are beyond the slab width and, so, do not participate in
Columbus Stainless Steel (Middleburg, South Africa),[70] Siderar heat extraction:
(Buenos Aires, Argentina),[85] and China Steel (Taiwan, Repub-
lic of China); thin slabs at Nucor (Crawfordsville, IN)[75] and wch 3 dch 3 W/L ch
DTwater_total channel 5 DTwater_hot channels [30]
POSCO (Seoul, South Korea);[76] blooms at BHP RBPD total channel area
(Newcastle, New South Wales);[77] and billets at POSCO Using reported slag properties and consumption rate
(Pohang, South Korea).[78] In order to calibrate the model, it (Table II), the heat flux was calibrated to match the mea-
is simply run several times, using trial and error, to find val- sured temperature rise of 7.1 °C by adjusting the solid-slag
ues of the model parameters that allow the model predictions speed ratio, fv, to 0.175. The corresponding temperature rise
to match all of the known measurements. Those measurements in just the hot channels is predicted to be 7.5 °C.
can include the cooling-water temperature rise, the time-
averaged temperature of any thermocouples embedded in the
B. Mold Temperatures
mold, the thickness profile of breakout shells, the thickness
of solidified-mold powder layers and slag rims, and the tem- The next step in calibration of CON1D is to further
perature histories of any thermocouples embedded in the strand. adjust the model parameters to match the measurements

of thermocouples embedded in the walls of the operating should not be confused with the location of the peak mold
casting mold. This step is very constrained, however, as temperature, which is usually about 35 mm below the heat-
every change that causes a local increase in heat flux must flux peak (55 mm below the meniscus, in this case). Assuming
be balanced by a corresponding decrease elsewhere, in no air gap in the interface for this wide-face simulation, the
order to maintain the balance with the cooling water contact resistances and scale thicknesses are other adjustable
already achieved. input conditions to match the mold thermocouple measure-
In this example (Table II), the slag-rim shape in region I ments. Here, a 0.02 mm scale layer was assumed for the top
was chosen to decrease linearly from 0.8 mm at the meniscus 305 mm, where specially designed inserts had been installed
to 0.5 mm at 15 mm below the metal level, which is near to to increase the local cooling-water velocity,[79] and a 0.01 mm
the position of peak heat flux. The peak heat-flux position scale layer was assumed for the bottom remainder of the mold.
These thicknesses are in accordance with plant observations
that the hot region had a thicker scale layer.[80]
Table II. Standard Input Conditions (Case 1) Figure 14 compares the predicted and measured tempera-
Carbon content, C pct 0.044 pct tures at several locations down the LTV mold. The thermo-
Liquidus temperature, Tliq 1529 °C couples were all 18.8 mm below the mold hot face. The
Solidus temperature, Tsol 1509 °C agreement indicates the calibration of the model for these typ-
Steel density, r steel 7400 kg/m2 ical casting conditions. This figure also shows the predicted
Steel emissivity, « steel 0.8 — hot-face and cold-face temperature profiles. The sharp change
Fraction solid for shell in temperature is due to a sudden increase in water-channel
thickness location, fs 0.1 — depth, produced by experimental inserts used in the trial.[79]
Mold thickness at top (outer Note that the observed scale layer greatly increased the mold
face, including water channel) 56.8 mm temperature, especially in the hot portion that contained the
Mold outer face radius, Ro 11.985 m insert. Based on this insight, steps were taken to improve
Total mold length, Z moldtotal 900 mm
water quality to prevent this scale and improve mold life.[79]
Total mold width 1876 mm
Scale thickness at mold cold
face (inserts region/below), dscale 0.02/0.01 mm C. Shell Thickness
Initial cooling water temperature,
Twater
Having calibrated the model, the predicted shell-thickness
30 °C
Water channel geometry, profile is compared with measurements down a breakout shell
dch 3 wch 3 Lch 25 3 5 3 29 mm3 that occurred under very similar castings conditions, as given
Cooling water velocity, Vwater 7.8 m/s in Figure 15. Shell thickness is defined in the model by inter-
Mold conductivity, k mold 315 W/mK polating the position between the liquidus and solidus isotherms
Mold emissivity, « mold 0.5 — with the temperature corresponding to the specified solid
Mold powder solidification fraction ( fs), according to the phase-fraction–temperature rela-
temperature, Tfsol 1045 °C tionship in Figure 10. In this sample case, fs 5 0.1, which is
Mold powder conductivity, the only adjustable parameter remaining for model calibration.
ksolid /k liquid 1.5/1.5 W/mK This is reasonable, as interdendritic liquid is held by surface
Air conductivity, kair 0.06 W/mK tension during draining of the breakout.
Slag layer/mold resistance, rcontact 5.0 3 1029 m2K/W To compare the predicted steady shell thickness with that
Mold powder viscosity at of a breakout shell, a correction is needed to account for the
1300 °C, m1300 1.1 Poise
Exponent for temperature
solidification time that occurred while the liquid metal was
dependence of viscosity, n 0.85 — draining during the breakout. Thus, the time in the steady
Slag density, rslag 2500 kg/m3
Slag absorption factor, a 250 m21
Slag refractive index, m 1.5 —
Slag emissivity, « slag 0.9 —
Mold powder
consumption rate, Q slag 0.6 kg/m2
Empirical solid slag
layer speed factor, fv 0.175 —
Casting speed, Vc 1.07 m/min
Pour temperature, T pour 1550 °C
Slab geometry, W 3 N 1780 3 225 mm 3 mm
Nozzle submergence
depth, d nozzle 265 mm
Working mold length, Z mold 810 mm
Oscillation mark geometry,
d mark 3 w mark 0.45 3 4.5 mm 3 mm
Mold oscillation
frequency 84 cpm
Oscillation stroke 10 mm
Time-step, dt 0.004 s
Mesh size, dx 0.5 mm
Fig. 14—Comparison of CON1D predicted and measured mold temperature.

