by
A THESIS
BIRMINGHAM, ALABAMA
2008
BRAKE DRUM PRODUCTION BY CENTRIFUGAL CASTING
MATERIALS ENGINEERING
ABSTRACT
The process of centrifugally casting a gray iron liner onto a steel shell for brake
drum applications was examined through experimental trials and computer simulations.
Casting trials were performed, and pour temperature, iron chemistry, caster spin speed,
and pour time were varied. The drums were machined and rated for the defects present.
Castings were examined microstructurally and defects that originated at the steel-
shell/gray iron interface were found. These defects were formed from an oxide on the
steel shell that reacted with the molten iron to form carbon monoxide. The oxide was
formed on the steel shell surface when it was preheated for one hr in an oxygen-rich
atmosphere. The carbon monoxide gas caused bubbles to form in the solidifying iron and
caused a failure in the steel shell-gray iron bonding layer. Calculations of the gas
evolution rate revealed that the carbon monoxide gas was converted from the iron oxide
for 64 seconds.
the iron, but the long run-times of the model on a desktop computer made it impractical
for studying multiple variables. A static solidification model was developed to examine
the solidification of the casting after the iron was poured. This model showed that, 64
seconds after pouring, some of the iron at the mold-metal interface solidified, which
The amount of oxide present on the steel shell could be reduced by induction
ii
reduce the time to convert this oxide to approximately 4 seconds. The solidification
model showed that the amount of iron solidified 4 seconds after pouring was small and
iii
TABLE OF CONTENTS
Page
ABSTRACT ........................................................................................................................ ii
CHAPTER
1 INTRODUCTION ...........................................................................................................1
2 BACKGROUND .............................................................................................................4
3 PROCEDURE ................................................................................................................10
iv
TABLE OF CONTENTS (CONTINUED)
v
LIST OF TABLES
Table Page
vi
LIST OF FIGURES
Figure Page
8 The relationship between carbon equivalent and defect rating in the gray
iron lining casting trial ...................................................................................................22
9 The relationship between caster spin speed and defect rating in the gray
iron lining casting trial ...................................................................................................22
13 Microstructure of a good bonding layer between the iron and steel shell ....................26
16 Cooling of the shell as measured from the backside of the shell ..................................30
vii
LIST OF FIGURES (CONTINUED)
23 2-dimensional slice of the casting showing the fraction of solid after 64 seconds .......41
24 Percentage of fully solid shell liner interface length vs. shell temperature
after 64 seconds.............................................................................................................42
25 2-dimensional slice of the casting showing the solidification after 4 seconds .............43
viii
CHAPTER 1
INTRODUCTION
Automotive brake drums are typically manufactured from gray iron using the
sand casting process. One of the advantages of gray iron for this application is the high
thermal conductivity produced by the graphite flake network in the metal. Unfortunately,
this network also produces low toughness, which requires the drum to have thick wall
sections to reduce the stresses created during braking. Any reduction in brake drum
weight will reduce unsprung and rotational mass of the vehicle, which in turn will
A brake drum that combines the high toughness of steel with the thermal
characteristics of gray iron would allow lower weight designs while maintaining part
performance. A method of making such a drum is to use a formed steel shell with a gray
iron wear surface. The challenge in creating this part is developing a method to make a
uniform iron lining surface that will weld with the steel shell. To achieve this, a
centrifugal casting process was investigated. Centrifugal casting provides a fast cycle
time and a uniform gray iron lining thickness, and adds centripetal force to help with the
joining process. With the centrifugal casting process, increased cooling rates and
maximize properties.
Another advantage to this process is the flotation of any lower density oxides that
form during pouring to the inner diameter due to the centrifugal force (Janco, 1988).
1
This force pushes the liquid iron outwards onto the shell wall and displaces the lighter
oxide material toward the center of the drum, where it can be machined away after the
casting is complete.
This process has several challenges, however. The steel shells are preheated to
promote the joining process, which is problematic, as the heating of a steel shell in an
atmosphere containing oxygen can cause the formation of iron oxide on the shell surface
(Iordanova, Surtchev, Forcey, & Krastev, 2000). This iron oxide can react with the
The formation of carbon monoxide, CO, gas may result in gas defects at the steel/iron
interface in the resulting casting (Sun & Easman, 2007). In the case of iron-lined steel
brake drums, the bonding surface can be interrupted and the mechanical properties of the
drum adversely affected. Another possible defect is pinholes, which can be caused by
both mold and metal variables, such as variations in carbon content and high hydrogen
feasibility of producing an iron /steel brake drum. The quality of the drums was
inconsistent, and the primary causes of rejection were gas holes and delaminations at the
steel/iron interface. This thesis theorizes that an iron oxide layer at the interface is the
primary source of the gas and delamination. This thesis will combine a numerical
2
Computer simulations can be an effective tool in investigating defect formation in
the centrifugal casting. Dynamic simulations can examine the flow and fill portion of the
casting process, while a static model can be used to examine the solidification process of
the liquid iron in the steel shell and how it is affected by the pouring temperature and the
shell preheat temperature. These factors determine the time that the liquid iron is in
The method used to preheat the shell is also important. Time at elevated
temperature and the heating atmosphere affect the thickness of the iron oxide scale on the
shell. The amount of iron oxide present determines the time of gas production and
volume of gas produced. The combinations of CO gas generation and iron solidification
characteristics predict if the gas will be entrapped and cause a defect or escape through
Defect analysis of selected drums was used to determine the possible cause of the
defects and where the defects originate. Experimental data was used to verify the
numerical simulation. Computer simulations were then used to determine how the iron
cools and solidifies and how long it takes for a layer of solidified iron to form. The
pouring temperature of the liquid iron, the shell preheat temperature, and the amount of
oxide on the preheated shell, which is dependent upon the time it takes to preheat the
3
CHAPTER 2
BACKGROUND
Centrifugal Casting
applications. Centrifugal casting is a process where liquid metal is poured into a mold
that is spinning, either at the time of casting or immediately following the completion of
the pour. The centrifugal force from the spinning mold forces the metal to solidify with a
and the air (Janco, 1998). Typically, the products produced are cylindrical, such as
bearings, ductile iron pipes, and rolls used in the paper and textile industries. These
products are usually made using the horizontal or true centrifugal casting method, where
the rotation occurs about a horizontal axis. The vertical casting method can be used to
make asymmetrical parts, such as valve bodies, jewelry, and dental prosthetics.
related to turbulent liquid flows due to the high revolutions per minute (RPM) used and
centrifugal casting will always encounter difficulties because of the physics involved
forces are not true forces, but this name was adopted and is used to describe the process.
