Chapter 5

Modeling of Continuous Casting

Brian G. Thomas, Professor of Mechanical Engineering, University of Illinois

The high cost of empirical investigation in an operating steel plant makes it prudent to use all
available tools in designing, troubleshooting and optimizing the process. Physical modeling, such
as using water to simulate molten steel, enables significant insights into the flow behavior of liq-
uid steel processes. The complexity of the continuous casting process and the phenomena which
govern it, illustrated in Figs. 5.1 and 5.2, make it difficult to model. However, with the increasing

Fig. 6.1 Schematic of continuous casting process.

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 1
Casting Volume

power of computer hardware and software, mathematical modeling is becoming an important tool
to understand all aspects of the process.

Fig. 5.2 Schematic of phenomena in the mold region of a steel slab caster.

5.1 Physical Models

Previous understanding of fluid flow in continuous casting has come about mainly through exper-
iments using physical water models. This technique is a useful way to test and understand the
effects of new configurations before implementing them in the process. A full-scale model has the
important additional benefit of providing operator training and understanding.
Construction of a physical model is based on satisfying certain similitude criteria between the
model and actual process by matching both the geometry and the force balances that govern the
important phenomena of interest.1–4 Some of the forces important to flow phenomena are listed in
Table 5.1. To reproduce the molten steel flow pattern with a water model, all of the ratios between

2 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

the dominant forces must be the same in both systems. This ensures that velocity ratios between
the model and the steel process are the same at every location. Table 5.2 shows some of the impor-
tant force ratios in continuous casting flows, which define dimensionless groups. The size of a
dimensionless group indicates the relative importance of two forces. Very small or very large
groups can be ignored, but all dimensionless groups of intermediate size in the steel process must
be matched in the physical model.

Table 5.1 Forces Important to Fluid Flow Phenomena.

Inertia ρ L2 V2 L = length scale (m)

Gravity ρ g L3 V = velocity (m/s)

Buoyancy (ρ – ρp) g L3 ρ = fluid density (kg/m3)

Viscous force µLV µ = viscosity (kg/m-s)

Thermal buoyancy ρ g L3 β ∆T g = gravity accel. = 9.81 m/s2

Surface tension σL σ = surface tension (N/m)

β = thermal exp. coef. (m/m-°C)

∆T = temperature difference (°C)

p = particle of solid or gas

Table 5.2 Dimensionless Groups Important to Fluid Flow Phenomena.

Force Ratio Definition Name Phenomena

Inertial VLρ Reynolds Fluid momentum

Viscous µ

Inertial V2 Froude Gravity-driven flow; Surface waves

Gravitational gL
Inertial V2 Froude* Natural convection
Thermal buoyancy gLβ∆T
Inertial ρLV 2 Weber Bubble formation; Liquid jet atomization
Surface Tension σ
* Modified Froude number

An appropriate geometry scale and fluid must be chosen to achieve these matches. It is fortunate
that water and steel have very similar kinematic viscosities (µ/ρ). Thus, Reynolds and Froude
numbers can be matched simultaneously by constructing a full-scale water model. Satisfying these
two criteria is sufficient to achieve reasonable accuracy in modeling isothermal single-phase flow
systems, such as the continuous casting nozzle and mold, which has been done with great success.
A full-scale model has the extra benefit of easy testing of plant components and operator training.
Actually, a water model of any geometric scale produces reasonable results for most of these flow
systems, so long as the velocities in both systems are high enough to produce fully turbulent flow
and very high Reynolds numbers. Because flow through the tundish and mold nozzles are gravity
driven, the Froude number is usually satisfied in any water model of these systems where the
hydraulic heads and geometries are all scaled by the same amount.

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 3
Casting Volume

Physical models sometimes must satisfy heat similitude criteria. In physical flow models of steady
flow in ladles and tundishes, for example, thermal buoyancy is large relative to the dominant iner-
tial-driven flow, as indicated by the size of the modified Froude number (Froude* in Table 5.2),
which therefore must be kept the same in the model as in the steel system. In ladles, where veloc-
ities are difficult to estimate, it is convenient to examine the square of the Reynolds number divided
by the modified Froude number, which is called the Grashof number. Inertia is dominant in the
mold, so thermal buoyancy can be ignored there. The relative magnitude of the thermal buoyancy
forces can be matched in a full-scale hot water model, for example, by controlling temperatures
and heat losses such that β∆T is the same in both model and caster. This is not easy, however, as
the phenomena that govern heat losses depend on properties such as the fluid conductivity and spe-
cific heat and the vessel wall conductivity, which are different in the model and the steel vessel. In
other systems, such as those involving low velocities, transients or solidification, simultaneously
satisfying the many other similitude criteria important for heat transfer is virtually impossible.
When physical flow models are used to study other phenomena, other force ratios must be satisfied
in addition to those already mentioned. For the study of inclusion particle movement, for example,
it is important to match the force ratios involving inertia, drag and buoyancy. This generates several
other conditions to satisfy, such as matching the terminal flotation velocity, which is:5

g (r - rp )d p2
VT = (Eq. 5.1)
18m(1 + 0.15 Re0.687 )

where:
VT = particle terminal velocity (m/s),
ρ, ρp = liquid, particle densities (kg / m3),
dp = particle diameter (m),
µ = liquid viscosity (kg / m-s),
g = gravity accel. = 9.81 m/s2,
Re = particle Reynold’s number = ρVT dp / µ.
In a full-scale water model, for example, 2.5-mm plastic beads with a density of 998 kg/m3 might
be used to simulate 100-µm 2300 kg/m3 solid spherical inclusions in steel because they have the
same terminal flotation velocity (equation 5.1), but are easier to visualize.
Sometimes, it is not possible to match all of the important criteria simultaneously. For example, in
studying two-phase flow, such as gas injection into liquid steel, new phenomena become impor-
tant. The fluid density depends on the local gas fraction, so flow similitude requires additional
matching of the gas fraction and its distribution. The gas fraction used in the water model must be
increased in order to account for the roughly fivefold gas expansion that occurs when cold gas is
injected into hot steel. Adjustments must also be made for the local pressure, which also affects
this expansion. In addition to matching the gas fraction, the bubble size should be the same, so
force ratios involving surface tension, such as the Weber number, should also be matched. In
attempting to achieve this, it may be necessary to deviate from geometric similitude at the injec-
tion point and to wax the model surfaces to modify the contact angles, in order to control the ini-
tial bubble size. If gas momentum is important, such as for high gas injection rates, then the ratio
of the gas and liquid densities must also be the same. For this, helium in water is a reasonable
match for argon in steel. In many cases, it is extremely difficult to simultaneously match all of the
important force ratios. To the extent that this can be approximately achieved, water modeling can
reveal accurate insights into the real process.
To quantify and visualize the flow, several different methods may be used. The easiest is to inject
innocuous amounts of gas, tracer beads or dye into the flow for direct observation or photography.

4 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

Quantitative mixing studies can measure concentration profiles of other tracers, such as dye with
colorimetry measurement, salt solution with electrical conductivity, or acid with pH tracking.4 For
all of these, it is important to consider the relative densities of the tracer and the fluid. Accurate
velocity measurements, including turbulence measurements, may be obtained with hot wire
anemometry,6 high-speed videography with image analysis, particle image velocimetry (PIV)7, 8 or
laser doppler velocimetry (LDV).9 Depending on the phenomena of interest, other parameters may
be measured, such as pressure and level fluctuations on the top surface.10
As an example, Fig. 5.3 shows the flow modeled in the mold region of a continuous thin-slab
caster.11 The right side visualizes the flow using dye tracer in a full-scale physical water model.
This particular caster features a 3-port nozzle that directs some of the flow downward in order to
stabilize the flow pattern from transient fluctuations and to dissipate some of the momentum to
lessen surface turbulence. The symmetrical left side shows results from the other important analy-
sis tool: computational modeling, which is discussed in the next section.

Fig. 5.3 Flow in a thin-slab casting mold visualized using (a) K-ε computer simulation and (b) water model with dye injec-
tion. From Ref. 11.