simulation corresponds to distance down the breakout shell Variation in the superheat flux is critical to shell growth
according to the relation down the narrow face and off-corner regions, where problems
z such as inadequate taper sometimes produce significant
t5 1 td [31] air gap(s). Together, the large superheat combined with
Vc decreased heat transfer across the interfacial gap can reduce
where the “drainage time” (td) is the time for the metal level shell growth. This was the subject of a significant study
to drop from the meniscus to the breakout slice of interest using the model, which was reported elsewhere.[82]
(z). The drainage time is calculated based on the Bernoulli
equation and a mass balance:[81] D. Powder-Layer Thickness
Ö Zb 2 Ö Zb 2 z The model predicts the thickness and velocity profiles
td 5 [32] expected in the powder layers in the interfacial gap. For
pd b2 g example, Figure 16 shows the solid and liquid slag-layer
CD
4NW 2 thickness profiles expected for the standard conditions inves-
tigated here (Table II). It shows that the liquid slag layer runs
where the drainage coefficient is CD 5 1. For the present case,
out at 380 mm below the meniscus, where the liquid slag-
the position of the breakout hole from the meniscus is Zb 5
layer/steel-shell interface temperature drops below the slag
1.524 m; the slab thickness is N 5 0.225 m; and the slab width
solidification temperature of 1045 °C, as shown in Figure 17.
is W 5 1.78 m. Assuming that steel flow to the mold was shut
The total slag thickness continues to increase while there
off simultaneously as the metal level started to drop below
is still liquid coming from the oscillation marks. This is
the meniscus and that the breakout hole diameter (db) began
indicated in Figure 17, where the shell-surface temperature
at 50 mm and linearly grew to 90 mm by the time all liquid
at the oscillation-mark roots still exceeds 1045 °C at mold
steel had drained, a transient shell profile can be calculated.
exit. Although no reliable slag samples were obtained from
Figure 15 gives the predicted shell thickness at both steady-
this caster, these slag-thickness predictions of 0.5 to 1.5 mm
state and transient conditions, compared with the breakout-
are consistent with samples measured at similar plants.[47,76]
shell measurements. The generally close match with the
transient predictions tends to validate the model. The under-
predicted shell thickness near the meniscus is likely due to a E. Shell-Surface Temperature
short interval of increased liquid flow into the mold after the Typical model predictions of the surface temperature in
breakout started and before level control and flow were shut the mold are shown in Figure 17 for standard conditions.
off. This would have allowed the liquid level to move down- When the liquid slag layer runs out at 380 mm below the
ward with the top of the breakout shell for a short time inter- meniscus (Figure 16), the liquid entrapped in oscillation
val (not included in the model), thus providing additional marks flows out and air fills in, which increases the resistance
solidification time at the very top of the breakout shell. This of the oscillation mark, so the temperature difference between
effect is commonly observed in breakout shells. the oscillation mark’s root and peak increases also, as shown
Growth of the shell naturally depends on both the inter- in Figure 17.
facial and superheat fluxes. The superheat distribution is After existing the mold, the slab surface quickly reheats,
important to the narrow face, as Figure 2 shows that the two and then it fluctuates greatly as it travels through the spray
curves are of the same magnitude low in the mold, where zones. Heat is extracted rapidly during contact with the
the hot molten steel jet impinges against the solidifying shell. support rolls and when passing the impingement zone of the
Figure 15 shows the shell thinning of the narrow face due cooling water from the spray nozzles, which each cause great
to this jet-impingement effect. temporary drops in surface temperature.
Fig. 15—Comparison of CON1D predicted and measured shell thickness. Fig. 16—Predicted slag layer thickness profiles.

Lacking accurate spray and roll contact heat-transfer coef- Internal temperature histories measured at three places
ficients, calibration of temperature predictions below the mold beneath the surface are included in Figure 18. Both surface
can be calibrated by adjusting the model parameters froll and thermocouples needed about 500 mm to heat to their sur-
the spray coefficients (Table III) to match measurements such rounding shell temperatures and later suffered from internal
as roll cooling-water heat-extraction rate and thermocouple debonding, so their results are reliable only between 500 and
temperatures embedded in the strand. An example of such 3000 mm. The centerline thermocouple needed almost 2 m
calibration is shown in Figure 18 for casting conditions to heat, and appears to be accurate within 10 °C. Both the
measured at the China Steel #1 slab caster in Taiwan, Republic internal temperatures and the amplitude of their wiggles
of China, given in Tables III and IV. The temperature are roughly matched, indicating the degree of calibration.
measurements were achieved by feeding a block containing Temperature fluctuations at the thermocouple location are
several thermocouples into the mold just before “tail-out” at quite small, compared with the surface, which varies over
the end of casting. The thermocouple tips extending through 100 °C over a single roll pitch. Near the top of the caster,
the bottom of the block were soon frozen into the strand. The the greatest surface-temperature drop occurs beneath each
last several meters of steel before the end of the cast ensured spray jet, while a tiny dip occurs at each small region of
that the recorded temperature histories would be typical, while direct contact with a support roll. Lower in the caster, the
allowing the insulated tube of thermocouple wires extending relative size of the dips becomes closer, with deep sharp
from the top of the block to follow the strand through the caster drops caused by the high local heat-extraction rate during
with minimal damage. The distance of each thermocouple from roll contact under high ferrostatic pressure.
the surface was measured after sectioning the final product. Optical pyrometers are also useful for model calibration,[61]
but are adversely affected by intermittent changes in surface-
scale emissivity and steam density from evaporating spray
water, so are most accurate when located below the spray
chamber. Attaching thermocouples directly to the strand
surface is another difficult experimental method that can
be used for model calibration.[19]
VI. SAMPLE APPLICATIONS

The calibrated model has many applications for both design
and operation of continuous casting machines. First, it can help
to investigate the effect of various process conditions on the
fundamentals of mold heat transfer. Most parameters, such as
oscillation practice, powder type, casting speed, and steel grade,
affect heat transfer in several different ways, which can only
be isolated and quantified independently using a model.
The model can make predictions of potential quality
problems, which have more relevance in practice than simple
heat transfer. For example, a warning of possible boiling in
Fig. 17—Predicted shell surface temperature. the cooling-water channels is issued when the mold surface
temperature exceeds the pressure-dependent water boiling
Table III. Input Conditions for Submold Calibration temperature. The model is currently being extended to make
(China Steel Case) other warnings such as breakout danger from excessive shell
Carbon content, C pct 0.45 pct

Mold thickness at top
(including water channel) 51 mm
Initial cooling water
temperature, Twater 35 °C
Water channel geometry,
dch 3 wch 3 Lch 25 3 5 3 28 mm3
Cooling water velocity, Vwater 7.62 m/s
Casting speed, Vc 0.55 m/min
Pour temperature, Tpour 1510 °C
Slab geometry, W 3 N 1560 3 270 mm 3 mm
Nozzle submergence Depth, d nozzle 200 mm
Working mold length, Z mold 600 mm
Mold oscillation frequency 120 cpm
Oscillation stroke 4 mm
Cooling water temperature
in spray zones, Tspray 35 °C
Spray length of spray zone
nozzle, L spray 0.05 m
Spray zone roll contact angle 7 deg
Fig. 18—Shell temperature (China Steel Case).

