Anda di halaman 1dari 10

Materials and Design 30 (2009) 35923601

Contents lists available at ScienceDirect

Materials and Design


journal homepage: www.elsevier.com/locate/matdes

Melt characteristics and solidication growth direction with respect to gravity


affecting the interfacial heat transfer coefcient of chill castings
No Cheung a, Ivaldo L. Ferreira b, Moiss M. Pariona c, Jos M.V. Quaresma d, Amauri Garcia a,*
a

Department of Materials Engineering, University of Campinas, UNICAMP, P.O. Box 6122, 13083-970 Campinas, SP, Brazil
Department of Mechanical Engineering, Fluminense Federal University, UFF, Av. dos Trabalhadores 420, 27255-125 Volta Redonda, RJ, Brazil
c
Department of Mathematics and Statistics, State University of Ponta Grossa, UEPG, 84030-900 Ponta Grossa, PR, Brazil
d
Federal University of Par, UFPA, Augusto Correa 1, Guam, 66075-110 Belm, PA, Brazil
b

a r t i c l e

i n f o

Article history:
Received 15 January 2009
Accepted 26 February 2009
Available online 5 March 2009
Keywords:
Non-ferrous metals and alloys (A)
Casting (C)
Thermal analysis (G)

a b s t r a c t
For purposes of an accurate mathematical modeling, it is essential to establish trustworthy boundary
conditions. The heat transfer that occurs at the casting/mold interface is one of these important conditions, which is a fundamental task during unsteady solidication in permanent mold casting processes.
This paper presents an overview of the inverse analysis technique (IHCP) applied to the determination
of interfacial heat transfer coefcients, hi, for a number of alloy solidication situations. A search algorithm is used to nd the transient metal/mold interface coefcient during solidication which is reported
either as a function of the casting surface temperature or time. Factors affecting hi such as the direction of
gravity in relation to the growth interface, the initial melt temperature prole, the wettability of the
liquid layer in contact with the mold inner surface, were individually analyzed and experimental laws
for hi have been established.
2009 Elsevier Ltd. All rights reserved.

1. Introduction
Inverse problems are encountered in various branches of science and engineering. Mechanical, materials, aerospace, chemical
and metallurgical engineers, astrophysicists, geophysicists, statisticians and specialists of many other disciplines are all interested in
inverse problems, each with different application in mind. In the
eld of heat and mass transfer, the use of inverse analysis for the
estimation of surface conditions such as temperature and heat ux,
thermal gradient, or the determination of thermal properties such
as thermal conductivity, heat capacity, enthalpy, latent heat and
densities of solid and liquid by utilizing transient temperature
measurements taken within the medium has a wide range of practical applications. The determination of transient metal/mold heat
transfer coefcients as a function of position and time during solidication of multicomponent alloys is an example of difcult
numerical treatment. In such situations, the inverse method of
analysis, using transient temperature measurements taken within
the medium can be applied for the estimation of such quantities.
However, difculties associated with the implementation of inverse analysis should be also recognized. The main difculty arises
from the fact that inverse solutions are very sensitive to changes in
the input data resulting from measurements and modeling errors,
hence may not be unique. Mathematically, the inverse problem be* Corresponding author. Tel.: +55 19 35213320; fax: +55 19 32893722.
E-mail address: amaurig@fem.unicamp.br (A. Garcia).
0261-3069/$ - see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2009.02.025

longs to the class of problems called the ill-posed problems, that is,
their solution does not satisfy the general requirement of existence, uniqueness and stability under small changes to the input
data. In order to overcome such difculties, a variety of techniques
for solving inverse heat transfer problems have been proposed [1].
The way heat ows through the casting/mold interface affects
the evolution of solidication, and is of notable importance in characterizing the ingot cooling conditions, mainly for the majority of
high heat diffusivity casting systems such as chill castings [2].
When the metal comes into contact with the mold, at the metal/
mold interface, the solid bodies are only in contact at isolated
points and the actual area of contact is only a small fraction of
the nominal area, as shown in Fig. 1.
Part of heat ow follows the path of the actual contact, but the
reminder must pass through the gaseous and nongaseous interstitial media between the surface peaks. The interstices are limited in
size, so that convection can be neglected. If temperature differences are not high, radiation does not play a signicant role and
most of the energy passes by conduction across the areas of actual
physical contact. The heat ow across a casting/massive mold
interface, can be characterized by a macroscopic averaged metal/
mold interfacial heat transfer coefcient (hi) given by,

hi

q
AT IC  T IM

where q (W) is the global heat ux of the interface; A (m2) is the


area and TIC and TIM are the surface casting temperature and the

3593

N. Cheung et al. / Materials and Design 30 (2009) 35923601

Tp

Mold

TIC

TIM
Tmold

Casting

Fig. 1. Heat ux at the metal/mold interface.

