Computers and Mathematics with Applications 59 (2010) 585594

Contents lists available at ScienceDirect

Computers and Mathematics with Applications

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

Finite element analysis of the steel quenching process: Temperature

field and solidsolid phase change
P. Carlone , G.S. Palazzo, R. Pasquino
Department of Mechanical Engineering, University of Salerno, Via Ponte Don Melillo 1, Fisciano, 84084, Italy

article

info

Article history:
Received 24 April 2009
Received in revised form 27 May 2009
Accepted 10 June 2009
Keywords:
Quenching process
Finite element analysis
Temperature
Phase change

abstract
This paper deals with a finite element analysis of the steel quenching process; regarding the
transient temperature field and the thermally induced solidsolid phase transformations.
All the process steps, i.e. heating, holding, and cooling, have been considered, modeling
both the austenite formation and decomposition and taking into account nucleation and
growth processes. The final hardness distribution into the quenched sample has been
predicted according to the rule of mixtures, taking into account the chemical composition
of the processing material, the final distribution of each phase and the local cooling rate.
2009 Elsevier Ltd. All rights reserved.

1. Introduction
The quenching process is a heat treatment widely used to improve the mechanical properties of steel products, such as
hardness, stiffness, and strength, by the means of opportune solidsolid phase changes, induced by a heating, holding, and
cooling thermal cycle. The main purpose of the heating and the holding stages is the transformation of the starting material
structure into a homogeneous austenitic phase, while, during the last stage of the process, forced cooling of the workpiece is
used to induce the opportune decomposition of austenite into several microstructures, such as martensite, pearlite, bainite,
ferrite, and Fe-carbide, depending on the chemical composition of the processing steel and the local cooling rate.
The timing of each step, as well as the heating and cooling rates and the holding temperature, are very important.
Indeed, relatively high heating rates could induce excessive temperature gradients into the workpiece, resulting into internal
stresses, deformations, or cracks, and into non-homogeneous austenite formation; the temperature and duration of the
holding stage should be carefully planned to obtain a fully austenitic structure, avoiding excessive grain size, and reducing
energy consumption and costs, due to extended holding at relatively high temperature. Finally, the opportune planning of
the cooling stage, relative to the choice of the cooling medium (water, quench oils, polymer solutions), cooling temperature,
and cooling technique (bath or spray), is crucial to induce the austenite decomposition into desired microstructures,
reducing residual stresses and part distortions.
Actually the process design is mainly based on industrial experience and on trial and error procedures, often leading to
non-optimal products, which require expensive destructive or non-destructive tests.
The effect of the specific thermal cycle on the hardness of the workpiece can be investigated using several hardness and
microhardness tests; however these tests are generally destructive, require a specific preparation of the workpiece, and
they have to be used on statistical base resulting sometimes in excessive scrap. Non-destructive methods, mainly based
on ultrasonic wave propagation into the quenched part, can be effectively used to evaluate the elastic properties of the
processed material, however, if the whole stress strain diagram is required, the tensile test on an opportune specimen is

Corresponding author. Tel.: +39 089 964320; fax: +39 089 964003.
E-mail address: [email protected] (P. Carlone).

0898-1221/$ see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.camwa.2009.06.006

