Modeling of Stress, Distortion,and Hot TearingBrian G. Thomas, University of Illinois (UIUC)Michel Bellet, Ecole des Mines de Paris, Centre de Mise en Forme des Matériaux (CEMEF)

COMPUTATIONALMODELING of mecha-nical behavior during solidification is becomingmore important. Thermal and microstructuralsimulations alone are insufficient to predict thequality of the final product that is desired bythe casting industry. Accurate calculation of dis-placements, strains, and stresses during the cast-ing process is needed to predict residual stressand distortion and defects such as the formationof cracks such as hot tears. It also helps predictporosity and segregation. As computing powerand software tools for computational mechanicsadvance, it is becoming increasingly possible toperform useful mechanical analysis of castingsand these important related behaviors.The thermomechanical analysis of castings pre-

sents a formidable challenge for many reasons:

� Many interacting physical phenomena areinvolved in stress-strain formation. Stressarises primarily from the mismatch of strainscaused by large temperature gradients anddepends on the time- and microstructure-dependent inelastic flow of the material.

� Predicting distortions and residual stresses incast products requires calculation of the his-tory of the cast product and its environmentover huge temperature intervals. This makesthe mechanical problem highly nonlinear,involving liquid solid interaction and com-plex constitutive equations. Even identifyingthe numerous metallurgical parametersinvolved in those relations is a daunting task.

� The coupling between the thermal and themechanical problems is an additional diffi-culty. This coupling comes from the mechani-cal interaction between the casting and themold components, through gap formation orthe buildupof contact pressure, locallymodify-ing the heat exchange. This adds some com-plexity to the nonlinear heat transfer resolution.

� Accounting for the mold and its interactionwith the casting makes the problem multido-main, usually involving numerous deform-able components with coupled interactions.

� Cast parts usually have very complex three-dimensional shapes, which puts great demands

on the interface between CAD design and themechanical solvers and on computationalresources.

� The important length scales range frommicrometers (dendrite arm shapes) to tensof meters (metallurgical length of a continu-ous caster), with similarly huge order-of-magnitude range in time scales.

This article summarizes some of the issues andapproaches in performing computational analy-ses of mechanical behavior, distortion, and hottearing during solidification. The governingequations are presented first, followed by a briefdescription of the methods used to solve them,and a few examples of recent applications inshape castings and continuous casting.

Governing Equations

The modeling of mechanical behavior req-uires solution of 1) the equilibrium or mome-ntum equations relating force and stress, 2)the constitutive equations relating stress andstrain, and 3) compatibility equations relatingstrain and displacement. This is because theboundary conditions specify either force or dis-placement at different boundary regions of thedomain O:

u ¼ u_ on @�usn ¼ T_ on @�T ðEq 1Þwhere u_ are prescribed displacements on bound-

ary surface portion @Ou, and T_

are boundarysurface forces or “tractions” on portion @OT.The next sections first present the equilibriumand compatibility equations and then introduceconstitutive equations for the different materialstates during solidification.Equilibrium and Compatibility Equations.

At any time and location in the solidifyingmaterial, the conservation of force (steady-stateequilibrium) or momentum (transient condi-tions) can be expressed by:

r � sþ rg� r dvdt

¼ 0 (Eq 2)

where s is the stress tensor, r is the density,g denotes gravity, v is the velocity field, andd/dt denotes the total (particular) time deriva-tion. Stress can be further split into the deviatoricstress tensor and the pressure field. The differ-ent approaches for simplifying and solvingthese equations are discussed in the section“Implementation Issues.”Once the material has solidified, the internal

and gravity forces dominate, so the inertiaterms in Eq 2 can be neglected. Furthermore,the strains that dominate thermomechanicalbehavior during solidification are on the orderof only a few percent, otherwise cracks willform. With small gradients of spatial displace-ment, ru = @u/@x, and the compatibility equa-tions simplify to (Ref 1):

« ¼ 12

ruþ ðruÞT� �

(Eq 3)

where « is the strain tensor and u is the dis-placement vector. This small-strain assumptionsimplifies the analysis considerably. The com-patibility equations can also be expressed asa rate formulation, where strains become strainrates, and displacements become velocities.This formulation is more convenient for a tran-sient computation with time integration involv-ing fluid flow and/or large deformation.In casting analysis, the cast material may be

in the liquid, mushy, or solid state. Each ofthese states has different constitutive behavior,as discussed in the next three sections.Liquid-State Constitutive Models. Metallic

alloys generally behave as Newtonian fluids.Including thermal dilatation effects, the consti-tutive equation can be expressed as:

_« ¼ 12ml

s� 13r

drdt

I (Eq 4)

The strain-rate tensor _« is split into two compo-nents: a mechanical part, which varies linearlywith the deviatoric stress tensor s, and a thermalpart. In this equation, ml is the dynamic viscos-ity of the liquid, r is the density, and I is theidentity tensor. Taking the trace of this expres-sion, tr _« ¼ r � v, the mass conservation equa-tion is recovered:

5115/4G/a0005238

drdt

þ rr � v ¼ @r@t

þr � ðrvÞ ¼ 0 (Eq 5)In casting processes, the liquid flow may be

turbulent, even after mold filling. This mayoccur because of buoyancy forces or forcedconvection such as in jets coming out of thenozzle outlets in continuous casting processes.The most accurate approach, direct numericalsimulation, generally is not feasible for indus-trial processes, owing to their complex-shapeddomains and high turbulence. To compute justthe large-scale flow features, turbulence modelsare used that increase the liquid viscosityaccording to different models of the small-scalephenomena. These models include the simple“mixing-length” models, the two-equationmodels such as k-e, and large eddy simulation(LES) models, which have been compared witheach other and with measurements of continu-ous casting (Ref 2–4).Mushy-State Constitutive Models. Metallic

alloys in the mushy state are two-phase liquid-solid media. Their mechanical response dependsgreatly on the local microstructural evolution,which involves several complex physical phe-nomena. An accurate description of these phe-nomena is useful for studying hot tearing ormacrosegregation. Knowledge of the liquid flowin the mushy zone is necessary to calculate thetransport of chemical species (alloying ele-ments) (Ref 5). Knowledge of the deformationof the solid phase is important when it affectsliquid flow in the mushy zone by “sponge-effects” (Ref 6). In such cases, two-phasemodels must be used. Starting from microscopicmodels describing the intrinsic behavior of theliquid phase and the solid phase, spatial averag-ing procedures must be developed to express thebehavior of the compressible solid continuumand of the liquid phase that flows through it(Ref 7–9).If a detailed description is not really needed,

such as in the analysis of residual stresses anddistortions, the mushy state can be approxi-mated as a single continuum that behaves as anon-Newtonian (i.e., viscoplastic) fluid, accord-ing to Eq 6 to 8. Thus, the liquid phase is notdistinguished from the solid phase, and the indi-vidual dendrites and grain boundaries are notresolved.

_« ¼ _«vp þ _«th (Eq 6)

_«vp ¼ 32K

ð _«eqÞ1�ms (Eq 7)

_«th ¼ � 13r

drdt

I (Eq 8)

K is the viscoplastic consistency and m the strain-

rate sensitivity. Denoting seq ¼ffiffiffiffiffiffiffiffiffiffiffiffiffi32sijsij

qthe

von Mises equivalent stress scalar, and

_eeq ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffi23_evpij _e

vpij

qthe von Mises equivalent

strain-rate scalar, Eq 7 yields the well-knownpower law: seq ¼ Kð _eeqÞm . Note that the pre-ceding Newtonian liquid model is actuallya particular case of this non-Newtonian one:

Eq 4 can be derived from Eq 6, 7, and 8 takingm = 1 and K = 3ml. The solidification shrinkageis included in Eq 8: writing r = gs rS + gl rLin the solidification interval (rS, rL densitiesat the solidus and liquidus temperatures, respec-tively, gs, gl volume fractions of solid and liq-uid, respectively), the thermal strain rate isdefined as:

tr_eth ¼� 1rdrdt

¼ � 1rðrS � rLÞ

dgsdt

� rL � rSrL

dgsdt (Eq 9)

Solid-State Constitutive Models. In thesolid state, metallic alloys can be modeledeither as elastic-plastic or elastic-viscoplasticmaterials. In the latter class of models, one ofthe simpler models is expressed as follows,but it should be mentioned that a lot of modelsof different complexity can be found in the lit-erature (Ref 10, 11).

