Professional Documents
Culture Documents
Heat Treat Simulation: Benefits of DANTE
Heat Treat Simulation: Benefits of DANTE
Introduction
Figure 2: (A) Full, 3D model of geometry and (B) Quarter Figure 4 shows the ambient gas temperature as a function of
model used for study. time used for the modeling study. The curve describes a dual
chamber system, whereas the single component, or small load,
The 3-dimensional solid model of the quarter section geometry is heated in one chamber under vacuum or a protective
was then meshed in Abaqus for finite element analysis (FEA). atmosphere, transferred in a vacuum chamber, and loaded into
Abaqus Standard was used as the FE solver, and DANTE was a pressure vessel at room temperature. The ambient gas
used to predict the material’s response during heat treatment. temperature in the vessel rapidly increases as the hot load is
Figure 3 shows the quarter model properly meshed for heat introduced, then cools as the heat exchanger continually cools
treatment simulation. Figure 3(A) shows the fine mesh near all the quench gas, reducing the component’s temperature.
surfaces which would be exposed to the atmosphere. Figure
3(B) shows a close-up of the mesh near a corner. The fine mesh 350
near the surface is required to properly describe the steep
Ambient Temp., °C 300
temperature and phase transformation gradients that occur from
the surface towards the core during the heat treating process. 250
200
150
100
50
0
0 10 20 30 40 50 60
time, sec
Figure 3: (A) Mesh used for quarter model and (B) Close-up of Distortion from Common Gas Quenching Rates
surface mesh. Using the DANTE heat treatment simulation software, the
Ferrium C64 disk was subjected to a range of possible HPGQ
Process Description HTCs, using the time-temperature curve in Fig. 4. The
The study examined a range of heat transfer coefficients (HTC) following heat transfer coefficients were evaluated: 10, 25, 50,
typical of HPGQ and the distortion response of the coupon 100, 200, 300, 400, 500, 600, 700, 800, 900, and 1000 W/m2K.
shown in Fig. 1 made from Ferrium® C64TM. The chemical Generally, HTCs of 200, 400, 700, and 1000 W/m2K
composition of Ferrium C64 is shown in Table 1. correspond to quench pressures of 2, 6, 10, and 20 bar,
respectively. This relationship will vary depending on the flow
Table 1: Chemical composition of Ferrium C64. pattern and velocity profile of the quench gas in the vessel and
around the component.
229
400
350
300
250
200
150
100
0 100 200 300 400 500 600 700 800 900 1000
HTC, W/m2K Figure 7: Snapshot of heating, just prior to austenite
transformation, showing austenite phase fraction (left) and
radial displacement of bore in mm (right).
Figure 6: Out-of-round distortion of the coupon bore as a
function of the heat transfer coefficient. Austenite transformation begins in the thinnest section first.
The increase in density from martensite to austenite reduces the
radial dimension at the thin section of the bore. This makes the
It is commonly assumed that if a component is quenched bore slightly larger in the horizontal direction than in the
slower; i.e., a lower HTC value, less distortion will occur. As vertical direction. As the austenite progresses up the sides of the
can be seen in Fig. 6, the highest HTC studied nearly produced bore, the increase in density begins to reduce the radial
the least amount of out-of-round distortion. The only HTCs that dimension around the bottom half of the bore. This increases
produced less distortion in this case can be associated with an the horizontal dimension of the bore even further, as shown in
air cool (10, 25, and 50 W/m2K). The most distortion occurred Fig. 8.
from an HTC of 400 W/m2K. This observation contradicts the
above assumption and leads to the question: How can
quenching slower produce more distortion?
231
HTC Comparison
Referencing Fig. 6, since 100 W/m2K (HTC100) produces more
distortion than 1000 W/m2K (HTC1000), but produces a
significantly slower cooling rate in the component, these two
conditions will be compared. These two conditions will also be
Figure 15: Snapshot during quenching after the martensite has compared to the HTC that causes the most out-of-round
completed in thin section and has begun around the top half of distortion of the bore, 400 W/m2K (HTC400).
the bore. Shown is martensite phase fraction (left) and radial
displacement of bore in mm (right). Table 2 outlines the out-of-round distortion for each of the three
conditions caused by Heating Only, Quenching Only, and from
As the martensite transformation begins to rapidly progress the Entire Process. Also included is the sum of Heating Only
through the thick section, the reduction in density pushes the and Quenching Only, compared to the distortion from the Entire
thick section out in the horizontal direction. This expansion Process. Since the heating was the same for all quenching
reduces the oblong shape of the hole somewhat, but it is not HTCs, the difference in distortion must be from the quenching
process. It should also be noted that the total out-of-round is the
232
HTC100 HTC400 HTC1000 To determine how the highest HTC evaluated can produce less
From From From distortion than one of the lowest HTCs evaluated, a comparison
Quenching Quenching Quenching between these three HTC values during the martensite
out-of-round from
245.6 μm 319.1 μm 178.1 μm transformation is warranted. Figures 19 – 23 show martensite
Quenching phase fraction comparisons between the three HTC values: 100,
400, and 1000 W/m2K. Figure 18 describes the two views
HTC100 HTC400 HTC1000 shown in Figs. 19 – 23.
