8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
1/12
TF Combo bag - Mathematical simulation results and up-date
Presented at the 3rd International Melt Quality Workshop,
Dubai, UAE, 14-16th November 2005
Daniel Larouche
Laval University
Dept. of Mining, Metallurgy and Materials Engineering
Adrien-Pouliot Bldg
Québec, (Québec)Canada, G1K 7P4
And
Sylvain P. Tremblay
Pyrotek High-Temperature Industrial Products Inc.Chicoutimi (Québec), Canada
Abstract
TF combo bags were introduced in 2002. Since then, many plants have been converted and betterresults have been reported than with regular sewn combo bags. The first results were presented at
TMS 2002.
This paper will deal with a direct comparison between a regular sewn combo bag and a TF bag
using mathematical simulations. Using ProCast simulation program package, the heat and flowdistributions of molten aluminum exiting combo bags will be presented. The mesh technique used
as well as boundary conditions will be described. Parameters such as bag deformation, fabric
opening and bag size have been simulated and will be presented.
Finally, an up-date on TF combo bag implementation around the world will be given.
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
2/12
1.0 Introduction
Since their introduction at TMS 2002 (1), TF combo bags have been a technical and commercial
success. The TF combo is a rigid bag featuring very accurate dimensions bag after bag. Figure 1
shows a TF combo bag (brown) besides a regular sewn combo bag (white). There are now more
than 25 casthouses using TF combo bags and each of these customers has reported better results
in terms of bottom block filling, turbulence, metal flow, butt curl, scalping, etc., compared to aregular sewn combo bag.
The aim of this project is to try to understand why the TF combo bag gives better production
results. Pyrotek’s water model results have given some direction in term of flow pattern
differences that might explain the best TF combo bag performance. The development of a
mathematical model should reinforce the previous water model flow results and bring a new
dimension regarding thermal fluid flow patterns.
Figure 1. Regular sewn combo bag and TF combo bag
In this communication, thermal-fluid flow mathematical simulations of DC casting including a
thermally formed or a regular sewn combo bag will be described and preliminary results will be
presented. A detailed description of the methodology used to estimate the pressure drop occurring
when the metal flows through fabric cloth will be given. The basic idea leading to this work was
that the permeability of fabric cloth is a key parameter in the distribution pattern of metal.
2.0 The models
2.1 Geometry and volume properties
Since the main goal of this project is to give more accurate visualisations of the flow and thermal
patterns obtained with different combo bags, it was decided to include the entire distribution
system in the model and to add filter type materials simulating the influence of fabric cloth. Bag
dimensions are 125 x 100 x 300 mm and the ingot shape is 596 x 1660 mm. Figure 2 shows the
geometrical models used in the mathematical simulations.
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
3/12
34L cloth
40F cloth
Model 1
Model 2
Figure 2: Geometrical models. The meshed volumes are thin permeable materials.
The models differ by their combo bag geometry, which are modelled according to a thermally
formed combo bag (model 1) and a regular sewn combo bag (model 2), the latter being shaped
according to the deformed geometry obtained after a casting. Both models show a dip tube,
control pin and a skim dam surrounding the combo. A spout sock is also part of each combo bag.
In the TF combo bag, the spout sock is a rigid square inner TF bag while in the regular combo, it
is a soft sewn sock. The novelty of the approach is to take into account the permeability of the
fabric cloth in the calculation of the flow pattern. Previous mathematical modelling published in
literature simply adjusted the velocity magnitude at permeable windows to satisfy continuity (2) or
added momentum source to describe the resistance of the fabric cloth (3). In this work, the
permeability of fabric cloth was first calculated by simulating the flow of metal through grids
made of rectangular holes, similar in size to the openings found in combo bags. Figure 3 presents
the model used to calculate the pressure drop in the fabric style 34L of Pyrotek.
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
4/12
Sz
wz
Sx
wx
Figure 3: Geometrical model of the fabric type 34L
Pressure drop were calculated with different nominal velocities for fabrics 34L and 40F. The size
specifications of these two grid types are given in Table I.