Table IV. Spray Zone Variables (China Steel Case)
Zone Starts at Roll in Zone Roll Radius Water Flow Rate Spray Width
Zone (mm) (mm) (mm) (L/min/row) (m) froll
1 600.0 2 70 27.5 1.3 0.05
2 906.9 5 70 14.86 1.2 0.05
3 1840.8 5 100 14.86 1.2 0.05
4 3034.3 5 125 11.84 1.2 0.2
5 4520.5 10 150 8.8 1.2 0.2
6 7977.9 10 175 7.15 1.2 0.2
7 11883.1 11 210 2.5 1.56 0.2
8 17050.7 18 240 0.0 1.56 0.2
26440.7 end of last spray zone
Table V. Parametric Study Conditions
Casting Speed Oscillation Frequency Total Consumption Lubrication Consumption Oscil0 lation Mark Size
Case Vc (m/min) (cpm) Rate Qslag (kg/m2) Rate Q lub (kg/m2) dmark 3 wmark (mm2)
1 1.07 84 0.60 0.4 0.45 3 4.5
2 1.07 84 0.56 0.36 0.45 3 4.5
3 1.25 98 0.56 0.4 0.4 3 4.0
thinning at the mold exit, solid slag-layer fracture from Thus, the total consumption rate of slag (Qslag ) depends
excessive mold friction and the accompanying heat-flux greatly on the oscillation-mark shape, while lubrication depends
variations, and crack formation. Finally, the model should mainly on Qlub , and mold heat transfer depends on both.
predict optimum casting conditions to avoid problems, To investigate the effect of mold powder consumption
whenever possible. Initial features of the model toward this rate, an intermediate case of a standard (low) casting speed
goal include a prediction of ideal mold taper. Together with with decreased consumption rate is also included. The three
other resources, CON1D is a powerful tool to investigate the cases in this study are listed in Table V, with other conditions
cause and prevention of quality problems and to investigate given in Table II.
potential design and operation improvements prior to costly Figure 19 presents the heat-flux profiles down the mold
experimental implementation. wide face, calculated for all three cases. Decreasing the
powder consumption rate at a constant casting speed (case
2) is seen to increase heat flux in the top portion of the mold,
A. Parametric Studies: Effect of Casting Speed relative to standard conditions (case 1). This is because the
As an example to illustrate the use of the model to under- average thickness of the slag layers decreases, thus lowering
stand fundamental phenomena in the mold, simulations were the interfacial resistance. This effect diminishes with distance
performed to investigate just two of the many interdependent down the mold (as the importance of interfacial resistance
parameters: casting speed and mold powder consumption. to heat transfer decreases relative to that from increasing
It is well known that increasing the casting speed causes steel shell thickness).
changes to other parameters, such as decreased mold powder The practical case of increasing the casting speed and simul-
consumption rate and shallower oscillation marks. To taneously decreasing the total powder consumption rate and
investigate the effect of increasing casting speed in a typical oscillation-mark depth (case 3) also increases heat flux toward
real caster, the oscillation frequency was increased pro- the bottom of the mold. This is due to the lower thermal resist-
portionally with speed, according to plant practice, and the ance of a thinner steel shell produced with less solidification
oscillation-mark depth was decreased, such that the negative time, which becomes increasingly important with distance
strip ratio and the lubrication consumption rate remained down the mold. The net result of increasing the casting speed
constant. The lubrication consumption rate (Qlub ) is a useful (comparing case 3 with case 1) is to increase heat flux almost
concept for comparing different powder consumption rates. uniformly down the mold. This is reflected in uniformly
It is introduced here as the rate of slag consumption, neglect- higher mold temperatures, as seen in the model predictions
ing the slag carried in the oscillation marks: in Figure 20. This prediction also matches mold thermocouple
measurements obtained for case 3 conditions, as included in
Qlub 5 Qslag 2 Qosc [33] Figure 20. The higher speed leads to a thinner steel shell and
higher steel-surface temperature, so the liquid slag layer per-
Oscillation marks filled with slag and moving at the sists further down the mold, as shown in Figures 21 through
casting speed consume slag at the following rate (Q osc ): 23, respectively. But, the higher heat flux for higher casting
speed also lowers the shell-surface temperature, which partially
0.5rslag 3 dmark 3 wmark cancels the effect of higher temperature due to the thinner
Qosc 5 [34]
Lpitch shell. For these cases, the surface temperatures at the oscillation

Fig. 19—Effect of casting speed and powder consumption on the heat Fig. 21—Effect of casting speed on shell thickness.
flux profile.
Fig. 20—Effect of casting speed on mold temperature. Fig. 22—Effect of casting speed on steel shell temperature.
mark’s root near the mold exit are almost the same, as shown
in Figures 22 and 24. Figure 24 compares the shell-temperature
profiles at the mold exit.
The model is suited to many further fundamental parametric
studies of this kind. For example, the steel grade affects the
average oscillation-mark size, powder consumption rate, air-
gap size due to thermal contraction (narrow face), and steel
strength. Mold powder properties and oscillation practice have
similar interdependent effects. The effect of oscillation-mark
depth, for example, is quantified in a model application
reported elsewhere.[73]
B. Boiling Prediction
The model issues a warning that boiling is possible, if the
mold cold-face temperature exceeds the boiling temperature
for the given operating pressure in the cooling-water
channels:[83]
Boiling if: Tcold ( C) . 100 (P (MPa)/0.10135)0.27 [35] Fig. 23—Effect of casting speed on slag layer thickness.