From From From
Heating Heating Heating
out-of-round from
135.6 μm 135.6 μm 135.6 μm
Heating
Sum of Quenching
& Heating 381.2 μm 454.7 μm 313.7 μm
Distortion
Percent Difference
of Sum and 6% 7% 0%
Complete Process
Figure 18: Description of views shown in Figures 19 – 23.
The out-of-round values reported in Table 2 include distortions
from phase transformations and thermally induced plastic Each quarter section of the geometry will behave the same due
strain. Using the DANTE software, it is possible to disable the to the geometric symmetry and the uniform HTC applied to all
phase transformations during the analysis. Table 3 shows the surfaces exposed to the atmosphere. The comparisons between
out-of-round distortion of the bore for heating and quenching the three HTC values are made when the maximum martensite
without the phase transformations to austenite and martensite, phase fraction is approximately equal in particular sections of
respectfully. The mechanical properties of austenite were used the coupon. The times at which these values occur will vary
for the predictions reported in Table 3. depending on the HTC value.
Table 3: Comparison of Thermally Induced predicted distortion Figure 19 shows a comparison when the martensite phase
of bore for HTCs of 100, 400, and 1000 W/m2K. reaches 80% for the first time. For all HTC values, this occurs
on the outer edge of the thin section. At this snapshot in time,
HTC 100 No HTC 400 No HTC 1000 No all three HTC values show a similar pattern. The maximum
Heat, No Heat, No Heat, No amount of martensite is at the thinnest section of the part on the
Phase Phase Phase
surface of the end face. However, a few differences can be seen.
out-of-
round from 5.8 μm 5.8 μm 5.8 μm Firstly, HTC100 has a significant amount of martensite,
Heating approximately 50%, already half way around the bore, with
out-of- approximately equal amounts of martensite at the surface and
round from 25.5 μm 119.3 μm 112.6 μm in the core. Secondly, HTC400 also has an area of 50%
Cooling martensite, but it is concentrated in the thin section of the
coupon. The martensite fraction for HTC400, like HTC100, is
The thermally induced distortion from heating is very small, almost equivalent on the surface and in the core. Thirdly, the
5.8μm, due to the ramped heating. This small value can be HTC1000 martensite has progressed approximately the same
considered insignificant and indicates that the nonuniform distance around the bore as HTC400, but there is almost no
austenite transformation is responsible for the distortion during martensite transformation occurring in the core at this time.
233
The next snapshot, Figure 22, shows the moment when the
martensite transformation completes in the thin section of the
coupon. It is now obvious as to why HTC400 produces the
greatest amount of out-of-round distortion: A uniform
transformation from the surface to the core, combined with a
steep transformation gradient circumferentially around the
bore, results in the bore being elongated in the vertical direction
with no offsetting effect from transformations occurring in the
thick section of the coupon. Although the transformation has
started near the surface of the thick section of HTC400 in Fig.
22, the damage has already been done by the transformation
completing in the thin section and nearly completing up the
sides of the bore before starting in the thick section.
234
The martensite transformation behavior from the HTC100 trial Additional Observations Concerning the Eccentric Bore
is in stark contrast to that of the HTC400 trial. The Due to differences in transformation timing at critical locations
transformation for HTC100 progressed with a very shallow of the part, the distortion response did not behave as anticipated.
transformation gradient around the circumference of the bore, The model presented used a simple geometry and a perfectly
with the martensite beginning in the thick section well before uniform heat transfer coefficient during heating and quenching.
the transformation was complete in the thin section. This Reality is never this kind. In most cases the ambient
allowed the elongating effect in the vertical direction, created temperature and/or the HTC in the bore would vary from the
by the martensite transformation progressing up the sides of the outer diameter and end faces of the coupon used in the study.
bore, to be offset by the stretching effect in the horizontal Table 4 looks at five (5) scenarios of differing boundary
direction, created by the transformation to martensite in the conditions placed on the bore of HTC400, during quenching
thick section. Figure 22 shows that there is approximately 90% only, and compares them to the quench only HTC400 result.
martensite halfway up the side the bore and nearly 60% The ambient temperature profile and HTC from the above study
martensite in the thick section of HTC100. were used for the outer diameter and end face surfaces of the
coupon.
HTC1000’s transformation around the bore progresses much
like HTC400, but with one critical difference: As the martensite While gas quenching offers a more uniform method of heat
progresses up the sides of the bore, martensite is forming in the extraction than liquid quenching, geometry can alter this
thick section. While Fig. 22. shows martensite in the thick uniformity. Generally, the ambient temperature in the bore of
section of HTC400 and HTC100, martensite in HTC400 only the geometry under study would be higher than other surfaces,
began forming after it had progressed a substantial amount due to radiant surfaces facing each other. The second and third
around the bore. HTC1000, on the other hand, had martensite rows of Table 4 show the out-of-round distortion results of the
transforming in the thick section before it had progressed bore if the ambient temperature of the bore was 1.5 and 2 times
halfway around the bore in the core, as shown in Fig. 21. higher than the outer diameter and end faces, respectively. The
HTC is the same for all surfaces. The increase in ambient
Figure 23 shows the process as it is completing. It is worth temperature essentially slows down the heat transfer through
noting the difference between HTC400 and HTC1000. HTC400 the bore wall and has a minimal effect on the total out-of-round
has a shallow gradient in the thick section from the surface to distortion compared to the situation where the ambient
the core. This means there is a large amount of volume temperature is the same for all surfaces.
