Materials and Design 32 (2011) 1508–1514

Contents lists available at ScienceDirect

Materials and Design

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

Short Communication

Thermal fatigue of cast irons for automotive application

Dhouha Mellouli a,⇑, Nader Haddar a, Alain Köster b, Alexandra Marie-Louise Toure c
a
Unité de Recherche de Chimie Industrielle et Matériaux (URCIM), ENIS, BPW 1173, Sfax, Tunisia
b
Centre des matériaux Pierre Marie Fourt, Ecole des Mines de Paris, France
c
PSA Peugeot Citroën, La Garenne-Colombes Cedex, France

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

Article history: The thermal fatigue damage constitutes the major problem of the parts subject to a variation in temper-
Received 27 April 2010 ature during their operation. Several materials have been used for such structures in order to limit their
Accepted 22 October 2010 damage. The facility of implementation and low cost of cast iron constitute an advantage for its use for
Available online 28 October 2010
the Diesel motors.
The aim of this work is to study the thermal fatigue damage of four nodular cast irons for automotive
application. For this purpose, experimental test including crack growth tests and metallographic obser-
vations were achieved.
The study has revealed the effects of temperature, microstructure and chemical composition of the
studied materials on the initiation and the crack growth under thermal fatigue loading.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction
The cast iron presents the best compromise between the sim-
plicity of implementation and cost, then remains the privileged
material for the Diesel motorizations [1–4].
The exhaust lines are subject to mechanical and thermal loading
[5]. The thermal solicitations during acceleration–decelerations
and starting up or shutdowns of the vehicle produces thermal fati-
gue damage coupled with high temperature effects (oxidation and
viscoplasticity).
Thermal fatigue facility consists in studying a representative
specimen in laboratory. Thermal stresses are the result of both
structure and heat gradients effects.
Several tests were developed to appreciate the thermal fatigue
effect on the structure among them the test which was developed
by Glenny et al. [6] at the end of the Fifties to simulate the thermal
fatigue of the turbine blades. A disc simulating the turbine blade is
subject to thermal fatigue by the technique of the fluidized beds. In
order to appreciate the resistance of the different grades used for
exhaust manifolds application, Bucher et al. [7,8] developed a ther-
mal fatigue test. The test consists of generate alternate cycles of
heating and cooling on a specimen. The specimen simples, as sheet
tape, were curved by cold folding and attached by two fixed bits.
The thermal cycle consists of two phases: heating (Joule effect) fol-
lowed by natural cooling. Meyer et al. [9] developed a thermal
fatigue test for light alloy bolts (typically aluminum alloy) of
combustion engines. This kind of bench simulates the thermal
fatigue process of such bolts. The test consists in heating the local
⇑ Corresponding author.
E-mail address: mellouli_dhouha@yahoo.fr (D. Mellouli).
zones of the simples in order to simulate the exposed parts of the
combustion room of the engine.
In this paper, we have proposed to investigate the thermal fati-
gue damage of four nodular cast irons (A, B, C and D) which have
different adding elements. We have started with studying the
effect of microstructure of ‘A’ cast iron (which was considered as
a standard cast iron). After that, we propose to examine adding ele-
ments effects as: silicon, molybdenum and nickel, on the crack
growth and oxidation of the studied samples. In the end, we have
test the effects of the thermal cycle, imposed on the same grade
cast iron specimens (‘A’ cast iron), on the thermal fatigue damage.
2. Materials
The Chemical analysis has revealed that the cast iron contains
an important silicon quantity (3–6%) which improves the oxidation
resistance of these materials [10].
The ‘D’ grade is a nodular cast iron, charged by nickel (Ni) which
is a graphitizing element, less energy than the silicon (Si). It refines
the microstructure and improves the tensile strength as well as the
heat strength and corrosion resistance in very aggressive condi-
tions [11–14].
The molybdenum (Mo) content in A and B cast irons is higher
than in C and D. Indeed, it improves the mechanical characteristics
at high temperatures and increases the temperature of allotropic
transformation [15,16].
The microstructures of the cast irons studied indicate the pres-
ence of a ferritic matrix. These micrographs show a reduction of
the grains size in B, C and D cast irons compared to A. The graphite
nodules have regular forms, the overall spheroid and their average

