Fatigue b e h a v i o u r and m e a n

e f f e c t s in g r e y c a s t iron
J. Fash and D. F. Socie

Strain cycle fatigue concepts are well integrated into fatigue life prediction methodology for
wrought components. Concepts developed for wrought materials cannot be directly extended
to cast materials because of differences in the fatigue mechanisms, but the framework of a life
prediction method for cast iron components will be similar. Observations and results of
constant-amplitude completely reversed fatigue tests performed in strain and load control are
reported for a pearlitic grey cast iron. Mean amplitude tests in both control modes have been
conducted to evaluate the effects of mean stresses and strains. A parameter of the form
(Omax A e) gives good correlation of all the fatigue tests performed and provides a simple
relationship to fatigue life.

Key words: fatigue life; fatigue tests; constant amplitude + completely reversed; grey cast iron;
strain control + load control; fatigue-life prediction

Design methodology to evaluate the fatigue resistance of

components manufactured from wrought materials sub-
jected to cyclic loads are in daily use by design engineers.
Strain cycle fatigue and the local strain approach 1,2 are
used extensively for predicting the crack initiation lifetime
of components. Several variables such as material selection,
component geometry and loading conditions, among others,
must be understood for a successful design procedure. To S
be a useful design tool, the analytical techniques and experi-
mentation required to complete the design procedure must
be easily implemented. In the local strain approach, five
major divisions can be made to describe the basic units of
the fatigue analysis (Fig. 1). The following five areas must,
in general, be considered. They are not totally independent \ ./
and are by no means complete. In some situations, other
factors will complicate the analysis.
i) Laboratory characterization of the fatigue behaviour
of the material of interest provides the foundation of
the analysis. Typically, completely reversed, strain
cycle fatigue tests are performed on smooth, axial Damage Model
tension/compression specimens. The cyclic stress/
strain curve is determined as representative of the
material's behaviour under cyclic loading. A relation- Component Design
ship between cyclic strain amplitude and the number Procedure
of cycles to failure is also determined. A Coffin-
Manson type relationship is often fitted to this data. Fig. 1 Basic steps for applying local strain concepts to evaluate the
fatigue performance of wrought components
ii) Geometric discontinuities such as holes, fillet radii,
and notches often initiate failure. To extend the
laboratory materials characterization to the actual can be incorporated into the fatigue analysis is
component, a stress/strain simulation in the critical required. Several methods are available for reducing
location where failure occurs is required. Both the load sequence into damaging events. Of these
material behaviour and the local geometry must be methods, rainflow counting has been shown to give
considered. Several stress analysis techniques are better life estimates. Typically, the damaging events
available to determine the local response to the are determined as closed hysteresis loops which have
applied loading conditions. Finite element analysis a strain range with a mean stress possibly present. To
and Neuber notch analysis are techniques frequently determine these values the cyclic response of the
employed. material must be modelled on a reversal-by-reversal
iii) Actual loading conditions vary considerably from the basis through the complex history of either stress,
completely reversed, constant amplitude loading strain, or load values.
applied in the laboratory. A technique to reduce a iv) Mean stresses, which are frequently present and can
complex load history into a series of events which be determined by the material response model,

0142-1123/82/030137-06 $03.00© 1982 Butterworth & Co (Publishers) Ltd INT. J. FATIGUE July 1982 137
 influence the fatigue life. F o r a given strain range in a
strain controlled fatigue test, increasing the mean
stress causes a decrease in the fatigue life. Several ~ ..,,,<\~ ;J iv. /,. I
parameters which characterize the effect of mean
stresses on fatigue life have been proposed. Most of
.:" ~.l'.,,'.'~-,(/'~,- ,,. - .~y~
these are empirical relationships which best fit experi-
mental data. The effects of mean strains are usually
t (.~: , .j~: i,- .... ~ L' ..-", -~
ignored.
v) A damage model is necessary to accumulate the
degrading effects of the closed hysteresis loops or
damaging events. When the amount of damage
accumulated equals some critical value, failure is "Z '.' .... .,, "",-T:h ,
' ~ -,2',."x'-- ,-"-~1~% ~"". ~. . . . . ' /
predicted. Miner's linear damage rule is widely used
in crack initiation analysis. For particular situations, t .---. t t" t~ill /. _l;-.~\J' ,~,..\ ~ <
variations o f Miner's rule have been suggested.
Damage models such as those based on continuum
damage concepts are also available. This is usually
i