Grid TypeSx
(mm)
wx
(mm)
Sz
(mm)
wz
(mm)
t
(mm)34L 3.4 0.94 3.0 0.86 0.5
40F 2.7 1.1 2.4 0.69 0.5
Table I: Grid specifications.
Notice that the grid shown in Figure 3 was not explicitly modelled with such details as in the DC
casting models. In the latter, filter materials were modelled as continuous sheet having
permeability determined according to the pressure drops calculated with the grid models.
The number of tetrahedral elements in models 1 and 2 was respectively 410119 and 415815. The
software package ProCAST v. 2004.1 was used to solve the conservation equations in the laminar
regime.
2.2 Boundary and initial conditions
Simulations of DC casting in models 1 and 2 were conducted with identical boundary and initial
conditions.
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
5/12
2.2.1 Fluid flow
The velocity of metal at the bottom of the model was set to 5 cm/min and a pressure inlet
condition was prescribed at the top entry of the metal. Wall type conditions were given to the side
of casting and a zero vertical velocity was prescribed at the top surface of the liquid pool. These
conditions insured that the volumetric flow of metal at the entry was equal to the casting speed
multiplied by the cross section of the ingot.
2.2.2 Heat transfer
Primary and secondary cooling conditions were modelled as in reference (4). Table II presents the
heat transfer used at different positions over the external surface of the ingot. A natural air
cooling condition was given to all other surfaces. Contact heat transfer coefficients between
different volumes were constant and equal to 1000 W/m2/K.
Heat flux = h·(Tsurface – T ref )
Distance from the
surface of the pool
(mm)
Tsurface
(K)
Heat transfer
coefficient h
(W/m2/K)
T ref (K)
Primary cooling (h is constant) 0 – 87 * 500 373Air gap (h is constant) 87-111 * 10 293
273 2000 283
373 5000 283
403 27000 283
573 8500 283
Water cooling
(h is function of Tsurface)111 and below
873 0 283
Table II: Heat transfer conditions applied at the external surface of the ingot.
2.2.3 Initial conditions
The pool of metal was artificially separated in such a way that one of the volumes had the
geometry of a bottom block. This precaution was taken to give the possibility to eventually model
the filling stage of the start-up. In this work, the pool of metal was considered as one volume with
no interface. However, the virtual separation allowed prescribing different initial temperatures to
the different regions so delimited. The initial temperatures were set as follow:
Metal in the dip tube and in the combo bag: 690°C
Metal between the virtual bottom block and the combo bag: 650°C
Metal in the virtual bottom block: 600°C.
Combo bag materials: 690°C
All other materials: 200°C
2.3 Material properties
The aluminium alloy considered in the simulations was the 5182. An enthalpy formulation was
used to avoid taking into account the rate of solid phase apparition to compute the release of
latent heat, since the metal was defined “liquid” at all temperatures for the purpose of solvingfluid flow even in regions where the temperature was below solidus. A penalty method was
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
6/12
therefore used, giving a high arbitrary viscosity to the metal below its coherency temperature. The
thermo-physical properties of the 5182 alloy were calculated using the rule of mixture. The solid
and liquid phases were considered as dilute solutions. Thermophysical properties of fiberglass
fabric cloth, dip tube and control pin were assumed to be those of fused silica. For the skim dam,
thermo-physical properties were taken from Pyrotek’s N14 product data sheet, except for the
specific heat, that was assumed to be equal to the specific heat of CaO-SiO2.
3.0 Results of mathematical simulations
3.1 Flow simulation through grids
Figure 4 presents the pressure drop obtained with a flow having a nominal velocity of 0.2 m/s, perpendicular to the grid 34L. Average pressure drops were calculated from the simulation results
for nominal velocities of 0.03, 0.2, 1.0 and 5.0 m/s. Figure 5 presents plots of pressure drop
versus nominal velocities obtained on grids 34L and 40F.
Figure 4: Pressure calculated in a through holes section of the 34L grid. Liquid phase is 5182
aluminium alloy at 700ºC. Boundary conditions for the fluid flow simulation are also
indicated.