586

P. Carlone et al. / Computers and Mathematics with Applications 59 (2010) 585594

needed. Other non-destructive techniques are available to investigate the residual stress status in quenched parts, such
as X-ray diffraction, neutron diffraction, and magneto-acoustic methods, which are mainly restricted to the surface of the
material. A complete investigation can be obtained using mechanical techniques, such as hole drilling or material sectioning.
In recent years, a remarkable interest has been focused on the analysis and optimization of the quenching process, in
order to obtain the desired phase changes, to reduce distortion and residual stress, to develop relatively shorter and less
expensive thermal cycles. Computational analysis, based on the finite element, finite difference, or finite volume methods,
is more and more used to effectively investigate and optimize complex manufacturing processes and thermal treatments,
allowing the evaluation of properties not always experimentally measurable [14]. In the literature several approaches
to quenching process modeling can be found. A finite element two-dimensional procedure for the FEA of the quenching
process of a 1080 cylinder has been proposed by Woodard et al. in [5], evidencing the relevance of the latent heat due to
phase change. The same benchmark has been used for the validation of the models proposed in [6] by Huiping et al. and
in [7,8] by Kang et al. A FE model of the Jominy end-quench test for the evaluation of the transient temperature, including
phase transformations, has been proposed and tested in [9] by Homberg. However, the aforesaid works are restricted only
to the cooling stage of the quenching process, assuming the initial structure as being totally austenitic. No consideration has
been reported on the specific heating and holding cycle to be followed.
This paper deals with the modeling and the computational analysis of the transient temperature field and the microstructure transformations related to the quenching process of a eutectoid plain carbon steel. The literature on mathematical
aspects related to the topics of this paper offers various useful hints. For instance, interpolation problems for elliptic equations are dealt with in [10] as a tool to be used to refine the approximation of solutions, while convergence results are
dealt with in [11,12]. All the process stages, i.e. the heating, the holding, and the cooling stages, have been considered,
modeling both austenite formation and successive decomposition into product phases. Nucleation and growth processes of
diffusive, as well as displacive transformations have been modeled according to the Scheils Additivity Rule and the JohnsonMehlAvramiKomogorov model, using a fictitious time procedure. The Yu parabolic model has been adopted to handle
the diffusionless austenite-martensite decomposition. In Section 2, the statement of the considered problem, including the
considered balance equation, the phase transformation models, as well as the numerical procedure are described, while
Section 3 deals with verification and validation of the proposed model and with some numerical results, outlining relevant
conclusions.
2. Process modeling
2.1. Introduction
This Section deals with the statement of the initial-boundary value problem. The complexity of the quenching process
modeling is related to the reciprocal coupling between the heat transfer, the deformation, and the solidsolid phase
transformations, schematically shown in Fig. 1, following a diagram proposed by other authors [6,8]. The microstructure
changes are temperature and stress dependent, however, solidsolid phase transformations imply enthalpy variation and
the different structures are characterized by different mechanical and physical properties, influencing the temperature
distribution, and by different volumes, resulting into mechanical strains. Transformation induced plasticity phenomena,
characterized by plastic strains developed at stresses below the yield stress, can occur during solidsolid phase changes,
depending on the transformation rate and instantaneous deviatoric stress state. The temperature and stress-strain
distributions are also coupled taking into account that thermal gradients into the processing part result in different thermal
expansions and then in the generation of stresses which can lead to plastic flow and part distortions; finally plastic
deformations are characterized by heat generation, influencing the temperature distribution.
In the present model only the main coupling factors have been considered; in particular, the effects of transformation
plasticity, phase change strain increment, and stress induced phase change, have been left for further analysis. Taking into
account that the heat generated by the plastic deformation occurring into the processing material is negligible with respect
to the heat transfer related to the applied thermal cycle, a sequential solution procedure has been adopted: the temperature
transient field, due to the applied thermal cycle, to the transformation latent heat and to heat diffusion, has been used as a
thermal load to evaluate the transient and residual stress-strain distribution and to evaluate eventual geometrical distortions
of the processing part. This paper is focused on the analysis of thermal and microstructural aspects; deformation analysis
will be described in a successive paper. The considered boundary value problem has been solved by the means of opportune
subroutines implemented into the finite element commercial package ANSYS.
2.2. Thermal model
Assuming the processing material as isotropic and considering that there are no mass transport phenomena, the
conservation of the thermal energy, according to the first law of thermodynamics and the Fourier model, writes as follows:

T
= (K T ) + ugen ,
t

(1)

where T is the temperature, t is the time, and c are, respectively, the temperature dependent material density and specific

P. Carlone et al. / Computers and Mathematics with Applications 59 (2010) 585594

587

HEAT
TRANSFER

Latent heat
Physical properties
variation

Heat
generation
Thermal
strain

Temperature induced
transformation
Stress induced
transformation

STRESS
STRAIN

PHASE
CHANGE
Transformation induced plasticity
Mechanical properties variation

Fig. 1. Metallographic, thermal, and mechanical coupling effects in the quenching process.

heat capacity, K is the thermal conductivity, and ugen is the internal heat generation rate. Multiplying both members of
Eq. (1) by a virtual temperature increment T and integrating on the considered domain V , defined by the boundary surface
S, yields:

c
V

T
T dV =
t

(K T ) T dV +

ugen T dV .

(2)

Taking into account that:

( TK (T )) = T (K T ) + T (K T ) ,

(3)

it follows that:

c
V

T
T dV =
t

( TK (T )) dV

T (K T ) dV +
V

ugen T dV .

(4)

Rearranging Eq. (4), indicating as qn the thermal flow acting over the surface S characterized by the unit outward normal
vector n, and taking into account the divergence theorem, one finally has:

c
V

T
T dV +
t

T (K T ) dV =
V

Tqn dS +
S

ugen T dV ,

(5)

which represents the Virtual Temperature Principle, solved, in the present investigation, using a finite element scheme,
according to the following initial and boundary conditions and loads:
- the initial temperature of the processing part is assumed as uniform and equal to room temperature;
- the heating and holding stages have been modeled by an imposed temperature on the external surfaces of the processing
workpiece, according to the specific thermal cycle;
- the cooling stage has been modeled considering convective and radiative heat fluxes, particularized on the quenched and
non-quenched surfaces of the processing part by the means of a unique heat transfer coefficient, used to describe both
fluxes;
- the latent heat due to solid phase transformations has been taken into account and included into the energy balance
equation, using the following heat generation rate form:
ugen =