_« ¼ _«el þ _«in þ _«th (Eq 10)

_«el ¼ 1þ nE

_s� nEtrð _sÞIþ _T @

@T

1þ nE

� �s

� _T @@T

nE

� �trðsÞI ðEq 11Þ

_«in ¼ 32seq

seq � s0K

D E1=ms (Eq 12)

_«th ¼ � 13r

drdt

I (Eq 13)

The strain-rate tensor _« is split into an elasticcomponent, an inelastic (nonreversible) compo-nent, and a thermal component. Equation 11 isthe hypoelastic Hooke’s law, where E isYoung’s modulus, n the Poisson’s coefficient,and _s a time derivative of the stress tensor s.Equation 12 gives the relation between the

inelastic strain-rate tensor _«in and the stressdeviator, s, in which s0 denotes the scalar staticyield stress, below which no inelastic deforma-tion occurs (the expression between brackets isset to 0 when it is negative). In these equations,the temperature dependency of all the involvedvariables should be considered. The effect ofstrain hardening may appear in such a modelby the increase of both the static yield stresss0 and the plastic consistency K with the accu-mulated inelastic strain eeq, or with anotherstate variable that is representative of the mate-rial structure. The corresponding scalar equa-tion relating stress and inelastic strain-rate vonMises invariants is:

seq ¼ s0 þKð _eeqÞm (Eq 14)Inserting this into Eq 12 simplifies it to:

_«in ¼ 3 _eeq2seq

s; or; in incremental form;

d«in ¼ 3deeq2seq

s ðEq 15ÞAlthough metallic alloys show a significant

strain-rate sensitivity at high temperature, theyare often modeled in the literature using elas-tic-plastic models, neglecting this importanteffect. In this case, Eq 15 still holds, but theflow stress is independent of the strain rate. Itmay depend on the accumulated plastic strainbecause of strain hardening.

Implementation Issues. One of the majordifficulties in the thermomechanical analysisof casting processes is the concurrent presenceof liquid, mushy, and solid regions that movewith time as solidification progresses. Severaldifferent strategies have been developed,according to the process and model objectives:

� A first strategy consists in extracting thesolidified regions of the casting domainbased on the thermal analysis results. Then,a small-strain thermomechanical analysis iscarried out on just this solid subdomain,using a standard solid-state constitutivemodel. Besides difficulties with the extrac-tion process, especially when the solidifiedregions have complex unconnected shapes,this method may have numerical problemswith the application of the liquid hydrostaticpressure onto the new internal boundary ofthe solidified region. However, this simplestrategy is very convenient for many practi-cal problems, especially when the solidifica-tion front is stationary, such as the primarycooling of continuous casting of aluminum(Ref 12) and steel (Ref 13, 14). For transientproblems, such as the prediction of residualstress and shape (butt-curl) during startupof the aluminum DC continuous casting pro-cess, the domain can be extended in time byadding layers (Ref 12).

� A second strategy considers the entire cast-ing, including the mushy, and liquid regions.The liquid, mushy, and solid regions aremodeled as a continuum by adopting theconstitutive equations for the solid phase(see the previous section “Solid-State Con-stitutive Models”) for all regions by adjust-ing material parameters such as K, m, E, n,s0, and r according to temperature. Forexample, liquid can be treated by settingthe strains to 0 when the temperature isabove the solidus temperature. This ensuresthat stress development in the liquid phaseis suppressed. In the equilibrium equation,Eq 2, acceleration terms are neglected, anda small-strain analysis can be performed.The primary unknowns are the displace-ments, or displacement increments. Thispopular approach can be used with structuralfinite element codes, such as MARC (Ref15) or ABAQUS (Ref 16), and with com-mercial solidification codes or special-purpose software, such as ALSIM (Ref 17)/ALSPEN, (Ref 18), CASTS, (Ref 19),CON2D (Ref 20, 21), Magmasoft (Ref 22),and Procast (Ref 23, 24). It has been appliedsuccessfully to simulate deformation andresidual stress in shape castings (Ref 25,26), direct chill casting of aluminum (Ref12, 17, 18, 27–29), and continuous castingof steel (Ref 20, 30). Despite its efficiency,this approach may suffer from several draw-backs. First, it cannot properly account forfluid flow and the volumetric shrinkage thataffects flow in the liquid pool, fluid feedinginto the mushy zone, and primary shrinkage

2/Modeling of Stress, Distortion, and Hot Tearing

5115/4G/a0005238

depressions that affect casting shape. Inaddition, incompressibility of the metal inthe liquid state is accounted for by increas-ing Poisson’s ratio to close to 0.5 whichsometimes makes the solution prone tonumerical instability (Ref 31, 32).

� A third strategy has recently been developedthat addresses the above issues. It still simu-lates the entire casting, but treats the massand momentum equations of the liquid andmushy regions with a mixed velocity-pres-sure formulation. The primary unknownsare the velocity (time derivative of displace-ment) and pressure fields, which make iteasier to impose the incompressibility con-straint (see next section “ThermomechanicalCoupling”). Indeed, the velocity-pressureformulation is also applied to the equilib-rium of the solid regions to provide a singlecontinuum framework for the global numer-ical solution. This strategy has been imple-mented into codes dedicated to castinganalysis such as THERCAST (Ref 30, 33,34) and VULCAN (Ref 35). If stress predic-tion is not important so that elastic strainscan be ignored, then this formulation sim-plifies to a standard fluid-flow analysis,which is useful in the prediction of bulgingand shape in large-strain processes.

Example of Solid-State Constitutive Equa-tions. Material property data are needed for thespecific alloy being modeled and in a form suit-able for the constitutive equations just discussed.This presents a significant challenge for quantita-tive mechanical analysis, because measurementsare not presented in this form and only rarelysupply enough information on the conditions toallow transformation to an alternate form. As anexample, the following elastic-viscoplastic con-stitutive equation was developed for the austen-ite phase of steel (Ref 36) by fitting constantstrain-rate tensile tests (Ref 37, 38) and con-stant-load creep tests (Ref 39) to the formrequired in Eq 10 to Eq 13.

_eeq ¼ f%C seq � s0� �1=m

exp � 4:465� 104

T

� �(Eq 16)

wheref%C ¼ 4:655� 104 þ 7:14� 10ð%CÞ þ 1:2� 104ð%CÞ2s0 ¼ ð130:5� 5:128� 10�3T Þeeqf2f2 ¼ �0:6289þ 1:114� 10�3T1=m ¼ 8:132� 1:54� 10�3Twith T ½K�: �eq: �0½MPa�

This equation, and a similar one for delta ferrite,have been implemented into the finite-elementcodes CON2D (Ref 20) and THERCAST (Ref40) and applied to investigate several problemsinvolving mechanical behavior during continuouscasting.Elastic modulus is a crucial property that

decreases with increasing temperature. It is dif-ficult to measure at the high temperaturesimportant to casting, owing to the susceptibilityof the material to creep and thermal strain

during a standard tensile test, which results inexcessively low values. Higher values areobtained from high strain rate tests, such asultrasonic measurements (Ref 41) Elastic mod-ulus measurements in steels near the solidustemperature range from �1 GPa (145 ksi)(Ref 42) to 44 GPa (6400 ksi) (Ref 43) withtypical modulus values �10 GPa (1450 ksi)near the solidus (Ref 44–46).The density needed to compute thermal

strain in Eq 4, 8, or 13 can be found froma weighted average of the values of the differ-ent solid and liquid phases, based on the localphase fractions. For the example of plain low-carbon steel, the following equations were com-piled (Ref 20) based on solid data for ferrite(a), austenite (g), and delta (d) (Ref 47, 48)and liquid (l) measurements (Ref 49).

rðkg=m3Þ ¼ rafa þ rgfg þ rdfd þ rlflra ¼ 7881� 0:324TðCÞ � 3� 10�5T ðCÞ2

rg ¼100½8106� 0:51T ðCÞ�

½100� ð%CÞ�½1þ 0:008ð%CÞ�3

rd ¼100½8011� 0:47T ðCÞ�

½100� ð%CÞ�½1þ 0:013ð%CÞ�3rl ¼ 7100� 73ð%CÞ � ½0:8� 0:09ð%CÞ�

½T ðCÞ � 1550� ðEq 17ÞSpecialized experiments to measure mechan-

ical properties for use in computational modelswill be an important trend for future research inthis field.

Thermomechanical Coupling

Coupling between the thermal and mechani-cal analyses arises from several sources. First,regarding the mechanical problem, besides thestrain rate due to thermal expansion and solidi-fication shrinkage, the material parameters ofthe preceding constitutive equations stronglydepend on temperature and phase fractions, asshown in the previous section. Second, in theheat-transfer problem, the thermal exchangebetween the casting and the mold stronglydepends on local conditions such as the contactpressure or the presence of a gap between them(as a result of thermal expansion and solidifica-tion shrinkage). This is illustrated in Fig. 1 anddiscussed in this section.Air Gap Formation: Conductive-Radiative

Modeling. In the presence of a gap betweenthe casting and the mold, resulting from their rel-ative deformation, the heat transfer results fromconcurrent conduction through the gas withinthe gap and from radiation. The exchanged ther-mal flux, qgap, can then be written:

qgap ¼ kgasg

ðTc � TmÞ þ sðT4c � T 4mÞ

1ecþ 1em � 1

(Eq 18)

with kgas (T) the thermal conductivity of thegas, g the gap thickness, Tc and Tm the localsurface temperature of the casting and mold,respectively, ec and em their gray-body emissiv-ities, s the Stefan-Boltzmann constant. It is tobe noted that the conductive part of the fluxcan be written in more detail to take into

account the presence of coating layers on themold surface: conduction through a medium ofthickness gcoat, of conductivity kcoat (T). It canbe seen that the first term tends to infinity as thegap thickness tends to 0; this expresses a perfectcontact condition, Tc and Tm tending towarda unique interface temperature. The reality issomewhat different, showing always nonperfectcontact conditions. Therefore, the conductiveheat-exchange coefficient hcond = kgas/g shouldbe limited by a finite value h0, corresponding tothe “no-gap” situation, and depends on theroughness of the casting surface. Specific exam-ples of these gap heat-transfer laws are providedelsewhere for continuous casting with oil lubri-cation (Ref 13) and mold flux (Ref 50).Effective Contact: Heat Transfer as a

Function of Contact Pressure. With effectivecontact, the conductive heat flux increases withthe contact pressure according to a power law(Ref 51). Still denoting h0 as the heat-exchangecoefficient corresponding to no gap and no con-tact pressure, the interfacial heat flux is:

qcontact ¼ ðh0 þApBc ÞðTc � TmÞ (Eq 19)with pc the contact pressure, A and B two param-eters that depend on the materials, the presenceof coating or lubricating agent, the surfaceroughness, and the temperature. The parametersand possibly the laws governing their evolutionneed to be determined experimentally.