5.2 Computational Models

In recent years, decreasing computational costs and the increasing power of commercial modeling
packages are making it easier to apply mathematical models as an additional tool to understand
complex materials processes such as the continuous casting of steel. Computational models have the
advantage of easy extension to other phenomena such as heat transfer, particle motion and two-
phase flow, which is difficult with isothermal water models. They are also capable of more faithful

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 5
Casting Volume

representation of the flow conditions experienced by the steel. For example, there is no need for the
physical bottom that interferes with the flow exiting a strand water model, and the presence of the
moving solidifying shell can be taken into account.
Models can now simulate most of the phenomena important to continuous casting, which include:
• fully-turbulent, transient fluid motion in a complex geometry (inlet nozzle
and strand liquid pool), affected by argon gas bubbles, thermal and solutal
buoyancies,
• thermodynamic reactions within and between the powder and steel phases,
• flow and heat transport within the liquid and solid flux layers, which float
on the top surface of the steel,
• dynamic motion of the free liquid surfaces and interfaces, including the
effects of surface tension, oscillation and gravity-induced waves, and flow
in several phases,
• transport of superheat through the turbulent molten steel,
• transport of solute (including intermixing during a grade change),
• transport of complex-geometry inclusions through the liquid, including the
effects of buoyancy, turbulent interactions, and possible entrapment of the
inclusions on nozzle walls, gas bubbles, solidifying steel walls, and the top
surface,
• thermal, fluid and mechanical interactions in the meniscus region between
the solidifying meniscus, solid slag rim, infiltrating molten flux, liquid steel,
powder layers and inclusion particles,
• heat transport through the solidifying steel shell, the interface between shell
and mold (which contains powder layers and growing air gaps), and the cop-
per mold,
• mass transport of powder down the gap between shell and mold,
• distortion and wear of the mold walls and support rolls,
• nucleation of solid crystals, both in the melt and against mold walls,
• solidification of the steel shell, including the growth of dendrites, grains and
microstructures, phase transformations, precipitate formation, and
microsegregation,
• shrinkage of the solidifying steel shell due to thermal contraction, phase
transformations and internal stresses,
• stress generation within the solidifying steel shell due to external forces
(mold friction, bulging between the support rolls, withdrawal, gravity),
thermal strains, creep, and plasticity (which varies with temperature, grade
and cooling rate),
• crack formation,
• coupled segregation, on both microscopic and macroscopic scales.
The staggering complexity of this process makes it impossible to model all of these phenomena
together at once. Thus, it is necessary to make reasonable assumptions and to uncouple or neglect
the less-important phenomena. Quantitative modeling requires incorporation of all of the phe-
nomena that affect the specific issue of interest, so every model needs a specific purpose. Once
the governing equations have been chosen, they are generally discretized and solved using finite-
difference or finite-element methods. It is important that adequate numerical validation be con-
ducted. Numerical errors commonly arise from too coarse a computational domain or incomplete
convergence when solving the nonlinear equations. Solving a known test problem and conduct-
ing mesh refinement studies to achieve grid independent solutions are important ways to help val-
idate the model. Finally, a model must be checked against experimental measurements on both

6 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

the laboratory and plant scales before it can be trusted to make quantitative predictions of the real
process for a parametric study.

5.2.1 Heat Transfer and Solidification

Models to predict temperature and growth of the solidifying steel shell are used for basic design,
troubleshooting and control of the continuous casting process.12 These models solve the transient
heat conduction equation,

∂ ∂ ∂ ∂T
(rH) + (rv i H) = (k eff ) +Q (Eq. 5.2)
∂t ∂x i ∂x i ∂x i

where:
∂/∂t = differentiation with respect to time (s-1),
ρ = density (kg/m3),
H = enthalpy or heat content (J/kg),
xi = coordinate direction, x, y or z (m),
vi = velocity component in xi direction (m/s),
keff = temperature-dependent effective thermal conductivity (W/m-K),
T = temperature field (K),
Q = heat sources (W/m3),
i = coordinate direction index which, when appearing twice in a term, implies the sum-
mation of all three possible terms.
An appropriate boundary condition must be provided to define heat input to every portion of the
domain boundary, in addition to an initial condition (usually fixing temperature to the pouring tem-
perature). Latent heat evolution and heat capacity are incorporated into the constitutive equation
that must also be supplied to relate temperature with enthalpy.
Axial heat conduction can be ignored in models of steel continuous casting because it is small rel-
ative to axial advection, as indicated by the small Peclet number (casting speed multiplied by shell
thickness divided by thermal diffusivity). Thus, Lagrangian models of a horizontal slice through
the strand have been employed with great success for steel.13 These models drop the second term
in equation 5.2 because velocity is zero in this reference frame. The transient term is still included,
even if the shell is withdrawn from the bottom of the mold at a casting speed that matches the
inflow of metal, so the process is assumed to operate at steady state.
Heat transfer in the mold region is controlled by:
• convection of liquid superheat to the shell surface,
• solidification (latent heat evolution in the mushy zone),
• conduction through the solid shell,
• the size and properties of the interface between the shell and the mold,
• conduction through the copper mold,
• convection to the mold-cooling water.
By far the most dominant of these is heat conduction across the interface between the surface of
the solidifying shell and the mold, although the solid shell also becomes significant lower down.
The greatest difficulty in accurate heat flow modeling is determination of the heat transfer across

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 7
Casting Volume

this gap, qgap, which varies with time and position depending on its thickness and the properties of
the gas or lubricating flux layers that fill it:

Ê k gap ˆ
q gap = Á h rad + ˜ (T0shell - T0mold ) (Eq. 5.3)
Ë dgap ¯

where:
qgap = local heat flux (W/m2),
hrad = radiation heat transfer coefficient across the gap (W/m2K),
kgap = effective thermal conductivity of the gap material (W/mK),
dgap = gap thickness (m),
T0shell = surface temperature of solidifying steel shell (K),
T0mold = hot face surface temperature of copper mold (K).
Usually, qgap is specified only as a function of distance down the mold, in order to match a given
set of mold thermocouple data.14 However, where metal shrinkage is not matched by taper of the
mold walls, an air gap can form, especially in the corners. This greatly reduces the heat flow
locally. More complex models simulate the mold, interface, and shell together, and use shrinkage
models to predict the gap size.15–17 This may allow the predictions to be more generalized.
Mold heat flow models can feature a detailed treatment of the interface.18–23 Some include heat,
mass, and momentum balances on the flux in the gap and the effect of shell surface imperfections
(oscillation marks) on heat flow and flux consumption.23 This is useful in steel slab casting opera-
tions with mold flux, for example, because hrad and kgap both drop as the flux crystallizes and must
be modeled properly in order to predict the corresponding drop in heat transfer. The coupled effect
of flow in the molten metal on delivering superheat to the inside of the shell and thereby retarding
solidification can also be modeled quantitatively.23,24 Mold heat flow models can be used to iden-
tify deviations from normal operation and thus predict quality problems such as impending break-
outs or surface depressions in time to take corrective action.
During the initial fraction of a second of solidification at the meniscus, a slight undercooling of the
liquid is required before nucleation of solid crystals can start. The nuclei rapidly grow into den-
drites, which evolve into grains and microstructures. These phenomena can be modeled using
microstructure models such as the cellular automata25 and phase field26 methods. The latter
requires coupling with the concentration field on a very small scale so is very computationally
intensive.
Below the mold, air mist and water spray cooling extract heat from the surface of the strand. With
the help of model calculations, cooling rates can be designed to avoid detrimental surface temper-
ature fluctuations. Online open-loop dynamic cooling models can be employed to control the spray
flow rates in order to ensure uniform surface cooling even during transients, such as the temporary
drop in casting speed required during a nozzle or ladle change.27
The strand core eventually becomes fully solidified when it reaches the “metallurgical length.”
Heat flow models that extend below the mold are needed for basic machine design to ensure that
the last support roll and torch cutter are positioned beyond the metallurgical length for the highest
casting speed. A heat flow model can also be used to troubleshoot defects. For example, the loca-
tion of a misaligned support roll that may be generating internal hot-tear cracks can be identified
by matching the position of the start of the crack beneath the strand surface with the location of
solidification front down the caster calculated with a calibrated model.