Hi,j

Fi,j
,
tj

(6)

where Hi,j is the heat of transformation of the phase i formed during the time step j, at the temperature Tj ; Fi,j is the
volume fraction of the transformed i phase, relatively to the same time step, and ti,j is duration of the time step j.
2.3. Phase change model
The quenching process is characterized by remarkable temperature variations, leading to microstructure transformations
into the workpiece. At each temperature, after the completeness of the nucleation and growth processes, stable microstructures can be observed, as reported on the well known FeFe3 C equilibrium diagram. During the heating and holding stages,
as the local temperature exceeds the A1c temperature (the subscript 1 refers to the pearliteaustenite transformation and the
subscript c specify the heating stage), austenite formation from the preexisting structures is observed; on the other hand, the
forced cooling leads to austenite decomposition into martensite, pearlite, bainite, ferrite, and iron carbide, whose formation
is strongly influenced by the chemical composition of the steel and the local cooling rate. Different transformation mechanisms, on heating as well as on cooling, have been proposed and investigated, regarding the formation of a specific product

588

P. Carlone et al. / Computers and Mathematics with Applications 59 (2010) 585594

phase from a parent phase, considering both the cases of single phase formation, as well as simultaneous and interacting
decomposition [1324]. A detailed discussion of the transformation mechanisms is behind the aim of this work, however,
it will be useful to recall that solidsolid phase changes are characterized by nucleation and growth processes, the former
defined by the time needed to obtain a detectable volume fraction of the product phase, the latter characterized by the progressive decomposition of the parent phase into the product phase. It is generally accepted that some phase transformations,
such as the austenite decomposition into ferrite, pearlite, or bainite, and vice versa, are diffusion controlled, temperature
and time dependent and are strongly influenced by the nucleation and growth mechanisms. The austenite decomposition
into martensite, and vice versa, are characterized by negligible nucleation time and diffusionless growth mechanisms, resulting in being very fast and, with optimal approximation, only temperature dependent. It is also relevant to underline [25]
that the well know Time-Temperature Transformation (TTT) diagrams, or, using a more actual terminology, the Isothermal
Transformation (IT) diagrams, should be correctly interpreted as being composed of at least two separate and partially overlapped C curves, describing, respectively, diffusional transformations (austenite-equiaxed ferrite and austenite-pearlite on
cooling) and displacive transformations (austeniteWidmansttten ferrite and austenitebainite on cooling), characterized,
therefore, by specific nucleation and growth processes. In isothermal conditions, the amount of the product phase, formed
according the diffusional or displacive phase changes, can be evaluated according to the JohnsonMehlAvramiKomogorov
model, as follows:
Fi (T ) = 1.0 exp a(T ) t (T )n(T ) ,

(7)

where Fi is the volume fraction of the product phase i, t is the time elapsed since the transformation beginning, i.e. after the
nucleation process, a and n are material parameters, dependent on the isothermal temperature and the forming phase. The
diffusional coefficient a and the transformation exponent n can be evaluated by the IT diagram of the considered material,
taking into account the time s and f , needed for the conversion, in isothermal conditions, of detectable start and finish
volume fractions, indicated, respectively, as Fs and Ff , of the forming phase, as follows:
a(T ) = ln(Fs )s (T )n(T ) ,

n(T ) =

ln(ln(Ff ) ln(Fs ))
ln(s (T )) ln(f (T ))

(8)

where the subscripts s and f refer to the starting and finishing conditions of the phase change. In the present investigation
Fs and Ff have been assumed, respectively, as 0.01 and 0.99.
Eq. (7) is strictly valid only for isothermal transformation, however the quenching process is characterized by continuous
heating and cooling of the processing material. The technique developed by Tzitzelkov [26] has been adopted to take into
account temperature variations. The temperature-time curves have been discretized into a series of isothermal steps; on
each step of the microstructural model the volume fraction of product phase is calculated using isothermal kinetics and the
Additivity Rule.
The Additivity Rule was proposed by Scheil for the evaluation of the time needed for nucleation and growth of the product
phase under non-isothermal conditions. Several researches have been performed in the past to define the conditions to be
satisfied by a given phase change to be considered additive. In [1719] Avrami suggested that a phase transformation can
be considered additive if it is an isokinetic reaction, i.e. if the rate of nucleation is proportional to the rate of growth over
the considered temperature range. A less restrictive criterion, called site saturation was formulated by Chan [20], who
proposed that the transformation can be considered as additive if the nucleation sites were saturated early in the reaction,
and if the growth rate is related to temperature only. The above criterion states that a phase transformation is additive
if the reaction rate depends only on the formed phase and on the temperature. Kuban et al. [21] proposed the effective
site saturation criterion, which states that a phase transformation is additive if the nuclei formed in the early stages of
the transformation dominate the subsequent transformation kinetics, while the nucleation controlled reaction condition,
proposed by Umemoto et al. [2224], states that the reaction is additive if the growth of nuclei is very fast at the beginning
of the reaction but ceases when the nucleating phase reaches a certain limiting size. More detailed discussions on the Rule
of Additivity and its validity can be found in [27,28] and in the bibliography therein cited. In particular, in [28] Zhua et al.
showed that the rule of additivity does not hold for the incubation time for transformation, but the error caused by using it
is negligible.
According to the modified JMAK model, for each time step j, corresponding to the temperature Tj , the cumulative volume
fraction of the specific i phase, indicated as Fi,j1 , transformed up to the last time step j 1, results in a fictitious time tj,fict ,
which represent the time needed, at the temperature Tj , to obtain the same amount of cumulative transformed phase. Some
manipulations of Eq. (7), allow one to evaluate the fictitious time tj,fict , as follows:

"

tj,fict Tj =

ln(1 Fi,j1 )


a Tj

1
n Tj

( )

(9)

As a consequence, the total permanence time tj at the temperature Tj can be written as the sum of the fictitious time tj,fict
and the duration tj of the time step j, as follows:

"

tj = tj + tj,fict

ln(1 Fi,j1 )


= tj +
a Tj

1
n Tj

( )

(10)

P. Carlone et al. / Computers and Mathematics with Applications 59 (2010) 585594

t+

1 exp (a(Tj+1) tn(Tj+1))

f(Tj+1)

T+

589

1 exp (a(Tj) tn(Tj))

f(Tj)

tj+1

tj+1,fict

F
1 exp (a(Tj1) tn(Tj1))
tj,fict
+

s(T )

f(T )

s(Tj+1)

tj
s(Tj)

s(Tj1)

Fig. 2. JMAK model with additivity rule.

while the amount of the product phase at the end of the considered time step j, assuming the aforesaid meaning for the
subscripts i and j, writes:

Fi,j = 1 exp a Tj

( )

n Tj

tj

(11)

The above procedure is shown in Fig. 2.

It should also be considered, moreover, that the JMAK model, in the form described by Eqs. (7)(11) is valid only to analyze
the decomposition of the parent phase into a single product phase, which is not the general case of continuous cooling
relative to the conventional quenching process. Indeed, as the local temperature decreases, different phases can nucleate
and grow at specific locations of the parent phase grains, but only the new product phase formed into the untransformed
parent phase should be considered. To handle this aspect, the following procedure has been applied: for each time step j of
the microstructure model, the existing volume fraction Fi,j1 of the potential forming phase i has been divided by the sum
of the volume fraction Fp,j1 of the parent phase p and the existing volume fraction Fi,j1 is as follows:
Fi =

Fi,j1
Fi,j1 + Fp,j1

.

(12)

The volume fraction Fi has been adjourned according to Eqs. (7)(11), and finally the volume fraction of the i phase Fi,j ,
transformed at the end of time step j is provided by:
Fi,j1 = Fi Fi,j1 + Fp,j1 .

(13)

Scheils Additivity Rule has also been adopted to take into account the nucleation stage. It is assumed that, for each time
step j, the time tj , spent at temperature Tj , divided by the incubation time at that temperature, s (Tj ), is a fraction of the
total nucleation time required. As a consequence, the transformation begins when the following condition is satisfied:
n
X
tj

1,

Tj
s
j=1

(14)

n being the present time step. The above procedure has been applied to consider nucleation and growth processes of
diffusional and displacive transformations, in each node of the finite element model, between the temperature Ar3 , which
represents the starting temperature for the incubation of the austenitepearlite (subscript 3) transformation on cooling
(subscript 3), and the temperature martensite start, up to a complete conversion.
Austenite conversion into martensite, during the cooling stage, is a diffusionless transformation, characterized by a
negligible nucleation time. As a consequence, it can be assumed that the amount of the formed martensite is not time or
temperature history dependent but only temperature dependent. Several approaches have been proposed in the literature
to model the diffusionless transformation, using opportune techniques to model the irreversibility of the martensite
transformation, in particular, the empirical Koistinen-Marburger model has been widely used [58]. However, it has been
evidenced that the simple form of the Koistinen-Marburger model, is not capable of handling the complete austenitemartensite decomposition in the temperature range defined between the start and finish transformation temperatures.
In this paper the Yu model [29] has been implemented to consider diffusionless transformations: the austenitemartensite phase change is described, on cooling, by a parabolic law defined between the temperatures Ms and Mf , where
subscripts s and f refer to the transformation start and finish; as a consequence the total decomposition of austenite
into martensite can be achieved if the final temperature is less than Mf . To model the irreversibility of the martensite
transformation, it is assumed that, at the end of each microstructural time step, the total martensite fraction is equal to the
maximum volume fraction evaluated at the present and the previous time steps. In this way the reduction of the transformed

