Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: Simulation is a very helpful and valuable work tool in the field of heat treatment of steels. It allows
Received 15 June 2012 behavior laws and algorithms to be learned and tested. Simulation provides a low cost, secure and fast
Received in revised form 17 September 2012 analysis tool. For a fine assimilation of the microstructure and distortion evolution concepts which char-
Accepted 1 October 2012
acterize the surface hardening process, a simulation of the thermo-metallurgical and mechanical cou-
Available online 7 December 2012
pling is elaborated by using ABAQUS software linked to phase transformation and elastic–plastic
modules which we developed taking into account of interactions between these fields. The application
Keywords:
of this method makes it possible to highlight the metallurgical and mechanical behavior laws and proce-
Surface quenching
Phase transformation
dures used to calculate phase fractions, austenite grain size, hardness, microstructure effect on physical
Hardness properties, transformation latent heat, deformation progress, and stress genesis during a surface harden-
Residual stress ing; which cannot be feasible by using standard software based simulation. Predicted thermo-metallur-
ABAQUS gical results and residual stress have been compared with measured ones; the coincidence between
simulated and experimental values confirms the validity of the chosen simulation tool.
Ó 2012 Elsevier B.V. All rights reserved.
0927-0256/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.commatsci.2012.10.001
298 M. Yaakoubi et al. / Computational Materials Science 68 (2013) 297–306
gradients induce thermal dilatations, stresses and phase transfor- length and time scales. Modeling of these processes necessitates
mations into the workpiece. Phase transformations affect the tem- dealing with inherent complexities such as large material proper-
perature field through latent heat of transformation and ties variations, complex couplings and complex boundary condi-
microstructure dependent thermo-physical properties [12,14]. tions. In this study, a finite element method based mathematical
They are also at the origin of additional deformations: volume framework capable of predicting temperature evolution, solid
changes (spherical deformations) due to the difference in compact- phase histories and internal stress genesis during the whole heat
ness between microstructures a and c and transformation induced treatment process of low alloy carbon steel is developed. In the fol-
plasticity (TRIP) which is considered proportional to both produced lowing sections, mechanical modeling is especially treated explic-
phase fractions during cooling and deviatoric stress tensor what- itly. The overall modeling is integrated into the commercial FE
ever the type of transformation is [15]. Furthermore, stresses and software ABAQUS via user subroutines.
strains affect phase transformations through modifying transfor-
mation temperatures. For example, temperature Ms decreases with 2.1. Thermal modeling
applying hydrostatic pressure on the workpiece during cooling but
it increases with applying uniaxial stress. Applied stresses lead also The distribution of the temperature into the workpiece under-
to time temperature transformation curves (TTT) moving to small going heat treatment process can be calculated by the resolution
times and an increasing of transformation rate [15,16]. of the heat transfer equation with Dirichlet boundary condition
Thermomechanical behavior of a material undergoing phase by using ABAQUS software:
transformation includes thermo-elastic and plastic behavior of
@T
multiphase material and the effects of phase transformations qC p ¼ divðk gradðTÞÞ þ q_ ð1Þ
@t
[15]. Material properties (Young modulus, Poisson ratio, yield
stress, plastic modulus, thermal conductivity, density, etc.) are to
T s ¼ f ðtÞ ð2Þ
be considered as temperature and microstructure dependent; mix-
ture rule is commonly used. with T is the temperature, t the time, k the thermal conductivity, q
There are many codes which can be used to simulate concepts the density, Cp the specific heat, and q_ is the thermal power released
characterizing surface hardening process. However, when utilizing or absorbed during phase transformation
standard software such as SYSWELD, HEARTS, and DEFORM on a The initial condition is the following one:
one hand, user cannot assimilate precisely the implemented mod- For t = 0 and for each material point:
els and does not know exactly how is carrying out the processing.