4
Centripetal force, or force acting around an axis toward the center of the axis, is given by
where FC is the centripetal force holding the liquid metal to the mold wall, m is the mass
of the liquid metal, v is the velocity, r is the radius, and is the rotational speed in
radians per second. In order for a droplet of liquid metal to be held to the mold surface by
spinning force alone, the centripetal force must be greater than the force of gravity acting
Fc > Fg
This gives a minimal rotational speed, which prevents raining of liquid iron during
casting liquid droplets falling from the mold wall. When converted to linear velocity on
the mold wall for a 16.5-inch diameter brake drum, the liquid velocity is much greater
than the 0.5-m/s threshold velocity for free surface envelopment, as derived by John
Campbell, and is thus a very turbulent flow (Campbell, 1991). The turbulent flow of the
liquid metal can introduce new defects due to the entrainment of air and unwanted
Large temperature gradients can form in the molten metal during the casting
process. These gradients can be affected by many different variables, such as mold
irregularities, irregular mold coatings, improper mold preheating, and turbulent flow. A
quick one-dimensional thermal analysis (Incropera & DeWitt, 1990) of a droplet of metal
5
on the mold surface shows that liquid on the mold surface freezes instantaneously, but
liquid just a few microns into the metal does not freeze for more than one second. The
very thin layer of metal that solidifies can trap gas produced by mold-metal reactions and
accumulate until a bubble forms and breaks through this layer and causes defects to form
in the still mushy metal. Defects can also be amplified due to the turbulent fluid flow and
irregular mold coating, which cause severe thermal gradients on the scale of a few
Computer Simulations
Computer simulations of casting processes have been successfully used for years
to predict, reduce, and/or eliminate defects in static castings. Simulating the centrifugal
casting process requires additional computational power to deal with the high metal
program that uses computational fluid dynamics to simulate metal flow in a mold. This
method uses standard flow equations that are solved for each individual computational
cell. This program can solve physical models to predict solidification rate, turbulence, air
The governing equations for dynamic free-surface fluid flow have been available
for nearly a century, but computer simulation of dynamic free-surface fluid flow has been
technically feasible only for the past decade. Figure 1 shows an infinitesimal volume cell
6
y
y y + dy
y
yx
yx + dy
y
yz
yz + dy
y
z
xy
xy + dx
x
xz zx x
x + dx
xz x
x zy xz + dx
x
zy
xy zy + dz
z
zx
zx + dz
z
x
z yz
z + dz
z
yx
z y
Figure 1. Fluid unit cell showing all accelerating forces. (FlowScience, Inc., 2007).
The mass of liquid that moves into and out of the control volume shown in Figure
VF p uA x vA y wA z RSOR
+ + + =
c 2 t x y z
(4)
where VF is the fractional volume open to flow; is the fluid density; c is the speed of
sound in the liquid; p is the pressure; u, v, and w are the component velocities in each of
the cardinal directions; Ax, Ay, and Az are the areas open to flow in each of the cardinal
directions; and RSOR is a density source with no momentum (FlowScience, Inc., 2007).
7
This equation, in simple terms, is equivalent to counting the population in one region
during one moment in time it is a function of the current regional population (the first
term in Equation 4), the rate of immigration and emigration (terms 2-4 in Equation 4),
and the number of people that are born and the number that die (the right side of Equation
4). In terms of the centrifugal casting process, the first term of the equation is the
compressibility of the fluid, which does not affect this process since the liquid iron is
incompressible. The second through fourth terms describe the volume flow of the liquid
metal in terms of the three directions of the unit cell. The right side of the equation is any
mass created within the cell that does not occur during the basic model but comes into
play when the iron oxide on the steel shell surface forms carbon monoxide.
The mass balance equation must be solved simultaneously with the Navier-Stokes
equations of momentum in order to fully describe the liquid being simulated. This is
given in equation 5.
u 1 u u u 1 p RSOR
+ uA x + vA y + wA z = G x + f x bx u
t VF x y z x VF
(5)
porous media (FlowScience, Inc., 2007). There is a similar equation for momentum in
the y- and z-directions, as well. In simpler form, and to further extend the previous
metaphor, the Navier-Stokes relationship describes the forces that make people move in a
particular region at a particular moment in time. The geography (the first term in
Equation 5), transportation speeds (the second term in Equation 5), and environmental
factors (the right side of Equation 5) all cause the masses to move in a particular
direction. In terms of fluid flow, the right side of the equation describes the unexpected
or uncontrollable events such as gravity and viscous acceleration of the fluid. Gx is the
8
gravity and the centrifugal force of the shell spinning. Viscosity forces that can cause
shear forces as the viscosity increases in the fluid are represented by the fx term. The third
term on the right side is the compressibility term, and bx is flow through a porous media,
which does not affect this particular model. The last term on the right side is the source
term that describes any mass created or destroyed within the unit cell. The first term on
the left side describes a geometry that causes high acceleration, such as a nozzle or a
weir. The rest of the left side describes the movement of the liquid due to the spinning of
the shell.
mass conservation and accelerations occur on a wall that has non-zero momentum. The
spinning mold adds an additional traction to the right side of Equation 5 (GX) and makes
tracking fluid momentum more difficult because of the relatively high mold speeds.
In simulating castings that use heated permanent molds, there may be a need to
account for mold surface reagents that can produce gas upon contact with the liquid metal
stream. This phenomenon can be simulated by the proper manipulation of the basic flow
approximation equations. Mold temperatures are governed by the general heat flow, as in
Equation 6.
T
(1 VF ) w C w Tw
t
x k w (1 A x )
T
x
y
( )
k w 1 A y
y
z k w (1 A z )
T
z
= TSOR
(6)
where w, Cw, kw, and Tw are solid material values for density, specific heat capacity,
thermal conductivity, and temperature, respectively, and TSOR is an energy source term
due to liquid-solid heat transfer and external heat sources (FlowScience, Inc., 2007)
TSOR would also include all heat energy input from phase transformations and radiative
9
CHAPTER 3
PROCEDURE
Casting Trials
and a gray iron liner that was centrifugally cast onto the shell to obtain a bond by
diffusion at the steel/iron interface. The thickness of the steel shell is approximately 5
mm, while the gray iron liner is approximately 19 mm. Figure 2 shows the drawings of
the steel shell before and after the gray iron liner was cast.
