Casting Simulation Methods: TS Prasanna Kumar, Indian Institute of Technology Madras, Chennai, India

5.
12 Casting Simulation Methods

TS Prasanna Kumar, Indian Institute of Technology Madras, Chennai, India
Ó 2014 Elsevier Ltd. All rights reserved.
5.12.1 Introduction 235

5.12.2 A Holistic Approach to Casting Simulation 236
5.12.3 Mathematical Modeling 237
5.12.3.1 Formulation 237
5.12.3.1.1 Modeling Latent Heat Release Through the Source Term 238
5.12.3.1.2 Enthalpy Method 239
5.12.3.1.3 Temperature Recovery Method 239
5.12.3.2 Alloy Solidification Models 240
5.12.3.2.1 Equilibrium Freezing with Diffusion in Both Liquid and Solid States 240
5.12.3.2.2 Nonequilibrium Freezing with Diffusion Only in the Liquid State 240
5.12.3.2.3 Solute Redistribution with Solid-State Diffusion 240
5.12.3.2.4 Mixing in the Liquid 241
5.12.3.2.5 Interface Stability 241
5.12.3.2.6 Modeling of Columnar Structures 241
5.12.3.2.7 Microscopic Modeling of Equiaxed Structures 241
5.12.3.3 Nucleation Kinetics 242
5.12.3.3.1 Instantaneous Nucleation 242
5.12.3.3.2 Continuous Nucleation 242
5.12.3.4 Metal–Mold Interface Heat Transfer 242
5.12.3.5 Numerical Methods 244
5.12.3.5.1 The Finite Element Method 244
5.12.3.5.2 FEM for Nonlinear Problems in Phase Change 247
5.12.3.5.3 FEM for Flow Modeling During Solidification 248
5.12.3.5.4 Advanced Applications 248
5.12.4 Simulation Models 249
5.12.4.1 Deterministic Models 249
5.12.4.2 Stochastic Models 249
5.12.5 Special Topics in Casting Simulation 249
5.12.5.1 Mold Filling 249
5.12.5.2 Mushy Zone Modeling 250
5.12.5.3 Porosity 250
5.12.5.4 Shrinkage Defects 251
5.12.5.5 Micro–Macro Segregation 251
5.12.5.6 Microstructure Modeling 252
5.12.6 Recent Applications and Future Directions 253
Appendix A 253
Finite Element Formulation 254
References 256
5.12.1 Introduction
Metal casting has grown as an art several thousand years back to a technology a few hundred years to a science since a few tens of
years. Rapid strides in understanding the physics of solidification were possible when mathematical modeling became a research
area. In spite of the voluminous work both in modeling and experimental verification, the theoretical models are restricted to
l Highly simplified alloy systems – mostly binary,
l Idealized equilibrium conditions of solidification,
l Alloy phase constitution, and
l Solid/liquid interface equilibrium.
In spite of these restrictions, the engineering applications of solidification simulation have taken the metal casting industry to
high levels of sophistication. The technology components of solidification simulation may be considered to be made of the
following components:
l Pre/postprocessors
l Efficient and robust solvers
Comprehensive Materials Processing, Volume 5 http://dx.doi.org/10.1016/B978-0-08-096532-1.00507-0 235

l Material database
l Quality assessment models (criteria functions)
l Computer technology
Solidification phenomena affect structure and mechanical properties of the castings at multiple scales ranging from the macro
(macrostructure, 103 m), the meso (mushy zone, 104 m), and the micro (microstructure, 106 m) to the nano (growth kinetics,
109 m) levels. A practical approach for considering the multiscale phenomena during solidification could be to identify macro-
and micromodeling and integrate them suitably based on the details required at each stage. The topics under these two broad heads
of modeling with different connotations may be listed as
l Macromodeling
B Conservation of mass, energy, species, and momentum
B Diffusion of species and energy
B Nucleation and growth in the absence of atomic level phenomena
B Mathematical description of macrosegregation and macroshrinkage
l Micromodeling
B Interface dynamics of grain nucleation and growth
B Transformation kinetics
B Dendrite arm spacing (DAS), nucleation rate, etc. based on empirical/experimental models
This chapter tries to summarize the important developments in solidification simulation. A full picture of the varieties of
approaches for modeling the solidification phenomena, mathematical formulations, the tools used, their limitations, etc. is not
possible in a single chapter of limited scope. An attempt is made here to give a cogent account of the subject keeping continuity and
connectivity in mind. In spite of the developments in computer technology, the application of full-scale simulation for industrial
castings may not be possible for a couple more decades (1).
5.12.2 A Holistic Approach to Casting Simulation
Modeling of metal casting processes have evolved over the decades, from being limited to an ideal alloy under controlled laboratory
conditions to include the complexity of industrial processing of commercial alloys. The initial efforts in solidification modeling
were limited to the closed form solution using the Bernoulli equation. This was exploited by Ruddle (2) to the design of gating
systems. Later, these calculations were coupled with empirical and engineering rules used by foundry engineers. With the advent of
digital computers, many computer-aided design programs incorporated these calculations in the computer software. Szekely et al.
(3) applied continuum mechanics for continuous casting. These models were effectively used for turbulent flow for continuous
casting models and gas-stirred ladles.
With the advances in fluid flow modeling techniques like the Marker-and-Cell (MAC) and Solidification Algorithm (SOLA),
researchers started looking at the macro aspects of solidification. Micromodeling for understanding the scientific principles of
solidification was separately developed. Coupled heat transfer and solidification effects for complex geometrics were developed in the
1980s and early 1990s. Simultaneously with the development of mathematical models, physical studies of solidification phenom-
enon were also carried out resulting in important observations, which would influence the modeling strategies. The role of convection
in grain size morphology in ingot casting were first shown by physical studies and modeled as inverse segregation (4). Similarly,
columnar to equiaxed transition, influence of oscillation, rotation, etc. were first observed during experiments and then explained by
modeling. The role of forced convection in producing thixocast metals with round grains was also first obtained by experiments.
The possible applications of process modeling for casting simulation and their theoretical basis include mainly the following (7):
l Casting soundness
B Macroshrinkage Criteria functions
B Volume deficit Mass and energy transport
B Microshrinkage/porosity Criteria functions/pressure maps
B Pore formation Pressure balance
l Casting composition
B Macrosegregation (macromodels) Convective-diffusive energy and mass transport
B Microsegregation (micromodels) Convective-diffusive energy and mass transport
l Surface quality
B Penetration index Pressure balance at metal–mold interface and interfacial reactions
(Continued)
Casting Simulation Methods 237
l Casting distortions
B Stress map Residual stress
l Casting microstructure
B Phase fractions Criteria functions
B Phase transition Criteria functions
l Mechanical properties of castings
B Macro/microscale composition Criteria functions based on composition, cooling rate, and microstructures
l Mold filling
B Filling time, velocity fields Convective mass transport
l Misruns, coldshuts Criteria functions, convective-diffusive energy, and mass transport
The understanding and analysis of transport phenomena is at the core of solidification simulation. Transport phenomenon
refers to the transport of heat, mass, and momentum, all three of which are responsible for the production of sound metal casting.
For an appreciation of the subject with specific reference to material processing and solidification modeling, readers are referred
to Ref. (5).
The governing equations for transport phenomena are obtained by applying conservation principles (6) but solutions are
dependent on robust algorithms particularly for solidification modeling as the material nonlinearities, thermophysical data,
microphenomena, etc. affect the solution procedure. Techniques like finite element method (FEM), finite difference method (FDM),
and boundary element method (BEM) and more recently meshless methods have been established for solution of these equations
under various conditions.
It is the mathematical description of the solidification process, which is a challenge for the researchers. Solidification of an alloy
involves phase change with attendant release of latent heat (LH), nonlinear properties, unobservable phenomena in microscale,
unquantifiable material behavior near solidus temperature, and multicomponent alloys that require ab initio thermodynamic
quantifications, all of which are special areas by themselves. This makes solidification simulation difficult for practical use even with
the accumulated knowledge over the decades.
A typical casting process can be described in terms of modeling complexity at two levels: (1) the mold filling stage and (2) the
phase change stage. The mold filling stage involves the flow of molten metal and the coupled phenomena of energy, mass, and
momentum transport. The computational fluid dynamics (CFD) principles have been successfully used for modeling flow of
various types in the mold cavity. The variety of metal casting methods, however, require different formulations of the same basic
equations and solution strategies for each one of the processes such as gravity-assisted castings, low-pressure die castings, high-
pressure die castings, use of centrifugal force, etc.
In solidification simulation, the main concerns are the prediction of temperatures, species concentration, and fluid
velocity within the solidifying casting. These quantities are obtained by solving equations derived from the conservation
laws of heat, mass, and momentum. Along with the conservation laws, the flux laws are also used. The initial and
boundary conditions differentiate the processes. The mathematical model is formulated with initial and boundary
conditions with the domain descriptions characterizing the physical process. These equations have to be solved
numerically.
The challenge in solidification modeling lies in handling of different length scales involved in the process. The micro- and
macroscale phenomena require different approaches. The microscale solute redistribution manifests in the macroscale as macro-
segregation. Excellent accounts of the several microsegregation equations starting with Scheil in 1942 to Stefanescu in 1993 can be
seen in Ref. (7). These developments spanning five decades, take into account the actual diffusion, interface morphology, etc. to
different levels of sophistication. The morphological aspects at the interface can give rise to different microstructures driven by
thermal, solutal, and surface energies. They require adequate models for explaining the microprocess that can be embedded in the
solution to the macroscale.
5.12.3 Mathematical Modeling

5.12.3.1 Formulation
Mathematical modeling of casting solidification involves combining fluid flow and heat transfer with phase change. The initial heat
transfer during mold filling takes place due to the hot liquid metal coming in contact with the mold at lower temperature. A
mathematical description of the heat transfer during this initial phase involves computation of the flow velocities and heat transfer.
The flow equations obtained by momentum balance will have to be solved along with the energy balance equations simultaneously
to obtain the thermal and flow fields.
The following set of equations in two-dimensional (2D) Cartesian coordinates can be written for the initial mold filling stage.
Momentum equations

vu vu vu v vu v vu vp
r ¼ r u r v þ m þ m þ ðr r0 Þgx [1]
vt vx vy vx vx vy vy vx

vv vv vv v vv v vv vp
r ¼ r u r v þ m þ m þ ðr r0 Þgy [2]
vt vx vy vx vx vy vy vy
Energy equation

v v v v vT v vT
ðrcTÞ ¼ ðurcTÞ ðvrcTÞþ k þ k þQ [3]
vt vx vy vx vx vy vy
Continuity equation
vu vv
þ ¼0 [4]
vx vy
with the initial condition
T ¼ Ti at t ¼ 0
and the boundary conditions
u¼b u ; v ¼ bv ; p ¼ b
p and T ¼ T b on GD t > 0

vT vT
k nx k ny ¼ q on G1
vx vy

vT vT
k nx k ny ¼ hðT TN Þ on G2
vx vy

vT vT
k nx k ny ¼ lε T 4 TN
4 on G3
vx vy
By the time the mold is filled with liquid metal, therefore, a nonuniform thermal field both in the mold and in the liquid metal
would have been set up, due to the initial transients during mold filling. The subsequent heat transfer from the liquid metal, which
has now filled the cavity to the surrounding mold wall, would be dictated by the energy balance equation. However, some natural
convection effects might add to the overall heat transfer in the liquid metal; the degree of its influence on the solidification
phenomena, particularly on the metallurgical aspects like morphology, etc., depends on the process. For most cases, however, the
convection effects may be neglected and the heat transfer from the moment of filling the mold can be largely defined by the same
equations governing heat transfer in solids. The equations are taken to be equally applicable for both the liquid metal and the mold.
The system of equations (eqns [1]–[4]) now reduces to

v v vT v vT
ðrcTÞ ¼ k þ k þQ [5]
vt vx vx vy vy
with appropriate initial and boundary conditions.
The heat transfer process during casting solidification is accompanied by the release of LH of fusion at the solid/liquid interfaces.
Several methods have been devised to take the release of this LH into account. These methods are generally divided into fixed and
moving mesh methods. Fixed mesh methods involve the solution of a continuous system with an implicit representation of the
phase change, while in the moving mesh or the front tracking methods, the solid/liquid regions are treated separately and the phase
change interface is explicitly determined as a moving boundary. The fixed grid solutions offer a more general solution as they
account for the phase change implicitly without attempting to establish the position of the front.
An extension of the fluid flow model to continuous casting equipped with electromagnetic stirring requires modeling of fluid
flow under an electromagnetic field. Unlike the flow phenomena encountered in mold filling, which require tracking the free
surface, the flow simulation in the continuous casting molds under electromagnetic stirring require turbulent flow model. Further
discussion on flow modeling is restricted to shaped castings.
5.12.3.1.1 Modeling Latent Heat Release Through the Source Term