0261-3069/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2010.10.025
 D. Mellouli et al. / Materials and Design 32 (2011) 1508–1514 1509

size are 50 lm in the case of the A cast iron and does not exceed Y 42
the 20 lm in the case of the ‘D’ grade (Fig. 1). Z 13,3
X

3. Thermal fatigue test 60

Fig. 2 shows the geometry and dimensions of the tested speci-

men. It had a prismatic form in shape of corner and it was elabo-
rated by electro-erosion technique. This geometry was used in 6 24°
order to localize the crack initiation on the thin edge. R= 0,25
The thermal fatigue test consists in alternating heating and
Fig. 2. Geometry of thermal fatigue specimens.
cooling of the thin specimen edge. Through thermal cycle, the
specimen is free to deform, so that, only mechanical strains result-
ing from the thermal gradients are taken into account.
Heating was provided by a radiation furnace with six 1500-W
light bulbs (Fig. 3). The thin specimen edge was placed in the cen-
ter of the furnace on an adjustable holder that kept the tested
structure in position without additional mechanical loading. The
furnace was designed to focus lamp radiation on a small area of
about 6 mm in diameter.
The cooling system consists of a removable nozzle, which
moves during the thermal cycle.
When the furnace is heating, the nozzle is in a backward posi-
tion, and when the cooling sequence begins, a pneumatic actuator
moves the cooling system near the specimen (3 mm). The cooling
was carried out by forced cold air [17].
The thermal cycle imposed on the specimen takes account of
the temperature of its thin edge, as well as that of its massive part Fig. 3. Schematic configuration of thermal fatigue facility.
(Fig. 4). The thermal cycle can be summarized in the following
way:

 At first, the thin edge of the specimen was carried quickly to the
maximum temperature of 850 °C (t  90 s).
 Then, the temperature of the thin edge was maintained until
that of the beginning of the massive part reached 780 °C
(t  150 s).
 After that, a phase of forced cooling was applied to the speci-
men until the temperature of the massive part was lower than
100 °C (t  120 s). At this moment, the temperature of the thin
edge was 25 °C.

Fig. 4. Thermal cycle imposed on the specimen.

4. Results and discussion

Usually, the thermal fatigue test was led with initiation and
propagation of multiple cracks on the specimens.

4.1. Example of thermal fatigue damage of nodular cast iron: ‘A’

The results of thermal fatigue damage of ‘A’ cast iron specimen

showed the initiation and the propagation of eight cracks situated
in different positions of the thin edge of this one (Fig. 5). Initially,
six cracks (n°1, n°2, n°3, n°5, n°6 and n°7) were initiated and dur-
ing the thermal cycle, other cracks (n°4, n°8) were observed. The
propagation of these cracks was relatively fast. The cracks n°5,
n°6 and n°8 were stopped with lengths lower than those of the five
other cracks. The shielding effect between neighboring cracks that
slowed down and then stopped the propagation of the smaller
ones. For the dominant cracks, their slowing had something to do
Fig. 1. Microstructures of the various studied cast irons (etched with the nital 4%); with the decrease of the thermal gradient along the structure
OM. and by the shielding effect as observed by Haddar et al. [18,19].
1510 D. Mellouli et al. / Materials and Design 32 (2011) 1508–1514

Fig. 7. Microstructure observed on specimen (2) from exhaust manifold; OM.

Fig. 5. Evolution of the cracks length of the ‘A’ cast iron specimen during the
thermal fatigue test.

The same thermal fatigue damage test was carried out on ‘B’, ‘C’
and ‘D’ cast irons specimens, on which one, we have observed the
initiation and propagation of several cracks distributed on their
thin edges.