temperature of the mold inner surface (K), respectively. In watercooled molds, the global equivalent heat ux is affected by a series
of thermal resistances, as shown in Fig. 2,
The global thermal resistance 1/hi can be expressed by:

1
1
e
1

hi hW kM hM=M

where hi is the global heat transfer coefcient between the casting


surface and the cooling uid (Wm2 K1), e is the mold thickness
(m), kM is the mold thermal conductivity (Wm1 K1), and nally,
hW is the mold/cooling uid heat transfer coefcient (Wm2 K1).
The averaged heat ux casting/cooling water is given by [3]:

q hi T IC  T 0

where T0 is the water temperature (K).


The thermal resistance at the mold/air interface, RM/A , can be
calculated as a function of the measured mold wall temperatures
(TEM) and the free-stream air temperature (T0), as shown in
Fig. 3, and is given by:

RM=A

1
hR hC AT

where AT is the chill cross-section area (m2) and hR and hC are the
radiation and convection heat transfer coefcients, respectively,
given by:

hR r  eT EM T 0 T 2EM T 20

where r is the StefanBoltzmann constant (5.672  108


Wm2 K4) and e is the mold emissivity.
The convection heat transfer coefcient is given by [4]:

hC

kNu

where k is the uid thermal conductivity (W m1 K1) and hC is represented in terms of the Nusselt number (Nu). For free convection
Nu can be calculated as a function of Grashof (Gr) and Prandtl (Pr)
numbers, as follows:

Nu CGr  Prn

where C and n are constants, and v is a characteristic length of the


solid surface (m), which in the particular case of Fig. 3 is the chill
vertical length . Gr and Pr are given respectively by:

TIC

Fig. 3. Thermal resistances in a chill mold.

Gr
Pr

g  c  v3 T EM  T 0
hg
k

c

g2

q2s

8
9

where g is the gravitational acceleration (m s2), c is the volume


coefcient of expansion (for ideal gases c = 1/ T0 (K1)), g is the uid
viscosity, q is the uid density and c is the uid specic heat [5].
For successful modeling of casting processes, reliable heat
transfer boundary conditions are required, in particular the metal/mold heat transfer coefcient. The accurate knowledge of this
coefcient is necessary for accurate modeling of casting dimensions and casting microstructure [6,7]. Many investigations concerning the heat transfer coefcient between metal and mold in
casting systems have been carried out, and pointed out the importance of the development of appropriate tools to predict the heat
transfer coefcient, hi. Most of the methods of calculation of hi
existing in the literature are based on temperature histories at
points of the casting or mold together with mathematical models
of heat transfer during solidication. Among these methods, those
based on the solution of the inverse heat conduction problem
(IHCP) have been widely used in the quantication of the transient
interfacial casting/mold heat transfer [814]. In general, hi is not
constant but varies during solidication and depends upon a number of factors. These factors include the thermophysical properties
of the contacting materials, the casting geometry, mold temperature, pouring temperature, the roughness of mold contacting surface, mold coatings, etc [15].
The purpose of the present study was to investigate the inuence of three important factors on the interfacial heat transfer
coefcient: the initial melt temperature prole, the wettability of
the liquid layer in contact with the mold inner surface, and the
direction of gravity in relation to the growth interface. Temperature readings, recorded by a bank of thermocouples distributed inside the casting, were used as input data for an inverse heat
conduction method in order to determine the time-varying interfacial heat transfer coefcient, hi. Casting experiments were carried
out with AlCu, AlSi, AlSn, SnPb, and PbSb alloys, which were
unidirectionally solidied in a massive chill mold and in a watercooled mold under different parametric solidication conditions.
Simulations were performed using a two-dimensional version of
a numerical heat transfer solidication model.

R3 = 1/ hM/M

2. Numerical modeling

Liquid

R2= e / k
Solid

e
Water

2.1. Governing equations


R1 = 1 / h w
To

Fig. 2. Thermal resistances in a water-cooled metal/mold system.

The numerical model used to simulate the thermal eld during


alloy solidication is based on that previously proposed by Voller
[16]. Modications to this numerical approach have been incorpo-

3594

N. Cheung et al. / Materials and Design 30 (2009) 35923601

where sub-indices S and L refer to solid and liquid phases,


respectively, TF is the fusion temperature of the pure solvent
in (K) and C S is the solid concentration at the interface;
(vi) The specic heats, CS and CL, thermal conductivities kS and
kL, and the densities qS and qL, are constants within each
phase, but discontinuous at the solidliquid boundary. The
latent heat of fusion is taken as the difference between
phases enthalpies DH = HLHS.
(vii) The metal/mold thermal resistance varies with time, and is
incorporated in a global heat transfer coefcient dened as
hi [18].
Using the above assumptions, the mixture equations for binary
alloys solidication read:
 Energy
Fig. 4. Schematic casting initial melt temperature distribution (t = 0).