In what can be seen as the most general and straightforward treatment of LH, the source term in eqn [5] is written as
vfs vfs vT
Q ¼ rL ¼ rL [6]
vt vT vt
where L is the LH of freezing and fs is the fraction solid. We can now write the energy balance equation as

vT v vT v vT vfs vT
rc ¼ k þ k þ rL [7]
vt vx vx vy vy vT vt

vfs vT v vT v vT
r cL ¼ k þ k [8]
vT vt vx vx vy vy
It is observed that the term in parenthesis on the left-hand side of the equation can be treated as an ‘equivalent specific heat,’
when compared with the original equation. This formulation of LH liberation during phase change is quite generic and is capable of
accounting for the several alloy solidification models, particularly for long freezing range alloys.
5.12.3.1.2 Enthalpy Method

In a slight variation of the above treatment of LH, the net enthalpy of the alloy at any temperature is written as
ZT
H¼ rcdT þ ð1 fs ÞrL [9]
T0
Differentiating the above equation with respect to T, we obtain

vH vfs
¼r cL [10]
vT vT
The energy equation (eqn [5]) becomes

vH v vT v vT
¼ k þ k [11]
vt vx vx vy vy
The numerical solution of the above equation is explained in a later section. This method assumes that a continuous function for
H over the entire computational domain is available.
5.12.3.1.3 Temperature Recovery Method

The temperature recovery method consists of two steps. The steps involved in the temperature recovery method are available in
Ref. (8). Briefly, in step 1, the energy equation is solved without considering the LH effect, to obtain an intermediate temperature.
Then, in step 2, the actual temperature is calculated by considering the effect of LH liberated during phase change. The source term
for these two steps can be written as
0
Zt1 ZT2
vT
Step 1 : Q1 ¼ rc dt ¼ rc dT [12]
vt
t1 T1
Zt1 ZT2 ZT2

vT vfs
Step 2 : Q2 ¼ rc dt þ DQ ¼ rc dT rL dT [13]
vt vT
t1 T1 T1
Depending upon the initial and intermediate temperatures, three cases can be realized as shown in Figure 1 above. Corre-
sponding to each of these cases, the corrected temperatures can be obtained by the equations.
0
ZT2 ZT2 ZT2
vfs
Case ðaÞ: rc dT rL dT ¼ rc dT [14a]
vT
T1 T1 T1
T1
Tl 2
T1 2 T2
T2
1
T2
1 T2 T1 2
TS
T2
1
t T2
t1 t2 t1 t2 t1 t2
(a) (b) (c)
Figure 1 Illustration of the temperature recovery algorithm. Reproduced from Arunkumar, T. S.; Prasanna Kumar, T. S. “Numerical Study of Freezing of
Water using Taylor-Galerkin based Non-Linear Temperature Recovery Method”. In 9th AIAA/ASME Joint Thermophysics and Heat Transfer Conference, San
Francisco, California, 5–8 June 2006.
0
ZT2 ZT2 ZT2
vfs
Case ðbÞ: rc dT rL dT ¼ rc dT [14b]
vT
T1 T1 T1
0
ZT2 ZTsol ZT2
vfs
Case ðcÞ: rCc dT rL dT ¼ rc dT [14c]
vT
T1 T1 T1
The two steps are used to find the temperature of node in a solidifying element till the solid fraction fs ¼ 1.
If the specific heat in phase change region is a function of temperature, the integration of the terms in the above integrals is not
straightforward. Hence, an iterative technique is employed to obtain the fraction solid (9).
5.12.3.2 Alloy Solidification Models

Alloy solidification is characterized by solute redistribution, which results in coring and variation in mechanical properties
over a microscopic scale. It has a predominant effect on freezing characteristics and the microstructure of the cast material. The binary
alloy has been extensively studied because with reasonable assumptions it is possible to predict the solute redistribution analytically (10).
5.12.3.2.1 Equilibrium Freezing with Diffusion in Both Liquid and Solid States
Cs
The equilibrium partition ratio, k0, is defined as the ratio of equilibrium solid and liquid compositions, , where Cs is the
Cl
concentration of the solid phase and Cl is the concentration of liquid phase in equilibrium at the interface in a binary phase
diagram. Equilibrium solidification requires very slow growth rates with L2 Ds t, where L is the length of the growing crystal, Ds is
the diffusion coefficient of the solute in the solid, and t is the time. It assumes complete diffusion in both the solid and liquid
phases. For constant k0 less than 1, a simple mass balance at the interface leads to the equation
!
1 Tl T
fs ¼ [15]
1 k0 Tf T
where Tf is the freezing temperature of the pure metal and T the temperature at any instant of time. The fraction solid is thus
obtained as a function of temperature. Since the temperature is obtained by solving the heat conduction equation, this formula for
fraction solid gives us a means for computing the increment in fraction solid at any location for every time step.
5.12.3.2.2 Nonequilibrium Freezing with Diffusion Only in the Liquid State

A case of much more practical interest than the foregoing is the one where there is no solid-state diffusion. In such a case, as the
solidification progresses, the liquid becomes richer in solute and so, the solid that forms is of higher solute content. The compo-
sition of the solid formed, however, remains unchanged. Considering a one-dimensional (1D) model of a binary alloy initially at
a uniform temperature above the liquidus, the following equation may be derived for the fraction solid as a function of temperature
! 1
k 1
Tf T 0
fs ¼ 1 [16]
Tf Tl
This equation is derived from the Scheil equation, considering the solidus and liquidus as straight lines.
5.12.3.2.3 Solute Redistribution with Solid-State Diffusion

The Scheil equation frequently gives a reasonable approximation to the real situation, both during unidirectional solidification and
cellular/dendritic solidification in mushy zone. In the absence of complicating factors, the freezing behavior is expected to lie
somewhere between the above two extreme cases, i.e., equilibrium (lever rule) and the nonequilibrium (Scheil equation) condi-
tions, depending on the importance of solid-state diffusion. The effect of solid-state diffusion is small in many of the cases of
interest. However, the above models behave incorrectly, when the solid-state diffusion is quite predominant.
The possible contributions for solute redistribution can come from dendrite tip undercooling, extensive arm coarsening,
dendrite geometry effects, incomplete mixing in the liquid, solute exchange from outside the element, etc.
Such factors are important for substitutional solute types. Solid-state diffusion becomes far more important than the cases listed
above for interstitial solute types like carbon in steels and cast irons. The Brody and Flemings model (11) is an improvement on the
Scheil equation, which considers finite solid-state diffusion also. This is intermediate between the Scheil and lever rules. The model
is based on the 1D solute redistribution model.
Again, like in the previous models, the liquidus and the solidus are assumed straight lines so that the partition coefficient is
a constant. Basing the analysis on the conservation of solute within the element, Brody and Flemings (11) derived the equation
" ! 1 #
k 1
Tf T 0 2Ds
fs ¼ ð1 þ ak0 Þ 1 ; a¼ [17]
Tf Tl vl
For the case of primary interest in which we are treating a small volume element in the mushy zone, the interface advance
L l
velocity is not known a priori. We can make the simplest assumption of constant velocity, which gives v ¼ ¼ . This assumption
tf 2tf
redefines in terms of local freezing time and DAS as
4Ds tf
a¼ [18]
l2
In practice, cells and dendrites tend to grow in such a manner that the velocity decreases in a parabolic manner with increasing
time. Any such equation predicts the changing composition at the interface during solidification and the variation of the local liquid
fraction with local temperature. Thus, a numerical analysis based on these equations allows prediction of solute distribution within
the element at any temperature below the liquidus.
A number of critical observations have been made on this model. The major deficiency in the model is that the solute is not
conserved in the analysis. Large values of a (>2, in low carbon steels) lead to results that are physically impossible, predicting that
the last parts to solidify will have composition less than C0. The model therefore suggests that a limiting value of a could exist, at
a fraction solid equal to 1, which gives the limiting value of a as
" #
1 1
aL ¼ 1 [19]
k0 1 ðk0 Þk011
The value of aL decreases form 1 to 0 as k0 increases from 0 to N.

