You are on page 1of 13

International Journal of Fatigue 26 (2004) 10951107

www.elsevier.com/locate/ijfatigue

Simulation and evaluation of thermal fatigue cracking


of hot work tool steels
Anders Persson a,, Sture Hogmark b, Jens Bergstrom a
a

Department of Materials Engineering, Karlstad University, SE-651 88 Karlstad, Sweden


b
The Gngstrom Laboratory, Uppsala University, SE-751 21 Uppsala, Sweden

Received 30 June 2003; received in revised form 2 February 2004; accepted 3 March 2004

Abstract
Die casting is a very cost ecient method to manufacture near net-shaped and complex cast products. One limitation for further cost reduction is fatigue cracking of the tool due to thermal cycling, which is observed as a crack network on the tool surface. Hot work tool steels are commonly used as die material.
In this study, an experimental test machine for simulation of thermal fatigue is described. The test is based on cyclic induction
heating and internal cooling of hollow cylindrical test rods. The surface strain is continuously recorded during the thermal cycling
through a non-contact laser speckle technique. The applicability of the test is demonstrated on two hot work tool steel grades,
v
v
hardened and tempered to dierent conditions, and heat cycled between Tmin 170 C and Tmax 600850 C.
It is shown that the test method can simulate surface cracking of tools exposed to thermal fatigue. The surface strain recordings
proved to give sucient information to successfully deduce the strains and stresses behind the mechanism of thermal fatigue surface cracking, without knowledge of the temperature distribution below the surface. It was also found that low-cycle fatigue
v
occurs for the tests with Tmax 600 and 700 C, although the estimated tensile stress after cooling does not exceed the initial yield
strength of the steel. Most probably, the reason is the gradual softening of the tool steels during the thermal cycling. Additionally, the presence of stress concentrators play a critical role during these conditions.
# 2004 Elsevier Ltd. All rights reserved.
Keywords: Thermal fatigue; Heat checking; Surface strain; Hot work tool steel; Die casting

1. Introduction
Die casting is a very cost-ecient method of forming
near net-shaped cast products of, for example, aluminium, zinc, magnesium, and copper based alloys of
almost any shape [13]. Prior to casting aluminium and
copper alloys, the die is normally preheated to a temv
perature within the range of 250300 and 300350 C,
respectively, to reduce the thermal shock, and the average tool temperature is usually kept at those levels
through internal cooling. During a casting cycle, molten metal is forced into the mould by the application of
pressure, the peak of which can exceed 70 MPa. A distinguishing characteristic of the process is that the

Corresponding author. Tel.: +46-54-700-18-21; fax: +46-54-70014-49.
E-mail address: anders.persson@kau.se (A. Persson).

0142-1123/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2004.03.005

liquid metal ows with high velocity during injection


and provides rapid lling of the die cavity, typically
within milliseconds. For aluminium alloys, the entrance
velocity of the melt is usually 2060 m/s and the melt
v
temperature is approximately 700 C, whereas those for
v
copper alloys is about 110 m/s and 970 C. The high
melt velocity is necessary to completely ll the mould
of thin-walled and complex shaped products. Continuous internal cooling of the tool during the process
makes the solidication of the casting ecient, and
high rate manufacturing of typically 100 castings per
hour is possible. When the casting has solidied, the die
is opened and the casting ejected. Thereafter, the die
may be externally cooled and lubricated by spraying.
Thermal fatigue cracking, gross fracture, erosion,
corrosion and local adherence of the casting alloy (soldering) are important life-limiting tool failure mechanisms in aluminium and brass die casting [13]. Thermal

1096

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

fatigue cracking results from rapid alternations in temperature of the die surface during the casting process.
The temperature cycling may induce stresses high
enough to impose an increment of plastic strain in the
tool surface during each casting cycle. Surface cracks
develop generally within a few thousand cycles, or even
earlier, and are, consequently, formed in the low-cycle
fatigue range (<103104 cycles) [4,5]. Oxidation and
creep may contribute signicantly to cracking [48].
Thermal fatigue damage is often observed as a network
of ne cracks on the tool surface, and the cracks
usually penetrate only a limited surface layer. This type
of crack pattern is often named heat checking. The
crack network degrades the surface quality of the tool
and, consequently, the surface nish of the cast products. It may, ultimately, increase the production costs
through expensive maintenance, die failures, and rejection of castings.
High levels of hot strength (hot hardness), tempering
resistance, creep strength, and ductility, along with low
thermal expansion and high thermal conductivity are
some essential properties for high resistance to thermal
fatigue cracking [2]. Hot work tool steels oer several
good candidates for die components. Steel grades such
as AISI H11, H13, H20, H21, or H22 are frequently
used as die material.
Field tests of die casting tools are very expensive and
time consuming. Furthermore, they do not enable the
possibility to perform isolated studies on thermal fatigue cracking without any inuence from other failure
mechanisms. Thus, there is a need for a simplied
method to isolate and evaluate fatigue cracking of
materials exposed to thermal cycling. Since this cracking is a low-cycle fatigue process, the surface strain
response is of particular interest. To the authors best
knowledge, no laboratory test for simulation of thermal fatigue cracking, where the surface strain response
during heat cycling can be recorded, has been reported
in the literature. Therefore, the aim of this study was to
develop a simplied test method to imitate and evaluate thermal fatigue by deducing the surface strains
responsible for failure. A comparison between the two
hot work tool steel grades, quenched and tempered to
various conditions, were also made.