Figure 1. Shell before and after casting. Drawings of the steel shell before (on the left) and after the gray
iron liner is cast (on the right).
pouring speed (tilting speed of the ladle), pouring temperature (temperature was
measured in the transfer ladle before pouring the iron into the pouring ladle), chemistry
A hot alkaline wash was applied to the shells in order to remove any residual oil
from machining. Then the shells were cold rinsed and dipped in a heated solution of a
10
flux, borax pentahydrate at 82 C to help with the bonding of the iron to the steel shell
and to minimize the oxidation that occurs during the preheating process. The shells were
then placed in an oven at 677 C for an hour to preheat the shell before casting, as seen in
Figure 3.
A heat of iron was produced by charging an induction melting furnace with scrap
steel and alloys, then heating until melted. When the heat was brought to tapping
temperature, a thermal arrest cup was poured to measure the carbon equivalent, and the
chemistry was adjusted as needed. The iron was brought back up to tapping temperature
and poured into a pocket ladle in order to inoculate the gray iron. The shell was then
taken out of the oven and placed in the casting spinner while the iron was transferred to
the pouring ladle. Once the shell was secure and brought to the desired rotational speed,
the pouring ladle was rotated into the shell and the pour began. The ladle was rotated
manually for 410 seconds, according to the predetermined matrix. The ladle was then
rotated out of the shell, and the shell continued to spin until the metal was fully solidified
11
and cooled. The shell was then removed for examination. Figure 4 shows the pouring of
the iron, and Figure 5 shows the spinning after the ladle is rotated away.
Figure 5. Shell spinning after pouring until the liner is solidified and cooled.
The cast brake liners were machined, and the surface was examined for defects
according to an established standard. This was a visible inspection, and the defects were
12
classified as light or heavy, depending on the depth of the defect. The rating system was
a qualitative system that rated each machined casting from 1.0 to 4.0.
1.1 to 1.9 - light defects seen (less than 3 defects to be considered a sound
casting).
Any casting rated 2.0 or above was considered not acceptable. When defects
were present, the drums were sectioned through the defect to see the depth of the defect,
but no ratings were made on these sections. Conditions of the casting were recorded for
each trial. Temperatures were taken using immersion thermocouples, pour times were
analyzer, while the silicon concentration was measured using a spectrometer. Carbon
equivalent (carbon + 1/3 silicon), liquidus [1669-(124 x % carbon)-(% silicon x 31)] x 1.8
Analysis of Defects
defect causes. These sections were first examined macroscopically and photographed
before being further sectioned. These sections were mounted, ground, and polished and
examined under a light microscope. The samples were etched with Nital (4% nitric acid
in methanol) etchant, and the gas defects were examined to determine whether the defects
were caused by carbon monoxide gas or hydrogen gas. A graphite lining would be
13
present if the defects were caused by hydrogen, while a defect caused by the carbon
monoxide gas would have no graphite lining present (Carter et al., 1989).
shell after the pouring event. Measurements were taken using a fiber optic non-contact
temperature system (Williamson model number PRO 91-65) aimed at the backside of the
shell from the time of the pour and continued for 5 min of cooling time after the pour was
energy optically and uses this energy to calculate the target surface temperature. The
instrument used was rated for work in the 871C 1760C range. Temperatures were
recorded every 10 seconds manually and every second using a DaqBook data acquisition
thermocouple.
From experimental observation, the time from removal of the preheated shell
from the oven to pouring into the shell was 80 seconds. The shell was preheated for 1 hr
in a furnace at 677C, but the shell cooled in the 80 seconds between removal of the shell
from the furnace and the pour. To estimate shell temperature at the time of pouring, the
cooling rate of a 5-mm steel plate (the thickness of the steel shell) was measured. First, a
furnace was preheated to 677C. While the furnace was heating, the steel plates were
14
prepared by coating two thermocouples with boron nitride (Lubricoat-ZV) and spot
welding the thermocouples to the steel plate. The coating was used to protect the
insulation on the thermocouples from the high temperature of the furnace. The plates
were then placed in the oven and allowed to reach temperature for 30 min, then taken out
of the oven and allowed to cool in air at room temperature. A DaqView data acquisition
Computer Simulations
Centrifugal Model
A model of the centrifugal casting of a gray iron drum was developed using Flow-
3D by FlowScience Inc. The physical models that were implemented within Flow-3D
were heat transfer, solidification, viscosity, turbulence, surface tension, and such scalars
as air entrainment, velocity of the liquid metal, and surface defect tracking. Heat transfer
was determined with the inclusion of heat transfer between the fluid and shell, and the
conduction of the shell and fluid to the surrounding atmosphere. Solidification was
modeled in this program as resistance to flow. For low solid fractions, the viscosity was
a function of the solid fraction, but at solid fractions larger than the coherent solid
fraction, a DArcy type drag force was implemented with a drag coefficient proportional
to the solid fraction. The coherent solid fraction is where the amount of liquid solidified
is large enough to form a coherent structure. The DArcy-type drag force was calculated
by adding a term to the right side of the Navier-Stokes equation (equation 5). When the
solid fraction reached the point of rigidity, the drag became infinite, and there could be
no fluid flow.
15
Viscosity and turbulence were evaluated as a sum of assigned constant molecular
values for both the viscosity and turbulence for the fluid as well as a function of solid
fraction. Surface tension was applied in Flow 3D by the amount of free surface area in
the unit cell. If the volume fraction of the unit cell was one or zero, no surface tension
was applied. If the volume of the cell was between one and zero, the pressure of the
surface tension was applied toward the direction of the unit cell where there was more
liquid present and away from the direction of the unit cell where there was no liquid
present.
calculated within the code of the program when the turbulence was sufficient to disturb
the surface enough to allow air into the fluid. Entraining air is undesirable and may lead
The velocity magnitude of the liquid metal was tracked as it moved through each
and mass balance equations and also relates to turbulence in the casting process.
The surface defect tracking model records the time that a volume of liquid metal
is exposed to the air. This gives an oxide formation level that can be used to predict the
Solidification Model
In order to examine the thermal properties of the iron and steel shell and the effect
environment was important to examine due to its impact on the entrapment of gas. This
16
model was built axisymmetrically to examine a 30 slice of the casting in order to make
run-times shorter. This model was static, since the force from the spinning of the casting
was greater than the force due to gravity and should force the iron to stay against the
mold wall as soon as the mold was filled. The centrifugal model confirmed that the
centrifugal force was greater than the force due to gravity, since once the pouring was
complete and the shell was filled, the motion of the fluid within the shell was minimal.