Numerical Solution

The thermomechanical modeling equationsjust presented must be solved numerically,owing to the complex shape of the casting pro-cess domain, and the highly nonlinear materialproperties. The calculation depends greatly onthe numerical resolution of time and space.Although finite-difference approaches are popu-lar for heat-transfer, solidification, and fluid-flow analyses, the finite element formulation isusually preferred for the mechanical analysis,owing to its historical advantages with unstruc-tured meshes and accurate implicit solution ofthe resulting simultaneous algebraic equations.The latter are discussed below.

Finite Element Formulation andNumerical Implementation

In the framework of the small-strain appro-ach presented previously (see the section“Implementation Issues”), having displace-ments for primitive unknowns, the weak formof the equilibrium equation, Eq 2, neglectinginertia terms, is written as:

8u

ð�

s : d«ðuÞdV �ð@�

T � udS

�ð�

rg � udV ¼ 0 ðEq 20Þ

where T is the external stress vector. The vectortest functions u can be seen as virtual displace-ments in a statement of virtual work.

Modeling of Stress, Distortion, and Hot Tearing/3

5115/4G/a0005238

If the third strategy described in the section“Implementation Issues” is adopted, with veloc-ity and pressure as primary unknown variables,the weak form of the momentum equation(Eq 2) is written as (Ref 52):

8v

ð�

s : _«ðvÞdV �ð�

pr � vdV

�ð@�

T � vdS �ð�

rg � vdV

þð�

rdv

dt� vdV ¼ 0

8p

ð�

ptr _«indV ¼ 0

8>>>>>>>>>>>>>>>><>>>>>>>>>>>>>>>>:

(Eq 21)

The first equation contains vector test func-tions v*, which can be seen as virtual velocitiesin a statement of virtual power. Unlike Eq 20,the pressure p is a primary variable, and onlythe deviatoric part of the constitutive equationsis involved (to determine the stress deviator s).This is why the second equation is needed,which consists of a weak form of the incom-pressibility of inelastic deformations.Equations 20 and 21 are spatially discretized

using the standard finite-element method, asexplained in detail in many references (Ref52). Combined with time discretization usingfinite differences, this leads to a set of nonlinearequations to be solved at each time increment.In the context of the displacement strategy,Eq 20, this leads to:

RðUÞ ¼ 0 (Eq 22)where R is the vector of the nodal equilibriumresidues (number of components: 3 � numberof nodes, in three dimensions), and U is thevector of nodal incremental displacements(same size).Adopting the velocity-pressure strategy,

Eq 21 leads to a set of nonlinear equations:

R0ðV;PÞ ¼ 0 (Eq 23)where R0 is the vector of the nodal residues(number of components: 4 � number of nodes,

in dimension 3), V is the vector of nodal veloc-ities (size: 3 � number of nodes), and P is thevector of nodal pressures (size: number ofnodes).The global finite-element systems Eq 22 or

Eq 23 are usually solved using a full or modi-fied Newton-Raphson method (Ref 16, 31),which iterates to minimize the norm of the res-idue vectors R or R0. Alternatively, explicitmethods may be employed at this global level.At the local (finite element) level, an algo-

rithm is also required to integrate the constitu-tive equations, when they depend on strainrate or strain. When the constitutive equationsare highly nonlinear, an implicit algorithm isuseful to perform time integration at eachGauss point to provide better estimates ofinelastic strain at the local level (Ref 53–55).

Boundary Conditions: Modeling ofContact Conditions. MultidomainApproaches

At the interface between the solidifyingmaterial and the mold, a contact condition isrequired to prevent penetration of the shell intothe mold, while allowing shrinkage of the shellaway from the mold to create an interfacial gap:

s n � n � 0g � 0ðsn � nÞg ¼ 0

8<:

(Eq 24)

where g is the local interface gap width (positivewhen air gap exists effectively, as in section 0)and n is the local outward unit normal to thepart. Equation 24 can be satisfied with a penaltycondition, which consists of applying a normalstress vector T proportional to the penetrationdepth (if any) via a penalty constant wp:T ¼ sn ¼ ��p �gh in (Eq 25)Here again, the brackets denote the positive

part; a repulsive stress is applied only if g isnegative (penetration). Different methods oflocal adaptation of the penalty coefficient wp

have been developed, including the augmentedLagrangian method (Ref 56). More complexand computationally expensive methods, suchas the use of Lagrange multipliers may also beused (Ref 57).The possible tangential friction effects

between part and mold can be taken intoaccount by a friction law, such as a Coulombmodel for instance. In this case, the previousstress vector has a tangential component, Tt,given by:

Tt ¼ �mfpc1

v� vmoldk k ðv� vmoldÞ (Eq 26)where pc ¼ �sn ¼ �sn � n is the contact pres-sure, and mf the friction coefficient.The previous approach can be extended to the

multidomain context to account for the deforma-tion of mold components. The local stress vec-tors calculated by Eq 25 can be applied ontothe surface of the mold, contributing then to itsdeformation. For most casting processes, themechanical interaction between the cast productand the mold is sufficiently slow (i.e., its charac-teristic time remains significant with respect tothe process time) to permit a staggered schemewithin each time increment: the mechanicalproblem is successively solved in the cast prod-uct and in the different mold components.A global updating of the different configurationsis then performed at the end of the time incre-ment. This simple approach gives access to a pre-diction of the local air gap size g, or alternativelyof the local contact pressure pc, that is used inthe expressions of the heat-transfer coefficient,according to Eq 18 and 19 (Ref 58).

Treatment of the Regions in the Solid,Mushy, and Liquid States

Solidified Regions: Lagrangian Formula-tion. In casting processes, the solidified regionsgenerally encounter small deformations. It isthus natural to embed the finite element domaininto the material, with each node of the compu-tational grid corresponding with the same solidparticle during its displacement. The boundaryof the mesh corresponds then to the surface ofthe casting. This method, called Lagrangianformulation, provides the best accuracy whencomputing the gap forming between the solidi-fied material and the mold. It is also the morereliable and convenient method for time integra-tion of highly nonlinear constitutive equations,such as elastic-(visco)plastic laws presented inthe section “Solid-State Constitutive Models.”Mushy and liquid regions: ALE modeling.

When the mushy and liquid regions are mod-eled in the same domain as the solid (see thediscussion in the section “ImplementationIssues”), they are often subjected to largedisplacements and strains arising from solidifi-cation shrinkage, buoyancy, or forced convec-tion. Similar difficulties are generated incasting processes such as squeeze casting,where the entire domain is highly deformed.In these cases, a Lagrangian formulation would

Heat-exchange coefficienth (W/m2•K)

Gap width g (m)

Effective gapEffective contact

Contact pressure pc (Pa)

h0

h = h0 + ApCB h =

kgas

g 1 1 −1++

εc εm

s (Tc2 + Tm

2 )(Tc + Tm)

Fig. 1 Modeling of the local heat transfer coefficient in the gap and effective contact situations

4/Modeling of Stress, Distortion, and Hot Tearing

5115/4G/a0005238

demand frequent remeshings to avoid meshdegeneracy, which is both computationallycostly and detrimental to the accuracy of themodeling. It is then preferable to use a so-calledArbitrary Lagrangian Eulerian formulation(ALE). In a Eulerian formulation, materialmoves through the computational grid, whichremains stationary in the “laboratory” frame ofreference. In the ALE formulation, the updatingof the mesh is partially independent of thevelocity of the material particles to maintainthe quality of the computational grid. Severalmethods can be used, including the popular“barycentering” technique, which keeps eachnode at the geometrical centroid of a set of itsneighbors. This method involves significantextra complexity to account for the advectionof material through the domain, and the statevariables such as temperature and inelasticstrain must be updated according to the relativevelocity between the mesh and the particles. Indoing this, some surface constraints must beenforced to ensure mass conservation, expres-sing that the fluxes of mesh velocity and offluid particle velocity through the surface ofthe mesh should remain identical. A review onthe ALE method in solidification modeling isavailable, together with some details on itsapplication (Ref 33).