2. Experimental
2.1. Test equipment
The test equipment is based on cyclic induction heating and internal cooling of test rods. Surface strain
measurements through a non-contact laser speckle
technique, makes it possible to calculate the strains
induced in the specimen surface during thermal cycling
(see Fig. 1). The test specimens are hollow cylinders

Fig. 1. Schematic of the experimental set-up used in the thermal


fatigue test.

with a diameter of 10 mm, a length of 80 mm and have


a 3 mm axial hole for internal cooling. An induction
unit (25 kW, 3 MHz) heats approximately 20 mm of
the middle of the test rod. Continuous cooling is performed by circulating silicon oil or water through the
specimen, and also externally with argon or air. A thermal cycle includes generally a steep ramp to the
maximum temperature, followed by a hold time, and
subsequent cooling to the minimum temperature.
Consequently, the thermal conditions and test environment can be varied to simulate specied conditions.
The surface temperature of the specimen is monitored
by a pyrometer, but also measured by a thermocouple,
which is spot welded to the surface. The former
method is used for temperature control during heating,
and the latter to obtain recordings of the surface temperature during thermal cycling, which later is used in
surface strain estimations. The specimen surface represents the tool surface and the generated heat cycle
simulates the temperature cycle during die casting.
A surface area illuminated by a laser beam displays a
characteristic pattern of small visually observed dots.
This speckled pattern can be used to estimate the surface strains during thermal cycling [912]. In this study,
a 17 mW HeNe laser is used to illuminate a spot on the
diusely reecting specimen surface. The spot size was
about 1.5 mm and the laser beam hits the specimen surface at right angle. Two pairs of symmetrically and perpendicularly arranged CCD-array sensors (Fig. 1) are
used to record the movement of the speckle pattern
during thermal cycling, and from this movement the
axial and tangential surface strain E can be approximately calculated at any time using Eq. (1) [11]:
e

DA
2L0 tan h0

DA is the dierence between the speckle displacements


detected by each pair of sensors, L0 ( 0:17 m) is the
distance from the specimen surface to the CCDv
sensors, and h0 ( 45 ) is the direction between the incident laser beam and each CCD-sensor. The vertical pair

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

of sensors give the axial surface strain and the horizontal pair the tangential. The speckle displacements are
acquired with a sampling rate of 500 Hz.
2.2. Materials
Two Uddeholm hot work tool steels, QRO 90
Supreme and Hotvar, were used as test materials. QRO
90 has a nominal chemical composition (wt.%) of 0.38
C, 0.30 Si, 0.75 Mn, 2.6 Cr, 2.25 Mo, 0.9 V, and Fe
balance. Hotvar has a nominal chemical composition
(wt.%) of 0.55 C, 1.0 Si, 0.75 Mn, 2.6 Cr, 2.25 Mo,
0.85 V, and Fe balance. For the QRO 90 specimens,
two variants of heat treatment were used: hardening
v
(austenitizing 30 min at 1030 C, followed by air
v
quenching) and tempering 2  2 h at 640 C to a hardness of 430  10 HV30 and hardening as above folv
lowed by tempering 2  2 h at 625 C to a hardness of
510  10 HV30, respectively. The Hotvar specimens
v
were hardened (austenitizing 30 min at 1050 C, followed by air quenching) and tempered 2  2 h at 575
v
C to a hardness of 640  10 HV30. All heat treatments
resulted in microstructures of tempered martensite. The
hardness of the steels was assessed by Vickers indentations on polished cross-sections, using a load of 30 kg.
For all materials, the austenitizing treatment gives a
nominal austenite grain size of about ASTM 9, and no
signicant microstructural dierences between the
materials could be detected. The heat treatment of the
specimens was followed by grinding to a surface roughness (Ra) of 0:38  0:05 lm, as obtained using optical
surface prolometry.
Relevant mechanical properties of the two tool steels
are given in Fig. 2. The true yield strength values
(Fig. 2f) are obtained from visual inspections of tensile
curves such as that of Fig. 2e, and are dened as the
stresses when the stressstrain curves deviate from a
linear relation. Note that the true yield stress values are
well below Rp0.2 in Fig. 2b,e.
2.3. Thermal fatigue testing
Three temperature cycles were used to simulate various die casting conditions (see Table 1), and designated
according to their maximum temperature. The maximum temperatures for the three temperature cycles
v
were set to 600, 700, and 850 C, respectively, to
include the thermal conditions for aluminium and brass
die casting, respectively. The latter temperature corresponds to the maximum tool surface temperature during actual brass die casting [13]. The temperature cycles
included a steep ramp to the maximum temperature,
no holding time at maximum temperature, and subsequent cooling to the minimum temperature. Continuous cooling was performed by circulating silicon oil
v
(ow rate  2.5 l/min) of 60 C through the specimen,