5.2.2 Fluid Flow Models

Mathematical models of fluid flow can be applied to many different aspects of the continuous
casting process, including ladles, tundishes, nozzles and molds.12,28 A typical model solves the

8 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

following continuity equation and Navier Stokes equations for incompressible Newtonian fluids,
which are based on conserving mass (one equation) and momentum (three equations) at every
point in a computational domain:29, 30

∂v i
=0 (Eq. 5.4)
∂x i

∂ ∂ ∂P ∂ Ê ∂v ∂u j ˆ
ru j + ru i u j = - + m eff Á i + + a(T0 - T)rg j + Fj (Eq. 5.5)
∂t ∂x i ∂x i ∂x i Ë i ∂x i ˜¯
∂ x

where:
∂/∂t = differentiation with respect to time (s-1),
ρ = density (kg/m3),
vi = velocity component in xi direction (m/s),
xi = coordinate direction, x,y, or z (m),
P = pressure field (N/m2),
µeff = effective viscosity (kg/m-s),
T = temperature field (K),
T0 = initial temperature (K),
α = thermal expansion coefficient, (m/m-K),
gj = magnitude of gravity in j direction (m/s2),
Fj = other body forces (e.g., from eletromagnetic forces),
i, j = coordinate direction indices; which when repeated in a term, implies the summa-
tion of all three possible terms.
The second-to-last term in equation 5.5 accounts for the effect of thermal convection on the flow.
The last term accounts for other body forces, such as due to the application of electromagnetic
fields. The solution of these equations yields the pressure and velocity components at every point
in the domain, which generally should be three-dimensional. At the high flow rates involved in
these processes, these models must incorporate turbulent fluid flow. The simplest yet most com-
putationally demanding way to do this is to use a fine enough grid (mesh) to capture all of the tur-
bulent eddies and their motion with time. This method, known as “direct numerical simulation,”
was used to produce the instantaneous velocity field in the mold cavity of a continuous steel slab
caster shown in Fig. 5.4.7 The 30 seconds of flow simulated to achieve these results on a 1.5 mil-
lion-node mesh required 30 days of computation on an SGI Origin 2000 supercomputer. The cal-
culations are compared with particle image velocimetry measurements of the flow in a water
model, shown on the right side of Fig. 5.4. These calculations reveal structures in the flow pattern
that are important to transient events such as the intermittent capture of inclusion particles.
To achieve more computationally-efficient results, turbulence is usually modeled on a coarser grid
using a time-averaged approximation, such as the K-ε model,31 which averages out the effect of
turbulence using an increased effective viscosity field, µeff:

K2 (Eq. 5.6)
m eff = m 0 + m t = m 0 + rCm
e

where:
µo , µt = laminar and turbulent viscosity fields (kg/m-s),

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 9
Casting Volume

Fig. 5.4 Instantaneous flow pattern in a slab casting mold comparing LES simulation (left) and PIV measurement (right).
From Ref. 7.

ρ = fluid density (kg/m3),

Cµ = empirical constant = 0.09,
K = turbulent kinetic energy field, m2/s2,
ε = turbulent dissipation field, m2/s3.
This approach requires solving two additional partial differential equations for the transport of tur-
bulent kinetic energy and its dissipation:
∂K ∂ Ê m t ∂K ˆ ∂v j Ê ∂v i ∂v j ˆ
rv j = Á ˜ + mt Á + ˜ - re (Eq. 5.7)
∂x j ∂x j Ë s K ∂x j ¯ ∂x i Ë ∂x j ∂x i ¯

10 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

∂e ∂ Ê m t ∂e ˆ e ∂v j Ê ∂v i ∂v j ˆ e
rv j = Á ˜ + C1 m t Á + ˜ - C2 re (Eq. 5.8)
∂x j ∂x j Ë s e ∂x j ¯ K ∂x i Ë ∂x j ∂x i ¯ K

where:
∂/∂xi = differentiation with respect to coordinate direction x,y or z (m),
K = turbulent kinetic energy field, m2/s2,
ε = turbulent dissipation field, m2/s3,
ρ = density (kg/m3),
µt = turbulent viscosity (kg/m-s),
vi = velocity component in x, y or z direction (m/s),
σK, σ× = empirical constants (1.0, 1.3),
C1, C2 = empirical constants (1.44, 1.92),
i, j = coordinate direction indices, which, when repeated in a term, implies the sum-
mation of all three possible terms.
This approach generally uses special “wall functions” as the boundary conditions in order to
achieve reasonable accuracy on a coarse grid.31,32 Alternatively, a “low Reynold’s number” turbu-
lence model can be used, which models the boundary layer in a more general way but requires a
finer mesh at the walls.9, 34 An intermediate method between direct numerical simulation and K-ε
turbulence models, called “large eddy simulation,” uses a turbulence model only at the sub-grid
scale.35
Most previous flow models have used the finite difference method, owing to the availability of very
fast and efficient solution methods.36 Popular general-purpose codes of this type include CFX,37
FLUENT,38 and PHOENICS.39 Special-purpose codes of this type include MAGMASOFT40 and
PHYSICA,41 which also solve for solidification and temperature evolution in castings, coupled
with mold filling. The finite element method, such as used in FIDAP,42 can also be applied and has
the advantage of being more easily adapted to arbitrary geometries, although it takes longer to exe-
cute. Special-purpose codes of this type include PROCAST43 and CAFE,44 which are popular for
investment casting processes.
Flow in the mold is of great interest because it influences many important phenomena that have
far-reaching consequences on strand quality. Some of these phenomena are illustrated in Fig. 5.2.
They include the dissipation of superheat by the liquid jet impinging upon the solidifying shell
(and temperature at the meniscus), the flow and entrainment of the top-surface powder layers, top-
surface contour and level fluctuations, and the entrapment of subsurface inclusions and gas bub-
bles. Design compromises are needed to simultaneously satisfy the contradictory requirements for
avoiding each of these defect mechanisms, as discussed in Chapter 14.
It is important to extend the simulation as far upstream as necessary to provide adequate inlet
boundary conditions for the domain of interest. For example, flow calculations in the mold should
be preceded by calculations of flow through the submerged entry nozzle. This provides the veloc-
ities entering the mold in addition to the turbulence parameters, K and ε. Nozzle geometry greatly
affects the flow in the mold and is easy to change, so it is an important subject for modeling.
The flow pattern changes radically with increasing argon injection rate, which requires the solu-
tion of additional equations for the gas phase, and knowledge of the bubble size.6,37,45 The flow pat-
tern and mixing can also be altered by the application of electromagnetic forces, which can either
brake or stir the liquid. This can be modeled by solving the Maxwell, Ohm and charge conserva-
tion equations for electromagnetic forces simultaneously with the flow model equations.46 The

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 11
Casting Volume

great complexity that these phenomena add to the coupled model equations makes these calcula-
tions uncertain and a subject of ongoing research.

5.2.3 Superheat Dissipation

An important task of the flow pattern is to deliver molten steel to the meniscus region that has
enough superheat during the critical first stages of solidification. Superheat is the sensible heat
contained in the liquid metal above the liquidus temperature and is dissipated mainly in the mold.
The transport and removal of superheat is modeled by solving equation 5.2 using the velocities
found from the flow model (equations 5.4 to 5.8). The effective thermal conductivity of the liquid
is proportional to the effective viscosity, which can be found from the turbulence parameters (K
and ε). The solidification front, which forms the boundary to the liquid domain, can be treated in
different ways. Many researchers model flow and solidification as a coupled problem on a fixed
grid.25,47,48 Although very flexible, this approach is subject to convergence difficulties and requires
a fine grid to resolve the thin,
porous mushy zone next to the
thin shell.
An alternative approach for
columnar solidification of a
thin shell, such as found in the
mold for the continuous cast-
ing of steel, is to treat the
boundary as a rough wall fixed
at the liquidus temperature
using thermal wall laws.49 Fig.
5.5 compares calculations using
this approach with measured
temperatures in the liquid pool.50
Incorporating the effects of
argon on the flow pattern was
very important in achieving the
reasonable agreement observed.
This figure shows that the tem-
perature drops almost to the
liquidus by mold exit, indicat-
ing that most of the superheat
is dissipated in the mold. Most
of this heat is delivered to the
narrow face where the jet
impinges, which is important
to shell solidification.51
The coldest regions are found
at the meniscus at the top cor-
ners near the narrow face and
near the SEN. This is a concern
because it could lead to freez-
ing of the meniscus, and
encourage solidification of a
thick slag rim. This could lead
Fig. 5.5 Temperature distribution in mold showing superheat dissipation. From to quality problems such as
Ref. 50. deep oscillation marks, cracks

12 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

and other surface defects. In the extreme, the steel surface can solidify into a solid bridge between
the SEN and the shell against the mold wall, which often causes a breakout. To avoid these prob-
lems, flow must reach the surface quickly. These calculations should be used, for example, to help
design nozzle port geometries that do not direct the flow too deep.