590

P. Carlone et al. / Computers and Mathematics with Applications 59 (2010) 585594

1.0
0.9

Austenite volume fraction

0.8
0.7
0.6
0.5
0.4

Reference [30]

0.3

Present Model

0.2
0.1
0.0
700

720

740

760

780

800

Temperature [C]

Fig. 3. Austenite conversion on heating (heating rate 0.5 C/s).

martensite fraction due to an eventual temperature increase, related to the phase change latent heat, has been avoided. The
implemented algorithm finally writes:

Fm,i =

"

Fa 1
Fm,i1

T Mf

2 #

Ms Mf

if Fm,i Fm,i1

Mf < T < Ms ,

(15)

if Fm,i < Fm,i1

where Fm is the martensite volume fraction formed at the temperature T , Ms and Mf are, respectively, the martensite start
and martensite finish temperatures and Fa is the residual fraction of austenite at temperature Ms .
To optimize the solution routine, reducing computational time, the thermal and phase change models have been
decoupled, and a different time step, for each model, has been adopted. To improve accuracy, a relatively finer time step has
been used for the microstructure model. Each thermal step is divided into several microstructure steps, assuming a linear
temperature variation inside the thermal step, and for each microstructure step the mean temperature has been evaluated
and used to evaluate the transformed volume fractions and phase change enthalpies, according to the above procedure. The
volume fractions and phase change enthalpies, at the end of each thermal step are obtained as the sum of the contributions
due to all microstructure steps. The cumulative heat generation rate due to the evaluated phase transitions is then used as
additional internal heat source in the successive thermal steps.
3. Results and discussion
3.1. Computational simulations: continuous heating of a 1080 eutectoid steel
The continuous heating of an AISI 1080 eutectoid steel has been simulated to validate and verify the implemented
microstructural model, relatively to the austenite formation on heating, by comparison with results provided in [30], based
on the data reported in [31]. In particular, a zero dimensional model has been used, assuming the heating rate as constant
and equal to 0.5 C/s, the initial temperature as 700 C and the final temperature as 800 C [30]. The initial microstructure
has been assumed to be fully pearlitic. The time step of the microstructural model has been assumed as 1 s. In Fig. 3 the
austenite conversion on heating provided by the performed computational simulation and reported in [30] are shown. Good
agreement between numerical results and reference data has been found, with respect to the evaluation of the austenite
incubation and growth. The end of the incubation process and the start of the austenite growth have been detected, by the
present model, after 82 s heating, corresponding to 739 C, against the 78 s heating, corresponding to 741 C reported in [30].
Perfect matching between the obtained numerical results and the reference data, relatively to the full pearliteaustenite
transformation, has been found, after 104 s heating, corresponding to 752 C. Several simulations of austenite formation
on heating have been performed to test model sensitivity and to obtain good accuracy, reducing computational time. In
particular, three different heating rates have been considered, respectively 0.5, 1, and 5 C/s; for each heating rate three
different time steps of the microstructural model have been assumed, corresponding to temperature variations, between
the beginning and the end of each step, equal to 0.5, 1, and 5 C.
As evidenced in Fig. 4, good accuracy has been obtained in simulations performed assuming the maximum temperature
variation inside each load step as 1 C, for all the considered heating rates, corresponding to the 50% reduction of the
computational time required to solve the microstructural model, with respect to simulations performed assuming 0.5 C
as maximum temperature variation inside the load step. Assuming 5 C as the maximum temperature variation between
the beginning and the end of each step, the time required for the solution of the microstructural model is strongly reduced,
however a remarkable loss in result accuracy has been found.

P. Carlone et al. / Computers and Mathematics with Applications 59 (2010) 585594

591

1.0

Austenite volume fraction

0.9 Heating rate

0.5C/sec
0.8

Heating rate
1C/sec

0.7
0.6
0.5

Heating rate
5C/sec

0.4
0.3
0.2

T=0.5C

0.1

T=1C

0.0
740

T=5C
750

760

770

780

790

800

Temperature [C]

Fig. 4. Austenite formation model: sensitivity analysis.

