Proceedings of the 28th ASM Heat Treating Society Conference Copyright © 2015 ASM International®

October 20–22, 2015, Detroit, Michigan, USA All rights reserved

asminternational.org

Behavior of a Complex steel part during Hardening Heat Treatment

Mr. Vishal R Marje, Mr. Srivatsa Kulkarni, Dr. G Balachandran and Mr. V Balasubramanian
R & D, Kalyani Carpenter Special Steels Ltd., Pune, India,

Abstract increased distortion due to non-uniform cooling of various

sections during hardening [3]. There are limited studies on
Heat treatment response of a complex shaped cylindrical distortion associated with heat treatment of cylindrical shapes
component with varying sections of thickness and length has with varying cross-section components.
been studied in a 15CDV6 bainitic steel. The effect of cooling
rate at every location has been predicted using an FEM based In the present study, the heat treatment response of a concept
simulation. The distortion associated with the experimental component made of a low alloy Cr-Mo steel, with varying
component was matched with the theoretical FEM based heat section diameter and length was examined. The study brings
treatment model. The component distortion has been attributed out the effect of cooling rates at various locations of the
to the strain associated with volume changes associated with component and relates the phase constitution and associated
thermal gradient and phase transformation at various distortion. The experimental study was further modelled using
locations. The heat flux on the surface was modelled to arrive a finite element based mathematical modeling to understand
at heat transfer by inverse technique. The study validates the the distortion behavior during hardening. The heat flux during
definition of boundary conditions for such complex hardening at the experimental process condition was used to
components. generate the heat transfer boundary condition by inverse
simulation. FEM based modelling based on FORGE software
1. Introduction was used to predict the theoretical cooling rates, distortion,
phase changes and micro-structure.
Heat treatment of steel components is carried out to improve
the material properties. For a given composition, these
properties are derived from the micro structure in the steel 2. Finite Element Modeling
which is again a function of cooling rate the component
experiences at various sections. The process of heat treatment FORGE software used for simulation is based on multi-
involves phase transformation and thermal gradient associated physics phenomenon involving coupling between
volume changes, which ultimately result in distortion and transformation kinetics, mechanical behavior, and thermal
residual stress formation during the process. Cylindrical evolution occurring during quenching, as shown in Figure 1.
components such as rotors used in power plant need to be
modelled and distortion in that need to be predicted. At times
improper heat treatment can cause a crack in a value added
forged component at final stages.
This issue is more severe in case of cylindrical specimens
with varying cross sections. Such type of cylindrical
components like rotors are more prone to distortions and
residual stresses due to its complex shape. The issue of
distortion in steels during heat treatment has been tackled for
decades by various authors. Some of authors have used
mathematical models to predict the distortions with input
parameters such as steel grade, quenchants, component size,
shape, agitation etc. It is well known that distortion is
associated with volume changes associated with thermal Figure 1: Coupled model used in FORGE software
gradient and phase transformation. Arimoto et al. [1] studied
the distortion behavior on a steel shaft with a keyway during During heating, the model considers endothermic reactions
hardening process with due importance given to the effect of corresponding to phase transformation from ferrite to
size of steel shaft on heat transfer coefficient. Further Totten austenite. On the contrary, during cooling, an exothermic
[2] explained the change in residual stress distributions during reaction is initiated transforming austenite to ferrite. A phase
quenching of steel cylinders with different diameters. It has transformation also has effect on temperature distribution by
been emphasized that residual stresses can be shifted from generating latent heat during process. Further phase
transformation type to shrinking type with increasing cylinder transformation and deformation coupling were considered by
diameter of same steel shaft. There are studies that report that using phase change expansion parameter. This considers the
asymmetric design of a cylindrical component can result into distortion due volume variation during phase change.

