You are on page 1of 9

V. C.

Prantil

M. L. Callabresi2
J. F. Lathrop2
Sandia National Laboratories,
Livermore, CA 94551-0969

G. S. Ramaswamy
The General Electric Company,
Niskayuna, NY 12309

M. T. Lusk
Colorado School of Mines,
Golden, CO 80401

Simulating Distortion and


Residual Stresses in Carburized
Thin Strips
This paper illustrates the application of a new multiphase material model for simulating
distortion and residual stresses in carburized and quenched gear steels. Simulation is
focused on thin, metallic strips that are heat treated to introduce a through-thickness
carbon gradient. Because the material properties are strongly dependent on the carbon
content, quenching causes significant transverse out-of-plane distortion. The material
model accounts for a multiphase alloy structure where inelasticity in the individual phases
is temperature and rate dependent. The model is fit to an extensive matrix of experimental
data for low carbon steels (0.20.8 percent) whose transformation kinetics and mechanical response are similar to 4023 and 4620 alloys used in experiments. While residual
stress data are limited, reasonable agreement with X-ray diffraction measurements was
obtained. Comparisons of transverse deflections predicted numerically showed excellent
agreement with those measured experimentally for all five thicknesses reported. Accurate
transformation and lattice carburization strains are critical to correctly predict the sense
and magnitude of these transverse distortions and in-plane residual stresses.
DOI: 10.1115/1.1543973

Introduction
During heat treatment, furnace and quench schedules produce
complex thermal histories in which parts simultaneously experience significant thermal as well as phase transformation-induced
volume changes. The complex interaction and relative timing of
these competing factors determine the residual stresses and displacements that result. Often thermal processing results in detrimental shape changes. Examples include helical unwinding and
potato chipping in quenched gears and plate warping in multipass
welding. While beneficial residual stresses can be tailored, conditions do arise in which internal stresses are sufficient to cause
fatigue and quench cracks. For parts with complex shapes, undesirable residual stresses can also result from locally constrained
deformation.
In the austenitic, low carbon steels of interest in this work,
successful analysis of these processing steps requires an accurate
material characterization and constitutive description of the material. When quenching low alloy steels, the thermal and mechanical
responses are inherently coupled with evolution of the underlying
material microstructure. In particular, phase transformations drive
microscopic plasticity that, in part, determines the residual
stresses and distortion that result. In order to develop an accurate
numerical simulation capability for carburization and quenching,
one must take care to properly account for the coupling between
the thermal, mechanical, and microstructural physical mechanisms
at work.
The modeling strategy adopted here couples differential equations for phase evolution with a multiphase internal state variable
material model. The kinetic rate equations for each product phase
are derived using a thermodynamic formulation and fit to experimental data. The kinetics model is validated using time temperature transformation TTT data, continuous cooling transformation
CCT data, and by studying the influence of stress on the kinetics
through compression and tension experiments. The elastic-plastic

constitutive behavior is highly nonlinear. In particular, at high


temperatures, the material has very low yield strength. During
transformation, product phases such as martensite are substantially harder than the parent austenite. The volume expansion accompanying phase transformation on cooling drives further plastic
straining in the austenite when under stress, a phenomenon referred to as transformation induced plasticity TRIP. The yield
and hardening behavior depend on the carbon content, temperature and strain rate. The transformation strains also depend upon
carbon content and temperature. Finally, the austenite lattice volume expansion upon carburization is of the same order as the
transformation strains and contributes significantly to the plate
distortion.
This work addresses application of the heat treatment simulation software package, DANTE, developed for the Heat Treatment
Distortion Project at the National Center for Manufacturing Sciences NCMS 1. The mechanical model is based on a mixture
theory wherein we track the behavior of individual phases using
an internal state variable formulation for elasto-plasticity. This
model is joined with a new approach to simulating phase transformation kinetics recently proposed by Lusk et al. 2. The macroscopic manifestation of the interaction between phases is given
by their relative strain accommodation during transformation.
First, the volume difference associated with the phase change imparts a purely dilatational deformation. In addition, the deviatoric
transformation plastic straining will result in permanent shape
change when phase change occurs in the presence of a deviatoric
macroscopic stress field. The TRIP strains are parameterized by
stressed dilatometry tests.
An extensive series of tests were performed to fit the single
phase plasticity and transformation kinetics models. These models
have been implemented as a User Material UMAT Interface to
the ABAQUS finite element code 3. Here, we present a loosely
coupled simulation to predict the residual stresses and distortions
that occur during carburization and quenching of a thin, carburized strip 4,5.