The goal of the solidification model was to determine how long the fluid was in the liquid
state and was verified by matching the cooling curve of the model to the experimental
cooling curve.
One of the problems encountered in building this model was the variation in the
data for thermal properties of gray iron. Thermal properties from a review by Pehlke
(1982) yielded the closest match to the experimental curve. Using the Pehlke thermal
properties data, models were run to examine the solidification time as a function of two
variables: temperature of the liquid and temperature of the shell. This data was then
analyzed against the rate of gas formation data to determine the probability of defects
To study the range of possible conditions for the casting, two different pouring
temperatures (1316 C and 1482 C) and two different shell preheat temperatures (510 C
and 677 C) were used. The temperatures were selected as best case (1482 C and 677
C) and typical conditions (1316 C and 510 C) that may be seen for the two variable
factors. In Flow3D, history points are points selected by the user to collect data, such as
temperature, pressure, and fluid fraction, in a specific location. Data collection points
were included in the model at the outer diameter of the shell and also at three equally
17
spaced points through the thickness of the liquid. The point on the outer diameter of the
shell was used in order to compare against the experimental while the points spaced
through the thickness of the liquid were used to determine the temperature of the iron.
18
CHAPTER 4
Casting Trials
After each drum lining was cast, the inside diameter of the casting was machined
and examined for defects. Each casting was then rated to determine if the casting was
acceptable. The rating system was from 1.0 to 4.0. If the casting was determined to be
unacceptable (2.04.0), the casting was sectioned through the defects and further
examined. The results of the gray iron lining casting trial can be seen in Table 1.
Almost all of the defects seen were identified as gas defects. Although not all
conditions were recorded for each drum produced, there was a single significant
statistical correlation found within the results. The biggest factor in casting an acceptable
part was casting pouring speed. The shorter pour time of 4 seconds resulted in more
acceptable castings than the longer pour time of 10 seconds. It is possible that the shorter
pour time resulted in more turbulent pouring and thus more entrainment and reoxidation
defects that can break down to form more gaseous defects. This correlation can be seen
in Figure 6.
All other factors examined, including tap temperature, turning speed, temperature
in the transfer ladle before pouring, carbon equivalent (carbon content plus one third the
silicon content), carbon concentration, and silicon concentration, did not show a
statistically significant correlation with the defect rating. Typical results for transfer ladle
temperature, carbon equivalent, and caster spin speed are shown in Figures 79.
19
Table 1
Equivalent (wt %)
Tap Temperature
Temperature in
Turning Speed
Defect Rating
Pour Time (s)
Liquidus (C)
Solidus (C)
Si (wt %)
C (wt %)
Material
Carbon
(RPM)
(C)
(C)
4 GI 2 3.790 1.870 1583 1419 4.413 1141 1116
4 GI 2 3.790 1.870 1583 1419 4.413 1141 1116
4 GI 4 3.790 1.870 1583 1419 4.413 1141 1116
4 GI 3 3.790 1.870 1583 1419 4.413 1141 1116
4 GI 3 3.697 1.890 1581 1444 4.327 1152 1115
4 GI 2 3.697 1.890 1581 1444 4.327 1152 1115
4 GI 2 3.697 1.890 1583 1444 4.327 1152 1115
4 GI 2 3.697 1.890 1583 1444 4.327 1152 1115
350 GI 4 3.500 2.060 1599 1412 4.187 1171 1113
420 GI 3 3.500 2.060 1599 1412 4.187 1171 1113
350 GI 3 3.500 2.060 1611 1459 4.187 1171 1113
420 GI 3 3.500 2.060 1611 1459 4.187 1171 1113
350 GI 1.5 3.470 2.300 1598 1442 4.237 1167 1110
420 GI 4 3.470 2.300 1598 1442 4.237 1167 1110
350 GI 3 3.470 2.300 1602 1416 4.237 1167 1110
420 GI 3 3.470 2.300 1602 1416 4.237 1167 1110
550 GI 1.5 3.412 2.230 1599 1461 4.155 1177 1111
420 GI 3 3.412 2.230 1599 1461 4.155 1177 1111
550 GI 3 3.412 2.230 1614 1441 4.155 1177 1111
420 GI 3 3.412 2.230 1614 1441 4.155 1177 1111
550 GI 1.5 3.480 2.230 1603 1441 4.223 1168 1111
420 GI 3 3.480 2.230 1603 1441 4.223 1168 1111
550 GI 3 3.480 2.230 1610 1442 4.223 1168 1111
420 GI 4 3.480 2.230 1610 1442 4.223 1168 1111
10 GI 3 3.509 2.250 1598 1448 4.259 1164 1111
10 GI 3 3.509 2.250 1598 1448 4.259 1164 1111
10 GI 1.5 3.509 2.250 1594 1494 4.259 1164 1111
10 GI 1 3.515 2.210 1602 1451 4.252 1164 1111
10 GI 1 3.515 2.210 1602 1451 4.252 1164 1111
10 GI 1 3.515 2.210 1599 1488 4.252 1164 1111
10 GI 1.5 3.515 2.210 1599 1488 4.252 1164 1111
10 GI 1 3.550 2.280 1598 1349 4.310 1158 1110
10 GI 1.5 3.550 2.280 1598 1349 4.310 1158 1110
10 GI 1.2 3.550 2.280 1597 1491 4.310 1158 1110
10 GI 1 3.550 2.280 1597 1491 4.310 1158 1110
10 GI 1.2 3.535 2.210 1597 1373 4.272 1162 1111
10 GI 1.2 3.535 2.210 1597 1373 4.272 1162 1111
10 GI 1.5 3.535 2.210 1597 1373 4.272 1162 1111
10 GI 1.5 3.535 2.210 1597 1373 4.272 1162 1111
20
4.0
3.5
3.0
Quality Rating
2.5
2.0
1.5
1.0
0.5
0.0
0 2 4 6 8 10 12
Pour Time (seconds)
Figure 6. Correlation between pouring time and defect rating. H pouring time was ten seconds to
complete the pour while L was four seconds. A defect rating over two is deemed as unacceptable.
4.0
3.5
3.0
Defect Rating
2.5
2.0
1.5
1.0
0.5
0.0
2400 2450 2500 2550 2600 2650 2700 2750
Temperature in the Transfer Ladle
Figure 7. The relationship between pouring temperature and defect rating in the gray iron lining casting
trial. This shows no direct relationship between pouring temperature and the likelihood of defect
occurrence.