transforming in the thick section all at once, acting to stretch the
hole in the horizontal direction, after the material around the A uniform HTC on all surfaces of this geometry is also highly
bore had already transformed. HTC1000, on the other hand, has unlikely in actual practice. There are two scenarios possible.
a steep transformation gradient and has much less material to First, air has a difficult time moving through the bore and the
transform before completion. This is due to the transformation HTC in the bore is reduced compared to the other surfaces. The
in the thick and thin sections occurring simultaneously. fourth row in Table 4 shows the results if the HTC in the bore
is 0.5 times the HTC applied to other surfaces. Like the
increased ambient temperature, this boundary condition acts to
235
110
out-of-round distortion, μm
105
100
95
90
85
80
0 100 200 300 400 500 600 700 800
HTC, W/m2K
236
LOW PRESSURE
CARBURIZATION
MODELING
WITH CARBIDE FORMATION
AND DISSOCIATION
26 OCTOBER 2020
With the advent of high-strength steels for the aerospace industry,
most of which contain substantial amounts of strong carbide-
forming elements such as chromium, molybdenum, and vanadium,
the LPC process can be extremely challenging to control.
By JUSTIN SIMS
AEROSPACE CHALLENGES capability and was used to fit the diffusivity and primary carbide
For steel alloys that do not contain a significant amount of strong kinetics of Ferrium C64 steel alloy from LPC experiments.
carbide forming elements, this process can be relatively easy to LPC experiments were conducted whereby 4-inch diameter cyl-
control. However, with the advent of high-strength steels for the inders made of Ferrium C64 were subjected to three different LPC
aerospace industry, most of which contain substantial amounts of cycles with the only difference in the cycles being the final diffuse
strong carbide-forming elements such as chromium, molybdenum, time. Each run used a progressively longer final diffuse time. Carbon
and vanadium, the LPC process can be extremely challenging to measurements were made using a LECO burn. Unfortunately, this
control. These elements bond with the carbon deposited on the method of carbon analysis gives no indication as to the amount
surface to form primary carbides, which, if not properly dissolved, of carbon in solid solution and the amount of carbon in primary
can damage fatigue performance. Long boost times, as well as very carbide form.
short diffuse times, can lead to the thermodynamic stabilization Regardless, the data was fit to diffusivity and primary carbide for-
of these primary carbides. mation/dissolution kinetics parameters used in the DANTE carburi-
While Fick’s Second Law describes the diffusion of carbon through zation model and compared to the experimental data. The results
a low-alloy steel with reasonable accuracy, the same is not true of are shown in Figure 1. As can be seen, the match between prediction
medium- and high-alloy steels. This is due to the presence of primary and experiment is reasonable, with a significant amount of scatter
carbides forming and dissolving during the LPC process. When the in the experimental data. With the ability to predict carbon diffu-
boost step is occurring, the primary carbides take carbon away from sion and carbide formation/dissociation, the DANTE model can now
the austenite solution, allowing more carbon to enter the solid solu- be used to predict and design LPC schedules for Ferrium C64 steel.
tion from the surface. During the diffuse step, as the carbon that is in
solid solution diffuses into the part effectively reducing the carbon CASE STUDY
in austenite, the primary carbides can dissolve to provide more car- The following is a case study for the redesign of an LPC schedule for
bon to the solid-state solution. If the primary carbides are not allowed a ring gear using the DANTE heat-treatment simulation software.
to fully dissolve or shrink to a significantly small amount before the The original LPC schedule, consisting of six boost-diffuse steps, was
next boost step begins, they will continue to grow. producing too many primary carbides during the process, and, as
a consequence, rolling contact fatigue performance was very poor.
PRIMARY CARBIDE FORMATION AND DISSOLUTION Since the carbon potential should act uniformly around the cir-
In order to properly predict the carbon profile of medium- and high- cumference of the gear and each tooth should behave the same
alloy steels, the primary carbide formation and dissolution must be with respect to carbon diffusion, only a single tooth was modeled.
considered. The heat-treatment simulation software DANTE has this Figure 2 shows the full CAD model of the gear (left), the single
thermalprocessing.com 27
tooth meshed model (center), and a closeup of the gear tip corner
mesh (right). An extremely fine surface layer of elements is required
to capture the steep chemical gradient that exists near the surface
during an LPC process. Only the cross-section is shown, though the
gear does have a height. The gear has a 5.5-inch outer diameter, a
4.5-inch inner diameter, a 0.60-inch height, and 40 teeth. The gear is
made of Ferrium C64 with a base carbon of 0.1%. The model contains
233,850 hexagonal linear elements and 245,055 nodes.
The case depth was originally designed for 0.75 mm on the flank
of the tooth, with a carbon value of 0.3% resulting in a hardness
value of 50 HRC for Ferrium C64 when tempered at 495°C. Figure Figure 2: CAD model (left), single tooth meshed model (center), and close-up of
3 shows a plot of the carbon in the austenite matrix (carbon) and tooth tip edge fine mesh (right).
the carbon in primary carbide form (carbon in carbide) from the
surface of the flank toward the core. As can be seen, a great deal of
carbon is tied up in primary carbide form — approximately 0.75%.
These primary carbides, if not dissolved either during a reheating
process to form austenite or at the end of the carburization cycle,
can be extremely detrimental to fatigue performance. Figure 3 also
shows that the case is deeper than it needs to be, at 1.1mm.