Thermomechanical Coupling

Because of the interdependency of the ther-mal and mechanical analyses, as presented inthe section “Thermomechanical Coupling,”their coupling should be taken into account allthroughout the cooling process. In practice,the cooling time is decomposed into time incre-ments, each increment requiring the solution oftwo problems: the energy conservation and themomentum conservation. With the highly non-linear elastic-viscoplastic constitutive equationstypical of solidifying metals, the incrementalsteps required for the mechanical analysis toconverge are generally much smaller than thosefor the thermal analysis. Thus, these two analy-ses are generally performed in succession andonly once per time increment. However, in thecase of very rapid cooling, these solutionsmight preferably be performed together (includ-ing thermal and mechanical unknowns in a sin-gle set of nonlinear equations), or else

separately but iteratively until convergence ateach time increment, otherwise the time stephas to be dramatically reduced.

Model Validation

Model validation with both analytical solu-tions and experiments is a crucial step in anycomputational analysis, and thermomechanicalmodeling is no exception. Weiner and Boley(Ref 59) derived an analytical solution for uni-directional solidification of an unconstrainedplate with a unique solidification temperature,an elastic perfectly plastic constitutive law andconstant properties. The plate is subjected tosudden surface quench from a uniform initialtemperature to a constant mold temperature.This benchmark problem is ideal for estimat-

ing the discretization errors of computationalthermal-stress models, as it can be solved witha simple mesh consisting of one row elements,as shown in Fig. 2. Numerical predictionsshould match with acceptable precision usingthe same element type and mesh refinementplanned for the real problem. For example, thesolidification stress analysis code, CON2D(Ref 20) and the commercial code ABAQUSwere applied for typical conditions of steel cast-ing (Ref 21).Figures 3 and 4 compare the temperature and

stress profiles in the plate at two times. Thetemperature profile through the solidifying shellis almost linear. Because the interior cools rela-tive to the fixed surface temperature, its shrink-age generates internal tensile stress, whichinduces compressive stress at the surface. Withno applied external pressure, the average stressthrough the thickness must naturally equal 0,

and stress must decrease to 0 in the liquid.Stresses and strains in both transverse direc-tions are equal for this symmetrical problem.The close agreement demonstrates that bothcomputational models are numerically consis-tent and have an acceptable mesh resolution.Comparison with experimental measurementsis also required to validate that the modelingassumptions and input data are reasonable.

Example Applications

Sand Casting of Braking Disks. The finiteelement software THERCAST for thermome-chanical analysis of solidification (Ref 34) hasbeen used in the automotive industry to predictdistortion of gray-iron braking discs cast insand molds (Ref 60). Particular attention hasbeen paid to the interaction between the defor-mation of internal sand cores and the cast parts.This demands a global coupled thermomechani-cal simulation, as presented previously. Figure 5illustrates the discretization of the differentdomains involved in the calculation. The actualcooling scenario has been simulated: cooling inmold for 45 min, shakeout, and air cooling for15 min. Figure 6 shows the temperature evolu-tion at different points in a horizontal cross sec-tion at midheight in the disc, revealing:solidification after 2 min, and solid-state phasechange after 20 min. The calculated deforma-tion of the core blades shows thermal bucklingcaused by the very high temperature, and con-straint of their dilatation, as shown in Fig. 7.This deformation causes a difference in thick-ness between the two braking tracks of the disc.Such a defect needs heavy and costly machin-ing operations to produce quality parts. Instead,

Molten metal

z

x

y

Fig. 2 One-dimensional slice domain for modelingsolidifying plate

Fig. 3 Temperatures through solidifying plate at different times comparing analytical solution and numericalpredictions

Modeling of Stress, Distortion, and Hot Tearing/5

5115/4G/a0005238

process simulation allows the manufacturer totest alternative geometries and process condi-tions in order to minimize the defect.Similar thermomechanical calculations have

been made for plain discs, leading to compari-sons with residual stress measurements bymeansof neutrons and x-ray diffraction (Ref 61). Asshown in Fig. 8, calculations are consistent withmeasurements to within 10 MPa (1.5 ksi).Continuous Casting of Steel Slabs. Ther-

momechanical simulations are used by steel-makers to analyze stresses and strains both inthe mold and in the secondary cooling zone

below. One goal is to quantify the bulging ofthe solidified crust between the supporting rollsthat is responsible for the tensile stress state inthe mushy core, which in turn induces internalcracks and macrosegregation (Ref 62, 63).Two- and three-dimensional finite elementmodels have been recently developed, for theentire length of the caster using THERCAST,as described elsewhere (Ref 40, 64). The con-stitutive models were presented in the section“Governing Equations.” Contact with support-ing rolls is simulated with the penalty formula-tion discussed in the section “Boundary

Conditions: Modeling of Contact Conditions.Multidomain Approaches” adapting penaltycoefficients for the different rolls continuouslyto control numerical penetration of the strand.Figure 9 shows results for a vertical-curved

machine (strand thickness 0.22 m, or 0.75 ft,casting speed 0.9 m/min, or 3 ft/min, materialFe-0.06wt%C) at around 11 m (36 ft) belowthe meniscus. The pressure distribution revealsa double alternation of compressive and depres-sive zones. First, the strand surface is in a com-pressive state under the rolls where the pressurereaches its maximum, 36 MPa (5 ksi). Con-versely, it is in a depressive (tensile) statebetween rolls, where the pressure is minimum(�9 MPa, or �1 ksi). Near the solidificationfront (i.e., close to the solidus isotherm), thestress alternates between tension (negative pres-sure of about �2 MPa, or �0.3 ksi) beneath therolls and compression in between (2–3 MPa, or0.3–0.4 ksi). These results agree with previousstructural analyses of the deformation of thesolidified shell between rolls, such as those car-ried out in static conditions by Wünnenberg andHuchingen (Ref 65), Miyazawa and Schwerdt-feger (Ref 62), or by Kajitani et al. (Ref 66)on small slab sections moving downstreambetween rolls and submitted to the metallurgi-cal pressure onto the solidification front.The influence of process parameters on the

thermomechanical state of the strand can thenbe studied using such numerical models. Anexample is given in Fig. 10, presenting the sen-sitivity of bulging to the casting speed. It canalso be seen that bulging predictions are sensi-tive to the roll pitch, a larger pitch betweentwo sets of rolls inducing an increased bulging.These numerical simulations can then be usedto study possible modifications in the designof continuous casters, such as the replacementof large rolls by smaller ones to reduce thepitch and the associated bulging (Ref 67).

Fig. 4 Transverse (Y and Z ) stress through solidifying plate at different times comparing analytical solution andnumerical predictions

Fig. 5 Finite element meshes of the different domains: part, core, and two half molds

6/Modeling of Stress, Distortion, and Hot Tearing

5115/4G/a0005238

Fig. 6 Temperature evolution in the part at different points located in the indicated section

Fig. 7 Deformation of core blades in a radial section, after a few seconds of cooling. On the left, displacements are magnified (100�). The temperature distribution issuperimposed. On the right, the difference in thickness between the two braking tracks is shown.

70 MPa(10 ksi)

−40 ± 20 MPa(−6 ± 3 ksi)

Measured σθθ

Calculated σθθ 10 MPa(1.5 ksi) Calculated σrr

Measured σrr20 ± 22 MPa

(3 ± 3 ksi)−25 ± 20 MPa

(−4 ± 3 ksi)

−10 MPa(−1.5 ksi)

60 ± 8 MPa(9 ± 1 ksi)

−30 MPa(−4 ksi)

Fig. 8 Residual hoop stresses (left) and radial stresses (right) in a radial section on as-cast plain discs made of gray iron. Top line, calculated values; bottom line, measured values

Modeling of Stress, Distortion, and Hot Tearing/7

5115/4G/a0005238

Fig. 9 Predictions in the middle of the secondary cooling zone, about 11 m (36 ft) below the meniscus. The finite element mesh, (top left) features a fine band of 20 mm (0.8 in.).The pressure distribution (right) reveals alternating stress, including tension near the solidification front (the mushy zone is materialized by 20 lines separated by an interval

Dgl = 0.05).

