Heat

Transfer
Fundamentals
for Metal
Casting

D.R. Poirier and E.J. Poirier

 TABLE OF CONTENTS

l. TYPES OF HEAT TRANSFER 1

A. Conduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
B. Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
C. Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

11. HEAT TRANSFER IN SOLIDIFYING METAL . . . . . . . . . . . . . . . . . . . . 30

A. Solidification of Pure Metal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

B. Solidlfication of Alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

111. HEAT TRANSFER IN MOLDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

A. Conduction in Mold Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

B. Heat Transfer across Casting Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

IV. SOLIDIFICATION HEAT TRANSFER . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

A. Solidijication in Thick Molds . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . . 41

B. Modulus Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
C. Solidification in Molds of Finite Thickness . . . . . . . . . . . . . . . . . . . . . . . . 48

V. RISER DESIGN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

A. Riser Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
B. Feeding Distance and Riser Location . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

VI. NUMERICAL METHODS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

A. Conduction in Mold or Core Surrounded by Metal . . . . . . . . . . . . . . . . . . 70

B. Conduction in Shell Mold witn Heat Loss at the Surface . . . . . . . . . . . . . . 77
C. Closing Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

SUGGESTED READINGS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

IX
 I. TYPES OF HEAT TRANSFER

Heat is transferred by three basic mechanisms: conduction, convection and

radiation. In the context of metal casting, conduction is the mechanism by which heat
is transferred internally within the solidifying metal and the mold. Convection within a
solidifying metal (i.e., the all-liquid zone) can have sorne important consequences, e.g.,
the heating of gates during pouring and macrosegregation in thick castings. In investment
castings, convection relates to the rate of heat loss from the outside surface of a shell-
mold to the surroundings, and it is in thís context that heat transfer by convection is
considered, herein. Radiation heat transfer can be important, because heat loss from
open risers or from the surfaces of preheated shell molds is largely by radiation.

Each of the three types of heat transfer is described in the following sections.

A. Conduction

There can be steady state conduction or transient conduction. In casting

applications transient conduction is by far the more prevalent scenario. However, it is
important to understand steady state conduction as a background to understanding
transient conduction.

Consider the plane wall shown in Fig. l. Toe surface at x = O is hotter than the
surface at x = L, so heat is conducted from left-to-right according to

(1)

X
x=O x=L
Figure 1: Steady state temperature distribution
in a plane wall.

Heat Transíer Fundamental• for Metal Cuting

D.R. Poirier and E.J. Poirier
The Mineralo, Metal• & Material• Soc:iety, 1992
where

Q = heat flow rate, Btu/h

A = area of the plate, ft2

L = thickness of the plate, ft

To,Ti = temperatures, ºF

k = thermal conductivity of the plate, Btu/h·ft· ºF

Toe thermal conductivity is a physical property, with a value that depends upon the
material and its temperature. Sorne data are given in Table 1 to illustrate typical values
for solids, liquids, and gases.

Examp/e 1

An insulating wall of glass wool is 6 in. thick with an arca of 100 ft2•
Calculate the rate of heat loss (Btu/hr) through the wall if its two surfaces are
maintained at SOºF and 80ºF, respectively.

From Table 1, the thermal conductivity is 0.022 Btu/h·ft· ºF. Toen, Eq. (1)
gives

0.022 Btu 100 ft 2 (80-50) ºF

Q
h·ft·ºF 0.05 ft

Q = 132 Btu/h

This is low value, indeed, which is not surprising because glass wool is a very good
insulator.

A better way to write Eq. (1) is

q -k dT (2)
dx

where

q = Q/A = heat flux, Btu/h·ft2

<1J; = temperature gradient, ºF /ft.

2
 Table 1

Thermal Conductivity of Various Materials at OºC (32ºF)

Material W/m·ºC Btu/h·ft· ºF

Metals:
Silver (pure) 410 237
Copper (pure) 385 223
Aluminum (pure) 202 117
Nickel (purc) 93 54
Iron (pure) 73 42
Carbon stcel, 1 % C 43 25
Lead (pure) 35 20.3
Chrome-nickel steel 16.3 9.4
(18% Cr, 8% Ni)

Nonmetallic Solids:
Quartz, parallel to axis 41.6 24
Magnesite 4.15 2.4
Glass wool 0.038 0.022

Liquids:
Mercury 8.21 4.74
Water 0.556 0.327
Lubricating oil, SAE 50 0.147 0.085

Gases:
Hydrogen 0.175 0.101
Helium 0.141 0.081
Air 0.024 0.0139
Water vapor (saturated) 0.0206 0.0019
Carbon dioxide 0.0146 0.00844

From J. P. Holrnan, Heat Transfer, sixth edition, McGraw-Hill,

New York, NY, 1986, p. 8.

Equation (2) is a form of Fourier's rate law, which is the basic equation for analyzing heat
conduction. Equation (2) is written for steady state and for unidirectional heat flow.
Therefore, temperature is only a function of x and the gradient can be written as a ful!
derivative. Toe negative sign is needed so that the heat is conducted from "hot" to "cold"
and never vice-versa.

3
 Consider Fig. 1 again. The gradient is

dT (3)
dx

Then Eqs. (2) and (3) give

or

Q To - TL (4)
q = - = +k --=--
A L
Of course, Eqs. (1) and (4) are the same. Furthermore, with T0 > TL the flux, q, is
positive so that heat is conducted from left-to-right (i.e., from "hot" to "cold").

The temperature distribution is linear only when steady state prevails. However,
when the situation is transient the temperature distribution is nonlinear (Fig. 2).
Consider a thin slice of the material with a thickness of ÁX. Notice that the gradient at
x is somewhat greater than the gradient at x + .1x. Therefore, the heat conducted into
the thin slice across the surface at x is greater than the heat conducted out of the slice
across the surface at x + .ix. Therefore, the energy (and temperature) within the slice
must increase. Ali of this can be expressed by simply making a mathematical statement
of the conservation of energy. This is written as

+ A.1xpC
ar
- (5)
Aq 1 = Aq I
éJt
x x+.1x p

Figure 2: Nonlinear temperature distribution

in transient heat conduction.

4
where

p = density, lbm/ft3

CP = heat capacity, Btu/lb., · ºF

f = rate of temperature increase, ºF /h.

At this point the two terms with the flux, q, should be familiar. Let's examine the
remaining term in more detail. AAxp is simply the mass of the slice; therefore the entire
third term represents a rate of increase of energy within the slice.

Equation (5) is used to derive the energy equation for unidirectional heat
conduction. After dividing each term by AAx, Eq. (5) is rearranged to the following
form:

q 1 x+áx - q Ix
_ e ar (6)
lim
,1.c-0 áx
p p -¡¡¡

Toe left side of Eq. (6) is the definition of the first derivative; so after the limit is taken,
we have
aq -pC
ar
-. (7)
ax p a,
Notice that partía! derivatives are used, because both q and T are functions of x and t.
Similarly, Eq. (2) should be written as

q -k ar (8)
ax
and after combining Eqs. (7) and (8), the energy equation for conduction heat transfer
resul ts. It is

ar _ 1 a [ k ar) (9)
a, pCP ax ax
lf k is uniform, then
ar (10)
Tt

5
where a is the thermal diffusivity and defined as
k
Ot = (11)

Toe units of a are ft2 /hr.

Actually Eq. (10) can be used as the starting point for solving either steady state
or transient problems. For example, in the steady state plane wall problem of Fig. 1,
there is no change of temperature with time so Eq. (10) reduces to

(12)

If the second derivative is zero, then the first derivative must be a constant. Hence

dT = C1
dx

and another integration gives

(13)

where C1 and C2 are integration constants. Notice that the temperature distribution is
linear, a result that was previously assumed.

There are two arbitrary constants in Eq. (13) because two integrations were
required; they are evaluated by two boundary conditions, A boundary condition is a
known value of the temperature or the temperature gradient at a particular value of x.
For example, suppose we know that
at X = 0 (14)

and
at X= L. (15)

By applying Eqs. (14) and (15) to Eq. (13), we get C2 = T0 and C1 = (TL - T0)/L.
Therefore,

(16)

Immediately we see that the slope or temperature gradient is (TL - T0)/L so that the flux
(q) and heat flow rate (Q) are

q = -k dT (17)
dx

6
and
kA
Q = qA = L (To - Ti) (18)

Notice that with Eq. (18), we are back to Eq. (1). A complete circle has been
made, but that has not been the intent. Rather by recognizing that Eq. (10) is valid for
transient and steady state conduction, it serves as a logical starting point. Toe other
necessary relationship is Eq. (8), which enables us to connect the temperature distribution
and the flux.

In Example 1 you might bave wondered how there could be a plane wall of glass
wool. Realistically, of course, the glass wool would be between two supporting walls, as
depicted in Fig. 3. At steady state, Eq. (12} applies to each of the three Iayers so that
temperature must be linear in each. However, the thermal conductivities are not equal
so that the gradients within eacb are different. With contact thermocouples, we could
measure T1 and T4, but it is unlikely that we would measure T2 and T3• We seek those
interna! temperatures and the heat loss (Q) through the composite wall.

T1
y(k2)
(k,)~ (k)
T4
81 L 82

Figure 3: Steady state temperature in a composite wall.

A convenient way to attack the problem is to use Ohm's law as an analog to

Eq. (18). Ohm's law is
llE (19)
l= -
R
where lis the current (i.e., flow of coulombs), llE is the voltage (í.e., potential
difference) and R is the resistance. In Eq. (18) Q is flow of heat so it is analogous to
I, and (T0 - TL) is a temperature drop and analogous to llE. That leaves the thermal
resistance, which is defined as
L (20)
RT = -.
kA

7
 Now returning to Fig. 3, there are three thermal resistors: 81/k1A, L/kA and
82/k,.A.. Toe three resistors are in series so that the total thermal resistance is

(21)

Toe temperature drop across the three resistors is (T1 - T4) so that the heat flow rate
is
(22)

Because T1 is known, we can then determine T2 because

(23)

and similarly T4 is known so that T3 can be deterrnined from

(24)

Example 2

Toe 6 in. of glass wool in Example 1 is supported by 1/4 in. thick sheets of
stainless steel. Toe externa! surface temperatures are SOºF and 50ºF; determine
the rate of heat loss through the composite wall and the interna! temperatures.

From Table 1, the thermal conductivity of stainless steel is

9.4 Btu/(h·ft· ºF). Toe thermal resistance of the wall is

2 h·ft·ºF 0.25 in 1 ft 1 h·ft·ºF 0.5 ft

RT = ----------,,--+--- + -----t----=-
9.4 Btu 100 ft 2 12 in 0.022 Btu 100 ft 2

RT = (4.43 X 10-5 + 10-1)h·ºF/Btu

(Notice that practically ali of the resistance is within the glass wool, itself, because
it is an excellent insulator.)

Now

Q (80-50)ºF I Btu
2.27 X 10-1 h·ºF

Q 132 Btu/h

8
 (Toe result is the same as Example l.)

(Ti _ T2) = _1_32...,...h_B_tu_t_4_._43_x--=-10_-s_h·_º_F 0.003 ºF

2 Btu

and

(Toe temperature drop through each sheet of stainless steel is very small because
the thermal resistance of each sheet is negligible, when compared to the thermal
resistance of the glass wool.)

Now let's considera transient problem that has application in materials processing,
including metal casting. A very thick solid is initially at a uniform temperature, T;, and
then its surface temperature is abruptly raised to and maintained at T0 (see Fig. 4a).
Immediately a very steep temperature gradient is set up at the surface so there must be
heat conduction into the solid. Temperature varies with distance x and time t as shown
schematically in Fig. 4b.

Temperature must satisfy Eq. (10), which contains a first derivative with respect
to time (t) and a second derivative with respect to distance (x). Therefore, to get a
particular solution for T(x,t) without any integration constants, we specify an initial
condition and two boundary conditions, They are as follows:

lnitial Condition

T(x,O) = T1, x~O (25)

X X
(o) (b)

Figure 4: Transient temperature in a semi-infinite thick solid: (a) solid with uniform
temperature T¡, (b) temperature distributions for times t1 < t2 < t3•

9
 Boundary Conditions

T(O,t) = To, t > o (26)

T( oo,t) = T;, t ~ o (27)

Equation (25) expresses in mathematical form that, at t = O, the temperature is uniform

and equal to T¡. Equation (26) states that, at the surface of the solid (x = O), the
temperature is maintained at a constant temperature T0. Equation (27) simply states that
the temperature must decrease toward T1 with an increase in x. For mathematical
convenience, x -+ oo so the solid is considered to be semi-infinite.

Toe solution that satisfies Eq. (10) and Eqs. (25)-(27) is

(28)

where the error function is defined as

erf [-x2{;c ] 2
.¡;
sl2"Vr>J

f
o
e-~1d,¡. (29)

Readers with experience in applied statistics might be familiar with the error function.
For the uninitiated, simply think of it as you think of any other function ( e.g., sine,
logarithm, square root, etc.). In other words if the argument of the function is known
then the function can be determined from a tabulation.

Toe error function is given in Table 2. To satisfy ourselves that sensible results
are obtained, let's use Eq. (28) and calculate the temperature at t = O. Toen the
argument of the error function is oo, and Table 2 indicates that the error function must
be 1, its máximum value. Toen Eq. (28) gives
T - T0
1,
T¡ - T0

so that T = T1• Of course, we get back the initial condition. Now calculate the
temperature at the surface for t > O. Toe argument of the error function is O, Table 2
gives an error function of O, and Eq. (28) gives
T - T0
o.
T1 - T0

Therefore, T = T0 at x = O, which is one of the boundary conditions. Finally at x = oo,

it is easy to see that T = T1, which is the other boundary condition.

