International Journal of Fatigue 92 (2016) 166–178

Contents lists available at ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier.com/locate/ijfatigue

Energy-based approach to thermal fatigue life of tool steels for die

casting dies
Changrong Chen a,d, Yan Wang b, Hengan Ou c, Yueh-Jaw Lin a,⇑
a
Department of Mechanical, Materials and Manufacturing Engineering, University of Nottingham Ningbo China, Ningbo 315100, China
b
School of Computing, Engineering and Mathematics, University of Brighton, Brighton BN2 4GJ, UK
c
Department of Mechanical, Materials and Manufacturing Engineering, University of Nottingham, Nottingham NG7 2RD, UK
d
School of Mechanical and Automotive Engineering, Fujian University of Technology, Fuzhou 350118, China

a r t i c l e i n f o a b s t r a c t

Article history: Thermal fatigue cracking is one of the mostly encountered failure mechanisms for die halves in the die
Received 2 January 2016 casting industry. This is due to rapid alternating heating and cooling of die surfaces during the casting
Received in revised form 19 June 2016 process. In this paper, an experimental thermal fatigue test method based on cyclic induction heating
Accepted 20 June 2016
and water cooling is proposed for the evaluation of thermal cracking of the tool steel used in the industry.
Available online 23 June 2016
An energy-based fatigue life model is formulated by accounting the test period. Finite element models are
developed for better understanding of thermal loadings experienced by samples under the fatigue test-
Keywords:
ing. The results demonstrate that thermal cracking is closely related to inelastic energies dissipated at the
Die casting
Heat checking
material levels. The outcome of this study enables accurate evaluation of crack growth and, thus, evalu-
Thermal fatigue ation of thermal fatigue life of die casting dies by using the proposed energy-based model.
Energy life approach Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
Die casting is a cost-efficient method for forming geometrically
complex, near net-shaped products with close tolerances and
excellent surface finishes [1,2]. As a high volume production pro-
cess, it commonly reaches a production rate of 200 parts per hour
and a production batch of 300,000 parts [3]. Therefore, low cycle
times require high flow velocities and rapid solidification of molten
metal (large thermal gradients). For instance, during aluminium
die casting, molten aluminium (670–710 °C) is injected into the
mould at velocities of 30–100 m/s [3,2]. These severe conditions
limit the service life of die halves. Die casting dies mostly fail in
the modes, such as heat checking, soldering and corrosion. Heat
checking (also known as thermal cracking) is the principal mode
in hot working. It specifically accounts for 70% of failures in die
casting dies [4]. Die failures cause significant loss to the die casting
industry due to high cost of die halves and production suspension
as a result of die downtime [5]. Therefore, this has become a major
research topic for die workers to increase thermal fatigue resis-
tance of hot-work die steels.
Christopher [6] and Srivastava et al. [5] developed computa-
tional models to understand thermal and structural behaviour of
test samples and to predict fatigue cracking in die steel samples
⇑ Corresponding author.
E-mail address: Y.J.Lin@nottingham.edu.cn (Y.-J. Lin).
studied experimentally by the Wallace test [7]. Li et al. [8] defined
thermal fatigue crack initiation (TFCI) life from engineering point
of view and proposed an expression for the TFCI life of H13 and
H21 steels based on a modified Coffin–Manson expression. Velay
et al. [9], Persson [10] and Persson et al. [11] studied thermal fati-
gue cracking behaviour of die steels using an experimental thermal
fatigue test method based on induction heating and internal water
cooling. They developed a strain-based approach for thermal fati-
gue resistance. Klobčar et al. [12,13] and Klobčarand Tušsek [14]
established finite element models for understanding thermal stres-
ses experienced by die steel samples during immersion tests. The
thermal loadings were then used to compare resistances of differ-
ent tool steels, AISI H11, AISI H13 and maraging steel. Thermal fati-
gue resistance of surface engineered die steels was also popularly
investigated. Yatsushiro et al. [15] applied laser peening for hot
work die steel, AISI H13, for the improvement of the resistance
to crack growth. Moreover, Borrego et al. [16] studied the effect
of laser deposit welding on the fatigue resistance of two die mate-
rials in mould production (AISI H13 and P20).
In most of the literature, Coffin–Manson expression and its
modifications are frequently applied for analysing thermal fatigue
resistance. This approach correlates thermal fatigue life with cyclic
strains. Christopher [6] and Sakhuja and Brevick [17] used the
method of universal slopes to relate cyclic strain ranges to the
number of cycles necessary for fatigue crack initiation. Persson
[10] extended the strain-based approach to account for the crack

http://dx.doi.org/10.1016/j.ijfatigue.2016.06.016
0142-1123/Ó 2016 Elsevier Ltd. All rights reserved.
 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178 167

propagation stage by defining corresponding strain intensity. Sim- Table 2

ilarly, Li et al. [8] proposed a temperature-based approach for the Typical mechanical properties of the mould steel (Room temperature).

