a r t i c l e i n f o
a b s t r a c t
Article history:
Received 20 August 2013
Received in revised form
16 June 2014
Accepted 24 June 2014
Available online
The 3-D numerically simulated steady state results of an industrial size vertical Direct Chill (DC) slab
caster tted with a metallic-foam-plated combo bag melt distributor for AA-1050 aluminum alloy are
reported. The turbulence in the melt and the mushy region solidication of the alloy are modeled
through the popular low Reynolds number ke model of Launder and Sharma and the enthalpy-porosity
scheme, respectively. The transport of melt through the foam is modeled using the BrinkmaneForchheimer extended Darcy equation. The verication of the numerical model is performed by comparing the
predicted and measured solidication front for AA-3104 rolling ingot reported in the literature. The inlet
melt superheat and the porosity of the metallic foam of the combo-bag were kept xed at 32 C and 0.9,
respectively. Parametric studies are carried out by varying two important parameters of the process, viz.,
the casting speed and the heat transfer coefcient (HTC) at the metal-mold contact region. Specically,
the casting speed and the HTC are varied from 40 to 100 mm/min and from 750 to 3000 W/(m2-K),
respectively. With the increase of the casting speed, the solidication process is delayed while the solidshell thickness increased with the increase of the HTC. The predicted results are presented for the solidshell thickness, the sump depth, the mushy thickness, the surface temperature as well as the nondimensional turbulent viscosity contours. In addition the temperature elds and velocity proles with
streamlines are also provided.
2014 Elsevier Masson SAS. All rights reserved.
Keywords:
Modeling
Mushy-zone solidication
3-D DC casting
Rolling ingot
Aluminum alloy
Turbulent melt
Metallic foam
1. Introduction
The vertical direct chill casting, often referred to as DC casting
(DCC), is a semi-continuous casting process which is a relatively
efcient and economical process for casting large size rolling ingots
and round extrusion billets for nonferrous metals and their alloys.
Among the non-ferrous metals, aluminum is the most extensively
used metal spanning a wide range of applications including, in
aerospace, automotive, electronic, construction, etc. More than 90
pct of the large slabs and billets of aluminum alloys are nowadays
cast using this technology [1]. The cast slabs are rolled into plates,
sheets, strips and foils for various end uses. The round billets are
usually extruded in the form of bars, rods, pipes and tubes for
diversied practical uses. In this process, molten metal is poured
into a water cooled mold whose open bottom is initially kept closed
with a bottom block such that the mold wall(s) and the base bottom
* Corresponding author. Tel.: 1 514 398 2524, 1 514 924 8542 (mobile); fax: 1
514 398 4492.
E-mail addresses: Latifa.begum@mail.mcgill.ca (L. Begum), Mainul.hasan@
mcgill.ca, mnlhsn73@gmail.com (M. Hasan).
http://dx.doi.org/10.1016/j.ijthermalsci.2014.06.031
1290-0729/ 2014 Elsevier Masson SAS. All rights reserved.
69
70
2. Mathematical model
2.1. Model description
In this study, a rectangular rolling ingot having dimensions of
2500 mm 1730 mm 660 mm in the x, y, and z directions,
respectively, was considered as the physical model which is schematically represented in Fig. 1. A standard hot top mold having a
length of 210 mm was considered. The hot top consisted of an
insulated part of length 130 mm (not shown in Fig. 1) while the lower
part of 80 mm formed the active mold. Since the modeled domain
has a two-fold symmetry, hence only quarter (ash color region of
Fig. 1) of the domain was modeled to decrease the computational
time and associated costs. This gure also shows the origin and the
directions of the coordinate axes used. The melt was being delivered
to the mold through a newly proposed combo bag. The structure of
the combo bag is such that it formed a rigid channel which is closed
on the two rolling sides and the top while there are two openings
facing the two narrow sides of the caster. The bottom plate of the
combo bag is made of stainless steel porous foam so as to allow a
controlled delivery of the melt as well as to capture some of the solid
oxide inclusions that might had been generated in the launder/
trough assembly above the combo bag. Fig. 2 shows the enlarged
view of the newly designed combo-bag. A square-shaped nozzle
having a cross-sectional area of 900 mm2 supplies the melt into the
hot top from the launder/trough assembly. The nozzle walls, the
walls of the distributor bag, and the bottom plate were all assumed
to have a constant thickness of 5 mm [8] and were assumed to have
been made of stainless steel. The thermo-physical properties of
stainless steel are given in Table 1. The nozzle was immersed in the
liquid metal pool to a depth of 20 mm and placed at the center of the
distributor bag, and the whole nozzle-distribution bag assembly was
placed at the geometrical center of the caster. The dimensions of the
Fig. 1. Schematic of a vertical DC caster with the calculation domain represented by ash color for a submerged nozzle with a combo-bag.
Table 1
Physical properties of the stainless steel [14].
Properties (symbol)
Value (units)
0.036 (kW/m-K)
7850 (kg/m3)
0.690 (kJ/kg-K)
vrui F
v
vF
GF
vxi
vxi
vxi
SF
71
(1)
H h DH
(2)
Zh
ZT
dh
cp dT
(3)
Tref
href
For constant cp, and taking href 0 at Tref, the above equation can
be integrated to obtain the sensible heat as:
h cp T Tref
(4)
DH DHf fl
(5)
fl
8
>
>
>
>
<
T Tsolidus
whenTliquidus T Tsolidus 0Mushyregion
>
T
liquidus Tsolidus
>
>
>
:
0
whenT <Tsolidus 0Solidregion
(6)
72
Table 2
Physical properties of AA-1050.