1097

and also externally with either argon (forced convection) or air (natural convection).
Argon was used as cooling medium because the tools
are exposed to an environment with reduced oxygen
content during actual die casting. The oxygen in the die
cavity is partly consumed through oxidation of tool
material and casting alloy. For comparison, some tests
were performed in air.
Prior to testing, the specimens were pre-oxidised to
obtain a thin oxide layer, which facilitates the pyrometer temperature control during heating. This was
made by electrochemical oxidation in a NaOH-solution
(containing deionized water and 300 g/l NaOH) at
v
70 C for about 5 min, followed by 1 h heat treatment
v
at 200 C in air. A K-type ChromelAlumel thermocouple with thin wires (1 0.13 mm) was spot welded to
the specimen to measure the surface temperature during testing. The thin wires enable rapid response of any
change in temperature. Finally, to obtain a good
speckle pattern for the surface strain measurements, an
area of approximately 10  10 mm located in the middle of the specimens was roughened by a 1000 mesh
abrasive paper.
2.4. Evaluation techniques
During the heat cycling, the surface strain is continuously obtained by the laser speckle technique from the
change in the specimen dimensions, and represented as
surface strain vs. temperature. Any thermal fatigue
damage of the specimen surface was revealed using
scanning electron microscopy (SEM). It was further
characterised with respect to crack growth (crack
length vs. number of cycles) and crack density (number
of cracks per unit of length along the specimen surface)
by measurements on polished axial cross-sections performed in light optical microscopy (LOM). For each
specimen, all evaluation of cracks is based on cracks
longer than about 5 lm, detected along two lines, each
of 8 mm length.
Proles of surface hardness and hardness vs. depth
after exposure to the heat cycling were assessed by
Vickers indentations on polished specimen surfaces and
cross-sections, respectively, using a load of 25 g.
Finally, the residual stress state in the surface layer
was measured by X-ray diraction (XRD) using CuKa
radiation and the sin2 w method.
2.5. Tests for verication of the strain measurements
To investigate if the surface strains obtained by the
laser speckle arrangement correlate with the actual surface deformation, a cylindrical pressure vessel of aluminium (diameter 100 mm, wall thickness 2 mm, and
length 80 mm) was selected. The vessel was designed
and positioned in such a way that its external surface

1098

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

Fig. 2. Nominal mechanical properties of Hotvar and QRO 90 hardened as above and tempered to various conditions. (a) Temper resistance as
v
hardness at room temperature vs. holding time for temperatures within the range of 550650 C. (b) Hot yield strength as Rp0.2 vs. temperature.
(c) Modulus of elasticity vs. temperature. (d) Hot ductility as reduction of area at tensile fracture vs. temperature. (e) Schematic of sample tensile
curve. (f) Estimated true yield strength ry vs. temperature. The solid lines are the second order polynomial t to the data.

was kept in the same position as that of an ordinary


thermal fatigue specimen. To achieve a good speckle
pattern for the speckle measurements, the surface were
roughened as above. Consequently, the speckle movements during deformation of the pressure vessel were

detected by the CCD-sensors under equal optical conditions as those during heat cycling of the thermal fatigue specimens. A strain gauge was glued to the surface,
next to the spot illuminated by the laser beam, to
obtain reference values of the surface strain.

Table 1
Thermal cycles used in the thermal fatigue tests
v

Max. temperature [ C]

Min. temperature [ C]

Heating time [s]

Total cycle time [s]

External cooling

600
700
850

170
170
170

0.2
0.3
2.2

11.2
14.3
26.2

Argon or air
Argon or air
Argon or air

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

1099

The design of the pressure vessel enabled deformation under well-controlled conditions and simultaneous strain measurements by the laser speckle
technique and the strain gauge, respectively. For the
two methods, the tangential surface strain (hoop strain)
after loading or unloading of the pressure vessel was
used for the evaluation, since the induced deformation
is signicantly larger in the tangential than in the axial
direction. The results from the two techniques were
compared.

3. Results
3.1. Verication of the laser speckle technique
By pressuring the aluminium vessel, it proved possible to verify that the surface strains obtained by the
laser speckle technique correlates very well to the reference values measured by the strain gauge (see Fig. 3).
It is concluded that the laser speckle technique makes it
possible to detect surface strains with a resolution of
about 25  10-6 . Because of security reasons, it was
necessary to limit the maximum pressure exposure of
the pressure vessel, which as a result, restricted the
maximum and minimum surface strains of this verication to about 160  106 .
3.2. Recorded surface temperatures
Induction heating very rapidly increases the surface
temperature, as seen in Fig. 4a. After the short heating
time, the surface will cool down following a more gradual slope. The three heat cycles tested all start with a
v
specimen temperature of 60 C (= oil temperature). It
takes about four cycles to obtain equilibrium temperature conditions (see Fig. 4b and Table 1). No signi-

Fig. 3. Comparison of tangential surface strain (hoop strain)


obtained by the laser speckle technique and the resistance strain
gauge, respectively, after pressuring or depressurising the aluminium
vessel. Each circle represents the result from one loading (positive
values) or unloading (negative values) event. The solid line is the linear t to the experimental data.