TFCI life based on a modified Coffin–Manson model. The main Hardness (HRC) Tensile strength, Rm (MPa) Yield strength, Rp (MPa)
problem for this method is that either crack initiation and propaga- 52 1820 1520
tion are not distinguished or the length for crack initiation is usu- 45 1420 1280
ally subjectively defined. Based on the measurements made on
sections from failed die casting dies, according to Parishram [18],
cracks that cause noticeable heat checks for the die casting appli-
cation had an average of 0.5 mm in length. Therefore, for the die
casting industry, a life approach capable of accounting for cracking
procedure would be more appropriate.
In this paper, an energy-based approach is proposed to investi-
gate thermal cracking of die material. Firstly, discrete thermal
cracking data were obtained through thermal fatigue tests of die
steel samples using a dedicated test machine. Then, numerical
modelling was conducted to understand work loadings experi-
enced by the material points. And lastly, crack data were correlated
with the thermal/structural loadings and thermal fatigue life
model for the die material was established.

2. Experimental work

Experimental work is investigated in this section to understand

Fig. 1. Microstructure of 8407 supreme after heat treatment.
thermal cracking process. In this experimental work, the com-
monly used die material 8407S (produced by ASSAB company) is
used. A thermal fatigue test machine is purposely designed for this
research study. Temperature profiles of samples during testing are 02
± 0,
60° ± 0,5° R0,1
captured using an infrared camera for the calibration of thermal
0,2
analysis in the forthcoming section for finite element modelling.

2.1. Specimen design and preparation DETAIL A

1,5±0,05
SCALE 8,000
Test samples are made of the material 8407 supreme, which is a 20±0,02 5±0,02
premium high quality AISI H13 die steel. It is characterized by high
level of resistance to thermal shock and thermal fatigue. The elon-

0,4
gation at break is 10% at 500 °C and 600 °C and 20% at 700 °C. The
properties enable the material particularly suitable for tooling sub-
A
jected to high mechanical and thermal fatigue stresses, e.g. die
20±0,02

casting dies, forging tools and extrusion tooling [19]. The chemical
composition and mechanical properties of the tool steel are
depicted in Tables 1 and 2, respectively. The microstructure of
10±0,1

the tool steel after heat treatment is illustrated in Fig. 1.

With heat treatment to be harder than 54HRC as die casting
dies, the test materials were machined into specimens with dimen-
sions of 20
20
5 mm. As shown in Fig. 2, the specimens have a Fig. 2. Drawing of thermal fatigue testing sample.

notch of 0.1 mm in radius with a wire Electrical Discharge Machin-

ing (EDM) cut. The geometry of the specimen is defined to (i)
obtain an effective thermal gradient in the cross section of the
specimen; (ii) guarantee efficient cooling in the central part of
the specimen; (iii) generate a state of stress concentration; (iv)
limit manufacturing costs of the specimen; and (v) consider die
geometry effects on thermal cycling. Surfaces under major thermal
loads were also well finished to eliminate the effect of machining
on thermal fatigue resistance.
2.2. Experimental set-up and arrangements
Until recently, numerous methodologies are proposed for ther-
mal fatigue testing of hot-work mould steel materials [9,13,20].
Table 1
Chemical composition of the analysed mould steels (wt%).
Mould steel C Si Mn Cr
8407S (Premium AISI H13) 0.39 1.0 0.4 5.2
The major difference amongst them is in the heating and cooling
means obtaining rapid heating and cooling effects that are experi-
enced in die halves. Therefore, this study utilizes induction heating
based thermal fatigue test equipment, as schematically repre-
sented in Fig. 3. Each specimen was heated by an induction coil
and immediately cooled by flushing water to achieve thermal
shock effects as die casting dies experience. The controllable
parameters include frequency for heating current (F, Hz), heating
time (th , s), and cooling time (tc , s). Test samples were taken out
for the observation of cracking information in the vicinity of the
notch tip firstly at 800 cycles and then every 400 cycles. After elim-
ination of the oxide layer by ultrasonic cleaning in dilute
hydrochloric acid solution (10%) for 10–15 min, thermal cracks
within test samples were evaluated by optical microscopic obser-
vation. During thermal fatigue testing, temperature profiles of test
sample surface were measured using a FLIR T640 thermal imaging
Mo V
camera. The data will then be applied in calibration of finite ele-
1.4 0.9
ment modelling of thermal fatigue test as well.
168 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178

Table 5
Water inlet Parameters calibrated for the thermal infrared camera.