Variable (units)
Value
0.231
0.90
396.4
657.0
646.0
16.2 104
2700
"
#
v rui H
v rui DH
v
m mt vh
vxi
Pr st vxi
vxi
vxi
(7)
The last term in Eq. (7) represents the latent heat exchange rate
due to solideliquid phase change resulting from the turbulent
convective ow. The above equation was suitably nondimensionalized and its nal form is listed in Table 3.
2.4.3. Modeling of uid ow in the mushy region
In this study, in order to solve the complex mushy uid problem,
it was assumed that the solidication is predominantly equiaxed
and the solid and liquid move at the same velocity in the mushy
region which is different from the casting speed and is not necessarily equal to the adjacent melt velocity. When the melt is
completely solidied, the shell moves at the casting speed. In reality, the transport phenomena that prevail in the mushy region are
quite complex and difcult to describe by macroscopic equations.
Because of the simplicity of the Darcy law for porous media, in the
current model this law is adopted. This law is based on the
Table 3
Summary of the non-dimensional governing equations for the melt region.
*
G*F
S*F
Equation
Continuity
U-Momentum
1
U
0
1
Re 1
m*t
V-Momentum
1
Re 1
m*t
1
Re 1
m*t
W-Momentum W
Kinetic energy k*
vU A* U U Grm h* h*
G*F vX
s
a
Re
i
Re2
*
* vV
v
A
vXi GF vXi Re V Vs
v
A* W W
vX
G*F vW
Re
s
vXi
i
v
vP
vX vXi
*
vP
vY
vP
vZ
D*k
G*
Re Re Sk*
1 skt
E*
Rate of energy * 1
1
* *
*2
m*t
Re f1 C1 G k* C2 f2 k* Re2 S*
Re 1 s
dissipation
*
vVDH *
vWDH *
vUDH
Energy
h* 1 1 m*t
vX vY
vZ
st
Re Pr
vUi
vXj
!
vU
vUi
vXj ;
vXij
!
E* 2m*t
m*
(8)
"
#
v rui uj
v
vui
m
ml mt
SU 0 ui uis
i
vxj
K
vxj
vxj
(9)
where the coefcient m/K0 decreases from a large value in the solid
phase to zero in the liquid phase. Consequently, the Darcy term (the
third term in the right hand side of Eq. (9)) vanishes as the liquid
fraction becomes one. It is to be noted that the Darcy term is an
adhoc sink term. The CarmaneKoseny equation is adopted for the
relation between the permeability and the liquid fraction.
m
C1 fl 2
K0
f3 q
(10)
.
C 180 d2
(11)
x
y
z
u
v
w
; Y ; Z ; U
; V
; W
; P*
D
D
D
uin
uin
uin
P
k
D
h
DH
m
; k* 2 ; * 3 ; h*
; DH*
; m*t t
DHf
DHf
m
ro u2in
uin
uin
(12)
where.
Re rumin D :
G* m*t
1
Re
K 0 vP
ui
rgxi
m vxi
v2 U i
vXj vXk
p
*
D*k 2 vvXki
!
v2 Ui
vXj vXk
p
v k* :
vXi
3:4
*
*2
*2
t
2
f2 1 0:3eRet ; A*
C * 1 f1 2
:
fl3 q
30
v rUi F*
v
vF*
G*F
S*F
vXi
vXi
vXi
.
i 1; 2; 3
(13)
The values of F*, and the associated denition of G*F and S*F , for
all of the transport equations are listed in Table 3. The development
of a general-purpose program is considerably simplied because of
the casting of the governing equations in a general form.
The non-dimensional form of the boundary conditions becomes:
[1] Nozzle inlet:
.
U 1; V W 0; h* h*in ; k* 0:01; * cm 0:013=2 0:05
(14)
vV vW vk* v* vh*
0;
vX
vX
vX
vX
vX
U0
(15)
73
(22)
.
x x1
g x ggap
gmax ggap kW m2 K
x2 x1
210 x < 220 mm
(23)
.
x x2
g
gmax kW m2 K
g x gmax
x3 x2 film
220 x < 260 mm
(24)
g x gfilm
x 260 mm
(25)
vU vV vk* v* vh*
0;
vZ
vZ
vZ
vZ
vZ
W0
(16)
vU vW vk* v* vh*
0;
vY
vY
vY
vY
vY
V 0
(17)
[4] Outlet:
vU vk* v* vh*
0
vX
vX
vX
vX
(18)
V W 0
(18a)
U Us us =uin ; V W k* * 0;
vh* vh*
gD *
hs h*a
K
vY
vZ
(19)
Adiabatic section : gx 0: kW
.
m2 K 0 x < 130 mm
(20)
(21)
.
.
ggap 0:15 kW m2 K ; gmax 20:0 kW m2 K ;
.
gflim 10:0 kW m2 K :
The air gap distance between the mold and the semi-solid shell
was assumed to be 0.4 mm which is within the range of
0.2e0.5 mm suggested by Prasad and Bainbridge [1]. It should be
noted here that in the present model the non-dimensionalized
transport equations and associated boundary conditions were
solved, which therefore generated non-dimensional results. To
make the results easily understandable, all predicted results in this
paper are reported and discussed in their primitive (dimensional)
forms.