48
Quenching can be defined as a transient heat conduction The constant n is dependent on the nucleation mode and the
problem with convective and radiation boundary conditions form of the nucleus, ߟ is Avrami's characteristic time defined
with internal heat source and sink. The governing equation for by the probability of a nucleus being activated.
this problem has been formulated by Fourier's law (Eq.1):
In the case of martensitic transformation, Avrami equation
‫ ݍ‬ൌ െ݇‫………ܶ׏‬...................................................… (1) cannot be used as it is practically independent of time and its
kinetics is not influenced by cooling rate. The amount of
where q is the vector of heat flow per unit area, k is thermal martensite formed is often calculated as a function of
conductivity and ‫ ܶ׏‬is the gradient of the temperature field temperature using Koistinen and Marburger law, Eq. (6),
inside the material.
‫ݕ‬௠௔௥௧ ൌ ‫ݕ‬௔௨௦௧ ሾͳ െ ݁‫݌ݔ‬ሺെ‫ܣ‬ெ ‫ܯۃ‬ௌ െ ܶ‫ۄ‬ሻሿ ………………… (6)
The transient heat conduction during quenching is defined by,
The phase transformation rates ‫ݕ‬௠௔௥௧ and ‫ݕ‬௔௨௦௧ respectively
߲ܶ ͳ ߲ ߲ܶ ͳ ߲ ߲ܶ ߲ ߲ܶ corresponds to martensite and the remaining austenite. ‫ܣ‬ெ is a
ߩ‫ܥ‬௣ ൌ ൬݇‫ ݎ‬൰ ൅ ଶ ൬݇‫ ݎ‬൰ ൅ ൬݇ ൰ ൅ ‫ݍ‬௩ parameter that is inherent to the material, along with ‫ܯ‬ௌ , the
߲‫ݎ߲ ݎ ݐ‬ ߲‫ݎ‬ ‫׎߲ ݎ‬ ߲‫׎‬ ߲‫ݖ߲ ݖ‬
martensitic transformation start temperature.
………………… (2)

where ݇ is the thermal conductivity, ܶ is the temperature of 3. Numerical Simulation and Experiment
quenching part, qv is the latent heat of phase transformation, Procedure
ߩ is the density of material, ‫ܥ‬௣ is the constant pressure
specific heat, t is the time,‫ݎ‬, ‫ ׎‬and ‫ ݖ‬are the cylindrical The component chosen for the study with dimension in mm is
coordinates. shown in Figure 2. It has very slender section and also
proportionally very thick section in it. During hardening, the
The boundary condition for the convective heat transfer is volume changes are expected to vary significantly that would
given in Eq.3, cause un-equal phase constitution and volume changes that
would distort the component. The phase formed is a function
డ்
ൌ ݄ሺܶሻሺܶ௔ െ ܶஶ ሻ ………........................................… (3) of cooling rate encountered at every location in the
డ௡
component. Since slender section cools faster than thicker
where ݄ሺܶሻ is the surface heat-transfer coefficient, a function section the type of phases formed and the proportion of phases
of temperature. The heat transfer between steel component and formed is likely to differ. This is expected to cause distortion
quenchant through convection can be defined using Newton’s of one section relative to the other. The objective in the
law of cooling given in Eq.4, present study was to predict by simulation the distortion
theoretically using FEM analysis and validate the same with
ܳ ൌ ݄‫ܣ‬οܶ………........................................................… (4) experimentation.

where ܳ is the heat flux, h is the heat-transfer coefficient

which combines the influence of the convection and radiation,
‫ ܣ‬is the surface area of the solid component, and οܶ is the
temperature difference between the surface of the component
and the quenching medium.

The phase transformation during quenching process can be

described by TTT diagram, which gives relationship between
the transformation start, end, and the transformed volume
fraction during the isothermal process at different
temperatures. The iso-thermal kinetics is calculated using the
Johnson-Mehl-Avrami-Kolmogorov (JMAK) formalism and
the additivity principle [3]. This model considers thermo- Figure 2: Shape and Dimensions (in mm) of component heat
physical properties, latent heat and crystal structure change treated
during phase transformation which helps to predict distortions
due to heat treatment. The Avrami equation used is given by, The theoretical study was initiated with the input of 3D model
on the component, made using SolidEdge software in the FEM
‫ ݕ‬ൌ ͳ െ ݁‫݌ݔ‬ൣെܾሺܶሻߟሺ‫ݐ‬ሻ௡ሺ்ሻ ൧……….............................… (5) based computation model in FORGE software with a
tetragonal mesh comprised of 98101 elements and 20854
Where, ‫ ݕ‬is the rate of phase transformation, ‫ ݐ‬is the nodes. The model input conditions were matched with the
isothermal time duration, ܾ is a constant dependent on the experimental data. The material data base and TTT diagram
temperature, composition of parent phase, and the grain size. for the steel grade in the FORGE rheology data base for