Parameter Value Unit

Emissivity 0.34 –
Reflected apparent temperature 18 °C
Test sample
Atmosphere temperature 20 °C
Relative humidity 0.5 –
Distance 0.4 m
Infrared imaging
Induction coil

Inductor

Water tank
Water outlet

Fig. 3. A schematic representation of thermal fatigue test equipment.

Table 3
Experimental variables and levels for thermal fatigue testing.

Variables/levels 1 2 3
Frequency, F (Hz) 600 700 800
Heating time, th (s) 1 2 3 y
Cooling time, t c (s) 1 1.5 2

z x
Table 4 P1
Experimental arrangement for thermal fatigue testing.

No. of sample Factor A Factor B Factor C

N1 2 3 3 Fig. 5. Meshing of test specimens in FEM.
N2 1 3 2
N3 3 1 2
N4 3 2 1
N5 2 1 1
N6 1 2 3

Cooling
planes
Heating
planes

Symmetry
planes
y

z x

Fig. 6. Thermal analysis boundary conditions for thermal fatigue tested samples.

Tables 3 and 4. The experimental plan takes three levels of each

test variable into consideration.

2.3. Calibration of emissivity of test samples

Fig. 4. Calibration of emissivity.

The radiation measured by an infrared camera is influenced by

In order to establish the damage evolution model, a series of
thermal fatigue tests are conducted. This is considered to be neces-
sary in order to obtain a panorama view of the material behaviour.
Generally, at least six series of strain controlled fatigue tests were
undertaken with three duplications. Such an experimenting
method is usually condemned for being time-consuming and it
also becomes more complicated in the case of thermal fatigue. Uni-
form design offers efficient experimental arrays with acceptable
accuracy kept. The experiment design method was adopted for
the test and experimental runs were designed as illustrated in
many factors including the emissivity of the object, the reflected
apparent temperature, the distance between the camera and the
object, the relative humidity and temperature of the atmosphere
[21]. To measure temperature accurately, it is necessary to cali-
brate the most important parameter, emissivity. In this paper,
the reflected apparent temperature was determined in advance
as the temperature of a large piece of aluminium foil worked as a
reflector of the reflection source.
The calibration of the emissivity of test samples was then car-
ried out following the procedure suggested in FLIR R&D software
manual: (a) a piece of 3M type 88 black vinyl electrical tap with
 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178 169

3. Finite element modelling

Start
The main objective of numerical simulation is to investigate
stress–strain loops of test samples under different testing condi-
tions. As large spatial temperature gradients exist in the sample,
Initialization and the plastic strain rate with the body is small, heat generation
due to irreversible deformation can be neglected [22,6]. Therefore,
the thermo-mechanical analysis is performed by the ABAQUSÒ FE
package in sequentially coupled manner.
Set Variables Due to the geometrical symmetry in two planes, i.e. XY plane
Experimental and YZ plane, one quarter FE model of specimens is developed. This
Divergence quarter model is analysed using eight-node linear hexahedron ele-
Results
ments, as shown in Fig. 5. The meshing is refined near the notch tip
and the outer surface of the sample to capture thermal gradients.
Update ABAQUS
Python files
no 3.1. Thermal analysis
Minimized?
The thermal fatigue cycle is comprised of two stages: induction
heating and water cooling. The following boundary conditions dur-
Call Abaqus yes ing testing were considered in the finite element analysis:

Save Data
Read FEM
results End
Fig. 7. Thermal calibration procedure using ABAQUS-MATLAB interaction.
the emissivity of 0.96 was put on a sample; (b) the sample was
evenly heated to a temperature 20 K above room temperature;
(c) an infrared image of the sample was captured; (d) the emissiv-
ity was adjusted to obtain equal temperatures of the tap surface
and the sample surface. Fig. 4 shows an image for the calibration
of the emissivity of test samples. Therefore the emissivity of test
sample surface is calibrated as 0.34. Other related parameters are
also calibrated and illustrated in Table 5.
 Surface heating introduced at heating stage. For the frequency
range in this work, the induced energy flowing into the part is
most intense on the surface and decays rapidly below the sur-
face. Surface heat flux is effective to model the heating process.
For better fitting with the experimental thermal cycles, a Gaus-
sian distribution of heat flux density with respect to y axis is
assumed on the heating surface.
 Acute convective exchange condition to account for water cool-
ing. The heat transfer coefficient between the material and
water is initially assumed to be 0.028 W/mm2 K and then opti-
mized through thermal calibration process as depicted in Fig. 7.
 Gentle convective heat transfer to account for the air cooling. A
constant heat coefficient of 5
106 W/mm2 K is defined for all
the surfaces.
 The symmetry planes indicated in Fig. 6 are isolated.