5.2.4 Top Surface Powder/Flux Layer Behavior

The flow of steel in the upper mold may influence the top surface powder layers, which are very
important to steel quality. Mold powder is added periodically to the top surface of the steel. It sin-
ters and melts to form a protective liquid flux layer, which helps to trap impurities and inclusions.
This liquid is drawn into the gap between the shell and mold during oscillation, where it acts as a
lubricant and helps to make heat transfer more uniform. These phenomena are difficult to measure
or to accurately simulate with a physical model, so are worthy of mathematical modeling.
Fig. 5.6 shows results from a 3-D finite-element model of heat transfer and fluid flow in the pow-
der and flux layers, based on solving equations 5.2, 5.4 and 5.5.52 The bottom of the model domain
is the steel/flux interface. Its shape is imposed based on measurements in an operating caster. Alter-
natively, this interface shape can be calculated by solving additional equations to satisfy the force
balance at the interface, which involves the pressure in the two phases, shear forces from the mov-
ing fluids, surface tension and gravity.53 For the conditions in this figure, the momentum of the
flow up the narrow face has raised the level of the interface there. The shear stress along the inter-
face is determined through coupled calculations with the 3-D steady flow model. The model fea-
tures different temperature-dependent flux properties for the interior, during sintering before
melting, compared with the region near the narrow face mold walls, where the flux resolidifies to
form a solid rim.
When molten steel flows rapidly along the steel/flux interface, it induces motion in the flux layer.
If the interface velocity becomes too high, then the liquid flux can be sheared away from the inter-
face, become entrained in the steel jet, and be sent deep into the liquid pool to become trapped in
the solidifying shell as a harmful inclusion. If the interface velocity increases further, then the inter-
face standing wave becomes
unstable, and huge level fluc-
tuations contribute to further
problems.
The thickness of the benefi-
cial liquid flux layer is also
very important. As shown in
the model calculations in Fig.
5.6, the liquid flux layer may
become dangerously thin
near the narrow face if the
steel flow tends to drag the
liquid toward the center. This
shortage of flux feeding into
the gap can lead to air gaps,
reduced nonuniform heat
flow, thinning of the shell,
and longitudinal surface
cracks. Quantifying these Fig. 5.6 Comparison of measured and predicted melt-interface positions.52
phenomena requires model-
ing of both the steel flow and
flux layers.

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 13
Casting Volume

5.2.5 Motion and Entrapment of Inclusions and Gas Bubbles

The jets of molten steel exiting the nozzle may carry argon bubbles and inclusions such as alumina
into the mold cavity. These particles may create defects if they become entrapped in the solidify-
ing shell. Particle trajectories can be calculated using the Langrangian particle tracking method,
which solves a transport equation for each particle as it travels through a previously-calculated
velocity field.34,54,55
The force balance on each particle includes buoyancy and drag force relative to the molten steel.
The effects of turbulent motion can be modeled crudely from a K-ε flow field by adding a random
velocity fluctuation at each step, whose magnitude
varies with the local turbulent kinetic energy level. To
obtain significant statistics, the trajectories of several
hundred individual particles should be calculated,
using different starting points. The bubbles collect
inclusions, and inclusion clusters collide, so their size
and shape distributions evolve with time, which affects
their drag and flotation velocities and importance.
Models are being developed to include these effects.55
Fig. 5.7 shows the trajectories of several particles mov-
ing through a steady flow field, calculated using the K-
ε model.55 This simulation features particle trajectory
tracking that incorporates the influence of turbulence
by giving a random velocity component to the velocity
at each time step in the calculation, in proportion to the
local turbulence level.
Most of the argon bubbles circulate in the upper mold
area and float out to the top surface. A few might be
trapped at the meniscus if there is a solidification
hook, and lead to surface defects. A few small bubbles
Fig. 5.7 Sample trajectories of 0.3 mm argon bub-
manage to penetrate into the lower recirculation zone,
bles with turbulent motion. From Ref. 55.
where they move similarly to large inclusion clusters.
Particles in this lower region tend to move slowly in
large spirals, while they float toward the inner radius of the slab. When they eventually touch the
solidifying shell in this deep region, entrapment is more likely on the inside radius. Trapped argon
bubbles elongate during rolling and, in low-strength steel, may expand during subsequent anneal-
ing processes to create costly surface blisters and “pencil pipe” defects. Transient models are
likely to yield further insights into the complex and important phenomena of inclusion entrap-
ment.

5.2.6 Composition Variation During Grade Changes

Large composition differences can arise through the thickness and along the length of the final
product due to intermixing after a change in steel grade during continuous casting. Steel produc-
ers need to optimize casting conditions and grade sequences to minimize the amount of steel down-
graded or scrapped due to this intermixing. In addition, the unintentional sale of intermixed
product must be avoided. To do this requires knowledge of the location and extent of the inter-
mixed region and how it is affected by grade specifications and casting conditions.
Models to predict intermixing must first simulate composition change in both the tundish and in
the liquid core of the strand as a function of time. This can be done using simple lumped mixing-
box models and/or by solving the mass diffusion equation in the flowing liquid:

14 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

∂C ∂C ∂ ∂C
+ vi = (Deff ) (Eq. 5.9)
∂t ∂x i ∂x i ∂x i

In this equation, the composition, C, ranges between the old grade concentration of 0 and the new
grade concentration of 1. This dimensionless concept is useful because alloying elements inter-
mix essentially equally, owing to the much greater importance of convection and turbulent diffu-
sion Deff over laminar diffusion. In order to predict the composition distribution within the final
product, a further model must account for the cessation of intermixing after the shell has solidi-
fied.
Fig. 5.8 shows example composition distributions in a continuous cast slab calculated using such
a model.56,57 To ensure accuracy, extensive verification and calibration must be undertaken for each
submodel.57 The tundish mixing
submodel must be calibrated to
match chemical analysis of steel
samples taken from the mold in the
nozzle port exit streams, or with
tracer studies using full-scale water
models. Accuracy of a simplified
strand submodel is demonstrated in
Fig. 5.8 through comparison both
with composition measurements in a
solidified slab and with a full 3-D
model (equation 5.4).
The results in Fig. 5.8 clearly show
the important difference between
centerline and surface composition.
New grade penetrates deeply into
the liquid cavity and contaminates
the old grade along the centerline. Fig. 5.8 Predicted composition distribution in a steel slab cast during a
Old grade lingers in the tundish and grade change compared with experiments. From Ref. 57.
mold cavity to contaminate the sur-
face composition of the new grade. This difference is particularly evident in small tundish, thick-
mold operations, where mixing in the strand is dominant.
Intermix models such as this one are in use at many steel companies. The model can be enhanced to
serve as an on-line tool by outputting, for each grade change, the critical distances that define the
length of intermixed steel product that falls outside the given composition specifications for the old
and new grades.57 In addition, it can be applied off-line to perform parametric studies to evaluate the
relative effects on the amount of intermixed steel for different intermixing operations and for differ-
ent operating conditions using a standard ladle-exchange operation.58 Finally, it can be used to opti-
mize scheduling and casting operation in order to minimize cost.

5.2.7 Thermal Mechanical Behavior of the Mold

Thermal distortion of the mold during operation is important to residual stress, residual distortion,
fatigue cracks and mold life. By affecting the internal geometry of the mold cavity, it is also impor-
tant to heat transfer to the solidifying shell. To study thermal distortion of the mold and its related
phenomena first requires accurate solution of heat transfer, equation 5.3, using measurements to
help determine the interfacial heat flux. In addition, a thermal-mechanical model must solve the
equilibrium equations that relate force and stress, the constitutive equations that relate stress and
strain, and the compatibility equations that relate strain and displacement.