2.4.5. Modeling of porous bottom plate of the combo bag
Fluid ow through a porous media is a complex phenomenon
due to the presence of solid matrix and the interactions of various
physical elds. To simplify the problem, the following assumptions
were also made with the assumptions listed in Section 2.3.
[1]. The lter was considered as non-deformable, homogeneous
and isotropic porous media, which was saturated with the
incoming melt.
[2]. The solid matrix of the porous media was considered to be in
local thermal equilibrium with the surrounding melt.
To model the DC casting process with internally placed porous
foam, two different sets of transport equations (continuity, momentum and energy) for clear uid and porous regions are needed.
For clear uid region the time-averaged turbulent form of the
governing conservation equations are already listed in Table 3. The
conservation equations for the porous region were based on
BrinkmaneForchheimer extended Darcy model. Beckermann et al.
[22,23] have discussed the importance of the two extensions of the
Darcy model. These authors have found that for modeling porous
media with a high permeability (i.e., a high Darcy number) the
aforementioned two extensions are necessary. It should be
mentioned here, the linear additions of the above two terms in the
Darcy model are based on semi-empirical models from the experimental ttings of the results. In this study, the ow of the melt
inside the porous lter was considered to be turbulent and
incompressible. The mathematical description of the turbulent ow
in the porous region is still a controversial issue. The present
mathematical model followed the approach where a volume-
74
Continuity equation :
1 vui
*
0
f
vxi
(26)
v rui uj
1
*
vxj
f2
"
!
#
CF r uj ui
vP
1
v
vui vuj
m
0
0
*
f*ml
rui uj ui p
vxi
K
vxj vxi
f2 vxj
K
Momentum equation :
rgx
(27)
Energy equation :
v
vxj
1 v ruj H
*
f
vxj
vT
f1=fg*ru0j H0
kEffective
vxj
!
(28)
"
!
# .
1 v ruj k
v
mt vk
vui vuj vui
2
*
ml
1 f *mt
f
vxj
sk vxj
vxj vxi vxj
vxj
rk uj
Ck p r Dk
K
(29)
Transport equation for rate of dissipation of kinetic energy:
"
#
v ruj
v
mt v
1=f *
ml
c1 f1 G c2 r
vxj
k
s vxj
vxj
u j
2
rE
Ck p f 2
k
K
(30)
the
eddy
viscosity
concept
and
was
given
by:
rui uj mt vui =vxj vuj =vxi 2=3dij k; and the turbulent viscosity
2
6
6
fu exp6
4
3
7
3:4
7
2 7
5
1 RT=50
f2 1 0:3 exp R2T
(31)
(32)
where RT k2/n turbulent Reynolds number; and n is the kinematic viscosity. Following Launder and Sharma (1974), the model
constants used here are given by: Cm 0.09, C1 1.44, C2 1.92,
sk 1.0, s 1.3, and the value of f1 is set to 1.
Similar to the transport equations for the melt, the porous media
transport equations (Eqs. (26)e(30)) were non-dimensionalized
with the same dimensional variables and were cast in the general
form as Eq. (13) with appropriate coefcients, and source and sink
terms. It is to be noted here that the non-dimensional transport
equations for the porous bottom plate of the combo-bag contained
two additional parameters, porosity(f) of the medium and Darcy
number (Da K/D2) both of which needed to be prescribed before
solving the equations.
The control volume based nite-difference scheme was used to
solve the modeled equations and boundary conditions which
offered a great advantage by the fact that the discretized governing
equations were derived by integrating the partial differential
equations over each control volume of the calculation domain
which automatically handled the strong momentum and heat exchanges at the interface of the clear uid and porous lter when the
relevant thermo-physical properties of the porous plate were
appropriately prescribed. Therefore, the boundary conditions at the
porous-uid interface were not necessary to consider.
2.4.6. Solution procedure
A control volume (CV) based nite difference approach was
adopted to discretize the equations. Staggered control volumes
were employed in the computational domain. A hybrid difference
scheme [33] was used to discretize the convectionediffusion terms.
In the present study there are seven variables, namely, U, V, W, P*,
k*, *, h* which were solved sequentially to obtain a converged
solution [33]. For each of the transport variable, the discretized
equation was cast in the following general form:
ap fp aE fE aW fW aN fN aS fS aT fT aB fB bf
(33)
where aE, aW, aN, aS, aT, and aB are the coefcients of the six
neighboring nodes of node P of a box-shaped rectangular control
volume. The term bF represents the linearized source term. The
convection and diffusion terms are embedded in the coefcients,
depending on the transport variable, the source term may contain
various terms such as pressure gradient, buoyancy force, and Darcy
terms. To resolve the velocityepressure coupling in the three momentum equations, the SIMPLE algorithm [34,35] was used. In
actual implementation of the SIMPLE algorithm, an iterative solution scheme was adopted using an implicit relaxation technique to
solve the discretized equations. The following combination of
under-relaxation factors provided converged solutions:
RF
X
X
anb Fnb b
ap Fp
all nodes
(34)
nb
The CPU time per iteration was about 1.0 min. The computations
were performed on a personal computer having a speed of
2.66 GHz and tted with an RAM of 4 Gigabytes. For a typical case,
to obtain a fully converged solution it took about 25 days and
required 36,000 iterations. For various parametric studies, in order
to decrease the computational time, the runs were restarted from
the converged solutions of the variables obtained for different
parametric cases.