@ qcT
@g
r  qL cL uT r  krT  qS DH
@t
@T

rated to allow the use of different thermophysical properties for


the liquid and solid phases, as well as the mushy zone (it can deal
with temperature and concentration dependent thermophysical
properties), to treat variable metal/mold interface heat transfer
coefcient and to account for a space dependent initial melt temperature prole. A time variable metal/mold interface heat transfer
coefcient introduces a non-linearity condition at the z = 0 boundary. In addition, a variable space grid is used to assure the accuracy
of simulation results without considerably raising the number of
spatial nodes. Considering the previous exposed, the solidication
of binary alloys is our target problem. At time t < 0, the alloy is in
the molten state at the nominal concentration C0 and with an initial temperature distribution T0(z) = a  z2 + b  z + c, contained in
the insulated mold dened by 0 < z < Zb according to Fig. 4. Solidication begins by cooling the molten metal at the chill (z = 0) until
the temperature drops bellow the eutectic temperature TE. At times
t > 0, three transient regions are formed: solid, solid + liquid
(mushy zone) and liquid.
To develop a numerical solution for the equations of the coupled thermal and solutal elds, the following assumptions were
adopted:
(i) The domain is one-dimensional, dened by 0 < z < Zb, where
Zb is a point far removed from the chill.
(ii) The solid phase is stationary, i.e., once the solid has formed it
has zero velocity.
(iii) Due to the relatively rapid nature of heat and mass diffusion
in the liquid, within a representative elemental averaging
volume, the liquid concentration CL, the temperature T, the
liquid density qL and the liquid velocity uL are constants
[17].
(iv) The partition coefcient k0 and liquidus slope mL, are
obtained from the ThermoCalc software1.
(v) Equilibrium conditions exist at the solid/liquid interface, i.e.,
the temperature and concentrations fulll the equations:

T T F  mL C L

10

and

C S k0 C L

12

 Species

@ qC
r  qL uC L 0
@t

13

 Mass

@q
r  qL u 0
@t

14

where g is the liquid volume fraction and u is the volume averaged


uid velocity dened as:

u guL

15

 Mixture density

1g

qS da g qL

16

 Mixture solute density

qC

1g

qS C S da g qL C L

17

where qC is the volumetric specic heat, taken as volume fraction


weighted averages.
The boundary conditions at the domain are prescribed as:

u 0;
T ! Tp

@T
hi T 0  Tjz0 and
@z
and C ! C 0 at z Zb;

@C L
0 at z 0
@z

18
19

where Tp is the either a constant initial melt temperature or an initial melt temperature prole as a function of z.
The inverse problem consists on estimating the boundary heat
transfer coefcient at the metal/mold interface from experimental
temperatures in the casting. The inverse problem can be stated as
follows:
given M measured temperatures Tj (j = 1, 2, 3, . . . , N);
estimating the heat transfer coefcient given by its components hi (i = 1, 2, 3, . . . , N);

11

1
The ThermoCalc software [19] can be used to generate equilibrium diagrams and
through ThermoCalc interface for Fortran or C++ it is possible to recall those data
generated by the software in order to provide more accurate input values for model
simulations.

In order to solve the problem, the estimated temperature T est


i
(i = 1, 2, 3, . . . , N) computed from the solution of the direct problem
using the estimated values of the heat transfer coefcient components hi (i = 1, 2, 3, . . . , N), should match the measured temperatures
(i = 1, 2, 3, . . . , N), as close as possible, as shown by the
T exp
i
schematic representation of Fig. 5. This matching can be done by

N. Cheung et al. / Materials and Design 30 (2009) 35923601

3595

minimizing the standard least squares norm with respect to each


of the unknown heat transfer coefcient components.
This method makes a complete mathematical description of the
physics of the process and is supported by temperature measurements at known locations inside the heat conducting body. The
temperature les containing the experimentally monitored temperatures are used in a nite difference heat ow model to determine hi, as described in a previous article [5]. The process at each
time step included the following: a suitable initial value of hi is
assumed and with this value, the temperature of each reference
location in casting at the end of each time interval Dt is simulated
by the numerical model. The correction in hi at each interaction
step is made by a value Dhi, and new temperatures are estimated
[Test(hi + Dhi)] or [Test(hi  Dhi)]. With these values, sensitivity coefcients / are calculated for each interaction, given by:

T est hi Dhi  T est hi


Dhi

20

The procedure determines the value of hi, which minimizes an


objective function dened by:

Fhi

n
X
T est  T exp 2

21

i1

where Test and Texp are the estimated and the experimentally measured temperatures at various thermocouples locations and times,
and n is the iteration stage. The applied method is a simulation assisted one and has been used in recent publications for determining
hi for a number of solidication situations [2,2024].
The ow chart shown in Fig. 6 gives an overview of the solution
procedure.
3. Experimental procedure
Three different solidication apparatus have been used in the
experimental analysis and the assemblage details of these systems
are shown in Fig. 7.
In order to promote vertical upward solidication, an apparatus
designed in such a way that the heat was extracted by a watercooled bottom provoking upward directional solidication was
used (Fig. 7a). A stainless steel cylindrical mold was employed,
having an internal diameter (i.d.) of 50 mm, height of 110 mm
and wall thickness of 5 mm. The inner vertical surface was covered
with a layer of insulating alumina to minimize radial heat losses,
and a top cover made of insulating material was employed to reduce heat losses from the metal/air surface. The bottom part of
the mold was closed with a thin (3 mm) carbon steel sheet.
The use of a water-cooled stainless steel chamber at the top of
the casting has permitted experiments for downward directional
growth to be carried out (Fig. 7b). A stainless steel split mold
was used having an i.d. of 57 mm, height of 150 mm and wall
thickness of 10 mm. As mentioned before, alumina was applied

Fig. 6. Flow chart for the determination of metal/mold heat transfer coefcients.

at the mold inner surface in order to prevent radial heat losses.


The upper part of the split mold was closed by the cooling chamber
(3 mm thick wall).
In the upward and downward systems, the alloys were melted
in situ and the electric heaters had their power controlled in order
to permit a desired melt superheat to be achieved. To begin solidication, the electric heaters were disconnected and at the same
time, the water ow was initiated. Temperatures in the casting
were monitored during solidication via the output of a bank of
types J and K thermocouples accurately positioned with respect
to the heat extracting surface. In order to minimize temperature
eld distortions, the thermocouples were installed parallel to the
isotherms in the casting [7]. Further, the thermocouple tips were
placed as near as possible to the transversal geometric center of
the casting. The thermocouples were also calibrated at the melting

Fig. 5. Diagram showing domain for inverse heat conduction problems.

3596

N. Cheung et al. / Materials and Design 30 (2009) 35923601

Fig. 7. Experimental setups: (a) upward, (b) downward and (c) horizontal systems.

temperatures of aluminum and tin exhibiting uctuations of about


1.0 C and 0.4 C, respectively. Thermocouples readings (at intervals of 0.5 s) were collected by a data acquisition system and
stored in a computer.
Although the correct thermocouple positions with regard to the
heat extracting surface were veried before the experiments, a
deviation of about 1 mm from the nominal positions was observed for some of them as a result of interaction of sensors with
melt movement and casting shrinkage.
A third casting assembly was used for horizontal solidication
experiments (Fig. 7c). In order to promote unidirectional heat ow
during solidication, a low carbon steel chill with a wall thickness
of 60 mm was used, with the heat extracting surface being polished. Each alloy was melted in an electric resistance-type furnace
until the melt reached a predetermined temperature. It was then
stirred, degassed and poured into the casting chamber as soon as
the desired melt superheat was achieved. Temperatures in the chill
and in the casting were monitored during solidication via the output of a bank of thermocouples accurately located with respect to
the metal/mold interface. Unidirectional heat ow was achieved by
adequate insulation of the casting chamber.
4. Results and discussion
4.1. Inuence of melt temperature prole
Temperature les containing the experimentally monitored
temperatures were coupled to the numerical solidication pro-

gram for determining the transient metal/mold heat transfer


coefcient hi. Thermophysical properties of each alloy and solidication parameters are used as input data for simulations. Fig. 8
shows the temperature data collected in the metal during the
course of upward solidication of an Al 10 wt%Cu alloy casting
in the vertical water-cooled apparatus, with the bottom heat
extracting surface being polished. The experimental thermal responses corresponding to ve different positions inside the casting were compared with the predictions furnished by the
numerical solidication model. The best theoretical-experimental
t has provided appropriate transient hi prole for two different
approaches: (i) an average initial melt temperature has been
adopted (Fig. 8a), and (ii) a quadratic equation, based on experimental thermal readings, representing the initial melt temperature as a function of position in casting has been used (Fig. 8b).
A comparison between hi proles determined in each case is
shown in Fig. 8c. It can be seen that a signicant difference exists
between the two curves, with the assumption of a constant melt
temperature overestimating the metal/mold heat transfer coefcient. The two curves tend to approach each other with increasing
time.
In order to evaluate the real signicance of hi overestimation
additional simulations were conducted considering two-dimensional solidication. A regular geometry of an Al 10 wt%Cu alloy
square casting (100  100 mm2) was simulated by a 2D version
of the numerical approach described in Section 2, in order to evaluate the inuence of each hi prole previously determined, which
was imposed at the four faces of the square ingot. Fig. 9a and b