10
 Table 2

The Error Function

N erf N N erf N N erf N

0.00 0.00000 o.so 0.5205 1.1 0.8802

o.os 0.05637 O.SS 0.5633 1.2 0.9103
0.10 0.1125 0.60 0.6039 1.3 0.9340
0.15 0.1680 0.65 0.6420 1.4 0.9523
0.20 0.2227 0.70 0.6778 1.5 0.9661
0.25 0.2763 0.75 0.7112 1.6 0.9763
0.30 0.3286 0.80 0.7421 1.7 0.9838
0.35 0.3794 0.85 0.7707 1.8 0.9891
0.40 0.4284 0.90 0.7969 1.9 0.9928
0.45 0.4755 0.95 0.8209 2.0 0.9953
1.00 0.8427 OC> 1

For casting applications, it is more important to know the temperature gradient

than it is to know the temperature distribution given by Eq. (28). Without an exposition

l
of the cakulus involved, the temperature gradient obtainable from Eq. (28) is

et T¡ - To [ X 2 (30)
ax = JT<xt exp - 4at .

Notice that the gradient varíes with both x and t. Ata fixed time (e.g., t = t2 in Fig. 4b),
the gradient decreases with increasing x; at a particular location, the gradient decreases
with increasing time.

Example 3

The surface of a thick mold of plaster, initially at 80ºF, is abruptly heated

to lOOOºFand maintained at that temperature. If the surface has an area of 1 ft2,
determine the heat flow rate into the mold at 1, 5 and 15 minutes. The properties
of the plaster are k = 0.277 Btu/h·ft· ºF, p = 90 lbm/ft3, CP = 0.201 Btu/lb¿ · ºF,
and a = 0.0153 ft2/h.

Ali of the heat absorbed by the mold must pass through its surface, so we
should evaluate the flux at x = O. This is

% = -k. [ oT)
OX
,x=O

11
and with Eq. (30) we get

(31)

Notice that q0 varies with time. Toen, at t = 1 min. (1.667 X 10-2 h),

ºº = qo-4 = kA (T0
¡;¿;¡-
- T¡)

Qo = (0.227)(1)(1000-80) = 7.38 x 1()3 Btu/h

k0.0153· 1.667 X 10-2 )112

Similar calculations are done for t = 5 and 15 min. Toe results are plotted in
Fig. 5.

s: 8000

-
<,
::::,
CD
04000
o

4 8 12 16 20
Time, min
Figure 5: Rate of heat conducted into a plaster mold.

Example 4

Estimate the thickness of plaster mold in Example 3 so that it can be

considered to be serni-infinite thick.

Toe heated surface of the mold is at x = O. Let the thickness of the mold
be L; then when the temperature at x = L increases, the mold can no longer be
considered serni-infinite thick. Reference to Table 2 shows that the error function
is practically at its maximum value of 1 when N = 2. Therefore,
X
= 2.
2..¡at

12
 We set x "' L and solve for L, when t = 1/60 h.
L (2)(2)(0.0153/60)1n

L 0.0639 ft = 0.766 in.

For t = 5 min and 15 min, the results are L = 1.714 in. and 2.969 in., respectively.

If the mold has a thickness Jess than these calculated values of L, then it is not
serni-infinite, and thc solution for the error function is invalid. With the argument
selected as 2, then at x = L

T - To = T - lOOO = 0.995322
T; - T0 80-1000

and T = 84.3ºF, which is sufficiently close to the initial temperature that the mold
is semi-infinite thick. If we had selected a more stringent requirement
(e.g., x/2.f;t = 3), then the calculated values of L would have becn smaller.

B. Convection

It is well known that a heated solid will cool faster when placed in front of a fan
than when it is placed in still air. Cooling is said to be by convection heat transfer. To
deal with convection quantitatively, we must specify or know the velocity, type of
convection, the thermal properties of the fluid, and the geometry of the solid surface.
For now we qualitatively describe convection heat transfer and relate it to conduction.

Consider the heated plate shown in Fig. 6, which is being cooled by the passing
fluid. The velocity of the fluid is shown as v_.(y); notice that the velocity is zero at the
surface of the plate because of the viscous nature of the fluid. Exactly at the surface,
therefore, heat is transferred only by conduction, and the flux at the surface (Qo) is

% = -k [ ,JT)
dX
.
x=O
(32)

where the gradient is in the fluid. If we are able to calcula te the gradient at the surface,
then we can get the heat transfer from the plate to the cooling fluid and, in turn,
calculate the temperature distribution within the plate. However, that is a "big if,"
because the gradient depends upon the details of the fluid flow, which can be analyzed
but only with great effort. Therefore, we almost always rely on empirical data or on
results of previous analyses of the fluid flow to get the information required to solve the
heat transfer.

To express the effect of convection we use

(33)

13
 v.,_(y)

x To Surface

Figure 6: Convection heat transfer to a surface.

where

T0 = surface temperature,

and

T.., = temperature of the fluid away from the surface.

The quantity h is called the convection heat transfer coefficient or simply heat transfer
coefficient. The units of h are Btu/h·ft2• ºF. Typical values are given in Table 3.

In Table 3, reference is made to forced convection and free convection. A fan

blowing air, water flowing in a pipe, and relative motion between a solid in a fluid are
examples of forced convection. If a heated plate is exposed to ambient air, there is a
movement of air as a result of density gradients near the plate. This type of convection
is called free or natural convection.

Example5

Air at 75ºF blows pasta heated face of a slab with dimensions of 6 in. by
6 in. Toe slab is maintained at 300ºF and h = 10 Btu/h·ft2• ºF. Calculate the
heat transfer rate.

Equation (33) is used as follows:

10 Btu 0.25 ft 2 (300-75)ºF

(34)
h·ft 2·F

562.5 Btu/h
14
 Table 3

Convection Heat Transfer Coefficients

Mode W/m2·ºC Btu/h·ft2• ºF

Free convection, AT = 30ºC

Vertical plate 0.3 m [ 1 ft] high in air 4.5 0.79
Horizontal cylinder, 5-cm diameter, in air 6.5 1.14
Horizontal cylinder, 2-cm diameter, 890 157
in water

Forced convection
Airflow at 2 m/s over 0.2-m square plate 12 2.1
Airflow at 35 m/s over 0.75-m square plate 75 13.2
Air at 2 atm flowing in 2.5-cm-diameter 65 11.4
tube at 10 m/s
Water at 0.5 kg/s flowing in 3,500 616
2.5-cm-diameter tube
Airflow across 5-cm-diameter cylinder 180 32
with velocity of 50 m/s

Boiling water
In a pool or container 2,500-35,000 440-6,200
Flowing in a tube 5,000-100,000 880-17,600

From J. P. Holman, Heat Transfer,sixth edition, McGraw-Hill, New York, NY, 1986,
p. 13.

Figure 7 depicts an interaction between convection and conduction. In Fig. 1 the

surface temperatures were known a priori and so they could be specified. More often
than not, however, the temperatures of the fluids (T001 and T002) are known, and then we
seek the temperature distribution within the salid that depends upon the temperatures
T"" 1 and T002 and the corresponding heat transfer coefficients.

At steady state, the temperature distribution within the plate is linear so that

(35)

and

(36)

15
 X
x=O x=L
Figure 7: Steady state temperature distribution in a plane wall
between two fluids at different temperatures.

Toe two unknown temperatures, T0 and Ti, can be determined by simultaneously solving
Eqs. (35) and (36).

An easier solution method is to use the electrical analog for Eq. (33); i.e.,

Q = (To - T.. )
Rr
where Rr is the thermal resistance for convection heat transfer and is defined as

(37)

In Fig. 7, there are three resistors in series, so the total resistance is

L 1
+ - + --· (38)
kA h2A

Toe temperature difference is T .. , - T.. 2, so that

T,,.1 - T.. 2
Q (39)
RT
and the unknown temperatures are

T °"' - To (40)
T,,.1-T .. 2

16
and

+ h2
h¡
i-· (41)

Example 6

A 1/4 in. thick plate of stainless steel (k = 9.4 Btu/h·ft· ºF) is at steady state
with moving air on one side (h = 5 Btu/h·ft· ºF, T = 80ºF) and stagnant air on
the other (h = 1 Btu/h·ft·ºF, T = 50ºF). Calculate the heat flux and the surface
temperatures.

The thermal resistance is given by Eq. (38).

Rr = [ .!_ + (1/48) + .!_]

5 (9.4) 1 A

º
1.2 22 h · ºF /Btu.
A

Then the flux is

Q Toot - T 002
q
A AR7

(80-50) ºF· Btu

24.954 Btu/h.
1.2022 h·ºF

Equations (40) and (41) can be used to calculate the surface temperatures.

T 001 T0 = [ 1 + (5)~\48) + fJ (80-50) = 4.99ºF

r
- 1

Ti - T 002 = [1 + (l)~\48) + { (80-50) = 24.95ºF

Therefore, the surface temperatures are 7SºF and SSºF.

The field of heat transfer cuts across many disciplines and, consequently, over the
years much effort has been devoted to experiments and analyses on convection heat
transfer. A few of the many results are presented here.

17
 For laminar flow parallel to the surface of a plate, the heat transfer coefficient is
cakulated from

Nul = 0.664 Re¿'2 Pr 113• (42)

Equation (42) is typical for forced conveetíon, in that the heat transfer coefficient is
correlated in terms of three dimensionless groups: a Nusselt number (Nui), a Reynolds
number (Rei), and a Prandtl number (Pr). They are defined as follows:
hL
Nu¿ = -s-:
k
(43)

LV p
(44)
00

and

Pr (45)

where L is the length of the plate, V.. is the free-strearn velocity of the fluid, p is density,
µ. is viscosity, CP is heat capacity, k is thermal conductivity, and ali thermal properties (k,
p, µ., and C ) are for the fluid. Inspection of the dimensionless groups shows that
Eq. (42) embodies the effect of a particular fluid, the length of the plate, and the velocity
of the fluid on the heat transfer coefficient.

The properties of sorne fluids encountered in metal casting are given in Tables 4
through 6. Notice that the fluid properties vary somewhat with temperature; this is
especially so for gases. When there is an appreciable difference between the surface
temperature of the solid (T0) and the free-strearn temperature (T 00) of the fluid, it is
recommended that the properties be evaluated at the so-calledji/m temperature. The film
temperature is defined as

(46)

When using correlations, such as Eq. ( 42), one must be aware of its limitations.
Equation (42) is valid provided 0.6 ~ Pr ~ 50. It does not apply to fluids with very low
Prandtl nurnbers, like liquid metals, or to fluids with very high Prandtl nurnbers, like
heavy oils. Equation (42) is also restricted to laminar flows so that Re¿ < 106 (approx.).

l
If flow is turbulent, then

Nul = ..!. se¿ Pr 1/3 [ 0.455 - 3300 (47)

2 (log Rei)2·584 ReL

18
 Table 4

Thermal Properties of Air at Atmospheric Pressure

µ, a,
p CP' kg/m·s k, m2/s
T, K kg/m3 kJ/kg·ºC X 10S W/m·ºC X 104 Pr

300 1.1774 1.0057 1.8462 0.02624 0.22160 0.708

400 0.8826 1.0140 2.286 0.03365 0.3760 0.689
500 0.7048 1.0295 2.671 0.04038 0.5564 0.680
600 0.5879 1.0551 3.018 0.04659 0.7512 0.680
700 0.5030 1.0752 3.332 0.05230 0.9672 0.684
800 0.4405 1.0978 3.625 0.05779 1.1951 0.689
900 0.3925 1.1212 3.899 0.06279 1.4271 0.696
1000 0.3524 1.1417 4.152 0.06752 1.6779 0.702
1200 0.2947 1.179 4.69 0.0782 2.251 0.707
1400 0.2515 1.214 5.17 0.0891 2.920 0.705
1600 0.2211 1.248 5.63 0.100 3.609 0.705
1800 0.1970 1.287 6.07 0.111 4.379 0.704
2000 0.1762 1.338 6.50 0.124 5.260 0.702
2200 0.1602 1.419 6.93 0.139 6.120 0.707

From F. P. Incropera and D. P. DeWitt, Fundamentals of Heat and Mass Transfer, third
edition, John Wiley & Sons, New York, NY, 1990, p. AlS.

The values of µ, k, Cp, and Pr are not strongly pressure-dependent and may be used up
to 10 atm.

Heat transfer coefficients for free convection are represented in the following form
for many circumstances:
Nu = C(GrPrr (48)

where C and m are constants that are given in Table 7.

In general, the Nusselt number contains a characteristic length of the solid surface.
For free convection next to vertical planes and vertical cylinders, the characteristic length
is the vertical length, L. For horizontal cylinders, the diameter, D, is the characteristic
length.