Apart from the above, Clyne and Kurz, Ohnaka, Kobayashi, Nastac and Stefanescu (12–15) have developed models with
increasing sophistication.
5.12.3.2.4 Mixing in the Liquid

In the derivation of normal freezing equation, the assumption of uniform liquid composition due to convection was employed.
However, even in turbulent flows, the boundary layer inhibits uniform liquid composition and a solute buildup takes place in the
boundary layer. Mathematical analysis of this solute buildup at the solidification front leads us to the effective distribution coef-
ficient, ke, given by
k0
ke ¼ [20]
k0 þ ð1 k0 ÞeRd=D
Rd
which varies from a minimum value of k0 and a maximum of 1 with increasing .
D
5.12.3.2.5 Interface Stability
During solidification, a solute-rich boundary layer builds up in front of the interface, giving rise to constitutional undercooling, which
results in instability of the plane front since any protuberance forming on the interface would find itself in supercooled liquid and
therefore would be stable. A mathematical analysis of the stability of interface leads us to the equation for the critical growth rate given by

mR 1 k0
Gcr ¼ ðCs Þi [21]
D k0
This equation predicts the stability of planar interfaces during alloy solidification. Many controlled experiments have been
carried out that have shown that the planar–dendritic transition depends upon the composition of the alloy. Based on this analysis,
the following three modes of interface may be identified:
G Gcr: planar interface.
G slightly less than Gcr: cellular interface.
G Gcr : dendritic interface.
5.12.3.2.6 Modeling of Columnar Structures

The macroscopic heat flow analysis can be reasonably applied to columnar solidification because the growth rate of the micro-
structure (eutectic front or dendrite tip) is more or less equal to the speed at which the corresponding isotherms move. Therefore,
microstructural parameters and undercooling can be directly calculated from the temperature field in this case.
The criterion for obtaining a fully columnar structure can be defined by

1=3 DTN 3
G > AN0 1 DTE [22]
DTC
5.12.3.2.7 Microscopic Modeling of Equiaxed Structures

For binary alloys, macromodels based on phase diagrams are generally used to solve solidification problems. These models can give
only rough predictions of the solidification time, isotherms, etc., which do not have any direct relation with the microstructures and
physical properties of the solidified alloys. Using models that consider the formation of microstructures, we can obtain more
information than by employing macromodels. Besides a more accurate solidification time and cooling curve, the models can also
obtain data on undercooling, number of grains, grain size, nucleation, growth rates, etc.
Macro–micro models differ from macromodels in the way they handle LH. Before solidification, there is no difference between
these models. Once the temperature falls below the melting point, the methods used to compute the source term, fs, are not the
same. Considering spherical nuclei and their growth, the following equation may be derived for rate of solidification:
dfs ðtÞ dRðtÞ
¼ 4pR2 ðtÞNðtÞ fi ðtÞ [23]
dt dt
To predict the evolution of the solid fraction fs(t), one must therefore relate three variables N(t), R(t), and fi(t) to the under-
cooling DT. This can be done by considering nucleation kinetics, growth kinetics, and solute diffusion.
5.12.3.3 Nucleation Kinetics

5.12.3.3.1 Instantaneous Nucleation
In this model, nucleation is assumed to occur instantaneously, which means N(t) is a constant. Hence, the equation for solid
fraction becomes
dfs ðtÞ dRðtÞ
¼ 4pR2 ðtÞN [24]
dt dt
For modeling the growth step, we use a model of the type
dR
¼ mðDTÞ2 [25]
dt
where m is the growth constant and DT is the undercooling, the difference between the local temperature, T, and the eutectic
temperature, Te. Upon integrating, the growth term yields the expression for radius of the grain as a function of temperature:

1
R ¼ R0 þ m Te2 t Te T 2 þ T 3 [26]
3
where R0 is the critical radius. It is assumed that nucleation is instantaneous at time equal to zero with radii equal to the critical
radius. Taking the impingement factor as (1 fs), when growth stops, we get the equation for solid fraction as

4
fs ¼ 1 exp pR3 N [27]
3
5.12.3.3.2 Continuous Nucleation

The basic assumption used in the continuous nucleation model is that nucleation occurs continuously over a period of time. Once
the nucleation starts, a Gaussian distribution of nucleation rate is assumed. The nucleation rate is written as
" #
dN Nmax ðDT DT0 Þ2 dT
¼ pffiffiffiffiffiffiffiffiffiffiffi exp [28]
dt 2DTs 2DTs dt
where DT0 is the undercooling at the middle of the nucleation rate curve (corresponding to the maximum rate), DTs is the standard
deviation of the distribution, and Nmax is the total grain density. The main difference between the instantaneous nucleation model
and the continuous nucleation model is in computing the number of nuclei, N(t). The growth models and the correction for
dT
impingement are identical. When > 0, i.e., when recalescence occurs, dN ¼ 0, marking the end of the nucleation step.
dt
5.12.3.4 Metal–Mold Interface Heat Transfer

The quantification of heat flow from casting has not matched with the developments in modeling phase change, micro–macro
segregation, pore nucleation and growth, microstructure formation, etc. The inaccuracies in the description of heat flow at metal–
mold interface may not cause large error in the analysis of sand castings but in metallic molds where significant air gap forms, it will
have pronounced effect on the computations originating from an inaccurate thermal field. The air gap that forms at the initial stages
will have a bearing on heat conduction during later stages of solidification.
The major factors that affect the interface heat transfer can be listed as
l Time
l Casting alloy
l Thermomechanical behavior of skin
l Thermophysical properties of mold material
l Casting/mold configuration
l Gravity-assisted/desisted separation
l Local curvatures, corner effects

l External pressure
l Casting/mold volume ratio
First principles approach to modeling interface heat transfer will have to take into account the nature of solid–solid contact in
terms of pressure applied and surface irregularities. At this stage, heat transfer is mainly through conduction at the points of contact
between the just solidified skin and the mold surface. As solidification proceeds, the solidified shell shrinks away from the mold
creating an air gap. At this stage, the heat transfer is predominantly by convection and radiation.
Models based on the inverses heat conduction problem, IHCP (16,17) provide a powerful technique for quantifying the metal–
mold interface heat transfer, as it circumvents the need to model the boundary heat flow phenomena. The IHCP algorithm is based
on measuring the thermal history inside the mold wall during solidification and solves the model equations for unknown boundary
conditions. Irrespective of the complexity of heat transfer mechanism between solidifying shell and mold, the solution to the IHCP
gives the rate of heat transfer from the metal to the mold.
Mathematical modeling of interface heat transfer during casting is a well-researched subject. The methods adopted by the
researchers for the assessment of interface heat transfer were mostly based on the solution to IHCP in general with different
implementation strategies. Table 1 summarizes some studies in the area of metal–mold interface heat transfer under varying
conditions of mold/casting configuration, alloys, external pressure, etc.
The IHCP theory was first developed with single known temperature history, which was limited to the estimation of single heat
flux. The principle was then extended to multiple sensors, which led to the simultaneous estimation of multiple heat fluxes (17).
Tseng and Zhao (35) developed the direct sensitivity coefficient algorithm in the context of FEM for estimating multiple heat flux
components at the boundary, which required simultaneous estimation of the multiple boundaries.
One of the initial attempts at characterizing the interface heat transfer in die casting of aluminum alloys can be found in Pra-
sanna Kumar and Narayan Prabhu (36), who developed a finite element-based code for inverse solution of the heat conduction
equation for estimating the transient heat flux during chill casting of aluminum base alloys, considered as an average quantity over
the entire interface. The algorithm was extended by Prasanna Kumar (37) to cases where multiple transient heat sources at the
boundary could be determined by solving for the heat sources in a serial fashion. This algorithm was then applied to aluminum
casting in metallic molds. A brief outline of the formulation is given below.
Table 1 Some studies in modeling metal–mold interface heat transfer characteristics
Sl. no. Casting–mold configuration Alloy system(s) Thermal condition Comments
1 Cylindrical castings; various Various aluminum One dimensional Heat transfer coefficient (HTC) at the metal–mold
chill materials alloys interface estimated using measured mold
temperatures, as function of time and casting
parameters (18–21)
2 Cylindrical casting; insulating Copper-base alloy One dimensional HTC was calculated by whole domain method for
sleeve the inverse solution to the heat conduction
differential equation with phase change (22)
3 Cylindrical casting; Cast iron One dimensional Heat flux transients showed a double peak during
sand/ceramic molds solidification of the metal due to moisture in the
mold (23)
4 Hollow cylindrical Ti alloy Two dimensional, HTCs at the core and the outer surfaces were
permanent mold axisymmetric determined as functions of casting surface
temperature (24)
5 Cylindrical permanent mold; Aluminum alloys One dimensional, axisymmetric The effect of mold coating and preheating
refractory at bottom and top in the radial direction temperature on HTC was investigated (25)
6 Cylindrical stainless steel– Pure aluminum One dimensional HTC obtained by parameter estimation considering
coated mold; water-cooled radiation and conduction in the gap (26)
copper chill
7 Solid shell formed by dipping Aluminum silicon One dimensional HTC and heat flux transients during early stages of
water-cooled chills into alloys solidification were modeled using the inverse
melts algorithm (27)
8 Squeeze casting Aluminum alloys NA The effect of pressure on heat transfer at the
metal–mold interface under pressure (28–31)
9 Cylindrical castings; copper, Pure aluminum One dimensional (radial) Correlations were developed for HTC–air gap
graphite, and sand molds and its alloys relationships (32)
10 Square bar and rectangular Aluminum – 3% Cu Two-dimensional Cartesian Estimation of multiple heat flux components at the
plate castings in cast iron 4.5% Si alloy coordinates metal–mold interface in corners (33)
molds
11 Vertical wall of a Aluminum alloys Two-dimensional Cartesian Estimation of multiple heat fluxes during mold
metallic mold coordinates filling (34)
Reproduced from Prasanna Kumar, T. S.; Kamath, H. C. Estimation of Multiple Heat-Flux Components at the Metal/Mold Interface in Bar and Plate Aluminum Castings. Metall.
Mater. Trans. 2004, 35, 3.
The basic equation for the IHCP is the set of equations obtained for the transient heat transfer within the mold, viz.,

v vT v v vT
l þ l ¼ rc [29]
vx vx vy vy vt
with the initial condition T(x,y) ¼ T 0 , t ¼ 0, and the boundary conditions
vT vT
l nx l ny ¼ qk ðx; y; tÞ on Gk ; k ¼ 1; 2; .p.l
vx vy
where the right-hand side of the boundary condition are the unknown independent flux boundaries. The IHCP is to find (qp)m;
assuming (qk)i, k ¼ 1, ., l; i ¼ 1, ., m 1, and k ¼ 1, ., p 1; i ¼ 1, 2, ., m are known. To estimate the unknown boundary heat
flux (qp)m, temperature measurements at known locations inside the mold are obtained. The objective function for minimization is
chosen as
Xs X r 2
Sk ¼ Yj;mþi1 Tbþ
j;k;mþi1 ; k ¼ 1.p.l [30]
j¼1 i¼1
Minimization of the objective function leads us to

PJ Pr
b j;k;mþi1 fj;k;i
Yj;mþi1 T
j¼1 i¼1
ðDqk Þm ¼ PJ Pr 2 [31]
j¼1 i¼1 fj;k;i
where f is the sensitivity coefficient and is given by

bþ T
T b j;i
j;i
fj;i ¼ [32]
Dqi
The terms in the numerator can be represented by the following notation and are computed by solving direct heat conduction
equation using FEM.
b þ ¼ Tj;i qm;.;mþr1 ¼ qm1 0

T [33]
j;i qm;.;mþr1 ¼ ð1 þ εÞqm1
b j;i ¼ Tj;i qm;.;mþr1 ¼ qm1

T [34]
The application of this algorithm has brought out clearly the nature of heat transfer during solidification at corners in plate
castings, continuous casting molds, variation of the effect of the heat flux along the vertical side of a metallic mold, air gap formation
in metallic molds, etc. Prasanna Kumar and Kamath (33) used the algorithm for simulating the early air gap formation at the corners
of a plate casting. They measured the temperatures inside the corner of a plate mold and computed the spatial distribution of the
interface heat flux. The multiple heat flux model brings out the complex heat flow at the corner because of the earlier air gap
formation at the corner, as shown in Figure 2.
In another work, Arunkumar et al. (34) measured the temperatures inside the vertical wall of a metallic mold at six locations
during pouring of an aluminum alloy and solved for the heat flux distribution along the mold–metal interface by inverse heat
conduction modeling. The resultant simulated thermal field evolution in the mold wall is given in Figure 3(a). The simulation
showed that during the time taken for the mold to fill, viz. about 25 s, a temperature gradient develops along the interface, making
the heat transfer 2D. The interface heat transfer gradually becomes 1D as the solidification progresses. The initial temperature
variation along the metal–mold interface and its influence on the initial stages of solidification are yet to be studied. The heat flux at
the interface as the metal starts filling the mold is shown in Figure 3(b). The IHCP algorithm (37) was shown to delineate such
spatial and temporal distributions of heat fluxes during mold filling.
A straightforward approach for the heat transfer coefficient at the metal–mold interface is still not available.
5.12.3.5 Numerical Methods