T ¼ T0 ð3Þ
On the other hand, user cannot change modeling or formulation
because he has not access to the code source. This article deals where T0 is the initial temperature of the workpiece. The boundary
with the extension of the ABAQUS software (which does not have condition is given by the Eq. (2) where Ts is the surface temperature
of standard phase transformation module) to simulate concepts which evolves through time. Temperature Ts describing the thermal
characterizing the quenching process of steels. Indeed, Thermo- cycle applied on the surface; is obtained by experimental measure.
metallo-mechanical modeling and interactions between physical Thermal properties of multiphase material depending on tempera-
phenomena is linked to ABAQUS standard through external rou- ture and microstructure are evaluated by linear mixture law. They
tines so the processing covers the whole numerical problem. are introduced into the ABAQUS Standard via user subroutine
Therefore post-processing enables the simulation of thermal, met- USDFLD.
allurgical and mechanical variables. The proposed general proce-
dure of simulation is applied to the steel sample. Numerical 2.2. ABAQUS software extension for phase transformation modeling
results show that the proposed model and the suggested simula- and its effects on mechanical behavior
tion tool are able to capture the general behavior depicting the
quenching process of steels. Besides, numerical results present a In order to increase the functionality of several ABAQUS capa-
good agreement with those of experimental data. bilities, many user subroutines which they furnish an extremely
powerful and flexible tool for analysis are provided. Numerical
modeling of heat treatment process must include laws and algo-
2. Modeling rithms describing thermal evolution, phase transformation, genesis
of internal stress and interactions between them. Indeed, user sub-
During thermal processing, parts are usually subjected to con- routines allow developing special constitutive behavior of a mate-
tinuous heating and cooling cycles during which microstructural rial which is not predefined in ABAQUS Standard to be available.
and mechanical evolutions occur simultaneously at different Moreover, user subroutines permit especially to define solution-
dependent state variables (SDVs), user defined fields (FIELD(n); n
is the required number of defined fields) and heat flux (FLUX)
due to internal heat generation in a material; for example as might
be associated with phase changes occurring during the solution
[17]. User subroutines are typically written as FORTRAN code and
must be included in a model when executing the analysis. They
should be written with great care and must include the statement
‘‘include ‘aba_param.inc’’’ as the first statement after the argument
list in order to ensure the linkage with ABAQUS Standard analysis
case. The file ‘‘aba_param.inc’’ is installed on the system by the
ABAQUS installation procedure and contains important installation
parameters. This statement tells the ABAQUS execution procedure;
which compiles and links the user subroutine with the rest of ABA-
QUS, to include the aba_param.inc file automatically. It is not nec-
essary to find the file and copy it to any particular directory;
Fig. 1. Physical fields and interactions during quenching of steels. ABAQUS will know where to find it.
M. Yaakoubi et al. / Computational Materials Science 68 (2013) 297–306 299
2.2.1. User subroutine ‘‘UMAT’’ equation of compressibility, the mechanical loading imposed on
The modeling of the heat treatment residual stresses, distor- surface (boundary conditions) and the constitutive equations
tions and phase transformation during the quenching process is which provide relationship between stresses and strains. Some
governed principally by time and temperature evolution. So, user assumptions are made in this study such that small strain incre-
subroutine ‘‘UMAT’’ provides several variables ‘‘passed in for infor- ment theory is applicable, viscous effect during high temperature
mation’’ such as time, time increment, temperature at the start of is neglected, the yielding of the material is governed by Von Mises
increment, increment of temperature, step number, and increment yield criterion with the associated hardening rules, the material is
number which characterize a transient analysis. Therefore, hardened isotropically and kinematically (Chaboche hypothesis),
mechanical and metallurgical variables such as thermal strain, all material parameters (yield stress, Young modulus, hardening
elastic strain, plastic strain, phase fractions, transformation tem- parameters, etc.) are to be considered as temperature and micro-
peratures, carbon proportion, and austenite grain size are defined structure dependent (mixture rule is generally assumed). The
as solution dependent state variables ‘SDVs’. At each time incre- integration of the behavior law on each time increment during
ment, user subroutine UMAT is called at all material integration mechanical computing will be done at unvarying temperature;
points of elements then solution dependent state variables ‘SDVs’ the temperature being able to evolve in each point from a time
will be updated by following the algorithm imposed by the user. increment to another. Strain increment composition is given by
Therefore, this user subroutine is used to model most physical phe- Prandtl–Reuss rule. During the heat treatment process, phase
nomena which characterize quenching process and interactions transformation is a source of deformation so the phase transforma-
between them: tion–stress interaction leads to an additional deformation which it
can be decomposed in two parts: spherical part and deviatoric part.