The Grashof number, Gr, in Eq. (48) is a measure of the strength of the free
convection, whereas the Reynolds number, Re¿, is a measure of the strength of forced
convection. The Grashof number is
Gr = g(jp2 (To - Too) L 3 (49)
µ.2

19
 o o o o o o - >.
~
-
00020000
"'l)~r~ ...~
1""""'..-tl""""'IO.....,..,...._..._.
~
=l. s
X X X X X X X e:
~ X .e:
C>o ~ o:_.;_.;_.;M"-t\Oex>....:
~Ó:<'l~~~g ..,o
¿-
.g
:-a
~~~~~~c;~:=:~vi~
- ....
<I)

t C"')T""""lt'-V"IVt"")rt1NN ..... _. ...... -e

....
:g
.:

-
~
u
.:; e
o ::8 V'I
V)
r-V)
V'I '<t O
O\ - <'\
V) \O \O
i '<t V'I <'\ 00 V'I V'I
V) \O !' r- 00 00
\O \O \O \O \O '° ~
¡.;.;

....
~ ºººººººººººº ~
ij
~
~"'
§
...b
.,.,
<I)
:E
....o
·!"'....
<I)
. "'e
=l.-
b.()
.lo:
-
X
~~~
-
b
X
\Olxl ~r='.8~ 8 ~g¿
~
~
'o-
~
i
Q.
.....:,.....;.....;oo\O..,..;"lf!t~rri
r"'iN_;
~ ...
o