Now at Milwaukee School of Engineering, Milwaukee, WI 53202.


Now retired.
Contributed by the Materials Division for publication in the JOURNAL OF ENGINEERING MATERIALS AND TECHNOLOGY. Manuscript received by the Materials
Division September 28, 1999; revision received September 18, 2000. Associate Editor: G. Johnson.
2

116 Vol. 125, APRIL 2003

Background and Prior Work


In the past, attempts to model the effects of phase transformations have met with mixed success. Previous simulations of the
modified Almen strip by Henriksen, Larson and Van Tyne 4

Copyright 2003 by ASME

Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 12/17/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

were partially successful in the thick strip limit, but suffered from
a lack of quantitatively reliable constitutive data. In particular,
data available for rate-independent yield strength, recovery at
large strain and directional hardening have been insufficient to
accurately account for low temperature behavior and load reversals. The computational predictions of Henriksen et al. 4 also
illustrated a significant sensitivity to both phase transformation
kinetics and the flexural rigidity of the thin strips. They correctly
focused attention on the five constitutive parameters most influential in analysis of quenched specimens: the elastic moduli, yield
strength, hardening moduli, thermal expansion and phase transformation volume change. While they relied predominantly on literature data, the work presented here draws on data collected from an
expansive test series on low carbon steels 1.
Empirical methods for incorporating a temperature dependent
yield strength to simulate a phase change do not account for the
volume change or any additional plasticity 6,7. Mimicking the
phase change with large changes in the thermal expansion coefficient can capture the effects of the spherical volume change, but
not the microplasticity that accompanies them 6,8,9. Calculations that account for the dilatational volume change explicitly,
but not the accompanying microscopic deviatoric TRIP strains
exhibit discrepancies with measured residual stresses 10,11. For
problems that are predominantly deformation driven or exhibit
large kinematic constraints, the predicted stresses can be substantially in error. Simulations with and without this augmented plasticity indicate that neglecting this strain contribution can result in
residual stresses that are incorrect in sign and magnitude
6,12,13. Because transformation strains are accounted for locally, their effect on global distortions and residual stresses will,
in general, be problem dependent 8,12.
Finally, while previous empirical models for phase transformation kinetics have proven their utility for reconstructing CCT and
TTT curves, a rate formulation capable of recovering smooth
transformation rates and accurate volume fraction time histories is
required to construct a robust numerical tool for multiphase
behavior.

Description of the Almen Strip Problem


The present work focuses on a simplified test geometry similar
to that developed by J. O. Almen of General Motors Research
Laboratories 5. The Almen strip, described in detail by Henriksen et al. 4, was originally developed as a calibration test for the
shot peening industry. The strip is subjected to a shot stream,
relieving the near-surface, in-plane residual stresses imparted by
previous working of the material. This causes the strip to deflect
out-of-plane. This net deflection can then be related to the residual
stress depth profile.
Tests on modified Almen strips were performed by Larson 5.
Thin, wide strips of SAE 4023 and SAE 4620 alloy steel were
carburized on a single side and subsequently quenched. In the
presence of the surface carbon gradient, the temperature and carbon dependence of material properties and phase transformation
strains result in a net transverse deflection of the strip similar to
the specimens in Almens original shot peening experiments.
The specimen geometry is illustrated in Fig. 1. Specimens were
all 100 mm long and 20 mm wide. The thickness was varied to
alter the flexural rigidity and thus amplify the effects of the thermal and transformation strains. Strips with thicknesses of 1.3,
1.85, 2.03, 2.44, and 3.2 mm were heat treated in repeated trials.
Relatively thin samples in which the carbon diffuses into 3050
percent of the strip depth tend to amplify the effects of the carburized case on the overall distortion. In this way, the distortional
response of thin strips is distinguished from that of bulk components in which the shape changes are significantly affected by the
uncarburized core material.
Journal of Engineering Materials and Technology