Fig. 4. Typical surface temperature recording for the 700 C test at


equilibrium (a), and of maximum temperature Tmax and minimum
temperature Tmin during the rst 10 cycles extracted from the temperature recordings (b).

cant dierences could be seen among the dierent


steels.
3.3. Surface strain during thermal cycling
Typically, the surface strain increases with the surface temperature, followed by an almost constant or a
slightly increasing strain level during the rst steep part
of the cooling phase, whereafter it decreases with temperature (see Fig. 5). As expected, the surface strain
level was strongly dependent on the maximum temperature during each thermal cycle (see Fig. 5). During
v
heating to 850 C, there is a sharp increase in the surv
face strain curve at about 760 C to a slope that is signicantly higher as compared to that at any lower
temperature (see Fig. 5c). After the cooling event and
at equilibrium temperature conditions, each surface
strain recording forms either a closed loop with practically no residual surface strain at the minimum
v
temperature (see Fig. 5a (Tmax 600 C) and b
v
(Tmax 700 C), or an open loop with tensile residual
surface strain at the minimum temperature (see Fig. 5c;
v
Tmax 850 C). Note also that the tangential strain e/

1100

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

Fig. 6. Overview of typical crack network after 10 000 thermal


v
cycles to 700 C in argon (SEM). The axial direction of the specimen
is horizontal.
v

to 700 C, the crack path was relatively straight


(Fig. 7a), whereas branched cracks (Fig. 7b) in
addition to the single cracks were observed after heat
v
cycling to 850 C.
The crack length was strongly dependent on the
number of cycles and the maximum temperature during each cycle (see Fig. 8). Since the crack propagation
v
rate was very high during the 700 and 850 C cycles, as
v
compared to the 600 C cycles, it was necessary to limit

Fig. 5. Typical surface strain recordings of QRO 90 at 510 HV30


during thermal cycling in air at equilibrium temperature conditions to
v
v
v
600 C (a), 700 C (b), and 850 C (c).

and axial strain ez are almost identical, and in the following only e/ is considered.
3.4. Surface cracking after thermal fatigue
3.4.1. Argon atmosphere
Thermal fatigue damage was typically observed as a
network of cracks on the specimen surface (see Fig. 6).
Polished cross-sections revealed that the appearance of
the cracks was strongly dependent on the maximum
temperature during each cycle. For thermal cycling up

Fig. 7. Polished cross-section revealing typical crack after thermal


v
cycling in argon (LOM). (a) 10 000 cycles to 700 C. (b) 1 000 cycles
v
to 850 C (QRO 90 at 510 HV30).

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

1101

Fig. 9. Length and density of thermal fatigue cracks after 1000


v
cycles to 700 C in air. One specimen of each material was tested.

about 5 lm were observed on Hotvar after 100 cycles


v
to 850 C.
3.4.2. Air atmosphere
The crack length and crack density after 100 cycles
v
v
to 850 C and 1000 cycles to 700 C, respectively, in
argon and air atmosphere were comparable.
v
Thermal cycling in air (1000 cycles to 700 C)
revealed that the crack length and crack density in general decrease with increasing hardness, see Fig. 9. For
all materials, no cracks larger than the evaluation criteria of about 5 lm were detected after 100 cycles to
v
850 C.
3.5. Hardness after thermal fatigue

Fig. 8. Length and density of thermal fatigue cracks after heat cycling in argon atmosphere. (a) Maximum crack length vs. number of
cycles. (b) Mean crack length vs. number of cycles. (c) Crack density
vs. number of cycles. Each pile is the mean value of three or four speciv
v
mens, except those for 100 cycles to 850 C and 1000 cycles to 700 C,
respectively, which are based on one specimen of each material. The
error bars indicate the maximum and minimum recording.

the number of cycles to 10 000 and 1000 cycles,


respectively, for these temperature levels. In a relatively
early stage, the density of cracks was also strongly
dependent on the number of cycles and the maximum
temperature level. Thereafter, the crack density averaged at almost constant levels. With few exceptions,
the crack length and crack density seem to have a tendency to decrease with increasing hardness. Note also
that no cracks exceeding the evaluation criteria of

As expected, the surface hardness decreased with the


number of thermal cycles for all materials and conditions, as exemplied by Fig. 10. It is seen that the
surface hardness is rapidly reduced initially, followed
by a more gradual loss of hardness with the number of
cycles. The hardness levels and the thickness of the softened surface layer were strongly dependent on the
number of heat cycles and the maximum temperature
during each cycle (see Fig. 11). Evidently, after
v
exposure to 600 C there is no notable eect on the
hardness levels (Fig. 11a), whereas a considerable
softening of the surface layer is detected after thermal
v
v
cycling to 700 C (Fig. 11b) and 850 C (Fig. 11c),
v
respectively. For the 700 C test, the hardness decreases gradually from a depth of about 0.5 mm towards
v
the surface (Fig. 11b). Thermal cycling at 850 C resulted in a dramatic hardness reduction throughout the
whole specimen (see Fig. 11c), which also reveals a
hardness maximum at about 0.3 mm depth for all
steels. Finally, the softening rate and the thickness of
the softened surface layer is larger for the Hotvar steel
than for the QRO 90 steel, so that the hardness levels
gradually evens for increasing number of heat exposures (see Figs. 10 and 11b,c).

1102

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

Fig. 10. Surface hardness vs. number of heat cycles to 700 C. (a)
For the rst 5000 cycles. (b) Close-up of the rst 100 cycles (note the
linear scale for the number of cycles).