The FDM (38) has wide applications in solving the transport equations. FDM offers computational ease but is not considered
suitable for complex geometries. The FEM originally developed for solving structural problems (39) was later applied to heat
transfer problems with phase change by Ref. (40). The BEM is based on integral formulations and weighted residual techniques have
been applied to phase change problems as well (8). The control volume or the finite volume method (FVM) has been extensively
used for the coupled fluid flow and heat transfer problems (41).
5.12.3.5.1 The Finite Element Method

The principles of numerical methods are extensively covered in the literature and no attempt is made here to elaborate them.
The principles of FEM are briefed here with reference to the transient, 2D energy balance equation (without the convective and
source terms).
Figure 2 Mold interface showing identification of heat flux components with corresponding boundary segments in plate casting (left) and heat
flux distribution at the metal–mold interface for uncoated molds (right). Reproduced from Prasanna Kumar, T. S.; Kamath, H. C. Estimation of Multiple Heat-
Flux Components at the Metal/Mold Interface in Bar and Plate Aluminum Castings. Metal. Mat. Trans. 2004, 35B, 3 (Figures 8(c) and 11(c)).

v vT v vT vT
k þ k ¼ rc in U [35]
vx vx vy vy vt
with the initial condition
T ¼ Ti ðx; yÞ at t ¼ 0
T ¼ Tðx; yÞ on G1 ; t > 0

vT vT
k nx k ny ¼ qðx; yÞ on G2
vx vy

vT vT
k nx k ny ¼ hðT TN Þ on G3
vx vy
where k, p, and c are the thermal conductivity, density, and specific heat of the material, which could be functions of temperature T.
All the types of boundary conditions may or may not be present in a single process.
Regardless of the physical nature of the problem, a standard FEM primarily involves the following steps.
l Constructing approximate functions for the elements
l Obtaining the integral or the weak form of the partial differential equation (PDE)
l Domain discretization
l Obtaining element matrices
l Assembly of element equations
l Introduction of boundary conditions
l Solution of the final set of simultaneous equations
l Interpretation of the results
The distribution of the dependent variable, T, within an arbitrary element is assumed to be a linear combination of polynomials
of the form
X
r
T e ðx; y; tÞ ¼ ji ðx; yÞTi ðtÞ [36]
i¼1
Figure 3 (a) Thermal histories in the vertical wall of a metallic mold during filling of aluminum alloy. Spatial Variation of Heat Flux at the Metal–mold
Interface Due to Mold Filling Effects in Gravity Die–casting. Reproduced from Arunkumar, S.; Prasanna Kumar, T. S. Numerical Study of Freezing of
Water using Taylor–Galerkin based Non–Linear Temperature Recovery Method, 9th AIAA/ASME Joint Thermophysics and Heat Transfer Conference 5–8
June 2006, San Francisco, CA.
where r is the number of nodes assigned to element e and the Ti are the nodal temperatures. ji are the interpolation or shape
functions, the form of which is governed by the order of the element.
The residual obtained after applying the approximate function to a single element domain is weighted by an arbitrary function,
integrated over the element and set to zero. Mathematically, we get the minimized residual function as
Z
v vT e v vT e vT e
x k þ k rc dU ¼ 0 [37]
vx vx vy vy vt
Ue
After integrating by parts the first two terms and regrouping, we get
Z I
vT e vx vT e vx vT e vT e vT e
k k xrc dU þ x k nx þ k ny ds ¼ 0 [38]
vx vx vy vy vt G vx vy
Ue
The boundary G, which is made up of G2,3 depending on the problem formulation, can be split into its component terms giving
I I I I
vT e vT e
x k nx þ k ny ds ¼ xq ds xhðT e Þ ds þ xhðTN Þ ds [39]
G vx vy G2 G3 G3
Using the approximate function for temperature and the shape function for the weights, the element equation in matrix form
results
e
vT
½Ce þ ½Ke fTge ¼ fFge [40]
vt
where the matrix terms are given by
Z
Cij ¼ rcji jj dU [41]
Ue
Z I
vji vjj vj vjj
Kij ¼ k þk i dU þ hji jj dG [42]
vx vx vy vy G3
Ue
I I
Fi ¼ qji dG þ hTN ji dG [43]
G2 G3
The element equations are assembled to get the global equations and the resulting set of first-order ordinary differential
equations is rendered into a set of linear algebraic equations by applying weighted average approximations to the quantities. The
interpolation is linear, defined by
fTgnþ1 fTgn
q T_ nþ1 þ ð1 qÞ T_ n ¼ [44]
Dt
where the subscript n refers to the time step number. The choice of the value of q leads us to the various difference schemes. With
q ¼ 0 and 1, we have the forward and backward difference schemes, respectively, both of which are conditionally stable. With q ¼ 1/2
and 2/3, we have the Crank–Nicholson and the Galerkin methods, respectively, both of which are unconditionally stable, and
largely used in transient heat transfer analysis.
The final matrix equation will be of the form
½AfTgnþ1 ¼ fBg [45]
where
½A ¼ ð½C þ qDt½KÞ [46]
fBg ¼ ð½C ð1 qÞDt½KÞfTgn þ DtfFgn [47]
which is solved for each time step using a suitable solver. The readers are referred to standard texts on FE algorithms for a detailed
account (42,43,108).
5.12.3.5.2 FEM for Nonlinear Problems in Phase Change

The mathematical formulations for handling the LH liberation dealt with in Section 5.12.3.1 applies to constant thermophysical
properties. The development of the global matrix equations in the form shown in the previous section requires that the stiffness and
capacitance matrices are linear (constant properties). This is not the case in solidification simulation, particularly when it comes to
LH liberation, which depends on the rate of solidification. All the alloy solidification models discussed previously are nonlinear
except the linear model, which is of no practical significance.
Swaminathan and Voller (44) developed an iterative scheme for handling the LH liberation for temperature-dependent
properties, which is known as the implicit enthalpy method (see Appendix A for detailed formulation). The formulation leads
to an iterative scheme within the time step obtained by the Taylor expansion of the enthalpy term in equation 9. The heat
conduction equation 5 is considered here for accounting the LH through the source term, which is rewritten in terms of
enthalpy as

v vT v vT vH
k þ k ¼ [48]
vx vx vy vy vt
on the same lines as in Section 5.12.3.1.2. Following the Galerkin method of discretization and implicit formulation, the above
equation leads to the standard form of the system of equations given by
mþ1
vH
½KfTgmþ1 þ ½C ¼ fFg [49]
vt
The unknown current iteration of the total enthalpy is replaced by its Taylor expansion
dHnm mþ1
Hnmþ1 ¼ Hnm þ Tn Tnm [50]
dT
An iterative formulation within each time step is thus obtained for computing the temperature given by

1 dHj m 1 dHi m
½C þ ½K fTgmþ1 ¼ ½C Him Hin þ fTgm [51]
Dt dT Dt dT
where i is the row index, j is the column index, m is the iterative step number in the current time step, and n is the previous time step.
5.12.3.5.3 FEM for Flow Modeling During Solidification

The FEM is known to be effective in solving nonlinear problems in conduction heat transfer. However, for nonlinear system of
equations arising out of fluid flow, FDM/FVM is preferred. When it comes to FEM in fluid flow modeling, spurious oscillation in the
solution occurs. There is a great body of literature dealing with the fundamentals of solving flow problems by FEM through various
stabilization schemes (41,45,46). The convective flows in the phase change regime is lamellar and hence a coupled analysis of
interdendritic, buoyancy/solute gradient-driven flows and heat transfer may be within the scope of easily implementable FEM code.
For solidification simulations, a combination of FDM/FVM for initial mold filling and the FEM for phase change (and thermal
stresses) analysis may be a good proposition.
Arunkumar and Prasanna Kumar (2004) implemented the Taylor–Galerkin-based split algorithm (47) in the FE code TmmFE for
coupled flow and thermal analysis with nonlinear properties. After validating the code for isothermal and nonisothermal flows by
benchmarking with standard problems in the literature, TmmFE was used for simulating freezing of water in a rectangular cavity of
5 cm length and 1.6 cm height. The circulation reversal due to the anomalous expansion of water simulated by this model is shown
in Figure 4.
5.12.3.5.4 Advanced Applications

Some of the more recent advances in these methods with reference to their application to solidification modeling are briefed here.
1. The understanding of advection-dominated problem using unstructured grids based on operator splitting has been reported
(48). The algorithm was validated against analytical solutions for high Peclet numbers in fully developed pipe flows.
2. Extending the meshless methods using radial basis functions, Guangming Yau et al. (49) compared the numerical results of
simple diffusion equation with Dirichlet boundary conditions. The accuracy and stability of the three meshless methods studied
in the paper were shown to be acceptable.
3. Combining cell-centered FVM and the control volume FEM, a vortex-based cell-centered technique was reported by McBride
et al. (50) claiming the method efficient for mold filling analysis in complex geometries.
4. In an effort to model air gap in the metal–mold interface with uneven mold surfaces, a coupled thermomechanical, thermal
transport, and segregation of aluminum alloys was performed by Deep Samantha and Nicholas Zabaras (51). The heat transfer at
the interface was modeled by them using the actual contact pressure and air gap size obtained by solving a subproblem at the
metal–mold interface.
Figure 4 Coupled analysis for solidification of water by FEM using Taylor–Galerkin algorithm; streamlines (top); thermal field (bottom) at steady state.
Note the reversal of circulation due to its minimum density at 4 C. Reproduced from Arunkumar, T. S.; Prasanna Kumar, T. S. “Numerical Study
of Freezing of Water using Taylor-Galerkin based Non-Linear Temperature Recovery Method”, In 9th AIAA/ASME Joint Thermophysics and Heat Transfer
Conference, San Francisco, California, 5–8 June 2006.
5. In the phase field model, the variable varies over thin interface, which requires an adaptive mesh. Rosam et al. (52) have reported
a fully implicit adaptive time and space distribution for binary alloy solidification in two dimension claiming h-independent
convergence of nonlinear algebraic equations reducing central processing unit (CPU) time increasing efficiency.
6. In a paper on characterizing the metal–mold interface in pressure die casting, an analytical model has been developed for
predicting time variant thermal conductance at the interface (53).
5.12.4 Simulation Models
Solidification is a complex process particularly with reference to industrial castings. None of the models that explain the basics of
thermal, mechanical, and metallurgical aspects involved during solidification considers all the physical phenomena because of the
complexities and hence are not suitable for direct use in industrial applications. These models can be roughly divided into (1)
deterministic and (2) stochastic.
5.12.4.1 Deterministic Models