E. Crack-Formation Analysis
As with previous continuous casting models, CON1D can
be used to locate where defects are formed. For example,
by accurately predicting the shell thickness existing the mold,
the model can identify whether a subsurface crack formed
in or below the mold. This can be difficult to tell, particu-
larly near the narrow face, where shell growth is slower.
Here, a crack forming below the mold might appear to have
formed in the mold, without an accurate calculation of shell
growth that incorporates superheat delivery. The model can
also simulate phenomena below the mold, such as reheat-
ing of the shell surface, which can lead to surface cracks.
Submold bulging and crack formation requires accurate tem-
perature variation between rolls, so the model is useful for
designing spray water-cooling systems.
Fig. 24—Effect of casting speed on shell temperature profile at mold exit. F. Calculation of Ideal Mold Taper
The narrow face of the mold should be tapered to match
the shrinkage of the steel shell, which is cooling against
the wide face. Previous work has determined that this shrink-
Boiling in the water channels changes the rate of heat age depends mainly on the surface temperature of the shell
removal and causes temperature fluctuations that together and the steel grade.[12] The model predicts an ideal average
pose a serious potential quality problem. Figure 14 shows taper, by dividing the thermal strain («) by distance down
that boiling is, indeed, possible for the conditions investigated the mold (instantaneous taper) or by the mold length (total
here. This is due to the 0.02-mm-thick layer of scale on the taper per meter). Thermal-shrinkage strain is estimated here
mold cold face near the meniscus, which raises the mold- in two different ways; first, for «th1, by
face temperature , 70 °C. On the other hand, adding a
«th1 5 TLE(Tsol )2TLE(Ts) [36]
0.5-mm-thick protective Ni coating to the hot face is
predicted to have only a minimal effect on heat flux and Another method to calculate shrinkage was developed
cold-face temperature. The CON1D model is ideal for by Dippenaar et al.[34,84] The strain «th2 is computed by
quantifying effects such as these. summing the average thermal linear expansion of the solid
portion of the shell between each pair of consecutive time
C. Breakout Analysis steps:
The model can be used to help understand how a breakout t solid nodes
may have arisen. Sticker breakouts are easily identified by 1
«th2 5 TLE(Tit )2TLE(Tit 1 Dt) [37]
their characteristic effect on mold thermocouple histories. t5 0 i i5 1
Other breakouts, such as those caused by inadequate taper,
can be more difficult to identify. For example, the model Here, TLE is the thermal linear-expansion function for
could be used to determine whether a given narrow-face the given steel grade, calculated form weighted averages of
breakout was more likely caused by excessive superheat the phases present.
resulting from a clogged nozzle, or from an insufficient For the sample cases, the higher speed causes a hotter
mold taper, causing an excessive gap. Either condition shell with less shrinkage, shown in Figure 25, so it needs a
could produce a narrow-face shell that is too hot and thin slightly less-narrow-face mold taper. The shrinkage « th1,
to have the hot strength needed to avoid rupture. Further based on surface temperature only, is generally less than
calibration may allow the model to accurately warn of a «th2 and is almost independent of casting speed, due to the
potential breakout when shell growth is predicted to fall cancellation effect discussed in Section VI–A. With a linear
below a critical value. Initial work toward this end is taper, the narrow-face shell attempts to shrink away from
reported elsewhere.[30] the upper portion of the mold, while it pushes against the
lower portion of the mold. To match the shrinkage, it is clear
that the taper should be increased high in the mold and
D. Lubrication Prediction
decreased lower down. Mold distortion, viscoplastic creep
The model is being extended to predict the consequences of of the steel, and other factors should also be taken into
interfacial heat transfer on mold friction and steel quality.[55] If account when designing a nonlinear mold taper. These
the mold slag, which fills most of the gap, is allowed to cool calculations require sophisticated thermal-stress models to
completely below its crystallization temperature, then it becomes calculate temperatures, stresses, and shrinkage, including the
viscous and is less able to lubricate the strand. This may increase formation of an air gap near the corners, and its effect on
mold friction, cause the solid slag layer to fracture, and lead to heat flow across the mold/shell interface. The calibrated
transient temperature changes, making problems such as surface CON1D model is currently being used to provide calibrated
cracks more likely. Figure 16 suggests that this might occur heat-transfer data to these models to evaluate and improve
below 400 mm, for the present conditions. taper optimization.

mold taper, and potential quality problems such as complete
slag solidification and boiling in the water channels. It has
many potential applications.
ACKNOWLEDGMENTS
The authors thank former students Bryant Ho, Guowei Li,
and Ying Shang for their work on early versions of the
CON1D program and to the Continuous Casting Consortium,
University of Illinois, and the National Science Foundation
(Grant Nos. MSS-89567195 and DMI-01-15486) for fund-
ing, which made this work possible. Some 3-D computations
for validation were performed at the National Center for
Supercomputing Applications, UIUC. Special thanks go to
Bill Emling and others, LTV Steel, and to Kuan-Ju Lin and
others, China Steel, for collecting the operating data and
experimental measurements used in model validation.
Fig. 25—Effect of casting speed on shell shrinkage.
NOMENCLATURE
G. Future Applications Cp specific heat (J/kgK)
d depth/thickness (m)
The model is based on conservation laws that must
hold, regardless of the complex phenomena present in the caster. db diameter of the breakout hole (m)
However, there are many more unknowns than equations. Thus, dosc volume-averaged oscillation-
the model requires extensive calibration, which includes the mark depth (mm)
values of many parameters not currently known. Preferably, froll fraction of heat flow per spray zone
some of the required input data should be predicted, such as going to roll (—)
fs solid steel fraction (—)
powder consumption rate and oscillation-mark size.
Much further work is needed before the model can realize fv empirical solid slag layer speed factor (—)
its full potential as a predictive tool for design, improvement, g gravity (9.81 m/s2)
and control of continuous casting operations. For example, h heat-transfer coefficient (W/m2K)
the model simulates only time-averaged behavior, while, in h conv natural convection h in spray zones (W/m2K)
reality, many phenomena, especially involving the slag layer, hrad_spray radiation h in spray zones (W/m2K)
vary greatly during each oscillation cycle. This requires a
hrad radiation h in slag layers (W/m2K)
detailed, transient treatment. When and how the solid slag k thermal conductivity (W/mK)
layer slides along the mold wall, the accompanying friction L length (m)
forces, and if and where the solid slag fractures are other Lf latent heat of steel (kJ/kg)
important issues. Below the mold, fundamental measurements L pitch distance between successive
of spray-zone heat transfer are needed. This work will require oscillation marks (m)
n exponent for temperature dependence
advanced 3-D model-strand calculations, in addition to
extensive calibration. of slag viscosity (—)
N slab thickness (m)
Prwaterw Prandtl number of water at mold cold
VII. CONCLUSIONS face temperature (Cpm /k)
Q average mold heat flux (kW/m2)
A simple but comprehensive heat-flow model of the Q slag mold slag consumption (kg/m2)
continuous slab-casting mold, gap, and shell has been devel- Q water water flow rate in spray zones (L/m2s)
oped. It simulates 1-D solidification of the steel shell and qint shell/mold interface heat flux (kW/m2)
features the dissipation of superheat, movement of the solid qsh superheat flux (kW/m2)
and liquid slag layers in the interfacial gap, and 2-D heat Rewaterf Reynolds number at average of mold cold
conduction within the copper mold wall. The model accounts face and cooling water
for the effects of oscillation marks on both heat transfer temperatures (DVr /m)
and powder consumption. It also accounts for variations in rcontact slag/mold contact resistance (m2K/W)
water-slot geometry and steel grade. It is user friendly and t time (s)
runs quickly on a personal computer. It has been validated td drainage time (s)
through numerical comparisons and calibrated with mea- T temperature (°C)
surements on operating casters, including cooling water tem- Tf sol mold slag solidification temperature (°C)
perature rise, mold thermocouple temperatures, breakout Thotc mold copper hot face temperature (°C)
shell thickness, slag layer thickness, and thermocouples Tmold mold hot face temperature with coating (°C)
embedded in the steel shell. In addition to heat transfer, the Tliq steel liquid temperature (°C)
model predicts thickness of the solidified slag layers, ideal Tsol steel solidus temperature (°C)