49
thermo-physical properties was used. The heat flux The heat treatment involved heating from room temperature to
measurement in the hardening process was used to input the 920ΣC at 100ΣC/h heating rate. The steel was held at 920oC for
heat transfer boundary condition. The model uses the inbuilt 4h. This was done to ensure that 100% austenite is formed
FORGE rheology database, where temperature dependent even in thickest section. The component was then subject to
thermo-physical properties like thermal conductivity, linear quenching with a 30s quench delay. The initial temperature of
coefficient of expansion; Young’s modulus, density etc. are quenchant water was maintained at 40ΣC. The component was
compiled. cooled in water for 30 minutes before it was removed for
measurement. The as-quenched sample was subjected to
Forge software generates TTT diagram using chemical CMM analysis again to record the changes in dimension and
composition of material and grain size using Kirkaldy method shape after hardening. Tempering was carried out by heating
[4]. But, at times the model has differences compared to the the component from room temperature to 650ΣC for 5h with a
published TTT diagram and cooling rates. To overcome this tempering time of 5 h. The as-quenched and tempered
limitation, TTT diagram from reliable published literature [5] hardness were recorded.
is input into the software and used for better accuracy. The
diagram used is shown in Figure 3. The gray color lines
represent the actual curves from literature while the green line
show bainite nose optimized with published data.

Figure 3: CCT diagram of 15CDV6 used in the present study Figure 4: CMM Measurement prior to heat treatment

In order to simulate the component heat treatment, the heat The heat treatment process parameters were input in the
transfer coefficient (HTC) was calculated under the given FORGE software based on FEM model. The model uses
cooling [hardening] condition. This is usually achieved by setting up of appropriate heat transfer coefficient that was
heat flux measurement by monitoring temperature drop during generated based on thermal flux measurement made with
hardening. Experiments were carried out by embedded embedment of thermocouples in equivalent sections followed
thermocouple to get the thermal history as a function of time, by inverse simulation using the same software. The distortion
which was used in the model inverse technique to generate the of the component after hardening was matched with the
convective heat transfer coefficient. prediction in the model.

3.1 Experimental Procedure and Simulation 4. Result and Discussion

The steel was made of 15CDV6 steel with the composition Heat treatment in general involves changes in volume
given in Table 1, measured using ARL 3460 make associated with thermal gradients and phase formation at
spectrometer. The steel component was machined from a 300 various location of a component. The volume changes
mm diameter bar made from a hot rolled bar of an ingot cast translate into plastic strain and associated stresses in the
steel. The component after machining was subjected to Co- component. Depending on the thickness of the section and the
ordinate measurement machine (CMM) to record the steel flow strength the stress can cause deformation in the
dimension before heat treatment as shown in Figure 4. sections leading to distortion or warpage. The scope of this
study is limited to assessment of distortion in hardening.
Table 1: Chemical composition of component studied
4.1Thermal History and Phase Transformations
% C Mn Cr Mo V Si S P
Wt. 0.15 0.95 1.38 0.90 0.25 0.20 0.002 0.013 Thermal history of the sample heated to a uniform temperature
of 920oC on hardening in water was analyzed. Different
sections in the component lose heat differently that results in

50
varying cooling rates. The heat transfer in the component is
influenced by the section surface area to volume ratio. Higher
surface enables the section to lose the heat more and section
with higher volume stores more quantity of heat and hence it
loses heat slowly. Hence, for given steel the cooling rate is
influenced by the surface area to volume of the section as
shown in Figure 5. It is seen that the cooling rate increases at
surface and core if the surface to area ratio increases.

(a) Temperature distribution

(b) Bainite

Figure 5: The influence of section surface area to volume

ratio on the cooling rate.