3.2. Computational simulations: Continuous cooling of a 1080 eutectoid steel

In this sub-section results provided by computational simulations of continuous cooling processes, performed according
to the implemented model and assuming a constant cooling rate for each simulation, have been reported to validate the
austenite decomposition model. In Fig. 5, computational results are shown and compared with the well known experimental
Continuous Cooling Transformation (CCT) diagram.
Curves A, B, C, D, E, F, G, H, I, L represent temperature histories obtained assuming cooling rates as, respectively, 150, 140,
120, 100, 60, 40, 35, 30, 20, 10 C/s; for each curve, the numerically evaluated points of austenitepearlite transformation
start and finish have been evidenced, respectively, by the symbols O and ; the experimental austenitepearlite
transformation start and finish curves are represented by Ps (dashed line) and Pf (dotted line), respectively. As evidenced,
excellent agreement with experimental data has been found for all performed simulations. Moreover, the implemented
computational model has shown a good capability to simulate the solid phase growth process, in particular relative to
cooling rates leading to decomposition of the parent phase into a single product phase, characterized by negligible growth
competition between different product phases. Good agreement with experimental data has been also found with respect
to the maximum cooing rate to complete the pearlite incubation process, which results equal to 140 C/s numerically
and experimentally, as well as to obtain the full austenitepearlite decomposition (33 C/s numerically predicted versus
35 C/s experimentally evaluated). Moreover, as confirmed by experimental data, for cooling rates between 35 and 140 C/s,
only partial austenitepearlite decomposition has been predicted, then, martensite diffusionless transformation has been
detected below Ms temperature. No austenitebainite decomposition has been detected, due to the insufficient time
available for the bainite nucleation process. For cooling rates major than 35 C/s, taking into account that for the considered
steel the Mf lies well below the room temperature, variable volume fractions of retained austenite have been found.
3.3. Computational simulations: Water quenching of a cylindrical 1080 eutectoid steel
Finally, the implemented FE model has been used to simulate the water quenching process of a 1080 eutectoid steel
cylindrical sample, as described by Woodard et al. in [5], and already used as a benchmark by Huiping et al. in [6], and by
Kang et al. in [7,8], for validation purpose and to investigate the thermal and metallographic effects of the process on the
workpiece. Specimen height and diameter have been assumed, respectively, as 76.2 and 38.1 mm. Taking into account the
cylindrical shape of the specimen, a two-dimensional axialsymmetric model has been adopted; for symmetry reasons only
one half of the workpiece has been modeled, applying opportune boundary conditions into the symmetry cross section. The
solid model has been discretized using 200 elements and 231 nodes, as shown in Fig. 6(a), including the applied boundary
conditions and loads. The physical material properties reported in [57] have been used in the performed computational
simulation. The initial temperature of the specimen has been considered as uniform and equal to 850 C; at this time a
fully austenitic structure has been assumed. A 90 s water quenching process has been simulated considering a temperature
dependent heat transfer coefficient [5,7,8], assuming the temperature of the quenching medium as 22.5 C. The time step
of the thermal model has been assumed as 0.1 s, and the time step of the microstructural model has been fixed as 0.01 s. In
Fig. 6(b) are shown the temperature profiles relative to the cooling curves of the points A, B, C, D, and E, placed, respectively
at radial distances equal to 0, 8.0793, 13.369, 16.252, and 19.05 mm from the cylinder axes. The temperature profiles relative
to the cylinder center (A) and surface (E), are compared with the results reported in [5,7,8]. Good agreement has been found
between the numerically evaluated cooling curves and the reference data. As evidenced, the negligible effect of the latent
heat due to a phase change on the cooling curves relative to the points D and E, in the proximity of the external quenched
surface, has been found, due to the minor enthalpy change related to the austenitemartensite decomposition and to the

592

P. Carlone et al. / Computers and Mathematics with Applications 59 (2010) 585594

800
700

Ps

Temperature [C]

600

Pf

500
400
300
200
100

AB C D

0
1.00E01

1.00E+00

FGH

L
1.00E+02

1.00E+01
Time [sec]

Fig. 5. Continuous cooling transformations: numerical results and experimental data comparison (Ps : pearlite transformation start, experimental; Pf :
pearlite transformation finish, experimental; O pearlite transformation start, numerical; : pearlite transformation finish, numerical).

b
900

q = hq(TT)

800
A (Present Model)
A (Woodard et al., [5])
A (Kang et al., [7])
B (Present Model)
C (Present Model)
D (Present Model)
E (Present Model)
E (Woodard et al., [5])
E (Kang et al., [7])

700
Temperature [C]

q = hq(TT)
Specimen
axes

X
A

B CDE

Symmetry
plane

600
500
400
300
200
100
0
0

10

20

30

40

50

60

70

80

90

Time [s]

Fig. 6. Water quenching of a 1080 steel cylinder: (a) FE model and mesh, including loads and boundary conditions; (b) temperature profiles on cooling
and comparison with reference data.