Ts9 steel shell surface temperature
(at oscillation mark root) (°C)
Ts liquid slag layer hot-side temperature (°C)
DTwater cooling water temperature rise (°C)
TLE thermal linear expansion (—)
Vc casting speed (m/s)
w width (m)
W slab width (m)
x shell thickness direction (m)
z casting-direction, distance
below meniscus (m)
Z mold working mold length (m)
a thermal linear expansion
coefficient (K21)
s Stefan Boltzman constant
(5.67 3 1028 W/m2K4)
« surface emissivities (—)
«th thermal strain of steel shell (pct)
r density (kg/m3) Fig. A.1—Simulation domain in shell.
m viscosity (Pa s)
Subscripts
steel steel slab (2) Interior nodes:
a, d, g, l a-Fe, d-Fe, g-Fe, liquid
steel phases Dt k
Tinew 5 Ti 1 (Ti21 22Ti 1 Ti 1 1)
mold copper mold Dx2rCp*
[A4]
coat mold coating layer Dt k
water cooling water (bulk) 1 2 * (Ti1 1 2Ti21 )2
4Dx rCp T
ch cooling water channel in mold
scale scale layer in mold (3) Shell surface node (with heat flux boundary):
cooling channel
mark oscillation mark 2Dt k
Tnnew 5 Tn 1 (Tn21 2Tn)
eff effective oscillation mark Dx2 rCp* [A5]
(based on heat balance)
2
air air gap Dt k qint 2Dt qint
gap shell/mold gap 1 2
rCp* T k DxrCp*
slag mold slag
solid, liquid solid slag layer, liquid slag layer The effect of superheat is included by adjusting Eq. [A4]
spray spray nozzle below mold for the first interior node with a temperature below the
liquidus temperature:
APPENDIX A Dt
Tinew 5 Tinew 1 qsh [A6]
rCp* dx
FDM solution of steel solidification model
Figure A1 shows the simulation domain in the solidifying where dx 5 Dx for interior nodes, and dx 5 Dx/2 for bound-
steel, which is a slice through the liquid steel and solid shell ary nodes.
and moves down at casting speed, Vc. Applying the bound- The preceding equations are solved at each time-step (Tinew )
ary conditions, based on properties evaluated at the previous step (Ti). This
simple explicit scheme is usually acceptable because property
T changes are generally gradual with temperature (Figures 11
centerline 50 [A1] and 12). However, the effective specific heat has a sudden
x
jump when temperature drops below the liquidus temperature.
T To improve accuracy, and allow a larger time-step, a postit-
k steel (Ts ) steel surface 5 2qint [A2] eration correction is applied to each node after the time-step
x
when it first drops below the liquidus temperature. Specifi-
cally, its temperature is increased to match the solid fraction
Equation [3] is solved at each time-step using the follow-
that should have been achieved, based on converting the
ing explicit central finite difference discretization:
sensible heat extracted from that node into latent heat, accord-
(1) Centerline liquid node (adiabatic boundary): ing to the solid fraction curve T( fs) defined previously:
2Dt k (Tliq 2 Tinew ) C p* (Ti)

T1new 5 T1 1 (T2 2 T1) [A3] Tinew 5 T fs 5 [A7]
Dx2 rCp* Lf

APPENDIX B APPENDIX C
Mold thickness Equilibrium partition coefficient, diffusion coefficient, and
For a curved mold, the mold thickness, dmold varies with liquidus line slopes of the solute element
distance down the mold, which is calculated for the outer The microsegregation model used here[68] evaluates den-
and inner radius mold faces separately: drite arm spacing and phase fraction evolution as a function
of cooling rate, CR , and steel composition, CO . Liquidus,
1 solidus, and peritectic temperatures depend on steel com-
d outer outer
mold 5 d moldo 1 RO2 2 (Z mold_total )2 position as follows:
4
[B1]
1 Tliq 5 Tpure 2 mi C0,i [C1]
2Ö R2o 2 (Z mold_total )2 2 Z mold_total
2 i
4
T sol 5 T pure 2 m i CL ,i C 0,i,k i,D i,C R , fs 5 1 [C2]
1 i
d inner inner
mold 5 d moldo 2 R I2 2 (Z mold_total )2
4
[B2] T d/g 5 Tpure
d/g
2 ni kdi /L CL,i
d
[C3]
i
1
1 RI2 2 (Z mold_total )2 2 Z 2mold_total
4 Extended data needed for this model are listed in the
following table and include the partition coefficients, k, and
where d moldo is the mold thickness at the top of the mold, diffusion coefficients, D, for each phase, and the slopes of
Zmold_total is the total mold length (sum of working mold length the equilibrium liquidus, m, and the slopes of Ar4 lines, n,
Zmold and distance of meniscus from top of the mold Zmen), for the pseudobinary alloy of each element with iron.
and RO, and RI are mold outer and inner radius of curvature, The results are not very sensitive to cooling rate, as the accom-
respectively. panying dendrite arm spacing change tends to compensate.
Element kd/L kg/L Dd (cm2/s) Dg (cm2/s) m (°C/pct) n (°C/pct)
C 0.19 0.34 0.0127 exp (219,450/RT) 0.0761 exp (232,160/RT) 78.0 21122
Si 0.77 0.52 8.0 exp (259,500/RT) 0.3 exp (260,100/RT) 7.6 60
Mn 0.76 0.78 0.76 exp (253,640/RT) 0.055 exp (259,600/RT) 4.9 212
P 0.23 0.13 2.9 exp (255,000/RT) 0.01 exp (243,700/RT) 34.4 140
S 0.05 0.035 4.56 exp (251,300/RT) 2.4 exp (253,400/RT) 38.0 160
Cr 0.95 0.86 2.4 exp (257,310/RT) 0.0012 exp (252,340/RT) 1.04 13.4
Ni 0.83 0.95 1.6 exp (257,360/RT) 0.34 exp (267,490/RT) 4.69 228.6
Cu 0.53 0.88 2.6 exp (257,360/RT) 0.7 exp (268,350/RT) 5.32 210.4
Mo 0.80 0.585 3.47 exp (257,690/RT) 0.068 exp (259,000/RT) 2.6 77.6
Ti 0.38 0.33 3.15 exp (259,200/RT) 0.15 exp (259,980/RT) 10.24 120.5
V 0.93 0.63 4.8 exp (257,360/RT) 0.284 exp (261,900/RT) 12.95 85.5
Nb 0.40 0.22 50.2 exp (260,220/RT) 0.83 exp (263,690/RT) 10.24 70.8
W 0.85 0.45 1.57 exp (258,200/RT) 0.13 exp (257,300/RT) 0.24 18.8
N 0.25 0.48 0.008 exp (218,900/RT) 0.91 exp (240,270/RT) 60.0 —
Note: R is the gas constant of 1.987 cal/mol K, and T is the temperature in Kelvin.
APPENDIX D
Carbon steel thermal properties functions
(1) Thermal conductivities
k 5 ka fa 1 kd fd 1 kg fg 1 kl fl
where
ka 5 (80.9129.9269 3 1022 T 1 4.613 3 1025 T 2) (1 2 a1 (C pct)a2)
kd 5 (20.14 2 9.313 3 1023 T ) (1 2 a1 (C pct)a2 ) [D1]
23
kg 5 21.6 2 8.35 3 10 T
k l 5 39.0
a1 5 0.425 2 4.385 3 1024 T
a2 5 0.209 1 1.09 3 1023 T