4.2. Effect of microstructure on thermal fatigue damage

The microstructure of the cast iron was related to addition ele-

ments and the cooling rate during the moulding process. These Fig. 8. Evolution of the main crack vs. cycles for 1 and 2 specimens.
parameters depended on the thickness of those parts. In literature,
many studies on various types of cast iron showed the influence of
these parameters on the mechanical proprieties like Zeytin et al. on the crack growth propagation was also observed by Liu et al.
[10], Jault [20], Baer et al. [21] or Choi et al. [22]. [23].
The most massive part of an exhaust manifold did not exceed
10 mm thickness, although our thermal fatigue specimens were ta-
4.3. Effect of chemical composition on thermal fatigue damage
ken in structures of which the thickness was at least of 40 mm.
A micrographic observation of two ‘A’ cast iron specimens from
Fig. 9 showed the evolution of the main cracks lengths as a func-
casting bar (1) and a new exhaust manifold (2) showed a notable
tion of the number of cycles for the two cast irons ‘A’ and ‘B’.
difference of microstructure (respectively Figs. 6 and 7).
We have observed that the initiation period was slightly earlier
The microstructure of the exhaust manifold is finer than the bar.
for the ‘B’ cast iron than the ‘A’ one. On the other hand, the cracks
The micrographs show a decrease in the number of spheroids and
growth rate as well as the number of initiated cracks was lower for
an increase in the average size of the ferrite grains on the bar.
‘B’ cast iron than the ‘A’.
Fig. 8 showed the evolution of the main cracks length of the two
It seemed that the addition of silicon in nodular cast iron did not
‘A’ cast iron specimens: 1 (bar) and 2 (exhaust manifold) as a func-
allow a reduction of the initiation period, but contributed to a
tion of the number of cycles.
reduction in the propagation rate of the cracks and the number
We have observed the earlier crack initiation for the specimen 1
of initiated cracks.
compared to the specimen 2. The initial propagation rate is more
The microstructures of the two cast irons (Fig. 1), shows that the
significant for the specimen 2 compared to the specimen 1. Thus,
dispersion of the graphite spheroids in the ferrite matrix can have a
the values of maximum propagation rates are lower for the speci-
significant role in the cracks propagation rate. Indeed, for graphite
men 2 compared to the specimen 1. The numbers of cracks ob-
of smaller size (‘B’ cast iron) the cracks propagation rate decreased
served on the 1 and 2 specimens were respectively eight and six.
A finer microstructure increased the initiation period, mini-
mized the number of initiated cracks as well as the propagation
rate and the main crack length. The Effect of the microstructure

Fig. 6. Microstructure observed on specimen (1) from bar; OM. Fig. 9. Evolution of the main crack vs. cycles for ‘A’ and ‘B’ cast irons specimens.
 D. Mellouli et al. / Materials and Design 32 (2011) 1508–1514 1511

Fig. 10. Fracture surface of the ‘A’ cast iron specimen (20–850 °C); OM.

Fig. 11. Fracture surface of the ‘B’ cast iron specimen (20–850 °C); OM.

compared to the same microstructure with big spheroids sizes (‘A’