The effect of field temperature on phase transformation (metal- The first is called phase transformation strain; it is one of main
lurgical modeling). factors causing local deformation of the steel part while the
The effect of field temperature on the carbon distribution (car- microstructural plastic strain (second part) is called transformation
bon diffusion). induced plasticity (TRIP). TRIP is a plastic deformation which
The evaluation of the new hardness profile. appears during quenching under external applied stresses even
The evolution of thermal strains and stresses (mechanical they are lower than the yield stress of the weak phase. It has been
modeling). well investigated experimentally and theoretically so it has
The evaluation of deformations due to phase transformation been observed that this transformation induced plasticity incre-
then introducing them to the mechanical modeling. ment is proportional to the generation rate of each obtained phase
The effect of carbon content and applied stresses on beginning during cooling and deviatoric stress tensor [15,28,29]. The total
and ending of transformation temperatures and on kinetics of strain rate increment generated during thermal processing can
phase transformation during quenching. be decomposed into various individual strain increments as
follows:
The summarization of metallurgical modeling is given briefly as
e_ ij ¼ e_ th _ e _ p _ tr _ pt
ij þ eij þ eij þ eij þ eij ð4Þ
follow: austenitization and diffusive kinetics are determined
respectively by using the algorithm of Farias et al. [8] basing on Eq. (4) defines the increment of total strain (left hand side). In the
the extension of (JMAK) isotherm kinetic to anisotherm transfor- right hand side, the first term is the thermal expansion rate:
mation law by applying the additive concept which consists in
e_ th _
making discretization of the continuous temperature time curve ij ¼ aðTÞTdij ð5Þ
to a succession of isothermal elementary transformations then cal-
The parameter a(T) is the linear thermal expansion coefficient
culating new phase fraction during each one of them by introduc-
related to the multiphase material, dij is the Kronecker delta. The
ing the fictitious time notion [18–20]. Except that for the diffusive
second term is the elastic strain rate which is related to the stress
transformations, the growth of the phases is preceded by an incu-
rate by Hooke’s law:
bation stage which its duration is evaluated by the Scheil sum
[21,22]. The formulation usually used to model martensitic trans- 1þm m
e_ eij ¼ r_ Trðr_ Þdij ð6Þ
formation is the one of Koistinen and Marburguer [21–23]. Carbon E E
content in the austenite is modeled according to the variation in The third term is related to the plastic strain rate. It is calculated
temperature DT = Tmax Ac1 where Tmax is the austenitization by using the classical theory of plasticity with the associated hard-
temperature and Ac1 is heating rate dependent [8]. Carbon heter- ening rules. A superposition of isotropic and kinematic hardening
ogeneity and applied stresses effects on transformation tempera- is supposed in this modeling since non monotonous loads are im-
tures and on transformation kinetics are modeled by the formula posed on the workpiece during the heat treatment because of the
given by Aliaga and Denis [12,15]. Austenite grain size (AGS) af- transition traction–compression. Furthermore, Chaboche proved
fects considerably martensitic start temperature (Ms) and mar- that in the field of small deformations, hardening is rather of kine-
tensite hardness; then the relation of Leblond and Devaux [24] is matic type with a weak isotropic expansion [30]. According to that
chosen to evaluate the growth of AGS during heating. Temperature preceded, the Von Mises yield function and the plastic strain rate
(Ms) is estimated by a developed empirical formula which intro- tensor are defined respectively as follows:
duces carbon proportion, austenite grain size and applied stress 1=2
as variables basing on approaches given by Denis et al. [15,25– 3
f ¼ req rp ¼ n:n ry ðTÞ Rðp; TÞ ð7Þ
27]. We suppose that martensite hardness depends on the carbon 2
proportion when prior austenite is heterogeneous and on (AGS)
when prior austenite is homogeneous. The dependence of martens- @f
ite hardness on carbon content in the parent phase is estimated by
e_ p ¼ k_ ð8Þ
@r
using the equation given by Aliaga [12]. While the reduction of
where n = s x, T is the temperature, req is the equivalent stress, s
homogeneous martensite hardness regardless the growth of AGS
is the deviatoric stress tensor, and x is the linear kinematic harden-
is modeled by using Hall–Petch relationship [13].
