International Journal of ISSI, Vol. 19(2022), No. 1, pp.

67-80

The Effect of Slab Thickness on the Solidification of Low Carbon Steel in

Continuous Casting Process: A Simulation Case Study

A. Pourfathi*1

Department of Materials Science and Engineering, Sharif University of Technology, Iran

Abstract

One of the most important and effective factors in the solidification of the steel continuous casting process
is the geometry of the strand. To study the effect of this geometry, the influence of slab thickness will be in-
vestigated. To this end, a thermal model is first proposed, and its reliability is verified by simulating another
paper in the literature and comparing the results with that research. In the model of the present work, thermo-
physical properties are calculated based on the computational thermodynamics model, CALPHAD technique.
Then three different thicknesses are chosen subject to the same cooling conditions and technological parame-
ters. Afterward, the metallurgical length and shell thickness for these thicknesses are compared. As the shell
thickness is approximated by a square root function of time, holding the coefficient K, finally, the K factor of
the mentioned thicknesses is extracted and compared with one another such that the higher the thickness, the
higher the K coefficient.

Keywords : Metallurgical length, Heat transfer simulation, Continuous slab casting, Slab Thickness, Stephan
problem, Coefficient K.

1. Introduction partially, and a thin solid shell forms. Then the strand
gradually moves down to the secondary cooling zone
In the field of materials engineering, steel continuous (SCZ) and later toward the torch cut-off point. In the
casting (CC) has surpassed the ingot casting process in SCZ, cooling water sprays sprinkle water droplets onto
recent decades, contributing to higher production rates, the surface of the strand and extract heat energy from it
fewer casting defects, less energy waste, and higher qual- until the last melt droplet is solidified, or the cross-sec-
ity of the semi-finished product. In vertical CC of steel, tion becomes completely solid (see Fig. 1a). The distance
which is widely used around the world, the molten steel from this point to the meniscus is called the metallurgical
is poured from a ladle into a tundish, and then from the length (or liquid pool) [1, 2]. To be more specific, ac-
tundish into the primary cooling zone (water-cooled bot- cording to Fig. 1a, the yellow arc-like region inside the
tom-opened mold). It should be noted that the top level strand is the liquid pool, the tip of which is the last melt
of melt in the mold is called the meniscus (see Fig. 1a). droplet to be solidified. The metallurgical length equals
From the meniscus to the mold’s exit, the melt solidifies the length of the arc on which the liquid pool is outlined
from the tip of the liquid pool to the meniscus. Finally, at
*Corresponding author the torch cut-off point, the strand is cut to be stored in the
Email: [email protected] storehouse (for more details of the process, see Fig. 1a).
Address:Department of Materials Science and Engineering, One of the most important parameters in the CC pro-
Sharif University of Technology, Tehran, P.O. Box 11365- cess is the quality of the final product which is affected
9466, Iran.
by quite a few factors, especially the geometry of the
1. M.Sc

67
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

strand, metallurgical parameters, and technological pa- the feedforward-feedback model and closed feedback
rameters in the continuous casting machine (CCM). To loop model, respectively are proposed to minimize
illustrate further, the geometry (especially thickness) the temperature fluctuation due to the velocity change.
and heat transfer parameters such as the temperature of In addition, works such as [6] and [7] have developed
cooling water sprays and the water flow rate in SCZ can open-loop control models. During recent years, the pro-
enormously influence the existence of defects both in portional-integral-derivative (PID) control algorithm
products and in the machine. Thanks to the increasing has become popular for regulating the water flow rate
computing power of computers, it has become feasible of sprays in SCZ and bringing down temperature distur-
to simulate and monitor heat transfer conditions in the bances, as seen in [4] and [8-13]. In [14], an alternating
CC process and to control both the shell thickness of the direction implicit algorithm (ADI), was used to maintain
solidified shell and the defects. In the remainder of this the surface temperature of the billet. In control methods,
section, several models based on this computing pow- the high-performance computing power of computers is
er will be reviewed. In the CC process, as the ladle is essential, but when computers have limited computing
changed, the casting velocity must be reduced and then power and low storage memory, offline models should
raised. As a result, it is highly probable that defects such be used to design the CC process and its variables. Spe-
as cracks and porosities, to name just two, may form. Ad- cifically, optimization models used to design the CC
ditionally, this change affects the size of the metallurgical process are based on the inverse heat transfer problem
length and the strand temperature. Indeed, the metallurgi- (IHTP). Reference (or target) temperatures are measured
cal length, as well as the temperature field of the strand, in the CC plant, and are used as target temperatures in
should be fixed during any change in the CCM because optimization algorithms of IHTP to determine the opti-
an increase or decrease may culminate in the formation of mal values of heat transfer parameters in the CC process.
breakout, porosity, cracks, and other defects. Therefore, With regards to these models, there are a great number of
controlling and monitoring the above-mentioned param- works in the literature; [4, 15, 16] the have applied genet-
eters has been a matter that researchers have investigated ic algorithms in IHTP to minimize transversal, internal,
in their research studies. For this purpose, and longitudinal cracks, as well as segregation. In [17],
control-based models (called real-time or online a genetic algorithm was used to calculate the heat trans-
models) and optimization models (also called of- fer coefficient in the mold and SCZ, but the boundary
fline models) have been proposed. In [4] and [5], condition that describes the target temperature is a