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 15
Casting Volume

∂s ij
Fi = (Eq. 5.10)
∂x i

s ij = Dijkl e elij (Eq. 5.11)

Ê
1 ∂u i
∂u j ˆ
e tot
ij = 2 Á + ˜ (Eq. 5.12)
Ë ∂x j ∂x i ¯

where:
∂/∂x = differentiation with respect to coordinate direction (m-1),
Fi = force component in i direction (N),
σij = stress component (N/m2),
xi = coordinate direction, x, y or z (m),
Dijkl = components of elasticity tensor (N/m2),
ε ijel = elastic strain component (–)
ε tot
ij = total strain component (–),
ui = displacement component in i direction (m),
i = coordinate direction index which, when appearing twice in a term, implies the
summation of all three possible terms.
Thermal strain is found from the tempera-
tures calculated in the heat transfer model and
accounts for the difference between the elas-
tic and total strain. Further details are found
elsewhere.
In order to match the measured distortion,
models should incorporate all the important
geometric features of the mold, which often
includes the four copper plates with their
water slots, reinforced steel water box assem-
blies, and tightened bolts. Three-dimensional
elastic-plastic-creep finite element models
have been developed for slabs59 and thin
slabs60,61 using the commercial finite-element
package ABAQUS,62 which is well suited to
this nonlinear thermal stress problem. Their
four-piece construction makes slab molds
behave very differently from single-piece
bloom or billet molds, which have also been
studied using thermal stress models.63
Fig. 5.9 illustrates typical temperature con-
tours and the displaced shape calculated in
one quarter of the mold under steady operat-
Fig. 5.9 Distorted shape of thin slab casting mold during
ing conditions.61 The hot exterior of each cop-
operation (50X magnification) with temperature contours per plate attempts to expand but is
(°C). From Ref. 61. constrained by its colder interior and the cold,

16 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

stiff, steel water jacket. This makes each plate bend in toward the solidifying steel. Maximum
inward distortions of more than one millimeter are predicted just above the center of the mold
faces, and below the location of highest temperature, which is found just below the meniscus.
The narrow face is free to rotate away from the wide face and contact only along a thin vertical
line at the front corner of the hot face. This hot edge must transmit all of the clamping forces, so
it is prone to accelerated wear and crushing, especially during automatic width changes. If steel
enters the gaps formed by this mechanism, this can lead to finning defects or even a sticker break-
out. In addition, the wide faces may be gouged, leading to longitudinal cracks and other surface
defects.
The high compressive stress due to constrained thermal expansion induces creep in the hot exte-
rior of the copper plates that face the steel. This relaxes the stresses during operation but allows
residual tensile stress to develop during cooling. Over time, these cyclic thermal stresses and creep
build up significant distortion of the mold plates. This can contribute greatly to remachining
requirements and reduced mold life. Under adverse conditions, this stress could lead to cracking
of the copper plates. The distortion predictions are important for designing mold taper to avoid
detrimental air gap formation.
These practical concerns can be investigated with quantitative modeling studies of the effects of
different process and mold design variables on mold temperature, distortion, creep and residual
stress. This type of stress model application will become more important in the future to optimize
the design of the new molds being developed for continuous thin-slab and strip casting. For exam-
ple, thermal distortion of the rolls during operation of a twin-roll strip caster is on the same order
as the section thickness of the steel product.

5.2.8 Thermal Mechanical Behavior of the Shell

The solidifying shell is prone to a variety of distortion, cracking and segregation problems, owing
to its creep at elevated temperature, combined with metallurgical embrittlement and thermal stress.
To start to investigate these problems, models are being developed to simulate coupled fluid flow,
thermal and mechanical behavior of the solidifying steel shell during continuous casting.17,60,64 The
thermal-mechanical solution procedure is documented elsewhere.65 In addition to solving equa-
tions 5.2, 5.3, 5.10, 5.11 and 5.12, further constitutive equations are needed to characterize the
inelastic creep and plastic strains as a function of stress, temperature and structure in order to accu-
rately incorporate the mechanical properties of the material. For example,66

Ê -Q ˆ ne
e
p = C exp Á [s - a e e p ]n (Eq. 5.13)
Ë T ˜¯

where:
.
ε = inelastic strain rate (s–1),
σ = stress (MPa),
εp = inelastic strain (structure parameter),
T = temperature (K),
C,Q,aε ,nε ,n = empirical constants.
Constitutive equations such as these are a subject of ongoing research because the equations are
difficult to develop, especially for complex loading conditions involving stress reversals. The
numerical methods to evaluate them are prone to instability, and the experimental measurements
they are based upon are difficult to conduct.

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 17
Casting Volume

Thermal-mechanical models can be applied in order to predict the evolution of temperature, stress
and deformation of the solidifying shell while in the mold for both billets15 and slabs.16,67–70 In this
region, these phenomena are intimately coupled because the shrinkage of the shell affects heat
transfer across the air gap, which complicates the calculation procedure. The predicted tempera-
ture contours and distorted shape of a transverse region near the corner are compared in Fig. 5.10
with measurements of a breakout shell from an operating steel caster.16 This model tracks the
behavior of a two-dimen-
sional slice through the
strand as it moves down-
ward at the casting speed
through the mold and
upper spray zones. It
consists of separate
finite-element models of
heat flow and stress gen-
eration that are step-wise
coupled through the size
of the interfacial gap.
The heat transfer model
was calibrated using
thermocouple measure-
ments down the center-
line of the wide face for
typical conditions.
Fig. 5.10 Comparison between predicted and measured shell thickness in a horizontal Shrinkage predictions
(x-y) section through the corner of a continuous-cast steel breakout shell. From Ref. 16. from the stress model are
used to find the air gap
thickness needed in equation 5.3 in order to extend the calculations around the mold perimeter. The
model includes the effect of mold distortion on the air gaps, and superheat delivery from the flow-
ing jet of steel, calculated in separate models. The stress model includes ferrostatic pressure from
the molten steel on the inside of the shell and calculates intermittent contact between the shell and
the mold. It also features a temperature-dependent elastic modulus and an elastic-viscoplastic con-
stitutive equation that includes the effects of temperature, composition, phase transformations and
stress state on the local inelastic creep rate. Efficient numerical algorithms are needed to integrate
the equations.
As expected, good agreement is obtained in the region of good contact along the wide face, where
calibration was done. Near the corner along the narrow face, steel shrinkage is seen to exceed the
mold taper, which was insufficient. Thus, an air gap is predicted. This air gap lowers heat extrac-
tion from the shell in the off-corner region of the narrow face. When combined with high super-
heat delivery from the bifurcated nozzle directed at this location, shell growth is greatly reduced
locally. Just below the mold, this thin region along the off-corner narrow-face shell caused the
breakout.
Near the center of the narrow face, creep of the shell under ferrostatic pressure from the liquid is
seen to maintain contact with the mold, so much less thinning is observed. This illustrates the
tremendous effect that superheat has on slowing shell growth, if there is a problem that lowers heat
flow.
Fig. 5.11 presents sample distributions of temperature and stress through the thickness of the shell,
calculated with this model.70 To achieve reasonable accuracy, a very fine mesh and small time steps
are needed. The temperature profile is almost linear through the shell. The stress profile shows that
the shell surface is in compression. This is because, in the absence of friction with the mold, the
surface layer solidifies and cools stress free. As each inner layer solidifies, it cools and tries to shrink,

18 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

while the surface temperature

remains relatively constant.
The slab is constrained to
remain planar, so complemen-
tary subsurface tension and
surface compression stresses
are produced. Note that the
average stress through the shell
thickness is zero in order to
maintain force equilibrium. It
is significant that the maxi-
mum tensile stress is found
near the solidification front.
This generic subsurface tensile
stress is responsible for hot
tear cracks, when accompa-
Fig. 5.11 Typical temperature and stress distributions through shell thickness.
nied by metallurgical embrit- From Ref. 70.
tlement.
Thermal-mechanical models such as this one can be applied to predict ideal mold taper,71 to pre-
vent breakouts such as the one discussed here24 and to understand the cause of other problems such
as surface depressions72 and longitudinal cracks. When combined with transient temperature, flow
and pressure calculations in the slag layers, such models can simulate phenomena at the meniscus
such as oscillation mark formation.73
Computational models can also be applied to calculate thermomechanical behavior of the solidi-
fying shell below the mold. Models can investigate shell bulging between the support rolls due to
ferrostatic-pressure induced creep,74–78 and the stresses induced during unbending.79 These models
are important for the design of spray systems and rolls in order to avoid internal hot tear cracks and
centerline segregation. These models face great numerical challenges because the phenomena are
generally three-dimensional and transient, the constitutive equations are highly nonlinear, and the
mechanical behavior in one region (e.g., the mold) may be coupled with the behavior very far away
(e.g., unbending rolls).