Fig. 10 Slab bulging calculated at two different casting speeds: 0.9 and 1.2 m/min (3 and 4 ft/min). The slab bulging increases with the casting speed. Source: Ref 67

8/Modeling of Stress, Distortion, and Hot Tearing

5115/4G/a0005238

Hot-Tearing Analysis

Hot-tear crack formation is one of the mostimportant consequences of stress during solidifi-cation. Hot tearing is caused by a combination oftensile stress and metallurgical embrittlement. Itoccurs at temperatures near the solidus whenstrain concentrates within the interdendritic liq-uid films, causing separation of the dendritesand intergranular cracks at very small strains(on the order of 1%). This complex phenomenondepends on the ability of liquid to flow throughthe dendritic structure to feed the volumetricshrinkage, the strength of the surrounding den-dritic skeleton, the grain size and shape, thenucleation of supersaturated gas into pores orcrack surfaces, the segregation of solute impuri-ties, and the formation of interfering solid precip-itates. The subsequent refilling of hot tears withsegregated liquid alloy can cause internal defectsthat are just as serious as exposed surface cracks.The hot tearing of aluminum alloys is reviewedelsewhere (Ref 68). Hot-tearing phenomena aretoo complex, too small-scale, and insufficientlyunderstood tomodel in detail, so several differentcriteria have been developed to predict hot tearsfrom the results of a thermomechanical analysis.Thermal Analysis Criteria. Casting condi-

tions that produce faster solidification andalloys with wider freezing ranges are moreprone to hot tears. Thus, many criteria are solelybased on thermal analysis. Clyne and Daviessimply compare the local time spent betweentwo critical solid fractions gs1 and gs2 (typically0.9 and 0.99, respectively), with the total localsolidification time (or a reference solidificationtime) (Ref 69). The “hot-cracking susceptibility”is defined as:

HCSClyne ¼ t0:99 � t0:90t0:90 � t0:40 (Eq 27)

Classical Mechanics Criteria. Criteriabased on classical mechanics often assumecracks will form when a critical stress isexceeded, and they are popular for predictingcracks at lower temperatures (Ref 70–73). Thiscritical stress depends greatly on the local tem-perature and strain rate. Its accuracy relies onmeasurements, such as the submerged split-chilltensile test for hot tearing (Ref 74–76).Measurements often correlate hot-tear forma-

tion with the accumulation of a critical level ofmechanical strain while applying tensile load-ing within a critical solid fraction where liquidfeeding is difficult. This has formed the basisfor many hot-tearing criteria. That of Yamanakaet al. (Ref 77) accumulates inelastic deforma-tion over a brittleness temperature range, whichis defined, for example as gs 2 ½0:85; 0:99� fora Fe-0.15wt%C steel grade. The local conditionfor fracture initiation is then:Xgs2

gs1�ein � ecr (Eq 28)

in which the critical strain ecr is 1.6% at a typi-cal strain rate of 3 � 10�4 s�1. Careful mea-surements during bending of solidifying steelingots have revealed critical strains ranging

from 1 to 3.8% (Ref 77, 78). The lowest valueswere found at high strain rate and in crack-sen-sitive grades (e.g., high-sulfur peritectic steel)(Ref 77). In aluminum-rich Al-Cu alloys, criti-cal strains were reported from 0.09 to 1.6%and were relatively independent of strain rate(Ref 79). Tensile stress is also a requirementfor hot-tear formation (Ref 77). The maximumtensile stress occurs just before formation ofa critical flaw (Ref 79).The critical strain decreases with increasing

strain rate, presumably because less time isavailable for liquid feeding, and also decreasesfor alloys with wider freezing ranges. Won et al.(Ref 80) suggested the following empiricalequation for the critical strain in steel, basedon fitting measurements from many bend tests:

ecr ¼ 0:02821_e0:3131�T 0:8638B

(Eq 29)

where _e is the strain rate and DTB is the brittletemperature range, defined between the temper-atures corresponding to solid fractions of 0.9and 0.99.Mechanistically Based Criteria. More

mechanistically based hot-tearing criteriainclude more of the local physical phenomenathat give rise to hot tears. Feurer (Ref 81) andmore recently Rappaz et al. (Ref 82) have pro-posed that hot tears form when the local inter-dendritic liquid feeding rate is not sufficient tobalance the rate of tensile strain increase acrossthe mushy zone. The criterion of Rappaz et al.predicts fracture when the strain rate exceedsa limit value that allows pore cavitation to sep-arate the residual liquid film between thedendrites:

_e � 1R

�22 rTk k180ml

rL

rS

ðpm � pCÞ

�vTr

S� r

L

rS

H

ðEq 30Þ

in which ml is the dynamic viscosity of the liq-uid phase, l2 is the secondary dendrite armspacing, pm is the local pressure in the liquidahead of the mushy zone, pC is the cavitationpressure, and vT is the velocity of the solidifica-tion front. The quantities R and H depend onthe solidification path of the alloy:

R ¼ðT1T2

g2sF ðT Þgl3

dT

H ¼ðT1T2

g2sg2l

dT

F ðT Þ ¼ 1rTk kðTT2

gsdT ðEq 31Þwhere the integration limits are calibrationparameters that also have physical meaning(Ref 83). The upper limit T1 may be the liquidusor the coherency temperature, while the lowerlimit T2 typically is within the solid fractionrange of 0.95 to 0.99 (Ref 84).Case Study: Billet Casting Speed Optimi-

zation. A Lagrangian model of temperature, dis-tortion, strain, stress, hot tearing has been appliedto predict the maximum speed for continuouscasting of steel billets without forming off-corner

internal cracks. The two-dimensional transientfinite-element thermomechanicalmodel,CON2D(Ref 20, 21), has been used to track a transverseslice through the solidifying steel strand as itmoves downward at the casting speed to revealthe entire three-dimensional stress state. Thetwo-dimensional assumption produces reasonabletemperature predictions because axial (z-direc-tion) conduction is negligible relative to axialadvection (Ref 50). In-plane mechanical predic-tions are also reasonable because bulging effectsare small and the undiscretized casting directionis modeled with the appropriate condition of gen-eralized plain strain. Other applications with thismodel include the prediction of ideal taper of themold walls (Ref 85) and quantifying the effect ofsteel grade on oscillation mark severity duringlevel fluctuations (Ref 86).The model domain is an L-shaped region of

a two-dimensional transverse section, shownin Fig. 11. Removing the central liquid regionsaves computation and lessens stability prob-lems related to element “locking.” Physically,this “trick” is important in two-dimensionaldomains because it allows the liquid volumeto change without generating stress, whichmimics the effect of fluid flow into and out ofthe domain that occurs in the actual open-topped casting process. Simulations start at themeniscus, 100 mm (4 in.) below the mold top,and extend through the 800 mm (32 in.) longmold and below, for a caster with no submoldsupport. The instantaneous heat flux, given inEq 32, was based on plant measurements (Ref45). It was assumed to be uniform around theperimeter of the billet surface in order to simu-late ideal taper and perfect contact between theshell and mold. Below the mold, the billet

Fig. 11 Model domain

Modeling of Stress, Distortion, and Hot Tearing/9

5115/4G/a0005238

surface temperature was kept constant at its cir-cumferential profile at mold exit. This elimi-nates the effect of spray cooling practiceimperfections on submold reheating or coolingand the associated complication for the stress/strain development. A typical plain carbon steelwas studied (0.27% C, 1.52% Mn, 0.34% Si)with 1500.7 C (2733 F) liquidus temperature,and 1411.8 C (2573 F) solidus temperature.Constitutive equation and properties are givenin the sections “Solid-State Constitutive Models”and “Example of Solid-State ConstitutiveEquations.”

qðMW=m2Þ ¼ 5� 0:2444tðsÞ t � 1:0 s4:7556tðsÞ�0:504 t > 1:0 s

�(Eq 32)

Sample results are presented here for one-quarter of a 120 mm2s (0.2 in.2) billet cast atspeeds of 2.0 and 5.0 m/min (6.5 to 16.5 ft/s).The latter is the critical speed at which hot-tearcrack failure of the shell is just predicted tooccur. The temperature and axial (z) stress dis-tributions in a typical section through the wideface of the steel shell cast at 2.0 m/min (6.5 ft/s)are shown in Fig. 12 and 13 at four differenttimes during cooling in the mold. Unlikethe analytical solution in Fig. 3, the surface tem-perature drops as time progresses. The correspon-ding stress distributions are qualitatively similarto the analytical solution (Fig. 4). The stressesincrease with time, however, as solidificationprogresses. The realistic constitutive equationsproduce a large region of tension near the solidi-fication front. The magnitude of these stresses(as well as the corresponding strains) is notpredicted to be enough to cause hot tearing inthe mold, however. The results from two differ-ent codes reasonably match, demonstrating thatthe formulations are accurately implemented,convergence has been achieved, and the meshand time-step refinement are sufficient.Figure 14(a) shows the distorted temperature

contours near the strand corner at 200 mm(8 in.) below the mold exit, for a casting speedof 5.0 m/min (16.5 ft/min). The corner region iscoldest, owing to two-dimensional cooling. Theshell becomes hotter and thinner with increas-ing casting speed, owing to less time in themold. This weakens the shell, allowing it tobulge more under the ferrostatic pressure belowthe mold.Figure 14(b) shows contours of “hoop” stress

constructed by taking the stress component act-ing perpendicular to the dendrite growth direc-tion, which simplifies to sx in the lower rightportion of the domain and sy in the upper leftportion. High values appear at the off-cornersubsurface region, due to a hinging effect thatthe ferrostatic pressure over the entire faceexerts around the corner. This bends the shellaround the corner and generates high subsur-face tensile stress at the weak solidificationfront in the off-corner subsurface location. Thistensile stress peak increases slightly and movestoward the surface at higher casting speed.Stress concentration is less, and the surfacehoop stress is compressive at the lower casting

speed. This indicates no possibility of surfacecracking. However, tensile surface hoop stressis generated below the mold at high speed inFig. 14(b) at the face center due to excessivebulging. This tensile stress, and the accompa-nying hot-tear strain, might contribute to longi-tudinal cracks that penetrate the surface.Hot tearing was predicted using the criterion

in Eq 28 with the critical strain given in Eq 29.Inelastic strain was accumulated for the compo-nent oriented normal to the dendrite growth

direction, because that is the weakest directionand corresponds to the measurements used toobtain Eq 29. Figure 14(c) shows contours ofhot-tear strain in the hoop direction. The high-est values appear at the off-corner subsurfaceregion in the hoop direction. Moreover, signifi-cantly higher values are found at higher castingspeeds. For this particular example, hot-tearstrain exceeds the threshold at 12 nodes, alllocated near the off-corner subsurface region.This is caused by the hinging mechanism

Distance to chilled surface, in.