An added benefit of using simulation software like DANTE to
design LPC schedules is the ability to witness the effect of each indi-
vidual boost-diffuse step. It is therefore possible to know when to
reduce the boost time to avoid large primary carbide formation and
how long to diffuse to ensure the primary carbides are dissolved to
a sufficiently small size. Most medium- and high-alloy steels will
form primary carbides during the boost step, regardless of the time.
So, it is important to allow enough carbon to enter the part before Figure 3: Carbon in the austenite matrix and carbon in primary carbide form for
beginning the diffuse step. the baseline process from the surface of the flank.
Controlling the length of the diffuse step then becomes critical
in ensuring the primary carbides have properly dissolved before
beginning the next boost step. Figure 4 shows the predicted carbon
in the austenite matrix (carbon) and the carbon in primary carbide
form (carbon in carbides) at the surface of the flank for the baseline
model over the total time of the process. As can be seen, the carbon
in primary carbide form continues to increase as the process pro-
gresses. The long final diffuse begins to allow for decomposition,
but the time is not long enough.
Figure 4 also indicates that the diffuse times are much too short,
as the carbon in carbide does not dissociate back into the austenite
matrix, but remains more than 1.0% for most of the process. The
boost times seem reasonable, since the carbon in primary carbide
form rises approximately 1.2%. This is reasonable for Ferrium C64,
as the high level of chromium causes rapid formation of primary Figure 4: Carbon in the austenite matrix and carbon in primary carbide form for
the baseline process on the surface of the flank for the total process time.
carbides. The important insight gained from Figure 4 is that the dif-
fuse times are far too short to properly dissolve the primary carbides dissolving to a minor fraction of a percent. Comparing to Figure 4,
back into solid-state solution. this matches the first two boost steps, but the baseline carbon in
primary carbide form contained 1.8% carbon in later boost steps. So,
A NEW SCHEDULE the redesign had a significant effect on primary carbide formation.
To ensure the primary carbides dissolve completely before harden- The baseline model also showed the surface carbon near 1.0%,
ing, a new schedule was developed with the aim of reducing the car- which is much higher than what is needed to reach a hardness of
bon in primary carbide form. To reach this goal, three boost-diffuse 60 HRC. For Ferrium C64, a surface carbon of 0.55% is sufficient
steps were removed, and the diffuse times increased substantially. to reach a surface hardness of 60 HRC after tempering at 495°C.
This increase in diffuse time added approximately one-half hour to Therefore, the surface carbon was also reduced in the redesigned
the schedule, which is acceptable given the positive results. Figure schedule. Figure 6 shows the carbon in the austenite matrix (carbon)
5 shows the predicted carbon in the austenite matrix (carbon) and and the carbon in primary carbide form (carbide) from the surface of
the carbon in primary carbide form (carbon in carbide) for the total the flank toward the core for the redesigned schedule. By reducing
process time at the same surface location on the flank as Figure 4 the surface carbon, this material was able to meet the required hard-
for the baseline LPC schedule. ness, case depth, and avoid significant primary carbide formation
As can be seen in Figure 5, the primary carbides now have time during the LPC process. The carbon in primary carbide form on the
to dissolve during each diffuse step and are nearly eliminated by surface can now be easily removed with a light finishing operation,
the final diffuse step. It is also interesting to note that the carbon whereas a significant amount of grinding would have been required
in primary carbide form rises to 1.3% during each boost step before to remove the primary carbide layer from the baseline process.
28 OCTOBER 2020
CONCLUSION
It has been shown that it is possible to model a low-pressure carburi-
zation process, including the formation and dissociation of primary
carbides using the DANTE heat-treatment simulation software. The
primary carbide kinetics can be determined through fitting experi-
mental data to model parameters that include diffusivity of carbon
in austenite and the formation and dissociation of primary carbides.
In this case, LPC experiments were conducted on Ferrium C64
cylinders, and parameters inside the DANTE carburization model
were fit from the experimental data. The model was then used
to redesign an LPC schedule for a gear component. It was shown,
through modeling, that the original schedule was producing too
many primary carbides and also had a surface carbon that was too
Figure 5: Carbon in the austenite matrix and carbon in primary carbide form for high. A schedule was then successfully redesigned using DANTE in
the redesigned process on the surface of the flank for the total process time.
which the surface carbon was reduced to 0.55% and the primary
carbides reduced to a small amount in a shallow surface layer. The
same procedure of conducting LPC experiments and fitting the data
to a model that includes carbon diffusivity through austenite and
the formation/dissociation of primary carbides can be applied to
any medium- or high-alloy steel.
thermalprocessing.com 29
Thermal Processing in Motion 2018—Conference Proceedings of the Thermal Processing in Motion Copyright © 2018 ASM International®
June 5-7, 2018, Spartanburg, South Carolina, USA All rights reserved
L. Ferguson, L. Frame, R. Goldstein, D. Guisbert, D. Scott MacKenzie, editors www.asminternational.org
263
potential that is too high will result in massive carbides in the D(T,C)=0.47*exp[-1.6*C]*exp[-(37000-6600*C)/R/T] (2)
final product that will degrade performance.