1
e,
~
...
El _...s 0000\000000r'l<'INNI'-
<'i gg~f::~~~ ~ ~
¡:;
<I)
<:!.- ~ g:~vi
ee
.lo: °' °'°' ~~ °'°'°'°' °' °'°' ..:
-~N
¡.....
~~ <I
u ': ci. ~

- ~~~~~~~~~~~r=:
- u~Ii
o i::i.. • .-----,
u.._~ NN,......t_._._.. ...... _._.NNN
..¡. ..¡. ..¡. ..¡. ..¡. ..¡. ..¡. ..¡. ..¡. ..¡. ..¡. ..¡. ci~
~ ~e: • 1.
a.,,z>
N

...
&.~-
'° o o

--
r-- 00 °'
u
'<t
'<t V'I \O e-- 00 -- N
N <'\
<'\
\O 00 t >-
o o..fvi_¿r-:oc50....;csi<'ivir-:
-N<'l'<t\0!'000\-r'l .s :l t.....
¡:¡.,z
• <I)
c.,
¡¡,; vi'
El § ~
¡;¡, ~~@a68~~@a68~a6 oVl
o _._. ....... .,.......,..NNN
i.t ~ zo

20
 -... .o...o
"O
i::
r:! :::,
c.. z
E
-
O\
o
o

'° in '° s:t oó -oo-~r-

'<t\C .... 0\-
oósiin...;oc,;
T""""'T"""'(""'""~rf") ..... '<t <'} 00 V)

?;- "'
"ü; • b
o E,..... ~~~~~~~~~~~~ O\ ~00
.....,.. 'l""""I V)~
\C
<.> ........
.~
>.!.<: e.o X ..-.of'i..-.00..-.00000
ºººº
e,.. ...
·;;;eº
i:: .............
Q~X

e
:::,

~u

ª
ºº
¡!:

......
r-
N
.....
O\
-
....
'

.e
;
E
~"'

21
 Table 7

Constants for Use with Eq. (48) in Free Convection Situations

Geometry GrPr e m

Vertical planes and cylinders 104-109 0.59 1/4

109-1013 0.021 2/5
109-1013 0.10 1/3
Horizontal cylinders 0-10-s 0.4 o
104-109 0.53 1/4
109-1012 0.13 1/3
10-10_10·2 0.675 0.058
10-2-102 1.02 0.148
1ü2-104 0.850 0.188
104-107 0.480 1/4
101-1012 0.125 1/3
Upper surface of heated plates 2 X 104-8 X 106 0.54 1/4
or lower surface of cooled plates"
Upper surface of heated plates 8 X 106-1011 0.15 1/3
or lower surface of cooled plates"
Lower surface of heated plates lef-1011 0.27 1/4
or upper surface of cooled plates"
Vertical cylinder, 104-106 0.775 0.21
height = diameter
characteristic length = diameter

From J. P. Holman, Heat Transfer, sixth edition, McGraw-Hill, New York, NY,
1986, p. 333.

• L = {t:t;, for rectangular plates with dimensions t I by t2•

where fJ is the volume coefficient of expansion and g is gravitational acceleration. For

ideal gases

fJ = _!_ (50)
T
where T is the absolute temperature. Values of GrPr for water are conveníently
obtainable from the last column in Table 5.

22
 Examp/e 7

A vertical plate (1 ft by 1 ft) is maintained at 320ºF and is exposed to air

at 80ºF. Compare the heat transfer coefficients for free convection and forced
convection with a velocity of 20 ft/s.

First we gather the properties of air at T1= 1/2(320 + 80) = 200ºF (366 K).
From interpolations in Table 4:

p = 0.9828 kg/m3 = 0.0614 lbm/ft3

CP = 1.011 kJ/kg·K = 0.239 Btu/(lbm· ºF)

µ = 2.316 X 10·5 kg/m·s = 0.0517 lbm/ft·h

k = 0.03113 W/m·K = 0.0180 Btu/h·ft·ºF

Pr = 0.695

Also

{3 = 1/T1= 1/(200 + 460) = 1.515 X 10·3ºr1

g = 32.2 ft/s2 = 4.173 X 108 ft/h2

Toen

g{3p2(T0 - T,.)L3
Gr=

4.173 X 108 ft 1.515 X 10-J 0.06142 lb~ (320-80)ºF

ºF ft6

2
13 ft3 1 ft2·h
0.05172 lb~

1 ft 20 X 3600 ft 0.0614 lbm ft·h

h ft J 0.0517 lbm

= 8.55 X 1()4

23
 For free convection, Gr-Pr = 1.49 x 108 and Table 7 gíves C = 0.59 and m = 1/4.
Toen, using Eq. (48):
hL 0.59 ( 1.49 X 108)
114
k

65.16

11 (65.16)(0.0180) = 1.1? Btu/h·ft 2. ºF

1

For forced convection, Rei < 106 so that Eq. (42) applies. Toen

Nu¿ = hL = 0.664 (8.55 X 104 )11\0.695)113

k

= 259

h (259)(0.0l80) = 4.66 Btu/h·ft 2· ºF

Notice that the heat transfer coefficient for forced convection is almost four
times that for free convection.

C. Radiation

In contrast to conduction and convection, with heat transfer through a material,

heat may also be transferred through a vacuum. Toe mechanism is by electromagnetic
radiation, specifically thennal radiation.

An ideal thermal radiator, or blackbody, emits energy at a rate proportional to the

fourth power of the absolute temperature of the body and directly proportional to its
surface area. Thus
Q = uAT4 (51)

where u is called the Stefan-Boltzmann constant with the value of 0.1714 X 10·8
Btu/h·ft2· ºR4• Equation (51) is called the Stefan-Boltzmann law of thermal radiation,
and it applies only to blackbodies.

Toe net radiant exchange between two surfaces, expressed as a flux, is proportional
to the difference in absolute temperatures to the fourth power; i.e.,

q ex u(Tt - T/) (52)

24
 We have mentioned that a blackbody is a body which radiates energy according
to the T4 Jaw. We call such a body black because black surfaces, like a piece of metal
covered with carbon black, approximate this type of behavior. Real surfaces, like a
polished metal plate, do not radiate as much energy as the blackbody and are
approximated as gray surfaces. However, the total radiation emitted by these bodies still
generally follows the t; proportionality. To take account of the gray nature of such
surfaces we introduce another factor into Eq. (51), called the emissivity E, which is the
ratio of the radiation of the gray surface to that of an ideal black surface. In addition,
we must take into account the fact that not ali the radiation leaving one surface reaches
the other surface because electromagnetic radiation travels in straight Jines and sorne will
be Iost to the surroundings. We therefore introduce two new factors in Eq. (52) to take
into account both situations, so that
(53)

where F is a function of the emissivities and the geometries of the two surfaces. Toe
determination of the form of this function for specific configurations is a major subject
of radiation heat transfer. Here, we consider only a few cases that can be applied to
metal casting. In general, however, radiation heat transfer can be very complex.

A simple radiation problem is encountered when we have a surface, at

temperature Ti, completely enclosed by a much larger surface maintained at T2• Toe net
radian! exchange in this case can be calculated with
(54)

Another situation is depicted in Fig. 8, where there are two radíatíng circular discs
that are separated by a space bounded with an insulator. Toe net radiant exchange is
given by Eq. (53) whereA = area of each disc (1rd2/4) and

F = [ 2 - 2F~2 + _.!. + _.!. - 2]-1 (55)

1 - F12 E¡ E2

( 1)

lnsulator TX

1
----d----
.,,,,..,.,,------- ........
/ '
(2)

Figure 8: Parallel discs separated by an insulated space.

25
In Eq. (55), F12 is called the view factor, it is given in Fig. 9. Emissivities of various
materials are in Table 8.

-- --
(\J
1.0
o..1..
X
t.C o.a
.: ~9~
-
o
u
a
LL 0.4
0.6
/
V"
·~

/
Q)
a.
o
s: 0.2
(/)
/ 2 3 4 5 6 7
Ratio d/x
Figure 9: Shape factor for radiation between parallel circular disks.
(From J. P. Holman, Heat Transfer, sixth edition,
McGraw-Hill, New York, NY, 1986, p. 388.)

Example 8

A shell mold with a surface area of 1 ft2 is preheated to 1200ºF. When it is

removed from the preheat furnace, what is the rate of heat loss? Assume that the
emissivity of the mold is 0.5.

In this siruation, Eq. (54) applies with E1 = 0.5, A1 = 1 ft2, T1 =

{1200 + 460)ºR, and T2 = (80 + 460}°R. Toen

Q , (05)(0.1714)(1) [ ¡ ·.: r _ ¡ :: r l
= 6.435 X lü3 Btu/h

Notice that the value of c (0.1714 X 10-9 Btu/h·ft2• ºR) has been split into two parts:
0.1714 and each temperature divided by 100. At this relatively high temperature, the
radiation heat transfer is approximately 2 to 4 times that for convection heat transfer.
As a general rule, when T ~ lOOOºF (approx.) radiation heat transfer usually
dominates.

26
 Table 8

Ernissivities of Various Surfaces

Surfacc T, ºF Emissivity E

Alumioum:
Highly polished plate, 98.3% pure 440-1070 0.039-0.057
Commercial sheet 212 0.09
Hcavily oxidized 299-940 0.20-0.31
Brass:
Highly polishcd:
73.2% Cu, 26.7% Zn 476-674 0.028-0.031
62.4% Cu, 36.8% Zn, 0.4% Pb, 0.3% Al 494-710 0.033-0.037
82.9% Cu, 17.0% Zn 530 0.030
Hard-rolled, polished, but direction of polishing visible 70 0.038
Dull plate 120-660 0.22
Iron and steel (not including stainlcss):
Steel, polished 212 0.066
Iron polished, 800-1880 0.14-0.38
Cast iron, newly turned 72 0.44
tumed and heated 1620-1810 0.60-0.70
Mild steel 450-1950 0.20-0.32
Sheet stcel with strong, rough oxide layer 75 0.80
Magnesium, magnesium oxide 530-1520 0.55-0.20
Monel metal, oxidized at 1110°F 390-1110 0.41-0.46
Nickel:
Polished 212 0.072
Nickel oxide 1200-2290 0.59-0.86
Nickel alloys:
Coppcr nickel, polished 212 0.059
Nichrome wire, bright 120-1830 0.65-0.79
Nichrome wire, oxidized 120-930 0.95-0.98
Stainless steels:
Polished 212 0.074
Typc 301; 8 450-1725 0.54-0.63
Alumina (85-99.5%, Al20,, 0-12% SiO,, 0-1% GtiO,);
effect of mean grain size, microns (µm):
10 µm 0.30-0.18
50 µm 0.39-0.28
100 µm 0.50-0.40
Asbestos, board 74 0.96
Brick:
Red, rough, but no gross irregularities 70 0.93
Fireclay 1832 0.75
Quartz, rough, fuscd 70 0.93

Selected from J. P. Holman, Heat Transfer,sixth edition, McGraw-Hill, New York, NY,
1986, pp. 648-649.

27
 In the development of convection heat transfer, we found it convenient to define
a heat transfer coefficient by

Because radiation is often closely associated with convection, and the total heat transfer
by both convection and radiation is sought, it is worthwhile to put both processes on a
common basis by defining a radiation heat-transfer coefficient h, as
(56)

where T1 and T2 are the absolute temperatures of the two bodies exchanging heat by
radiation. The total heat transfer is then the sum of the convection and radiation,

(57)

if we assume that the second-radiatíon-exchange surface is at the same ternperature as

the fluid (i.e., T2 = T .. and T1 = T0). For example, the heat loss by free convection and
radiation from a hot steam pipe passing through a room could be calculated from
Eq. (57).

In many instances the convection heat transfer coefficient is not strongly

dependent on temperature. However, this is not so with the radiation heat-transfer
coefficient. The value of h,, corresponding to Eq. (53) is
(58)

Obviously, the radiation coefficient is a very strong function of temperature.

Example 9

Estímate the rate of heat loss from the top of an open riser (2 in. dia.) of
molten steel. The freezing point of the steel is 2732ºF (1500ºC), and the
emissivity of molten steel is 0.28.

The total heat loss is by free convection and by radiation. Consider

convection first and gather the properties for air at T1 = (2732 + 80)/2 = 1406°F
(1036 K). From Table 8:

p = 0.3420 kg/m3 = 0.0214 lbm/ft3

µ = 4.249 x 10-s kg/m·s = 0.103 lbm/ft·h

k = 0.06944 W/m·K = 0.0401 Btu/h·ft·ºF

Pr = 0.703

28
Also

fj = l/T1 = 1/(1406 + 460) = 5.36 x 10·4 ºr1

L = diameter = 0.167 ft

g = 4.173 X 108 ft/h2

Toen

Gr = (4.173 X lü8)(5.36 X 10-4) (0.0214)2 (2732-80)(0.167)3

(0.103)2

= 7.14 X 105

Gr·Pr = 5.02 x 105

Table 7 gives C = 0.54 and m = 1/4.

Nu = hL = 0.54 (5.02 X 105)114 = 14.37

k

and

h = (l4.37) (0.040l) = 3.45 Btu/h·ft 2• ºF

0.167
For the radiation heat transfer coefficient, we use Eq. (58) with F = E = 0.28,
T1 = (2732 + 460)ºR and T2 = (80 + 460)°R. Toen

h, = (0.28)(0.1714 X 10-8)(31922 + 54Q2)(3192 + 540)

= 18.77 Btu/h·ft 2·ºF

The total heat loss is

= (3.45 + 18.77) [-¡ · 0.1672) (2732-80)

= 1,290 Btu/h

Notice that 84.5% of the heat loss from the top of the riser is by radiation. With
a metal with a lower melting point (e.g., Al at 1220ºF), the heat loss is only
141 Btu/h and the percentage byradiation is reduced to 63.4%.

29
11. HEAT TRANSFER IN A SOLIDIFYING METAL

A. Solidification of Pure Metal

To determine the time for a casting to solidify, one can estimate the energy
absorbed by the mold and equate that energy to the energy extracted from the solidifying
metal. Thus there must be an accounting of the energy in the metal. In this section we
consider a pure metal.

Toe energy of the metal is expressed in terms of its enthalpy, Toe enthalpy of a
substance has no absolute value, so we use its enthalpy re lative to a reference state. As
an example, consider iron. Like many elernents, iron is allotropic. Below 1184 K
{1672ºF) its crystalline form is body centered cubic (b.c.c.). Upon heating at 1184 Kit
transforms to face centered cubic (f.c.c.), and with continued heating to 1665 K {2538ºF),
it transforms back to b.c.c. Finally, it melts at 1809 K (2797ºF).

With the reference state taken as b.c.c. at 298 K, the enthalpy is O. Toen the
enthalpy of iron can be represented as shown in Fig. 10. Notice at each transfonnation
temperature, there is a jump in the enthalpy. For example, consider the transformation
at 1184 K; the reaction is
ex (b.c.c.) = -y (f.c.c.)

and upon heating through 1184 K the change in the enthalpy is only

MI = H.., - Ha = 225 cal/mol

= 7.25 Btu/lbm

Now consider the solidification reaction at 1809 K. Upon cooling liquid transforms to
solid, labeled ó, and the associated change in enthalpy is

MI = H6 - HL = -3630 cal/mol

= -117 Btu/lbm

This MI is called the latent heat of solidification. Toe negative sign tells us that when
iron solidifies, latent heat is evolved from the iron, and this heat is available as a source
of energy. If we think about melting (i.e., fusion), then

and this MI is called the laient heat of fusion and given the symbol H1.

Now suppose we fill a small ladle with liquid iron exactly at its freezing point
(2797ºF). Toen there will be heat losses by convection and radiation from the top of the
melt and by conduction through the wall of the ladle. When the total heat lost equals

Heat Tranofer Fundament&lofar Metal Casting

D.R. Poirier and E.J. Poirier
The Mineral•, Metal• & Material• Society, 1992 30
 20 L
""o
1

X
o
E 15
<,
e
u
~
~ 10
<t
I
1-
z
w 5

O'--~..c--'-~~~_,._~~~-'--~~--'~~~-'
o 500 1000 1500 2000 2500
TEMPERATURE,K

Figure 10: Toe enthalpy of iron.

the product of the mass and latent heat of fusion of the metal, then the iron in tbe ladle
will be completely solidified.

Toe heat to be removed from the metal is

(59)

where

(¿M = heat removed from metal, Btu

m = mass of metal, lbm

H1 = latent heat of fusion, Btu/lbm.

Toe heat loss from the top surface is

,2r = QTt (60)

31
where

Ór = heat loss from top, Btu

QT = rate of heat loss from top, Btu/h

t = time, h.

QT is by convection and radiation, so that

(61)

where

Ar = area of the top surface of metal, ft2

TM = freezing point of the metal, ºF

T00 = temperature of the surroundings.

To QT we must also add the heat lost through the side wall of the ladle. To sirnplify our
thought process, however, let's assume that the side wall of the ladle is a perfect insulator
so that Eq. (61) accounts for ali of the heat lost. Then, we merely make QM = Ór, and
substitute Eq. (61). Then we solve for the time required to solidify ali of the iron. The
result is

t = (62)

Now consider a modification in the above scenario. The ladle is filled with
superheated liquid iron, ata ternperature T, (i.e., the pouring temperature). In this case
the heat to be removed from the metal comprises sensible heat and latent heat. Then
(63)

where c,.L is the specific heat of the liquid metal. Any specific heat is defined as

e,= -dH
dT

and is approximately constant over a range of temperature. For example, the slope of