(a) (b)
Fig. 1. (a) Schematic outline of continuous casting process [3], and (b) 2D computational domain (thickness-length
cross-section from the meniscus to torch cutoff point) used in this study for the simulation of heat transfer in steel slab.

68
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

mathematical function. To maintain, the temperature 2) The slab is assumed to be vertical without any curve
field during any change in CCM, by setting corner and/ or bending; therefore, plastic strain or elastic strain is not
or central temperatures as target points, other algorithms considered in the model.
have been used such as: Nelder-Mead algorithm [18], 3) There is no roller in contact with the strand.
multi-objective optimization algorithm [19], particle 4) The length of the slab (along the casting direction)
swarm optimization (PSO) algorithm [20], predictive is higher than its thickness and width. The width of the
GPU-based weighted least square model [21], neural net- strand is larger than the thickness, and as a result, the
work (NN) which showed that the PID method is less ef- thermal gradient along the width direction is ignored in
ficient [22], artificial intelligence heuristic search models the heat transfer equation (simulation will be conducted
[23, 24], modified proximal bundle algorithm [25], Broy- just for the length-thickness cross-section).
den–Fletcher–Goldfarb–Shanno (BFGS) algorithm [26], 5) The effect of fluid flow is ignored.
Levenberg−Marquardt [21, 27, 28], and sequential qua- 6) The casting velocity is assumed to be constant (but
dratic programming (SQP) [29]. In [30], the influence of slab thickness is changed, and its effect will be studied
changing casting velocity on thermomechanical behavior in this study).
while maintaining metallurgical length was studied. To be 7) The density, thermal conductivity, and specific heat
more precise, the method used in [30] is based on fixing capacity for both liquid and solid phases are not assumed
transported heat energy to maintain a constant metallurgi- to be identical or constant. Instead, the thermophysical
cal length for different casting velocities with heat trans- properties are extracted directly as a function of tempera-
fer coefficients calculated according to an empirical rela- ture based on a computational thermodynamics method
tion. As previously mentioned, geometry is an important called CALPHAD approach, which stems from global
factor that influences the temperature field, metallurgical minimization of Gibbs energy. More details are described
length, and the quality of the strand, but there have been in section 3.
no studies on the effect of strand thickness in the literature. The correspondent computational domain of the
In the present work, we study the effect of different strand is denoted as Ω , and its boundary, ∂Ω , is di-
strand thicknesses on the metallurgical length, tempera- vided into four partitions: lateral boundary (or cooling
ture field, and solidification of a low-carbon steel grade. region), Γl , top boundary (or meniscus), Γ t , bottom
To elaborate, section 2, expresses the mathematical mod- boundary (or torch cut-off point), Γb , (in other words,
el used in this study will be expressed. In section 3, the ∂Ω=Γl∪Γt∪Γb), see Fig. 1b Following [4, 29, 31-35], the
mathematical model is implemented and validated us- quasi-steady state heat transfer equation based on the en-
ing a reference including the simulation of the low-car- thalpy method is as follows:
bon steel. The thermophysical properties are obtained
from a thermodynamic computation method called  cpe V . T .(k  T ) in  Eq. (1)
CALPHAD2. Next, three different thicknesses are as-
sumed, and the simulation is run for these thicknesses.  In T Tp on t Eq. (2)
section 4, the results will be discussed, showing that the Eq. (3)
metallurgical length computed using the proposed sim- k T .n 0 on b
ulation model conforms well with an analytical square k T . n h (T  Ti ) on l Eq. (4)
root formula with a coefficient. Furthermore, as this for-
mula is commonly used in continuous casting plants to Where V⃗ , T, k, cpe, ρ, Ti , h, n⃗ and Tp denote casting
predict metallurgical length easily, the coefficient will be velocity, temperature, thermal conductivity, effective
calculated for each test case, and their correlation with specific heat capacity, density, cooling water tempera-
slab thickness will be discussed and investigated. ture, heat transfer coefficient, outward unit normal, and
pouring temperature, respectively  (both Ti and h vary
2. Mathematical Description spatially). The casting velocity, V = (0, -u ) , is such that