D(T,C)=(0.04+0.08*C)*exp[-31350/R/T] (3)
From Table 1, steel alloys that are gas carburized have levels
of strong carbide formers, other than iron, that total less than 1 D(T,M,C)=(0.146-0.036*C*(1-1.075*Cr)+Σk1*M)*
weight percent. With typical carburizing temperatures of 900 exp[-(144.3-15.0*C+0.37*C2+Σk2*M)/R/T] (4)
C to 950 C (1650 F to 1750F), gas atmosphere carbon
potentials typically range from 0.8% to 1.3%, with the lower where T is temperature, °K,
values being more common. As the alloy content increases, as C is weight percentage of carbon,
for 9310 steel, the carbon potential may be adjusted from high k1 and k2 are multiplying factors for specific elements,
at the beginning of the process, to a lower value for the M is weight percentage of Mn, Si, Ni, Cr, Mo, or Al,
remainder of the process to help guard against carbide R is the universal gas constant, 1.986 cal/mol/°K, and
formation. Steps like preoxidation may be added to the RkJ is the gas constant expressed as kJ/mol/°K.
process to further reduce the potential for alloy carbide
formation and allow greater carbon diffusion into the base These equations are valid within particular chemistry ranges,
metal. as discussed in references [3] for eq.(2), [4] for eq.(3), and [5]
for eq.(4). A graphical comparison of these equations is
Table 1: Carburizing Steels Showing Base Carbon Level and shown in Figure 2, where it is obvious that diffusion
Amount of Strong Carbide Forming Elements. coefficient increases markedly with temperature, and also with
carbon content. What is not as clear is the effect that alloying
Alloy %C %Cr %Mo %V Other elements, especially chromium, have on carbon diffusion.
Typical AISI 1020 0.2 - - -
Carburiz- AISI 4120 0.2 0.5 0.17 -
ing Steel AISI 4320 0.2 0.5 0.25 -
Grades AISI 5120 0.2 0.8 - -
AISI 8620 0.2 0.5 0.2 -
Historic AISI 9310 0.1 1.2 0.12 -
Aerospace
Gear Steel
High Pyrowear 53 0.1 1.0 3.0 2.0 Mn,
Strength Si, Cu
Carburizi Pyrowear 0.07 13.0 1.8 0.6 Co
ng Steels 675
Ferrium C64 0.11 3.5 1.75 - Co,
W
CSS-422L 0.12 14.0 4.75 0.6 Co
M50NiL 0.12 4.5 4.0 1.2 Ni
Vasco X2-M 0.15 5.0 1.4 0.5 Mn,
Si
Figure 2: Calculated diffusion coefficient as a function of
The equations used to predict the development of the carbon percentage at three temperatures, 800 C, 950 C and
carburized case in terms of time, temperature, and atmosphere 1100 C. ‘T’is for eq.(2), ‘W&M’ is for eq.(3), and ‘CSM’ is
carbon potential are quite simple, as differences between for eq. (4).
boundary diffusion and bulk diffusion are not considered and
carbon potential is held below austenite saturation level. Low Pressure Carburization: Low pressure carburization,
Because the gas carburizing times are long, i.e. several hours, as the name implies, is conducted in vacuum furnaces at
the full carbon potential is immediately applied to the surface pressures of about 0.1 to 1 torr. During heating to the
of the part. The well known Harris equation can be used to carburizing temperature, a nonreactive gas such as nitrogen
calculate total case depth for gas carburization[2]: may be added to provide convection to speed heating through
Case depth = f * √t (1) the lower temperatures where radiation is less effective. As
where t is time is hours and f is a temperature dependent factor the temperature increases, the parts must be protected from
that is related to diffusion. Notice that there is no relation to surface oxidation, so the vacuum must be below 0.1 to 0.3 torr.
different modes of diffusion (boundary or bulk), and there is A partial pressure of a “surface cleaning” gas such as hydrogen
no provision for alloy effects. Yet, this equation works fairly may be added during this stage to make the steel alloy more
well for the commonly carburized grades of carbon and low receptive to carbon absorption. Once the parts have been
alloy steel. heated to the desired temperature, carburization commences as
a series of boost and diffuse steps. For reasons discussed
Many equations for carbon diffusion coefficient in iron and below, the boost steps are short, typically less than two
low alloy steels have been published. Three equations for minutes, and the diffuse steps may be initially similar time, but
diffusion coefficient D expressed as cm2/s are: become progressively longer as the case develops.
264
Acetylene (C2H2) is commonly used as the carbon source gas.
During a boost step, the hydrocarbon gas may be blended with
another gas such as nitrogen or hydrogen to assist mixing of
the hydrocarbon gas in the vacuum chamber. The boost pulse
time, gas flows and chamber pressure for two similar furnaces
but of different sizes are shown in Figure 3 a and b. Due to
size differences, the effective gas flows and chamber pressure
are different for a boost step with the same common aim.
266
Figure 9 shows results for a computer simulation that had nine
boost and diffuse steps for a LPC run. There are two curves
each for two locations, position 501 on the surface and
position 492 at a depth of 0.25 mm below the surface. The
solid lines are carbon weight fraction predictions, and the
dashed lines are normalized carbide size predictions. This
boost-diffuse schedule is predicted to produce carbides at the
surface that form during the first boost and persist through the
entire LPC process. The subsurface location is predicted to
form carbides during the last long diffuse step, but these
dissolve. From the plot, carbide formation is predicted to
occur when the carbon weight fraction hits 0.010. To prevent
any carbide formation, the carbon weight fraction should
remain below this level for the steel properties used in this
Figure 7: Hardness and carbon profiles measured for a LPC model. The emphasis on this last phrase is important as the
trial. plot in Figure 8 is not for C64, but it is for a competing high
strength steel alloy.
of being high at the surface and lower with depth from the
surface, it has a lot of variation and is more difficult to
interpret. Complicating the issue is the possibility of carbide
formation and interference with pure diffusion of carbon
through the austenite lattice during carburization.