The thermal history of the sample at time intervals [5, 250,

720 & 1800 seconds] as a function of temperature distribution,
bainite phase formation, martensite phase formation, principle
stress distribution and distortion is shown in Figure 6. The
temperature distribution shows that immediately after 5 s, the
thinnest section has temperature fall to 300 oC at which critical (c) Martensite
cooling rate for martensite formation is exceeded as per the
Figure 2 and Fig.6(c) where surface martensite is evident. The
principle stress build up is seen on the surface layers in thick
section. This stress build up has started distortion as in Fig.6
(e).
At a time of 250 s hardening, the slender sections B and D has
reached quenchant temperature with full martensite
transformation. The section A attains about 300oC and there is
a bainite formation at core in Fig.6 (b). The top disk section A
has started showing warpage. The warpage is associated with
unequal heat transfer along section A top surface and bottom
surface which shows compressive stress [6]. The temperature (d) Principle stress distribution
in thickest section still retains high temperature of about
400oC. The cooling rate decrease is retarded due to section
thickness. This section E and also G & H show bainite
formation due to slow cooling rate. The principle stress shows
compressive stress at the surface and tensile stress at the core.
This results in a distortion that distorts the section to decrease
the diameter in section E. The end section A and G both show
warpage from the direction perpendicular to the axis along the
free zone not joined to the adjacent section.
(e) Distortion
At a time instant of 720 s, the thickest section E show a
maximum temperature of 400oC still prevailing. Bainite and Figure 6: Variation of (a)Temperature distribution, (b)Bainite
martensite fractions are seen in sections A, C and F. The formation, (c) Martensite formation, (d)principle stress and
sections B and D show complete martensite. The principle (e) distortion at various instant of quenching time i) 5sec ii)
stresses predominate in section C due to the volume changes 250sec iii) 720 sec iv) 1800sec

51
associated with the transformation. The phase transformation stage of transformation, when the core gets transformed from
in section A and C causes the slender section to accommodate austenite to low temperature phases, the volume change that
stresses developed. This results in the slender section to distort take place against a fully transformed colder outer layers leads
from its axis towards bending in one side. At a time instant to high tensile stress. This means the core is ready to rupture.
of 1800 s the entire component has attained the quenchant In big components such as thermal power plant rotors this sort
temperature. The final microstructure consists of predominant of large section may lead to rupture.
martensite in section B and D. Mixed bainite & martensite in

dĞŵƉĞƌĂƚƵƌĞ;ΣͿ͕
section A, C D and E and also core section of B. The principle
stress predominates in core of sections C and E. The stresses ϵϬϬ

ĂŝŶŝƚĞWŚĂƐĞsŽůƵŵĞ
Ϭ͘ϲ
in section in D relative to section A act on the slender section ϳϬϬ
B in between which distorts by bending. The distortion is

^ƚƌĞƐƐ;DWĂͿ

&ƌĂĐƚŝŽŶ
ϱϬϬ
maximized at section A and B as in Fig.6 (e). Ϭ͘ϰ
ϯϬϬ
4.2 Residual Stress and Phase Transformation ϭϬϬ Ϭ͘Ϯ
ͲϭϬϬ
The volume changes associated within a given phase
associated with thermal gradient and the volume changes ͲϯϬϬ Ϭ
associated with phase changes differ in proportion at different Ϭ ϯϬϬ ϲϬϬ ϵϬϬ ϭϮϬϬ ϭϱϬϬ ϭϴϬϬ
section thicknesses generates plastic strain. When the strain is dŝŵĞ;ƐĞĐͿ
restrained by the section thickness there is development of
residual stresses in the component. In the case of cylindrical (a) Section A
components with varying cross section, heat extraction is
mainly in the radial direction for sections B, C, D, E and F. In
dĞŵƉĞƌĂƚƵƌĞ;ΣͿ͕
ϭ͘Ϯ
the sections A and D there is faster heat extraction from the ϵϬϬ