Pressure = 0
Vy= 0.2 m/s
Direction of flow
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
7/12
y = 2666.5x1.9165
R 2 = 0.9998
1.0E-02
1.0E-01
1.0E+00
1.0E+01
1.0E+02
1.0E+03
1.0E+04
1.0E+05
0.01 0.1 1 10
Nominal velocity (m/s)
A v e r a g e P r e s s u r e D r o p ( P a )
Grid 34L
y = 3033.1x1.9
R 2 = 0.9993
1.0E-02
1.0E-01
1.0E+00
1.0E+01
1.0E+02
1.0E+03
1.0E+04
1.0E+05
0.01 0.1 1 10
Nominal velocity (m/s)
A v e r a g e P r e
s s u r e D r o p ( P a )
Grid 40F
Figure 5: Calculated pressure drop versus the velocity of A5182 at 700ºC for grids 34L and 40F.
The pressure drop is highly non-linear with velocity, which indicates that pressure drop at high
velocities is more a consequence of the impacting effect on the screen than a consequence of
friction effects. It is interesting to see that a power law fits very well the pressure drop associatedwith the velocity of metal. Unfortunately, the filter model in ProCAST assumes that pressure drop
has a linear dependency with velocity. Such a model is suitable for thick porous media wherePoiseuille type of flow mainly occurs. This linear dependency can be expressed as follow:
K
V μ
Δ x
Δ P ⋅= (1)
where Δ P/ Δ x is the pressure gradient, µ is the viscosity of the fluid, V is the nominal velocity and
K is the permeability of the porous media. Since the metal velocity obtained through the fabric
cloth was in general less than 0.2 m/s, it was decided to use equation (1) as a linear
approximation. For cloths 34L and 40F, the estimated values for the permeability ( K ) were
respectively 8.2x10-6
cm2
and 7.0x10-6
cm2
. Figure 6 shows the difference obtained at lowvelocities with grid 34L, assuming a linear approximation.
y = 2666.5x1.9165
0
500
1000
1500
2000
2500
3000
0 0.2 0.4 0.6 0.8
Nominal ve locity (m/s)
P r e s s u r e d r o p
( P a )
1
Linear approximation
Grid 34L
Figure 6: Calculated pressure drop with A5182 at 700ºC for grid 34L, assuming a power law or a
linear fit at low velocities.
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
8/12
As one can see, the approximation is quite acceptable for velocity below 0.2 m/s.
3.2 Simulations of DC casting
Because of the very large number of elements in the models, only 22 seconds of simulations were
done at the time of writing the paper. The results are therefore preliminary and indicate only some
trends that will necessitate more investigations. The temperature and velocity profiles calculated
in the longitudinal symmetry plane (parallel to the rolling faces) are presented in Figure 7 and 8.
The velocity pattern is clearly different at the exit of the two combo bags. For the thermally
formed combo bag, the arrows indicate that the flow is directed upward next to the side exit
window and is confined in a narrow channel following the walls of the combo bag. In the regular
sewn combo bag, the metal is pushed farther from the side exit window and also deeper from the
bottom window. There is a net higher flow in the regular sewn combo bag along the longitudinal
symmetry plane, which explains why the high temperature isotherms extend larger in this plane.
The two models have identical volume flow rate, so if a higher flow occurs along the longitudinal
symmetry plane for the regular sewn combo bag, it is because the flow in the thermally formedcombo bag is off-centred. Figure 9 shows what happens in the thermally formed combo bag in the
transverse symmetry plane. A non negligible portion of the metal exiting the dip tube flows
directly through the permeable fabric cloths (spout sock and inner shell) and gets confined
between them and the impermeable outer shell surrounding the combo bag. Once confined, the
metal flows in the longitudinal direction (z axis) and creates a relatively strong stream at the exit
of the combo bag (Figure 10).
Figure 7: Calculated velocity magnitude at t = 22s in the longitudinal symmetry plane. Arrows
indicate direction of flow only.
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
9/12
Figure 8: Calculated temperature profile t = 22s in the longitudinal symmetry plane. Arrows
indicate direction of flow only.
Figure 9: Calculated velocity magnitude at t = 22s in the transverse symmetry plane of model 1.
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
10/12
Figure 10: Calculated z-component velocity at t = 22s in a transverse section located just after
the extremity of the combo bag in model 1.