Fig. 1 Almen strip geometry a and illustration of the carbon


gradient through the thickness b

Physical Mechanisms
Distortion and residual stresses can result from carburizing and
quenching alloy steel components. Both are caused by internal
loading from strains that evolve as a consequence of interstitial
lattice accommodation, cooling and phase transformation-induced
dilatation. In many cases and especially at high temperature, internal stresses are sufficient to exceed the local yield strength.
This gives rise to path dependent plastic flow. The levels of stress
are determined primarily by the timing of phase transformation
strains relative to the thermal contraction and the degree to which
these dilatations are suppressed by the surrounding material.
The carburizing process is performed at high temperature where
the diffusivity of carbon in steel is high. Upon brief exposure to a
carbon-rich atmosphere, carbon diffuses into the alloy lattice interstitial sites expanding the lattice and introducing a carbon gradient through the strip thickness. The carbon rich surface is
harder, more wear resistant and, upon quenching, results in a localized compressive residual stress state that strengthens the part
against imposed tensile stresses that accompany sustained fatigue
loading.
At cooling rates specified in the quench history, time scales are
too small for carbon diffusion to take place and austenite transforms via a shear mechanism to a body centered tetragonal structure known as martensite. This athermal transformation induces a
volumetric expansion because the lattice volume of the martensite
is larger than that of the austenite from which it formed. Further,
higher carbon contents reduce the temperature at which the martensite first forms and result in a larger volume change. The net
deflections of the carburized strip result from a series of competing strains imposed by the evolving microstructure.
First, the strip is uniformly heated and carburized resulting in
thermal expansion locally augmented by interstitial accommodation of carbon in the lattice. This results in an initial convex deflection of the strip due to carburization. Upon cooling, the low
carbon side of the strip reaches its transformation temperature
first. Martensite forming on this back side imposes a dilatational
strain that bends the strip back. Once the lower start temperature
of the carbon rich surface is reached, a larger dilatational strain
accompanying the transformation of high carbon martensite
changes the sense of deflection once more.
This series of alternating flexures results in a net convex distortion of the carburized surface and imbedded compressive residual
stresses at both surfaces. The experimental arc deflections reported by Larson 5 are illustrated in Fig. 2. The amplitude of the
arc deflections and severity of residual stress profiles are the result
of a complex path-dependent deformation history resulting from
thermal and phase transformation dilatational strains that are functions of the cooling and deformation rates, temperature and carbon
content.
APRIL 2003, Vol. 125 117

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 12/17/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 2 Distortion of Almen strip set as a function of strip thickness from 5. The top convex surface of the
strip is the carburized surface.

Model Formulation
The modified Almen samples were subjected to a heat treat
schedule composed of furnace carburizing followed by oil and air
quenches, respectively. The process schedule is specified by the
temperature history pictured in Fig. 3 reproduced from Larson
5. The strip is carburized in a three step furnace hold, quenched
in oil, and, finally, air cooled to room temperature. The heat treat
process is broken down into a series of three uncoupled analyses:
carburization, thermal analysis and structural analysis. The interaction of the three modules is schematically represented in Fig. 4.
A diffusion-based carburization analysis is performed by applying a specified carbon potential to one side of the strip. This carburization process is performed at high temperature 900C
where the diffusivity of carbon in steel is high. Carbon diffuses
into interstitial sites expanding the lattice and introducing a car-

bon gradient through the strip thickness. A set of nodal carbon


concentrations is then fed as input to a transient thermal analysis
wherein prescribed surface heat fluxes are applied corresponding
to time-temperature histories measured on other part geometries
quenched in similar oils. The analysis is coupled with a differential phase transformation model to account for latent heating when
computing the temperature field. The transformation kinetics
model parameters are temperature and carbon dependent. The
thermophysical properties are assumed to depend on phase alone.
Finally, a structural analysis is performed for the resulting phase
mixture. Initially the strips are heated to temperature, undergoing
thermal expansion and phase volume contraction on austenitization. On cooling, the model includes transformation-induced plasticity and an internal state variable description for the mechanical
response of individual phases.

An Internal State Variable Model


Internal state variable models have been successfully utilized
previously to describe the response of single phase metals over
large temperature and strain rate ranges and have been formulated
in a manner consistent with the kinematics of large deformation
elastic-plastic response 14. Generally, the internal state variables
are associated with underlying micro-mechanisms that are assumed to dominate the macroscopic mechanical response. Isotropic mechanisms are characterized by scalar internal state variables
while directional effects are modeled with tensor internal state
variables. At the macroscopic level, the introduction of these variables results in the prediction of strain rate, temperature and history dependent material response.
Bammann, Jin, Odegard and Johnson have proposed a model of
this type 14. In their formulation, a hypoelastic relation governs
the Cauchy stress in each phase, i, in a manner consistent with an
assumption of linear elasticity giving

i i W ei i i W ei 2 De i
Fig. 3 Heat treatment schedule for the experimental Almen
samples

118 Vol. 125, APRIL 2003

(1)

where () denotes the co-rotational derivative, De is the elastic


symmetric part of the deviatoric velocity gradient, is the tem(i)

Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 12/17/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 4 Uncoupled analyses available within the NCMS heat treatment process model

perature dependent shear modulus, assumed to be the same for all


phases, and W(ei ) is the skew part of the elastic velocity gradient
for each phase given by
Wei WWpi .