3.6. Residual stresses


Residual tensile stresses are gradually building up in
the surface layer during the thermal cycling, whereas
the residual shear stress is virtually unaected (see
Fig. 12). It is seen that the residual tangential and axial
stresses increase rapidly from a compressive to a tensile
state, followed by a more gradual increase in tensile
stresses with the number of heat exposures. Note that
the tangential and axial stresses follow the same tendv
ency. For the 600 C tests, the tensile stress state is
reached after about 10 heat cycles, whereas that for
v
the 700 C experiments is obtained immediately after
the rst cycle. Generally, the levels of the tensile stresv
ses are higher after the 700 C tests than after those of
v
600 C. The maximum scatter in residual stress at all
number of cycles, except at the rst where it is highest
(about 6070%), is about 1040%.

Fig. 11. Hardness vs. depth after exposure to thermal cycling. (a) 20
v
v
000 cycles to 600 C. (b) 10 000 cycles to 700 C. (c) 1000 cycles to
v
850 C.

the mechanisms behind thermal fatigue. Please note


that this is possible just by recording the surface temperature and strain. Knowledge of the temperature
prole below the surface is not needed!
4.1. Characteristics of the induction heating

4. Discussion
The laser speckle technique was shown to give the
true surface strains (see Fig. 3 and Ref. [14]), and could
thus be used to obtain information necessary to verify

Induction heating of steel using a frequency of


3 MHz results in very fast heating of only a thin surface layer, of the order of 10 lm (skin-eect). When
applying this technique to the tool steel test rods, the
thermal expansion of the surface material is retained by

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

Fig. 12. Residual tangential stress ru, axial stress rz, and shear
stress s, respectively, in the surface layer vs. number of thermal cycles
v
(QRO 90 at 510 HV30). (a) Heat cycling to 600 C. (b) Heat cycling
v
to 700 C.

the cooler bulk material (see Fig. 5). During the initial
phase of the cooling, the surface contracts but the bulk
material still expands due to heat conduction. Thereby,
the decrease in surface strain with temperature is
delayed (see Fig. 5).
v
When heating to 850 C, the heating rate is dramatically reduced above the Curie temperature of Fe
v
(768 C) (see Table 1), as a result of the change in the
magnetic properties of the material. This promotes the
temperature distribution through the specimen to even
out by heat conduction. Hence, thermal cycling by
induction heating above the Curie temperature does
not ideally generate the temperature prole representative for die casting of e.g. brass [13]. Consequently, the
v
850 C test in this investigation is not as representative
v
for die casting as those of 600 and 700 C, and is not
included in the strain estimations below.
4.2. Mechanical surface conditions during thermal
cycling
4.2.1. Deduction of mechanical surface strain
Since the thermal strain of the surface layer is constrained during the thermal cycling, the surface

1103

Fig. 13. Typical surface strain response during thermal cycling


(QRO 90 at 510 HV30, equilibrium temperature conditions). (a) Heat
v
v
cycling to 600 C. (b) Heat cycling to 700 C.

material will be exposed to cyclic stresses. The hypothetical strains corresponding to these stresses are
dened as mechanical strains Emech. Crack nucleation
and growth during thermal cycling is determined by
uctuations in Emech. Similarly, we dene thermal
strains Eth from the thermal expansion coecient of the
tool material a(T) and the minimum temperatures Tmin
without any constraint as:
eth T aTT  Tmin

In the calculations, a(T) is the nominal value for the


steels as the mean value between room temperature
and the temperature of interest.
Thus, emech at any part of the thermal cycle is possible to deduce from the corresponding values of the
surface temperature cycle, thermal expansion coecient
of the tool material, and the measured surface strain
Etot according to Ref. [15]:
emech T etot T  eth T

Obviously, knowledge of the temperature distribution


below the surface is not necessary.
The three types of surface strain during thermal cycling can be deduced from results such as those of
Figs. 4 and 5, and using Eqs. (2) and (3) (see Fig. 13).
Note that the surface layer is not totally constrained

1104

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

Fig. 14. Mechanical surface strain vs. number of thermal cycles durv
ing heat cycling to 600 and 700 C, respectively (Hotvar 640 HV30).

during the heat cycling because Etot is not represented


by perfectly horizontal lines. The situation should
rather be considered as a partially constrained case.
From mechanical surface strain loops such as those
of Fig. 13, the minimum mechanical surface strain Emin
and any residual mechanical surface strain at the minimum temperature eresidual of each cycle could be extracted (see Fig. 14). For all tests, these two strains seem to
be independent of the number of thermal cycles. For
v
the tests at 600 and 700 C, eresidual was about zero for
all cycles, and emin was about 0.47% and 0.56%,
respectively, for the two temperature levels at all equilibrium cycles. No dierences were seen between the
tested steels. These values of the mechanical surface
strain can be used for numerical simulations and
experimental tests of thermal fatigue processes.
4.2.2. Estimated surface stress and risk of plastic strain
Simplied, the elastic stress r(T) in the surface layer
at any temperature within the thermal cycles of the test
rods can be estimated by Hookes law in plane stress
using the mechanical surface strain emech from this
investigation and Poissons ratio m(T) and modulus of
elasticity E(T) of the tool materials (isotropic and elasticideal-plastic) as:
rT