3597

700

700

600

600

500

500

Temperature [C]

Temperature [C]

N. Cheung et al. / Materials and Design 30 (2009) 35923601

400
300
200
100
0

5 mm
10 mm
15 mm
30 mm
50 mm
Numerical simulation

Al-10wt%Cu - Polished mold


Tp = 653.5 C (mean)
hi = 10800 . t

20

-0.075

[W/m K]

40

60

400
300
200

5 mm
10 mm
15 mm
30 mm
50 mm
Numerical Simulation

Al-10wt%Cu - Polished mold


2

100

80

100

Tp(z) = -4267.14 z + 734.04 z + 910.83 [K]


hi = 9000. t

-0.039

[W/m K]

20

40
60
Time [s]

Time [s]

(a)

80

100

(b)

11500
-0.039

hi = 9000.t

11000

-0.075

hi = 10800.t

[W/m 2K] - quadratic melt temperature profile


[W/m 2K] - constant melt temperature

10000

hi [W/m K]

10500

9500
9000
8500
8000
7500
0

20

40

60

80

100

Time [s]

(c)
Fig. 8. (a) Simulated and measured temperature responses for an Al 10 wt%Cu alloy casting at 5, 10, 15, 30 and 50 mm from the metal/mold interface adopting an average
melt temperature; (b) Simulated and measured temperature responses for an Al 10 wt%Cu alloy casting at the same positions adopting a melt temperature prole; and (c)
Evolution of metal/mold interface heat transfer coefcient (hi) as a function of time for an Al 10 wt%Cu alloy casting (polished mold).

Fig. 9. Isotherms (C) distribution for t = 13.75 s obtained considering (a) hi = 9000  t0.039 and (b) hi = 10,800  t0.075.

show some isotherms at the casting cross-section for t = 13.75 s


considering hi = 9000  t0.039 and hi = 10,800  t0.075, respectively.

It can be noticed that the liquid core is larger when the more accurate melt prole was adopted as can be seen by comparing Fig. 9a

3598

N. Cheung et al. / Materials and Design 30 (2009) 35923601

Fig. 10. (a) Comparison of the resultant experimental hi proles as a function of time for the PbSb alloys experimentally examined and (b) uidity behavior of PbSb alloys.

Fig. 11. Isotherms (C) distribution for t = 48 s considering (a) Pb 2.5 wt%Sb; hi = 4500  t0.11 and (b) Pb 3.0 wt%Sb hi = 3700  t0.11.

N. Cheung et al. / Materials and Design 30 (2009) 35923601

and Fig. 9b, i.e., the adoption of a simplied constant melt prole
will provide a quicker solidication evolution.
4.2. Effect of melt uidity
Fig. 10a shows the time dependence of the metal/coolant interface heat transfer coefcient (hi) during the course of different
experiments of upward directional solidication of PbSb alloys,
including the prole obtained for the eutectic composition. In order to permit more accurate values of hi to be determined, a quadratic function has been used to characterize the initial melt
prole, as discussed in the preceding section. The thermophysical
properties, the solidication range and the melt uidity are some
of the factors affecting hi. The surface roughness of the steel sheet
which separates the metal from the cooling uid has been
parameterized.
Although a single exponent 0.11 has been found for the power
laws characterizing the variation of hi with time, different multipliers have been obtained. Such multipliers seem to be mainly linked
to the wettability of the liquid layer in contact with the mold inner
surface, i.e., connected with the molten alloy uidity. Both liquid
metal and mold characteristics are involved in determining uidity
[25,26]. Fig. 10b shows the uidity superimposed to the PbSb
phase diagram. The uidity of PbSb alloys decreases from pure

3599

lead up to a range of compositions between 3.5 wt%Sb and


8.0 wt%Sb increasing again with increasing Sb content toward the
eutectic composition. The two extremes of the composition range
experimentally examined, i.e., the Pb 2.2 wt%Sb alloy and the eutectic composition are associated with the highest hi proles as
shown in Fig. 10. By observing Fig. 10 a correlation between the
multiplier (A) of the experimentally determined hi = f(t) equations
and the uiditys values can be established.
In Fig. 11, different locations of the isotherms, at t = 48 s, can be
realized for the simulation of the two-dimensional solidication of
two PbSb alloys (Pb 2.5 wt%Sb and Pb 3.0 wt%Sb). Although the
composition between the two alloys is quite close, the melt uidity
is signicantly different which means that specic hi proles have