the last step of the analysis and it includes all the

previously mentioned interactions.
When the general areas described above are incorporated
into a fatigue life analysis, the crack initiation life of a
component can be estimated. Similar analytical techniques
are not widely accepted for life predictions of cast iron
components in which the effects of metallurgical notches
such as free graphite, porosity and inclusions must be con-
sidered. The phenomena occurring during the fatigue of
cast irons must be understood in order to develop the units
of a life prediction technique which models the actual
rraterial behaviour.
To develop fatigue design methodology for cast iron
components, the fatigue behaviour of grey iron has been
observed during research conducted in the Materials
-,;., ,
Engineering Research Laboratory at the University of
Illinois at Urbana-Champaign. Observations made during
constant-amplitude completely-reversed fatigue tests
performed in strain and load control have been reported t,~, ~ - ,rt
previously 3'4 and will be reviewed below. Results from Fig. 2 M m r o s t r u c t u r e o f p e a r h t i c grey iron investigated ( t o p - Type
fatigue tests with mean values of the control parameter A ~ 60%, T y p e D ~ 40%, size 4; b o t t o m - 2% N I T A L )
are reported. A previously proposed parameter for
incorporating mean stress effects, is shown to give good
correlation of the results of the completely reversed and Table 2. Mechanical properties
mean amplitude tests for a pearlitic grey iron. Use of this
parameter will satisfy section (iv) of the fatigue analysis Property Pearlitic iron
procedure described above.
Modulus of elasticity 84/108 (GPa)
E (tension/compression)
MA TERIA L
Yield strength, 0.2%Sy 185 (MPa)
Grey cast iron test bars 30 mm in diameter and 200 mm Ultimate strength 228 (MPa)
long were cast in green sand moulds to give a pearlitic (true fracture strength, of)
microstructure in the matrix steel. Chemical composition is True fracture ductility (of) 0.0122
given in Table 1. Classification of the size and distribution Bulk hardness 180 BHN
of the free graphite structure is given by ASTM standard (3000 kg)
A2475 as approximately 60% type A, and approximately Eutectic cell size 0.14 (mm)
40% type D, size 4 (Fig. 2). Steads reagent was used to ASTM A247 flake designation ~60% Type A,
preferentially etch metallographic specimens to determine ~40% Type D
the eutectic cell size, which is approximately 0.14 ram. Size 4
Mechanical and monotonic properties are given in Table 2.

Tables 1. Chemical c o m p o s i t i o n

C Si Mn S P Ti Ni Cr Mo Cu Sn

Pearlitic iron 3.30 2.20 0.44 0.02 0.01 0.01 0.06 0.03 0.01 0.40 -

138 INT. J. FATIGUE July 1982

EXPERIMENTAL PROG RA MME
o 100
Low cycle fatigue specimens (Fig. 3) were machined from
the as-cast test bars. 6 All fatigue tests were performed on
100 kN, closed loop, servo-hydraulic materials testing equip- N=I
ment interfaced to a digital computer which in some tests J N = 1024
recorded data and in other tests performed test control as o.ool N 4og6
well as data acquisition. Load/strain response versus applied
cycles was monitored. no
Deformation and crack initiation processes in cast
irons have previously been reported in the literature. 7-9
Cracking occurs first in the free graphite structure. This
structure is very brittle and weak in tension but will carry
much larger compressive loads. Observations during fatigue
loadinglO, 11 have shown that cracks grow from the graphite
structure which causes large stress concentrations in the : 2o°o
matrix steel. This crack growth behaviour occurs very early
in the life of the fatigue specimens.
Observations of the initiation and growth of cracks - _
in the smooth specimens have been accomplished using a
surface replication technique. 3'4 Standard metallographic Strain