5.2.9 Crack Formation

Although obviously of great interest, crack formation is particularly difficult to model directly and
is rarely attempted. Very small strains (on the order of 1%) can start hot tear cracks at the grain
boundaries if liquid metal is unable to feed through the secondary dendrite arms to accommodate
the shrinkage. Strain localization may occur on both the small scale (when residual elements seg-
regate to the grain boundaries) and on a larger scale (within surface depressions or hot spots). Later
sources of tensile stress, including constraint due to friction and sticking, unsteady cooling below
the mold, withdrawal forces, bulging between support rolls, and unbending all worsen strain con-
centration and promote crack growth. Microstructure, grain size and segregation are extremely
complex, so modeling of these phenomena is generally done independently of the stress model. Of
even greater difficulty for computational modeling is the great difference in scale between these
microstructural phenomena relative to the size of the casting, where the important macroscopic
temperature and stress fields develop.
Considering this complexity, the results of macroscopic thermal-stress models are linked to the
microstructural phenomena that control crack initiation and propagation through the use of frac-
ture criteria. To predict hot tear cracks, most fracture criteria identify a critical amount of inelastic
strain (e.g., 1 – 3.8%) accumulated over a critical range of liquid fraction, fL, such as 0.01< fL <
0.2.80,81 Recent work suggests that the fracture criteria should consider the inelastic strain rate,

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 19
Casting Volume

which is important during liquid feeding through a permeable region of dendrite arms in the mushy
zone.82 Careful experiments are needed to develop these fracture criteria by applying stress during
solidification.83–85 There experiments are difficult to control, so detailed modeling of the experi-
ment itself is becoming necessary, in order to extract more fundamental material properties such
as fracture criteria.

5.2.10 Centerline Segregation

Macrosegregation near the centerline of the solidified slab is detrimental to product properties, par-
ticularly for highly alloyed steels, which experience the most segregation. Centerline segregation
can be reduced and even avoided through careful application of electromagnetic forces, and soft
reduction, where the slab is rolled or quenched just before it is fully solidified. Computational
modeling would be useful to understanding and optimizing these practices.
Centerline segregation is a very difficult problem to simulate because such a wide range of cou-
pled phenomena must be properly modeled. Bulging between the rolls and solidification shrinkage
together drive the fluid flow necessary for macrosegregation, so fluid flow, solidification heat
transfer and stresses must all be modeled accurately (including equations 5.2, 5.4, 5.5, 5.10, 5.12
and 5.13). The microstructure is also important, as equiaxed crystals behave differently than
columnar grains, and so must also be modeled. This is complicated by the convection of crystals in
the molten pool in the strand, which depends on both flow from the nozzle and thermal/solutal con-
vection. Increasing superheat tends to worsen segregation, so the details of mold superheat transfer
must also be properly modeled. Naturally, equation 5.8 must be solved for each important alloying
element on both the microstructural scale (between dendrite arms), with the help of microsegrega-
tion software such as THERMOCALC,86 and on the macroscopic scale (from surface to center of
the strand), using advanced computations.48,87 The diffusion coefficients and partition coefficients
needed for this calculation are not currently known with sufficient accuracy. Finally, the application
of electromagnetic and roll forces generate additional modeling complexity. Although the task
appears overwhelming, steps are being taken to model this important problem.88,89

Much further work is needed to understand and quantify these phenomena and to apply the results
to optimize the continuous casting process. In striving towards these goals, the importance of com-
bining modeling and experiments together cannot be overemphasized.

5.3 Conclusion
The final test of a model is if the results can be implemented and improvements can be achieved,
such as the avoidance of defects in the steel product. Plant trials are ultimately needed for this
implementation. Trials should be conducted on the basis of insights supplied from all available
sources, including physical models, mathematical models, literature and previous experience.

As increasing computational power continues to advance the capabilities of numerical simulation

tools, modeling should play an increasing role in future advances to high-technology processes
such as the continuous casting of steel. Modeling can augment traditional research methods in gen-
erating and quantifying the understanding needed to improve any aspect of the process. Areas
where advanced computational modeling should play a crucial role in future improvements include
transient flow simulation, mold flux behavior, taper design, online quality prediction and control,
especially for new problems and processes such as high-speed billet casting, thin slab casting and
strip casting.

Future advances in the continuous casting process will not come from either models, experiments,
or plant trials. They will come from ideas generated by people who understand the process and the

20 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

problems. This understanding is rooted in knowledge, which can be confirmed, deepened, and
quantified by tools that include computational models. As our computational tools continue to
improve, they should grow in importance in fulfilling this important role, leading to future process
advances.

References
1. J. Szekely, J.W. Evans and J.K. Brimacombe, The Mathematical and Physical Modeling of Pri-
mary Metals Processing Operations (New York: John Wiley & Sons, 1987).
2. J. Szekely and N. Themelis, Rate Phenomena in Process Metallurgy (New York: Wiley-Inter-
science, 1971), 515–597.
3. R.I.L. Guthrie, Engineering in Process Metallurgy (Oxford, UK: Clarendon Press, 1992), 528.
4. L.J. Heaslip and J. Schade, “Physical Modeling and Visualization of Liquid Steel Flow Behav-
ior During Continuous Casting,” Iron & Steelmaker, 26:1 (1999): 33–41.
5. S.L. Lee, “Particle Drag in a Dilute Turbulent Two-Phase Suspension Flow,” Journal of Mul-
tiphase Flow, 13:2 (1987): 247.
6. B.G. Thomas, X. Huang and R.C. Sussman, “Simulation of Argon Gas Flow Effects in a Con-
tinuous Slab Caster,” Metall. Trans. B, 25B:4 (1994): 527–547.
7. S. Sivaramakrishnan, H. Bai, B.G. Thomas, P. Vanka, P. Dauby and M. Assar, “Transient Flow
Structures in Continuous Cast Steel,” in Ironmaking Conference Proceedings, 59, Pittsburgh,
Pa. (Warrendale, Pa.: Iron and Steel Society, 2000), 541–557.
8. M.B. Assar, P.H. Dauby and G.D. Lawson, “Opening the Black Box: PIV and MFC Measure-
ments in a Continuous Caster Mold,” in Steelmaking Conference Proceedings, 83 (Warrendale,
Pa.: Iron and Steel Society, 2000), 397–411.
9. X.K. Lan, J.M. Khodadadi and F. Shen, “Evaluation of Six k-Turbulence Model Predictions of
Flowin a Continuous Casting Billet-Mold Water Model Using Laser Doppler Velocimetry
Measurements,” Metall. Mater. Trans., 28B:2 (1997): 321–332.
10. J. Herbertson, Q.L. He, P.J. Flint and R.B. Mahapatra, “Modelling of Metal Delivery to Con-
tinuous Casting Moulds,” in Steelmaking Conference Proceedings, 74 (Warrendale, Pa.: Iron
and Steel Society, 1991), 171–185.
11. B.G. Thomas, R. O’Malley, T. Shi, Y. Meng, D. Creech and D. Stone, “Validation of Fluid
Flow and Solidification Simulation of a Continuous Thin Slab Caster,” in Modeling of Cast-
ing, Welding, and Advanced Solidification Processes, IX, Aachen, Germany, Aug. 20–25, 2000
(Aachen, Germany: Shaker Verlag GmbH, 2000), 769–776.
12. B.G. Thomas, “Mathematical Modeling of the Continuous Slab Casting Mold: A State of the
Art Review,” in 74th Steelmaking Conference Proceedings, 74 (Warrendale, Pa.: Iron and Steel
Society, 1991), 105–118.
13. J. Lait, J.K. Brimacombe and F. Weinberg, “Mathematical Modelling of Heat Flow in the Con-
tinuous Casting of Steel,” Ironmaking and Steelmaking, 2 (1974): 90–98.
14. R.B. Mahapatra, J.K. Brimacombe and I.V. Samarasekera, “Mold Behavior and its Influence
on Quality in the Continuous Casting of Slabs: Part I. Industiral Trials, Mold Temperature
Measurements, and Mathematical Modelling,” Metallurgical Transactions B, 22B (Dec.
(1991), 861–874.
15. J.E. Kelly, K.P. Michalek, T.G. OConnor, B.G. Thomas, J.A. Dantzig, “Initial Development of
Thermal and Stress Fields in Continuously Cast Steel Billets,” Metallurgical Transactions A,
19A:10 (1988): 2589–2602.
16. A. Moitra and B.G. Thomas, “Application of a Thermo-Mechanical Finite Element Model of
Steel Shell Behavior in the Continuous Slab Casting Mold,” in Steelmaking Proceedings, 76
(Dallas, Texas: Iron and Steel Society, 1993), 657–667.
17. J.-E. Lee, T.-J. Yeo, K.H. Oh, J.-K. Yoon and U.-S. Yoon, “Prediction of Cracks in Continu-
ously Cast Beam Blank Through Fully Coupled Analysis of Fluid Flow, Heat Transfer and
Deformation Behavior of Solidifying Shell,” Metall. Mater. Trans. A, 31A:1 (2000):
225–237.