3. Grid independency test
The grid independency tests were performed using three sets of
grid systems, namely, 60 42 24, 70 52 34, and 80 62 44
(x, y, z directions), which constituted 51,040, 108,800, and 196,560
control volumes, respectively. The local surface heat uxes at
y 430.5 mm, and z 322.5 mm for x varying from 135 to 400 mm
from the top free surface were compared for the above three grid
arrangements. It was found that the local surface heat ux varied by
less than 5% between the coarse and ne grid systems. The details
of the predicted results along with the comparisons are provided in
Begum [17]. In order to conserve space, those results are not reported here. To optimize CPU resources with an acceptable level of
accuracy and reasonable computational time, all production runs
reported in this paper were performed with a grid distribution of
60 42 24 grid points (or 51,040 control volumes).
4. Validation of the present developed code
Since a comprehensive 3-D DC casting work for a slab of
aluminum alloy AA-1050 similar to the one simulated in this study
is not available in the literature, hence the thermo-physical
properties of AA-3104 were considered and modeled the vertical
DC casting for the latter alloy in order to compare with the
experimental measurements of Jones et al. [36]. The latter authors
provided the measured temperatures at the ingot wide symmetry
plane for a rolling ingot of AA-3104 alloy. The simulated ingot had
a dimension of 2500 1320 660 mm in the x, y and z directions,
respectively and was identical to that used in the real casting
experiment by the above authors. The positions of the solidication front from the experimental measurements and predicted
from the present numerical model are compared in Table 4. It is
apparent from the above comparison of the depth of the solidication front that closer to the slab center, the predicted and
measured distances matched quite well, while further the thermocouples were away from the center and near the side wall the
match is not that satisfactory. This could be due to the fact that
Jones et al. did not measure the thermal boundary conditions
experimentally, and as a result, the employed thermal boundary
conditions used in the model might not have been appropriate.
Closer to the ingot center, a good agreement obtained with regard
to the vertical distances of the solidication front is an indication
that the heat transfer, namely, conductive, convective, and solidication which are at play in the caster as well as the turbulent
aspects of the process are well incorporated in the mathematical
model. Further details about the verications of the code are
available in Begum [17].
75
Table 4
Comparison of experimentally measured and numerically predicted vertical distance of the solidication front from the top of the mold at the wide symmetry plane
(z 0) for aluminum (AA-3104) alloy.
Jones et al.
Number
Horizontal distance
of sacricial
from the narrow face expts. [36]
(mm)
thermocouple of the mold wall
at the wide symmetry
plane (mm)
Relative
Present
numerical percent error
prediction
(mm)
1
2
3
4
140
281
382
510
45
147
249
351
63.8
191.3
418.2
568.6
119.4%
46.89%
8.66%
10.31%
76
Table 5
Summary of the various parametric conditions studied for the combo bag with a metallic foam bottom plate.
Case
Inlet melt
superheat, [ C]
Casting speed,
us, [mm/min]
Metal-mold contact,
HTC [W/m2-K]
1
2
3
4
5
6
7
8
32
32
32
32
32
32
32
32
40
60
80
100
40
40
80
80
1500
1500
1500
1500
750
3000
750
3000
42,367
63,550
84,734
105,918
42,367
42,367
84,734
84,734
267
401
535
669
267
267
535
535
Fig. 3. 3-D surface plots of the temperature contours and velocity elds with streamlines for complete solution domain for cases: (i) case-1 (a, b); (ii) case-2 (c, d); (iii) case-3 (e, f);
(iv) case-4 (g, h).
active just beneath the nozzle area where the jet momentum is
high. This is reected by the vertical arrows reaching a greater
distance in the axial (x) direction in the nozzle zone of the combo
bag. The horizontal jet of hot melt, after reaching the narrow side of
the slab, is seen to split into two streams. One stream is forming a
counter clockwise vortex at the wide symmetry plane, hence forth,
will be called the lower recirculation zone. Another part of the
steam moves upward and then it travels along the narrow and wide
faces and nally reaches at the mid-plane of the wide face where it
divides into two streams. One stream forms a clockwise recirculation zone at the top free surface, which is referred to as upper
recirculation zone, while the other stream moves in an inclined
manner, termed as angle ow, and meets with the ow of the lower
recirculation zone. In order to depict clearly the above description
of the ow eld, an enlarged view of the velocity vectors with
streamlines are portrayed in Fig. 4(a, b) for low and high casting
speeds of 40 and 100 mm/min, respectively for only the top part of
the caster. Since the velocity vectors inside the nozzle and combo
bag are very large compared to the velocity eld outside the above
two regions, hence while plotting the ow elds the velocity vectors inside the nozzle and combo bag were set to zero to overcome
the scaling problem.
One objective of the melt distributor in DCC is to minimize
turbulence at the top free surface so that the ow does not promote
the generation of the oxide inclusions there. Another objective of
having reduced ow there is to prevent the carryover of the inclusions into the cast. The proposed melt feeding system shows
that an upper recirculation zone forms at the free surface which
may not be desirable for a DCC process. A comparison of Fig. 3(b, d,
f, h), clearly shows that with the increase of the casting speed the
strength of the ow is increased. This can be observed by
comparing the indicated maximum resultant velocities reported in
the above four gures. This is mainly due to the additional
incoming melt for the higher casting speeds.