18000
-0.1

hi = 10,500.t

16000

-0.1

hi = 6,000.t

14000

-0.1

hi = 12,500.t

- Al-20wt% Sn alloy
- Al-30wt% Sn alloy
- Al-40wt% Sn alloy

-2

-1

hi [W.m .K ]

12000
10000
8000
6000
4000
2000

100

200
Time [s]

300

400

Fig. 12. Evolution of metal/coolant interface heat transfer coefcient (hi) as a


function of time (t) during vertical upward solidication.

10000
Sn-5wt%Pb

9000
8000

-0.47

hi = 18000.t

hi [W/m K]

7000

[W/m K] - upward solidification

-0.001

[W/m K] - downward solidification

hi = 1650.t

6000

[W/m K] - horizontal solidification

-0.12

hi = 6000.t

5000
4000
3000
2000
1000

50

100

150
Time [s]

200

250

300

Fig. 13. Evolution of metal/mold interface heat transfer coefcient (hi) as a function
of time for a Sn 5 wt%Pb alloy solidied vertically upwards, downwards and
horizontally.

Fig. 14. Isotherms (C) distribution during solidication (for t = 80 s) of a Sn


5 wt%Pb alloy casting: hi = 1650  t0.001 over the upper surface; hi = 6000  t0.12
over the bottom surface; hi = 18,000  t0.47 over the lateral surfaces (a) considering
heat transfer in the liquid metal only by conduction (b) considering also uid ow.

3600

N. Cheung et al. / Materials and Design 30 (2009) 35923601

to be considered, i.e., the adoption of a same hi prole for both alloys can induce important differences. Indeed, the interfacial heat
transfer coefcient does inuence solidication behavior as it is
evident from the simulated isotherms in Fig. 11. Whilst for the
Pb 2.5 wt%Sb alloy (Fig 11a) the solidication is almost complete
the Pb 3.0 wt%Sb alloy casting is not ready to be unmolded.
Fig. 12 shows the time dependence of the overall metal/coolant
heat transfer coefcient (hg) during the course of different experiments of upward directional solidication of AlSn alloys in uncoated cooled molds. Although a same exponent 0.1 has been
found for the power laws characterizing the hg variation with time,
very different multipliers have been obtained. Such multipliers are
mainly linked to the wettability of the liquid layer in contact with
the mold inner surface, i.e., connected with uidity. Both liquid
metal and mold characteristics are involved in determining uidity. The lowest hg prole refers to the Al 30 wt%Sn alloy, while
the other two alloys present higher hg proles. It has been demonstrated that when uidity is superimposed to binary constitution
diagrams, the best uidity is attained for pure components, eutectics or phases that freeze congruently [26]. It seems that for AlSn
alloys the uidity decreases from pure aluminum up to a composition about 30 wt%Sn increasing again with increasing Sn content
toward the eutectic composition. This is reected by the multipliers of the experimentally determined hg = f(t) equations, shown in
Fig. 12.
4.3. Effect of growth direction with respect to gravity

4900

4900
4200

- Al-5wt% Si

-0.09

- Al-7wt% Si

-0.09

- Al-9wt% Si

- Al-5wt% Si

hi = 4500 (t)

-0.001

- Al-7wt% Si

hi = 3900 (t)

hi = 2100 (t)

-0.001

hi = 1100 (t)

hi = 3300 (t)

- Al-9wt% Si

4200
3500

3500

-0.09

-0.001

hi = 2400 (t)

hi (W/m K)

Metal/Coolant Heat Transfer Coeficient

The inuence of the direction of growth on hi during solidication has been experimentally examined for opposite conditions
with respect to the gravity vector (upward and downward solidication) and by using alloys of quite different thermal responses
during solidication (SnPb and AlSi). For the SnPb alloy the
horizontal conguration has also been examined.
The best theoretical-experimental cooling curves t has provided an appropriate transient hi prole during solidication of a
Sn 5 wt%Pb alloy. Fig. 13 shows such proles during the course
of different experiments involving downward, upward and horizontal directional solidication. The heat transfer coefcient is
clearly dependent on the orientation of solidication with respect
to gravity. In the upward vertical solidication the effect of gravity
causes the casting to rest on the mold surface, but during downward solidication, this action causes the solidied portion of the
casting to retreat from the mold surface. It is well known that
the reduction in the contact pressure between casting and mold