21
4.0
3.5
3.0
Defect Rating
2.5
2.0
1.5
1.0
0.5
0.0
4.10 4.15 4.20 4.25 4.30 4.35 4.40 4.45
Carbon Equivalent
Figure 8. The relationship between carbon equivalent and defect rating in the gray iron lining casting trial.
This shows no direct relationship between carbon equivalent and the likelihood of defect occurrence.
4.0
3.5
3.0
2.5
Defect Rating
2.0
1.5
1.0
0.5
0.0
300 350 400 450 500 550 600
Caster Spin Speed
Figure 9. The relationship between caster spin speed and defect rating in the gray iron lining casting trial.
This shows no direct relationship between carbon equivalent and the likelihood of defect occurrence.
The transfer ladle temperature and spinning speed were expected to have an effect on the
occurrence of defects in the casting, but the data did not reveal any correlation. However,
22
the transfer ladle temperature was probably not a good indication of the actual pouring
temperature. There were two pouring stations, and the pouring ladle from each station
was filled from a single transfer ladle, but the temperature was only taken before the first
pouring ladle was filled. The metal in the second pouring ladle would be at a lower
temperature, as the transfer ladle sat unheated during the first pour. In addition, as the
pouring ladles were not preheated well, the actual metal pouring temperature could be
Analysis of Defects
Sections of drums that had been rejected due to the presence of surface defects
were analyzed to investigate the cause of the defects. There were two areas where the
defects occurred: on the inside diameter of the drum and on the interface between the
steel shell and the iron lining. Examples of these defects can be seen in Figure 10.
(a) (b)
Figure 10. Two types of defects on the casting. Photographs showing defects on the inside diameter of the
casting (a) (2X) and on the steel-iron interface (b) (1.5X)
The defects present on the inside diameter of the casting could come from several
possible sources. This defect could be the result of oxidation of the iron during pouring
or from the inclusion of slag material from the melt. The material could be trapped in the
23
fluid and cause oxidation as the fluid solidifies. The centrifugal force should displace
this material to the inside diameter of the casting where it can be machined away, unless
the metal solidifies too quickly or the turbulence causes the defect to be entrapped.
Another possible cause of the defect present at the inside diameter of the casting
is the creation of gas from the fluid-mold interaction. The conditions in which the shells
are preheated cause iron oxide scale to form on the surface. When the liquid iron comes
in contact with this oxide, the carbon reacts with the oxide and is reduced to iron and
carbon monoxide gas. If the iron freezes before the gas can travel through the liquid by
virtue of the centrifugal force, defects are seen at the inside diameter of the casting, either
The gas created at the mold-metal interface can result in a poor bond between the
iron and steel, and gas defects can form as the gas travels through the fluid and is trapped
if the liquid freezes too quickly. The preheated shell and the liquid iron were at elevated
temperatures for a period of time to allow diffusion to take place and the bond to form.
Figure 11 shows a bond that was disrupted due to the presence of gas. Figure 11a shows
the surfaces of the iron (on the left) and the steel shell (on the right) and the gap in the
interface from the presence of the gas. Figure 11b shows the two materials as they would
be in the as cast condition with a gas bubble extending into the iron. Figure 11c is a
polished section of the defect. When there was no gas present or the gas was allowed to
escape through the molten iron, a good, continuous bond was created. A good section of
24
Iron Steel
Iron Steel
(a) (b)
Iron
Steel
(c)
Figure 11. Views of the defect seen in 10b. (a)(2X) is the two pieces separated, showing the defected
bonding layer. (b)(4X) are the two pieces put back together showing a gas defect that is seen polished and
magnified in (c)(20X).
Figure 12. Section of casting showing a good continuous bond between the iron and steel shell. (1X)
25
The microstructure of the bond between the iron and the steel shell showed a
sharp layer separating the two materials, with a diffusion layer into the steel shell from
the molten iron being poured onto the shell. This microstructure can be seen in Figure
13.
Figure 13. Microstructure of a good bonding layer between the iron and steel shell. (40X)
atmosphere containing oxygen, the dominant oxide layer, Fe3O4, grows to a thickness of
about 15 microns (Ioranova et al., 2000). The growth of the oxide on low carbon steel at
600 C according to Ioranova et al., can be seen in Figure 14. Figure 14 shows that, at
600 C (the shells are preheated at 677 C), the oxide layer will grow to about 15 m
after 60 min in an atmosphere containing oxygen. This layer could be reduced by heating
in an inert atmosphere or, as the figure shows, reducing the time that the steel shell is
preheated. If the shells were preheated using an induction furnace, the time it takes the
shell to reach the desired temperature would be reduced, which would reduce the amount
26
16
14
y = 4.52505Ln(x) - 3.48248
R2 = 0.99807
12
Oxide Thickness (m)
10
0
0 10 20 30 40 50 60 70
Time (min)
Figure 14. Growth of oxide layer at 600 C. Results of a study of high temperature oxidation of steel by
Ioranova et al. with curve fitted to data points.
Using the known oxide layer thickness of 15 m, the density of the oxide (5.4
g/cm3), and molecular weight of the oxide, the total amount of Fe3O4 present was
calculated. The Fe3O4 is reduced by the carbon in solution to form iron plus carbon
Using the calculated amount of Fe3O4, the amount of CO that is formed can be
calculated. The rate of gas formation was calculated from the Arrhenius dependence as
seen in Equation 7:
Q
Rate = k exp (7)
RT
where k is the rate constant (approximately 1x10-7 from reference 3), Q is the total energy
for the reaction in equation 1 (Perry, Green, & Maloney, 1997), R is the universal gas
27
constant, and T is the temperature in Kelvin. The rate of gas formation was calculated to
be 0.0136 cm3/cm2/s from previous calculations. This rate was then used to calculate the
total time of gas evolution, depending upon the thickness of the oxide on the shell.
The amount of gas that evolved was dependent upon the total amount of oxide
present on the surface. The amount of oxide present on the shell was dependent upon the
time that the steel is heated in an oxygen-rich atmosphere, which is displayed in Figure
14. Since the amount of oxide present determines the total amount of gas evolved and
the time that it takes to evolve all of the gas, a relationship can be made between the time
it will take to preheat the shell and the time it will take to convert all of the oxide present
70
65
Time for Conversion Completion (seconds)
60
55
50
45
40
35
30
25
20
15
10
0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70
Figure 15. Time of conversion completion versus pre-heat time. The relationship between the time the
shell is preheated in an oxygen rich atmosphere and the time it will take to convert all of the oxide present.