28
Conductivity (W/m-k)

Coefficient of Thermal

14.0
Expansion (μ/K)

27
13.5
26
13.0
25
12.5
24
0 200 400 600 800 400 600 800
Temperature (°C) Temperature (°C)

1.0
Specific Heat (J/g-K)

20C
True Stress (MPa)

1600
550C
0.8
1200
700C

0.6 800

400
0.4
0
0 200 400 600 800 1000 0.00 0.05 0.10 0.15 0.20
Temperature (°C) True Strain (1)
Fig. 8. Material properties of ASSAB 8407S mould steel.
170 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178

450 450

Temperature (°C)

Temperature (°C)
400 400

350 350

300 300
FEM predictions FEM predictions
Measurements Measurements
250 250
95 96 97 98 85 86 87 88 89
Time (s) Time (s)
(a) N1 (b) N2
350 450
Temperature (°C)

Temperature (°C)
400
300

350

250
300
FEM predictions FEM predictions
Measurements Measurements
200 250
47.6 47.8 48 48.2 48.4 48.6 66.5 67 67.5 68 68.5
Time (s) Time (s)
(c) N3 (d) N4
350 400
Temperature (°C)

Temperature (°C)

350
300

300

250
250
FEM predictions FEM predictions
Measurements Measurements
200 200
76 76.5 77 77.5 78 76 76.5 77 77.5 78
Time (s) Time (s)
(e) N5 (f) N6
Fig. 9. Comparison of calculated and measured thermal profiles of P1.

Fig. 7 illustrates the procedure used to calibrate heat flux due to
induction heating and heat transfer coefficient between coolant
and specimen. The main function in MATLABÒ modifies the vari-
ables and updates them in Python files accordingly. The Python
files are then called in the MATLABÒ environment to execute finite
element analysis and postprocessing. The results from ABAQUSÒ
analyses are compared with experimental data. Finally, the diver-
gence is minimised to get optimal variables.
Other temperature-depended material properties of ASSAB
8407S tool steel are summarized in Fig. 8. In order to develop a
quasi-steady-state heat transfer profile of the specimen, a series
of twenty thermal cycles were run. Each cycle is modelled as one
heat transfer step in ABAQUS with two amplitudes defined to stim-
ulate two stages. The initial temperature of specimens is assumed
equivalent to ambient temperature, i.e. 23 °C.
Temperature profile of the point P1, 1.5 mm below the notch
tip, was measured by the thermal infrared camera for the
comparison of data calculated by FEM. The divergence is mini-
mized to optimize the non-uniform surface heat flux and the heat
transfer coefficient between the sample and cooling water. In order
to exclude the effects of cooling water, any temperature data less
than 100 °C during heating process are not considered in the opti-
mization stage. The parameters of optimized thermal analysis
enabled a close correlation between the measured and computed
thermal fields, as illustrated in Fig. 9.
3.2. Mechanical analysis
With the heat flux and the heat transfer coefficient calibrated,
temperature history profiles of the sample can be computed via
the thermal analysis. This is also described in the previous section.
The computed thermal fields throughout the transient are then
imported as the predefined field for the mechanical analysis. Since
this analysis investigates the structural behaviour of test speci-
 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178 171

UZ= 1-paper
0 3-paper
Sprin 2-paper
gs 600 4-paper
3-fem
4-fem

Temperature [°C]
1-fem
2-fem
400

UX=0
200

y
0
0 10 20 30
z x Time [s]
UY=
0 Fig. 12. Comparison of transient temperature from literature and FEM.

Fig. 10. Structural analysis boundary conditions for thermal fatigue tested samples.
1-paper
3-paper
mens after achieving quasi-steady-state thermal profile, only the 2-paper
thermal data of last step is introduced. The mechanical analysis 800 4-paper
also uses eight-node linear hexahedron elements. The elastic–plas- 3-fem
4-fem
tic constitutive model is applied, with the yield stress, the ultimate
1-fem
stress, and the work hardening properties of the material defined. 400
2-fem
A strain rate independent model with the Von Mises yield criterion,
s33 [MPa]

isotropic hardening, and Poisson’s ratio of 0.3 is used in the analy-

sis. Coefficient of thermal expansion of the material is considered 0
temperature-dependent in the analysis.
Boundary conditions for the mechanical part are illustrated in
Fig. 10. The symmetrical XY plane and YZ plane of the specimen -400
are restrained in z-axis and x-axis, respectively. Displacement in
y-axis is fixed at the bottom plane, while engineering springs with
stiffness of 0.5 N/mm are assumed on the top surface.
-800
0 10 20 30
3.3. Verification of modelling methodology Time [s]