Acknowledgments
The authors wish to acknowledge the support of the US Army
ADD for their support of this work and Solar Atmospheres,
Figure 8: Framework for LPC simulation software. Inc. for low pressure carburizing the test bars.
267
References
268
Heat Treat 2021: Proceedings of the 31st ASM Heat Treating Society Conference Copyright © 2021 ASM International®
September 14–16, 2021 All rights reserved.
America’s Center, St. Louis, Missouri, USA www.asminternational.org
DOI: 10.31399/asm.cp.ht2021p0263
Jason Fetty
Combat Capabilities Development Command Aviation & Missile Center, Fort Eustis, Virginia, U.S.A.
Abstract
Traditionally, liquid quenchants were used to quench most steel
A gas quenching method was developed by DANTE Solutions, parts to obtain a martensitic microstructure. Liquid quenchants
in conjunction with the U.S. Army Combat Capabilities undergo a unique phenomenon, comprised of three stages,
Development Command Aviation & Missile Center when a red-hot part is immersed into the liquid [1]. First, a thin
(DEVCOM AvMC), to control distortion in difficult to quench vapor film is formed around the red-hot part, with extremely
geometries. This new method addresses the nonuniform cooling slow heat transfer rates. Nucleate boiling commences as the
inherent in most gas quenching processes. A prototype unit was vapor blanket breaks down. Nucleate boiling results in the
constructed and tested with the aim of controlling the martensite fastest heat transfer due to a combination of the latent heat of
formation rate uniformity in the component being quenched. vaporization and aggressive convection. Convective cooling,
With the ability of the DANTE Controlled Gas Quenching the final stage, begins as the nucleate boiling subsides [2].
(DCGQ) unit to control the temperature of the quench gas
entering the quench chamber, thermal and phase transformation The continually changing heat transfer rates associated with
gradients are significantly reduced. This reduction in gradients liquid quenching can severely affect the cooling uniformity of
yields a more uniform phase transformation, resulting in a given part. First, the breakdown of the vapor blanket rarely
reduced and predictable distortion. Being able to minimize and occurs evenly on all part surfaces, being dependent on the part
predict distortion during gas quenching, post heat treatment surface temperature, local flow behaviour, and the liquid
finishing operations can be reduced or eliminated, and as such, properties, creating brief periods of nonuniform heat transfer.
fatigue performance can be improved. This paper will discuss The chaotic nature of this phenomenon is difficult to predict and
the prototype unit performance. Mechanical testing and can lead to inconsistent distortion within a single load of parts.
metallographic analysis were also performed on Ferrium C64 Part geometry and immersion orientation also play a significant
alloy steel coupons and will be discussed. The results obtained role in nonuniform cooling when quenching in liquids [3 – 5].
showed that the slower cooling rate provided by the prototype
did not alter the microstructure, hardness, strength, ductility, High pressure gas quenching (HPGQ) does not involve a phase
toughness, or residual stress of the alloy. change of the quenching media, and therefore, has a more stable
heat transfer rate. However, due to its low density and specific
heat, gas is unable to absorb energy as well as liquids, and will
Introduction suffer a temperature change as heat is removed from the part.
Gases’ low density also make it more susceptible to local flow
variations. HPGQ equipment can also significantly contribute
Distortion can generally be divided into two categories: size to local flow variations [6].
change and shape change. Solid-state phase changes occurring
in steel alloys during thermal processing can result in In response to large distortion during HPGQ of complex
permanent size change, due to the difference between the geometries, DANTE Solutions devised a novel process, termed
starting microstructure and the microstructure after heat DANTE Controlled Gas Quenching (DCGQ), by which the
treatment. Size change is unavoidable but can easily be martensitic phase transformation is controlled during gas
predicted and accounted for in part design. Permanent shape quenching [7]. Since the transformation from austenite to
change is a result of nonuniform plastic strain, caused by martensite is driven by a reduction in temperature, and is not
nonuniform phase transformations, thermal strains, or creep time dependent like the diffusive phase transformations, the
while at high temperature, and is more difficult to predict and simplest way to control the martensite transformation is to
control. The nonuniformities can be a result of alloy control the rate of temperature change within the component.
segregation, uneven heating or quenching, poor support while By controlling the uniformity of martensitic transformation
at high temperature, thermal expansion or contraction throughout the part, distortion can be significantly reduced,
restrictions, or residual stresses from prior manufacturing easily predicted, and consistently reproduced. This paper will
operations. examine the DCGQ prototype unit design and operation.