the enthalpy curve for liquid iron in Fig. 10 is constant so that c,.L = 0.189 Btu/lbm·ºF.

Again there is heat lost by convection and radiation from the top of the metal, but
its temperature decreases from T, to TM as the liquid metal cools and then remains fixed
at TM as the liquid solidifies. For the first period t I

(64)

32
where f = l/2(Tp + TM) and the heat transfer coefficients are evaluated at f. During
this period the sensible heat is removed from the metal, so we can write
(65)

or
m Cp,i(Tp - TM)
(66)
(h + h,)1Ar(f - r .. )
For the second period, we use Eq. (62) with t2 substituted for t. Thus
mH1
'2. =
(h + h,)Ar(TM - T 00)
(67)

and the total time is

(68)

Example JO

A ladle with 500 lb., of liquid iron is initially at 2950°F. Estímate (i) the
time required for the liquid iron to lose all of its superheat (i.e., sensible heat)
and (ii) the total time for it to solidify. The exposed area is 1.25 ft2• Assume that
the ladle wall is a perfect insulator.

For the first period, f = 2874 ºF and we gather the properties for air at
T1= 1/2(2874 + 80) = 1477ºF (1076 K) from Table 4:

p = 0.0205 lbm/ft3

µ = 0.106 lbm/ft·h

k = 0.0415 Btu/h·ft· ºF

Pr = 0.704

Also

/3 = 1/T¡ = 1/(1477 + 460) = 5.16 X 10-• ºF1

L ::: diameter = 1.26 ft

g = 4.173 X 108 ft/h2

33
Then

Gr = (4.173 X 108) (5.16 X 10-4) (0.0205)2 (2874-80)(1.26)3

(0.106)2

= 4.50 X 107

Gr·Pr = 3.17 X 107

Table 7 gives C = 0.15 and m = 1/3.

Nu = hL = 0.15(3.17 X 107¡i'3 = 47.47

k

h = (47.47)(0.04l5) = 1.56 Btu/h·ft2·ºF

(1.26)

The radiation heat transfer coefficient is evaluated with e = 0.28, T1 =

(2874 + 460)ºR and T2 = (80 + 460)ºR. Then

h, = (0.28) (0.1714 X 10-8) (33342 + 54a2) (3334 + 540)

2•
= 21.21 Btu/h·ft ºF

Now Eq. (66) can be used; with CP,L = 0.189 Btu/lbm and TM = 2797°F:

(500)(0.189)(2950-2797)
(1.56 + 21.21)(1.25)(2874-80)

= 0.182 h = 10.9 min

During the solidification period, Eq. (67) applies, but now the temperature
is slightly less than f for the first period. The reduced temperature slightly
changes the properties of air so that the convection heat transfer coefficient
becomes h = 1.57 Btu/h·ft2• ºF. (To save space, this calculation is omitted.) The
radiation heat transfer coefficient is evaluated with T1 = (2797 + 460) ºR and
T2 = 540°R; the result is h, = 19.86 Btu/h·ft2• ºF. Now Eq. (67) is applied with
H¡ = 117 Btu/lbm.
t _ (500)(117)
2
- (1.57 + 19.86)(1.25)(2797-80)

= 0.804 h = 48.2 min

The total time is t = t 1 + t2 = 59.1 min. By far most of the time is taken up
while the iron solidifies.

34
 It is instructive to plot temperature versus time when initially the liquid is at the
freezing point and when the liquid has superheat. These cases are shown in Figs. l la
and b, respectively. In Fig. l la, the temperature does not change while the metal
solidifies. Only when solidification is complete does the temperature decrease. In
Fig. llb, temperature decreases as the liquid metal loses sensible heat, remaíns constant
as latent heat is evolved during solidification, and then decreases as the solid cools. In
Chapter IV, we wíll apply these principies to the solidification of metal castings.

Time Time
(o) (b)

Figure 11: Solidification of apure metal: (a) with no superheat

and (b) with superheat.

B. Solidificasion of Alloys

Casting alloys solidify over a range of tempera tu re so that the latent heat of fusion
is not released at one temperature. Figure 12a illustrates the fraction solidified vs.
temperature for an alloy that solidifies to a single phase. The enthalpy of the alloy
during solidification, H, is

where

H,,HL = the enthalpy of the solid and liquid, respectively

and

f,,ÍL = the weight fraction of solid and liquid, respectively.

The enthalpy can also be written as

H = H¿ - f,Hf' (69)

For the alloy shown in Fig. 12a, enthalpy vs. temperature is shown in Fig. 12b.

35
 Figure 12c is drawn for an alloy that solidifies as a single phase with an eutectic
constituent at the end of solidification. Toe eutectic solidifies ata fixed temperature (or
overa small range). This accounts for the step equal to fEH in Fig. 12d, where fE is the
fraction of the eutectic constituent. The added complex.ityoEthe evolution of latent heat
over a range of temperatures usually requires numerical computations for estimating
temperatures and heat transfer during solidification.

o fs
(a)

TL

T H

....¡ f E
o fs T
(e) (d)
Figure 12: Fraction solid and enthalpy of alloys during solidification. (a) and (b) for an
alloy that solidifies as one phase; (e) and (d) for an alloy that solidifies with
an eutectic at the end of solidification.

36
III. HEAT TRANSFER IN MOLDS

A. Conduction in Mold Materials

Mold materials can be classified as insulators or as good conductors. The former

include plaster full molds, ceramic shell molds in investment castings, and silica sand
molds in sand castings. Permanent mold casting and die castings use metallic molds that
are good conductors.

The conduction heat flux in a mold is expressed with the usual rate equation:

q = -k éJT (70)
ax
The thermal conductivity of the mold, k, is an effectivethermal conductivity, as contrasted
to the intrin.sic thermal conductivity of the individual particles that make up the mold.
On a microscopic leve!, the mold is porous so heat is actually transferred by intrinsic
conduction through each particle and conduction and convection in the gas in the pores.
In addition, there is radiation from particle to particle by radiation across the pores.
Therefore, the thermal conductivity, used in Eq. (70), is an effective conductivity because
it depends on severa! factors: particle material, particle size, binder, volume fraction of
pores, gas in the pores, emissivity of the particles, and temperature.

The effect of temperature on the thermal conductivity in insulating molds is shown

in Fig. 13. The curves labeled with AFS sand numbers are for sand molds. In the sand
molds, thermal conductivity increases with increasing temperature, because the fraction
of energy transferred by radiation from particle to particle increases.

B. Heat Tran.sfer across Casting Gaps

As an approximation, the temperature at the surface of an insulating mold

adjacent to solidifying metal is equal to the freezing point of the metal. Because the
thermal conductivity of the mold is only - 0.01 to 0.02 that of the metal, then practically
ali of the thermal resistance to the heat transfer is within the mold, itself. The
temperature distribution in this situation is illustrated in Fig. 14a.

Early in the solidification process the resistance to heat transfer across the gap
that forms at the mold-casting interface can also influence the overall heat transfer
(Fig. 14b). If we knew precisely the characteristics of the gap, including its thickness
variation with time, composition of the gas within the gap and the emissivíties of the
mold and casting surfaces, then the heat transfer across the gap could be estimated.
However, such attempts usually fail so heat transfer associated with a gap is usually
treated empirically with a gap heat tran.sfer coefficient (h8), which is defined by the flux
across the gap. The flux across the gap is
(71)

Heat Transfer Fundamental• for Metal Casting

D.R. Poirier and E.J. Poirier
The Minerals, Metals & Material& Society, 1992 37
 IJ...
o
~o.a~-~---~---~-~
s:

-
<,
:::,

CD 0.6
>="
1-
>
1-
u 0.4
~
a
z
o
UQ.2
.....J
<t
~
a:
W O.O....__..___....__...__..._ _ _..__..
I O 800 1600 2400
1-
TEMPERATURE, ºF

Figure 13: Thermal conductivity of ínsulating mold

materials. Curves for various AFS numbers
refer to sand molds. Toe cross hatched area is
for ceramic shell molds with various binders.

t 1
l..
1
w
a:: 1 T
....
1
1 1
:::> MOLO 1 <l 1
~ 1 1
o:: souo]- 1
w 1
a.. I LIOUID 1
~ 1
w 1
1
f- 1
(o) (b)

Figure 14: Temperature distributions when casting in an insulating

mold: (a) with no thermal resistance at the casting-mold
interface; (b) with thermal resistance at the casting mold
interface.

38
where Te and T M are the temperatures at the casting surface and mold surface,
respectively. Because the characteristics of the gap are not that well know, actual values
of h8 can only be determined empirically.

Castings freeze relatively quickly in metal molds (i.e., conducting molds), and
temperature changes drastically in both the mold and the casting. An understanding of
the factors affecting solidification is important because permanent mold castings and die
castings are made in metal molds. Also many castings are made in insulating molds with
metal chills inserted at strategic positions to increase solidification rate.

Toe analysis of heat transfer when metal is poured against a chill is more
complicated than when metal is poured against an insulating mold. Toe added
complexities are illustrated in Fig. 15. At the metal-mold interface, the thermal
resistance of the gap ( or mold casting) causes a rather large temperature drop. Toe
condition of no thermal resistance at this interface would exist only if the casting solders
or welds to the rnold, which obviously is not desirable.

1
CHILL 1
1
MOLO 1
1
1
1
l
1
SOLIDI-
: LIQUID

Figure 15: Temperature distribution when casting in

a conducting mold with thermal resistance
at the casting-mold interface.

In addition to the rather Iarge temperature drop across the gap, there are other
differences between solidification in insulating and chill molds:

(a) Toe thermal resistance of the metal being cast forms an important portian of
the overall resistance to heat transfer.

(b) More total heat is removed during solidification because the solidified metal
is appreciably below the melting point.

Toe value of AT across the gap depends upon the thermal properties of both the
mold and the casting and the gap heat transfer coefficient. Gap heat transfer coefficients
are given in Table 9.

39
 Table 9

Gap Heat-Transfer Coefficients

h,
Casting Situation Btu/h·ft2• ºF

Ductile iron in cast iron mold coated with amorphous carbon 300

Steel in cast iron mold 180

Aluminum alloy in small copper mold 300-450

Steel chilled by steel mold

before gap forms • 70-180
after gap forms 70

Aluminum die castings

before gap forrns" 440-880
after gap forms 70

"With no gap, thermal contact between the casting and the mold is not perfect
because of surface tension effects, oxide layers and mold coatings.

40
 IV. SOLIDIFICATION HEAT TRANSFER

Toe solidification rate of alloys is an important processing variable. For example,

solidification rate relates directly to the coarseness (or fineness) of dendritic structures
and hence controls the spacing and distribution of microheterogeneities, such as dendritic
microsegregation, second phases, inclusions and microporosity. Although most types of
macrosegregation are not directly related to solidification rate, it is known that Iarger
castings (i.e., castings with relatively slower solidification rates) are more prone to
macrosegregation than are smaller castings. For these metallurgical reasons and from
a manufacturing viewpoint (e.g., production rate and riser design), metal casting engineers
should recognize solidification heat transfer as an important tapie.

A. Solidification in Thick Molds

Consider pure liquid metal with no superheat poured against a flat surface of an
insulating mold (Fig. 14a). We make the following assumptions:

(1) no thermal resistance at the casting-rnold interface;

(2) because ali of the resistance to heat transfer is almost entirely within the
mold, there is no temperature gradient in the solidified portion of the casting;

(3) the mold is serni-infinite in extent and its thermal properties are constant and
uniform.

By assumptions 1 and 2, the temperature at the surface of the mold (x = O) is

equal to the freezing point of the metal (TM)· Toen with assumption 3, Eq. (31) applies
so the rate of heat flow into the mold is

Q = ~ .¡:; JkpCp (T M - T)r

I
112 (72)

where A is the area of the mold-castíng interface and T; is the initial temperature of the
mold. Toe product kpCP. represents the ability of the mold to absorb heat at a certain
rate. It is called heat diffusivity and is not to be confused with the thermal diffusivity
(o = k/pCp).

With p" as the density of the solid metal and M as the thickness solidified,
pAdM /dt represents the rate of increase of salid (Btu/Ibm). If it is multiplied by the
latent heat we get the heat extracted from the casting:

(73)

Toen Eqs. (72) and (73) are set equal; this gives

dM 1 (74)
JkpCP r112
dt .¡:;
Heat Transfer Fundamentals forMetal Casting
D.R. Poirier and E.J. Poirier
The Minerals, Metal• & Material• Sociely, 1992 41
 l
Integration of Eq. (74) follows with M = O at t = O; the result is

M = 2_ [ TM, - T¡ Jkp CP t 111. (75)

.¡; P Hf

We see that the amount of solidification depends on the properties of the metal (the
quantities in the parentheses) and the mold's heat diffusivity. Sorne mold and metal data
are given in Table 10.

Table 10

Properties for Insulating Molds and Solidifying Metals

k p cp JkpCP
Mold Material" Btu/ft·h· °F lbm/ft3 Btu/lbm °F Btu/°F·ft2·h112

Silica Sand 0.3 100 0.28 2.9

Mullite 0.22 100 0.18 2.0
Plaster 0.20 70 0.20 1.7
Zircon Sand 0.60 170 0.20 4.5

TM H, p e·p o
k'
Casting Material ºF Btu/lbm lbm/ft3 Btu/lbm F Btu/h·ft· °F

Iron 2802 117 450 0.18 23

Nickel 2651 125 490 0.16 20
Aluminum 1220 168 150 0.25 150

*Only typical values can be given here. Actual properties depend on temperature,
particle size, binders, etc.

Example 12

Estímate the solidification time for iron (0.5 in. thick) cast in a zircon sand
mold.

Toe metal solidifies from both faces of the mold. Thus,

M = 0.25 in. = 0.0208 ft.

42
 Solving Eq. (75) for t:

r = ..!. M2
4
P 'H1
[ TM - T¡
1 2 (k e )-1
P P

t = [ ..!. ] (0.02082) [ 450

4
X 117
2802 - 80
12 (4.52r1
t = 0.0063 h = 23 s

Now we consider the effect of thermal resistance at the casting-mold interface with
an approximate analysis. Assume that the temperature in the solidified metal is linear.
Toen the therrnal resistance in the solid metal is M [Ak. By making the analogy between
Ohm's law and Eq. (72), the thermal resistance in the mold is

R r -- .¡;¡- (76)
A JkpCP

With the thermal resistance of the gap taken as 1/Ah8, we get the total thermal
resistance

Rr= -
M + _
1 +
.¡;¡- (77)
Ak' Ahg AJkpCP

Toen Eq. (72) can be replaced with

or

Q
(78)

To keep the problem tractable, we ignore the sensible heat extracted from the
solid and make Eqs. (73) and (78) equal. Further we recognize that 1/hg » M/k' and
neglect M/k'. Toe result is

l
-1
dM .¡;¡- (79)
dt
JkpCP

43
Equation (79) is integrated with M = O at t = O; the result is

The term in brackets can be looked upon as a factor, <f,(t), which accounts for
resistance to heat transfer because of the gap. Figure 16 is a plot of </, vs. time for plaster
and zircon sand molds (Table 10) with h, = 100 and 500 Btu/h·ft2• ºF. For small times