This relationship shows that when the shell was preheated in an oxygen rich
atmosphere for 1 hr, it took more than 60 seconds to convert all of the oxide present on
28
the surface. With both the rate and total CO gas formed calculated, the time for
completion was calculated to be 64 seconds, meaning that the gas can continuously form
for 64 seconds or until the metal is solidified. The formation of the CO gas can be a
major source of defects for an iron pour into a steel mold preheated at a high temperature.
Measurements were taken showing the heating and cooling of the shell during the
pouring process. The shell was at approximately 510 C when the metal was poured.
The metal was poured between 1316 C and 1482 C, so there was an initial heating of
the shell while the metal was being poured. Once the pouring event was over, the cooling
began and the spinning continued for at least 3 min, until the temperature was below the
liquidus of the iron and the metal was solidified. The cooling of the shell was measured
from the backside of the shell with an infrared thermometer and can be seen in Figure 16.
On this pour, the temperature was 1533 C in the transfer ladle and was taken
using an immersion thermocouple. There was a heat loss as the metal was transferred to
the pouring ladle, since the ladle could not be sufficiently preheated in this setup. The
bearings used in the rotating arm would seize up when the ladle was preheated
thoroughly, and the ladle could not be rotated to pour the iron into the shell. This graph
shows the temperature on the backside of the shell, so it is lower than the temperature of
the liquid iron. The infrared thermometer used had a range of 871 C 1760 C, so
temperatures could not be measured below 871 C, which is not important, since the
casting is removed once it reaches that temperature. Figure 16 shows the initial heating
of the shell due to the presence of the molten iron and the subsequent cooling. The initial
29
slope of the curve was steep and then leveled off as the iron reached the liquidus
temperature and began to solidify. Then the temperature began to drop more quickly
again until the solidus was reached and the slope leveled out again. Finally, the casting
1200
1150
1100
Temperature (C)
1050
1000
950
900
850
800
0 25 50 75 100 125 150 175 200 225 250
Time (s)
Figure 16. Cooling of the shell as measured from the backside of the shell.
To determine the temperature of the shell when the iron is poured, the cooling of a
5-mm steel plate (approximate thickness of the steel shell used in the process) was
observed. It took approximately 80 seconds to remove the shell from the furnace, place it
in the spinner, and begin to pour the iron into the shell. Temperatures were recorded
using DaqView data acquisition software and the results can be seen in Figure 17.
30
700
Plates taken out of furnace
675
650
625
Temperature (F)
600
575
550
525
When iron would be poured
500
0 20 40 60 80 100 120
Time (s)
Figure 17. Estimating the temperature of the shell at the time of pouring. The cooling of two 5mm plates
(approximate thickness of the steel shell used in the process) from the preheat temperature of 677 C.
From these results, it was predicted that the shell temperature would drop about 300 C in
80 seconds, from 677 C to 510 C. This temperature was used as an input in the
Computer Simulations
Centrifugal Model
Preliminary simulations for the casting problem used the General Moving Object
model in Flow3D. The General Moving Object model rotates the geometry used (a .STL
drawing file) through the computational mesh and applies the physical models to the
geometry. The complex geometry of the mold and the high velocity used to spin the
16.5-inch mold at 400 to 600 RPM resulted in run times of over a month on the desktop
31
computers. The general moving object model setup did not fully resolve the part as
shown in Figure 18, even with a very fine mesh size, causing inaccuracies in the
simulation.
Figure 18. Original .STL geometry. .STL geometry used for the general moving object model showing
poor resolution even with a very fine mesh. The white triangles are parts of the geometry that are not
resolved.
A different method was used in order to decrease run times on the desktop computers a
method known as the Phantom Obstacle model. The Phantom Obstacle model defines a
region to apply an axial velocity but does not give the region a volume or any real
obstacle blocking effects (FlowScience, Inc., 2007), allowing for less computational load
and a geometry that is more easily resolved. The Phantom Obstacle model can still have
all of the previous physical models applied to it and run on a desktop computer.
Unfortunately, imported .STL geometry cannot be used for this model, so the geometry
had to be created within Flow-3D. The first model tested was a simple cylinder that can
32
Figure 19. Simplified geometry used to test the phantom obstacle method.
With this method, the part was fully resolved and run times were reduced to 3 to 4 weeks
depending upon the input variables used e.g., pouring temperature, pouring speed, and
caster spin speed. Figure 20 is a series of two-dimensional slices through the center of
the casting and pouring ladle with the fluid colored by temperature in Kelvin. The initial
temperature of metal in the pouring ladle was 1760 K (1487 C), the shell was 783 K
(510 C), and the RPM was 550. After one second of pouring, the liquid being poured on
the shell was disorganized, with droplets raining from the shell surface and no significant
amount of cooling of the liquid, even though the cooling rate should be maximized at this
time. The liquid was still at the pouring temperature of 1482 C, and there was no
immediate chill effect from the contact with the shell. The raining phenomenon can also
lead to oxidation, since the liquid iron will be exposed to oxygen as the small droplets
solidify quickly.
33
At 2 seconds, the liquid started to form a layer inside the shell, although there was
still a large unfilled section. There was some raining seen, although it was drastically
reduced with the establishment of the liquid layer. After 3 seconds, there was a
continuous layer of liquid on all parts of the shell, but it was also clear that the layer
thickness was not uniform, which might make the thickness of the solidified iron uneven
around the shell. Also of note is how fast cooling began once a liquid layer was
established. This is evident by the liquid on the shell for a full rotation being at around
1427 C while the oncoming liquid was at 1482 C. Also, the liquid being poured onto
the shell was hot enough to reheat preexisting liquid, which further delayed solidification.
At the end of the pouring cycle, none of the liquid was near the liquidus of 1410 K
(1138 C), and the thickness was still very uneven around the shell. Experimental
observations show that, with these pouring conditions, the casting was not fully solidified
until at least 3 min after pouring. The simulation could not be run for this length of time
The Surface Defect Tracking model records the relative amount of time a volume
of liquid metal is exposed to the air (FlowScience, Inc., 2007). Figure 21 shows 2-D
plots of the predicted surface defect locations during the pour. As shown in Figure 21a,
liquid in the ladle had already formed a large amount of oxidation products on the
surface, and these oxides flowed into the shell mold and were incorporated into the
liquid. This is even more evident after 4 seconds, as shown in Figure 21b, where the
liquid oxides were significantly thicker, which means that the last liquid poured was the
most contaminated.