This off-centred stream is in accordance with the flow observed on a water model experiment
conducted with a TF combo bag. Figure 11 presents a snapshot of the dye dispersion obtained
after injection of the colorant in the water circulating in the dip tube. On the same figure, a z-
component velocity mapping is presented for comparison.
Figure 11: Comparison between the water model visualisation and the calculated z-component
velocity in a section crossing the stream formed in the outer shell of TF combo bag .
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
11/12
The flow rates in models 1 and 2 are clearly balanced in different ways, model 1 having extra off-
centred outputs while model 2 has only centred outputs. The hot zones are consequently located
at different positions as one can see in top views of temperature profiles (Figure 12).
Top view
19 mm
below the
surface
Model 1 : Thermally formed Model 2 : Regular
Off-centred hot zone Centred hot zone
Figure 12: Calculated temperature profiles at t = 22s at two different depths.
4.0 Discussion
The metal distribution in a DC caster is governed essentially by how the momentum gained by
the metal going down the dip tube is balanced by the different output resistances offered by the
system. Filters materials resist to flow according to the velocity of metal and as such, are passive
devices that don’t impose any velocity distribution. Momentum sources however are active
devices working like pumps and promote a velocity distribution. These sources may acceleratethe metal in certain locations where the velocity can be stagnant in reality. Imposing a pre-defined
velocity distribution at combo bag exit windows is also like imposing a solution that can be far
from reality. Considering the large area of exit windows in combo bags (where a uniform
distribution of velocity is unlikely) and the fact that one can have multiple outputs oriented
differently in space, it is clear that the flow rate in each of these windows will be decided by the
relative resistance encountered by the metal in the path to reach that window. With these
considerations in mind, it is clear that the balance of metal flow can only be estimated by a 3D
calculation involving filter materials having the right geometry, thickness and flow resistances. In
this work, filter materials 0.5mm thick were modelled, combined, meshed and included in the DCmodels. Their resistance to the flow of metal was estimated by numerical simulations, giving an
idea of the pressure drops associated with metal velocity below 0.2 m/s. This approach was
undertaken at the cost of computation time since the models must have a very large number ofelements to represent adequately thin filter cloths. It is why turbulence models were not activated,
computation times having been multiply by approximately a factor 10 doing so. It is expected thatturbulence will affect mainly the velocity pattern inside the combo bags, but much less outside.
These modelling efforts have enlightened how important is the geometry of the outer shell. In this
regard, the results shown for the regular sewn combo bag must be taken with care since the outer
and inner shells were supposed to form one shell with permeable windows. In reality, the outer
shell is very likely separated from the inner shell at some places when the metal flows in the
8/18/2019 11 TF combo bag_Tremblay_Pyrotek .pdf
12/12
distributor. The exact configuration of the outer shell in model 2 has not been taken into account,
having no indications how this shell deforms in a real casting. To get more accurate visualisations
of the flow with regular sewn combo bag, it will be important to better describe the geometry of
the system in the future.
5.0 Conclusions
The first results of the mathematical simulation program on TF combo bag are in agreement withthe previous water model results. The project is at the beginning but it will go on until the final
solution in laminar regime is completed. Then, turbulent regime will be added to see its influence
on the solution since it is well known that the flow is turbulent at the dip spout discharge.
The next step will be a validation step at one of our customer plant using these combo bag andDC mould dimensions.
6.0 References
(1) S. Tremblay, M. Ruel
"The Manufacturing, Use and Plant Test Results of TF Combo Bags for DC Sheet Ingot Casting”,Light Metals 2002.
(2) D. Xu, W.K. Jones Jr., J.W. Evans, D.P. Cook
" Mathematical and physical modeling of systems for metal delivery in the continuous casting of steel
and DC casting of aluminum", Appl. Mathematical Modelling, 22, 1998, pp. 883-893.
(3) 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
2005, pp.1019-1024.
(4) G.U. Gruen, A. Buchholz, D. Mortensen"3-D Modeling of Fluid Flow and Heat Transfer During the DC Casting Process – Influence of Flow
Modeling Approach–", Light Metals 2000, pp.573-578.