acetyl film, which softens when moistened with acetone, is Fig. 4 Hysteresis response during completely-reversed strain-control
used. When it is soft, the tape is placed around the test fatigue lest
section and it conforms to the surface topography. The
tape re-hardens and is then removed, giving a permanent A A_~ =0003 Nf = 813
2
replica of the surface. By taking a series of these replicas
180 ~ £ ~ _ ~ . ~ ~_ = 0.001 , NI = 1 05,106
during a fatigue test, the development and growth of cracks
can be observed. Results of these observations will be
included below. 120 \'~ Mox,mum stress

RESULTS A N D DISCUSSION
13_
~ 60
Strain control
Strain controlled, completely reversed fatigue tests were
performed. Tests were run to approximately 90% load drop
from the initial maximum load. Cyclic stress/strain response
0
was observed during the tests but the typical stable
hysteresis response found in wrought metals was not
observed. A series of hysteresis loops from a strain-control - 3 0 , ~
test are shown in Fig. 4. A continual decrease in the l 1 II,lllJ 1 I I i lllll , illlllll r i IIILlll I I ,111111 l .lllllll

maximum tensile load occurs but the ma~-num compres- 10 10 2 tO s 10 4 I0' 10 ~

sive load remains constant throughout the test. This
degradation in the load carrying capacity of the specimen is
attributed to the formation and development of multiple
crack systems. Development of the crack system causing
failure was observed from the surface replicas taken through-
out the test and it was found S that the crack system develops
slowly to a size of about 1 to 2 mm. Once a crack reaches
d,ameter
~
--I0
Apphed cycles, N
Fig. 5 Transient behaviour of the maximum stress and mean stress
during strain-controlled fatigue tests
this size in the smooth specimen, failure occurs very quickly
through a complex process of crack linking and crack
growth. During completely-reversed strain-control tests, a
compressive mean stress is present due to the anisotropy
between tensile and compressive properties. This mean
stress decreases as the strain amplitude increases. The value
of the compressive mean stress and the maximum stress
decrease throughout the test as shown in Fig. 5 for two
test conditions. Strain/life results are shown in Fig. 6 as well
as the results of mean strain tests which will be discussed
later.

Load control
i Gauge length Constant amplitude load-control tests were performed. As
in the strain-controltests,cyclic stabilitywas not observed.
~" 125 -
In these tests, the cyclic strain amplitude continually
increased (Fig. 7) due to the increased compliance o f the
specimen as crack systems developed. Maximum values of
Fig. 3 Low cycle fatigue specimen both the tensile and compressive strains increase but the

INT. J. FATIGUE July 1982 139

 pretation of the effects of mean stresses and simplifies the
Meon 5troln
:0;'[_ fatigue life analysis. A simple relationship has been proposed
o o0
• O. 0 0 2
by Smith, Watson & Topper. 12 This parameter (called the
• 0 0015 SWT parameter) was originally proposed by the authors to
E: 0 001 include crack growth and seems most appropriate for grey
O0 '~ O. 0 0 5
iron. Other parameters based on the reduction of static
• - O. 0 0 5
0@0
A -0. 001
or fatigue strengths are inappropriate for fatigue problems
that have been demonstrated to be crack growth dominated.
I O O O ~, z~
Several forms of the SWT parameter are possible and
QLI~] Q OOO •
a form which relates the product of the strain amplitude
~m D ~' O0 0 and the maximum stress to the fatigue life for a given test
P condition will be used to correlate experimental results.
The parameter is then of the form

(Omax A £/2)

P where A¢/2 = 1/2 total strain range and Omax = stress at

maximum strain.
A series of strain-control fatigue tests were conducted

i0.,.[
10 I IIIIIIII
10 ~
I J.Jllllil
103
I 1111111]
104
I I lllllll
10 s
I Illlllll
10 c:
in which a variety of strain amplitudes and mean strain
combinations were imposed. These tests resulted in a wide
•C y c l e s t o fo u r e Nf