cast iron). Graphite, as a barrier, inhibits the progression of the
cracks.
The microscopy observation of the surface of the two specimens
showed that the cracks were overall rectilinear and perpendicular
to the thin edge of the specimen (Figs. 10 and 11). Thus, the crack
developed in the perpendicular plans to the mechanical loading
and the propagation was carried out mainly in mode I.
At the end of the test, a contraction of the thin edge of the two
specimens was observed. It was due to the severity of the thermal
shock.
Fig. 12. Thickness of the oxide formed on the surface of the ‘A’ cast iron specimen
The oxidation rate of the specimen surfaces differed according (20–850 °C); MEB.
to the cast iron grade. On the ‘A’ specimen surface (Fig. 12) a thick
oxide was formed, it reached 200 lm. However a thick layer
140 lm, was observed on the ‘B’ specimen surface (Fig. 13). Indeed,
the carbon percentage compared to that in silicon, was less in the
‘B’ cast iron than in the ‘A’. In fact, it conferred more oxidation
resistance to the ‘B’ grade.
Some researchers claim that the oxidation rate can be decreased
by increasing the silicon amount from 4% to between 4.5% and 5%
[10].
Fig. 14 showed the evolution of the size of the main cracks as a
function of the number of cycles for two cast irons ‘A’ and ‘C’. The
results proved that the number of cycles necessary to detect the
initiation of the first crack is the same for the two grades. However,
Fig. 13. Thickness of the oxide formed on the surface of the ‘B’ cast iron specimen
the main crack propagation rate was much lower in the case of ‘C’ (20–850 °C); MEB.
compared to the ‘A’ grade.
Although the two microstructures were comparable (Fig. 1), the
difference of the kinetics of propagation between the two cast ir-
ons could be allotted to the differences of the thermal properties
of the two cast irons. Indeed, the thermal conductivity of the ‘C’
cast iron was higher than ‘A’ cast iron and thereafter the heat gra-
dient will be more significant in the latter. This can explain the
deceleration of the main crack growth rate in the case of cast iron
‘C’ in spite of the initiation of the same number of cracks i.e. the
presence of the same shielding effect in the two specimens [18,19].
The cracks as the two preceding specimens (‘A’ and ‘C’), are
overall rectilinear and perpendicular to the thin specimen edge
(Fig. 15). The thickness of the oxide formed was about 200 lm
(Fig. 16).
The Molybdenum (Mo) addition, in the ‘A’ cast iron, led to the
increase of the cracks growth rate and the reduction of their
lengths to saturation. The addition of Mo had a considerable influ-
Fig. 14. Evolution of the main crack vs. cycles for ‘A’ and ‘C’ cast irons specimens.
ence on the crack growth curves. But the initiation period, the
number of cracks initiated and the oxidation rate were similar
for the two cast irons grade. This result is in good accordance with Fig. 17 showed the results in which we compared the propaga-
those reported by Zeytin et al. [10]. tion rate of the main cracks between ‘D’ cast iron and ‘A’ grade.
1512 D. Mellouli et al. / Materials and Design 32 (2011) 1508–1514

Fig. 15. Fracture surface of the ‘C’ cast iron specimen (20–850 °C); OM.

Fig. 16. Thickness of the oxide formed on the surface of the ‘C’ cast iron specimen Fig. 19. Thickness of the oxide formed on the surface of the ‘D’ cast iron specimen
(20–850 °C); MEB. (20–850 °C); MEB.

Fig. 17. Evolution of the main crack vs. number of cycles for ‘A’ and ‘D’ cast irons Fig. 20. Evolution of the cracks length of the ‘A’ cast iron specimen during the
specimens. thermal fatigue test (20–730 °C).

The crack initiation period was earlier for the ‘D’ cast iron than
‘A’. The initial crack rate was however, more significant for the ‘D’
cast iron compared to ‘A’. On the other hand, the maximum prop-
agation rate was higher in the case of the ‘A’ cast iron than in the
case of the ‘D’ grade.
The thermal conductivity of the ‘D’ cast iron was lower
compared to the ‘A’ one. Thereafter, the thermal gradient was more
significant in the first case, moreover the thermal dilation coeffi-
cient of the ‘D’ cast iron was higher than of the ‘A’. The yield stress
of the ‘D’ cast iron was lower than that of the ‘A’. Thus, for the same
thermal gradient, the mechanical strain is more significant in the
first grade and even more the plastic strain which was the driving
force of the crack propagation.
After thermal cycling, the observation of the thin edge of the ‘D’
cast iron specimen shows a structural stability for this cast iron
grade (Fig. 18) as reported in previous work by Cavallini et al.
[24]. A very fine oxide was formed on the specimen surface; its
thickness does not exceed 50 lm (Fig. 19).
According to these results, we conclude that the addition of
Nickel and Chrome in the ‘D’ cast iron contributed to its oxidation
resistance and dimensional stability in thermal fatigue.
4.4. Influence of maximum temperature on the thermal fatigue
damage
A thermal fatigue test was carried out on ‘A’ cast iron specimen
following with temperature variation between 20 and 730 °C. The
cycle period was preserved and only the maximum temperature
was changed (730 °C). On this specimen, we observed the initiation
and the propagation of four cracks. Fig. 20 represents the evolution