The modelling methodology adopted in this study is verified by
conducting a benchmark study of published work in the literature.
Klobčar et al. [13] studied the finite element computation of test
specimens under thermal fatigue tests using immersion tests.
The specimens are subjected to cyclic heating in bath of molten
Aluminium Alloy 226 and cooling in bath of water-based coolant.
The specimens are also continuously cooled with cold water
Fig. 13. Comparison of transient stress from literature and FEM.
flowing through internal channels. In the literature, Klobčar et al.
[13] investigated the influence of immersion test parameters,
material, specimen edge geometry, and thickness of maraging steel
surfacing welds on thermal stresses.

y y Outer
2 wall

1 Symmetry
4 wall
3 x
φ 9.5
Cooling
channel
25 Symmetry x
wall
(a) complete (b) one-quarter model
Fig. 11. Cross section of test specimen in literature.
172 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178
Fig. 14. Thermal fatigue crack growth under different testing conditions.
The benchmark study only analyses the standard specimen
shown in Fig. 11. The results from the literature and the methodol-
ogy utilized in this paper are compared, as illustrated in Figs. 12
and 13. The data are in general agreement with those from the lit-
erature. Transient thermal results are slightly greater than data
from the benchmark methodology. The same case can be observed
in the comparison of stresses. The only exception is for the last two
stages where the data decrease more dramatically in the bench-
mark analysis. From this comparison, it can be concluded that
the methodology used in this paper is viable.
(a) 800 cycles
(c) 1600 cycles
Fig. 15. Thermal cracking observation on sample N2.
4. Results
4.1. Experimental testing
Test specimens cracked through the thickness and cracks of
both sides were observed. With conservative concerns, the larger
side was considered as the crack propagation data for the entire
specimen. The crack growth data (a  N) of six experimental runs
designed in previous section are depicted in Fig. 14. From the fig-
ure, one can observe that six test conditions lead to a roughly
three-area distributed a  N curves. This in turn proves that the
experimental array has produced a uniform distribution of test
points.
In addition, Fig. 15 illustrates an example of thermal cracking
observed on sample N2. Following the procedure, the cracks are
observed and measured using an optical microscope at 800,
1200, 1600, 2000 and 2400 cycles. From the figure, it is obvious
that thermal cracks initiate from the vicinity of the tip. Due to
the imperfection of the material and preparation of samples, there
is usually more than one crack within the specimen, as shown in
Fig. 15a. But when specimens undergo for more number of cycles,
one crack dominates the propagation until fracture of the struc-
ture, as illustrated in Fig. 15b–d.
Furthermore, Fig. 16 shows temperature field of one sample
during a thermal testing cycle with several frames. The same levels
of temperature have been assumed for the figures so that thermal
fatigue test cycle can be easily observed from the colour level. The
water-based coolant is flushed down to the notch when the cooling
starts. Therefore, the temperature is the lowest at the notch and
gets higher towards the edge, as illustrated in Fig. 16a. The hot area
is getting smaller when cooling continues (see also Fig. 16b). When
(b) 1200 cycles
(d) 2000 cycles
 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178 173

(a) t=20.433 s, start of cooling (b) t=20.800 s, end of cooling

(c) t=22.000 s, start of heating (d) t=24.800 s, end of heating

Fig. 16. Thermal infrared imaging on sample N2.

Temperature profiles of every material point on the specimen

p1.Average surface can be extracted from the temperature field. Fig. 17 illus-
400 trates the thermal profile of point P1 which is 1.5 mm below the
notch tip. The temperature varies between 430 °C and 30 °C. Great
thermal gradients can be found right at the instants of heating and
p1.Average [oC]

300 cooling. The reason is that thermal infrared cameras are only cap-
able of capturing radiation of the material surface. Therefore, when
the coolant has not vaporized from the specimen, the obtained
200 radiation would be from the coolant rather than the specimen
surface.

100 4.2. Numerical simulation

Fig. 18 illustrates the transient 20-cycle temperature of the

0 notch tip under designated six conditions. From the figures, one
30 32 34 36 38 40
can find that test specimens accomplish the thermal equilibrium
Time [s] within around 20 cycles. Therefore, the temperature field of the
20th cycle can be used for the study of structural behaviour.
Fig. 17. Temperature profile of P1 under F600-h3-c1p5.
The corresponding stress profiles of the notch tip under these
six conditions are depicted in Fig. 19. According to the stress pro-
file, test samples experience great stresses at the notch tip in the
heating begins, the specimen is heated at the centre of both sides. x-direction due to the restraint of neighbouring elements. While
The notch is then comparably cool within the sample (Fig. 16c). in other directions, little stresses are produced since the notch
And at the end of heating, the entire specimen gets heated up, as tip is free to expand or contract. One can also observe that maxi-
shown in Fig. 16d. mum compressive stress and maximum tensile stress occur at
174 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178