(2) Specific heat
Cp 5 Cpa fa 1 Cpd fd 1 Cpg fg 1 Cpl fl
where
504.8146 2 0.1311139T (K) 2 5.1875834 3 106T (K)22
T (K) # 800
1 4.486659 3 1024T (K)2
24720.324 1 4.583364T (K) 1 1.109483 3 109T (K)22 800 , T (K) # 1000
Cap 5
211501.07 1 12.476362T (K) 1000 , T (K) # 1042 [D2]
34871.21 2 32.02658T (K) 1042 , T (K) # 1060
210068.18 1 5.98686T (K) 1 5.217657 3 109T (K)22 1060 , T (K) # 1184
Cpd 5 441.3942 1 0.17744236T (K)
Cpg 5 429.8495 1 0.1497802T (K)
Cpl 5 824.6157
(3) Density used for thermal linear expansion calculation
r(T ) 5 ra fa 1 rd fd 1 rg fg 1 rl fl
where
ra 5 7880.76 2 0.3244T 2 2.7461 3 1025T 2
C pct
rd 5 (8010.71 2 0.4724T) 11 (1 1 13.43 3 1023 pct C)23 [D3]
100 2 C pct
C pct
r 5 (8105.91 2 0.5091T ) 11 (1 1 8.317 3 1023 pct C)23
100 2 C pct
rl 5 7965.98 2 0.619T
REFERENCES 16. E.A. Mizikar: Trans. TMS-AIME 1967, vol. 239 (11), pp. 1747-58.
17. J.E. Lait, J.K. Brimacombe, and F. Weinberg: Ironmaking and Steel-
1. F.M. Najjar, B.G. Thomas, and D.E. Hershey: Metall. Mater. Trans. making, 1974, vol. 1 (2), pp. 90-97.
B, 1995, vol. 26B, pp. 749-65. 18. E.A. Upton, T.R.S. Rao, P.H. Dauby, and R.C. Knechtges: Iron Steel-
2. K.C. Mills: Steel Technol. Int., 1994, pp. 161-66. maker, 1988, vol. 15 (5), pp. 51-57.
3. R.J. O’Malley: “An Examination of Mold Flux Film Structures and 19. R. Davies, N. Blake, and P. Campbell: 4th Int. Conf. Continuous
Mold Gap Behavior Using Mold Thermal Monitoring and Petrographic Casting, Preprints, Vol. 2, Brussels, Belgium, 17-19 May, 1988, VS,
Analysis at Armco’s Mansfield Operations”, Proc. METEC Congress Dusseldorf, FRG, 1988, vol. 2, pp. 645-54.
99, (Dusseldorf, Germany, 13-15 June, 1999), Verein Deutscher Eisen- 20. R.B. Mahapatra, J.K. Brimacombe, I.V. Samarasekera, N. Walker,
huttenleute, Dusseldorf, FRG, 1999. E.A. Paterson, and J.D. Young: Metall. Trans. B, 1991, vol. 22B,
4. R.J. O’Malley: 82nd Steelmaking Conf., Chicago, IL, 21-24 Mar., 1999, pp. 861-74.
ISS/AIME, Warrendale, PA, 1999, vol. 82, pp. 13-33. 21. J.K. Brimacombe: Can. Metall. Q., 1976, vol. 15 (2), pp. 163-75.
5. A. Yamauchi, K. Sorimachi, T. Sakuraya and T. Fujii: Iron Steel Inst. 22. M.R. Aboutalebi, R.I.L. Guthrie, and M. Hasan: Steel Res., 1994,
Jen. Int., 1993, vol. 33 (1), pp. 140-47. vol. 65 (6), pp. 225-33.
6. J. Cho, H. Shibata, T. Emi, and M. Suzuki: Iron Steel Inst. Jpn. Int., 23. F. Wimmer, H. Thone, and B. Lindorfer: Berg. Huttenmannische
1998, vol. 38 (5), pp. 440-46. Monatshefte, 1996, vol. 141 (5), pp. 185-91.
7. M.S. Bhamra, M.G. Charlesworth, S. Wong, D. Sawyers-Villers, and 24. H. Nam, H.S. Park, and J.K. Yoon: Iron Steel Inst. Jpn. Int., 2000,
A.W. Cramb: 54th Electric Furnace Conf., Dallas, TX, 9-12 Dec., vol. 40 (9), pp. 886-92.
1996, ISS/AIME, Warrendale, PA, 1996, vol. 54, pp. 551-64. 25. S.K. Choudhary and D. Mazumdar: Iron Steel Inst. Jpn. Int., 1994,
8. C. Orrling, A.W. Cramb, A. Tilliander, and Y. Kashiwaya: Iron Steel- vol. 34 (7), pp. 584-92.
maker, 2000, vol. 27 (1), pp. 53-63. 26. B.G. Thomas, R.J. O’Malley, T. Shi, Y. Meng, D. Creech, and D.
9. Y. Kashiwaya, C.E. Cicutti, and A.W. Cramb: Iron Steel Inst. Jpn. Stone: Modeling of Casting, Welding and Advanced Solidification
Int., 1998, vol. 38 (4), pp. 357-65. Process IX, Aachen, Germany, 20-25 Aug., 2000, TMS, Warrendale,
10. B.G. Thomas, G. Li, A. Moitra, and D. Habing: 80th Steelmaking PA, 2000, pp. 769-76.
Conf., Chicago, IL, 1997, vol. 80, pp. 183-201. 27. J.E. Lee, T.J. Yeo, K.H. Oh, J.K. Yoon, and U.S. Yoon: Metall. Mater.
11. X. Huang, B.G. Thomas, and F.M. Najjar: Metall. Mater. Trans. B Trans. A, 2000, vol. 31A, pp. 225-37.
1992, vol. 23B, pp. 339-56. 28. C. Ohler, H.J. Odenthal, H. Pfeifer, and I. Lemanowicz: Stahl Eisen.,
12. A. Moitra and B.G. Thomas: 76th Steelmaking Conf., Dallas, TX, 28-31 2002, vol. 122 (3), pp. 55-63.
Mar., 1993, ISS, Warrendale, PA, 1993, vol. 76, pp. 657-67. 29. B.G. Thomas, W.R. Storkman, and A. Moitra: IISC 6th Int. Iron and
13. B.G. Thomas: 1991 Steelmaking Conf., ISS, Warrendale, PA, 1991, Steel Congr., Nagoya, Japan, 21-26 Oct., 1990, ISIJ, Tokyo, Japan,
pp. 69-82. 1990, vol 3, Steelmaking I, pp. 348-55.
14. B.G. Thomas and L. Zhang: Iron Steel Inst. Jpn. Int. 2001, vol. 41 (10), 30. C. Li and B.G. Thomas: Brimacombe Memorial Symp., Vancouver,
pp. 1181-93. BC, Canada, 1-4 Oct., 2000, CIMMP, Montreal, PQ, Canada, 2000,
15. B.G. Thomas: Metall. Mater. Trans. B, 2002, vol. 33B, pp. 795-812. pp. 595-611.