Fig. 6 Cyclic life of strain-controlled fatigue tests ~) 004~

A A~. _ 1 ; ' f ) M P : : , #,'+ : "3"~+0
j
•I r') A" 7:J - if)r,,t,,IP,'.,. +vt - 2 2 0 5 9
N = 1032

500 Q O©3- _)
=

N : 3032

. . . . • c
_ z
0 001 ~_ 0 f:,02. / 2
g _ /
C_
u3 /

' I J .-4 ~ _ 0 _ o_~_(~__o_o~o_(>_4.,~ 7.< , .-c ,,~

•$ 0 0 1 " ~
"-- Meo n stro,r'
/
Ao" : 120 M P o
2
Nf = 3310 I I I lilLiz+L~.l IIItUJ_ I l_llltll i I II.[:II . I11~II'1
iC 102 +03 10 ~ ~C>5
ApDned c y c l e s ,'^~'

Fig. 8 T r a n s i e n t behaviour of the strain amplitude and mean strain

during load-control fatigue tests

StrOlP
200 7 .....................................
Fig. 7 Hysteresisresponseduring completely-reversed toad-control
fatigue tests 1801- Me(in stress
o 0

160- c
• 30
A 5C
tensile value increases much more than the compressive
1 4 0 -- OR
value. Crack initiation and growth behaviour similar to that
observed in the strain-control tests was observed in the load
1 2 0 -- • ooo
control tests. Multiple crack systems were found very early
in the fatigue life and were observed to grow slowly to C o o
6 100 - A •
about 1 to 2 mm. 4 Once a crack system reached this size, <
oO
the crack became dominant and grew rapidly to instability 8o F
and failure. In the completely-reversed load-control tests, ~A

+f
the mean strain and strain amplitude both increased with
number of cycles (Fig. 8). Stress versus life results are
40
shown in Fig. 9. Mean stress data are also shown in this
figure and will be discussed in the next section. 2O
I
I I I llitlJ+ I I IIJLLLL__I l t l t i J X [ . . . I I Iilllll l. JiJJllli
Mean parameter 10:+ 1C 3 104 105 106
CyCleS t o f o l l u r e , Nf
Ideally mean parameters collate all the results of fatigue
tests into a single relationship which provides an inter- Fig. 9 Cyclic life of load-controlled fatigue tests

140 INT. J. FATIGUE July 1982

range of mean stresses and the results are included in Fig. 6.
o 0 Straits-control tests
Transient behaviour, similar to that shown for the completely- o
reversed tests in Fig. 5 was observed. In order to apply the Load-control tests

SWT parameter a value o f the maximum stress must be

chosen at some point in the life. F o r convenience, the value
at the half life of the specimen was used. The value of the
maximum stress could be chosen at another point in the
life but this would not change the results significantly since
the change is small (see Fig. 5). Using the SWT parameter °

both the completely-reversed and mean strain-control data ×10 - -

can be collapsed to a single scatter band as shown in Fig. 10.

o
Similar mean tests were conducted in load-control.
During these tests, the magnitude of the strain range
increased throughout the test and the value at the half life
was used in the SWT parameter. The data in Fig. 9 are
collapsed into a single scatter band in Fig. 11 using this
parameter.
Implementing this form of the SWT parameter allows
10 I llllJIll I.llllllll 1 illlllll I I IIHJI] I IIIIIIll
a single graphical representation of both strain-control and 10' 10 2 1 0 :~ 10 a ~0 "~ 10 6
Cycles to foflure,Nf

Fig. 12 Results of all fatigue tests presented using theSWT parameter

~I--- o o Zero meon stroln
• Mean stroln present load-control fatigue tests. F o r the approximately 60 test
points shown in Fig 12, a power law relationship can be
b
• 0~0
•,.o •