strong cooling facility on the cylinder surface. A more and more evident effect of the latent heat has been observed in the
cooling profiles of points C, B, and A, moving towards the specimen axes. On the specimen axes, an evident increase of
temperature has been found at the end of the incubation time of the pearlite, due to the remarkable latent heat released. A
relatively minor effect can be observed in the cooling curves relative to the points B and C, due to the faster cooling, which
implies that the incubation process is completed at a relatively lower temperature, resulting in minor enthalpy variations.
In Fig. 7 the numerically evaluated distribution of the solid phases, obtained by the partial or total austenite decomposition on cooling, are shown and compared with results provided by Woodard et al. [5] and by Kang et al. [7,8]. Very good
agreement with numerical results reported in [7] has been found; in particular relative to the penetration depth of the considered heat treatment, given by the distance, below the processing material surface, where martensite can be found; both
models have evidenced increasing martensite volume fractions for a radial distance from the specimen axes major of than
12 mm and only martensite and retained austenite volume fractions for a radial distance greater than 17.1 mm. A relatively
major penetration depth of the heat treatment has been found by Woodard et al. [5], whose results evidenced an increasing
amount of martensite for a radial distance from the specimen axes major of than 7.5 mm and only martensite and retained
austenite for radial distance greater than 15 mm. Negligible differences, relative to the distribution of phases obtained by diffusive or diffusionless decomposition of austenite, have been found between the results provided by the present model and
reference data in [7]. In particular, a relatively minor volume fraction of pearlite and bainite formation has been predicted by
the present model with respect to the Kang model, however, in [7], it is not well specified if displacive austenitebainite decomposition has been considered. Moreover, differences between predicted and reference data relatively to the martensite
volume fractions can be well explained considering that austenitemartensite decomposition is only temperature dependent, so small differences between local predicted temperatures after 90 s cooling lead to some deviations of the evaluated
volume fractions.

P. Carlone et al. / Computers and Mathematics with Applications 59 (2010) 585594

593

1
0.9
Pearlite, Woodard [5]

0.8

Pearlite, Kang [7]

Volume fraction

0.7

Pearlite, Present Model

Martensite, Woodard [5]

0.6

Martensite, Kang [7]

0.5

Martensite, Present Model

Bainite, Present Model

0.4

Retained Austenite, Woodard [5]

0.3

Retained Austenite, Kang [7]

Retained Austenite, Present Model

0.2
0.1
0
0

10

12

14

16

18

20

Distance along specimen radius [mm]

Fig. 7. Solid phases distribution along specimen radius after 90 s cooling.

80
70

Hardness HRC

60
50
40
30
Experimental (Woodard et al. [5])
20

Numerical (Woodard et al. [5])

Present Model

10
0
0

10

12

14

16

18

20

Distance along specimen radius [mm]

Fig. 8. Hardness distribution along specimen radius after 90 s cooling.

In Fig. 8 the Rockwell C hardness distribution along the specimen radius is shown and compared with experimental and
numerical results reported by Woodard et al. in [5]. The local Rockwell C hardness has been evaluated by conversion of the
hardness Vickers, obtained according to the rule of mixtures, taking into account the volume fractions of the formed phases,
the chemical composition of the processing steel and the cooling rate Vr at 700 C, as proposed in [32] by Maynier et al. The
Vickers hardness of the retained austenite has been assumed to be equal to the Vickers hardness of martensite [5].
Taking into account its capability the proposed model can be used as an effective tool for not only for process analysis, but
also for optimal planning of the real process, resulting into energy, time and cost-saving thermal cycle, into and reducing the
need for parts finishing due to the reduction of the thermal induced distortions. Future perspectives of research are mainly
focused on deformation analysis and prediction of process induced distortions, which should be carefully considered in
process planning and optimization.
References
[1] X. Ai, Y. Shu, B.Q. Li, Discontinuous finite-element phase-field modeling of polycrystalline grain growth with convection, Computers and Mathematics
with Applications 52 (2006) 721734.
[2] M.F. Zhu, T. Dai, S. Lee, C.P. Hong, Modeling of solutal dendritic growth with melt convection, Computers and Mathematics with Applications 55 (2008)
16201628.
[3] P. Carlone, G.S. Palazzo, R. Pasquino, Pultrusion manufacturing process development: Cure optimization by hybrid computational methods, Computers
and Mathematics with Applications 53 (2007) 14641471.
[4] P. Carlone, G.S. Palazzo, R. Pasquino, Inverse analysis of the laser forming process by computational modelling and methods, Computers and
Mathematics with Applications 55 (2008) 20182032.
[5] P.R. Woodard, S. Chandrasekar, H.T.Y. Yang, Analysis of Temperature and Microstructure in the Quenching of Steel Cylinders, Metallurgical and
Materials Transactions 30B (1999) 815822.
[6] L. Huiping, Z. Guoqun, N. Shanting, H. Chuanzhen, FEM simulation of quenching process and experimental verification of simulation results, Materials
Science and Engineering A 452453 (2007) 705714.
[7] S.H. Kang, Y.T. Im, Three-dimensional thermo-elasticplastic finite element modeling of quenching process of plain-carbon steel in couple with phase
transformation, International Journal of Mechanical Science 49 (2007) 423439.
[8] S.H. Kang, Y.T. Im, Finite element investigation of multi-phase transformation within carburized carbon steel, Journal of Materials Processing
Technology 183 (2007) 241248.
[9] D. Hmberg, A numerical simulation of the Jominy end-quench test, Acta Materialia 44 (1996) 43754385.