(i.e., thin section castings), the gap retards solidification significantly. The retarding effect
is more in zircon sand molds than in plaster molds because the heat diffusivity of zircon
sand is greater than that of plaster.

1.0 -----------~-----

---
500 .... ---------

--
---
0.8
-- ------
/
/
0.6 /
/ --- Plaster
I
/ ---- Zircon sand
I

0.4------- ..........-----~----~
o 100 200 300
TIME, s
Figure 16: ¡/) for various values of the gap heat transfer coefficient in insulating molds.

Example 13

Repeat the problem of Examp/e 12 and assume that h1 = 500 Btu/h·ft2°F.

By our previous calculation, an estímate of the solidification time is 23 s. For

23 r, Fig. 16 shows 0.82. Inspection of Example 12 shows that <f,2t = 23 s.
é ""

23
t = = 34 s
(0.82)2

B-r,,/ ~ -/( re
44

I
; V e:
::::>
e 7_
b
"J._._\'. i
-.l (
Ce,,//. ,. -')
/ »t> ~
 B. Modulus Method

For metal castings made in insulating molds, the modulus is defined as the volume
to surface area ratio, V/A. Toe volume V can be the entire volume of the casting or
sorne portion of the casting, and A is the corresponding casting-mold interfacial area.
For infinite plates, V/A = M where Mis the semithickness. Toen we can write Eq. (80)
as

(81)

where t is the solidification time of the casting and an appropriate value of </> is used.

r
This relationship is solved for t with result:

(82)
t, = e (:
where

e ~
44,2 [ T: \ r [ ,.'e, ] ·
Equation (82) is often referred to as Chvorinov's role, and C, as Chvorinov's
constan t. It is in tended for comparison of freezing times of castings with different shapes
and sizes. The relationship works best for casting geometries in which none of the mold
material becomes saturated with heat, such as in interna) comers or interna) cores. Its
success hinges on the mold material absorbing the same amount of heat for every unit
area exposed to the casting. This is strictly true only for a group of castings, which have
similar geometries but different sizes.

To quantify effects of various geometries of castings, !et us examine differences

arnong three basic shapes: namely infinite plates, infinite circular cyíinders, and spheres.
First we define two dimensionless parameters, f3 and -y:
(83)

and

(84)

Toen the freezing times for the three basic shapes are given by

{J=-y[2.+.2..]
.¡; a/3
(85)

45
where a = oo for infinite plates, a = 4 for infinite circular cylinders and a = 3 for
spheres.

We can deduce Eq. (85) from Eq. (81) for infinite plates, with a = ee , Likewise
Eq. (85) can be deduced from available equations for cylinders and spheres, although
those equations are not given here.

Example 14

Compare the freezing time of an iron sphere, 2 in. diameter, to a plate that
is 2 in. thick. Both are cast in a zircon sand mold. Assume that </¡ = 0.9.

First Eq. (84) is used to estímate 'Y with thermal properties from Table 10.

'Y
= 0.9 [
º
28 2
-
80
450 X 117
) (170 X 0.20) = 1.582.

Toen Eq. (85) is used to get f3 for both the plate and the sphere.

{3P = {3(plate) = 1.582 [ ~ ]

f3, = f3(sphere) = 1.582 [ 2.

.¡; + .i, ] ·
3{3,

Solving:
1.78 and {3, = 2.04.

For the plate,

V = 1 in. 0.0833 ft
A

0·60
a = = 0.0176 ft 2 /h
(170)(0.20)

(V/A)2 0.08332
t(plate)
{l,,a (1. 782)(0.0176)

t(plate) = 0.124 h = 7.4 min.

46
 For the sphere (d = dia.),
V d
0.0278 ft
A 6

0.02782
t(sphere)
(2.042)(0.0176)

t(sphere) = 0.0106 h = 0.6 min.

Toe above example clearly illustrates that castings of the same thickness do not solidify
in the same time. Shapes that favor divergent heat flux Iines in the mold solidify in less
time than shapes made in molds with unidirectional heat flux lines or convergent heat flux
lines. Toe modulus should be used as the characteristic dimension of the casting when
solidification times are estimated.

Example 15

From Example 14 the solidification time of a 2 in. dia. sphere is 0.0106 h.

Estimate the freezing time of a 3.25 in. cube.

Using subscripts s and e for the sphere and cube, respectively, Chvorinov's
rule gives

te (V/A)~
t, (V/A);

For the sphere:

V ] = f!_ = 0.333 in.

[ A s 6

For the cube, with t as the cube edge length:

[~ ] = !.. = 0.541 in.

A e 6

Toen

te = ( º·0.333
541 ] 2 (0.0106) = 0.0280 h = 1.7 min.

(Notice that we have assumed a = 3, in Eq. (85), for the cube.)

47
 We can approximate the effect of superheat on solidification time by realizing that,
in addition to absorbing latent heat, the mold must also absorb superheat. Toe total
quantity of heat to be removed from the casting is
(86)

Toe subscript t denotes Iiquid phase properties, and tl.T, is the amount of superheat. In
order to make the analysis simple, yet sufficiently accurate, we assume that the interface
temperature of the mold is constant while the liquid loses its superheat. In addition,
p; :::: p' then

(87)

where Hj is the effectiveheat of fusion. That is

H; = H1 + CP,,tlT,

and Hj replaces H1in ali of the previous equations that are used to estimate solidification
time. Note that solidification time is still proportional to (V/A)2.

C. Solidiftcation in Molds of Finite Thickness

In this section we consider cases in which the mold cannot be assumed to be

infinitely thick. Figure 17 illustrates a situation that often leads to a hot spot in a casting.
Part of the mold or a core is surrounded by the solidifying metal. If this part of the mold
is thin enough, it can become saturated with heat and incapable of extracting further
energy from the solidifying metal.

Metal

t =O
1-----iT¡

Mold

Figure 17: Temperature distribution in a

mold or core surrounded by the
solidifying metal.

48
 Heat conduction in the mold must satisfy Eq. (10) and

initial temperature: T(x,O) = T1 (uniform)

surface (x = L): T(L,t) = TM

centerline (x = O): iJT (O,t) = O.

ax
Toe particular solution that satisfies these conditions can be found in many books on
conduction heat transfer. Because the mold has a finite thickness, the solution is in the
form of an infinite series, which is not repeated here. lnstead we write the equation for
the heat absorbed by the mold, which can be derived from the infinite series for the
temperature distribution. Toe equation is

C L (T
QM/A - T.) =
2 ~
L..J
11-2 [ 1
n -
( tf F )]
exp - n o
(89)
P p M I n•O

where

Fo

If A is taken to be the area of one face of the mold portion, then QM is the energy
absorbed by one half of the mold. If A represents the area of both faces, then QM is the
energy absorbed by the whole mold portion. Toe Fourier nurnber, Fo, is dimensionless
time and appears in conduction problems.

As usual the energy extracted from the solidified casting is p'Vl{_¡, and it is
substituted for QM. After rearranging, Eq. (89) is written in dimensionless form as

V/A
L
2-y
n=O
E 11: 2 [ 1 - exp { -{3~Fo) ] (90)

where 'Y is defined by Eq. (84).

Equation (90) is plotted in Fig. 18 for typical values of 'Y· Each curve in Fig. 18
rises to a plateau when the mold is saturated with heat, and there is no additional
solidification.

49
 1.3
1.2 r= 13

0.9
~O.B r=0.9
<{
<,
>
0.5
0.4 y=0.5

2 3
Fo= at/L2

Figure 18: Solidification against a mold or core surrounded by metal.

Example 16

Estimate the maximum thickness of aluminum that is solidified from both

sides of a plaster mold that is 3/8 in. thick. Toe mold is preheated to 400ºF.

First we calculate -y with properties from Table 10 (assume that ti> = 1).
1220 400
= (70)(0.20) { - ) = 0.455
'Y 150 X 168

Toe limiting values of (V/A)/L are equal to -y (see Fig. 18); this corresponds to
the situation when the mold is completely saturated with heat and t = oo in
Fig. 17. Toen
(V/A)
= 0.455.
L

Finally, with L = (1/2)(3/8 in.) = 3/16 in., the thickness solidified is

M = V/A = (0.455)(3/16) = 0.0853 in.

This simple calculation shows that a mold or core surrounded by metal can
easily become saturated with heat.

50
 Now we consider solidification in a shell mold with a thickness L and heat loss
from the outside surface to the surroundings. Again Eq. (10) applies. Toe initial
condition and the boundary conditions are
initial temperature: T(x,O) = T¡ (uniform)

mold-metal surface (x = O): T(O,t) = TM

mold-surroundings surface (x = L): aT (L,t) + !!.. [ T(L,t) - T ,,.] O.

ax k

Toe last condition is derived by equating the conduction heat flux in the mold to the flux
of heat lost to the surroundings.

Toe method of solution to this particular set of conditions is rather lengthy so only

¡
the final result is given. With dimensionless groups, it is

V/~
L'Y
= { BiFo.}
1 + B1
l + 2
Fo
t
n-i
s,
-
wn Bi 2
(Bi2
2
+

+ wn +
w,.2)
Bi
(
exp -wnFo
2 )} (91)

where the wns are the positive roots of

(92)

the Bns are given by

and

Examination of Eqs. (91)-(93) shows that there are four independent dimensionless
groups:

(1) (V/A)/L'Y'

(2) Bi (Biot nurnber) = hL/k

(3) Fo (Fourier nurnber) = at/L2

51
 Toe first and third dimensionless numbers give the modulus of the casting versus
solidification time. Properties of the metal and mold are evident in the first, second and
third dimensionless numbers. Finally the effects of the mold thickness, mold preheat, and
the heat transfer coefficient for the heat loss to the surroundings are also included in the
entire set.

Results for solidification in molds with preheat and with no preheat are shown in
Fig. 19. Except for relatívely short times, the results depend mainly on the Biot number
and are almost independent of the preheat. This can be seen in Fig. 19a. Toe
differences between preheat and no preheat are more apparent for short times (Figs. 19b
and 19c).

Provided Fo e? 1, the results can be represented as

V/A
L
= -,· [ ~
1 + B1
J Fo+ b (94)

12 T M-T¡ = 0.25
TM-Ta,
Nopreheat

¡.
1 4 8 12 16 20
Fo=at/L2
(o)

uo.0.4
Q..
_J

0.4 0.8 1.2 0.4 0.8 1 2 1.6

Fo= al /L2 Fo=al/L2
(b) (el
Figure 19: Solidification in molds of finite thickness; (a) long times; (b) short times with
preheat; (e) short times with no preheat.

52
where b is an intercept. In the linear part of the curves, the temperature distribution in
the mold is very close to steady state as the casting solidifies. With no preheat, more
time is required to achieve steady state; this can be seen by examination of Figs. 19b and
19c. The value of the intercept depends on Bi and the preheat; in ali cases b s 0.25.

Example 17

Estímate the solidification time for nickel cast with lOOºF superheat in a
ceramic shell mold (0.3 in. thick) preheated to 1800ºF. Assume that the
properties of mullite in Table 10 are appropriate for the shell mold, and that
h = 18 Btu/h·ft2·ºF. The modulus of the casting is 1 inch.

Assume that the effective laten! heat of fusion can be used to account for
both the latent heat and the superheat. Then HÍ = 125 + 0.16(100) =
141 Btu/lbm. Equations (91)-(93) are best solved with a computer code because
Eq. (91) contains a series and Eq. (92) cannot be solved explicitly, Evaluation of
the dimensionless groups is as follows:

Bi = hL ( 18)(0.3/ 12)
2.04
k (0.22)

TM - T; 2651 - 1800
0.331
TM - T,,. 2651 - 80

'Y = pCP [T.-T.l

..
pHJ
(100)(0.18) [ 2651 - 80
(490)(141)
J 6.70

(V/A) (1/12)
0.498.
L"f' (0.3/12)(6.70)

kt 0.22 t
Fo 19.56 t (t in h)
pCp L2 (100)(0.18)(0.3/12)2

The calculated results are listed as follows:

Fo (V/A)/Ly'
0.326 0.0215
1.956 0.1304
3.585 0.2402
5.215 0.3500
6.844 0.4591
8.474 0.5685
10.104 0.6779

53
 Interpolation between Fo = 6.844 and 8.474 gives Fo = 7.424.

7.424
:. t = = 0.3795 h = 22.8 min.
19.56

In Examp/e 17, b ~ O so that Eq. (94) could be approximated as the same

result obtained if the temperature in the mold had been assumed to be at steady state.
This result is
V (95)
Ct
A

where the constant C is defined as

(96)

Now suppose we consider solidification in a cylindrical shell mold, with the

mold in the domain R1 S r S R2• At the casting mold interface (r = R1), the
temperature is TM• and at the outer surface of the mold (, = R2) there is heat loss to
the surroundings at T... It can be shown that the steady state heat lost through the
inside area of the mold (A) in time t is

Q = Ah(TM - T .. )t (97)
m (R1/R2) + (hR1/k) In (R2/Ri)
As usual
(98)

so that with QM = Óc and t = t, (solidification time), we get

V = Ct (99)
A

where

and

(100)

54
 Example 18

Compare the values of C for the solidification of a plate and of a cylinder

in shell molds of the same thickness. Use the same conditions of Example 17.

For the plate, we substitute values into Eq. (96), with L = 0.025 ft.
Bi = hL = (18)(0.025) = 2_045
k 0.22

e = [ 0.22 ] [ 2651 - 80 ] [ 2.045 ] = 0_220

0.025 490 X 141 3.045

For the cylinder, we assume the same modulus. Toen

V __!_ ft = R¡
A 12 2

R1 = 0.1667 ft

and for same thickness of mold, R2 - R1 = 0.025 ft, Now substituting into
Eq. (100) we get

e = [ 0.22 ]
0.025
[ 2651 - 80 ]
490 X 141
[
0.8696 +
2.045
(0.1667)(ln l.150)(2.045)/(0.ü25)
J
e= 0.241.

For the same modulus, the cylinder solidifies about 10% quicker than does the
plate.

55
V. RISER DESIGN

A Riser Size

We have seen that the solidification time of a simple shape cast in a thick mold
of an insulating material can be estimated with Chvorinov's rule, which is

t = C(V/A)2 (82)

where

t = solidification time of the casting,

V = volume of the casting,

A = surface area of the casting,

C = Chvorinov's constant.

Let's first consider the case when Chvorinov's constants of the casting and riser
are equal. Toe most efficient riser from freezing time point of view is the one in which
solidification ceases simultaneously with the solidification in the casting. Because the
freezing times are equal for the riser and the casting, Chvorinov's rule requires that

(101)

where

VRJ = final volume of the riser (after complete solidification),

Ve = volume of the casting,

AR = surface area of the riser, and

Ac = surface area of the casting.

Conceptually, volumes are proportional to the amount of heat that must be

removed for complete solidification, and areas are proportional to the amount of heat
that can be absorbed by the mold or extracted through the mold. Notice that the final
volume of the riser is used because, this volume (and not its initial volume) is
proportional to the heat extracted from the riser through area AR.

A material balance on a riser-casting system, with all the shrinkages resulting in

the riser, is
(102)

Heat Tl"an1ferFundamental•for Metal Casting

D.R. Poirier and E.J. Poirier
The Mineral•,Metal• & Material•Society, 1992 56
in which

VR = initial volume of the riser, and

f = fraction of feed metal requirement.

The quantity f(VR + Ve) is simply the total feed metal required to feed both the casting
(Ve) and the riser (VR).

By combining Eqs. (101) and (102), the basic riser equation is obtained; it is

(103)

Toe feed metal requirernent is a result of metal shrinkage and mold dilation, and
as such f can be process dependent, depending upon the degree of mold dilation. If