Fig. 3(a, c, e, g) shows the temperature contours along with the
solidication front (solidus) and the liquidus isotherm. The effect of
casting speed on temperature elds can be seen by investigating
77
Fig. 4. 3-D enlarged view of the velocity elds with superimposed streamlines for the top part of the calculation domain for cases: (a) case-1; (b) case-4. The maximum resultant
velocity developed in the ingot outside the combo bag and the nozzle has also been indicated by the red color arrow above each gure. (For interpretation of the references to colour
in this gure legend, the reader is referred to the web version of this article.)
78
Fig. 5. Enlarged 2-D view of temperature contours and velocity vectors of the top domain for case-1 at: (a) wide symmetry plane at z 0 mm, (b) vertical plane parallel to the wide
face at z 62.5 mm, (c) vertical plane parallel to the wide face at z 312.5 mm.
140, 152, and 165 mm, respectively, from the slab center, as illustrated in those gures. The increased uniformity of the solidication front with the increase in the casting speed is a very desirable
metallurgical characteristic in DC casting for the following reasons.
First, increasing the casting speed implies an increase in the
production rate of aluminum slabs. Second, an increase in the at
part of the solidication front implies that uniform heat transfer is
taking place over the increased surface area of the solidication
front. The latter would allow the slab to have better homogeneous
metallurgical characteristics. For a lower casting speed, the reason
for getting an almost-horizontally-attened solidication front is
attributed mainly due to the inability of the weak vertical jet
coming from the bottom porous plate to penetrate the stronger
angle ow near the central region, thereby the thermal convection
effect is reduced in that region. An increase in the casting speed
causes the speed of the vertical jet to increase. Consequently, a
79
Fig. 6. Enlarged 2-D view of temperature contours and velocity vectors of the top domain for case-2 at: (a) wide symmetry plane at z 0 mm, (b) vertical plane parallel to the wide
face at z 62.5 mm, (c) vertical plane parallel to the wide face at z 312.5 mm.
collision is taking place near the central region between the vertical
jet and the angle ow, thus reducing the strength of the resultant
ow and hence the thermal convective effect there. Moreover, for a
higher casting speed, the melt temperature is higher near the
bottom of the porous plate compared to the temperature of the
angle ow, as a result, the melt is becoming thermally stratied
there and causing less thermal buoyancy effect.
The left hand panels of Figs. 5(b)e8(b) show temperature
contours for a plane parallel to the wide symmetry plane at
z 62.5 mm. The temperature isotherms in each panel are seen
to have lifted upward in comparison to the isotherms presented
in the wide symmetry plane. The upward movements of these
isotherms, is mainly due to the weak convection within the
melt itself, and also due to the comparatively nearer location
with respect to the wide face where greater heat transfer takes
place.
Similar to the earlier gures, the left hand panels of Figs. 5(c)e
8(c) show temperature contours for a plane parallel to the wide
80
Fig. 7. Enlarged 2-D view of temperature contours and velocity vectors of the top domain for case-3 at: (a) wide symmetry plane at z 0 mm, (b) vertical plane parallel to the wide
face at z 62.5 mm, (c) vertical plane parallel to the wide face at z 312.5 mm.
taken, which are at an axial distance of 100, 200, 300, 400, 500, 600,
and 700 mm from the top ingot surface. In these gures, the rst
slice is taken near the end of the adiabatic section of the hot top at
an axial distance of x 100 mm. The second cross-section is taken
at the end of the metal-mold contact region within the mold, at
x 200 mm. The latter ve sections are located in the secondary
cooling region. Examinations of the gures show how the solidication proceeds along the wide and narrow faces of the ingot. At
x 100 mm, only liquid melt exist, as shown in these gures. Near
the end of the mold, a thicker solidied layer is formed along the
ingot rolling face compared to the ingot narrow face. The slow
moving melt along the rolling section allows more heat to be
extracted by the primary coolant of the mold. Whereas, the high
momentum melt stream delivered to the narrow side of the hot-top
mold through the side opening of the combo bag has delayed the
formation of the solidied layer in that part. The mushy layer is
separated between the center of the ingot and near the narrow slab
81
Fig. 8. Enlarged 2-D view of temperature contours and velocity vectors of the top domain for case-4 at: (a) wide symmetry plane at z 0 mm, (b) vertical plane parallel to the wide
face at z 62.5 mm, (c) vertical plane parallel to the wide face at z 312.5 mm.
disproportionately by about 65.19%, 83.02%, and 124.54%, respectively compared to the casting speed of 40 mm/min. Upon comparisons of sump depth and mushy layer thickness for various
casting speeds one nds that the increase of the sump depth is not
that signicant with the change in casting speed while there is a
signicant increase in the mushy layer thickness with the casting
speed. Because of the collision between the heated vertical jet and
the relatively colder angle ow in the central area, the thermal
convection effect is markedly reduced. As a result, the effect of the
casting speed is not greatly manifested on the sump depth.