The deterministic models are based on the solution to the continuum equations. They solve macroscopic heat flow for computing
average microstructural features such as grain size DAS and segregation. The basic assumption is that the molten metal is a New-
tonian fluid and obeys the conservation laws. The models are based on transport equations, described elsewhere.
Analytical methods for solving the equations describing the transport of mass, energy, and momentum have been developed by
Carslaw and Jaeger (54). They suffer from a highly simplified approach to the real-world problem, like constant thermal gradient,
constant liquid feeding velocity, etc. Examples of such applications can be found in Ref. (55).
Numerical methods can handle practically any complex mathematical description of the process. Of the numerical schemes that
have been universally applied (FDM, FEM, and BEM), FDM is generally restricted to simple shapes. FEM originally developed for
stress analysis problems was later applied to heat transfer and phase change also.
The BEM is a combination of integral equation and weighted residual methods. In the BEM, the nodes are defined only on the
external surface. The equations are written satisfying the boundary conditions. From the solution to the equations, the internal
variables are calculated as functions of the boundary values. An example of the coupling of FDM with BEM for casting solidification
can be found in Ref. (8).
The phase field models based on thermodynamics are applicable over small computational domains only. They generate precise
understanding of the microscale processes, particularly of grain growth, dendrite branching, coarsening, etc. Phase field analysis
encompasses detailed investigation of flow in mushy zone, growth of dendrite tips, microsegregation, fragmentation, etc., which
cannot be addressed in segregation models. However, these models cannot be corroborated by experimental verification. The phase
field method has been used to predict complex growth patterns from first principles (56–58,110). A summary can be found in Ref. (1).
5.12.4.2 Stochastic Models

The stochastic models such as Monte Carlo (MC) and cellular automaton (CA) models are computationally efficient and can be
applied to large domains for practical problems. MC models have been applied for the simulation of cast structures (59). The
drawback of MC for solidification simulation is that it does not consider macro- and microtransport. Also, MC uses a pseudo-time
step and hence the process cannot be mapped onto real time. The CA technique has also been applied to solidification modeling
(60). Compared to the MC models, the CA model can quantify the time-dependent microstructure evolution, in which the indi-
vidual grains can be shown graphically as they evolve in shape and size. But the CA algorithm cannot describe the finer aspects of
grain growth such as dendritic branching, segregation, eutectic phase formation, etc. Recently, a modified CA model has been re-
ported (61), in which the quantitative description of dendritic growth, coarsening, branching, etc. has been captured.
5.12.5 Special Topics in Casting Simulation

5.12.5.1 Mold Filling
The nature of flow during mold filling results in casting defects such as gas entrapment, oxide film, misruns, cold shut, etc., making
the analysis of mold filling important in the production of quality castings. The development of these and other models has been
reported at a series of conferences on modeling of castings, welding, and advanced solidification processes. A bench mark problem
was evolved by Cross and Campbell (62). A successful filling algorithm should be able to handle free surface flow and turbulent
flow in the mold filling stage. Of the algorithms that have been used for modeling mold filling, the volume of fluid (VOF) method is
computationally more effective than the MAC method. Although the FEM can generate complex gird system more effectively than
FDM, it is known to be unsuitable for predicting flows. The semi implicit method for pressure-linked equations (SIMPLE) algorithm
coupled with VOF method (63) based on body-fitted coordinate has been reported for modeling the mold filling of thin-walled and
curved castings (64).
For a preliminary discussion of transport phenomena, the reader is referred to Ref. (6). The methodology is briefly described
here. The integral form of the continuity equation and momentum balance equations are discretized using any of the several
methods like MAC, SOLA, or simplified marker-and-cell (SMAC). The discretized equations are combined with MAC or VOF
methods and solved to analyze to free surface during mold filling. A description of scheme can be found in Ref. (5), where the free
surface tracking by MAC and VOF for three-dimensional (3D) flows is considered.
Some of the effective techniques for modeling the free surface are MAC and VOF methods. The MAC algorithm has the following
features: primitive variable and finite difference scheme in a cell structure and use of marker particles. The Eulerian mesh is used for
solving velocity field and the instantaneous positions of markers in the Lagrangian coordinate system. As the marker particles move
with the fluid, the cells that are empty get to fill over a period of time depending on flow field. An improvement in the MAC method
is a simplified MAC or SMAC. As against the MAC, VOF method solves for fractional volume of fluid occupying the cells. The
fraction may vary from 0 (empty cell) to 1 (cell filled with fluid), while anything in between represents free surface. A detailed
account of these algorithms as applicable to mold filling can be found in Ref. (5). The SIMPLE method can be adopted for body-
fitted coordinate systems, which have been developed to mold filling of thin-walled and curved castings (64).
5.12.5.2 Mushy Zone Modeling

Mathematical modeling of mushy zone where the physical nature of the casting evolves is highly complex. These models are so
distinct from the macroscopic models that it is impossible to make justice to both these topics in one single article. There is
tremendous advancement in understanding the physics of the mushy zone due to pioneering works of Brody, Flemings, Clyne,
Stefanescu, and others. These models are closed form equations for interface instability, dendritic tip velocity, growth mechanisms,
coarsening, columnar to equiaxed transition, eutectic solidification, divorced eutectic, equiaxed eutectic, peritectic solidification
monotectic solidification, multicomponent systems, etc. Integration of these models with the commercial software packages have
not been very effective as on date and is expected to take several decades to make them available to the practicing foundrymen. One
such success case that can be cited is the incorporation of the thermodynamic calculations to combine multicomponent pseudo-
binary approach in Ref. (14).
Solidification simulation is characterized by its importance for handling phase change. From the energy transport angle, the LH
during solidification has to be removed from the mushy zone to the ambient through the solidified shell and the mold metal
interface. From the mass transport angle, suitable solute redistribution models will have to be employed for explaining the specific
alloy characteristics. From the momentum balance equations perspective, several phenomena can be occurring simultaneously in
the mushy zone: bulk liquid movement due to thermal and solute gradient, liquid flow due to shrinkage, two phase movement in
the mushy zone, etc. The thermal and physical properties vary with temperature making solidification problem nonlinear.
The numerical methods for solving phase change can be classified into fixed grid, variable grid, and transformed grid methods. In
the fixed grid method, the grid is uniform and the solid–liquid interface is obtained by interpolation using thermal field data, while
the variable grid method tries to track solidification front (moving boundary problem). They are limited to 1D model of pure
substances. The adaptive grid method is based on coordinate transformation. The irregular region separating solid–liquid boundary
is mapped into regular regions by suitable grid techniques (65). The grids are generated at successive time step making the method
highly CPU intensive.
Alloy solidification is modeled from the first principles with respect to the binary equilibrium diagram. How much of this is
valid for multicomponent system is a mute point. Diffusion models based on lever rule, Scheil model, Brody and Flemings are built
into fraction solid estimation at any time step which is then written in the form of heat liberated during the time interval. The
resulting formulation is plugged into the source term of the governing equation. The recovery of LH in the model itself can be
classified into (1) equivalent specific heat method, (2) enthalpy method, and (3) temperature recovery method. The details can be
found in Ref. (5).
5.12.5.3 Porosity
Pellini (66) related the size of riser to the feeding distance through a function and then onward many functions have been proposed
in the literature for porosity formation in various alloy castings. The works of Tynelius et al. (67), Roy et al. (68), and Irani and
Kondic (69) have been summarized by Lee et al. (70). The Niama criterion (71) is one of the best known for predicting the feeding
of castings. However, the Niyama criterion is not applicable to formation of micropores in Al–Si alloys (72).
Porosity development in aluminum alloys has been addressed in a number of models over the past several decades. The
solutions range from analytical solution to highly complex simulations of the stochastic nucleation and growth of pores during
solidification of the alloy. Excellent analytical models are available in Ref. (70). Analytical models suffer from an accurate prediction
of porosity for the complete range of conditions found in industrial-shaped castings. Shrinkage porosity and gas porosity are the
driving forces for pore prediction, resulting in an ideal model being too complex for industrial use. Porosity in aluminum castings
are classified based on the size of pores as macro- and micro-porosity and by the cause of formation (70).
Considering the heat transfer, gas diffusion, and simultaneous growth of gas formation in an ingot, the gas pore nucleation was
modeled by Ludmill Drenchev (73). The paper brings out the relationship between the size of the gas nucleus and the number of
nucleation sites on the solid metal interface. The enthalpy method was modified and successfully applied to simulation experi-
mental results of a bismuth–tin alloy performed on board US space shuttle ‘Columbia.’ The effect of thermal and solute convection
in microgravity and concentration-dependent melting temperature was included in the algorithm (74). In another application of
the modeling technique, the solidification of binary alloy in Bridgman furnace under microgravity conditions was reported by
Timchenko et al. (111).
In a paper by Lee and Gokhale (75), heat transfer simulations were performed to study the effect of the presence of air gap in the
local heat transfer rate, which would affect the local solidification. The authors showed that air in the gas pore acts as heat-insulating
medium and retards heat transfer in the melt leading to shrinkage around the gas pores in the case of magnesium alloys high-
pressure die casting.
Another type of model that has been successfully applied to explain the microfeatures of a solidified casting is the stochastic
models. Fundamentally, the method depends on determining the functional form of the relationship between, say nucleation and
the processing parameters. They also depend on the experimental observations of the process. As an example, Lee and Hunt (76)
observed the hydrogen pores grow from entrapped soluble gas pockets and hence they developed a stochastic model based on the
level of supersaturation.
In a more recent paper, Lee and Hunt (77) combined continuum and stochastic models to simulate and the track nucleation and
growth of hydrogen porosity in directionally solidified Al–Cu alloys. The individual pores were tracked and the models are coupled
with in situ observations where size and distribution of pores were measured as a function of temperature/time, which compared
well. The importance of incorporating the diffusion of hydrogen for accurate modeling of pore formation in aluminum alloys was
brought out.
For a comprehensive account of the criteria functions for porosity estimation, readers are referred to Ref. (7).
5.12.5.4 Shrinkage Defects

In casting solidification simulation, the solidus isotherm will outline a liquid–solid interface. If the amount of gas in the liquid is
within its solubility limit in the solid, only shrinkage cavity will form. The position of these isotherms will depend on heat transfer
and fluid flow. In the presence of gravity, the location of shrinkage cavity will move toward the top of the casting. The shrinkage
cavities resulting from the difference in specific volume of liquid and solid will manifest in the last parts to solidify. The models that
have been proposed to date are in principle, only macromodels.
For predicting shrinkage cavity from heat transfer analysis, the following assumptions are made:
l The evolution of fraction solid is only a function of temperature and is independent of solidification kinetics.
l There is no significant mass transfer by fluid flow between different parts of the casting.
l The mold and the casting are rigid.
With the above assumptions, the macro heat conduction equation is solved by FEM (or FDM) and the path of the isotherms is
calculated throughout the casting. Once the isotherms are known, a number of criteria are used to predict the position of the
shrinkage cavity. The main criteria are as follows:
1. Isotherm tracking: shrinkage is formed in the loop
2. Isofraction-solid tracking. This method can be used under two different assumptions: fs ¼ 1, equivalent to tracking solidus
isotherm, and fs ¼ fcritic < 1. The latter is based on the assumption that decreasing permeability through the mushy zone
increases the pressure required for mass transfer. It might then be possible that at a certain critical fraction solid (<1), mass
feeding might entirely stop.
3. Temperature gradient method: a critical gradient value is assumed; the maximum temperature gradient at fs ¼ 1 or fs ¼ fcritic must
be smaller than this critical gradient to avoid shrinkage cavity formation.
Imafaku and Chijiiwa (78) developed solidification shrinkage model based on the following assumptions: (1) gravity feeding
occurs instantly, (2) liquid metal–free surface is flat, and (3) volume shrinkage cavity is equal to volume of contraction. Macroscopic
fluid flow exists as long as fs > fcritic. If the volume loss is in the surface, it is compensated by lowering the liquid level; if it is inside
the casting, it is compensating by a void. Their model was validated using observed results from ‘T’-shaped castings.
Bounds et al. (79) developed a more comprehensive model on quality features of casting including pipe shrinkage. Using
modified transport equations based on mixture theory and avoiding pore nucleation and growth complexities, their model replaces
the shrinkage volume by gas. They predicted macroporosity based on pressure drop using Chang–Stefanescu model (80). They
could bring out the known behavior of shrinkage defects in short and long freezing range alloys through their models.
5.12.5.5 Micro–Macro Segregation