necessary, it can be measured empirically by titrating water into the shrinkage cavities of
risers during the development stage of a production run.

Recalling that the solidification times of simple shapes can be represented as

{3 = 'Y [-2...¡; + ___!_]

a/3
(85)

we can replace Eq. (101) with

(104)

(V/A)c is the volume-to-surface area ratio of the casting or the portion thereof fed by the
riser, and subscripts R and e refer to the properties associated with the riser and the
casting, respectively. Toen by combining Eqs. (102) and (104), we get

(105)

This equation is more general than Eq. (103) because it allows us to consider cases when
the casting and riser have different geometries and/or different molding materials
surrounding them. However, its use presumes that the areas of each are uniformly
molded with its particular mold material.

57
Examp/e 19

An aluminum plate casting, 1 in. X 5 in. X 5 in., is fed with an end riser;
both are entirely encased in silica sand. Calculate the height and diameter of the
riser.

1 in.

Neglect the contact area between the casting and the riser; this adds sorne
conservatism to our analysis. Toe edges and comers of the plate will freezc more
quickly than the plate as a whole, but the riser must rcmain molten long enough
so that it does not freeze before the entire casting. Therefore, (V/A)c should be
selected as that of an infinite plate. Toen

(:L = 1h inch (i.e., semithickness)

Toe shrinkage of pure aluminum is 0.06; to be safe we add a fraction for mold
dilation and assume that f = 0.075. Toe volume of the casting is Ve = 25 in.3

Now we must estimate the values of the (js in Eq. (105). First consider the
plate (Table 10):

'Y = (100)(0.28) [ 1;i2~-~8 ] 1.27

Then (with a = oc for plates), Eq. (85) gives

(je = (2)(1.27) = 1.43.

.¡;
Now consider the riser. It is certainly notan infinitely long riser, so we can expect
that 3 :s; a :s; 4. For our estimatc, let's assume that a = 3.5. Toen

(3R = 1.27 [.2...¡; + _l_]'

3.5 (3R

from which, we find (jR = 1.65. Now with ªR = ªe• we have

(3R.¡;; 1.65 = 1.16.
1.43
(je.¡;;:

58
Let's summarize Eq. (105), by inserting known values up to this point:

0.925 VR - (0.075)(25) = (1.16)(1/i)AR.

For the cylindrical riser (D = dia. and H = hgt.),

·•D2H
VR = -4-

and
AR = A(top) • A(bottom) + A(side)

= 2 ...o 2
[ -4- l + 1íDH.

We are left with two unknown dimensions, D and H. The minimum diameter of
the riser corresponds to a riser that is infinitely tall. Of course, this is not a
practica) solution, but it is interesting to determine the minimum diameter. As
H-+ =. fVc-+ 0 and (VR/AR)-+ D/4. Then

Dmln = 4 [ ~) (V2) = 2.51 in.

0.925

To obtain a practica) result, let's assume that H = D. Then

1( DJ,
VR = -
4

and

(0.925) [ -¡) D3 - (1.16) [

3;
) D 2 = (0.075)(25)

We have a choice: either solve the cubic equation directly or solve it by tria! and
error. Toe latter is easier, particularly because we already know Dm1n· Toe
solution is D = H = 3.93 in.

This might seem like a rather large riser, but we should keep in mind that
a pure metal will, in fact, exhibit piping. Alloys that freeze in a mushy manner,
do not pipe very much and the shrinkage takes the form of microporosity
dispersed in both the riser and the casting. Hence, less feed metal is required.

59
 Example 20

Repeat Example 19 for a mushy freezing aluminum alloy (f = 0.04) and

selecta riser with H/D = l.

By following the solution to Example 19, but usingf = 0.04, we get

0.960 VR - (0.04)(25) = (1.16)(1/2)AR

Now we substitute H = D, so that

-,;DJ 3-,;D;
VR = and = -2-·
4

Toen
34-,;
(0.960) [ ~ ) D 3 - (1.16) [ ) D2 (0.04)(25).

D =H = 3.72 in.

When an insulating sleeve is used around the riser, the riser is in contact with
different materials on its sides, top and bottom. A top riser can be in contact with
insulating material on its sides, with insulating, exothermic or sand material on its top and
with the casting on its bottom. A side riser on the other hand, can be in contact with
similar materials on its sides and top, but it can also be in contact with sand or insulating
material on its bottom. To account for these various possibilities, therefore, we split the
riser into three components for its side, top and bottom, then
(106)

Toe subscripts represent the top (t}, bottom (b ), and side (s) of the riser. By combining
Eqs. (105) and (106), we arrive at

(107)

where

e = fJ, ¡;;: the apparentsurface alterationfactor(asa) for the sides of the riser,
PcFc
/3 ¡;;:
-1
-1- the asa factor for top of the riser,
f1cFc
60
and

z2 = f3b .¡;;:- the asa factor for bottom of the riser.

f3c¡;;:

Let us define x = H/D for the riser, with H its height and D its diameter. Now,

VR = ~4 D3x

(108a,b,c)

and

A I = A b = ~4 D2

Substituting Eqs. (108a,b,c) into Eq, (107) we arrive at the riser design equation:

(109)

with the constants K1 and K2 defined as

(V/A)c
K, = (1 - f)

4/~
~(1 - !)

Example 21

An aluminum bar casting, 2 in. x 2 in. X 5 in., is fed with a top riser that
is encased in an insulating sleeve with a top. Assume that f = 0.04 and H / D = 2.
The sleeve has the following properties: k = 0.07 Btu/h ft ºF; p = 15 lbm/ft3;
CP = 0.26 Btu/lbm ºF.

[: L = 0.5 in. neglecting the ends

-y ( casting) 1.27 (from Example 19)

61
As a long square bar, the casting is approximated as a cylinder. Therefore, a = 4

l;
and

{3 e = 1.27 [ 2_ _l
.¡; +
4{3<

solving f3c = 1.63.

Now we deal with the riser (side and top are encased in the insulator).
1220-80
"Y(riser) = (15)(0.26) [ ) = 0.176
150 X 168

l
With a :::: 3.5,

{3R = 0.176 [
.¡; + .z.,
2_
3.5 {3R

and solving {3R = 0.342.

Toe thermal diffusivities for the materials in contact with the casting and
the riser are:

ae = (0.3) = 0.01707 ft 2/h

( 100) (0.28)

and

(0.07) = 0.0179 ft2/h

(15)(0.26)

Toe asa factors are

e = z1 = 0.342 .[ofi9 = 0.272.

1.63 {0107

Toe contact arca between the riser and the casting (i.e., the bottom of the riser)
is assumed to be an adiabatic surface; i.e., there is no heat flow by conduction
from the riser to the casting. Thus, z2 = O.

Toe parameters K1 and K2 are determined next.

(0.5)
= 0.521 in.
(1-0.04)

Ki = (4)(0.04)(20) 1.061 in.3

T(l-0.04)

62
 With x = 2, Eq. (109) becomes

D3 - 0.521[(4)(0.272)(2) + 0.272 + O]D2 - 1.061 O.

Solving D = 1.66 in. and H = 3.32 in.

Now we consider casting-riser designs in ceramic shell molds. If we start by

selecting a riser that freezes an instant after the casting, then

(110)

We can follow the steps leading to Eé¡. (109), but now the asa factors are defined as
e = C./Cc for the side, z1 = C,/Cc for the top, and z2 = Cb/Cc for the bottom of the
riser.

Radiation and convection heat losses from the tops of open risers should also be
considered in their design. The heat loss from the top of an open riser in solidification
time t is

Q, = (h + h,)A,(TM - T )t, 00
{111)

where the subscript t represents the top of the riser. The asa factor is the ratio of the
heat loss to that through an equivalent arca of the casting surface. For a flat casting
surface, the heat loss can be deduced from Eqs. (95) and (96); it is

(112)

Toen by comparing Eqs. (111) and (112), we get

z1 = Bi,(1 + Bi)p'H;/Bi (113)

where Bi, is the Biot number associated with the heat loss from the top of the riser,
defined as
Bi, ~ (h + h,)L/k.

Example 22

Derive the asa factor for an open riser attached to a bar-like casting,
produced in a ceramic shell mold.

63
 Compare Eqs. (97) and (111) and solve for their ratio. Then

Z¡ =
(h + h,)A,(TM - T,.)t,
hcAc(TM - T,.)t,
h R
+ _c_lIn
k [ :: ] ]
With A, = A"' this simplifies to

h R
+ _c_lIn
k [ :: l l (114)

Example 23

Refer to Example 9 and calculate the size of an open end riser. The end
riser is attached to an iron casting that is bar-like with dimensions 1.5 X 1.5 x 6
(inches). The ceramic shell mold is 0.4 in. in thickness, and it has thermal
properties similar to mullite in Table 10. Assume that he = 10 Btu/h ft2 ºF and
f = 0.05.
From Example 9, h = 3.45 and h, = 18.77 Btu/h ft2 ºF.

[; L = 0.375 in.

Both the riser and the casting are bar-like and encased in the same shell material;
therefore e = 1. Also, because an end riser is specified, then z2 = 1. The asa
factor for the top of the open riser is given by Eq. (114).

z• (3.45 ;o 18.77) [ !:! + (1~~~.5) In [ !:; ] ]

z1 = 37.6

This extremely high value of the asa factor clearly indicates that alloys or metals
with high melting points should not be cast with open risers.

In order to use Eq. (109), we calculate K1 and K2•

K1 = o.375 = 0.395 in.

1-0.05

K2 = (4)(0.05)(13.5) = 0.905 in.3

(T){0.95)

64

L J
 Then (assume x = 1)

D3 - 0.395(4 X 1 X 2 + 1 + 37.6)D2 - 0.905 0

and
D = 18.41 in. and H = 36.82 in.

This is obviously an absurd design! The top of the riser must be insulated.

Example 24

Repeat Example 23 but assume that the top of the riser is perfectly
insulated (i.e., z2 = O).

Now the cubic equation for the diameter is

D3 - 0.395 ( 4 x 1 X 2 + 1 + O) D 2 - 0.905 0

then
D = 3.62 in. and H = 7.24 in.

This is still a rather large riser but vastly improved over the open top riser. The
yield can be irnproved more by insulating the side of the riser.

B. Feeding Distance and Riser Location

Alloys that solidify in molds of insulating materials freeze over a range of

temperatures so that a portian of the casting or the entire casting exists in the solid plus
liquid condition for a substantial part of the solidification period. Feed metal, which
must compensate for the solidification shrinkage, must flow through a tortuous maze of
the dendritic solid. When the resistance to the flow of this metal becomes too great,
interna! porosity results. This is most evident in alloys with a narrow freezing range that
are cast into shapes with uniform cross sections. In these instances, centerline shrinkage
may result particularly if a single riser is used to feed an excessive length of the casting.
Therefore risers must be located such that the feeding distance of the alloy is not
exceeded.

Feeding distance depends on the specific solidification characteristics of the alloy,

the mold, and the extent of acceptable porosity in the casting. Feeding distances for
sorne alloys can be found in metal casting textbooks, and metal casters, themselves,
develop their own rules based upon experience. Figure 20 shows examples of feeding
distances in steel plates and bars made in silica sand molds. Provided that the feeding
distance is not exceeded and the risers are properly designed, sound castings result.

65
 1-r.,F- ¡-,.1r--
.,.
r
'--------+--_.
__ _J_j_
1r-l-1:,r-l
ll1:-rr rnntril,r,tion .. Al:t" mnuilmtiun H,,,., 1:<1·:••
,,,111111 .. ,i,,,11 11>n1rilt1111••n
l.111¡:,th J(.íl'.1ll'f 111:in 1111, ,j,.f:11,··r

- 1.sr ,,, 2r¡-

T
..._ __ _¡__j,._ __ i.._ r
:0-.!im•I__.L

\l:1:c. ,li .. c:,nf"f" 1111

--11"1,,-11·-

T r
r ._ 1'
_;_J_

'--------------'~ ¡.,.,,1tth ¡:r,·;1ln 111.111 m.,, ,11-t,,11,,•

l.4·11'="1h ~rt·:1!c-r 1l1.1n rn:1'1:. ,l, .. 1:111<'1'

(a) (b)

Figure 20: Feeding distances in bars (a) and plates (b). (From R. A Flinn,
Fundamentals of Metal Casting, Addison Wesley, Reading, MA, 1963,
pp. 54-55.)

In addition to feeding distance, the placement of risers depends on the freezing

arder in the casting. Castings are seldom simple elemental shapes as discussed so far, but
complex castings can be subdivided into simple elements, which joined together make the
casting. Then the freezing order of the elements must be determined and the riser
attached to the element that freezes last, with no element isolated from the riser by an
earlier freezing element. Toe best way to illustrate the placement of risers is with sorne
examples.

Example 25

An aluminum investment casting comprises two bar-like sections connected

by a plate. Dimensions are in inches. Design (a) riser(s) for the casting.

66
 Ti'----~-iJ2f
-*-- • I'
LJ~
1-A---s--·l e 1-

Toe casting is divided into three elements: A, B and C. Toen we estimate

the freezing order. For element A:

VA = 12 in.3

CA == 0.241 (from Example 18)

12
(V/A)A = = 0:393 in.
2(2x3) + 2(2x2} + 2(3 x2) - 1/zx3

For element C:

ve = 4.5 in.3

ce = 0.241

45
(V/A)e = = 0.272 in.
2(1x3) + 2(11/zXl) + 2(1'hx3) - 1/zx3

For element B:

Va = 6 in.3

Ca = 0.220

(V/A)a = 0.25 in. (semithickness)

67
Toen the freezing order is

(V/A)A
> >
CA

1.63 > 1.12(9) > 1.13(6)

With this simple model our ability to distinguish between the freezing times of elements
C and B is not as precise as the numbers indicate. To be safe, we assume it is in the
order shown and attach risers to elements A and C.

Toe riser attached to A must be designed to solidify after a portion of the casting
with the modulus of element A, but to feed a portion that has the volume of element A
and more than one-half of the volume of element B. So in the riser design equation we
use the following values:

J,;'. ~ + [ 1.6/~\14 ] VB

= 15.53 in.3

(Toe ratio in the parenthesis is the proportion of element B that is fed by this
riser.)

We assume that the top of the riser is perfectly insulated (z¡ = O) and that a
side riser is employed (e = z2 = 1). Also
0.393
K1 = = 0.414 in.
(1 - 0.05)

(4 )(0.05)( 15.53)
1.041 in.'
T(l - 0.05)

Toen, Eq. (107) gives (assume H = D):

D3 - 0.414(4 x 1 x 1 + 1 + O)D2 - 1.041 = O

Solving: D =H = 2.27 in.

68
 Now for the riser attached to element C:

l : L = 0.272 in.
V
e
= 4.5 + [ 1.14
1.63 + 1.14
] 6

= 6.97 in.3

(Notice that the sum of the volumes fed by both risers equals the total
volume of the casting.)

Again z = O, e = 1 and z2 = l. Also

0.272
K l- - = 0.286 in.
(1 - O.OS)

(4)(0.05)(6.97)
0.467 in.3
..-(1 - O.OS)

Then, with H = D:
D3 - 0.286 ( 4 x 1 x 1 + 1 + O) D 2 - 0.467 O

Solving: D = H = 1.61 in.

From a riser design point-of-view, the two risers should be adequate. However,
the feeding distance in the plate is probably exceeded (see Fig. 20b), and perhaps
another riser (or two) would have to be attached to the plate, itself. A better solution
would be to increase the thickness of the plate, and then taper it from element C
toward element A.

69
VI. NUMERICAL METIIODS

Toe solutions in the previous chapters apply to situations of simple geometry. Of

course, casting geometry is typically complex so other means of estimating the
temperature field during solidification must he sought. It was also mentioned in
Chapter II that alloys freeze over a range of temperatures, and in Chapter III we saw that
thermal properties generally vary with temperature. When these added complexities are
to be included in solving solidification problems, numerical methods should be used. One
popular and often used method is called the finite difference approximation (FDA).

In this chapter two problems, that were previously discussed in §IV.e are
reviewed, and then calculated results based on FDA are compared to the exact solutions.