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 21
Casting Volume

18. R. Bommaraju and E. Saad, “Mathematical modelling of lubrication capacity of mold fluxes,”
in Steelmaking Proceedings, 73 (Warrendale, Pa.: Iron and Steel Society, 1990), 281–296.
19. B.G. Thomas and B. Ho, “Spread Sheet Model of Continuous Casting,” J. Engineering Indus-
try, 118:1 (1996): 37–44.
20. B. Ho, “Characterization of Interfacial Heat Transfer in the Continuous Slab Casting Process”
(Masters Thesis, University of Illinois at Urbana–Champaign, 1992).
21. J.A. DiLellio and G.W. Young, “An Asymptotic Model of the Mold Region in a Continuous
Steel Caster,” Metall. Mater. Trans., 26B:6 (1995): 1225–1241.
22. B.G. Thomas, D. Lui and B. Ho, “Effect of Transverse and Oscillation Marks on Heat Trans-
fer in the Continuous Casting Mold,” in Applications of Sensors in Materials Processing,
edited by V. Viswanathan, Orlando, Fla. (Warrendale, Pa.: TMS, 1997), 117–142.
23. B.G. Thomas, B. Ho and G. Li, “Heat Flow Model of the Continuous Slab Casting Mold, Inter-
face, and Shell,” in Alex McLean Symposium Proceedings, Toronto (Warrendale, Pa.: Iron and
Steel Society, 1998), 177–193.
24. G.D. Lawson, S.C. Sander, W.H. Emling, A. Moitra and B.G. Thomas, “Prevention of Shell
Thinning Breakouts Associated with Widening Width Changes,” in Steelmaking Conference
Proceedings, 77 (Warrendale, Pa.: Iron and Steel Society, 1994), 329–336.
25. C.-A. Gandin, T. Jalanti and M. Rappaz, “Modeling of Dendritic Grain Structures,” in Model-
ing of Casting, Welding, and Advanced Solidification Processes, VIII, edited by B.G. Thomas
and C. Beckermann (Warrendale, Pa.: TMS, 1998), 363–374.
26. I. Steinbach and G.J. Schmitz, “Direct Numerical Simulation of Solidification Structure using
the Phase Field Method,” in Modeling of Casting, Welding, and Advanced Solidification
Processes, VIII, edited by B.G. Thomas and C. Beckermann (Warrendale, Pa.: TMS, 1998),
521–532.
27. R.A. Hardin, K. Liu and C. Beckermann, “Development of a Model for Transient Simulation
and Control of a Continuous Steel Slab Caster,” in Materials Processing in the Computer Age,
3 (Warrendale, Pa.: Minerals, Metals, & Materials Society, 2000), 61–74.
28. J. Szekely and R.T. Yadoya, “The Physical and Mathematical Modelling of the Flow Field in
teh Mold Region of Continuous Casitng Units. Part II. Computer Solution of the Turbulent
Flow Equations,” Metall. Mater. Trans., 4 (1973): 1379.
29. S.V. Patankar, Numerical Heat Transfer and Fluid Flow (New York: McGraw Hill, 1980).
30. S.V. Patankar and B.D. Spalding, “A calculation procedure for heat, mass and momentum
transer in three-dimensional parabolic flows,” Int. J. Heat Mass Transfer, 15 (1992): 1787-
1806.
31. B.E. Launder and D.B. Spalding, “Numerical Computation of Turbulent Flows,” Computer
Methods in Applied Mechanics and Engr, 13 (1974): 269–289.
32. B.G. Thomas and F.M. Najjar, “Finite-Element Modeling of Turbulent Fluid Flow and Heat
Transfer in Continuous Casting,” Applied Mathematical Modeling, 15 (1991): 226–243.
33. D.E. Hershey, B.G. Thomas and F.M. Najjar, “Turbulent Flow through Bifurcated Nozzles,”
International Journal for Numerical Methods in Fluids, 17 (1993): 23–47.
34. M.R. Aboutalebi, M. Hasan and R.I.L. Guthrie, “Coupled Turbulent Flow, Heat, and Solute
Transport in Continuous Casting Processes,” Metall. Mater. Trans., 26B:4 (1995): 731–744.
35. J. Smagorinsky, “General Circulation Experiments with the Primitive Equations” (Washing-
ton: Report, Weather Bureau, 1963), 99–121.
36. N.C. Markatos, “Computational Fluid Flow Capabilities and Software,” Ironmaking and Steel-
making, 16:4 (1989): 266–273.
37. “CFX 4.2” (Pittsburgh, Pa.: AEA Technology, 1998).
38. “FLUENT 5.1” (Lebanon, N.H.: Fluent Inc., 2000).
39. “PHOENICS” (London: CHAM, 2000).
40. E. Flender, “MAGMASOFT” (Aachen, Germany: Magma Gmbh, 2000).
41. M. Cross, “PHYSICA” (Greenwich, UK: Computing & Math Sciences, 2000).
42. M.S. Engleman, “FIDAP 8.5” (Evanston, Ill.: Fluent, Inc., 2000).
43. M. Sammonds, “PROCAST” (Annapolis, Md.: UES Software, 2000).
44. P. Thevoz, “CAFE” (Lausanne, Switzerland: CALCOM, EPFL, 2000).

22 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.
 Modeling of Continuous Casting