As is common in the literature, the heat transfer equa- the value of u > 0, and the strand is cast along the - y
tion is applied for thermal simulation. The heat transfer direction (see Fig. 1b). In many works, the effective spe-
equation is used in the present work to simulate solid- cific heat capacity, c pe , (using the enthalpy method) is
ification and the distribution of temperature in a slab. obtained as follows:
Beforehand, several assumptions are considered in this df (T )
cpe cp  Lf l Eq. (5)
study as follows: dT
1) The process is assumed to be a in quasi-steady state;
Where fl(T), cp, and Lf denote the liquid volume frac-
thus, the quasi-steady state heat transfer equation will be
solved because it is computationally less costly. tion, heat capacity, and latent heat of fusion, respectively.
Although liquid volume fraction, fl(T), and latent heat of
fusion, Lf, as well as other thermophysical properties, are
2 CALculation of PHAse Diagrams simply interpolated and assumed constant in some re-

69
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

search studies in the literature (for example [31]), these idus temperature, liquidus temperature, water flow rates
thermophysical properties should not be simplified as of nozzles in SCZ, and cooling water temperature of the
constant or as an explicit function of temperature because nozzles are extracted from reference [37], see Table 1 and
in practice, phenomena such as solid-state phase trans- Table 2. lz, lx, ly, lm, nx, ny, u, Tl, and Ts, respectively de-
formation, the peritectic reaction in peritectic grades, note the width of the slab, the thickness, machine length
and so on, occur. Because of these restrictions, to con- (the length of the slab, precisely from the meniscus to the
sider the effects of these phenomena, the thermophys- torch cutoff), the mold length, mesh size on the x⃗ axis,
ical properties are calculated based on the CALPHAD the mesh size on the y⃗ axis, the value of casting velocity,
technique which is explained in more detail in the next liquidus temperature, and solidus temperature. Also, the
section [36]. chemical composition of the low-carbon steel grade used
in this study is listed in Table 3. The thickness of test case
3. Numerical Implementation #1 (which is simulated in this subsection), as well as test
cases #2 and #3 (which will be studied in the next sub-
In this section, the presented model is implemented section) are listed in Table 4. As far as heat transfer on the
using a reference in the literature, and our simulation is interphase between the mold and the strand is concerned,
conducted for a thickness of 23cm, and the results of our according to reference [37], the following heat flux is used
simulation are compared with those of the reference to
validate the simulation model. Then, the proposed model  q 2400000  346000 lm u Eq. (6)
is applied for thicknesses of 18cm and 28cm. Henceforth,
the simulations supposed to be implemented for thick- Where q denotes heat flux.
nesses 23cm, 18cm, and 28cm are symbolized as test It should be noted that in [37] only the water volume
case #1, test case #2, and test case #3, respectively. in SCZ is given. However, in this study using Nozaki’s
To find the temperature distribution of test case #1, relation [38], the heat transfer coefficients are computed
a heat transfer simulation based on the mathematical in each segment of SCZ as Eq. (7):
model described previously will be conducted. First, the 1570 w0.55 (1  0.0075 Ti )
thermodynamic properties are extracted from the CAL- h Eq. (7)

PHAD model using the JMatPro package, see Fig. 2. The
CALPHAD approach is an interesting thermodynamic Where w and β denote the water flow rate and the ma-
technique in the field of materials science and engineer- chine-dependent calibration factor. The water volume and
ing that can predict thermophysical properties as a func- machine-dependent calibration factor are extracted from
tion of temperature. In fact, given an alloy with a definite [37], listed in Table 5. By the given water volume, wa-
chemical analysis, it is possible to globally minimize ter flow rate, w, can be obtained; then, using the relation
the free energy by employing a database including free (7), the heat transfer coefficient in each segment of SCZ
energies of each phase. As a result, this method paves is calculated. The computed heat transfer coefficient for
the way to calculate thermodynamics equilibria such as all spray segments is symbolized as hworking hereafter. The
phase diagrams and thermophysical properties for com- heat transfer coefficient and cooling water temperature (of
mercial steel grades as well as for multi-component al- cooling nozzles in SCZ) are the same (as Ti) for all three
loys [36]. Then the geometry, pouring temperature, sol- cases.

Table 1. Geometry, mesh size parameters, casting velocity, and cooling water temperature of nozzles used in this study
and [37].

𝑙𝑙𝑧𝑧 (m) 𝑙𝑙𝑥𝑥 (cm) 𝑙𝑙𝑦𝑦 (m) 𝑙𝑙𝑚𝑚 (m) 𝑛𝑛𝑥𝑥 𝑛𝑛𝑦𝑦 u (m/min) 𝑇𝑇𝑖𝑖 (oC)
1.25 23 30 0.9 40 100 1.1 25

Table 2. Pouring Temperature, solidus Temperature, and liquidus

Temperature of the steel grade used in this study [37].