Here the major intent is to introduce the reader to the rnethod of FDA, so that simple
geometry and constant thermal properties are maintained.

A. Conduction in Mold or Core Surrounded by Metal

Let us look again at the example of a series of alternate platelike sections of

solidifying metal and mold and look at the temperature in the mold versus time. We
assume that the metal is at its solidification temperature and that the thermal properties
of the mold are constant, This situation is depicted in Fig. 17.

Once again the governing differential equation is given by Eq. (10), which is

iJT (10)
Tt
The boundary conditions are

iJT (L,t)
iJx
o t '<!: O, (115a)

and
T(O,t) ~ o (115b)

Toe initial condition is

T(x,O) = T;. (115c)

In Eqs. (115a,b,c), x = L is at the centerline of the mold, L is the sernithickness of the

mold, TM is the solidification temperature of the metal, and T¡ is the initial temperature
of the mold. • Toe domain O ~ x ~ L is subdivided into small segments of length .1x,
as shown in Fig. 21. Nodes are numbered consecutively from Jeft to right from O to N.

"In §IV.C, z = O was at the centerline. Toe two different coordinate systerns can be
reconciled with a simple coordinate transformation.
Heal Tranefer Fundamental• for Metal Caalíng
D.R. Poírier and E.J. Poírier 70
The Mineral,. Meole & Material, Sociely, 1992
 ; 1
1
1 1
1 1 1
1 1 1
1 1 1 1
1
1 1 -'6x ¡-- 1
1 1 1 1 1 1 Centerline

• • t !1 t
1 1
+1
1
1
1
1
1
1
1
:
1
1
'
1
1
1
1 1
1 1 t
o '
1 1
m-1 m m+I N
(x=Ol (x=U

Figure 21: The domain O ~ x ~ L subdivided into segments.

At a particular time, the gradient at node m and between m and m + 1 is

Similarly,

For the conduction equation, the second derivative is required; this is

t [ !L ~ L [ [ !L - [ !L] ·
Therefore,

and Eq. (10) for an interior node becomes

arm a (116)
[T 2T Tm.i ]·
_
a, = --
(11x)2 m-
1 -
m
+

At the centerline (m = N), the finite difference equation is derived in a different

manner. Toe flux to the centerline is

71
and the accumulation of energy is
áx aTN
-pC -,
2 P ar
where áx/2 is the volume associated with nade N. Toen at this insulated surface, the
finite difference equation is written as

(117)

For convenience, normalized variables are introduced. For time and position, these
are exactly the same as the dimensionless numbers that were introduced in Chapter IV.
Toe normalized temperature (u), time (8) and position (r') are defined as follows:
T - T1
u =
TM - T/

at
8 Fo
L2'

and

X = LX

With the normalized coordinates, Eqs. (10) and (l lSa,b,c) are replaced by
au (118)
a8

~ (1,8) o (119a)
ax'

u(0,8) (119b)

and
u(x ·.o) = o. (119c)

For the interior nades, the finite difference equation, Eq. (116), becomes
éJUM 1 (120)
a6 ""' (áx ')2 [ Um-1 - 2um + Um•I ],

72
and at the insulated surface (m = N) Eq. (117) becomes
iJuN 2
a8 = (ÁX ')2 [ UN-1 - UN l· (121)

For an actual application, the number of nodes should be Iarge, but in order to
illustrate the calculation process, we take N = 4. Equation (119b) requires u0 = 1; then
the difference equations for nodes 1, ... N are as follows:

(122)

(123)

(124)

(125)

Toe initial temperatures (fJ = O) are u1 = u2 = u3 = u4 = l.

There are severa! methods for proceeding; here the so-called Euler method is
used, in which

(126)

It is necessary to add superscripts to indicate time; e.g., u; is the present time and u;+1
is the future temperature after one time step of duration ÁfJ.

By combining Eqs. (122) through (126) for m = 1, ... N, with the appropriate
superscripts, we get

u;-+1 1 + (1 - 2P) uj' + P u2

p uj' + (1 · 2P) Uz + p u;
p U2 + ( 1 • 2P) U3 +

2P u3 + (1 - 2P) uJ (127a,b,c,d)

where P = ÁfJj(Áx')2. Initially each u;;. is known (u~ = O), so that Eqs. (127a,b,c,d)
comprise four equations with the four unknowns, u;+ 1• After the simultaneous equations

73
are solved, the temperatures at m = 1, ... 4 are known for time step v+ 1 (8 = t..8) and
then the process is repeated foras many time steps (8 = vt..8) as desired.

With the Euler method, numerical oscillations in the solution result if either t..fJ
or t..x' are too large. Specifically, the Euler method requires that P = t..8/(t..x')2 :5 0.5,
in order to prevent numerical oscillations. Table 11 gives calculated temperatures al the
insulated surface for N = 5 (t..x' = 0.2) and N = 10 (t..x' = 0.1) with P = 0.25. Also
shown are the temperatures based on the exact solution. With more nodes the results
are closer to the exact solution. With even more nodes, the agreement between the
approximated temperatures and the exact temperatures would be even closer. However,
with smaller values of t..x', t..8 has to be correspondingly reduced in order to meet the
criterion of P s 0.5. Accordingly, more computer time would be required to execute the
calculations.

Table 11

Normalized Temperature at the Insulated Surface

with P = 0.25

With With Exact

8 t..x' = 0.2 t..x' = 0.1 Solution
0.2 0.225 0.229 0.235
0.4 0.522 0.527 0.531
0.6 0.710 0.711 0.714
0.8 0.822 0.824 0.826
1.0 0.892 0.893 0.894

The numerical results can also be used to estimate the amount of energy taken
up by the mold (i.e., the domain O s x s L) as a function of time. There are two
methods for carrying out this estimate.

The first method is to track the gradient at the mold surface in contact with the
metal. This surface is at x = O in Fig. 21. For each time step the energy absorbed by
the mold can be approximated as the product of the flux and the duration of the time
step. By summing these products for ali of the time steps, the integrated or total energy
absorbed by the mold can be estimated. Toe accuracy of this method depends on the
accuracy of the estimate of the gradient at x = O.

The second method is based upon calculations of the average temperature in the
mold. For this particular problem, this method is more accurate than the first because
the estimates of temperature are more exact than the estima tes of the gradient at x = O.

74
 In the first method, the estimated gradient, in terms of normalized variables, is
(u1 - u0)/t.x'. In terms of the original dimensional variables, this is

l[
Therefore, the flux at the surface is approximated as

q0v = -k: T¡ - T; = -k: [ TM - T¡ _t.u ]' , (128)

t.x L Ax' 0

and the energy absorbed by the mold of area A for one time step is q~AAt. Therefore,

= Ak [ T -T.]
M
L
1 -
L2 •
a
E
I [ ::. [ AB

= ALpCp(T M - Tt) E1
(129)

As before the energy given up by the solidified portion of the casting (assumed to be
pure metal) is

(130)

where V is the volume solidified next to the casting-mold surface. With Óc = Ó,,.,
Eqs. (129) and (130) finally give

(V/A) = E
-y;¡- 1 (131)

where -y is defined by Eq. (84). Equation (131) is the approximate counterpart to the
exact solution given by Eq. (90).

75
 In the second method, the average normalized temperature at a given normalized
time is calculated, Toe volume of material associated with the nodes m = O and m = N
is 0.5 llx', so the average norrnalízed temperature is

u
-V
= -
N
1
[
-
2
1
(u0 +
V
uN) E u,. .
Y
+
m=I
N-1 V ] (132)

The average dimensional temperature is

f v = u•(TM - T;) + T1,

so that the energy absorbed by one-half of the mold must be

(133)

with Q,. = Óeo we combine Eqs. (130) and (133); the result is

(V/A) _ -. (134)
-- -u.
L"Y
Of course, Eq. (134) is another approximate counterpart to the exact solution given by
Eq. (90).

Approximations of V/AL"Y based on Eqs. (129) and (134), with P = 0.25 and
N = 10, are given in Table 12. Also shown are values of V/AL"'( calculated from the
exact solution, Eq. (90). Toe table shows that the estimates based on calculating the
average temperature are more accurate than are those based on estimating the gradient
at the mold-casting interface.

Table 12

Amount of Solidification as V/AL'( versus (J

with P = 0.25 and N = 10

Finite Difference Approx.

(J Therm. Grad. Avg. Temp. Exact Solution

0.04 0.163 0.231 0.229
0.2 0.434 0.506 0.505
0.4 0.626 0.699 0.698
0.6 0.742 0.817 0.817
0.8 0.814 0.888 0.886
1.0 0.857 0.932 0.930

76
 B. Conduction in She/l Mold with Heat Loss at the Surface

The shell mold has a thickness L. In this case, the surface in contact with the
metal is located at x = O, and the outer surface of the mold is at x = L (see Fig. 22).
Also, the normalized temperature is selected so that u = 1 and u = O correspond to
T = TM and T = T.. , respectively; thus

u =

By selecting (J = at/L2 and x' = x/L, then the heat conduction equation is Eq. (118).
The boundary conditions and initial condition are as follows:
u(0,6) = 1,

!!:_ (1,6) + Bi u(1,6) = O,

ax'
, T¡ - T ..
u(x ,O) = = u;. (135a,b,c)
TM - T ..
The temperature at nodes 1 through N-1 are approxirnated by Eq. (120). By
writing an energy balance for node N (with a volume of Ax'/2), we get

1 [
_
auN
= --- -2(1 + Bi Ax . ) UN + 2uN-I
] (136)
iJ(J (Ax')2

with Bi = hL/k.

In order to advance the solution in time, we again utilíze the Euler method and
combine Eq. (126) with Eqs. (120) and (136). For nodes 1 through N-1, we get
(137)

and for node N

u•/ = 2P u~_1 + [ 1 - 2P(1 + Bi Ax') J u~. (138)

As before P = A6/(Ax')2, and the superscripts v and v+ 1 represent the present and next
time, respectively.

By selecting P = 0.25 and Ax' = 0.1, then A6 = 0.0025; with u0 = 1 the set of
equations becomes:

ur+I = 0.25 + 0.5 U~ + 0.25 u; ;m = 1

0.25 u;_1 + 0.5 u; + 0.25 u:;.+1 ; m = 2, 3, ... 9

v+I
U 10 0.5 u9 + (0.5 - 0.125 Bi) u10 ; m =N= 10

77
 1 ' 1 • 1 1
t 1 _J I 1 1
1 ,t.xt-- 1 1
T=TM}
u=I
1
1
1
I ¡ 1 1 t h
1 1 1 1 I
1
1
1 1 1 1 1

'
1 +t t1 +1 t •1
1
1
1
l1 1
1
1
1
1
1
1 1 1 1 1
1 1 1 1 1
O I m-1 m m+I N
(x=O) (x=L)

Figure 22: Shell mold of thickness L subdivíded into segments

for estimating temperature during heat conduction.

Sorne calculated temperature distributions are shown in Fig. 23 for two initial
values of u~ and two Biot numbers. Notice that steady state is achieved in a relatively
short period of 8 = 1, which corresponds to a real time of only a few or severa] minutes.
Notice, too, that the temperature of the exterior surface of the mold decreases and then
increases to its steady state value in sorne circumstances (e.g., Fig. 23a).

Again the thermal gradients at x' = O derived from the temperature distributions
can be used to calculate the total energy extracted from tl-e casting. Hence to estímate
the thickness solidified from the mold wall, Eq. (131) applies. These estimates are shown
in Fig. 24.

C. Closing Remarks

In the two previous sections, a numerical method, called finite difference

approximations (FDA), was used to calculate temperature distributions in shell molds of
simple geometry. The examples were selected with the intention of introducing the
numerical method; hence, the examples were kept simple. The thermal properties of the
mold and the heat transfer coefficient at the exterior surface of the mold were constants,
In practice, however, these properties depend on temperature.

It should be noted that variable properties can be included in the FDA method
by updating the properties with the current temperatures as the calculation proceeds from
one time step to the next. The FDA method can also be extended to two dimensional
and three dimensional heat conduction problems so that complex casting geometries can
be analyzed.

As mentioned in §11.B, alloys solidify over a temperature range; in such cases,

nodes are assigned in the solidifying casting as well as in the mold. Sophisticated
computer codes can be used to show the positions of isotherrns in both the mold and
casting, to assist the casting engineer in locating "hot spots," designing and locating risers,
estimating solidification times, and in predicting microstructural features and physical
properties that depend upon the temperature history during solidification.

78
 u
0.4

0.2 Bi=0.5 Bi =2

O.O L---L---'---_,___ _ _._ _ __,

(a) (b)

u
0.4

0.2 Bi=0.5 Bi=2

0·0a.o
0.2 0.4 o.6 0.0 1.0 o.o 0.2 0.4 o.6 o.8 1.0
X.' x'
(e) (d)

Figure 23: Calculated temperature profiles in a shell mold of finite thickness with heat
loss from the surface. (a) Preheat temperature u~ = 0.536 and Bi = 0.5;
(b) u~ = 0.536 and Bi = 2; (e) u~= 0.778 and Bi = 0.5; (d) u~= 0.778 and
Bi = 2.

79
 0.8

---Bi=2
0.6 ---- Bi=0.5

>,.
_J

--
<{
<,

-- -
>0.4

02 --- -- - -- -
o.o ~-_.___...._ _ _.__ _._ _ __.__ __,__ __.__ ,____..____.J
O.O 0.2 0.4 0.6 0.8 1.0
8=at/L2

Figure 24: Estímate of thick.ness solidified (V/A) versus time in shell mold of finite
thickness with heat loss from the exterior surface.

80
SUGGESTED READINGS

J. P. Holman: Heat Transfer, 6th ed., McGraw-Hill, New York, NY, 1986.
This is a widely used textbook in many mechanical engineering programs.

F. P. Incropera and D. P. DeWitt: Fundamentals of Heat and Mass Transfer, 3rd ed.,
John Wiley & Sons, New York, NY, 1990.
Thís, too, is widely used in mechanical engineering programs.

G. H. Geiger and D. R. Poirier: Transport Phenomena in Metallurgy, Addison-Wesley,

Reading, MA, 1973.
This text comprises fundamentals of fluid dynamics, energy transport and mass
transport with applications in metallurgical engineering. One chapter is devoted
to solidification heat transfer.

H. S. Carslaw and J. C. Jaeger: Conduction of Heat in Solids, 2nd ed., Oxford University
Press, Oxford, U.K., 1959.
"Carslaw and Jaeger" is a mathematical treatise of solutions to the conduction heat
transfer equation. It is well organized and can also be used a catalog of analytical
solutions. It is considered to be a classic by many workers in heat transfer.

G. E. Myers: AnalyticalMethods in Conduction Heat Transfer,McGraw Hill, New York,

NY, 1971.
Myers wrote this text in a "self-teaching" style. His chapters on numerical methods
are excellent, especially for readers who need a review of matrices.

R. D. Pehlke, A. Jeyarajan and H. Wada: Summary of Thermal Properties for Casting

Alloys and Mold Materiats, National Science Foundation, Grant No. DAR78-26171, Dec.
1982.
This report gives thermal properties as a function of temperature for metals,
alloys, and mold materials.

D. R. Poirier and N. V. Ghandi: AFS Trans., vol. 84, 1976, pp. 577-584.
The riser design equation is reviewed and applied in an optimization scheme for
minimizing the manufacturing cost associated with risers.

H. Huang, J. T. Berry, X. Z. Zheng and T. S. Piwonka: "Thermal Conductivity of

Investment Casting Cera mies," 37th Annual TechnicalMeeting,lnvestment Casting Institute,
1989.
Thermal conductivity data on ceramic shell molds are not generally available.
From its title, it is obvious that this is an important reference for investment
casting engineers.

H. F. Taylor, M. C. Flemings and J. Wulff: Foundry Engineering, John Wiley & Sons,
New York, NY, 1959.
Although this book is rather old, the chapter on risering and riser design is still
very appropríate.

81