Material property testing was also conducted, and discussed in
263
264
Equipment Operation
The unit was constructed at Atmosphere Engineering and
shipped to Akron Steel Treating in Akron, OH, where it was
installed and tested. This first prototype unit requires heating to
be performed in a separate furnace, with the part transferred by
rail cart to the DCGQ unit after austenization is complete; future
production equipment should integrate controlled heating, as
well as cooling. The DCGQ unit is preheated to a predefined
temperature during the austenitization process in preparation
for quench; the preheat temperature is alloy and part geometry
dependent. The DCGQ recipe programmed into the HMI begins
once the front-loading door is closed. Figure 4 shows a
comparison of the temperature of the quench gas entering the
chamber (“Chamber Inlet PV”) and the recipe setpoint
temperature (“Chamber Inlet SP”) with no hot part in the
chamber. The recipe consists of 50° C temperature reductions
Figure 5: DCGQ Prototype unit temperature comparison
over two minutes, starting at 425° C, with twenty-minute holds
between quench gas entering the quench chamber and the
at each temperature step.
recipe setpoint temperature, with an empty chamber.
To reduce recipe design time, and ensure an optimal recipe is For DCGQ processes, the ambient temperature as a function of
achieved, with respect to distortion minimization and time is dependent on total mass, surface area, and initial
processing time, the equipment must be thermally characterized temperature, just as it is with HPGQ, but the ambient
so heat treatment simulation and design software, such as temperature for DCGQ is also dependent on the temperature
DANTE, can be used for DCGQ recipe design. The following ramp and hold times. Figure 8 shows the results of a
section describes the characterization of the DCGQ prototype comparative analysis between the setpoint and actual
unit. temperatures at the ends of the holds (left) and ramps (right) for
various DCGQ schedules. Ideally, to make design and analysis
Equipment Characterization of the DCGQ process simpler and more intuitive, the setpoint
To properly characterize the DCGQ prototype unit for and actual temperatures should be equal at the end of each
modeling and process design, or any other type of thermal temperature ramp and hold. For the processes shown in Figure
equipment, it is necessary to determine the heat transfer 8, HR4 and HR7 abide to this equality, but as the ramp and hold
coefficients (HTCs) and ambient temperatures acting on the times get shorter, these two temperatures begin to deviate; HR5
component being treated. For liquid quenching operations, the had the shortest ramp and hold times. This fact does not negate
convective HTC should be described in terms of part surface the use of modeling in recipe design, but rather solidifies the
temperature, due to the significant difference in heat transfer fact that processing equipment behavior should be thoroughly
between the vaporization, nucleate boiling, and convective understood, within normal operating conditions, if the process
stages of liquid quenching. The ambient temperature is assumed is to be modeled. Once the behavior is understood, it can be
to remain constant, due to liquids’ large specific heat and incorporated into the models for more accurate results.
overall volume in the quench vessel.
Figure 7(A) shows a cylindrical quench probe, made of AISI Figure 8: Relationship between recipe setpoint and ambient
304 stainless steel, used to characterize the DCGQ equipment. temperatures at the end of the Holds (left) and Ramps (right)
The cylinder has a 100 mm diameter and 100 mm height. There used for DCGQ recipe design and analysis.
are five holes drilled to mid-height, four approximately 3 mm
from the outer diameter and one in the center, which are fitted Figures 9 and 10 show time-temperature results comparing the
with K-type thermocouples. Thermocouples are also located in DANTE simulation, which used the single HTC and relevant
the DCGQ chamber at various distances from the probe to ambient temperatures determined from experiments, and the
measure the ambient temperature. Using the time-temperature actual time-temperature data. In addition to the two tests shown
history of the quench probe and recorded ambient temperature in this paper, five more recipes were executed, with their data
for a given DCGQ recipe, along with a DANTE model of the also used in characterizing the DCGQ equipment. Figures 9(A)
quench probe, shown in Figure 7(B), the HTC and ambient and 10(A), 1000° C to 400° C, and Figures 9(C) and 10(C), 400°
temperatures can be determined. C to 100° C, show the average temperature history for the near
surface thermocouples during the HR4 and HR5 tests and
DANTE simulation of these two tests. Figures 9(B) and 10(B),
1000° C to 400° C, and Figures 9(D) and 10(D), 400° C to 100°
266
Microstructural Evaluation
Previous experiments had been conducted on base carbon
Ferrium C64 prior to the design and construction of the DCGQ
Figure 9: Comparison of experimental and simulation results prototype unit to evaluate a slow transformation rate on C64’s
for one DCGQ recipe. microstructure. Those preliminary results showed no noticeable
difference between slowly and rapidly transformed martensitic
microstructures. Figure 11 shows the microstructure, magnified
1000X, of a carburized (A) DCGQ processed coupon and (B)
HPGQ processed coupon. As with previous experiments, there
is no discernable difference between the two microstructures.
The thin grey film on the DCGQ coupon is oxidized copper,
which was removed prior to testing. The microhardness profiles
for the coupons are shown in Figure 12 and confirm that the
slow transformation rate did not alter the microstructure of the
DCGQ processed coupons, compared to coupons processed
using a standard 2-bar HPGQ process.