ETemech T
1  mT

Starting with a stress-free state at cyclic equilibrium


temperature and solving Eq. (4) with mT 0:3, and
E(T) and emech(T) according to Figs. 2c and 13b,
respectively, the stress on the surface layer during two
thermal cycles varies as shown in Fig. 15. It is seen that
during the rst cycle, the compressive surface stress
r(T) increases with temperature, and the surface
material behaves elastically until the stress reaches the
true yield strength of the material ry(T) in compression
(see Fig. 2f). Thus, the critical surface temperature Tc
during heating at which the material begins to deform
plastically is the temperature when r(T) equals the

Fig. 15. Surface stress vs. time during thermal cycling to 700 C at
equilibrium temperature conditions (QRO 90 at 510 HV30). r(T) is
estimated using Eq. (4) and the true yield stress ry(T) is according to
Fig. 2f. The conventional yield strength Rp0.2 according to Fig. 2b,e is
included for reference. (a) Two whole cycles. The residual stress
rresidual after the rst cycle is dened. The arrows indicate the development of r(T). (b) Close-up of the rst 10 s of the rst thermal cycle
(linear-logarithmic scale).

compressive yield strength ry(T). For QRO 90 at


510HV30 and Hotvar at 640 HV30, Tc was found to be
v
approximately 420 and 460 C, respectively. Thereafter, the surface material accumulates compressive
plastic strain until the maximum cycle temperature is
reached. The maximum plastic strain Ep is obtained at
Tmax and can be expressed by Eq. (5) below:
ep emech Tmax  emech Tc
ry Tc 1  mT
emech Tmax 
ETc

During this plastic strain the stress level is determined by the compressive yield strength curve. During
the subsequent cooling, the surface stress is reversed
towards the tensile direction by the thermal contraction. It is seen in Fig. 15 that the developing tensile
stress r(T) during the cooling part of the cycle never
reaches the corresponding ry(T). The magnitude of the
v
tensile residual stress rresidual at Tmin 170 C equals
the maximum rresidual and occurs at Tmin (Fig. 15). It is

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107


Table 2
Estimated plastic strains [%] and residual stresses [MPa] after thermal
v
cycling with preheating (Tmin 170 C)
Material

ep =rresidual
v

QRO 90 at 510 HV30


Hotvar at 640 HV30

Tmax 600 C

Tmax 700 C

0.20/600
0.15/500

0.30/880
0.26/830

deduced by Eq. (6) below:


rresidual ry Tmax 

ETmax emech Tmax


1  mT

Consequently, the residual tensile stress at Tmin is simply the dierence between the elastic stress that the
material would have experienced at Tmax if there was
no plastic deformation (r(Tmax), Eq. (4)) and the yield
stress at Tmax.
Solving Eqs. (5) and (6) with m(T) and E(T) as above,
and Emech(T) and ry(T) according to Figs. 13 and 2f,
respectively, the values shown in Table 2 on the plastic
surface strain after the rst thermal cycle and the
residual stress in the surface layer after each cycle were
obtained. Due to lack of true yield stress values (such
as those in Fig. 2f) for QRO 90 at 430 HV30, no estimations were made on this material. However, the
plastic surface strain magnitude and the tensile residual
stress in the surface layer is expected to be even higher
for this steel. Generally, the magnitude of the plastic
strain and the residual tensile stress increases more
than proportional to the maximum surface temperature
level, since the yield strength decreases at the same time
as r(Tmax) increases with temperature.
During the following thermal cycle, the maximum
compressive stress on the surface material is equal to
the compressive yield strength and, therefore, the
material behaves elastic during the whole cycle. The
reason is the presence of residual tensile surface stress,
which at the end of each cycle, except for the rst one,
is equal to that of the previous one. Consequently, the
surface material is already exposed to the worst mechanical conditions during the rst cycle.
On the other hand, if the surface material only
would experience elastic deformation during the rst
cycle, there would be zero residual stress at the end of
the rst as well as the following cycles.
As seen in Fig. 15 and by comparing the data in
Table 2 with Fig. 2f, the calculated residual stresses
v
after the 600 and 700 C cycles do not exceed the true
v
yield stress of the tool materials at 170 C. In spite of
this fact, numerous cracks appear during the very rst
few cycles. However, the stress estimations are based
on simplied elasticideal-plastic materials and the
information given in Fig. 2. Any thermally induced
alternations in yield stress or any presence of stress