0.2 0.4 0.6

2800

2600

2400

2200

2000

Tem

pera

ture

, �F

Tem

pera

ture

, �C

Abaqus 1 s

CON2D 1 s

Abaqus 5 s

CON2D 5 s

Abaqus 10 s

CON2D 10 s

Abaqus 21 s

CON2D 21 s

10000 1 2 3 4 5 6 7 8

Distance to chilled surface, mm

9 10 11 12 13 14 15

1050

1100

1150

1200

1250

1300

1350

1400

1450

1500

1550

Fig. 12 Temperature distribution along the solidifying slice in continuous casting mold

Distance to chilled surface, in.0.2 0.4

Abaqus 1 s

CON2D 1 s

Abaqus 5 s

CON2D 5 s

Abaqus 10 s

CON2D 10 s

Abaqus 21 s

CON2D 21 s

4

2

0

−2

−4

−6

−8

−10

−12

−14

0.6

0.5

0

−0.5

−1

−1.5

0 1 2 3 4 5 6 7 8Distance to chilled surface, mm

Str

ess,

ksi

Str

ess,

MP

a

9 10 11 12 13 14 15−2

Fig. 13 Lateral (y and z) stress distribution along the solidifying slice in continuous casting mold

10/Modeling of Stress, Distortion, and Hot Tearing

5115/4G/a0005238

around the corner. No nodes fail at the centersurface, in spite of the high tensile stress there.The predicted hot-tearing region matches thelocation of off-corner longitudinal cracksobserved in sections through real solidifyingshells, such as the one pictured in Fig. 15. Thebulged shape is also similar.Results from many computations were used

to find the critical speed to avoid hot-tearcracks as a function of section size and workingmold length, presented in Fig. 16 (Ref 46).These predictions slightly exceed plant prac-tice, which is generally chosen by empiricaltrial and error (Ref 87). This suggests that plantconditions such as mold taper are less thanideal, that other factors limit casting speed, orthose speeds in practice could be increased.The qualitative trends are the same.This quantitative model of hot tearing pro-

vides many useful insights into the continuouscasting process. Larger section sizes are moresusceptible to bending around the corner andso have a lower critical casting speed, resultingin less productivity increase than expected. Thetrend toward longer molds over the past threedecades enables a higher casting speed withoutcracks by producing a thicker, stronger shell atmold exit.

Conclusions

Mechanical analysis of casting processes isgrowing in sophistication, accuracy, and phe-nomena incorporated. Quantitative predictionsof temperature, deformation, strain, stress, andhot tearing in real casting processes are becom-ing possible. Computations are still hamperedby the computational speed and limits of meshresolution, especially for realistic three-dimen-sional geometries and defect analysis.

ACKNOWLEDGMENT

The authors wish to thank the ContinuousCasting Consortium and the National Centerfor Supercomputing Applications at the

University of Illinois, the French Ministry ofIndustry, the French Technical Center of Cast-ing Industries (CTIF), and the companies Arce-lorMittal, Ascometal, Aubert et Duval,Industeel, and PSA Peugeot-Citroën, for sup-port of this work.

REFERENCES

1. G.E. Mase and G.T. Mase, ContinuumMechanics for Engineers, 2nd ed., CRCPress, 1999

2. B.G. Thomas, Q. Yuan, S. Sivaramakrish-nan, T. Shi, S.P. Vanka, and M.B. Assar,Comparison of Four Methods to EvaluateFluid Velocities in a Continuous Casting

Fig. 14 Distorted contours at 200 mm (8 in.) below mold exit. (a) Temperature. (b) Hoop stress. (c) Hot-tear strain

Fig. 15 Off-corner internal crack in breakout shellfrom a 175 mm (7 in.) square bloom

Section size, in.2

Section size, mm2

0.156

5

4

3

2

1

0100 120

CON2DWorking

mold length Plant

2000 data1000 mm (40 in.)700 mm (28 in.)500 mm (20 in.)

1990 data1970 data

140 160 180 200 220 240 260

5

10

15

20

Crit

ical

cas

ting

spee

d, ft

/min

Crit

ical

cas

ting

spee

d, m

/min

0.20 0.25 0.30 0.35 0.40

Fig. 16 Comparison of critical casting speeds, based on hot-tearing criterion (Ref 46), and typical plant practice(Ref 87)

Modeling of Stress, Distortion, and Hot Tearing/11

5115/4G/a0005238

Mold, ISIJ Int., Vol 41 (No. 10), 2001,p 1266–1276

3. Q. Yuan, B. Zhao, S.P. Vanka, and B.G.Thomas, Study of Computational Issues inSimulation of Transient Flow in Continu-ous Casting, Steel Res. Int., Vol 76(No. 1, Special Issue: Simulation of FluidFlow in Metallurgy), 2005, p 33–43

4. Q. Yuan, S. Sivaramakrishnan, S.P. Vanka,and B.G. Thomas, Computational andExperimental Study of Turbulent Flow ina 0.4-Scale Water Model of a ContinuousSteel Caster, Metall. Mater. Trans., Vol35B (No. 5), 2004, p 967–982

5. C. Beckermann, Macrosegregation, inEncyclopedia of Materials: Science andTechnology, Elsevier, 2001, p 4733–4739

6. G. Lesoult, C.-A. Gandin, and N.T. Niane,Segregation During Solidification withSpongy Deformation of the Mushy Zone,Acta Mater., Vol 51, 2003, p 5263–5283

7. L.C. Nicolli, A. Mo, and M. M’Hamdi,Modeling of Macrosegregation Caused byVolumetric Deformation in a CoherentMushy Zone, Metall. Mater. Trans. A, Vol36, 2005, p 433–442

8. V.D. Fachinotti, S. Le Corre, N. Triolet, M.Bobadilla, and M. Bellet, Two-PhaseThermo-mechanical and MacrosegregationModelling of Binary Alloys Solidificationwith Emphasis on the Secondary CoolingStage of Steel Slab Continuous Casting,Int. J. Num. Meth. Eng., Vol 67 (No. 10),2006, p 1341–1384

9. O. Ludwig, C.-L. Martin, J.-M. Drezet, andM. Suéry, Rheological Behaviour of Al-CuAlloys During Solidification: ConstitutiveModelling, Experimental Identificationand Numerical Study, Met. Mater. Trans.,Vol 36A, 2005, p 1525–1535

10. Y. Estrin, A Versatile Unified ConstitutiveModel Based on Dislocation Density Evo-lution, in Constitutive Modelling—Theoryand Application, Vol MD-vol26/AMD-vol121, ASME, 1991, p 65–75

11. C. Agelet de Saracibar, M. Cervera, and M.Chiumenti, On the Constitutive Modelingof Coupled Thermomechanical Phase-Change Problems, Int. J. Plast., Vol 17,2001, p 1565–1622

12. J. Sengupta, S.L. Cockcroft, D.M. Maijer,and A. Larouche, Quantification of Tem-perature, Stress, and Strain Fields Duringthe Start-up Phase of Direct Chill CastingProcess by Using a 3D Fully CoupledThermal and Stress Model for AA5182Ingots, Mater. Sci. Eng. A, Vol 397, 2005,p 157–177

13. J.E. Kelly, K.P. Michalek, T.G. O’Connor,B.G. Thomas, and J.A. Dantzig, InitialDevelopment of Thermal and Stress Fieldsin Continuously Cast Steel Billets, Metall.Trans. A, Vol 19A (No. 10), 1988, p2589–2602

14. A.E. Huespe, A. Cardona, and V. Fachi-notti, Thermomechanical Model of a Con-tinuous Casting Process, Comput. Meth.

Appl. Mech. Eng., Vol 182 (No. 3), 2000,p 439–455

15. MARC Users Manual, MARC AnalysisResearch Corp., Palo Alto, CA, 1991

16. “ABAQUS Theory Manual v6.0,” Abaqus,Inc., Pawtucket, RI, 2004

17. A. Hakonsen and D. Mortensen, Modelingof Casting, Welding and Advanced Solidifi-cation Processes VII, TMS, Warrendale,PA, 1995, p 963–970