Ferrium C64 was chosen as the candidate alloy, due to its high
hardenability, use in high stress powertrain applications, and
267
268
270
➢ DANTE is a commercial finite element based software tool for heat treatment
simulation validated for industrial processes
• Process design
• Process trouble shooting
• Design of experiments
➢ State-of-the-art mechanical and phase transformation kinetics models for use
with ANSYS Mechanical
➢ Heat Treatment Process Capabilities for Modeling:
• Carburizing; gas, low pressure • High pressure gas quench • Induction hardening
• Nitriding; gas, low pressure • Air cooling • Cryogenic treatments
• Heating; furnace, flame, laser, body • Spray quenching • Tempering; low and high
flux (induction/direct current) • Intensive quenching temperature
• Immersion liquid quenching; oil, • Press / fixture quenching • Normalizing
water, polymer
➢ Carburization model, thermal model, and stress ➢ Carburization model is sequentially coupled to
model are sequentially coupled. the fully coupled thermal and stress model
Dilatometry
350
Boiling
Phase 300
Gas is heated
by hot part
250
Temperature, C
Gas is cooled by
200
heat exchanger
150
Initial gas
100
temperature
50
0
0 20 40 60 80 100
time, s
Example of ambient gas quench
Example of liquid quenchant HTC as a function of temperature
temperature as a function of time
Copyright © 2021 DANTE Solutions. All rights reserved. 15
Methods for Heat Transfer Characterization
Roll Body Roll Body 5”
Journal Surface Depth
Surface
Equiv. Test
Position
Core
➢ For accurate HTC determination, thermocouples are attached to record heating and cooling steps
• In most cases, embedded thermocouples are preferred
• Finite Element model is developed to describe the thermocouple response effectively
• Thermocouples can be embedded in actual parts or simple cylinders
TC1
TC2
TC3
➢ Low mass of nitrogen gas vs. liquid means gas will heat up, but liquid heat up should be insignificant
• Tamb changes during gas quench, but can be ignored in most cases of liquid quenching
➢ With gas, no concern about boiling phenomena as with liquid
• HTC is more related to gas velocity than to surface temperature of the part
➢ Uniformity of gas flow pattern is more critical than that of liquid flow pattern on the part cooling
➢ Considerations of HPGQ process:
• Efficiency of heat exchanger
• Change in Tgas as it circulates through racked parts
• Uniformity of local gas velocity
Grain size:
0.0635mm
Ferrite
Transformation
20CrMnXX
Carbide Coarsening
This difference is
significant!
-160 µm
-900 µm
➢ DANTE modeling can be used to optimize crankshaft geometry to be more suitable to heat treat process, obtain
beneficial residual stresses, avoid fracture, and improve fatigue performance
Copyright © 2021 DANTE Solutions. All rights reserved. 36
Modeling Example
Press Quenching of AISI 9310 Carburized Bevel Gear
Model and Process Description
Model Description:
➢ Bevel gear made of AISI 9310
➢ OD of 420 mm and ID of 140 mm
➢ Bevel teeth removed since bore distortion was main concern
• Internal spline was modeled
➢ Single tooth sector was modeled
• Assumes that cooling is uniform in the circumferential direction
270° section of bevel gear with bevel teeth removed
Process Description:
➢ Part was carburized on all surfaces to a case depth of 1.3 mm
➢ Austenitized at 850° C for 3 hrs.
➢ Transferred to press quench unit and quenched for 10 min.
➢ Cooled to room temperature
➢ Tempering was not modeled for this study
• Tempering usually has a negligible effect on distortion
• DANTE can model the tempering process
Single tooth model with bevel tooth removed
Copyright © 2021 DANTE Solutions. All rights reserved. 39
Schematic Showing Press Quench Tooling
➢ All dies and expanders are
modeled as rigid surfaces
➢ Thermal interaction between
dies/expanders and part is ignored
➢ All die sizes and locations relative
to the part are maintained in the
model
Axial Disp.
Copyright © 2021 DANTE Solutions. All rights reserved. 42
Carburization Process Modeling
Carbide Formation during Low Pressure Carburization
Carbon and Carbide Evolution during LPC Process
of a High Alloy Steel
➢ Main challenges:
• Controlling the amount and size of primary carbides while
achieving the specified surface carbon and case depth
• Reduce furnace time
Example of Multiple Chamber HPGQ Furnace Example of Single Chamber HPGQ Furnace
With Residual
Stress
Copyright © 2021 DANTE Solutions. All rights reserved. Before Loading After Loading 59
DANTE Utility: GCarb
➢ Gas carburization scheduler
➢ Can be used to design recipes
• Predict total time of last step, given the inputs
➢ Can be used to predict predefined recipes
➢ Predict carbon profile in terms of depth
➢ Predict as-quenched martensite hardness in
terms of depth
➢ Can include multiple carburization steps
➢ Material inputs can include:
• Steel grade
• Base carbon
• Alloy composition
➢ Process inputs can include:
• Time and temperature
• Carbon potential
• Effective case depth and carbon value at
effective case depth
Copyright © 2021 DANTE Solutions. All rights reserved. 61
DANTE Utility: VCarb
➢ Low pressure carburization scheduler
➢ Can be used to design recipes
• Determine entire boost/diffuse schedule
➢ Can be used to predict predefined recipes
➢ Predict carbon profile in terms of depth
➢ Predict as-quenched martensite hardness in
terms of depth
➢ Material inputs can include:
• Steel grade
• Base carbon
• Alloy composition
➢ Process inputs can include:
• Time and temperature
• Final surface carbon
• Effective case depth and carbon value at
effective case depth
TTT_Fit
➢ Used to fit phase transformation behavior from TTT diagrams
• Good to predict the timing of the transformations from austenite to ferrite, pearlite, and/or bainite
• No information on the martensite transformation behavior; start and end times are generally provided
• No information on transformation strains
HTC_Fit
➢ Must be used in combination with an Abaqus-DANTE model; not available for use with Ansys
➢ Used to determine the heat transfer coefficient in terms of temperature or time