(2)

For the present purposes we choose a Jaumann derivative and all


W(pi ) 0.
The elastic deformation rate is defined as the difference between the total deformation rate and the sum of the dilatational
(i)
thermal deformation rate, Dth
, the phase transformation strain
(i)
rate, DPT , the inelastic deformation rate, D(pi ) , and phase
transformation-induced plastic strain rate, DTRI P
i
Dei DDTRI P Dpi Dthi DPT
.

(3)

The effective stress acting to cause plastic flow in each phase is


then given by
i i i i ,

(4)

The plastic flow rule is chosen to have a strong nonlinear dependence upon the effective stress
Dpi f i sinh

i Y i

V i

i i
,
i i

(5)

where the plastic deformation rate is deviatoric. Here, f ( ) and


V( ) describe a rate dependence of the yield stress at constant
temperature. The inelastic flow rule in 5 can be expressed as an
inverted yield condition
i i i i D ,
where
Journal of Engineering Materials and Technology

(6)

i D pi , V i sinh1


D pi

f i

(7)

is the instantaneous rate dependent yield stress of the material.


The function Y ( i ) ( ) is the rate-independent yield stress. The
function f ( i ) ( ) determines the cutoff strain rate below which the
material exhibits a regime of rate-insensitive yielding. The function V ( i ) ( ) determines the magnitude of rate-dependence on
yielding. These functions are determined from simple isothermal
compression tests at different strain rates and temperatures,
V i C 1 exp C 2 /
Y i

C3
1tanh C 19 C 20 exp C 4 /
2
f i C 5 exp C 6 /

(8)
(9)
(10)

Tensor and scalar internal stress variables, , and , represent


the directional and isotropic resistances to plastic flow, respectively. The tensor variable, , often referred to as the backstress or
the kinematic hardening variable, represents a short transient and
results in a smooth knee in the transition from elastic to elasticplastic response in a uniaxial stress-strain curve. What is more
important, this variable controls the unloading response and is
responsible for the apparent material softening upon unloading
termed the Bauschinger effect. Its importance in the prediction of
residual stresses in problems involving phase transformations was
detailed in 15. The scalar variable, , is an isotropic hardening
variable that predicts no change in flow stress upon reverse loading. This variable is a scalar measure of the dislocation density
that captures long transients and is responsible for the prediction
of continued hardening at large strains. Evolution of both state
APRIL 2003, Vol. 125 119

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 12/17/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 5 Mechanical response data points for 4120 austenite at


850C along with model fit solid lines using the multiphase
state variable plasticity model for 0.2 weight % carbon

Fig. 6 Mechanical response data points for 4120 martensite


at a strain rate of 0.001 s1 along with model fits solid lines
using the multiphase state variable plasticity model for 0.2
weight % carbon

variables is cast into a familiar hardening minus recovery format.


Both dynamic and thermal static recovery terms have been proposed for these variables.

parameter fits is important for determining residual stresses. The


formulation must also be generalized to handle multiple phases
undergoing transformation. Inclusion of a transformation-induced
plastic strain in the model internally loads the softer parent phase
material during transformation in the presence of applied stresses.
The plastic flow that accommodates the transformation and thermal strains is primarily confined to the austenitic phase. Thus, the
material fit to austenitic mechanical response data is the model
feature to which simulation results are most sensitive. Plastic flow
is often significantly less in the product phases.

i h i ,C DPi r Si r Di DPi i i (11)


i H i ,C DPi R Si R Di DPi i
r s( i ) ( )

(12)

R s( i ) ( )

and
describe the diffusionThe scalar functions
controlled static or thermal recovery, while r (di ) ( ) and R (di ) ( ) are
scalar functions describing dynamic recovery. The anisotropic
hardening modulus is h ( i ) ( ), and the isotropic hardening modulus is H ( i ) ( ). The temperature dependent hardening and recovery
parameters are given as
r di C 7 exp C 8 /

(13)

h i C 9 C 10

(14)

r si C 11 exp C 12 /

(15)

R di C 13 exp C 14 /

(16)

H C 15C 16

(17)

R si C 17 exp C 18 / .