Fig. 18. Fracture surface of the ‘D’ cast iron specimen (20–850 °C); OM.
 D. Mellouli et al. / Materials and Design 32 (2011) 1508–1514
Fig. 21. Evolution of the main crack vs. cycles for ‘A’ cast iron specimens for cycles
(20–730 °C) and (20–850 °C).
of the cracks length during thermal cycle. Initially, a first crack
(n°4) initiated and during the thermal cycling, three other cracks
(n°1, n°2, and n°3) were observed. Those cracks propagation were
slower than the main crack. The deceleration of the main crack was
explained by the reduction of the thermal gradient. This result is in
good agreement with those found by Haddar et al. [19].
In order to illustrate the influence of the maximal tempera-
ture, we reported on the Fig. 21 the evolution of the main crack
length as a function of the number of cycles for the two maximal
temperatures 730 and 850 °C. Those results have proved that
when the maximal temperature of the thermal cycle increased,
Fig. 22. Fracture surface of the ‘A’ cast iron specimen (20–730 °C); OM.
(a) Cast iron 'A', N=370 cycles
(c) Cast iron 'C', N=440 cycles
Fig. 23. The main crack growth (20–850 °C); OM.
1513
the period of crack initiation decreased. The maximal tempera-
ture seemed to have an effect on the maximal crack growth rate.
But did not seem to have an influence on the main crack length at
the saturation.
The specimen geometry was not modified at low temperature
(730 °C) compared to tested specimen at high temperature
(850 °C) (represented respectively on Figs. 22 and 10). Conse-
quently, the permanent plastic strain in the first specimen
(730 °C) was less significant. The total number of cracks observed
on the first specimen (730 °C) was less than on the second speci-
men (850 °C).
4.5. Metallographic observation
In order to reveal the cast irons microstructure, metallographic
observations was carried out in a parallel plan to the crack. Fig. 23
showed that the cracks propagation was mainly intergranular and
that it is supported by oxidation under stresses of grain
boundaries.
5. Conclusion
Anisotherme crack growth tests were carried out on prismatic
specimens in the shape of corner.
Four nodular cast irons were studied (‘A’, ‘B’, ‘C’ and ‘D’), under
thermal solicitation. The thermal fatigue damage was quantified
from crack growth curves and metallographic observation.
(b) Cast iron 'B', N=430 cycles
(d) Cast iron 'D', N=440 cycles
1514 D. Mellouli et al. / Materials and Design 32 (2011) 1508–1514
The results show an important influence of the microstructure
of the tested materials on the crack growth and thermal fatigue
damage. In fact, a finer microstructure has increased the number
of cycles to initiation and decreased the crack numbers and the
main crack growth rate.
The chemical composition of the nodular cast iron had a consid-
erable influence on the crack curves. Indeed, the addition of molyb-
denum in the ‘A’ cast iron increased the cracks growth rate while
decreasing their lengths at saturation.
The metallographic observation of the cracked specimens has
showed that the crack growth is mainly intergranular and that
it’s supported by oxidation under stresses of grain boundaries.
The increase in the maximal temperature of thermal fatigue cy-
cle has led to a significant reduction in the initiation period and an
increase in the cracks rate as well as the number of cracks.
In order to estimate the mechanical loading, it is necessary to
make a numerical study to the thermal fatigue loading. This calcu-
lation can enable us to validate the experimental results found dur-