surfaces leads to a consequent reduction in the interfacial heat


transfer efciency.
The heat transfer coefcients for both upward and horizontal
solidication are high at the initial stages of solidication, as a result of the good surface conformity between the liquid core and the
solidied shell. The mold expands while solidication progresses
due to the absorption of heat and the solid metal shrinks during
cooling. As a consequence, a gap develops because pressure becomes insufcient to guarantee a conforming contact between
the surfaces. Once the air gaps forms, the heat transfer across the
interface decreases rapidly and a relatively constant value of hi is
attained.
In the upward vertical solidication the casting weight will contribute to a good metal/mold thermal contact when the lateral contraction is effective, i.e., when the ingot is able to detach from the
lateral walls. This will happen only after a determined solid shell is
formed. In contrast, at the early stages of solidication in the horizontal apparatus the good thermal contact is assured by the liquid
metal pressure exerted over the solid shell. When the solid shell is
able to contract, the air gap is formed and the thermal contact
decreases.
It is a common practice to assume the same interfacial heat
transfer coefcient over the whole casting surface when using
solidication simulation softwares. In order to highlight the importance of using real values of hi according to the gravity vector inuence, three different hi proles were simultaneously applied on the
simulation of solidication of a Sn 5 wt%Pb square casting
(100  100 mm2). The hi prole determined from the downward
solidication was applied over the casting upper surface, the one
from the upward solidication over the casting bottom surface,
and the one from the horizontal solidication over the lateral surfaces. Fig. 14 shows the isotherms shapes, for t = 80 s. It can be seen
that their shapes are not anymore that of concentric circles as
shown previously in Figs. 9 and 11. Higher cooling rates at the lateral faces changed the isotherm format from circle to ellipse
shaped. Similarly, one can conclude that for complex geometries,
which are widely used in industrial applications, there is a need
for a realistic description of these coefcients which are used as input parameters in softwares for simulation and control of industrial casting processes.
In the simulations of Fig. 14a, only heat transfer by conduction
in the melt was assumed. In contrast, if the liquid ow during
solidication is signicant but is not taken into account in the simulations, the accuracy of the calculated isotherms will be reduced.
Fig. 14b shows results of simulations with the same conditions

2800

2800

2100

2100

1400

1400

700

20

40

60

80

100 120 140 160 180 200

20

40

60

80

700
100 120 140 160 180 200

Time (s)

Time (s)

(a)

(b)

Fig. 15. Evolution of metal/coolant interface heat transfer coefcient (hi) as a function of time (t) for AlSi alloys during vertical (a) downward and (b) upward directional
solidication.

N. Cheung et al. / Materials and Design 30 (2009) 35923601

considered previously for the solidication of the Sn 5 wt%Pb alloy,


for t = 80 s, including now the effect of liquid ow during solidication. It can be seen that the ow inside the mushy zone gives rise
to instabilities in the solidication evolution which are responsible
for changes on the isotherms shape and location. As a consequence,
effects on the segregation distribution along the casting are also
expected.
The results obtained for three different hypoeutectic AlSi alloys for solidication carried out both vertically upwards and
downwards are shown in Fig. 15. Fig. 15a (downward solidication) shows constant values of hi along solidication. As the casting
moves away from the chamber surface very rapidly due to the casting weight during downward solidication, the sprouting of interfacial gap is faster than for upward solidication, which causes
lower and constant hi values.
5. Conclusions
The following major conclusions can be derived from the present study:
 When a non-uniform initial melt temperature prole is used as
input data of the IHCP technique in order to derive the corresponding interfacial heat transfer coefcient, a more realistic
simulation of the solidication evolution can be achieved.
 The wettability of the liquid layer in contact with the mold inner
surface, which is associated to the alloys uidity, was shown to
be important in the characterization of the interfacial heat ow.
In this context care should be exercised in the determination of
hi even for small variation of alloy solute content.
 Experimental evidence has shown that hi is strongly dependent
on the direction of solidication with respect to the gravity vector. Accurate simulation of freezing patterns in castings will
depend on the experimental determination of hi for important
growth directions. The uid ow when signicant was also
shown to affect the isotherms shape during solidication, and
has also to be included with accurate hi values for a realistic
description of solidication.
Acknowledgements
The authors acknowledge nancial support provided by FAPESP
(The Scientic Research Foundation of the State of So Paulo, Brazil), CNPq (The Brazilian Research Council) and FAEPEX UNICAMP.
References
[1] zisik MN, Orlande HRB. Inverse heat transfer:
applications. New York: Taylor & Francis; 2000.