(a) F = 700, th = 3.0, tc = 2.0 (b) F = 600, th = 3.0, tc = 1.5

(c) F = 800, th = 1.0, tc = 1.5 (d) F = 700, th = 2.0, tc = 1.5

(e) F = 600, th = 2.0, tc = 2.0 (f) F = 800, th = 2.0, tc = 2.0

Fig. 18. Transient thermal profile of the notch tip under different conditions.

the notch tip at the beginning of heating and cooling, respectively.
However, the magnitude of compressive stress is much smaller
than that in tension. This is probably as a result of the samples
heated from the centre rather than from the notch. Induced energy
was firstly introduced to the central part and then spread to the top
and bottom parts.
fatigue life to failure ranges between 100 and 10,000 cycles. The
Coffin–Manson fatigue model, proposed independently by Coffin
[23] and Manson [24], has been widely used for the evaluation of
low cycle fatigue failure [25].
The total number of cycles to failure, N f , is dependent on the
plastic strain range, D
p , the fatigue ductility coefficient,
0f , and
the fatigue ductility exponent, c, given in the expression below:

5. Thermal fatigue modelling

D
p
¼
0f ð2Nf Þc ð1Þ
2
5.1. Fatigue criterion
The fatigue ductility coefficient,
0f , is approximately equal to
Thermal fatigue failure in dies and moulds subject to thermal the true fracture ductility,
f . The fatigue ductility exponent, c, var-
cycling tend to fall in Low Cycle Fatigue (LCF) field where the ies between 0.5 and 0.7 [26], and experimental data are
 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178 175

(a) F = 700, th = 3.0, tc = 2.0 (b) F = 600, th = 3.0, tc = 1.5

(c) F = 800, th = 1.0, tc = 1.5 (d) F = 700, th = 2.0, tc = 1.5

(e) F = 600, th = 2.0, tc = 2.0 (f) F = 800, th = 2.0, tc = 2.0

S11 - Stress in the x direction; S22 - Stress in the y direction; S33 - Stress in the z direction.

Fig. 19. Stress profiles of the notch tip under different conditions.

required to determine the constants. The constants in the Coffin– to predict the crack initiation and crack growth behaviour using
Manson model are dependent on temperature and cyclic fre- the equations given below:
quency, which is a big difficulty for thermal fatigue situation where
temperature fluctuates.
N 0 ¼ C 1 DW C 2 ð2Þ
Energy-based fatigue criteria form another group of models. The
da
advantage of energy models is the capability of describing influ- ¼ C 3 DW C 4 ð3Þ
ences due to the interaction of stress and strain and allowing for dN
a generalisation to multi-axial loading. Energy criteria are repre- where DW is the plastic work, a is the area of the solder joint where
sentative for the cyclic behaviour of materials and are linked to a fatigue crack will grow till failure, C i ; i ¼ 1; . . . ; 4 are constants fit-
macroscopic crack initiation [27]. ted by the test data.
An empirical plastic work model was developed by Darveaux In this research, the plastic work density per unit time was used
[28,29] for solder joint subject to thermal cycling tests and is used as a parameter when evaluating thermal fatigue failure. Fig. 21
176 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178

1400 1400

1200 1200

1000 1000

Stress (MPa)
Stress (MPa)

800 800

600 600

400 400

200 200

0 0
s1 vs e1 s1 vs e1
−200 −200
0 0.002 0.004 0.006 0.008 0.01 0.012 0 0.002 0.004 0.006 0.008 0.01
Strain (1) Strain (1)
(a) F = 700, th = 3.0, tc = 2.0 (b) F = 600, th = 3.0, tc = 1.5

800 1200

1000
600
800
Stress (MPa)

Stress (MPa)

400
600

200 400

200
0
0
s1 vs e1 s1 vs e1
−200 −200
0 1 2 3 4 5 0 1 2 3 4 5 6 7 8
−3 −3
Strain (1) x 10 Strain (1) x 10

(c) F = 800, th = 1.0, tc = 1.5 (d) F = 700, th = 2.0, tc = 1.5

800 1400

1200
600
1000
Stress (MPa)

Stress (MPa)

400 800

600
200
400

200
0
0
s1 vs e1 s1 vs e1
−200 −200
0 1 2 3 4 5 0 1 2 3 4 5 6 7 8
−3 −3
Strain (1) x 10 Strain (1) x 10

(e) F = 600, th = 2.0, tc = 2.0 (f) F = 800, th = 2.0, tc = 2.0

Fig. 20. Maximum principal stress and strain loops of different conditions.