DĂƌƚĞŶƐŝƚĞWŚĂƐĞ
sŽůƵŵĞ&ƌĂĐƚŝŽŶ
flat side surfaces compared to the diameter causing phase ϳϬϬ
ϭ
^ƚƌĞƐƐ;DWĂͿ
changes from end surfaces. This distorts the end disks. Ϭ͘ϴ
ϱϬϬ
The component has volume changes differing at different ϯϬϬ Ϭ͘ϲ
locations due to different proportions of the phases formed and ϭϬϬ Ϭ͘ϰ
thermal gradient. In the present study, the steel 15CDV6 forms ͲϭϬϬ Ϭ͘Ϯ
bainite over a wide range of cooling rates especially in thicker
section. Martensite formation is limited to very thin section. If ͲϯϬϬ Ϭ
the component were an alloy steel, the component has a Ϭ ϯϬϬ ϲϬϬ ϵϬϬ ϭϮϬϬ ϭϱϬϬ ϭϴϬϬ
potential to crack and hence the bainitic steel is a good choice dŝŵĞ;ƐĞĐͿ
to study distortion.
(b) Section B
The variations of temperature, phase fraction, and residual
stress generated in section A, B and E in the surface and core
dĞŵƉĞƌĂƚƵƌĞ;ΣͿ͕

ϭϮϬϬ
of the component are shown in Fig 7(a) to (c) respectively. In

ĂŝŶŝƚĞWŚĂƐĞsŽůƵŵĞ
the section A in Fig.7 (a), slower cooling rate leads to bainite ϵϬϬ
Ϭ͘ϲϬ
formation. As phase transformation proceeds, tensile stresses
^ƚƌĞƐƐ;DWĂͿ

are developed at surface and core, which causes warpage. In Ϭ͘ϰϱ

&ƌĂĐƚŝŽŶ
ϲϬϬ
the case of section B as in Fig.7 (b) the tensile stresses are
ϯϬϬ Ϭ͘ϯϬ
very high especially in the case of core, compared to the
surface. The net effect of the stress variation acting on the Ϭ͘ϭϱ
Ϭ
section B, results in bending.
ͲϯϬϬ Ϭ͘ϬϬ
In the case of thickest part, section E, the surface stresses vary Ϭ ϯϬϬ ϲϬϬ ϵϬϬ ϭϮϬϬ ϭϱϬϬ ϭϴϬϬ
from a value of 200 MPa till a time duration of 520s, where
dŝŵĞ;ƐĞĐͿ
after it reverses to a compressive stress regime [Fig.7(c)]. The
core stress is initially compressive till 250s then between 250
to 430 s it is under tensile stress regime followed by a (c) Section E
compressive stress regime between 430 and 810s. Beyond
810s, the core is showing a very high tensile stress close to
1200 MPa. The above sequence of stress generation takes
place due to low temperature phase transformation proceeding
from cooler surface which causes surface shrinkage against a Figure 7: Variation of temperature and principle stresses with
high temperature core with low flow strength. The stress time (a) Section A (ø180*15) (b) Section B (ø10*84)
reversals are due to an inter-play of thermal stress and the (c) Section E (ø100*114)
stresses associated with phase transformation. During the final

52
4.2 Distortion Analysis The distortion behavior of the component can be better
understood from Fig.9. Initially, the thermal gradient and
Distortion as observed in the experimentation when compared phase transformation between surface and core is high, which
with the simulation show a close match as shown in Fig.8. The induces thermal stresses in the thinner sections section A,
distortion trends in simulation appear similar to that in which causes it to bend inwards from one end, while the other
experiment. The distortion that occurred relieves the stress in surface end bends outward. This distortion behaviour after
the steel. The CMM measurement, before and after heat 150sec of quenching is shown in Figure 9(a). Due to this
treatment is compiled and given in Table 2. volume change associated with phase transformation after 250
s (Fig.7), distortion changes its direction as in Fig. 9(b).
Table 2: Dimensional analysis – experiment and simulation Bending of the slender section B, is captured by the arrows
comparison moving outwards. As the quench duration progresses, phase
transformation is completed and temperature becomes uniform
throughout the section, there is permanent distortion in the
Dimensions material due to the plastic strain induced, that exceed flow
Section Dimensions After Heat
Before Heat stress to cause mechanical yielding during quenching process.
Diameter Treatment
Treatment This permanent distortion is shown in Fig. 9(c). It can be
observed that the larger section E of 100 mm diameter shrinks
Diameter Experiment Simulation at mid portion while ends of the cylinder are expanded. This is
A 180.013 177.77 172.78 due to the higher volume fraction of bainite at mid portion
surface of section E. Due to positive volume change
B 10.082 10.003 9.45
associated with bainite formation, the middle portion tries to
C 49.985 49.81 48.14 expand [7], but due to constraints, it will push theend material
D 9.833 9.75 9.2 which results into expansion of ends.