5.2.2. Predicted solid-shell thickness at two locations
In Fig. 10, the predicted shell thickness from the center of the
narrow face at the wide symmetry plane and from the center of the
wide face at the narrow symmetry plane at an axial distance of
210 mm from the top free surface is given in the form of a bar chart
for four cases (1e4) of Table 5. The predicted values of the shell
thickness in mm are also provided at the top inside each bar of the
82
Fig. 9. Contours of solidus and liquidus temperatures at various transverse cross-sectional planes (yez planes) of the top part of the ingot for four cases: (a) case-1; (b) case-2; (c)
case-3; (d) case-4.
chart for easy comprehension. For both locations, with the increase
of the casting speed the shell thickness decreases, as expected. For a
change of casting speed from 40 to 60 mm/min, the shell thickness
is seen to decrease rapidly whereas for the casting speeds above
60 mm/min the trend of the retardation of the solid-shell is
comparatively slow and gradual. This disproportionate change of
the solid-shell formation with the change in casting speed may be
due to the complex ow patterns produced by this melt feeding
scheme.
A signicant variation of the thickness of the solid-shell along
the narrow and wide faces at the exit of the mold is found. At the
mid-point of the narrow face, a thinner shell is seen to have
developed compared to the mid-point of the rolling face, and this is
true for all four simulated casting speeds. The reason behind this
phenomenon is the fact that the part of the melt coming out from
60
80
100
466.715
7.01
488.897
11.58
494.413
12.83
501.728
15.74
Fig. 10. Solid-shell thickness from narrow and wide slab faces at the wide and narrow
symmetric planes respectively at an axial distance of x 210 mm from the top free
surface versus casting speed for cases (1e4).
83
melt coming out through the side window of the bag is directed
towards the narrow face and as a result the surface temperature is
higher at that location. On the contrary, near the mid-point of the
rolling face, a relatively colder and slow moving melt generates a
lower thermal convective ow and has resulted in the reduction of
the surface temperature there. As a consequence of the two
different thermal convective effects in the said two regions there is
approximately 1.21e1.53 times thicker solid-shell is seen to
develop at the mid-point of the wide face compared to the midpoint of the narrow face (refer to Fig. 10) at the mold exit. This
trend is observed up to an axial length of 335 mm for the low
casting speed of 40 mm/min while for the higher casting speed of
100 mm/min this trend prevails up to a length of approximately
495 mm. Beyond the above axial distances, the trend in the temperature distributions is seen to have been reversed. In other
words, a comparatively higher value of the surface temperature is
predicted along the cast at the wide face compared to the value
predicted for the same axial location but at the narrow face. The
reasons behind this phenomenon are due to the higher thermal
resistance from the thicker solid-shell around the wide face and
also for the proximity of the mid-point of the narrow face to the
corner region where higher rate of heat extraction is taking place
from the two faces of the cast surfaces.
Fig. 11. Variations of surface temperature along the axial direction of the strand at four
locations of the caster for: (a) us 40 mm/min (case-1) (b) us 100 mm/min (case-4).
84
Fig. 12. 3-D surface contours plots of non-dimensional turbulent viscosity of the top part of the ingot for four cases: (a) case-1; (b) case-2; (c) case-3; (d) case-4.
takes place from the slab surface to the cooling water under this
condition.
5.2.4. Non-dimensional eddy viscosity
The non-dimensional eddy/turbulent viscosity contours on the
top free surface and two vertically symmetric planes for four cases
(1e4) are shown in Fig. 12(a)e(d). Turbulent viscosity is a ow
property which represents the enhancement of molecular viscosity
of the melt due to the turbulent motion of melt there. In the present
study, the eddy viscosity model was used to calculate quantitatively
the level of turbulence. For this delivery system the turbulent viscosity varies from zero to approximately 1000 times the molecular
viscosity. A zero value of the turbulent viscosity appears in the solid
regions, and it increases with the increase of convective intensity
into the bulk liquid phase. The highest value of the turbulent viscosity is obtained at the exit of the side window of the combo bag,
as expected. From these gures, the higher value of turbulent viscosity is observed at the ingot top surface which is caused by the
upper recirculation vortex produced due to the studied melt
feeding scheme. This may contaminate the melt by carrying the
solid oxide scales and inclusions generated at the top free surface
into the cast. The above situation can seriously deteriorate the
quality of the as-cast products. A comparison among the four gures indicates that the level of turbulence increases with the
increasing casting speed which is consistent with the progressively
increased Reynolds number of the ow presented in Table 5 for
cases (1e4).
5.3. Inuence of effective heat transfer coefcient (HTC) on
solidication process
For vertical DCC processes in the mold region a higher value of
HTC ranging from 1000 to 5000 W/(m2-K) and a lower value
ranging from 100 to 500 W/(m2-K) have been reported in the
literature [3,41]. It is to be noted that not many experimental
studies concerning HTCs in the mold region for industrial-sized
slab casters of different alloys are available in the literature. In
this section, a sensitivity analysis has been carried out by changing
the value of HTC in the mold-metal contact region only. For this
purpose, three different arbitrary values of constant HTC for the
mold-metal contact region, ranging from 750 to 3000 W/(m2-K),
have been considered and the shell-thickness as a function of HTC
is plotted for two casting speeds, 40 and 80 mm/min as shown in
Fig. 13. Solid-shell thickness from the narrow slab wall at the wide symmetry plane at
an axial distance of x 210 mm from the top free surface versus effective heat transfer
coefcient [W/(m2-K)] for casting speeds of 40 and 80 mm/min for six cases (1, 3, and
5e8).