Macrosegregation is a phenomenon that affects the quality of industrial castings and therefore attracted many research groups
for several decades. It is caused by relative movement of solid–liquid flow during solidification. Of the many causes of
segregation, the feeding of solidification shrinkage, thermal and solutal gradients in the liquid, buoyancy-driven force, flow
during pouring, magnetic stirring, rotation, vibration, movement of equiaxed grains due to heterogeneous nucleation, etc. may
be cited.
Modeling of macrosegregation has focused on the basic flow mechanism involved considering heat transfer and solute transport,
fluid flow, solid movement, and solid deformation. In addition, phase equilibrium, nucleation at microscopic level will have to be
considered. For a detailed review of macrosegregation models, readers are referred to Ref. (1). The initial effort began by Flemings
and coworkers in 1960 who considered interdendritic flow of liquid as flow through a fixed solid dendritic network. They arrived at
local solute redistribution equation basically starting from Scheil’s equation. Their equation can explain the conditions under which
macrosegregation occurs as in the case of Al alloys (inverse segregation). The LSRE equation could explain inverse segregation in
aluminum alloys, negative macrosegregation, positive macrosegregation in steel castings, freckling, etc.
Simultaneously, experiments were performed to measure the permeability of mushy zone. Ridder et al. (81) solved the coupled
set of equations given by Darcy’s law, the energy equation and the LSRE in the mushy zone, and the momentum and energy
equation in the fully liquid region. The single-domain model based on mixture theory (82,83) was based on mass, momentum,
solute segregation, and convection equations that are valid in all the regions of the alloy during solidification.
Fixed grid single-domain numerical models were developed for steel ingot, continuous casting of steel, aluminum direct chill
casting, nickel base superalloy casting, and shape metal casting (1).
5.12.5.6 Microstructure Modeling

The coupling between the macroscopic heat flow equation and the microscopic models of equiaxed solidification can be achieved,
principally by two methods: The LH method and the microenthalpy (ME) method.
The LH method is quite straightforward. The macro heat flow equation is solved by FDM or FEM. The incremental solidification
Dfs between t and t þ Dt at all nodes is calculated according to the microscopic model of solidification, by first calculating the
undercooling at each node. The source term Q in the macro heat flow equation couples the macroscopic heat flow and the
microscopic growth kinetics. The LH evolved is calculated for each element depending on the fraction solidified.
The ME method is also essentially based on the LH method. Since the enthalpy of the system is independent of the solidification
path, the macro heat flow calculations and microsolidification calculations can be decoupled. At the macro level, the heat flow
equation is solved and the change in enthalpy at all nodes is calculated. The solidification path is then computed within the macro
time step, by dividing the macro time step further into smaller intervals.
A comparison of the LH and ME methods demonstrated that the LH method is more accurate. Regardless of the method used for
coupling the macro heat transfer (HT) and micro solidification kinetics (SK) models, the procedure must allow for calculation of
recalescence, which is an important feature of the cooling curve.
One of the key issues that are relevant to the metallurgical solidification phenomena and its impact on the success in the
commercial arena is microstructural modeling. A very fine treatise on the modeling of microstructural evolution during solidifi-
cation can be found in Ref. (84). The underlying physical mechanism of microstructure evolution during alloy solidification is well
known (10). Models based on the basic phenomena of energy, mass, and momentum conservation have been established (7).
Deterministic models to predict the as cast microstructure have been detailed (85–88). The ideal physical models have been
simplified and have been shown to work well for many alloy systems. However, these models have to be tuned with experimental
data on microstructures. The computational space is divided into a series of interconnected volume elements with the heat transfer
taking place across the elements. But the metallurgical phenomenon is restricted within the volume element. The internal state
variable approach has been proved to be a versatile tool for studying nonisothermal microstructural evolution. They encompass
both the coupled thermal and diffusional transformation of microstructure formation during solidification and in the solid state.
However, for the metal casting specialist, the former is more important.
Mullins and Sekarka (89) introduced the approach for handling the constrained and unconstrained grain growth considering the
temperature, solute filed, and interface stability, which was later modified by Kurz and Fisher (10). It is also known that the
convection can influence interface stability as given by microstructural morphology. It is important for any solidification modeling
to be able to predict the cellular and dendritic nature of the solidified grains.
The mathematical formulation is well brought out in Ref. (84). The model system used in the investigation reference provides
a link between thermal analysis and casting. The melt containing fixed number of nucleation sites, which are randomly distributed,
are brought under the influence of thermal gradient, G, within the liquid due to heat diffusion through the wall of the container. The
thermal model is based on the Fourier heat conduction equation as applicable in a liquid with the LH term accounting for phase
change. The chapter considers both the isothermal and nonisothermal behaviors. The model is applied to Al–Si foundry alloys using
boron as grain refiner in the form of silicon boride. The models are shown to capture the behavior of growth of equiaxed structures
under the influence of variable amounts of nucleating agents.
The coupled analysis of microstructure with convection in macrosegregation is still lacking in many respects (1). The solute
rejection process at the microlevel influences macrosegregation. The solute diffusion depends on the dispersion of microstructure.
There has been some literature on the coupled micro- and macrosegregation (90–93), which can deal with multicomponent alloy
where multiple phases are forming.
Segregation models based on transport phenomenon cannot describe the microphenomenon of nucleation and growth.
Recently, CA models have been developed because of their potential to application in understanding the fundamentals of nucle-
ation and growth phenomena. However, for practical applications deterministic and stochastic models are more helpful. A review of
these models can be found in Ref. (5).
Deterministic models couple macroscopic heat flow to solidification phenomenon targeting grain size, dendritic arm spacing,
etc. at a scale of practical metal casting. Stochastic models like MC and CA provide efficient solutions and can be applied to large
domains. The MC procedures take into account diffusion, kinetics of grain formation based on lowest free energy change algorithm
(109). However, the MC method ignores the macro- and microtransport and time step is not correlated with real time. Hence, the
evolution of solidification cannot be related to real-time situations.
The CA technique is also a stochastic method based on nucleation and growth kinetics and crystallographic orientations. In the
CA model, the individual grains are identified and their shape and sizes can be made to describe branching of dendrites, solute
segregation, and eutectic phase formation. In a modified CA model, an account of nucleation and growth kinetics was implemented
into stochastic modeling and the results obtained are explained in Ref. (61).
Microstructure modeling has been extended using new techniques like Voronoi tessel on algorithm (94). Using finite element
mesh for a description of the intergranular space and considering liquid phase as incompressible Newtonian fluid, the permeability
was computed based on Darcy’s law. A good agreement is reported with available literature data. Another example of Voronoï
tessellation can be found in Ref. (95), which describes the development of network of local feeding. The heat transfer during
solidification of metal foams has been modeled using 3D tetrahedral finite element mesh for both the voids and solid phases of the
foam. Model validation is performed against pressure drop and heat transfer coefficients. The Navier–Stoke and energy equations
are solved to obtain the pore structure.
Dendritic solidification was modeled by Stewart and Weinberg (96) and Stoehr (97) who included the effect of natural and
forced convection. Glickman group modeled the solid–liquid interface in transparent systems like succinonitrile. The morpho-
logical instability in mushy zone was modeled by Stewart and Hellawell (98).
5.12.6 Recent Applications and Future Directions
Some of the more recent successful industrial applications of mathematical modeling techniques are mentioned below:
1. Application of 3D heat transfer solidification for continuous casting of wire using fluent package (99).
2. A transient 3D heat transfer model for continuous casting solidification control (100).
3. Direct chill casting model of Mg–Li alloys using 3D finite element considering the heat transfer between dummy block and the
slab along with casting parameters (101).
4. Polymer vaporization, gas diffusion at the interface during mold filling in the lost foam casting (102).
5. A front tracking algorithm that does not require the crystal data for analyzing the columnar to equiaxed transition (103).
6. MC technique for modeling the semisolid metal processing for manufacturing of near net-shaped components (104).
7. Coupling of finite element heat flow calculations with 3D CA model for grain structure prediction in nickel base superalloy
castings (105).
8. Application of modified SOLA-VOF 3D technique applied to filling of cast iron (106).
9. Directional solidification of bladelike castings under varying electromagnetic fields of different strengths and frequencies (107).
Further research in solidification modeling can be listed as nucleation in a convectional environment, fragmentation,
transport of fragments, effect of flow rate on dendrite tip, effect of flow rate in dendritic arm spacing, and rheology of semisolid
mix. Such modeling will heavily depend in first principles using direct numerical solutions and availability of computational
facilities.
Appendix A
The phase change from liquid to solid during casting solidification involves latent heat liberation, which will have to be considered
for the liquid part of the casting. The implicit enthalpy method, mentioned briefly in Section 5.12.3.5.2 is a very powerful algorithm
for taking into account the liberation of latent heat during alloy solidification. The enthalpy formulation considers the temperature
dependency of all the properties including thermal conductivity. The finite element formulation of the scheme is therefore elab-
orated here.
The partial differential equation (PDE) describing heat transfer in a solidifying alloy, neglecting the flow effects is written in 2D
Cartesian coordinates as:

v vT v vT vT
k þ k þ Q ¼ rc in U [A.1]
vx vx vy vy vt
Here, the term Q represents the rate of heat generation due to latent heat release with the initial condition
T ¼ Ti ðx; yÞ at t ¼ 0

vT vT
k nx k ny ¼ qðx; yÞ on the heat flux boundary, G1
vx vy

vT vT
k nx k ny ¼ hðT TN Þ on the convective boundary, G2, and
vx vy

vT vT
k nx k ny ¼ lεðT 4 TN
4
Þ ¼ hr ðT TN Þ on the radiation boundary, G3 with hr ¼ lεðT 2 þ TN
2 ÞðT þ T Þ
N
vx vy
where k, r, and c are the thermal conductivity, density, and specific heat of the material, which could be functions of temperature
T. All the types of boundary conditions may or may not be present in a single casting.
In what can be considered as the most general and straightforward treatment of LH, the source term in eqn [A.1] is written as
vfs vfs vT
Q ¼ rL ¼ rL [A.2]
vt vT vt
where L is the LH of freezing and fs is the fraction solid. We can now write the energy balance equation as:

v vT v vT vfs vT vT
k þ k þ rL ¼ rc [A.3]
vx vx vy vy vT vt vt

v vT v vT vfs vT
k þ k ¼r cL [A.4]
vx vx vy vy vT vt
The net enthalpy of the alloy at any temperature is now written as
ZT
H¼ rcdT þ ð1 fs ÞrL [A.5]
T0
where T0 is any reference temperature. Differentiating the above equation with respect to T, we obtain

vH vfs
¼r cL [A.6]
vT vT
The energy equation (eqn [A.4]) becomes:

v vT v vT vH
k þ k ¼ [A.7]
vx vx vy vy vt
Finite Element Formulation
The finite element formulation of eqn [A.7] can be conveniently obtained from the Galerkin formulation. The distribution of the
dependent variable, T, within the element is assumed to be a linear combination of polynomials of the form,
X
r
T e ðx; y; tÞ ¼ ji ðx; yÞTi ðtÞ
i¼1
where r is the number of nodes assigned to element e and the Ti are the nodal temperatures. ji are the interpolation or shape
functions, the form of which is governed by the order of the element. With the assumed distribution of the temperature over the
element, the elemental equation becomes:

v vT e v vT e vH
k þ k ¼ [A.8]
vx vx vy vy vt
Multiplying eqn [A.8] by weight functions ji, integrating over an element domain and equating the weighted residual to zero,
we get
Z
v vT e v vT e vH
ji k þ k dU ¼ 0 [A.9]
vx vx vy vy vt
Ue
Reducing the order of differentiation by applying integration by parts to the first and the second terms of the integral equation,
we get:
Z I
vT e vji vT e vji vH vT e vT e
k k ji dU þ ji k nx þ k ny ds ¼ 0 [A.10]
vx vx vy vy vt G vx vy
Ue
The boundary G, which is made up of G1,2,3 depending on the problem formulation, can be split into its component terms as
below:
I I I I
vT e vT e
ji k nx þ k ny ds ¼ ji qds ji hðT e TN Þds ji hr ðT e TN Þds [A.11]
G vx vy G1 G2 G3
Substituting eqn [A.11] in eqn [A.10], we get the element equation

Z Z I I
vH vT e vji vT e vji
ji dU þ k þ k dU þ ji hT e dG þ ji hT e dG
vt vx vx vy vy G2 G3
Ue Ue [A.12]
H H H
¼ G1 ji qdG þ G2 ji hTN dG þ G3 ji hTN dG
Using the polynomial functions for developing the element matrix equations gives us the elemental equation

_ þ ½Ke fTge ¼ fFge
½Ce H [A.13]
where the matrix terms are given by:

Z
Cij ¼ ji jj dU
Ue
Z I I
vji vjj vj vjj H H H
Kij ¼ k þk i dU þ hji jj dG þ hr ji jj dG and Fi ¼ G1 qji dG þ G2 hTN ji dG þ G3 hr TN ji dG
Ue vx vx vy vy G2 G3
The set of first order ordinary differential equations in eqn [A.13] is rendered into a set of linear algebraic equations by applying
weighted average approximations to the differentiated quantity. The interpolation is linear, defined by:
fHgnþ1 fHgn
_
q H _
þ ð1 qÞ H ¼ [A.14]
nþ1 n Dt
Where, the subscript n refers to the time step number. The choice of the value of q leads us to the various difference schemes. With q
¼ 0 and observing that [C] is independent of temperature, we get the implicit scheme, i.e.,

_
½Knþ1 fTgenþ1 þ ½C H ¼ fFgnþ1 [A.15]
nþ1
Substituting for the differentiated quantity by the linear interpolation function, we get the global set of equations:
fHgnþ1 fHgn
½Knþ1 fTgnþ1 þ ½C ¼ fFgnþ1 [A.16]
Dt
The unknown enthalpy term in eqn [A.16] is defined by its Taylor series expansion with iterations within the time step:
vHnm mþ1
Hnmþ1 ¼ Hnm þ Tn Tnm [A.17]
vT
The global set of equations now take the form for iteration within the time step:

dHm
Dt½Km mþ1
n fTgn þ ½C Hnm þ n Tnmþ1 Tnm fHgn ¼ DtfFgmn [A.18]
dT
m m
1 vHn 1 vHn
K Tnm þ ½C fTgmþ1 ¼ ½C fHgm m
n fHg þ fTgm þ fFg [A.19]
Dt vT n
Dt vT n
The matrix ½K is computed for each itreation using the converged values of temperatures at the previous iteration. The iterations
are continued till the end condition defined by the tolerance

fHgmþ1 fHgm
abs n
m
n
ε [A.20]
fHgn
is achieved when the temperature field would converge. It is noteworthy here that the enthalpy term is continuously differentiable
with respect to T as is evident from eqn [A.6], which gives us an opportunity to model different alloy solidification models described
in Section 5.12.3.2. The specific forms of this function are given below for two of the simpler cases.
Case 1. Equilibrium freezing with diffusion in both liquid and solid phases
" !#
vH v 1 Tl T rL Tf Tl
¼ rc rL ¼ rc 2 [A.21]
vT vT 1 k0 Tf T ð1 k0 Þ Tf T
Case 2. Nonequilibrium freezing with diffusion only in the liquid state
" ! 1 # 2k0
k0 1
vH v Tf T k0 1
rL Tf Tl
¼ rc rL 1 ¼ rc [A.22]
vT vT Tf Tl 1
k0 1
ðk0 1Þ Tf Tl
References
1. Beckermann, C. Int. Mater. Rev. 2002, 47 (5), 243–261.