ing tensor (back-stress); it is defined versus plastic strain tensor:
The mathematical model of the mechanical problem is based on
the dynamic equilibrium equation in the continuous mediums, the
300 M. Yaakoubi et al. / Computational Materials Science 68 (2013) 297–306
Fig. 2. Overall procedure for obtaining the solution for mechanical problem.
conditions, boundary conditions, etc. to ABAQUS by using an input modulus, hardening modulus, and yield stress of the material were
file. Then, the analysis results such as temperature field, micro- dependent on the temperature and microstructure. In this analysis,
structure repartition, carbon proportion, hardness profile, stresses, material parameters given by Bailey and de Oleivera [2,19] for the
and distortion fields are stored in the output database (o.d.b.) file. AISI 4140 steel were used. Tables 1–8 summarize these parameters
which were interpolated from experimental data and described by
polynomial expressions. Simulations consider initial conditions of
3. Simulations and discussions volume fraction distributions as 45% of pearlite and 55% of ferrite
and initial bulk material hardness as 240 Vickers.
In order to validate the numerical modeling procedure pro- The workpiece which is a rectangular block measuring
posed in this study to simulate the heat treatment phase transfor- 50 50 18 mm using a Nuvonyx 4 kW high power direct diode
mations and internal stresses during the quenching process, the laser. The laser has a rectangular beam profile with a size of
predicted temperature history, the solid phase history and stress 12 8 mm. The laser track coincides with the center of the work-
distribution were compared to the experimental data available in piece as shown in Fig. 4. So maximum temperature reached in the
the literature for a parallelepiped massive solid of AISI 4140 steel center of the top surface (point a) is about 1215 °C. Thermal cycles
which was laser surface hardened [2]. Thermal and mechanical recorded on points ‘a’, ‘b’, ‘c’, ‘d’, ‘e’, ‘f’, and ‘g’ which were given by
properties such as thermal conductivity, density, specific heat, la- Bailey [2] and shown in Fig. 5, were used as boundary conditions to
tent heat of transformation, thermal expansion coefficient, Young’s carry out the simulation. Indeed, thermal evolutions in these
302 M. Yaakoubi et al. / Computational Materials Science 68 (2013) 297–306
Table 1
Specific heat.
Table 2
Thermal conductivity.
Table 5
Yield stress.
Table 3
Latent heat. ry (Pa) Temperature range
Table 4
Young modulus.
Table 6
Hardening modulus.
Table 7 registered in the point ‘a’ then it decreases progressively with mov-
Coefficient of thermal expansion. ing away from the center of the top surface. Simulation shows also
Phase a (1/°C) that the temperatures of beginning and ending of austenitization
Ferrite 1.61 10–5
increase with increasing heating rate; a fact which characterize
Pearlite 1.53 10–5 short time austenitization [8]. Central zone limited by the ellipse
Austenite 2.20 10–5 that going by point ‘f’ was totally austenitized as shown in Fig. 8.