(18)

The material constants, C 1 C 20 , had been fit previously to data


from uniaxial tension tests of 5120, 5140, 5160, and 5180 alloy
steels using a nonlinear regression algorithm 16. The test specimens were heat treated to isolate the four phase fields of austenite,
ferrite-pearlite, bainite and martensite. Each single phase specimen was tested over a series of temperatures and strain rates.
These data indicate that the mechanical behavior is most sensitive
to the carbon content and less sensitive to additional alloying elements. In the absence of a similarly full data set for the 4000
series alloys, several select tests were performed on 4120 alloy
specimens. We then evaluated the ability of the 5100 model parameter fit to predict the response of the 4120 alloy specimens.
Figure 5 shows a comparison of the model predictions for 0.2%
carbon austenite with experimental results for an SAE 4120 alloy.
Figure 6 shows a similar model prediction for the response of
0.2% martensite compared with measured 4120 behavior. While
the model was fit for the 5100 series of steel alloys, the parameterization predicts the 4100 alloy series mechanical response reasonably well.
In reverse loading tests performed subsequent to the uniaxial
tensile data, the directional nature of hardening was evident. Other
coupled, thermo-mechanical simulations performed for welding
15 have shown that including reverse loading data in the model
120 Vol. 125, APRIL 2003

Transformation Plasticity
For multiphase materials, the effects of phase transformation
induced plastic strain must be included. When steel alloys are
cooled from above the austenization temperature, a solid state
phase transformation occurs, resulting in product phases that are
larger in volume and harder than the surrounding parent phase.
This volume difference between the crystal structures causes microscopic plastic flow as the transformation proceeds. In the presence of a deviatoric stress field the straining from the volume
misfit exhibits a deviatoric part proportional to the surrounding
stress. Since the product phases are generally stronger, this results
in an internal stress in the austenite. In this pre-loaded state, the
multiphase composite can yield when the applied stress is less
than the yield strength of the virgin austenite. Thus, the multiphase steel exhibits an apparent softening during phase transformation. Accounting for the resulting inelastic straining in the flow
rule provides a natural way to model the multiphase yield drop
and additional plasticity characteristic of transforming alloys.
Leblond et al. 17,18 have outlined an extensive theoretical
treatment of transformation plasticity. Leblond et al. 19,20 further provided a mathematical overview wherein the transformation plastic strain rate is given by
TRI P

K V


y V

(19)

The transformation plastic strain rate is proportional to the ap,


plied stress deviator, , the rate of change of volume fraction,
the volume misfit, V/V, and is inversely proportional to the
yield strength of the austenite, y . What is interesting is that the
transformation plastic strain rate is proportional to the deviatoric
stress 6. This resembles the phenomenological flow law for internal state variable viscoplasticity indicating that the transformation plasticity can be cast in a similar framework.
To generalize the internal state variable model for multiphase
materials, it is assumed that each point of the continuum is occuTransactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 12/17/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 7 Thermal and phase transformation dilatational strains


on cooling for low carbon steel alloys. Axial contraction is referenced arbitrarily to zero at 840C and plotted as the ordinate.
Fig. 9 Comparison of the Weibull parameterization for TRIP
with the previous formulation and experiments reported by
Leblond

pied simultaneously by all phases. The internal state variable


model is used to determine the stress response of an individual
phase. The transformation strain experienced by an individual
i)
phase can be computed by integrating the deformation rate, D (PT
in Eq. 3, over a time step, t. The dilatational transformation
strain increment given by
i
i j t
D PT

E M E A
ij
1E A

(20)

where
E M M 0 M 1 M 2 2 M 3 3

(21)

is the temperature dependence of dilatational strain in the martensite and


E A A 0 A 1

(22)

is the corresponding temperature dependence in the austenite. Figure 7 illustrates the extensional strain produced in dilatometer
specimens by thermal contraction and martensitic phase transformation as a function of temperature and carbon content. The deviation of each curve from the linear thermal contraction is precisely the transformation strain given in Eq. 18. The coefficients
appearing in Eqs. 21 and 22 are carbon-dependent as well.
This carbon dependence is illustrated for martensite in Fig. 8 for a
series of discrete temperatures.
We propose a new form for the deviatoric TRIP strain rate
motivated by limit behaviors of the Weibull probability distribution function

P
D TRI

Am
B B

m1

exp

(23)

where A, B, and m are material parameters which must be fit to


stressed dilatometry test data. Here the Weibull distribution function was chosen because it naturally achieves the limits necessary
for the volume fraction functional. The Weibull distribution, used
for in Eq. 19, ensures that the total integrated TRIP strain
is zero at zero volume fraction of transformed phase. Further, in
the limit of transformed volume fraction of unity, the TRIP strain
rate identically vanishes. We should point out that this choice for
the volume fraction function has no other inherent physical basis.
That is, it is not derived from first principles. Rather, it is purely
phenomenological and motivated by the qualitative comparison
and known limit behaviors. The Weibull function is fit to experimental data in Fig. 9 and compared with the functional representations reported in Leblond 19.
The state variable plasticity routine is used to determine the
respective phase stresses. Once the individual phase stresses are
calculated, the composite deviatoric Cauchy stress, , is given
by