85
6. Conclusions
The steady state phase of DCC process for an industrial-sized
rolling ingot with a combo-bag having a bottom porous foam
plate is simulated for the rst time. The following observations are
found:
[1] For an increase in casting speed a deeper sump, a thinner
shell, a higher surface temperature, and a greater level of
turbulence are attained due to the increased intensity of
thermal convection from the incoming melt.
[2] A thicker mushy region is observed at the ingot center with
the increase of the casting speed.
[3] For the studied melt feeding system two vortices are seen to
develop. One vortex forms in the bottom region and right
after the exit of the side window of the combo bag. Another
vortex forms at the ingot top free surface. With the increase
of the casting speed, the strength of these two vortices
increases.
[4] It is observed for the rst time that at the narrow symmetry
plane the melt is owing from the rolling face towards the
ingot center in an inclined manner, termed here as an angle
ow. The strength of the angle ow is increasing with the
increase of the casting speed.
[5] The vertical jet emerging from the bottom of the porous plate
of the combo bag near the central region is colliding with the
angle ow and the strength of the downward jet is reduced.
This has resulted in the attening of the solidication front in
the central area. With the increase of the casting speed the
extent of the atness of the solidication front in the central
region has progressively increased. Because of the above
colliding effects of the jet and angle ow there is only a
minor effect on the enhancement of the sump depth with
casting speed.
[6] For the rst time for a slab caster a cliff-shaped mushy region
is predicted right after the exit of the side window of the bag.
This is due to the enhanced thermal convection currents
resulting from the merging of the angle ow and the vortex
that forms right after the side window of the distribution
bag.
[7] With the increase of the casting speed the cliff-shaped
mushy region is moving downward and is taking different
shapes due to the stronger thermal convection.
[8] For the studied casting speeds, a thicker solid shell is predicted at the mid-point of the wide face compared to the
mid-point of the narrow face at the exit of the mold.
86
Darcy coefcient
coefcients in the discretized governing equations
empirical constants for low Reynolds number model
specic heat
morphology constant
nozzle hydraulic diameter
extra dissipation term in k-equation
Darcy number
extra generation term in -equation
empirical constants used in low-Re version of ke models
liquid fraction
solid fraction
production term in turbulent kinetic energy equation
Grashof number
modied Grashof number Gr/Ste
gravitational acceleration in the x-direction
sensible heat
total heat (sensible and latent)
permeability of porous media
turbulent kinetic energy
thermal conductivity of the liquid
thermal conductivity of the solid
effective thermal conductivity of the porous media
hydrodynamic pressure
Peclet number
Reynolds number
pore Reynolds number Re/f
turbulent Reynolds number based on the turbulent
quantities
source term
source term associated with F
non-dimensional source term associated with buoyancy
term in k-equation
non-dimensional source term associated with buoyancy
term in -equation
Stefan number
temperature
uctuation of temperature
inlet temperature
liquidus temperature
solidus temperature
surface temperature of the caster
velocity component in the i-th direction; corresponding
to u
time-average velocity component in the i-th direction
inlet velocity
casting speed
non-dimensional form of the u, v and w velocities
non-dimensional form of us
x
y
z
X, Y, Z
axial direction
horizontal direction parallel to the wide face
horizontal direction parallel to the narrow face
non-dimensional form of x, y, z
Greek symbols
nodal latent heat
latent heat of solidication
diffusion coefcient associated with F
alloy density
effective viscosity equal to mt ml
laminar viscosity
turbulent viscosity or eddy diffusivity
generalized dependent variable
effective diffusivity
DH
DHf
GF
r
me
ml
mt
F
Geff
Superscripts
*
non-dimensional variables
e
time-averaged variables
0
uctuation of variables.
References
[1] A. Prasad, I.F. Bainbridge, Experimental determination of heat transfer across
the metal/mold gap in a direct-chill (DC) casting mold-part-I: effect of gap size
and mold gas type, Metall. Mater. Trans. A 44A (2013) 456e468.
[2] J. Sengupta, B.G. Thomas, M.A. Wells, The use of water cooling during the
continuous casting of steel and aluminum alloys, Metall. Mater. Trans. A 36A
(2005) 187e204.
[3] J.F.R.C. Etienne, A.R. Baserinia, H. Ng, M.A. Wells, D.C. Weckman, Heat-transfer
measurements in the primary cooling phase of the direct-chill casting process,
Metall. Mater. Trans. B 43B (2012) 1202e1213.
[4] E.K. Jensen, W. Schneider, Investigations About Starting Cracks in DC-Casting
of 6063-Type Billets, II. Modeling Results, Light Metals, The Minerals, Metals
and Materials Society, Warrendale, PA, 1990, pp. 937e943.
[5] J. Jacoby, Direct chill casting defects, aluminum cast house technology, in: 5th
Australian Asian Pacic Conference, 1995.
[6] J.F. Grandeld, V. Nguyen, et al., Ingot caster productivity improvement
through examination of mould heat ow and deformation, in: Proceedings of
the Australasian Conference and Exhibition, Aluminum Cast House Technology, Sydney, 2007, pp. 147e154.
[7] D.G. Eskin, Physical metallurgy of direct chill casting of aluminum alloys, in:
J.N. Fridlyander, D.G. Eskin (Eds.), Advances in Metallic Alloys, CRC Press, 2008.