34
(a) (b)
(c) (d)
Figure 20. Temperature contrast of the phantom obstacle model. 2-D slices through the center of the mold
showing temperature (K) contours at (a) 1 second, (b) 2 seconds, (c) 3 seconds, and (d) 5 seconds from the
beginning of the pour.
(a) (b)
Figure 21. Surface defect concentration contrast of the phantom obstacle model. 2-D surface defect
concentration contours of the casting at (a) 1 second and (b) 4 seconds.
35
Figure 22 shows an image of liquid strain rate at 4 seconds from the beginning of
the pour. This plot is an indication of how well liquid is accelerated by the rotating shell
and, for ideal conditions, should show a very high shear at the shell surface and a gradual
decrease across the thickness of the fluid to low shear at the inside diameter of the liquid.
However, as shown in the image, the shearing layer is only at the shell surface, and there
is no gradual gradient in the liquid, which means that the liquid may not be efficiently
accelerated to the required velocity to prevent raining. Moreover, the impinging pouring
stream slows liquid down that is already on the shell, increasing the chance of liquid
pulling away from the shell. This may indicate that the shell may need some type of
Figure 22. 2-D plot of liquid strain rate after 4 seconds of pouring.
Solidification Model
Although the centrifugal model was run successfully, the run times required,
about 4 weeks in most cases, made it impractical to use this model to run a matrix in
order to identify the problems causing the defects. This model did show that the
36
centrifugal forces due to the spinning of the shell overcame the gravitational forces and
forced the liquid onto the shell once the pouring was complete. The analysis of the
defects showed that one source of the gas defects was the iron oxide decomposition on
the steel shell. This defect is also important, as it could be present without an indication
on the inner diameter of the drum. If the defect was severe enough, the liner and the shell
was developed to use in conjunction with the gas rate calculations performed previously
to determine the conditions that would lessen or eliminate the gas defects from the oxide
The previous calculations of the amount of oxide present from the pre-heating
methods used and the resulting gas converted from this oxide layer showed that the gas
would evolve for 64 seconds. The simulation results for the fraction solidified for a slice
in the drum after 64 seconds are shown in Figures 23a through 23d. Although the
chemistry of the iron used in the Pehlke data is not listed, the carbon equivalent for this
iron was calculated from the liquidus and solidus temperatures to be 3.39, with a carbon
concentration of 3.02 and a silicon concentration of 1.12. The solidus was 1127 C and
the liquidus was 1245 C. The experimental measurement had a carbon concentration of
liquidus in the experimental measurement was 1169 C, and the solidus was 1118 C.
Even though the chemistry of the iron used in the model was different from that of the
iron used in the experimental case, the small difference in the solidus temperatures made
the effect on the solid fraction measurements minimal. The units in the r and z directions
37
on the graphs in these figures are in meters. The red portions of the slice are fully solid,
The fraction of solid after 64 seconds is shown in Figure 23a for a pouring
about 25% of the iron in this slice was less than 50% solid. About 55% of the iron at the
interface was solidified (red in Figure 23a) and 100% of the interface iron was at least
50% solid. The fully solidified iron was thickest at the top and bottom corners of the
drum and extended from the shell all the way to the inner diameter of the liner at these
two locations. The drum had three smooth undulations around the circumference of the
drum to increase the bonding surface area and a large undulation that is known as a
squeal band used as a dampening feature in brake drums. The squeal band was the
largest both in the axial and circumferential directions. The other three undulations were
the same size. As would be expected, the iron located in these four areas showed a
greater depth of solidification than the recessed portions of the drum, as these regions
would cool faster. The squeal band was solidified to a depth of 3 mm. The three smaller
undulations all had about 1 mm of solid in the bottom portions. The top corner of the
casting had solidified about 7 mm, and the bottom corner of the casting had solidified
about 13 mm.
Figure 23b shows the effect of increasing the shell temperature to 677 C while
in the shell temperature only increased the fraction of liquid iron (blue) 2%. The shape of
the solidification front remained the same, while the fraction of the iron fully solid at the
38
interface decreased from 55 to 42%, and the depth that was fully solid remained almost
the same at all four undulations. All of the interface iron was at least 50% solid.
Results after 64 seconds and a 1482 C pour time are shown in Figures 23c and
23d for shell temperatures of 510 C and 677 C, respectively. The top and bottom
corners of the drum and the squeal band still showed regions that were fully solid but the
area decreased significantly. The fractions of liquid increased to 11 and 29% in the lining
poured in a shell at 510 C and 677 C, respectively. The entire surface of the lining/shell
A summary graph showing the effect of pouring and shell temperature on the
percentage lining length that is 100% solid is shown in Figure 24. The decrease in
pouring temperature had a greater effect than the same decrease in the shell temperature,
as shown in Figure 24. A 300 C decrease in the shell temperature decreased the fully
solidified length less than 13%, while the same drop in the pouring temperature led to a
28% reduction in length. The pouring temperature of 1482 C is attainable but is hotter
than the iron was poured in the experimental trials. The shell preheat temperature of 677
C was also not obtained, as the shell cooled considerably in the 80 seconds in took to go
In order to delay solidification long enough to avoid gas entrapment if using the
preheating methods described here, the iron and shell preheat temperatures would need to
be higher than are practical in conditions for a production setting. However, if induction
heating was used to preheat the shell, the time to preheat the shell would drastically
decrease, from one hour to less than half a minute. If the time it takes to move the shell
from the furnace to the spinner and pour the iron (about 80 seconds) is added, the total
39
time the shell will have to oxidize before the iron is poured will be less than 2 minutes.
The amount of oxide formed would be totally converted in less than 4 seconds. This can
also be said for preheating in an inert atmosphere, since the shell would not oxidize while
in this inert atmosphere. The simulation of the casting shows that even in the best case
(1482 C pour and 677 C shell preheat temperatures) very little solidification occurs
after 4 seconds, as shown in Figure 25d. In these conditions, all of the iron at the
shell/liner interface was less than 25% solid. The 1316 C pouring temperature and a
677 C shell preheat temperature and a 1482 C pour and 510 C shell preheat
Figures 25b and 25c, respectively. Again, all the iron at the interface was less than 25%
solid, although the fully liquid iron area decreased slightly. At the lowest temperatures
(1316 C pour and 510 C shell preheat temperatures), a very thin layer of iron that was
around 50% solid formed at the undulations and corners of the drum, as shown in Figure
25a. These areas could entrap some of the evolved gas and cause a defect to occur.