𝑇𝑇𝑙𝑙 (oC) 𝑇𝑇𝑠𝑠 (oC) 𝑇𝑇𝑃𝑃 (oC)

1519 1453 1577

70
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

Fig. 2. Thermophysical properties of the steel grade used in this study. All properties are based on the CALPHAD
method and extracted from the JMatPro package.

Table 3. Chemical analysis (in percent) of the low carbon steel grade studied both in [37] and in the present work.
Fe C Si Mn S P Al N
Base 0.15 0.15 0.25 0.015 0.02 0.002 0.016

71
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

Table 4. The thicknesses of three test cases studied in the present work.
Test case NO. Thickness (cm)
#1 23
#2 18
#3 28

Table 5. The length, water volume, and calibration factor of each spray cooling segment along SCZ [37].
Sprays Length (m) Water volume (L⁄min) Calibration factor
segment 1 2.96 367.8 4
segment 2 1.819 343.7 4.23
segment 3 1.819 293.9 4.78
segment 4 3.789 195.2 2.78
segment 5 6.094 260.3 2.89
segment 6 6.945 225.5 2.62
segment 7 4.62 140.6 3.06
segment 8 2.31 60 2.89

3.1 Model Validation Furthermore, Fig. 4 shows a close approximation of

the calculation of shell thickness among test case #1 in
By conducting the heat transfer simulation using the this study, simulation in [37], and measurement using an
data from the previous section for test case #1, the results experiment in [37]. The results of test case #1, shown in
will be explained in the following. In Fig. 3, the compar- Fig. 3 and Fig. 4 confirm the acceptable accuracy of the
ison of surface temperature and centerline temperature presented simulation model in this study. Next, using the
between research [37] and the simulation of the model in applied data in the simulation of test case #1, simulation
test case #1 in this study shows a reasonable and accept- is implemented for both cases by reducing and raising
able agreement between the simulation of the thermal the slab thickness to 18cm (test case #2) and 28cm (test
model in this study and research [37]. According to Fig. case #3), respectively and the results will be studied in
4 and the red graph as well as the blue graph in Fig. 3, the next section.
in our simulation, the calculated metallurgical length of
test case #1 equals 20.0317653m, which is close to that 4. Results and Discussion
obtained in [37] as 19.9m. In Fig. 4, the black profile is
the result of the simulation in the present work, and green This section aims to evaluate the performance and
and orange points are data presented in [37]. Fig. 4 shows functionality of the presented model. As per the previ-
that the shell thickness at three different distances from ous sections, the thicknesses of 18cm and 28cm corre-
the meniscus is approximately close to the simulated pro- spond to test case #2 and test case #3, respectively. Ad-
file in the present work. The difference in shell thickness ditionally, the interfacial boundary between the slab and
between our simulation and experimental results is 6%. both the mold and SCZ is divided into two partitions;
In addition, there is a little but acceptable difference be- in other words, Γl = Γlm ∪ Γls where Γlm is the contact re-
tween the surface temperature distribution simulated in gion between the slab and mold, and Γls is the contact
this study and that of [34] in the worst case of 7% (see region between the slab and SCZ. The heat transfer in
black and green graphs of Fig. 3). This difference is due these regions is the same as that in test case #1. In the
to the fact that CALPHAD thermodynamic calculations following, shell thickness (or solid) profiles, metallur-
mentioned in the previous section are used in this study, gical length, temperature distribution, and liquid vol-
but constant values for thermophysical properties are ap- ume fraction distribution will be discussed. Moreover,
plied in research [37]. the constant coefficient K in the square root function
mentioned at the end of section 1 will be determined.

72
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

1595

1495

1395 Centerline: [37]

Temperature [°C]

1295 Centerline: The present work

1195 Surfece: [37]

1095 Surface: The present work

995

895

795
0 5 10 15 20 25 30 35
Distance from meniscs [m]
Fig. 3. Comparison of simulated centerline temperature profiles (red and
blue) and simulated surface temperature profiles (black and green) be-
tween the present work and [37] for test case #1 with respect to distance
from the meniscus.

14

12

10
Shell thickness [cm]

6
Test case #1 in the present work: thickness 23cm
4
Experimental in [37]: thickness 23cm
2 Calculated in [37]: thickness 23cm

0
0 5 10 15 20 25 30 35
Distance from meniscus [m]

Fig. 4. Comparison of Shell thickness with respect to distance from

the meniscus. The black curve corresponds to test case #1 simulation
in this study; both the experimental data (orange points) and simulat-
ed data (green points) are extracted from [37].