18. H.G. Fjaer and A. Mo, ALSPEN—A Math-ematical Model for Thermal Stresses inDC-Cast Al Billets, Metall. Trans., Vol21B (No. 6), 1990, p 1049–1061

19. G. Laschet, J. Jakumeit, and S. Benke,Thermo-mechanical Analysis of Cast/Mould Interaction in Casting Processes, Z.Metallkd., Vol 95, 2004, p 1087–1096

20. C. Li andB.G.Thomas, Thermo-MechanicalFinite-Element Model of Shell Behavior inContinuous Casting of Steel,Metall. Mater.Trans. B, Vol. 35B (No. 6), 2004, p 1151–1172

21. S. Koric and B.G. Thomas, EfficientThermo-Mechanical Model for Solidifica-tion Processes, Int. J. Num. Meth. Eng.,Vol 66 (No. 12), 2006, p 1955–1989

22. MAGMASOFT, http://www.magmasoft.com, 2007

23. PROCAST, Procast, http://www.esi-group.com/SimulationSoftware/Die_Casting_So-lution/Products/ Casting_simulation/, 2007

24. M. Samonds and J.Z. Zhu, CoupledThermal-Fluid-Stress Analysis of Castings,Proc. MCWASP IX, Ninth Int. Conf. onModeling of Casting, Welding andAdvanced Solidification Processes, P.R.Sahm, P.N. Hansen, and J.G. Conley, Ed.(Aachen, Germany), Aug 20–25, 2000,Shaker Verlag, Aachen, 2000, p 80–87

25. J.M. Drezet, A. Bughardt, H.G. Fjaer, andB. Magnin, Thermomechanical Effects inD.C. Casting of Aluminum Alloy: ANumerical Benchmark Study, Mater. Sci.Forum, Vol 329–330, 2000, p 493–500

26. H.G. Fjaer and A. Hakonsen, The Mecha-nism of Pull-in During DC-Casting of Alu-minum Sheet Ingots, Light Metals 1997,TMS, Warrendale, PA, 1997, p 683–690

27. A. Mo, M. Rappaz, and L.L. Martin, Alumi-num, Vol 78 (No. 10), 2002, p 856–864

28. J.M. Drezet, A. Bughardt, H.G. Fjaer, andB. Magnin, Mater. Sci. Forum, Vol 329–330, 2000, p 493–500

29. A. Henry, Light Metal Age, Vol 58 (No. 7–8), 2000, p 66–67

30. M. Bellet, O. Jaouen, and I. Poitrault, AnALE-FEM Approach to the Thermomecha-nics of Solidification Processes with Appli-cation to the Prediction of Pipe Shrinkage,Int. J. Num. Meth. Heat Fluid Flow, Vol15, 2005, p 120–142

31. O.C. Zienkiewicz and R.L. Taylor, TheFinite Element Method, 4th ed., McGrawHill, 1988

32. J.-M. Drezet, B. Commet, H.G. Fjaer, andB. Magnin, Stress-Strain Computations of

the DC Casting Process of AluminumAlloy: A Sensitivity Study on MaterialProperties, Proc. MCWASP IX, Ninth Int.Conf. on Modeling of Casting, Weldingand Advanced Solidification Processes,P.R. Sahm, P.N. Hansen, and J.G. Conley,Ed., (Aachen, Germany), Aug 20–25,2000, Shaker Verlag, Aachen, 2000, p 33–40

33. M. Bellet and V.D. Fachinotti, ALEMethod for Solidification Modelling, Com-put. Meth. Appl. Mech. Eng., Vol 193,2004, p 4355–4381

34. Thercast, presentation, www.transvalor.com and www.scconsultants.com, 2007

35. M. Chiumenti, M. Cervera, and C.A.d. Sar-acibar, Coupled Thermomechanical Simu-lation of Solidification and Cooling Phasesin Casting Processes, Proc. MCWASP XI,11th Int. Conf. on Modeling of Casting,Welding and Advanced Solidification Pro-cesses, C.-A. Gandin and M. Bellet, Ed.,TMS, Warrendale, PA, 2006, p 201–208

36. P. Kozlowski, B.G. Thomas, J. Azzi, and H.Wang, Simple Constitutive Equations forSteel at High Temperature, Metall. Trans.A, Vol 23A (No. 3), 1992, p 903–918

37. P.J. Wray, Plastic Deformation of Delta-Ferritic Iron at Intermediate Strain Rates,Metall. Trans. A, Vol 7A, Nov 1976,p 1621–1627

38. P.J. Wray, Effect of Carbon Content on thePlastic Flow of Plain Carbon Steels at Ele-vated Temperatures, Metall. Trans. A, Vol13A (No. 1), 1982, p 125–134

39. T. Suzuki, K.H. Tacke, K. Wunnenberg,and K. Schwerdtfeger, Creep Properties ofSteel at Continuous Casting Temperatures,Ironmaking Steelmaking, Vol 15 (No. 2),1988, p 90–100

40. F. Costes, A. Heinrich, and M. Bellet, 3DThermomechanical Simulation of the Sec-ondary Cooling Zone of Steel ContinuousCasting, Proc. MCWASP X, Tenth Int.Conf. on Modeling of Casting, Weldingand Advanced Solidification Processes,D.M. Stefanescu, J.A. Warren, M.R. Jolly,and M.J.M. Krane, Ed., TMS, Warrendale,PA, 2003, p 393–400

41. D.L. Donsbach and M.W. Moyer, Ultra-sonic Measurement of Elastic Constants atTemperatures from 20 to 1100 C, Ultra-sonic Materials Characterization, SpecialPub. 596, H. Berger and M. Linzer, Ed.,National Bureau of Standards, 1980

42. O.M. Puhringer, Strand Mechanics forContinuous Slab Casting Plants, StahlEisen, Vol 96 (No. 6), 1976, p 279–284

43. D.R. Hub, Measurement of Velocity andAttenuation of Sound in Iron up to theMelting Point, Proc. IVth Int. Vong. Acous-tics (Copenhagen), 1962, paper 551

44. H. Mizukami, K. Murakami, and Y. Miya-shita, Mechanical Properties of Continu-ously Cast Steels at High Temperatures,Tetsu-to-Hagane, Vol 63 (No. 146), 1977,p S652

12/Modeling of Stress, Distortion, and Hot Tearing

5115/4G/a0005238

45. C. Li, and B.G. Thomas, Maximum Cast-ing Speed for Continuous Cast Steel BilletsBased on Sub-Mold Bulging Computation,Steelmaking Conf. Proc., Vol 85 (Nash-ville, TN), March 10–13, 2002, ISS, War-rendale, PA, 2002, p 109–130

46. C. Li and B.G. Thomas, Thermo-Mechani-cal Finite Element Model of Bulging andHot Tearing During Continuous Castingof Steel Billets, Modeling of Casting,Welding, and Advanced Solidification Pro-cesses, Vol X, D. Stefanescu, J. Warren,M. Jolly, and M. Krane, Ed. (San Destin,FL), May 25–30, 2003, TMS, Warrendale,PA, 2003, p 385–392

47. K. Harste, A. Jablonka, and K. Schwerdtfe-ger, Shrinkage and Formation of Mechani-cal Stresses During Solidification ofRound Steel Strands, Fourth Int. Conf. onContinuous Casting (Centres de RecherchesMetallurgiques and Verein DeutscherEisenhuttenleute), Stahl und Eisen, Brus-sels, 1988, p 633–644

48. K. Harste, “Investigation of the Shrinkageand the Origin of Mechanical Tension Dur-ing the Solidification and Successive Cool-ing of Cylindrical Bars of Fe-C Alloys,”Ph.D. dissertation thesis, Technical Univer-sity of Clausthal, 1989

49. I. Jimbo and A. Cramb, The Density ofLiquid Iron-Carbon Alloys, Metall. Trans.B, Vol 24B, 1993, p 5–10

50. Y. Meng and B.G. Thomas, Heat Transferand Solidification Model of ContinuousSlab Casting: CON1D, Met. Mater. Trans.,Vol 34B (No. 5), 2003, p 685–705

51. C.V. Madhusudana and L.S. Fletcher, Con-tact Heat Transfer—The Last Decade,AIAA J., Vol 24, 1985, p 510–523

52. M. Rappaz, M. Bellet, and M. Deville,Numerical Modeling in Materials Scienceand Engineering, in Springer Series inComputational Mathematics, Springer Ver-lag, Berlin, 2003

53. S. Nemat-Nasser, and Y.F. Li, An ExplicitAlgorithm for Large-Strain, Large-StrainRate Elastic-Viscoplasticity, Comput.Meth. Appl. Mech. Eng., Vol 48, 1992, p205–219

54. A.M. Lush, G. Weber, and L. Anand, AnImplicit Time-integration Procedure for ASet of Internal Variable Constitutive Equa-tions for Isotropic Elastic-Viscoplasticity,Int. J. Plast., Vol 5, 1989, p 521–549