ing this work and determine a damage model to predict crack
propagation under this kind of solicitation.
References
[1] O’Rourke R. Cast iron: the engineered metal. Adv Mater Process 2001;159:
65–8.
[2] Dawson S, Schroeder T. Practical applications for compacted graphite iron.
Trans Am Foundry Soc 2004;112:813–21.
[3] Shao S, Dawson S, Lampic M. The mechanical and physical properties of
compacted graphite iron. Materialwiss Werkstofftech 1998;29:397–411.
[4] Dawson S. Compacted graphite iron: mechanical and physical properties for
engine design. VDI-Ber 1999;1472:85–105.
[5] Höschler K, Bischof J, Koschel W. Thermo-mechanical analysis of an
automotive diesel engine exhaust manifold. Struct Integrity Soc 2002;29:
299–308.
[6] Glenny E, Northwood JE, Shaw SWK, Taylor TA. J Inst Met 1958–1959;87:
284–302.
[7] Bucher L. Etude de l’endommagement en fatigue thermique des aciers
inoxydables F17TNb et R20-12 pour application automobile. Thèse, Ecole
Nationale Supérieure des Mines de Paris; 2004.
[8] Bucher L et al. Thermal–mechanical fatigue behaviour of stainless steel for
automotive components. In: LCF 5: 5th international conference on low cycle
fatigue. Berlin (Germany); 9–11 September 2003.
[9] Meyer P, Guimbal F, Dupain A. Banc d’essai de fatigue thermique de culasses de
moteurs à combustion et procèdes associés. Organisation mondiale de la
propriété intellectuelle, WO 00/00807; 6 janvier 2000.
[10] Zeytin HK, Kubilay C, Aydin H, Ebrinc AA, Aydemir B. Effect of microstructure
on exhaust manifold cracks produced from SiMo ductile iron. J Iron Steel Res
Int 2009;16:32–6.
[11] Fatahalla N et al. C, Si and Ni as alloying elements to vary carbon equivalent of
austenitic ductile cast iron: microstructure and mechanical properties. Mater
Sci Eng A 2009;504.
[12] Zumelzu E et al. Wear and corrosion behaviour of high-chromium (14-30% Cr)
cast iron alloys. J Mater Process Technol 2002;128:250–5.
[13] Sinatora E et al. Some observations upon silicon effect on solidification of
white cast iron. In: Proceedings of the world foundry congress. Pekin (China);
1994.
[14] Goyos L, Castan~eda R, Wettinck E, Valera R, Moors M. Behaviour of nodular
iron with austempering in corrosion and wear. In: Proc IV iberam congress of
mech eng, CIDIM’99, vol. 1. Chile; 1999. p. 1–16.
[15] Tabrett CP, Sare IR, Ghomaschil MR. Int Mater Rev 1996;4:59–80.
[16] Carvalho SM, Macêdo MCS, Mello JDB, Scandian C. Mater Charact 2008.
[17] Köster A et al. Calcul d’un four à image: cas du four à lobes elliptiques. Rapport
interne, September 92, ENSMP/ARMINES/MPM; 1992.
[18] Haddar N. 2D simulation of the initiation and propagation of crack array under
thermal fatigue. ENSMP/ARMINES/MPM; 2004.
[19] Haddar N et al. Thermal fatigue crack networks: an computational study. Int J
Solids Struct 2005;42:771–88.
[20] Jault J. Fonte à graphite sphéroïdal – propriétés d’utilisation. Technique de
l’ingénieur, M4-610; 2004.
[21] Baer W, Wossidlo P, Abbasi B, Cassau M, Häcker R, Kossert R. Large scale
testing and statistical analysis of dynamic fracture toughness of ductile cast
iron. Eng Fract Mech 2009;76:1024–36.
[22] Choi JO, Kim JY, Choi CO, Kim JK, Rohatgi PK. Effect of rare earth element on
microstructure formation and mechanical properties of thin wall ductile iron
castings. Mater Sci Eng A 2004;383:323–33.
[23] Liu X, Xu J, Barbero E, Cao W, Kennedy RL. Effect of thermal treatment on the
fatigue crack propagation behavior of a new Ni-base superalloy. Mater Sci Eng
A 2008;474:30–8.
[24] Cavallini M et al. Fatigue crack propagation damaging micromechanisms in
ductile cast irons. Eng Fract Mech 2008;75:694–704.