73
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

4.1 Shell thickness 4.3 Analysis of Square Root Function

In Fig. 5, the variation of shell thickness as a func- In the literature, it is widely agreed that the solid pro-
tion of distance from the meniscus for all test cases #1, file of the strand approximately conforms to a square root
#2, and #3 is shown. According to the plot, raising the function. In fact, it is assumed that the solidification front
thickness of the slab from 18cm to 28cm, increase the is moving interphase between liquid and solid phases.
shell thickness profile correspondingly, especially from Considering the solidification front as a moving boundary,
segment 3 to the torch cut-off point; however, their dif- a two-dimensional Stephan problem has been solved in the
ference is minimal near the meniscus and segments literature [6,39-41], contributing to an analytical formula
1 and 2. for the calculation of metallurgical length as follows:

D
4.2 Metallurgical Length  t or xs K
xs K Eq. (8)
u
Using hworking, the metallurgical length for test cas- Where xs, D, and t denote the shell thickness, the
es #2 and #3 is approximately 12.5m and 28.78m, distance from the meniscus, and time, respectively. This
see Figs. 6c and 6a, respectively. With regards to test criterion expresses the approximate position of the sol-
cases #2 and #3, in which the thickness is reduced id shell; more importantly, the formula can determine
and raised respectively, the metallurgical length ap- the size of the metallurgical length. It is worth noting
proaches 12.5m and 28.78m in order. Additionally, that this formula is widely used to compute metallur-
Fig. 6 shows the contour plots of the temperature gical length and the location of segregation (after the
field corresponding to the liquid volume fraction traditional Baumann sulfur printing test) in the industry.
shown in Fig. 6. Accordingly, it is evident that the The most significant challenge for casting plants is de-
metallurgical length depends on the slab thickness termining the K factor in their casting case because K
such that by raising the slab thickness, the metallur- depends on machine variables, chemical composition,
gical length also increases. and the geometry of the strand, among others.

16

14

12
Shell thickness [cm]

10

8
Test case #3, thickness: 28cm
6
Test case #2, thickness: 18cm
4 Test case #1, thickness: 23cm

0
0 5 10 15 20 25 30 35
Distance from meniscus [m]
Fig. 5. Simulated shell thickness with respect to distance from meniscus for test case
#2 (18cm; green), for test case #1 (23cm; red), and for test case #3 (28cm; blue) in
this study.

74
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

(a) (b) (c)

Fig. 6. Contour plots of liquid volume fraction at the end of simulation: a) test case #3 b) test case #1, and c) test case
#2 in this study.

(a) (b) (c)

Fig. 7. Contour plots of temperature distribution at the end of simulation: a) test case #3 b) test case #1, and c) test case
#2 in this study.

75
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

In this section, coefficient K is extracted using the simulated solid profile, the coefficient K is determined
mathematical curve fitting technique for test cases #1, to be K23=26.8370 m. min-0.5 in order to predict the met-
#2, and #3, and the correlation between slab thickness allurgical length for test case #1. The relative absolute
and the coefficient K will be studied. In fact, K is deter- error equals 11 % which is acceptable because the size
mined such that criterion (8) predicts the metallurgical of the metallurgical length, not the exact solid profile in
length for continuous casting. To this end, the coefficient different positions, is important. Fig. 9, shows the shell
K will be calculated using the metallurgical length ob- thickness with respect to distance from the meniscus for
tained from shell thickness profiles. Regarding test case test case #2 (18cm); similar to test case #1, using the sim-
#1 (23cm), Fig. 8 shows the shell thickness with respect ulated solid profile (black curve) and xs function (red),
to the distance from the meniscus for both the simulated the coefficient K is calculated K18= 26.5226 m. min-0.5.
(black curve) and the xs function (red curve). Using the The relative absolute error equals 13% in test case #2.

0.14

0.12
𝐷𝐷
𝑥𝑥𝑠𝑠 = 26.837√
0.1 1.1
Shell thickness [m]

0.08

0.06
Simulated solid profile: 23cm
0.04
Squar root function: 23cm
0.02

0
0 5 10 15 20 25 30 35
Distance from meniscus [m]
Fig. 8. Simulated shell thickness with respect to distance from meniscus (black) and fitted square root function (red) for
test case #1 (23cm) in this study.

0.1
0.09 𝐷𝐷
𝑥𝑥𝑠𝑠 = 26.5226
0.08 1.1
0.07
Shell thickness [m]

0.06
0.05
Simulated solid profile: 18cm
0.04
0.03 Square root function: 18cm
0.02
0.01
0
0 5 10 15 20 25 30 35
Distance from meniscus [m]
Fig. 9. Simulated shell thickness with respect to distance from meniscus (black) and fitted square root function (red) for
test case #2 (18cm) in this study.