594

P. Carlone et al. / Computers and Mathematics with Applications 59 (2010) 585594

[10] L. Beirao da Veiga, C. Lovadina, An interpolation theory approach to shell eigenvalue problems, Mathematical Models and Methods in Applied Sciences
18 (2008) 20032018.
[11] P. Morin, K.G. Siebert, A. Veeser, A basic convergence result for conforming adaptive finite elements, Mathematical Models and Methods in Applied
Sciences 18 (2008) 707738.
[12] J. Xu, Y. Zhu, Uniform convergent multigrid methods for elliptic problems with strongly discontinuous coefficients, Mathematical Models and Methods
in Applied Sciences 18 (2008) 77106.
[13] C. Garca de Andrs, F.G. Caballero, C. Capdevila, H.K.D.H. Bhadeshia, Modelling of kinetics and dilatometric behavior of non-isothermal peralite-toaustenite transformation in an eutectoid steel, Scripta Materialia 39 (1998) 791796.
[14] F.G. Caballero, C. Capdevila, C. Garca de Andrs, Influence of scale parameters of pearlite on the kinetics of anisothermal pearlite-to-austenite
transformation in a eutectoid steel, Scripta Materialia 42 (2000) 11591165.
[15] S.J. Jones, H.K.D.H. Bhadeshia, Kinetics of the simultaneous decomposition of austenite into several transformation products, Acta Materialia 45 (1996)
29112920.
[16] H. Matsuda, H.K.D.H. Bhadeshia, Kinetics of the bainite transformation, Proceedings of the Royal Society 460 (2004) 17071722.
[17] M. Avrami, Kinetics of phase change I, Journal of Chemical Physics 7 (1939) 11031112.
[18] M. Avrami, Kinetics of phase change II, Journal of Chemical Physics 8 (1940) 212224.
[19] M. Avrami, Kinetics of phase change III, Journal of Chemical Physics 9 (1941) 177184.
[20] J.W. Cahn, Transformation kinetics during continuous cooling, Acta Metallurgica 4 (1956) 572575.
[21] M.B. Kuban, R. Jayaraman, E.B. Hawbolt, J.K. Brimacombe, An assessment of the additivity principle in predicting continuous-cooling austenite-topearlite transformation kinetics using isothermal transformation data, Metallurgical and Materials Transactions A 17 (1986) 14931503.
[22] M. Umemoto, K. Horiuchi, I. Tamura, Prediction of hardenability effects from isothermal transformation kinetics, Journal of Heat Treating 1 (1980)
5764.
[23] M. Umemoto, K. Horiuchi, I. Tamura, Transformation kinetics of bainite during isothermal holding and continuous cooling, Transactions of the Iron
and Steel Institute of Japan 22 (1982) 854861.
[24] M. Umemoto, K. Horiuchi, I. Tamura, Pearlite transformation during continuous cooling and its relation to isothermal transformation, Transactions of
the Iron and Steel Institute of Japan 23 (1983) 690695.
[25] H.K.D.H. Bhadeshia, Thermodynamic analysis of isothermal transformation diagrams, Metal Science 16 (1982) 159165.
[26] I. Tzitzelkov, H.P. Hougardy, A. Rose, Mathematische Beschreibung des Zeit-Temperatur-Umwandlungs-Schaubildes fr Isotermische und
Kontinuierliche Abkhlung, Arch. Eisenhttenwesen 45 (1974) 525532.
[27] M. Lusk, H.J. Jou, On the rule of additivity in phase transformation kinetics, Metallurgical and Materials Transactions A 28 (1997) 287291.
[28] Y.T. Zhua, T.C. Lowe, R.J. Asaro, Assessment of the theoretical basis of the Rule of Additivity for the nucleation incubation time during continuous
cooling, Journal of Applied Physics 82 (1997) 11291137.
[29] H.J. Yu, Berechnung von Abkhlungs-, Umwandlungs-, Schweiss-, sowie Verformungseigenspannungen mit Hilfe der Methode der finiten Elemente,
Ph.D. Thesis, Universitt Karlsruhe, Karlsruhe, 1977.
[30] T.C. Tszeng, G. Shi, A global optimization technique to identify overall transformation kinetics using dilatometry data-applications to austenitization
of steels, Materials Science and Engineering A 380 (2004) 123136.
[31] G.A. Roberts, R.F. Mehl, The mechanism and the rate of formation of austenite from ferritecementite aggregates, Transactions of the American Society
for Metals 31 (1943) 613650.
[32] P. Maynier, J. Dollet, P. Bastien, Hardenability Concepts with Applications to Steels, AIME, New York, 1978, 518544.