shows the definitions of plastic work density per unit time DP, the where the plastic strain energy density DW, is the area of the
total strain amplitude D
, plastic strain amplitude D
p , and stress stress–strain loop of thermal stable cycles, and Dt is the cycle
amplitude Dr for the cyclic stress–strain curve. DP can be period.
expressed as follows: Fig. 20 plots the maximum principal stress and strain loops of
the notch tip under proposed six different conditions. The stress–
DW strain hysteresis curves indicate different stress–strain routes from
DP ¼ ð4Þ
Dt those obtained by mechanical fatigues. The curve loops in the anti-
 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178 177

2 2400 Nf vs. dP
p
Power fit
2100

1800
W
2

Nf
1500

P= W/ t 1200

900
2
0.21 0.24 0.27 0.3 0.33
Fig. 21. Schematic representation of the plastic strain amplitude, stress amplitude,
and plastic work density per unit time.
dP
Fig. 23. N f  dP data and model fitting.

1. Stress–strain transients of material points along the central line

were calculated by sequentially coupled thermo-mechanical
analysis.
2. The plastic work density was computed for every material point
along the central line.
3. The plastic work density of material points were interpolated in
Crack direction
accordance with crack lengths observed by thermal fatigue
tests, as illustrated in Fig. 22.
4. The plastic work density per unit time of material points
observed by crack lengths were computed by dividing the plas-
No. of Cycles tic work density over test periods.
5. The model constants were fitted by the plastic work per unit
Crack length

time vs number of cycles to crack data.

Fig. 22. Thermal fatigue crack evolution demonstration.
Fig. 23 plots the mean plastic work per unit time of studied con-
ditions with regard to each number of cycles to crack. The data
were then used to fit the proposed fatigue model, as also depicted
in Fig. 23. The Rsquared value (0.9988) of coefficients indicates
good fitting of the model. Therefore, the fatigue criterion is given
by:

Nf ¼ 61:38DP2:412 ð6Þ
clockwise direction instead of the clockwise direction as structures
under mechanical fatigue. By introducing
By defining the plastic work density per unit time, the fatigue
criterion for this research is given as: C 1 ¼ C C
3
2
ð7Þ

 And then the model can be rewritten as

DW C 2
C2
Nf ¼ C 1 ¼ C 1 DP C 2 ð5Þ DP
Dt Nf ¼ ð8Þ
C3
pffiffiffiffiffiffi
where C 3 ¼ C 2 C 1 ¼ 5:512 is the plastic work density per unit time
5.2. Criterion fitting
due to one cycle fatigue.
The model can be validated by considering the simplest case of
The constants of the fatigue model (Eq. (5)) need to be cali-
fatigue loading, tensile test loading. Assuming the sample fractures
brated before the criterion implementation for fatigue prediction.
when strain reaches 0.1, then the plastic work density might be
For simplicity, the following assumptions are made to the pre-
approximated by the product of yield stress and the strain, i.e.
sented fatigue model:
DW ¼ 1520 MPa
0:1 ¼ 152
106 J=m3 . Thus, the period for a
sample to failure in a tensile test is about 27.6s which is a reason-
 Crack initiates from the notch tip and propagates along the cen-
able magnitude.
tral line below the notch.
 A material point is damaged when the criterion is met.
 Loading variations due to crack evolution are ignored. 6. Conclusions