76
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

Finally, Fig. 10 shows the simulated shell thickness cation of cooling parameters. It is crucial for steel con-
as a function of the distance from the meniscus and the tinuous casting plants to predict the shell thickness,
xs function. Similar to the previous test cases, the coeffi- metallurgical length, and the K factor in the square root
cient K is computed as K28= 27.6252 m. min-0.5 and the function, used to predict metallurgical length in steel
relative absolute error is approximately 11%. continuous slab casting as slab thickness varies. In this
By comparing the thicknesses and their correspond- study, a two-dimensional heat transfer model coupled
ing calculated factors in Table 6, it is obvious that by in- with the CALPHAD model to determine thermophysical
creasing the slab thickness, the coefficient K in Eq. (8), properties we used to investigate the effect of three slab
which determines the metallurgical length, rises propor- thicknesses on the metallurgical length and shell thick-
tionally. ness. Then the values of the K factor for each thickness
were extracted using curve fitting technique. According-
5. Conclusions ly, the higher the slab thickness, the longer the metal-
lurgical length. In order words, the coefficient K in the
Steel continuous casting process is sensitive to the square root function increases as far as the slab thickness
change in slab thickness, which can lead to the justifi- grows.

0.16

0.14

0.12 𝐷𝐷
𝑥𝑥𝑠𝑠 = 27.6252
Shell thickness [m]

0.1 1.1

0.08

0.06
Simulated solid profile: 28cm
0.04

0.02 Sqaure root function: 28cm

0
0 5 10 15 20 25 30 35
Distance from meniscus [m]
Fig. 10. Simulated shell thickness with respect to distance from
meniscus (black) and fitted square root function (red) for test case #3
(28cm) in this study.

Table 6. The calculated coefficient K, metallurgical length, and thick-

ness for test cases in this study.

Test case NO. Thickness (cm) Coefficient K (m. min-0.5) Metallurgical length (m)
#2 18 26.5226 12.50
#1 23 26.8370 20.03
#3 28 27.6252 28.78

77
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

Acknowledgments

The author would like to thank Professor Ruohollah Tavakoli for suggesting the writing of the present work.

Nomenclature
c pe Specific heat capacity

D Distance from meniscus

fl Liquid volume fraction

h Heat transfer coefficient

K The coefficient in the square root function of time and shell thickness
k Thermal conductivity
Lf Latent heat of fusion