1105

concentrators has not been accounted for. Nor is the


well-known eect that ry increases with strain rate considered. Locally, due to surface irregularities, the tensile stress during cooling may well exceed the yield
stress. It is indicated in Figs. 10 and 11 that the yield
stress values measured prior to testing also degrade signicantly during thermal cycling, causing further
accumulation of plastic strain. This is conrmed by the
measured residual stresses after temperature cycling to
v
600 and 700 C, respectively (see Fig. 12). Consequently, it is reasonable to believe that the surface
material will experience tensile stresses locally that
exceed the yield stress during thermal cycling to 600
v
850 C after a certain number of cycles. This is supported by the observation that thermal cracks are
formed and propagate within the low-cycle fatigue
range (see Figs. 8 and 9).
However, from Fig. 12 it is evident that the surface
layer material initially exhibits a compressive residual
stress of the order of 500 MPa. This stress should be
superimposed to the initial elastic stress during heating
as given in Fig. 15. The consequence is that the surface
material will reach the compressive yield stress earlier
during the rst cycle, and accumulate a larger amount
of compressive plastic strain until the maximum cycle
temperature is reached. However, the residual stress
after the rst and consecutive cycles is not inuenced.
It is also obvious from Fig. 12 that the magnitude of
the measured residual stresses is lower than calculated
ones (see Table 2). The reason, except those mentioned,
is the fact that the X-ray measurements give average
values of the stress within a surface layer of a few
micrometres. The stress in the supercial surface layer
which was calculated is probably much higher. The
limited depth resolution of XRD also explains the
observation that no tensile residual stresses are
v
observed after the rst 600 C cycle (see Fig. 12a).
Similarly, if the surface layer initially had been in a
tensile residual stress state, the sign and magnitude of
the residual stress after the rst heat cycle could also
been given by Table 2, provided that the initial tensile
stress did not completely prevent plastic deformation
during heating.
The conclusion is that an initial tensile residual stress
state above a certain magnitude in the tool material
surface layer may delay crack initiation, whereas initial
compressive residual stresses of any level would facilitate crack formation.
In the above calculations, it was assumed that the
specimens were preheated to the minimum equilibrium
v
temperature of about 170 C and that steady-state temperature conditions prevailed from the rst cycle itself.
v
However, the initial specimen temperature is 60 C,
and from Fig. 4b it is obvious that it takes about four
cycles to obtain cyclic equilibrium temperature conditions with constant minimum and maximum surface

1106

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

Table 3
Estimated plastic strains [%] and residual stresses [MPa] after thermal
v
cycling without preheating (Tmin 60 C)
Material

ep =rresidual
v

QRO 90 at 510 HV30


Hotvar at 640 HV30

Tmax 600 C

Tmax 700 C

0.17/530
0.12/390

0.28/820
0.23/720

temperature. For the 600 and 700 C tests, these conditions are obtained with an approximately constant
temperature range (TmaxTmin), whereas the range for
v
the 850 C experiments decreases slightly during the
rst cycles.
The fact that the mechanical surface strain is related
to the temperature range (see Figs. 13 and 4b), implies
v
that emech(T) for the rst 600 and 700 C cycle should
be approximately equal to that at equilibrium conditions. Solving Eq. (4) as above and using the temperature values for the rst cycle, the critical surface
temperature for plastic compression Tc was found to be
v
approximately 330 and 385 C, respectively, for QRO
90 at 510HV30 and Hotvar at 640 HV30. Solving
Eqs. (5) and (6) as above and using Tmax according to
Fig. 4b, the values given in Table 3 on the plastic surface strain and the residual stress in the surface layer
after the rst thermal cycle without preheating (initial
v
temperature 60 C) were obtained. Again, no estimations on QRO 90 at 430 HV30 could be made.
Obviously, preheating of the specimens to the minimum equilibrium temperature has a tendency to make
the mechanical conditions during the rst cycle worse
than without preheating. However, since the cyclic
equilibrium temperature conditions are obtained within
the rst few cycles (see Fig. 4b), there would probably
not be any notable eect on the thermal cracking.

Fig. 16. Schematic of crack length vs. number of cycles (after Ref.
[16]). The present experiments are conned to the shadowed area.

tures, and material inhomogeneities such as slag inclusions, carbides, and voids [4,17], play a critical role.
Even though the crack length increases within the
whole range of cycles tested, the crack density is saturated after a smaller number of cycles, between 1000
v
5000 and 100500, respectively, for the 700 and 850 C
tests (see Fig. 8). After some initial cycles during which
most cracks are formed, some cracks continue to grow
deeper into the tool material during the subsequent
cycles [4]. Since the material surrounding the cracks is
relieved, the growth of adjacent cracks is retarded [18].
Fig. 8 illustrates the well-known sensitivity to the
maximum temperature in thermal fatigue [19]. In
addition, Figs. 8 and 9 demonstrate that the crack
length and density decreases with increasing tool steel
hardness. Both phenomena are well explained by the
previous estimations of surface stress and risk of plastic
strain.
There is no remarkable inuence of the test environment on the crack characteristics during the early stage
of thermal fatigue (see Figs. 8 and 9).

4.3. Thermal fatigue cracking


5. Conclusions
When exposing the tested tool steels to thermal cycv
ling to 600850 C, it takes the order of <102104 heat
cycles to initiate and propagate the rst thermal fatigue
cracks (see Figs. 8 and 9). After initiation, both the
mean and maximum crack lengths increase substantially with the number of cycles. This agrees with the
experimental results obtained by Lieurade et al. [16]
(see Fig. 16). Initially, the crack length accelerates with
the number of cycles, since the geometrical eect dominates. Later, when the cracks have grown to a certain
depth the driving stress is reduced, and the growth rate
decreases.
In the previous paragraph it was suggested that the
crack initiation is associated with the gradual softening
of the tool steels during the thermal cycling. In
addition, stress concentrators due to topographical fea-

The following conclusions can be drawn from this


study.
. Induction heating and internal cooling of hollow
cylindrical test rods can be used to simulate the type
of surface cracking observed from die casting, forging, rolling, and other processes that involve thermal
cycling.
. The test can be useful for selection and development
of new tool materials in hot forming applications by
giving detailed information on crack nucleation and
propagation, inuence of material parameters such
as hardness and residual stresses, temperature variations, test environment, inuence from cooling
uids, etc.