45. N. Bessho, R. Yoda, H. Yamasaki, T. Fujii and T. Nozaki, “Numerical Analysis of Fluid Flow
in the Continuous Casting Mold by a Bubble Dispersion Model,” Iron and Steelmaker, 18:4
(1991): 39–44.
46. T. Ishii, S.S. Sazhin and M. Makhlouf, “Numerical prediction of magnetohydrodynamic flow
in continuous casting process,” Ironmaking and Steelmaking, 23:3 (1996): 267–272.
47. V.R. Voller, A.D. Brent and C. Prakash, “The Modelling of Heat, Mass and Solute Transport
in Solidification Systems,” Applied Mathematical Modelling, 32 (1989): 1719–1731.
48. M.C. Schneider and C. Beckermann, “The Formation of Macrosegregation by Multicompo-
nent Thermosolutal Convection During the Solidification of Steel,” Metall. Mater. Trans. A,
26A (1995): 2373–2388.
49. X. Huang, B.G. Thomas and F.M. Najjar, “Modeling Superheat Removal During Continuous
Casting of Steel Slabs,” Metallurgical Transactions B, 23B:6 (1992): 339–356.
50. B.G. Thomas, “Continuous Casting of Steel, Chap. 15,” in Modeling and Simulation for Casting
and Solidification: Theory and Applications, edited by O. Yu (New York: Marcel Dekker, 2000).
51. H. Nakato, M. Ozawa, K. Kinoshita, Y. Habu and T. Emi, “Factors Affecting the Formation of
Shell and Longitudinal Cracks in Mold During High-Speed Continuous Casting of Slabs,”
Trans. Iron Steel Inst. Japan, 24:11 (1984): 957–965.
52. R. McDavid and B.G. Thomas, “Flow and Thermal Behavior of the Top-Surface Flux/Powder
Layers in Continuous Casting Molds,” Metallurgical Transactions B, 27B:4 (1996): 672–685.
53. G.A. Panaras, A. Theodorakakos and G. Bergeles, “Numerical Investigation of the Free Surface
in a continuous Steel Casting Mold Model,” Metall. Mater. Trans. B, 29B (5) (1998): 1117–1126.
54. R.C. Sussman, M. Burns, X. Huang and B.G. Thomas, “Inclusion Particle Behavior in a Con-
tinuous Slab Casting Mold,” in 10th Process Technology Conference Proc., 10, Toronto (War-
rendale, Pa.: Iron and Steel Society, 1992), 291–304.
55. B.G. Thomas, A. Dennisov and H. Bai, “Behavior of Argon Bubbles During Continuous Cast-
ing of Steel,” in Steelmaking Conference Proceedings, 80, Chicaago (Warrendale, Pa.: Iron and
Steel Society, 1997), 375–384.
56. X. Huang and B.G. Thomas, “Modeling of Steel Grade Transition in Continuous Slab Casting
Processes,” Metallurgical Transactions, 24B (1993): 379–393.
57. X. Huang and B.G. Thomas, “Intermixing Model of Continuous Casting During a Grade Tran-
sition,” Metallurgical Mater Transactions B, 27B:4 (1996): 617–632.
58. B.G. Thomas, “Modeling Study of Intermixing in Tundish and Strand During a Continuous-
Casting Grade Transition,” ISS Transactions, 24:12 (1997): 83–96.
59. B.G. Thomas, G. Li, A. Moitra and D. Habing, “Analysis of Thermal and Mechanical Behav-
ior of Copper Molds during Continuous Casting of Steel Slabs,” Iron and Steelmaker (ISS
Transactions), 25:10 (1998): 125–143.
60. T. O’Conner and J. Dantzig, “Modeling the Thin Slab Continuous Casting Mold,” Metall.
Mater. Trans., 25B:4 (1994): 443–457.
61. J.-K. Park, I.V. Samarasekera, B.G. Thomas and U.-S. Yoon, “Analysis of Thermal and
Mechanical Behavior of Copper Mould During Thin Slab Casting,” in Steelmaking Conference
Proceedings, 83 (Warrendale, Pa.: Iron and Steel Society, 2000), 9–21.
62. “ABAQUS 5.8” (Pawtucket, R.I.: Hibbitt, Karlsson & Sorensen, Inc., 1999).
63. I.V. Samarasekera, D.L. Anderson and J.K. Brimacombe, “The Thermal Distortion of Contin-
uous Casting Billet Molds,” Metallurgical Transactions B, 13B (March 1982): 91–104.
64. A. Cristallini, M.R. Ridolfi, A. Spaccarotella, R. Capotosti, G. Flemming and J. Sucker,
“Advanced Process Modeling of CSP Funnel Design for Thin Slab Casting of High-Alloyed
Steels,” in Seminario de Aceria of the Instituto Argentino de Siderurgia (IAS) Proceedings,
(Buones Aires, Argentina: IAS, 1999), 468–477.
65. J.A. Dantzig, “Thermal Stress Development in Metal Casting Processes,” Metallurgical Sci-
ence and Technology, 7:3 (1989): 133–178.
66. P. Kozlowski, B.G. Thomas, J. Azzi and H. Wang, “Simple Constitutive Equations for Steel at
High Temperature,” Metallurgical Transactions A, 23A (March 1992): 903–918.
67. K. Sorimachi and J.K. Brimacombe, “Improvements in Mathematical Modelling of Stresses in
Continuous Casting of Steel,” Ironmaking and Steelmaking, 4 (1977): 240–245.

Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved. 23
Casting Volume

68. K. Kinoshita, H. Kitaoka and T. Emi, “Influence of Casting Conditions on the Solidification of
Steel Melt in Continuous Casting Mold,” Tetsu-to-Hagane, 67:1 (1981): 93–102.
69. I. Ohnaka and Y. Yashima, “Stress Analysis of Steel Shell Solidifying in Continuous Casting
Mold,” in Modeling of Casting and Welding Processes IV, 4 (Warrendale, Pa.: Minerals, Met-
als, and Materials Society, 1988), 385–394.
70. B.G. Thomas and J.T. Parkman, “Simulation of Thermal Mechanical Behavior During Initial
Solidification,” in Thermec 97 Internat. Conf. on Thermomechanical Processing of Steel and
Other Materials, 2, edited by T. Chandra (Wollongong, Australia: TMS, 1997), 2279–2285.
71. B.G. Thomas, A. Moitra and W.R. Storkman, “Optimizing Taper in Continuous Slab Casting
Molds Using Mathematical Models,” in Proceedings, 6th International Iron and Steel Con-
gress, 3, Nagoya, Japan (Tokyo: Iron & Steel Inst. Japan, 1990), 348–355.
72. B.G. Thomas, A. Moitra and R. McDavid, “Simulation of Longitudinal Off-Corner Depres-
sions in Continuously-Cast Steel Slabs,” ISS Transactions, 23:4 (1996): 57–70.
73. K. Schwerdtfeger and H. Sha, “Depth of Oscillation Marks Forming in Continuous Casting of
Steel,” Metallurgical and Materials Transactions A, in press (1999).
74. A. Palmaers, A. Etienne and J. Mignon, “Calculation of the Mechanical and Thermal Stresses
in Continuously Cast Strands, in German,” Stahl und Eisen, 99:19 (1979): 1039–1050.
75. J.B. Dalin and J.L. Chenot, “Finite Element Computation of Bulging in Continuously Cast
Steel with a Viscoplastic Model,” International Journal for Numerical Methods in Engineer-
ing, 25 (1988): 147–163.
76. B. Barber and A. Perkins, “Strand Deformation in Continuous Casting,” Ironmaking and Steel-
making, 16:6 (1989): 406–411.
77. K. Okamura and H. Kawashima, “Calculation of Bulging Strain and Its Application to Predic-
tion of Internal Cracks in Continuously Cast Slabs,” in Proc. Int. Conf. Comp. Ass. Mat. Design
Proc. Simul. (Tokyo: ISIJ, 1993), 129–134.
78. L. Yu, “Bulging in Continuous Cast Steel Slabs” (M.S. Thesis, University of Illinois, 2000).
79. M. Uehara, I.V. Samarasekera and J.K. Brimacombe, “Mathematical modelling fo unbending
of continuously cast steel slabs,” Ironmaking and Steelmaking, 13:3 (1986), 138.
80. T.W. Clyne and G.J. Davies, “The influence of composition on solidification cracking suscep-
tibiltity in binary alloy systems,” Br. Foundrymen, 74:4 (1981): 65–73.
81. A. Yamanaka, K. Nakajima and K. Okamura, “Critical strain for internal crack formation in
continuous casting,” Ironmaking and Steelmaking, 22:6 (1995): 508–512.
82. M. Rappaz, J.-M. Drezet and M. Gremaud, “A New Hot-Tearing Criterion,” Metall. Mater.
Trans. A, 30A:2 (1999): 449–455.
83. T. Matsumiya, M. Ito, H. Kajioka, S. Yamaguchi, Y. Nakamura, “An Evaluation of Critical
Strain for Internal Crack Formation in Continuously Cast Slabs,” Transactions of the Iron and
Steel Institute of Japan, 26 (1986): 540–546.
84. C. Bernhard, H. Hiebler and M.M. Wolf, “Experimental simulation of subsurface crack for-
mation in continuous casting,” Rev. Metall. (March 2000), 333–344.
85. C.H. Yu, M. Suzuki, H. Shibata, T. Emi, “Simulation of Crack Formation on Solidifying Steel
Sheel in Continuous Casting Mold,” ISIJ International, 36 (1996): S159–S162.
86. B. Sundman, B. Jansson and J.O. Anderson, “The Thermo-Calc Databank System,” CAL-
PHAD 9, 2 (1985): 153–190.
87. C.Y. Wang and C. Beckermann, “Equiaxed Dendritic Solidification with Convection,” Metall.
Mater. Trans. A, 27A:9 (1996): 2754–2764.
88. T. Kajitani, J.-M. Drezet and M. Rappaz, “Numerical simulation of deformation-induced seg-
regation in continuous casting of steel,” Metall. Mater. Trans. (2001), in press.
89. G. Lesoult and S. Sella, Solid State Phenomena, 3 (1988): 167–178.

24 Copyright © 2003, The AISE Steel Foundation, Pittsburgh, PA. All rights reserved.