A high enough pouring and shell temperature must be used to allow the gas
conversion of the iron oxide on the shell and movement of the gas through the liquid iron.
When a conventional air furnace was used, a large amount of oxide formed on the
surface. This would require a superheat greater than 430 C (the superheat used in the
1482 C pouring temperature, 677 C shell temperature model using typical class 30 gray
heating method that took less time to preheat the shell (induction furnace) so that the
shell would not be exposed to oxygen for an extended period, or using a furnace with an
inert atmosphere, the amount of oxide present would be minimal. In this case, a lower
40
superheat of about 180 C (based on the model at 4 seconds using the conditions of 2450
C pouring temperature, 677 C) would be needed to allow the conversion to take place.
(a) (b)
(c) (d)
Figure 23 2-dimensional slice of the casting showing the fraction of solid after 64 seconds. (a) initial
pouring temperature 1316 C, initial shell temperature 510 C. (b) Initial pouring temperature 1316 C,
initial shell temperature 677 C. (c) Initial pouring temperature 1482 C, initial shell temperature 510 C.
(d) Initial pouring temperature 1482 C, initial shell temperature 677 C.
41
100
90
Percentage Solidified at Shell Interface
80
70
60
50
Pour Temperature = 1482 C
40
30
Pour Temperature = 1316 C
20
10
0
500 520 540 560 580 600 620 640 660 680 700
Shell Temperature (C)
Figure 24 - Percentage of fully solid shell liner interface length vs. shell temperature after 64 seconds.
42
(a) (b)
(c) (d)
Figure 25 - 2-dimensional slice of the casting showing the solidification after 4 seconds. (a) is the casting
with the iron initially at 1316 and the shell initially at 510 C. (b) is the casting with the iron initially at
1316 and the shell initially at 677 C. (c) is the casting with the iron initially at 1482 and the shell
initially at 510 C. (d) is the casting with the iron initially at 1482 C and the shell initially at 677 C.
43
CHAPTER 5
Pouring trials were performed to develop a method to centrifugally cast gray iron
brake drum liners onto a steel shell. Transfer ladle temperature, pouring time, shell
rotation speed, and carbon equivalent were varied in the trials. There was no correlation
between transfer ladle temperature, shell rotation speed, or carbon equivalent on the
number of defects in the drum. Longer pouring times correlated with lower numbers of
defects.
The inner diameters of the drums were machined, and the drums were examined
and rated for the number of defects on the drums. Most of the defects in the drums were
identified as gas defects. Microstructural analysis of the defects showed that some of the
gas defects formed at the interface between the liner and the shell. The gas defects were
attributed to the conversion of iron oxide on the shell surface to gas during pouring.
The goal of the casting process is to form a continuous bond between the shell
and the iron liner by diffusion. Gas that forms from oxide on the shell can disrupt this
layer. The probability of defects occurring was shown to be dependent upon the time that
gas will evolve from the shell and the amount of time it took the liquid iron to solidify.
The time that gas evolved was dependent upon the amount of oxide present on the shell
to react with the liquid iron to form the gas. More oxide will form on the shell the longer
the shell is preheated in an oxygen-rich environment. The solidification time of the iron
was dependent upon the preheat temperature of the shell and the pouring temperature of
44
the iron. The gas that is formed from the oxide must be allowed to travel through the iron
before the iron solidifies or defects can form in the iron and disrupt the bond between
A dynamic computer simulation of the casting and solidification of the shell was
developed. This simulation allowed for prediction of the turbulence, solidification rate,
liquid strain rate, and surface defect concentration (likelihood of reoxidation) of the gray
iron lining. Unfortunately, even with modern computing technology, it still took about 4
weeks to model 4 seconds on a personal computer. To model the complete casting event,
of the iron shell. This simpler solidification model was a useful tool to examine the
environment of the casting once the pouring was completed and, using with the defect
analysis and gas formation calculations, provided a good prediction of the conditions
The centrifugal casting of an iron lining onto a steel shell could be a viable
process if care is taken to minimize the amount of oxide that is formed on the shell and to
control the temperature of the iron to maintain a sufficiently large superheat. The iron
liner has the thermal properties needed in a brake drum application, while the steel shell
has the strength to allow for a lighter product. The process also has a quicker casting
Future work on this problem would involve testing the production with a different
method of preheating the shell to minimize the oxide formed, to prove this product and
process can be superior to traditional methods of brake drum production. More thorough
45
data recording and defect analysis could also improve this casting trial to verify the
The solidification model used was verified by matching the cooling curves
measured using the infrared thermometer. A better model could be developed using
thermocouples and slip rings to more accurately measure the temperatures involved in the
casting.
to the chemistry of the iron. There was a great deal of discrepancy in the published data
and a lack of chemistry-specific data for gray iron. This affects many aspects of the
casting that are important for the model, such as the liquidus, solidus, enthalpy, viscosity,
and heat transfer coefficient. Using the information gained from the model to set up
more experimental trials could also improve and further verify the model.
46
LIST OF REFERENCES
2. Carter, S.F., Evans, W.J., Harkness, J.C., & Wallace, J.F. (1983). Factors Influencing
the Formation of Pinholes in Gray and Ductile Iron. AFS Transactions, 79-38, 245-
268.
4. Incropera, F.P., & De Witt, D.P., (1990). Introduction to Heat Transfer (2nd ed.).
New York: John Wiley & Sons.
5. Iordanova, I., Surtchev, M., Forcey, K.S., & Krastev V. (2000). High-temperature
surface oxidation of low-carbon rimming steel. Surface Interface Analysis. 158-168.
Retrieved June 11, 2007, from Compendex database.
7. Pehlke, R.D. (1982). Summary of Thermal Properties for Casting Alloys and Mold
Materials. National Technical Information Service. 53-60.
8. Perry, R.H., Green, D.W., & Maloney, J.O. (1997). Perrys Chemical Engineers
Handbook. (7th ed). New York: McGraw-Hill.
9. Sun, H., & Easman, W. (2007). Interfacial Phenomena and Reaction Kinetics
between the Carbon and Slag in the Ironmaking Process. Energy & Fuels. 413-418.
Retrieved June 11, 2007, from Compendex database.
47