(24)

and the volume fractions of phases are subject to the constraint


that they must sum to one

1.
i

(25)

Phase Transformation Kinetics

Fig. 8 Linear transformation strains for martensite formation


on cooling for low carbon steel alloys

Journal of Engineering Materials and Technology

We presume that athermal transitions are a special case of isothermal kinetics in which the mobility function approaches infinity. This implies that, for martensite formation isothermal martensite is not considered here, a time-dependent kinetic equation
is not necessary since equilibrium is always attained instantaneously upon changing the temperature. Figure 10 illustrates this
idea through a plot of free energy versus temperature. The energy
minimum corresponds to the equilibrium volume fraction of martensite. As temperature decreases during quenching, this minimum
shifts to higher martensite volume fractions. Experimental curves
depicting martensite volume fraction as a function of temperature
can therefore be interpreted as a description of the minimum value
of a temperature dependent energy function.
The most commonly used description of such well states is the
Koistinen-Marburger equation 21,
APRIL 2003, Vol. 125 121

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 12/17/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 12 Comparison of experimental data with the new kinetics DANTE model prediction and the prediction of the
Koistinen-Marburger K-M equation
Fig. 10 Schematic indicating the shifting of the equilibrium
volume fraction of martensite as a function of temperature

1exp b M S U M S ,

(26)

where M S is the martensite start temperature, b0.011 (C1 )


and U( ) is the Unit Step Functioni.e.,
U M S 1,

M S

U M S 0,

M S

M S
U M S ,
M S M F

d
d
M C C 1 C U M S
dt
dt

(27)

Another common equation of this sort is the Skrotzki equation


22:
1

rotzki equations is that the discontinuity in the transformation rate


at the start temperature can cause numerical difficulties.
Motivated by these issues, and guided by our previous internal
state variable modeling of diffusive phase transitions we recently
proposed a new equilibrium equation for athermal martensite
2,24:

(28)

where the exponent r is in the range of 2 to 3 and M F is the


martensite finish temperature. However, both of these equations
deliver concave curves no inflection point in graphs of martensite volume fraction versus temperature. As a result, they are most
suitable for burst-like transformation 22 and not for a smooth
transformation wherein the martensite volume fraction is known
to be a sigmoidal function of temperature. Grange and Stewart, for
instance, show a number of examples of such sigmoidal behavior
23. Another disadvantage of the Koistinen-Marburger and Sk-

where

C 0 1C
C 0 1C

122 Vol. 125, APRIL 2003

(30)

M C 0 1 C 2 C 2 3 C 3
and C is the carbon content. This equation represents a differential
relationship between martensite volume fraction and temperature,
but it is cast in a time-rate form so that it can be solved incrementally with the thermal simulation. The Koistinen-Marburger equation is recovered as a special case when 0, 1, and b.
The best match with dilatometry experiments was obtained using the Andrews nonlinear equation for the martensite start temperature, M S 25:
M S C 512453C16.9Ni9.5M o217C 2 71.5 C M n
15Cr67.6 C Cr 10Co7.5Si

Fig. 11 Comparison of theory and experiment for martensite


formation in low carbon steels. Predicted volume fractions
solid lines compare well with experimental data points.

(29)

(31)

Fig. 13 Almen strip carbon profile with depth from the carburized surface for h 1.30 mm

Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 12/17/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 14 Almen strip carbon profile with depth from the carburized surface for h 3.18 mm

Here the alloying elements are given in weight per cent. There are
three additional functions that need to be fitted, , and . As
indicated above, each of these functions is assumed to be a polynomial function of carbon level. A differential fitting routine was
developed to obtain the optimum values of these parameters for
low carbon alloys.
This model was used to generate the martensite volume fraction
thermal histories shown in Fig. 11. Note that the dilatometry data
was used to fit the kinetics model, so the results shown are not so
much a prediction of the model as much as a demonstration that
the model can capture the desired kinetic behavior. Figure 12
compares the predicted martensite volume fraction curve predicted by the new kinetics model with a similar prediction using
the Koistinen-Marburger equation. Note the burst-type behavior
predicted by the Koistinen-Marburger equation at the onset of the
transformation. This abrupt phase change behavior is not observed
experimentally.