lm The thickness of the mold

lx The thickness of the slab

ly The length of the slab

n The outward unit normal vector on a surface

nx Mesh size index along the thickness

ny Mesh size along the length

q Heat flux

T Temperature
Ti Cooling water temperature

TP Pouring temperature
t Time
u The absolute value of casting speed

V Casting velocity vector

w Water flow rate
xs Shell thickness

 Calibration factor of continuous casting machine

b The bottom boundary of the computational domain

l The lateral boundary of the computational domain

t The top boundary of the computational domain

x The size of the computational cell along the thickness direction

y The size of the computational cell along the length direction
 Density

 The computational domain corresponding to the slab

78
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

Reference Real-time Simulation and Control of the Continuous

Casting Process, In: J. Manley, S. McKee, D. Owens
[1] J. Hejazi, Ingot Casting, Irainan Foundarymen Soci- (eds), Proceedings of the Third European Conference on
ety, 1982, (in Persian). Mathematics in Industry, Springer-verlag Berlin Heidel-
[2] B. Petrus, D. Hammon, M. Miller, B. Williams, A. berg GmbH, 1990, pp.401–408.
Zewe, Z. Chen, J. Bentsman, B. Thomas, Newmethod to [13] L. Guo, Y. Tian, M. Yao, H. Shen, Temperature dis-
measure metallurgical length and application to improve tribution and dynamic control of secondary cooling in
computational models, Iron and Steel Technology Con- slab continuous casting, International Journal of Miner-
ference and 7th International Conference on the Science als, Metallurgy and Materials. 16 (2009) 626-631. https://
and Technology of Ironmaking, Cleavlad , USA, 2015. doi.org/10.1016/S1674-4799(10)60003-9.
[3] X. Huang, B. Thomas, Modeling of steel grade tran- [14] M. Jauhola, E. Kivela, J. Konttinen, E. Laitinen, S.
sition in continuous slab casting processes, Metallur- Louhenkilpi, Dynamic secondary cooling model for a
gical Transactions B. 24 (1993) 393-379. https://doi. continuous casting machine, Proceeding 6th Internation-
org/10.1007/BF02659140. al Rolling Conference, Dusseldorf, Germany, 1994.
[4] C. Santos, J. Spim, A. Garcia, Mathematical model- [15] Y. Zhai, Y. Li, B. Ma, C. Yan, Z. Jiang, The optimi-
ing and optimization strategies (genetic algorithm and sation of the secondary cooling water distribution with
knowledge base) applied to the continuous casting of improved genetic algorithm in continuous casting of
steel, Engineering Applications of Artificial Intelligence. steels. 19 (2015) 26-31. https://doi.org/10.1179/143289
16 (2003) 511-527. https://doi.org/10.1016/S0952- 1715Z.0000000001362.
1976(03)00072-1. [16] K. Worapradya, P. Thanakijkasem. Optimum spray
[5] M. Long, D. Chen, J. Zhang, Q. Ouyang, Novel on- cooling in continuous slab casting process under produc-
line temperature control system with closed feedback tivity improvement, IEEE International Conference on
loop for steel continuous casting, Ironmaking & Steel- Industrial Engineering and Engineering Management,
making. 38 (2011) 620-629. https://doi.org/10.1179/174 Hong Kong, 2009.
3281211Y.0000000042. [17] D. Słota, Identification of the cooling condition in
[6] L. Klimeš, J. Štětina, A rapid GPU-based heat trans- 2-D and 3-D continuous casting processes, Numerical
fer and solidification model for dynamic computer sim- Heat Transfer Part B: Fundamentals. 2 (2009) 155-176.
ulations of continuous steel casting, Journal of Materi- https://doi.org/10.1080/10407790802605232.
als Processing Technology. 226 (2015) 1-14. https://doi. [18] B. Filipic, E. Laitinen, Model-based tuning of pro-
org/10.1016/j.jmatprotec.2015.06.016. cess parameters for steady-state steel casting, Informati-
[7] S. Louhenkilpi, M. Mäkinen, S. Vapalahti, T. Räisänen, ca an international journal of computing and informatics.
J. Laine, 3D steady state and transient simulation tools 29 (2005) 2005 491-496.
for heat transfer and solidification in continuous casting, [19] K. Cho, B. Kim, Numerical analysis of secondary
Materials Science and Engineering: A. 413(2005) 135- cooling in continuous slab casting, Journal of Materials
138. https://doi.org/10.1016/j.msea.2005.08.153. Science and Technology. 24(2008) 389-390. https://doi.
[8] K. Zheng, B. Petrus, B. G. Thomas, J. Bentsman, org/10.1016/S0924-0136(01)00654-9.
Design and implementation of a real-time spray cooling [20] F. Camisani-Calzolari, I. Craig, P. Pistorius, Specifica-
control system for continuous casting of thin steel slabs, tion framework for control of the secondary cooling zone
Proceeding AISTech Steelmaking Conference, Indianap- in continuous casting, ISIJ international. 38 (1999) 7131-
olis, 2007. 7136. https://doi.org/10.2355/isijinternational.38.447.
[9] J. Yang, Z. Xie, Z. Ji, H. Meng, Real-time heat trans- [21] J. Zhang, D. Chen, C. Zhang, S. Wang, W. Hwang,
fer model based on variable non-uniform grid for dynam- Dynamic spray cooling control model based on the track-
ic control of continuous casting billets, ISIJ international. ing of velocity and superheat for the continuous casting
54 (2014) 328-335. https://doi.org/10.2355/isijinterna- steel, Journal of Materials Processing Technology. 229
tional.54.328. (2016) 651-658. https://doi.org/10.1016/j.jmatpro-
[10] B. Petrus, K. Zheng, X. Zhou, B. Thomas, J. Bents- tec.2015.10.015.
man, Real-time, model-based spray-cooling control [22] N. Cheung, A. Garcia, The use of a heuristic search
system for steel continuous casting, Metallurgical and technique for the optimization of quality of steel bil-
materials transactions B. 42 (2011) 87-103. https://doi. lets produced by continuous casting, Engineering Ap-
org/10.1007/s11663-010-9452-7. plications of Artificial Intelligence. 14 (2001) 229-238.
[11] T. Männikkö, E. Laitinen, P. Neittaanmäki, Re- https://doi.org/10.1016/S0952-1976(00)00075-0.
al-time simulation and control system for the continuous [23] D. Van der Spuy, I. Craig, P. Pistorius, An optimi-
casting process, in: H.-H. Sebastian, K.Tammer (eds), zation procedure for the secondary cooling zone of a
System Modelling and Optimization, Springer-verlag continuous billet caster, Journal of the Southern African
Berlin Heidelberg GmbH, 1990, pp.809-817. Institute of Mining and Metallurgy. 99 (1999) 49-54.
[12] E. Laitinen, P. Neittaanmäki, T. Männikkö, On the https://hdl.handle.net/10520/AJA0038223X_2613.