2. Ruddle, R. W. The Running and Gating of Castings. Report Series; Institute of Metals Monograph, 1956, 19.
3. Szekely, J.; Evans, J. W.; Brimacombe, J. K. The Mathematical and Physical Modeling of Primary Metals Processing; John Wiley and Sons, 1988.
4. Flemings, M. C. Solidification Processing; McGraw-Hill: New York, 1974.
5. Hong, C.-P. Computer Modelling of Heat and Fluid Flow in Materials Processing; Institute of Physics Publishing Ltd, 2004.
6. Bird, B.; Stewart, W. E.; Lightfoot, E. N. Transport Phenomena; Wiley, 1960.
7. Stefanescu, D. M. Science and Engineering of Casting Solidification, 2nd ed.; Springer, 2009.
8. Hong, C. P.; Umeda, T.; Kimura, Y. Metall. Trans. 1984, 15B, 101–107.
9. Arunkumar, S.; Prasanna Kumar, T. S. 9th AIAA/ASME Joint Thermophysics and Heat Transfer Conference, San Francisco, CA; 2006.
10. Kurz, W.; Fisher, D. J. Fundamentals of Solidification, 3rd ed.; Trans Tech Publications, 1989.
11. Brody, H. D.; Flemings, M. C. Trans. Met. Soc. AIME 1966, 236, 615.
12. Clyne, T. W.; Kurz, W. Metall. Trans. 1981, 12 A, 965.
13. Kobayashi, S. Trans. Iron Steel Inst. Jpn. 1988, 28, 728.
14. Nastac, I.; Stefanescu, D. M. Metall. Trans.. A 1993, 24, 2107.
15. Ohnaka, I. Trans. Iron Steel Inst. Jpn. 1986, 26, 1045.
16. Alifanov, O. M. Inverse Heat Transfer Problems; Springer, 1994.
17. Beck, J. V.; Blackwell, B.; St. Clair, C. R., Jr. Inverse Heat Conduction: Ill-Posed Problems; Wiley, 1985.
18. Griffiths, W. D. Metall. Trans. B 2000, 31B, 285–295.
19. Kai, Ho; Pehlke, R. D. Metall. Trans. B 1985, 16B, 585–594.
20. Krishnan, M.; Sharma, D. G. R. AFS Trans. 1994, 104, 769–774.
21. Narayan Prabhu, K.; Campbell, J. Int. J. Cast Met. Res. 1999, 12, 137–143.
22. Martorano, M. A.; Capocchi, J. D. T. Int. J. Heat Mass Transfer 2000, 43, 2541–2552.
23. Narayan Prabhu, K.; Griffiths, W. D. Int. J. Cast Met. Res. 2001, 14, 147–155.
24. Kobryn, P. A.; Semiatin, S. L. Metall. Trans. 2001, 32B, 685–695.
25. Michel, F.; Louchez, P. R.; Samuel, F. H. AFS Trans. 1995, 275–283.
26. EL-Mahallawy, N. S.; Assar, A. M. J. Mater. Sci. 1991, 26, 1729–1733.
27. Muojekwu, C. A.; Samarasekera, I. V.; Brimacombe, J. K. Metall. Trans. 1995, 26B, 361–381.
28. Cho, I. S.; Hong, C. P. Int. J. Cast Metals Res. 1996, 9, 227–232.
29. Nishida, Y.; Matsubara, H. The British Foundryman. 1976, 69, Part II, 274–278.
30. Nishida, Y.; Droste, W.; Engler, S. Metall. Trans. 1986, 17B, 833–844.
31. Okada AFS Trans. 1982, 135–146.
32. Trovent, M.; Argyropoulos, S. Metall. Trans. 2000, 31B, 87–96.
33. Prasanna Kumar, T. S.; Kamath, H. C. Metal. Mat. Trans. 2004, 35B, 575–585.
34. Arunkumar, S.; Sreenivas Rao, K. V.; Prasanna Kumar, T. S. Int. J. Heat Mass Transfer 2007.
35. Tseng, A. A.; Zhao, F. Z. Numer. Heat Transfer, Part B 1996, 29, 365–380.
36. Prasanna Kumar, T. S.; Narayan Prabhu, K. Met. Trans. B 1991, 22B, 717–727.
37. Prasanna Kumar, T. S. Numer. Heat Transfer Part B-Fundament. 2004, 45 (6), 541–563.
38. Dusinberre, G. Heat Transfer Calculations by Finite Differences, 1961.
39. Zienkiewicz, O. C. The Finite Element Method, 3rd ed.; TMH, 1979.
40. Comini, G.; Del Guidice S. Int. J. Numer. Methods Eng. 612, (8).
41. Chung, T. J. Computational Fluid Dynamics; Cambridge University Press, 2003. South Asian Edition.
42. Kardestuncer, H., Norrie, D. H., Eds. Finite Element Handbook; McGraw Hill, 1988.
43. Cook, R. D.; Malkus, D. S.; Plasha, M. E.; Witt, R. J. Concepts and Applications of Finite Element Analysis; Prentice Hall: India, 1987.
44. Swaminathan, C. R.; Voller, V. R. Metall. Trans. 1992, 23B, 651–664.
45. Baker, A. J. Finite Element Computational Fluid Mechanics; McGraw Hill, 1985.
46. Fries, T.-P.; Mathies, H. G. A Review of Petro-Galerkin Stabilization Approaches and an Extension to Meshfree Methods; Institute fur Wissenshaftliches Rechnen, 2003.
47. Zienkiewicz, O. C.; Taylor, R. L. The Finite Element Method, Chapters 2 and 3., 5th ed. In: Fluid Dynamics; Butterworth Heinemann, 2000; Vol 3.
48. Kaazempur-Mofrad, M. R.; Minev, P. D.; Ethier, C. R. Comput. Methods Appl. Mech. Eng. 2003, 192, 1281–1298.
49. Yao, G.; Sarler, B.; Chen, C. S. Engineering Analysis with Boundary Elements 2011, 35, 2011, 600–609.
50. McBride, D.; Croft, T. N.; Cross, M. Comput. Fluids 2008, 37, 170–180.
51. Samanta, D.; Zabaras, N. Mater. Sci. Eng., A 2005, 408, 211–226.
52. Rosam, J.; Jimack, P. K.; Mullis, A. J. Comput. Phys. 2007, 225, 1271–1287.
53. Hamasaiid, A.; Dour, G.; Loulou, T.; Dargusch, M. S. Int. J. Therm. Sci. 2010, 49, 365–372.
54. Carslaw, H. S.; Jaegar, J. C. Conduction of Heat in Solids, 2nd ed.; 1959.
55. Poirier, D. R.; Gieger, G. H. Transport Phenomena in Materials Processing; TMS, 1994.
56. Beckermann, C.; Diepers, H. J.; Steinbach, I.; Karma, A.; Tong, X. J. Comput. Phys. 1999, 154, 468–496.
57. Karma, A.; Rappel, W. J. Phys. Rev. E 1998, 57, 4323.
58. Wheeler, A. A.; Boettinger, W. J.; Mcfadden, G. B. Phys. Rev. A 1992, 45, 7424.
59. Zhu, P.; Smith, R. W. Acta Metall. Mater. 1992, 40, 683.
60. Rappaz, M.; Gandin Acta Metall. Mater. 1993, 41, 345.
61. Zhu, M. F.; Hong, C. P. ISIJ Int. 2001, 41, 436.
62. Cross, M.; Campbell, J. Modelling of Casting, Welding and Advanced Solidification Processes – VII, 1995.
63. Mok, J.; Hong, C. P.; Lee, J. ISIJ Int. 2003, 43, 1252.
64. Hong, C. P.; Lee, S. Y.; Song, K. ISIJ Int 2001, 41, 999.
65. Lacroix, M.; Voller, V. R. Numer. Heat Transfer, Part B 1990, 17, 25.
66. Pellini, W. S. AFS Trans. 1953, 61, 61–80.
67. Tynelius, K.; Major, J. F.; Apelian, D. AFS Trans. 1994, 101, 401–413.
68. Roy, N.; Louchez, P. R.; Samuel, F. H. J. Mater. Sci. 1996, 31 (18), 4725–4740.
69. Irani, D. R.; Kondic, V. AFS Trans. 1969, 77, 208–211.
70. Lee, P. D.; Chirazi, A.; See, D. J. Light Met. 2001, 15–30.
71. Niyama, E.; Uchida, T.; Morikawa, M.; Saito, S. AFS Int. Cast Met. J. 1982, 9, 52–63.
72. Spittle, J. A.; Almeshhedani, M.; Brown, S. G. R. Cast Met. 1994, 7 (1), 51–56.
73. Drenchev, L.; Sobczak, J.; Sobczak, N.; Sha, W.; Malinov, S. Acta Mater. 2007, 55, 6459–6471.
74. Timchenko, V.; Chen, P. Y. P.; Leonardi, E.; de Vahl Davis, G.; Abbaschian, R. Int. J. Heat Fluid Flow 2002, 23, 258–268.
75. Lee, S. G.; Gokhale, A. M. Scr. Mater. 2006, 55, 387–390.
76. Lee, P. D.; Hunt, J. D. Modeling of Casting, Welding and Advanced Solidification Processes, 1995; pp 585–592.
77. Lee, P. D.; Hunt, J. D. Acta Mater. 2001, 49, 1383–1398.
78. Imafaku, I.; Chijiiwa, K. AFS Trans. 1983, 91, 91.
79. Bounds, S.; Moran, G.; Pericleous, K.; Cross, M.; Croft, T. N. Metall. Mater. Trans. 2000, 31B, 515.
80. Chang, S.; Stefanescu, D. M. Metall. Mater. Trans. 1996, 27A, 2708.
81. Ridder, S. D.; Kou, S.; Mehrabian, R. Metall. Trans. 1981, 12B, 435–447.
82. Beckermann, C.; Viskanta, R. PhysicoChem. Hydrodyn. 1988, 10, 195–213.
83. Bennon, W. D.; Incropera, F. P. Int. J. Heat Mass Transfer 1987, 30, 2161–2170.
84. Gronge, O.; Shercliff, H. R. Prog. Mater. Sci. 2002, 47, 163–282.
85. Chen, Q.; Langet, E. W.; Hansen, P. N. Scand. J. Metall. 1995, 24, 48.
86. Maxwell, I.; Hellawell, A. Acta Metall. 1975, 23, 229.
87. Nastac, I.; Stefanescu, D. M. AFS Trans. 1995, 103, 329.
88. Wang, C. Y.; Beckermann, C. Metall. Trans. 1993, 24A, 2787.
89. Mullins, W. W.; Sekerka, R. F. J. Appl. Phys. 1964, 35, 444.
90. Rappaz, M.; Voller, V. R. Metall. Mater. Trans. 1990, 21A, 749–753.
91. Schneider, M. C.; Beckermann, C. Metall. Mater. Trans. 1995, 26A, 2373–2388.
92. Schneider, M. C.; Gu, G. P.; Beckermann, C.; Boettinger, W. J.; Kattner, U. R. Metall. Mater. Trans. 1997, 28A, 1517–1531.
93. Thevik, H. J.; Mo, A. Int. J. Heat Mass Transfer 1997, 40 (28B), 665–669.
94. Sun, Z.; Logé, R. E.; Bernacki, M. Comput. Mater. Sci. 2010, 49, 158–170.
95. Vernéde, S.; Jarry, P.; Rappaz, M. Acta Mater. 2006, 54, 4023–4034.
96. Stewart, M. J.; Weinberg, F. J. Cryst. Growth 1972, 12, 228–238.
97. Stoehr, R. A. Can. Metall. Quart. 1998, 37 (3–4), 179–184.
98. Stewart, J. R.; Hellawell, A. Metall. Trans. A 1988, 19A (7), 1861–1871.
99. Blasé, T. A.; Guo, Z. X.; Shia, Z.; Long, K.; Hopkins, W. J. Mater. Sci. Eng., A 2004, 365, 318–324.
100. Louhenkilpi, S.; Mäkinen, M.; Vapalahti, S.; Räisänen, T.; Laine, J. Mater. Sci. Eng., A 2005, 413–414, 135–138.
101. Shi, S.; Hao, H.; Zhang, X.; Fang, C.; Yao, S.; Jin, J. Rare Met. Mater. Eng. 2009, 38 (2), 203–208.
102. Barone, M. R.; Caulk, D. A. Int. J. Heat Mass Transfer 2005, 48, 4132–4149.
103. McFadden, S.; Browne, D. J. Appl. Math. Model. 2009, 33, 1397–1416.
104. Das, A.; Fan, Z. Mater. Sci. Eng., A 2004, 365, 330–335.
105. Seo, S.-M.; Kima, I.-S.; Jo, C.-Y.; Ogi, K. Mater. Sci. Eng., A 2007, 449–451, 713–716.
106. Babaei, R.; Abdollahi, J.; Homayonifar, P.; Varahram, N.; Davami, P. Comput. Methods Appl. Mech. Eng. 2006, 195, 775–795.
107. Daming Xu; Yunfeng, Bai; Hengzhi, Fu; Jingjie, Guo Int. J. Heat Mass Transfer 2005, 48, 2219–2232.
108. Bathe, K.-J. Finite Element Procedures; Prentice Hall: India, 1997.
109. Zhu, P.; Smith, R. W. Acta Metall. Mater 1992, 40, 683.
110. Kobayashi, S. Trans. Iron Steel Inst. Japan 1988, 28, 728.
111. Timchenko, V.; Chen, P. Y. P.; Leonardi, E.; de Vahl Davis, G.; Abbaschian, R. Int. J. Heat Fluid Flow 2002, 23, 258–268.

Casting Simulation Methods: TS Prasanna Kumar, Indian Institute of Technology Madras, Chennai, India

Uploaded by

Copyright:

Available Formats

Casting Simulation Methods: TS Prasanna Kumar, Indian Institute of Technology Madras, Chennai, India

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Casting Simulation Methods: TS Prasanna Kumar, Indian Institute of Technology Madras, Chennai, India

Uploaded by

Copyright:

Available Formats

5.

12 Casting Simulation Methods

5.12.1 Introduction 235

Comprehensive Materials Processing, Volume 5 http://dx.doi.org/10.1016/B978-0-08-096532-1.00507-0 235

5.12.2 A Holistic Approach to Casting Simulation

5.12.3 Mathematical Modeling

5.12.3.1.1 Modeling Latent Heat Release Through the Source Term

5.12.3.1.2 Enthalpy Method

Differentiating the above equation with respect to T, we obtain

5.12.3.1.3 Temperature Recovery Method

Zt1 ZT2 ZT2

5.12.3.2 Alloy Solidiﬁcation Models

5.12.3.2.2 Nonequilibrium Freezing with Diffusion Only in the Liquid State

5.12.3.2.3 Solute Redistribution with Solid-State Diffusion

The value of aL decreases form 1 to 0 as k0 increases from 0 to N.

5.12.3.2.4 Mixing in the Liquid

5.12.3.2.6 Modeling of Columnar Structures

5.12.3.2.7 Microscopic Modeling of Equiaxed Structures

5.12.3.3 Nucleation Kinetics

5.12.3.3.2 Continuous Nucleation

5.12.3.4 Metal–Mold Interface Heat Transfer

l Local curvatures, corner effects

Table 1 Some studies in modeling metal–mold interface heat transfer characteristics

Sl. no. Casting–mold configuration Alloy system(s) Thermal condition Comments

Minimization of the objective function leads us to

where f is the sensitivity coefﬁcient and is given by

b þ ¼ Tj;i qm;.;mþr1 ¼ qm1 0

b j;i ¼ Tj;i qm;.;mþr1 ¼ qm1

5.12.3.5 Numerical Methods

5.12.3.5.1 The Finite Element Method

fBg ¼ ð½C ð1 qÞDt½KÞfTgn þ DtfFgn [47]

5.12.3.5.2 FEM for Nonlinear Problems in Phase Change

5.12.3.5.3 FEM for Flow Modeling During Solidification

5.12.3.5.4 Advanced Applications

5.12.4 Simulation Models

5.12.4.1 Deterministic Models

5.12.4.2 Stochastic Models

5.12.5 Special Topics in Casting Simulation

5.12.5.2 Mushy Zone Modeling

5.12.5.4 Shrinkage Defects

5.12.5.5 Micro–Macro Segregation

5.12.5.6 Microstructure Modeling

5.12.6 Recent Applications and Future Directions

Finite Element Formulation

Substituting eqn [A.11] in eqn [A.10], we get the element equation

where the matrix terms are given by:

Case 2. Nonequilibrium freezing with diffusion only in the liquid state

1. Beckermann, C. Int. Mater. Rev. 2002, 47 (5), 243–261.

You might also like

fBg ¼ ð½C ð1 qÞDt½KÞfTgn þ DtfFgn [47]