E 99.961 100.08 97.24

F 59.966 59.908 58.93
G 179.627 179.372 174.85

It is observed that the thinner section B is more distorted than

thicker section as shown in Figure 8. In thin section, the
martensite formation is seen due to high cooling rates. In
addition, the thinner sections have lesser resistance to plastic
deformation compared to thicker sections. It can be observed
that the section A displaced by 17.5mm at the end of
quenching, while the section C bends by 8mm in the x-axis
direction.

(a) (b) (c)

Figure 9: Distortion behavior during quenching of rotor after

(a) 150 sec (b) 250sec (c) 720sec

4.3 Microstructural and Hardness Validation

The simulation results in terms of microstructure evaluation

were made at select locations. There is good agreement
between the model prediction and microstructure evaluated at
various locations as in Figure 10.

(a) (b) Hardness in tempered condition also shows good correlation

Figure 8: Distortion (mm) at the end of quenching (a) with simulation results as shown in Fig.10. Microstructure
Simulation (b) Experiment obtained in thinnest section B shows tempered martensite

53
whereas microstructure in thickest section shows tempered core sections show very high tensile stresses. The hardness
bainitic structure. profile showed a good match.

Acknowledgments

The authors wish to thank the constant encouragement and

support provided by Dr Baba Kalyani, Chairman & Managing
Director and Mr R K Goyal, Managing Director, Kalyani
Carpenter Special Steels Pvt. Ltd., Pune.

References

[1] Arimoto, K. et al., “Mechanism of Quench Distortion on

Steel Shaft with Keyway”, 19th Heat Treating
Conference, ASM International, 2001, p 144-151
[2] Totten, G. et al., Steel Heat Treatment: Metallurgy and
Technologies, pp. 324.
[3] Aliaga C., et al., “3D Finite element simulation of
residual stresses and distortions of cooling steel
workpieces”, Proc. of the 3rd International Conference
on Quenching and Control of Distortion, March 1999,
page 288.
[4] Micheli P., et al., “Towards the simulation of the whole
manufacturing chain processes with FORGE”,
International Conference on New Developments in
Forging Technology, NEMU 2015, Stuttgart, Germany,
May 2015.
[5] Metal Ravne Steel Selector v.4.0, Transformation
Figure 10: Simulated and experimental hardness (BHN) and Diagrams (CCT & TTT),
Microstructure at thinnest (section B) and thickest (section E) http://www.metalravne.com/steelselector/steels/PO795.ht
locations in tempered condition ml
[6] Shipley R. Analysis of Distortion and Deformation, ASM
Handbook Volume 11, Failure Analysis and Prevention
In summary, this study has enabled understanding the phase (ASM International) (2002),
transformations that differ as a function of varying section [7] Tszeng, T. et al., Proceedings of 16th Heat Treating
dimension and the cumulative effects that result into distortion Society Conference & Exposition, ASM International,
of the component. FEM modelling predicts the heat treatment 1996, pp 9-15.
responses correctly. However, efforts are needed to fine tune 
the heat flux inputs and definition of boundary conditions. The
present study throws open the capability to model heat
treatment of much larger components such as power plant
rotors. The prediction of distortion enables to initiate measures
to contain the distortion.

Conclusions

Heat treatment simulation of a cylindrical component of

varying section size was studied to understand distortion
behaviour during quenching. Simulated distortion behaviour
of the component showed excellent correlation with
experimental data. The stress development due to thermal
gradient between surface and the core of the various sections,
along with the stress development due to the extent of phase
transformation that takes place decides the overall distortion
behaviour. Slender section show higher distortion while thick

54