31. B.G. Thomas, A. Moitra, and R. McDavid: Iron Steelmarker, 1996, 60. M.M. Langeneckert: Master’s Thesis, University of Illinois, Urbana,
vol. 23 (4), pp. 57-70. IL, 2001.
32. J.K. Park, I.V. Samarasekera, B.G. Thomas, and U.S. Yoon: 83rd 61. J.K. Brimacombe, P.K. Agarwal, S. Hibbins, B. Prabhaker, and L.A.
Steelmaking Conf., Pittsburgh, PA, 26-29 Mar., 2000, ISS/AIME, Baptista: Continuous Casting, ISS/AIME, Warrendale, PA, 1984, vol. 2,
Warrendale, PA, 2000, vol. 83, pp. 9-21. pp. 109-23.
33. C.A.M. Pinheiro, I.V. Samarasekera, and B.N. Walker: Ironmaking 62. T. Nozaki: Iron Steel Inst. Jpn. Trans., 1978, vol. 18, pp. 330-38.
and Steelmaking, 2000, vol. 27 (1), pp. 37-54. 63. R.A. Hardin, H. Shen, and C. Beckermann: Modelling of Casting,
34. S. Chandra, J.K. Brimacombe, and I.V. Samarasekera: Ironmaking and Welding and Advanced Solidification Processes IX, 20-25 Aug., 2000,
Steelmaking, 1993, vol. 20 (2), pp. 104-12. TMS, Warrendale, PA, 2000.
35. F. Wimmer, H. Thone, and B. Lindorfer: Int. Conf. on Modelling and 64. R.D. Pehlke: Enginneering Summer Conf.: Continuous Casting of Steel,
Simulation in Metallurgical Engineering and Materials Science, Beijing, Univ. of Michigan, Ann Arbor, MI, 1980, Section VII, pp. 1-17.
China, 11-13 June, 1996, Metall. Industry Press, Beijing, China, 1996, 65. V.R. Voller, C.R. Swaminathan, and B.G. Thomas: Int. J. Num. Methods
pp. 366-71. in Eng., 1990, vol. 30, pp. 875-98.
36. N. Tiedje and E.W. Langer: Scand. J. Metall., 1992, vol. 21 (5), pp. 211-17. 66. B.G. Thomas and Y. Meng: CON1D Users Manual, University of Illinois,
37. R. Bommaraju and E. Saad: 73th Steelmaking Conf., Detroit, MI, 25-28, Urbana, IL, 2002.
Mar., 1990, ISS, Warrendale, PA, 1990, vol. 73, pp. 281-96. 67. A. Kagawa and T. Okamoto: Mater. Sci. Technol., 1986, vol. 2 (10),
38. J.A. DiLellio and G.W. Young: Metall. Mater. Trans. B 1995, vol. 26B, pp. 997-1008.
pp. 1225-1441. 68. Y.M. Won and B.G. Thomas: Metall. Mater. Trans. A, 2001, vol. 32A,
39. J.F. Chavez, A. Celaya, M.A. Barron, and R.D. Morales: 79th Conf. pp. 1755-67.
of the Steelmaking Division of the Iron and Steel Society, Pittsburgh, 69. G. Li: Ph.D. Thesis, University of Illinois, Urbana, IL, 1996.
PA, 24-27 Mar., 1996, ISS/AIME, Warrendale, PA, 1996, vol. 79, 70. B.G. Thomas and T. Morthland: “3-D Heat Transfer Analysis of Colum-
pp. 321-29. bus Slab Casting Mold,” CCC Report 2001, Univ. of Ill., Urbana, IL.
40. G.J.W. Kor: Continuous Casting of Steel, 2nd Process Technology 71. B.G. Thomas, B. Ho, and G. Li: Alex McLean Symp.: Process
Conf., Chicago, II, 23-25 Feb., 1981, ISS/AIME, Warrendale, PA, Fundamentals, Liquid Metal Processing for Cleanliness, Novel and
1981, vol. 2, pp. 124-32. Conventional Casting, and Novel Process Technologies, Toronto, ON,
41. K. Nakajima: Curr. Adv. Mater. Processing 1992, vol. 5 (4), pp. 1221-24. Canada, 1998, pp. 177-93.
42. T. Cimarelli: Metal Italiana, 1997, vol. 89 (9), pp. 31-37. 72. D.R. Poirier and G.H. Geiger: Transport Phenomena in Materials
43. D.R. Bland: IMA J. Appl Math., 1984, vol. 32, pp. 89-112. Processing, TMS, Warrendaler, PA, 1994, p. 339.
44. K. Harste: Ph.D. Dissertation, Technical University of Clausthal, 73. B.G. Thomas, D. Lui, and B. Ho: Sensors and Modeling in Mate-
Clausthal, 1989. rials Processing: Techniques and Applications, Orlando, FL, 1997,
45. R.D. Pehlke, A. Jeyarajan, and H. Wada: Report No. NSFMEA82028, pp. 117-42.
University of Michigan, Ann Arbor, MI, 1982. 74. M.M. Collur: “Use of CONID Model to Determine if the Casting Speed
46. H. Shibata, K. Kondo, and M. Suzuki: Iron Steel Inst. Jpn. Int., 1996, can be Increased for Chrome-Nickel Grades at 8-3 Caster,” Report,
vol. 35, Suppl., pp. S179-S182. Allegheny Ludlum Corporation Technical Center, Brackenridge, PA,
47. B. Ho: Master’s Thesis, University of Chicago, Urbana, IL, 1992. 1996.
48. K. Ichikawa, A. Morita, and Y. Kawabe: Shinagawa Techn. Rep., 1993, 75. J. Paidipati, B. Sharos, C. Slayton, and J. Wood: Report No. 6: ME280
vol. 36, pp. 99-107. Final Report, University of Illinois, Urbana, IL, 2002.
49. K. Watanabe, M. Suzuki, K. Murakami, H. Kondo, A. Miyamoto, 76. J.K. Park: Ph.D. Thesis, University of British Columbia, Vancouver,
and T. Shiomi: NKK Techn. Rev., 1997, vol. 77, pp. 20-26. BC, Canada, 2002.
50. R. Taylor and K.C. Mills: Ironmaking and Steelmaking, 1988, 77. B.G. Thomas: 79th Conf. of the Steelmaking Division of the Iron and
vol. 15 (4), pp. 187-94. Steel Society, Pittsburgh, PA, 24-27 Mar., 1996, ISS/AIME, Warrendale,
51. A.C. Mikrovas, S.A. Argyropoulos, and I.D. Sommerville: Iron Steel- PA, 1996, pp. 519-31.
maker, 1991, vol. 18 (12), pp. 51-61. 78. J.K. Park, B.G. Thomas, and I.V. Samarasekera: Ironmaking and
52. M.S. Jenkins: in Continuous Casting, vol. 9, Initial Solidification and Steelmaking, 2002, vol. 29(5), pp. 359-75.
Strand Surface Quality of Peritectic Steels, ISS-AIME, Warrendale, 79. R. Sobolewski, S.C. Sander, J.G. Kuczma, and A.J. Rumler: Steel-
PA, 1997, pp. 239-44. making Conf. Proc., Detroit, MI, 1990, vol. 73, pp. 275-80.
53. M.D. Lanyi and C.J. Rosa: Metall. Trans. B, 1981, vol. 12B (2), 80. I.V. Samarasekera and J.K. Brimacombe: Ironmaking and Steelmak-
pp. 287-98. ing, 1982, vol. 9 (1), pp. 1-15.
54. D. Larson: Ind. Heating, 1986, vol. 53 (4), pp. 16-17. 81. A. Moitra: Ph.D Thesis, University of Illinois, Urbana-Champaign, IL,
55. Y. Meng and B.G. Thomas: Metall. Mater. Trans. B, 2003, vol. 34B, 1993.
pp. 707-25. 82. G.D. Lawson, S.C. Sander, W.H. Emling, A. Moritra, and B.G. Thomas:
56. H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids, 2nd ed., 77th Steelmaking Conf., Chicago, IL, 20-23 Mar., 1994, ISS/AIME,
Oxford University Press, Oxford, United Kingdom, 1986. Warrendale, PA, 1994, vol. 77, pp. 329-36.
57. L.C. Burmesiter: Convective Heat Transfer, 2nd ed., Wiley, New York, 83. CRC Handbook of Chemistry and Physics, R.C. Weast, ed., Chemical
NY, 1993, p. 484. Rubber Co., Cleveland, OH, 1971.
58. C.A. Sleicher and M.W. Rouse: Int. J. Heat Mass Transfer, 1975, 84. R.J. Dippenaar, I.V. Samarasekera, and J.K. Brimacombe: ISS Trans.,
vol. 18 (5), pp. 677-83. 1986, No. 7, pp. 31-43.
59. F.W. Dittus and L.M.K. Boelter: Univ. Calif. Publ. Eng., 1930, No. 2, 85. B.G. Thomas, M. Langeneckert, L. Castella, M. Dziuba, G.D. Gresia and
pp. 443-61. W. Balante: Ironmaking and Steelmaking, 2003, vol. 30 (3), pp. 1-5.