55. H. Zhu, “Coupled Thermal-MechanicalFinite-Element Model with Application toInitial Solidification,” Thesis, Universityof Illinois, 1993

56. R. Glowinski and P.L. Tallee, AugmentedLagrangian and Operator-SplittingMethods in Non-linear Mechanics, Studiesin Applied Mathematics, SIAM, Vol 9,1989, 295 pages

57. P. Wriggers and G. Zavarise, On Contactbetween Three-Dimensional Beams Under-going Large Deflections, Commun. Num.Meth. Eng., Vol 13, 1997, p 429–438

58. O. Jaouen and M. Bellet, A NumericalMechanical Coupling Algorithm forDeformable Bodies: Application to Part/Mold Interaction in Casting Process, Proc.Eighth Int. Conf. on Modelling of Casting,Welding and Advanced Solidification Pro-cesses, B.G. Thomas and C. Beckermann,Ed. (San Diego, CA), June 7–12, 1998,TMS, Warrendale, PA, 1998, p 739–746

59. J.H. Weiner and B.A. Boley, Elasto-plasticThermal Stresses in a Solidifying Body, J.Mech. Phys. Solids, Vol 11, 1963, p 145–154

60. M. Bellet, C. Aliaga, and O. Jaouen, FiniteElements for a Thermomechanical Analy-sis of Solidification Processes, Modelingof Casting, Welding, and Advanced Solidi-fication Processes IX, Shaker VerlagGmbH, Aachen, 2000, p 10–17

61. S. David and P. Auburtin, Numerical Sim-ulation of Casting Processes. Benefits ofThermomechanical Simulation in Automo-tive Industry, Conference Matériaux 2002,Tours, France, Proc., Vol on CD, Univer-sité Technologique de Belfort-Montbéliard,2002, p 5 (in French)

62. K. Miyazawa and K. Schwerdtfeger,Macrosegregation in Continuously CastSteel Slabs: Preliminary Theoretical Inves-tigation on the Effect of Steady State Bulg-ing, Arch. Eisenh€utten., Vol 52 (No. 11),1981, p 415–422

63. G. Lesoult and S. Sella, Analysis and Pre-vention of Centreline Segregation DuringContinuous Casting of Steel Related toDeformation of the Solid Phase, Solid StatePhenom., Vol 3, 1988, p 167–178

64. M. Bellet and A. Heinrich, A Two-Dimen-sional Finite Element ThermomechanicalApproach to a Global Stress-Strain Analy-sis of Steel Continuous Casting, ISIJ Int.,Vol 44, 2004, p 1686–1695

65. K. Wunnenberg and D. Huchingen, StrandBulging between Supporting Rollers Dur-ing Continuous Slab Casting, Stahl Eisen,Vol 98 (No. 6), 1978, p 254–259

66. T. Kajitani, J.-M. Drezet, and M. Rappaz,Numerical Simulation of Deformation-Induced Segregation in Continuous Castingof Steel, Metall. Mater. Trans. A, Vol 32,2001, p 1479–1491

67. N. Triolet andM. Bobadilla,Mastering SteelSlab Internal Soundness and Surface QualityIssues through Thermomechanical Model-ling of Continuous Casting, Proc.MCWASP-XI, 11th Int. Conf. on Modellingof Casting, Welding, and Advanced Solidifi-cationProcesses,C.A.GandinandM.Bellet,Ed. (Opio, France), May 28 to June 2, 2006),TMS,Warrendale, PA, 2006, p 753–760

68. D.G. Eskin, Suyitno, and L. Katgerman,Mechanical Properties in the Semi-SolidState and Hot Tearing of Aluminum Alloys,Prog. Mater. Sci., Vol 49, 2004, p 629–711

69. T.W. Clyne and G.J. Davies, Comparisonbetween Experimental Data and Theoreti-cal Predictions Relating to Dependence of

Solidification Cracking on Composition,Solidification and Casting of Metals, TheMetals Society, London, 1979, p 275–278

70. K. Kinoshita, T. Emi, and M. Kasai, Ther-mal Elasto-plastic Stress Analysis of Solid-ifying Shell in Continuous Casting Mold,Tetsu-to-Hagane, Vol 65 (No. 14), 1979,p 2022–2031

71. J.O. Kristiansson, Thermal Stresses in theEarly Stage of Solidification of Steel, J.Thermal Stresses, Vol 5, 1982, p 315–330

72. B.G. Thomas, I.V. Samarasekera, and J.K.Brimacombe, Mathematical Model of theThermal Processing of Steel Ingots, PartII: Stress Model, Metall. Trans. B, Vol18B (No. 1), 1987, p 131–147

73. K. Okamura and H. Kawashima, Calcula-tion of Bulging Strain and its Applicationto Prediction of Internal Cracks in Continu-ously Cast Slabs, Proc. Int. Conf. Comp.Ass. Mat. Design Proc. Simul., ISIJ, Tokyo,1993, p 129–134

74. P. Ackermann, W. Kurz, and W. Heine-mann, In Situ Tensile Testing of Solidify-ing Aluminum and Al-Mg Shells, Mater.Sci. Eng. Vol 75, 1985, p 79–86

75. C. Bernhard, H. Hiebert, and M.M. Wolf,Simulation of Shell Strength Properties bythe SSCT Test, ISIJ Int. (Japan), Vol 36(Suppl. Science and Technology of Steel-making), 1996, p S163–S166

76. M. Suzuki, C. Yu, and T. Emi, In-SituMeasurement of Tensile Strength of Solid-ifying Steel Shells to Predict Upper Limitof Casting Speed in Continuous Caster withOscillating Mold, ISIJ Int., Iron Steel Inst.Jpn., Vol 37 (No. 4), 1997, p 375–382

77. A. Yamanaka, K. Nakajima, K. Yasumoto,H. Kawashima, and K. Nakai, Measure-ment of Critical Strain for SolidificationCracking, Modelling of Casting, Welding,and Advanced Solidification Processes,Vol V, M. Rappaz, M.R. Ozgu, and K.W.Mahin, Ed., (Davos, Switzerland), TMS,Warrendale, PA, 1990, p 279–284

78. A.Yamanaka,K.Nakajima, andK.Okamura,Critical Strain for Internal Crack FormationinContinuousCasting, Ironmaking Steelmak-ing, Vol 22 (No. 6), 1995, p 508–512

79. P. Wisniewski and H.D. Brody, TensileBehavior of Solidifying Aluminum Alloys,Modelling of Casting, Welding, andAdvanced Solidification Processes, Vol V,M. Rappaz, M.R. Ozgu, and K.W. Mahin,Ed., (Davos, Switzerland), TMS, Warren-dale, PA, 1990, p 273–278

80. Y.-M. Won, T.J. Yeo, D.J. Seol, and K.H.Oh, A New Criterion for Internal CrackFormation in Continuously Cast Steels,Metall. Mater. Trans. B, Vol 31B, 2000, p779–794

81. U. Feurer, Mathematisches modell derWarmrissneigung von binären aluminiumlegierungen, Giessereiforschung, Vol 28,1976, p 75–80

82. M. Rappaz, J.-M. Drezet, and M. Gremaud,A New Hot-Tearing Criterion, Metall.

Modeling of Stress, Distortion, and Hot Tearing/13

5115/4G/a0005238

Mater. Trans. A, Vol 30A (No. 2), 1999,p 449–455

83. J.M. Drezet and M. Rappaz, Prediction ofHot Tears in DC-Cast Aluminum Billets,Light Metals, J.L. Anjier, Ed., TMS, War-rendale, PA, 2001, p 887–893

84. M. M’Hamdi, S. Benum, D. Mortensen, H.G. Fjaer, and J.M. Drezet, The Importanceof Viscoplastic Strain Rate in the Forma-tion of Center Cracks During the Start-upPhase of Direct-Chill Cast Aluminium

Extrusion Ingots, Metall. Mater. Trans. A,Vol 34, 2003, p 1941–1952

85. B.G. Thomas and C. Ojeda, Ideal TaperPrediction for Slab Casting, ISSTech Steel-making Conference (Indianapolis, IN),April 27–30, 2003, Vol 86, 2003, p 396–308

86. J. Sengupta and B.G. Thomas, Effect ofa Sudden Level Fluctuation on Hook For-mation During Continuous Casting ofUltra-Low Carbon Steel Slabs, Modeling

of Casting, Welding, and Advanced Solidi-fication Processes XI (MCWASP XI) Con-ference, C.Z. Gandin, M. Bellet, and J.E.Allison, Ed. (Opio, France, May 28 to June2, 2006), TMS, Warrendale, PA, 2006, p727–736

87. E. Howard and D. Lorento, Developmentof High Speed Casting, 1996 Electric Fur-nace Conference Proceedings, (Dallas,TX), Dec 9–12, 1996, ISS, Warrendale,PA, 1996

14/Modeling of Stress, Distortion, and Hot Tearing

5115/4G/a0005238