Model Simulation & Comparison With Experiments


A series of uncoupled, plane strain finite element analyses were
performed using the DANTE User Material interface to the finite
element code ABAQUS. The finite element mesh of the strip consisted of 100 quadrilateral elements, twenty along the strip length
and five through the thickness. The meshes were self-similar for
the series of progressively thicker strips. The eight node quadratic

Fig. 15 Transverse distortion of Almen strip as a function of


strip thickness

Journal of Engineering Materials and Technology

Fig. 16 Predicted Almen strip transverse deflection histories.


Initial deflections result from carburization.

displacement element CPE8 was used for the mechanical analysis


and the corresponding heat transfer and mass diffusion element
DC2D8 was used for the carburization and thermal analyses.
Thermal boundary conditions were obtained by an inverse method
from data obtained on similarly quenched components. Measured
surface temperatures from these parts were applied to a finite element model. The calculated surface heat fluxes were then converted to an equivalent set of surface heat transfer coefficients.
The resulting thermal field is provided as input to the mechanical
analysis.
The results of the carburization analysis are highlighted in Figures 13 and 14 for two representative thicknesses. The calculated
carbon profiles compare well with estimates of the carbon content
derived from hardness measurements. These translate into similar
agreement for the martensite phase distribution. We obtained very
good agreement between calculated and measured transverse distortion as illustrated in Fig. 15. In addition, the predicted deflections, plotted as a function of temperature drop in Fig. 16, serve to
corroborate the transient scenario outlined in the previous section.
That is, the strips first deflect in one sense upon transformation of
the uncarburized side. The sense of deflection then reverses when
the carburized side reaches its martensitic start temperature,
nearly 200C above room temperature.
While excellent agreement was obtained for arc deflection,
a stronger test of the model lies in its ability to predict residual
stress profiles. Model predictions are shown in Figs. 17 and 18 for

Fig. 17 In-plane residual stress depth profile at the center


of the 2.44 mm thick Almen strip. Predicted stresses solid
lines compare well with X-ray measurements points from
Larson 5.

APRIL 2003, Vol. 125 123

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 12/17/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 18 In-plane residual stress depth profile at the center


of the 3.18 mm thick Almen strip. Predicted stresses solid
lines compare well with X-ray measurements points from
Larson 5.

the 2.44 mm and 3.2 mm strips, respectively. Larson and Van


Tyne report X-ray diffraction measurements of the near-surface
in-plane residual stress profiles for these strip thicknesses. In
both cases, considering the scatter of X-ray data, the agreement
is very good.

Conclusions and Recommendations


Henriksen, Larson, and Van Tyne 4 question a variety of assumptions in their model formulation including necessity of
stress-dependent kinetics and the applicability of the rule of mixtures. The present analysis appears to confirm that a rule of mixtures is reasonable. Based on the results obtained here, the model
parameters most significant to realistic simulations are the phase
transformation strains, TRIP parameterization and elasticity. Since
the elastic deformation can be a significant portion of the overall
plate bending in thicker sections, accurate temperature dependence of the elastic constants becomes essential.
We agree wholeheartedly with Henriksen et al. 4 in their assessment that simple problems, such as that described herein,
should be verified. This is a task best undertaken before more
involved two and three-dimensional simulations are performed on
more complex geometries where the singular effects of particular
model components can not be as readily isolated and explained. In
exercising the heat treat process model, DANTE, on the modified
Almen strip, we have shown that numerical simulation of
quenched components is no longer of questionable value. Rather,
the computational horsepower and modeling techniques are available to make such analyses practical provided the phase transformation parameterization and constitutive description of the material can be sufficiently characterized.

References
1 Predictive Model & Methodology for Heat Treatment Distortion, 1997
Phase I Project Summary Report, Constitutive Model Task Team; NCMS
Project, January 1994 December 1997.