The constants of the fatigue model were calculated by under- In this study, thermal fatigue resistance of hot work tool steel
taking the following steps: 8407S, which is commonly used in the die casting industry, was
178 C. Chen et al. / International Journal of Fatigue 92 (2016) 166–178
evaluated. Moreover, thermal fatigue life was modelled using an
energy-based approach. Based on the results in this work, the fol-
lowing conclusions are drawn:
 Through induction heating and external cooling of test plates
and recording of the surface temperature during thermal
cycling, the mechanism of thermal cracking can be reproduced.
 Thermal cracking is closely related to inelastic energies dissi-
pated at the material points and test period.
 A thermal fatigue life model based on dissipated inelastic
energy is suggested. The model is capable of predicting thermal
cracking of the tool steel.
Acknowledgements
The authors would like to acknowledge the financial support
from National Natural Science Foundation of China (Grant No.
51105212), Ningbo Science and Technology Bureau (Project ID
2011B81006) and Ningbo International Collaboration Fund (Project
ID 2011D10003).
References
[1] Garza-Delgado A. A study of casting distortion and residual stresses in die
casting. Ph.D. thesis, The Ohio State University; 2007.
[2] Klobčar D, Kosec LKBTJ. Thermo fatigue cracking of die casting dies. Eng Fail
Anal 2012;20:43–53.
[3] Shivpuri R, Chu Y-L, Venkatesan K, Conrad J, Sridharan K, Shamim M, et al. An
evaluation of metallic coatings for erosive wear resistance in die casting
applications. Wear 1996;192(12):49–55.
[4] Starling C, Branco J. Thermal fatigue of hot work tool steel with hard coatings.
Thin Solid Films 1997:436–42.
[5] Srivastava A, Joshi V, Shivpuri R. Numerical models and their validity in the
prediction of heat checking in die casting tooling. In: 6th International tooling
conference; 2004.
[6] Christopher R. Analysis of thermal fatigue and heat checking in die-casting
dies: a finite element approach. Thesis, Mechanical Engineering; 1992.
[7] Benedyk J, Moracz D, Wallace J. Thermal fatigue behavior of die materials for
aluminium die casting. In: Tool and die failures. American Society for Metals;
1970. p. 187–206.
[8] Li G, Li X, Wu J. Study of the thermal fatigue crack initial life of h13 and h21
steels. J Mater Process Technol 1998;74:23–6.
[9] Velay V, Persson A, Bernhart G, Penazzi L, Bergström J. Thermal fatigue of a tool
steel: experiment and numerical simulation. In: Proceedings of the 6th
international tooling conference: the use of tool steels: experience and
research; 2002. p. 667–85.
[10] Persson A. Strain-based approach to crack growth and thermal fatigue life of
hot work tool steels. Scand J Metall 2004;33(1):53–64.
[11] Persson A, Hogmark S, Bergstrom J. Thermal fatigue cracking of surface
engineered hot work tool steels. Surf Coat Technol 2005;191(2–3):216–27.
[12] Klobčar D, Tušek J, Taljat B. Finite element modeling of GTA weld surfacing
applied to hot-work tooling. Comput Mater Sci 2004;31(3–4):368–78.
[13] Klobčar D, Tušek J, Taljat B. Thermal fatigue of materials for die-casting tooling.
Mater Sci Eng A-Struct Mater Prop Microstruct Process 2008;472(1–
2):198–207.
[14] Klobčar D, Tušek J. Thermal stresses in aluminium alloy die casting dies.
Comput Mater Sci 2008;43(4):1147–54.
[15] Yatsushiro K, Sano M, Kuramoto M. Thermal fatigue properties of laser peened
hot work die steel (h13). In: Advances in X-ray Analysis, vol. 46. International
Centre for Diffraction Data; 2003.
[16] Borrego L, Pires J, Costa J, Ferreira J. Mould steels repaired by laser welding. Eng
Fail Anal 2009;16:596–607.
[17] Sakhuja A, Brevick JR. Prediction of thermal fatigue in tooling for die-casting
copper via finite element analysis. Materials processing and design: modeling,
simulation and applications, Pts 1 and 2 712; 2004. p. 1881–86+2298.
[18] Parishram P. Laser assisted repair welding of h13 tool steel for die casting.
Thesis, Mechanical Engineering; 2007.
[19] UDDEHOLM. ASSAB 8407 SUPREME. UDDEHOLM, 080808th Edition; 2012.
[20] Meng C, Zhou H, Cong D, Wang C, Zhang P, Zhang Z, et al. Effect of biomimetic
non-smooth unit morphology on thermal fatigue behavior of h13 hot-work
tool steel. Optics Laser Technol 2012;44:850–9.
[21] FLIR. Flir R&D Software 3.3-Users manual; 2012.
[22] Kovalenko AD, Alblas JB. Thermoelasticity: basic theory and
applications. Groningen: Wolters-Noordhoff; 1969.
[23] Coffin L. A study of the effects of cyclic thermal stresses on a ductile metal.
Trans Am Soc Mech Eng 1954;76:931–51.
[24] Manson S. Behavior of materials under condition of thermal stress. Tech rep.,
University of Michigan Engineering Research Institute; 1953.
[25] Sasaki K, Takahashi T. Low cycle thermal fatigue and microstructural change of
AC2B-T6 aluminum alloy. Int J Fatigue 2005;28(3):203–10.
[26] Kilinski T, Lesniak J, Sandor B. Modern approaches to fatigue life prediction of
SMT solder joints. In: Lau J, editor. Solder joint reliability. US: Springer; 1991.
p. 384–405.
[27] Christiner T, Eichlseder W, Gódor I, Reiser J, Trieb F, Stuehlinger R. Fretting
fatigue and wear: experimental investigations and numerical simulation. Tech
rep., SAE technical paper; 2011.
[28] Lau JH. Solder joint reliability – theory and applications. 1st ed. US: Springer;
1991.
[29] Darveaux R. Effect of simulation methodology on solder joint crack growth
correlation. In: Electronic components amp; technology conference, 2000.
2000 Proceedings. 50th; 2000. p. 1048–58.