[8] S.P. Tremblay, M. Vincent, Development and Use of a New Composite Material
for Aluminium Contact Applications, Light Metals, The Minerals, Metals and
Materials Society, Warrendale, PA, 2005, pp. 1007e1012.
[9] S.P. Tremblay, M. Lapointe, The Manufacturing, Design and Use of a New
Reusable Molten Metal Distributor for Sheet Ingot Casting, Light Metals, The
Minerals, Metals and Materials Society, Warrendale, PA, 2002, pp. 961e965.
[10] R. Kieft, J. van Oord, F. Frinking, D. Bal, H. van Schoonevelt, Detailed Modeling
of a Metal Distributor by Means of a Combined Numerical and Physical
Approach, Light Metals, The Minerals, Metals and Materials Society, Warrendale, PA, 2003, pp. 821e827.
[11] M. Fortier, A. Larouche, X.G. Chen, Y. Caron, The Effect of Process Parameters
on the Metal Distribution for DC Sheet Ingot Casting, Light Metals, The Minerals, Metals and Materials Society, Warrendale, PA, 2005, pp. 1019e1024.
[12] A. Arsenault, D. Larouche, S.P. Tremblay, J.P. Dube, DC Cast Thermal and Fluid
Flow Simulation Using a Semi-permeable Model of TF Combo Bag, Light
Metals, The Minerals, Metals and Materials Society, Warrendale, PA, 2008, pp.
781e785.
[13] F. Ilinca, J.F. Hetu, A. Arsenault, D. Larouche, S.P. Tremblay, 3-D Modeling of
the Flow and Heat Transfer During DC Casting with a Combo Bag, Light Metals,
The Minerals, Metals and Materials Society, Warrendale, PA, 2009, pp.
799e804.
[14] D.R. Poirier, G.H. Geiger, Transport Phenomena in Materials Processing, The
Minerals, Metals and Materials Society, Warrendale, PA, 1994.
[15] J. Zuidema Jr., L. Katgerman, I.J. Opstelten, J.M. Rabenberg, Secondary Cooling
in DC Casting: Modeling and Experimental Results, Light Metals, The Minerals,
Metals and Materials Society, Warrendale, PA, 2001, pp. 873e878.
87
[29] J.T. Hong, C.L. Tien, M. Kaviany, Non-Darcian effects on vertical plate natural
convection in porous media with high porosities, Int. J. Heat Mass Transfer 28
(11) (1985) 2140e2157.
[30] J.L. Large, The fundamental theory of ow through permeable media from
Darcy to turbulence, in: D.B. Ingham, I. Pop (Eds.), Transport Phenomena in
Porous Media, Elsevier Science, 1998, ISBN 0-08-042843-6, p. 446.
[31] Y.T. Yang, M.L. Hwang, Numerical simulation of turbulent uid ow and heat
transfer characteristics in a rectangular porous channel with periodically
spaced heated blocks, Numer. Heat Transfer Part A 54 (2008) 819e836.
[32] K. Vafai, R.L. Alkire, C.L. Tien, An experimental investigation of heat transfer in
variable porosity media, J. Heat Transfer Trans. ASME 107 (3) (1985) 642e647.
[33] S.V. Patankar, Numerical Heat Transfer and Fluid Flow, rst ed., Hemisphere
Publ. Corp., Taylor and Francis Group, New York, 1980, pp. 88e90, pp.
126e131.
[34] S.V. Patankar, Numerical prediction of three-dimensional ows, in:
B.E. Launder (Ed.), Studies in Convection: Theory, Measurement and Applications, Vol. 1, Academic Press, New York, 1975, pp. 125e127.
[35] H.K. Versteeg, W. Malalasekera, An Introduction to Computational Fluid Dynamics, the Finite Volume Method, rst ed., Longman Scientic Technical,
Longman Group Ltd., Essex, England, 1995, pp. 142e146.
[36] W.K. Jones Jr., D. Xu, J. Evans, W.E. Williams, D.P. Cook, Effects of Combo-bag
Geometry on the Thermal History and Sump Prole of a 3104 DC Cast Ingot,
Light Metals, The Minerals, Metals and Materials Society, Warrendale, PA,
1999, pp. 841e845.
[37] M. Can, A.B. Etemoglu, Porosity measurement of stainless steel lters produced by electrical discharge technique, Filtr. Sep. 41 (2004) 37e40.
[38] T.C. Dickenson, Filters and Filtration Handbook, Elsevier Advanced Technology, Oxford, U.K., 1992, pp. 123e126.
[39] L. Zhang, D.G. Eskin, A. Miroux, T. Subroto, L. Katgeran, Inuence of melt feeding
scheme and casting parameters during direct-chill casting on microstructure of
an AA7050 billets, Metall. Mater. Trans. B 43B (2012) 1565e1573.
[40] J.F. Grandeld, D.G. Eskin, I. Bainbridge, Direct-chill Casting of Light Alloys:
Science and Technology, Wiley, Hoboken, NJ, 2013, p. 244.
[41] A. Prasad, I.F. Bainbridge, Experimental determination of heat transfer with
the metal/mold gap in a direct-chill (DC) casting mold: part-II. Effect of casting
metal, mold material, and other casting parameters, Metall. Mater. Trans. A
44A (2013) 3099e3113.