79
 A. Pourfathi / International Journal of ISSI, Vol. 19(2022), No.1, 67-80

[24] T. Mauder, C. Sandera, J. Stetina, Optimal control org/10.1007/s00231-011-0822-8.

algorithm for continuous casting process by using fuzzy [33] E. Majchrzak, Numerical simulation of continuous
logic, Steel Research International. 86 (2015) 785-798. casting solidification by boundary element method, En-
https://doi.org/10.1002/srin.201400213. gineering Analysis with Boundary Elements. 11 (1993)
[25] Y. Wang, X. Luo, Y. Yu, Q. Yin, Evaluation of 95-99. https://doi.org/10.1016/0955-7997(93)90028-J.
heat transfer coefficients in continuous casting under [34] Z. Han, D. Chen, K. Feng, M. Long, Development
large disturbance by weighted least squares Leven- and application of dynamic soft-reduction control model
berg-Marquardt method, Applied Thermal Engineering. to slab continuous casting process, ISIJ international. 50
111 (2017) 989-996. https://doi.org/10.1016/j.applther- (2010) 1637-1643. https://doi.org/10.2355/isijinterna-
maleng.2016.09.154. tional.50.1637.
[26] Y. Yu, X. Luo, Estimation of heat transfer coeffi- [35] M. Alizadeh, A. J. Jahromi, O. Abouali, A new
cients and heat flux on the billet surface by an integrated semi-analytical model for prediction of the strand surface
approach, International Journal of Heat and Mass Trans- temperature in the continuous casting of steel in the mold
fer. 90 (2015) 645-653. https://doi.org/10.1016/j.ijheat- region, ISIJ international. 48 (2008) 161-169. https://doi.
masstransfer.2015.07.008. org/10.2355/isijinternational.48.161.
[27] T. Männikkö and M. Mäkelä, Nonsmooth penalty [36] A. Pourfathi, R. Tavakoli, Thermal optimization of
techniques in control of the continuous casting process, secondary cooling systems in the continuous steel cast-
in: P. Neittaanmaki (eds), Numerical Methods for Free ing process, International Journal of Thermal Sciences.
Boundary Problems, Springer-Basel AG, 1991, pp.297- 183 (2023) 107860. https://doi.org/10.1016/j.ijthermals-
307. https://doi.org/10.1007/978-3-0348-5715-426. ci.2022.107860.
[28] B. Lally, L. Biegler, H. Henein, Optimization and [37] Y. Yu, X. Luo, H. Y. Zhang, Q. Zhang, Dynamic op-
continuous casting: Part II Application to industrial cast- timization method of secondary cooling water quantity in
ers, Metallurgical Transactions B. 22 (1991) 649-659. continuous casting based on three-dimensional transient
https://doi.org/10.1007/BF02679020. nonlinear convective heat transfer equation, Applied
[29] S. Louhenkilpi, E. Laitinen, R. Nieminen, Real-time Thermal Engineering. 160 (2019) 113988. https://doi.
simulation of heat transfer in continuous casting, Metal- org/10.1016/j.applthermaleng.2019.113988.
lurgical Transactions B. 24 (1993) 685-693. https://doi. [38] S. Chaudhuri, R. Singh, K. Patwari, S. Majumdar,
org/10.1007/BF02673184. A. Ray, A. Singh, N. Neogi, Design and implementation
[30] M. Bellet, L. Salazar-Bbetancourt, O. Jaouen, F. of an automated secondary cooling system for the contin-
Costes, Modelling of water spray cooling Impact on ther- uous casting of billets, ISA transactions. 49 (2010) 121-
momechanics of solid shell and automatic monitoring to 129. https://doi.org/10.1016/j.isatra.2009.09.005.
keep metallurgical length constant, European continuous [39] S. Lalitha, S. Chattopadhyay, S. Das, K. Godiwal-
casting conference (8th ECCC), Austrian society for met- la, Simulation of heat transfer in the continuous casting
allurgy and materials, 2014. mold, Transactions of the Indian Institute of Metals. 44
[31] R. Tavakoli, Smooth modeling of solidification (1991) 89-92.
based on the latent heat evolution approach, The Interna- [40] J. Dantzig, Ch. Tucker, Modeling in materials pro-
tional Journal of Advanced Manufacturing Technology, cessing, Cambridge university press, 2001.
88 (2017) 3041-3052. https://doi.org/10.1007/s00170- [41] K. Spitzer, K. Harste, B. Weber, P. Monheim, K.
016-9012-7. Schwerdtfeger, Mathematical model for thermal tracking
[32] M. Sadat, A. H. Gheysari, S. Sadat, The effects of and on-line control in continuous casting, ISIJ interna-
casting speed on steel continuous casting process, Heat tional. 32 (1992) 848-856. https://doi.org/10.2355/isijin-
and mass transfer. 47 (2011) 1601-1609. https://doi. ternational.32.848.

80