124 Vol. 125, APRIL 2003

2 Lusk, M., Krauss, G., and Jou, H.-J., 1995, A Balance Principle Approach for
Modeling Phase Transformation Kinetics, Journal de Physique IV, C8, pp.
279284.
3 ABAQUS Users Manual V5.8, 1998, Hibbitt, Karlsson and Sorensen Pawtucket, RI.
4 Henriksen, M., Larson, D. B., and Van Tyne, C. J., 1992, On the Analysis of
Distortion and Residual Stress in Carburized Steels, ASME J. Eng. Mater.
Technol., 114, pp. 362367.
5 Larson, D., 1990, Finite Element Analysis of Residual Stress and Distortion
in Forged and Carburized Gear Steel, M.S. thesis T-3932, Colorado School of
Mines, Golden, CO.
6 Goldak, J. A., 1990, Modeling Thermal Stresses and Distortions in Welds,
Recent Trends in Welding Science and Technology, S. A. David and M. Vitek,
eds., ASM International, Gatlinburg, TN, pp. 71 82.
7 Dubois, D., Devaux, J., and Leblond, J. B., 1984, Numerical Simulation of a
Welding Operation: Calculation of Residual Stresses and Hydrogen Diffusion, Proceedings of the 5th International Conference on Pressure Vessel
Technology, 2, American Society of Mechanical Engineers, New York, pp.
12101239.
8 Sjostrom, S., 1985, Interactions and Constitutive Models for Calculating
Quench Stresses in Steel, Mater. Sci. Technol., 1, pp. 823 829.
9 Goldak, J. A., 1986, Computational Weld Mechanics, Proceedings of the
International Symposium on Computer Modeling of Fabrication Processes and
Constitutive Behavior of Metals, Ottawa, Canada.
10 Oddy, A. S., Goldak, J. A., and McDill, J. M. J., 1992, Transformation Plasticity and Residual Stresses in Single-Pass Repair Welds, ASME J. Pressure
Vessel Technol., 114, pp. 3338.
11 Josefson, B. L., and Karlsson, C. T., 1992, Transformation Plasticity Effects
on Residual Stresses in a Butt-Welded Pipe, ASME J. Pressure Vessel Technol., 114, pp. 376 378.
12 Oddy, A. S., Goldak, J. A., and McDill, J. M. J., 1990, Numerical Analysis of
Transformation Plasticity in 3D Finite Element Analysis of Welds, Eur. J.
Mech. A/Solids, 93, pp. 253263.
13 Goldak, J. A., 1991, Coupling Heat Transfer, Microstructure Evolution and
Thermal Stress Analysis in Weld Mechanics, Proceedings of the IUTAM Symposium, Lulea, Sweden, 131.
14 Bammann, D. J., Chiesa, M. L., and Johnson, G. C., 1996, Modeling Large
Deformation and Failure in Manufacturing Processes, in Proceedings of the
19th International Congress of Theoretical and Applied Mechanics, Tatsumi,
Wannabe and Kambe, eds., Kyoto, Japan, pp. 359376.
15 Bammann, D. J., and Ortega, A. R., 1993, The Influence of the Bauschinger
Effect and Yield Definition on the Modeling of Welding Processes, Welding
and Advanced Solidification Processes-VI, T. S. Piwonka, V. Voller, and L.
Katgerman, eds., The Minerals, Metals & Materials Society, Warrendale, PA,
pp. 543551.
16 Lathrop, J. F., 1997, BFITA Program to Analyze and Fit the Sandia Plasticity BAMMANN Model Parameters to Experimental Data, SAND978218.
17 Leblond, J. B., Mottet, G., and Devaux, J. C., 1986, A Theoretical and Numerical Approach to the Plastic Behavior of Steels During Phase Transformations: IDerivation of General Relations, J. Mech. Phys. Solids, 344, pp.
395 409.
18 Leblond, J. B., Mottet, G., and Devaux, J. C., 1986, A Theoretical and Numerical Approach to the Plastic Behavior of Steels During Phase Transformations: IIStudy of Classical Plasticity for Ideal-Plastic Phases, J. Mech.
Phys. Solids, 344, pp. 411 432.
19 Leblond, J. B., Devaux, J., and Devaux, J. C., 1989, Mathematical Modeling
of Transformation Plasticity in Steels I: Case of Ideal-Plastic Phases, Int. J.
Plast., 5, pp. 551572.
20 Leblond, J. B., Devaux, J., and Devaux, J. C., 1989, Mathematical Modeling
of Transformation Plasticity in Steels II: Coupling With Strain Hardening Phenomena, Int. J. Plast., 5, pp. 573591.
21 Koistinen, D., and Marburger, R., 1959, Acta Metall., 7, p. 59.
22 Skrotzki, B., 1991, J. de Physique IV Colloque, C81, p. 367.
23 Grange, R., and Stewart, H., 1946, Trans. AIME, 167, p. 467.
24 Bammann, D., Prantil, V., Kumar, A., Lathrop, J., Mosher, D., Lusk, M., Jou,
H.-J., Krauss, G., Elliott, W., 1996, Proceedings of the 2nd International Conference on Quenching and the Control of Distortion, G. Totten, M. Howes, S.
Sjostrom, and K. Funatani, eds., ASM, p. 367.
25 Andrews, K. W., 1965, Empirical Formulae for the Calculation of Some
Transformation Temperatures, J. Iron Steel Inst., London, pp. 721727.

Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/ on 12/17/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like