fundamentals

and

3601

[2] Cheung N, Santos NS, Quaresma JMV, Dulikravich GS, Garcia A. Interfacial heat
transfer coefcients and solidication of an aluminum alloy in a rotary
continuous caster. Int J Heat Mass Transf 2009;52:4519.
[3] Ferreira IL, Santos CA, Voller V, Garcia A. Analytical, numerical and
experimental analysis of inverse macrosegregation during upward
unidirectional solidication of AlCu alloys. Metall Mater Trans B
2004;35:28597.
[4] Spim Jr JA, Garcia A. Modied network approach for modeling solidication of
complex shaped domains. Numer Heat Transf B 2000;38:7592.
[5] Santos CA, Quaresma JMV, Garcia A. Determination of transient interfacial heat
transfer coefcients in chill mold castings. J Alloys Compd 2001;319:17486.
[6] Campbell J Castings, ButterworthHeinemann Ltd, Oxford, UK, 2003.
[7] Piwonka TS, Woodbury KA, Wiest JM. Modeling casting dimensions: effect of
wax rheology and interfacial heat transfer. Mater Des 2000;21:36572.
[8] Muojekwu CA, Samarasekera IV, Brimacombe JK. Heat transfer and
microstructure during the early stages of metal solidication. Metall Mater
Trans B 1995;26:36182.
[9] Grifths WD. A model of the interfacial heat-transfer coefcient during
unidirectional solidication of an aluminum alloy. Metall Mater Trans B
2000;31:28595.
[10] Browne DJ, OMahoney D. Interface heat transfer in investment casting of
aluminum alloys. Metall Mater Trans A 2001;32:305563.
[11] Hines JA. Determination of interfacial heat-transfer boundary conditions in an
aluminum low pressure permanent mold test casting. Metall Mater Trans B
2004;35:299311.
[12] Prabhu KN, Ravishankar BN. Effect of modication melt treatment on casting/
chill interfacial heat transfer and electrical conductivity of Al-13% Si alloy.
Mater Sci Eng A 2003;360:2938.
[13] Wang W, Qiu HH. Interfacial thermal conductance in rapid contact
solidication process. Int J Heat Mass Transf 2002;45:204353.
[14] Sahin HM, Kocatepe K, Kayikci R, Akar N. Determination of unidirectional heat
transfer coefcient during unsteady-state solidication at metal casting-chill
interface. Energ Convers Manage 2006;47:1934.
[15] Grifths WD. Modelled heat transfer coefcients for Al7wt%Si alloy castings
unidirectionally solidied horizontally and vertically downwards. Mater Sci
Technol 2000;16:25560.
[16] Voller VR, Sundarraj S. A model of inverse segregation: the role of
microporosity. Int J Heat Mass Transf 1995;38:100918.
[17] Ni J, Beckermann C. A volume-averaged two-phase model for transport
phenomena during solidication. Metall Trans B 1991;22:34961.
[18] Siqueira CA, Cheung N, Garcia A. The columnar to equiaxed transition during
solidication of SnPb alloys. J Alloys Compd 2003;351:12634.
[19] Sundman B, Chen Q. STT foundation (Foundation of computational
thermodynamics). Berlin: Springer; 1995.
[20] Ferreira IL, Spinelli JE, Pires JC, Garcia A. The effect of melt temperature prole
on the transient metal/mold heat transfer coefcient during solidication.
Mater Sci Eng A 2005;408:31725.
[21] Spinelli JE, Ferreira IL, Garcia A. Evaluation of heat transfer coefcients during
upward and downward transient directional solidication of AlSi alloys.
Struct Multidiscip Optimiz 2006;31:2418.
[22] Souza EN, Cheung N, Santos CA, Garcia A. The variation of the metal/mold heat
transfer coefcient along the cross section of cylindrical shaped castings.
Inverse Prob Sci Eng 2006;14:46781.
[23] Krishnan M, Sharma DGR. Determination of the interfacial heat transfer
coefcient h in unidirectional heat ow by Becks non linear estimation
procedure. Int Commun Heat Mass Transf 1996;23:20314.
[24] Boeira M, Ferreira IL, Garcia A. Alloy composition and metal/mold heat transfer
efciency affecting inverse segregation and porosity of as-cast AlCu alloys.
Mater Des 2009;30:20908.
[25] Rosa DM, Spinelli JE, Ferreira IL, Garcia A. Cellular/dendritic transition and
microstructure evolution during transient directional solidication of PbSb
alloys. Metall Mater Trans A 2008;39:216174.
[26] Heine RW, Loper CR, Rosenthal PC. Principles of metal casting. New York,
NY: McGraw-Hill Book Co.; 1967.

Anda mungkin juga menyukai