A. Persson et al. / International Journal of Fatigue 26 (2004) 10951107

. The surface strain recordings oered by the laser


speckle technique, proved to yield sucient information for calculation of surface strains and stresses
responsible for the thermal fatigue of the surface
layer, without knowledge of the temperature prole
below the surface.
. With the tool steels tested, low-cycle fatigue occurs
v
for the tests with Tmax 600 and 700 C even though
the calculated tensile stress after cooling does not
exceed the yield stress that the material exhibits
prior to testing. The reason is the gradual softening
of the tool steels during the thermal cycling, and the
presence of stress concentrators.
. Preheating of the tool does not improve the resistance to thermal fatigue cracking.
. An initial tensile residual stress state of a certain
magnitude in the tool material surface may delay
crack initiation, whereas an initial compressive
residual stress of any level facilitates crack formation.
. The hardness ranking between the materials was
maintained throughout the tests, even though all
steels suered a considerable softening.
. Unfortunately, the induction induced heating to 850
v
v
C is disturbed by the Curie transition at 768 C,
which dramatically reduces the heating rate above
that temperature. Thus, the results generated cannot
be directly transferred to the thermal operations
aimed at.

Acknowledgements
The authors are grateful to Uddeholm Tooling AB,
Tour & Andersson AB and Bodycote Heat Treatment
AB. The nancial support from the Swedish Knowledge Foundation is also acknowledged. Special thanks
to Mr Lars Carlsson at Karlstad University for assisting at the verication tests of the laser speckle arrangement.

References
[1] Sully LJD. Metals handbook, vol. 15, 9th ed. Metals Park (OH):
ASM International; 1988. p. 286.

1107

[2] Davis JR, editor. ASM speciality handbook, tool materials.


Materials Park OH: ASM International; 1995. p. 251.
[3] Allsop DF, Kennedy D. Pressure diecasting, Part 2: the technology of the casting and the die. Oxford: Pergamon Press Ltd;
1983.
[4] Danzer R, Sturm F, Schindler A, Zleppnig W. Thermal fatigue
cracks in pressure die casting dies. Gisserei-Praxis 1983;19(20):
287.
[5] Kovrigin VA, Starokozhev BS, Yurasov SA. Eect of oxidizing
processes on crazing of die-casting molds. Met Sci Heat Treat
1980;22:688.
[6] Nehrenberg AE. Metallurgical aspects of heat checking in brass
die casting dies. Die Castings 1949; Oct/Nov: 30.
[7] Zleppnig W, Danzer R, Fischer FD, Maurer KL. Inuence of
the structure and the temperatureeld on the formation and
propagation of thermal fatigue cracks. Proceedings of the 6th
Biennial European Conference on Fracture, Amsterdam, 1986.
p. 1139.
[8] Schindler AM, Danzer RB. Advanced materials for die casting
and extrusion applicationdie life in die casting. Proceedings of
the International Conference Step into the 90s, 3. Queensland,
1986. p. 905.
[9] Yamaguchi I. A laser-speckle strain gauge. J Phys E Sci Instrum
1981;14:1270.
[10] Yamaguchi I. Speckle displacement and decorrelation in the diffraction and image elds for small object deformation. Opt Acta
1981;28:1359.
[11] Yamaguchi I, Takemori T, Kobayashi K. Stabilized and accelerated speckle strain gauge. Opt Eng 1993;32:618.
[12] Yamaguchi I. Encoder and strain gauge using laser speckle. Opt
Lasers Eng 1989;11:223.
[13] Persson A, Hogmark S, Bergstrom J. Temperature proles and
conditions for thermal fatigue cracking in brass die casting.
J Mater Process Technol [in press]
[14] Velay V, Persson A, Bernhart G, Bergstrom J, Penazzi L. Thermal fatigue of a tool steel: experiment and numerical simulation.
Proceedings of the 6th International Tooling Conference.
Karlstad, 2002. p. 667.
[15] Sehitoglu H. ASM handbook, vol. 19. Materials Park (OH):
ASM International; 1996. p. 527.
[16] Lieurade HP, Dias A, Giusti J, Wisniewski A. Experimental
simulation and theoretical modelling of crack initiation and
propagation due to thermal cycling. High Temp Technol
1990;8:137.
[17] Hockanadel PW, Edwards GR, Maguire MC, Baldwin MD. The
eect of microstructure on the thermal fatigue resistance of
investment cast and wrought AISI H-13 hot work die steel.
Transactions of the 18th International Die Casting Congress and
Exposition, Indianapolis, 1995. p. 343.
[18] Baron HG, Bloomeld BS. Resistance to thermal stress fatigue
of some steels, heat-resisting alloys and cast irons. J Iron Steel
Inst 1961;197:223.
. Performance of hot-work tool steels. Scand
[19] Norstrom L-A
J Metall 1982;11:33.

You might also like