Martensite 1.15 10–5 Point ‘g’ with distance of 4 mm from the center underwent partial
austenitization since maximum temperature here (765 °C) did
not reach Ac3. The quenching was applied to the specimen using
Table 8 fast cooling rate. Consequently, austenite in heat affected zone
Volume change. (HAZ) was transformed mainly to martensite which makes this
Phase transformation f = DV/3V (%) zone harder than unaffected base material. Elliptical shape of iso-
thermal contours on top surface of the workpiece were inherited
Pearlite to austenite 0.110
Ferrite to austenite 0.126 for the distribution of both austenite and martensite during heat-
Austenite to martensite 0.342 ing and cooling respectively what led subsequently to obtain an
elliptical hardened region as shown in Fig. 9. Maximum hardness
value (720 Hv) was obtained in zone characterized by austenitiza-
tion temperature slightly greater than temperature Ac3; so the
kind material points located on the same elliptical line would have homogeneous austenite grain was still small (about 6 lm). The
the same temperature. In order to satisfy to the condition that the increasing of the AGS during heating and the subsequent decreas-
size of the rectangular beam profile was of 12 8 mm, ellipses ing of martensite hardness obtained after quenching was illus-
with large-axis equal to three half of small-axis were chosen to trated in central region corresponding to austenitization
be coaxial with the laser beam profile as shown in Fig. 6. This algo- temperature between 950 and 1215 °C; region where AGS grows
rithm is accessed into ABAQUS code through user subroutine DISP. from 8 to 22 lm while hardness declines from 700 to 650 Hv.
Thermal cycles recorded on deeper points ‘h’ and ‘j’ were utilized to Vickers microhardness tester is used to measure the microhard-
validate the numerical curves. So, Fig. 7 gives the temperature dis- ness in the HAZ. Fig. 10 illustrates the comparison of predicted
tribution in two different positions beneath the top surface: point hardness with measured one through the depth at the center of
‘h’ 0.5 mm and point ‘j’ 1.25 mm under higher surface of the block the specimen; the hardened depth is about 1.2 mm. The predicted
solid. The close conformity between the numerical results and results are in excellent agreement with the microhardness mea-
experimental measurements should be noted. Table 9 shows pre- surements. Great laser thermal power applied on top surface of
dicted transformation temperatures Ac1 and Ac3 and heating rates specimen during heating makes temperature in the middle of the
for points a, b, c, d, e, f, and g. The highest heating rate (405 °C/s) is top surface largely overshooting the temperature Ac3 that leads
Fig. 4. Schematic of laser hardening process showing the location of points for temperature survey.
304 M. Yaakoubi et al. / Computational Materials Science 68 (2013) 297–306
1200 Table 9
point a Predicted heating rate and austenitic transformation temperatures for points a, b, c, d,
point b e, f, and g.
point c
1000 Points Distance from the center Heating rate (°C/ Ac1 Ac3
Temperature (°C)
point d
(mm) s) (°C) (°C)
point e
point f a 0 405 758.5 860.5
point g b 1 400 758 860
800
c 2 390 757.5 859
d 2.5 367 756.5 855
e 3 330 755 850
600 f 3.5 235 753 846
g 4 170 751 –
400
10 12 14 16
Time (s)
100
Fig. 5. Recorded temperature history of superficial points shown in Fig. 4 given by
Bailey that used as boundary conditions for this study.
60
Top surface of
the block solid
40
20
0
10.5 11 11.5 12 12.5
Time (s)
Fig. 9. Distribution of the hardness on the top surface of the workpiece after surface laser quenching: (a) full model and (b) zoom.
800
measure
700 Simulation
600
Hardness (Hv)
500
400
300
200
100
0
0 0.25 0.5 0.75 1 1.25 1.5
Depth from surface (mm)
Fig. 10. Comparison between predicted and measured hardness into the depth
beneath the center of the top surface of the workpiece. Fig. 11. Residual stress value reporting scheme in the finite element model.
306 M. Yaakoubi et al. / Computational Materials Science 68 (2013) 297–306
Simulation
100
References