03 MTB Meng CON1D Reprint Post

Uploaded by

Copyright:

Available Formats

03 MTB Meng CON1D Reprint Post

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

03 MTB Meng CON1D Reprint Post

Uploaded by

Copyright:

Available Formats

Heat-Transfer and Solidification Model of Continuous

Slab Casting: CON1D

I. INTRODUCTION move downward, causing slag to be consumed at a rate

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—685

686—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

III. MODEL FORMULATION

­T ­2 T ­k steel ­T 2 dair dsolid

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—687

0.5 L mark d mark

688—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

(r slag 2 r steel ) gd2liquid Vc 1 Vsolid (n 1 1)

where ms is the slag viscosity at the liquid layer/steel shell

where W is the slab width and N is the slab thickness. Slag

The liquid- and solid-layer thicknesses are obtained

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—689

E. Heat Conduction in the Mold

F. Convection to the Cooling Water

690—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 8—Schematic of spray zone region. of memory) in about 30 seconds.

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—691

Fig. 10—Phase fraction variation with temperature in mushy zone.

692—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

Table I. Constants Used in Analytical Solution and

Conductivity, k steel 30 W/mK

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—693

A. Mold Cooling-Water Temperature Rise

Fig. 13—Comparison of model results and analytical solution: (a) shell

694—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—695

696—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

VI. SAMPLE APPLICATIONS

Carbon content, C pct 0.45 pct

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—697

Table V. Parametric Study Conditions

698—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—699

700—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—701

2Dt k (Tliq 2 Tinew ) C p* (Ti)

702—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

Element kd/L kg/L Dd (cm2/s) Dg (cm2/s) m (°C/pct) n (°C/pct)

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—703

704—VOLUME 34B, OCTOBER 2003 METALLURGICAL AND MATERIALS TRANSACTIONS B

METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 34B, OCTOBER 2003—705

You might also like

T 2 T k steel T 2 dair dsolid