• o
fitted which takes the form
(OmaxAe/2) -- 1.82 (Nf) 0.25
~o o The constants in this relationship were determined b y a
first order linear regression analysis.
2~1~ o. _.
0o 0
E SUMMA R Y
D

k In the finite life regime (less than 10 6 cycles to failure)

the fatigue resistance of grey cast iron is dominated b y
crack growth. The SWT parameter provides an effective
means for relating a wide range of loading conditions to the
cyclic life and can easily be incorporated into a fatigue life
prediction methodology. Further work must be done
10'i I llllllll 1 I IIIIIllI. I IIIIIIII. I I I I I I I l l _.J.. I I I I I I l l
~0 10 2 10 ~ 1 0 '~ 1 0 "~ 10 ~
before this parameter can be extended into life ranges
Cycles to farlure,Nf greater than 10 6 cycles where the mechanisms may differ.
Fig. 10 Results of strain-control fatigue tests presented using the
SWT paramater
A CKNOW LED GEMENTS
1k O Zero mean stress
Mr Furman and Ms Cleary are thanked for their assistance
in conducting the tests. Tests were performed in the Materials
L • Mean stress present
Engineering Research Laboratory at the University of Illinois
F O0
at Urbana-Champaign. Financial support was provided by
F
the Fracture Control Program and ~he test specimens were
i 000 cast by Deere & Co.
P.

lo-,!
E
0 O0
~o
oo
REFERENCES
1. Manual on L o w Cycle Fatigue Testing, ASTM STP 465
(American Society for Testing and Materials, Philadelphia, PA,

°i
1969)
2, Fatigue under Complex Loading (Society of Automotive
Engineers, Warrendale, PA, 1979)
L--
3. Fash, J. W., Socie, D. F. and Russell, E. S. 'Fatigue crack
I
initiation and growth in gray cast iron' Proc of Fatigue "81,
(Society of Environmental Engineers, Fatigue Group Con-
ference, Warwick University, England, 24--27 March 1981)
1 0 2I_ I .J- I I Illll I I IJIHll 1 lllJIIl[ I IILIHII I IlllIlll pp 40--51
10' 10; 10 { 10 4 10 5 10 e
4. Molinaro, L. 'Fatigue behavior and crack development in
Cycles to fQflure, Nf compacted graphite cast iron' Fracture Control Program,
Fig. 11 Results of load-control fatigue tests presented using the Report No 39 (College of Engineering, University of Illinois
SWT parameter (additional tests are currently in progress) at Urbana-Champaign, May 1981)

I N T . J. F A T I G U E J u l y 1982 141
 5. Method of Evaluating the Microstructure of Graphite in Iron
Castings ASTM Designation: A247, Plate I (Graphite Form
Types), Plate II (Graphite Flake Type Chart), and Plate III
(Graphite Size), (American Society for Testing and Materials,
1916 Race Street, Philadelphia, PA 19103)
6. Constant Amplitude Low-Cycle Fatigue Testing ASTM
Designation: E606, (American Society for Testing and
Materials, Philadelphia, PA 19103)
7. Gilbert, G. N. J. 'An evaluation of the stress/strain properties
of flake graphite cast iron in tension and compression' J o f
the British Cast Iron Research Association 7 (1959) pp
7 4 5 - 789
8. Gilbert, G. N. J. 'The stress/strain properties of nodular cast
irons in tension and compression' J o f the British Cast Iron
Research Association 12 No 2 (1964) pp 1 7 0 - 1 9 3
9. Gilbert, G. N. J. 'Factors relating to the stress/strain proper-
ties of cast iron' J o f the British Cast Iron Research Associa-
tion 6 No 11 (1957) pp 5 4 6 - 5 8 8
10. Ikawa, K. and Ohira, G. 'Fatigue properties of cast iron in
relation to graphite structure' Cast Metals Research J 3
(American Foundry Society, 1967) pp 11-21
142 I N T . J. F A T I G U E July 1982
11. Starkey, M. S. and Irving, P. E. 'The influence of m~crostru(;
ture on fatigue crack initiation in spheroidal graphite cast
iron' Proc of Int Symp on L o w Cycle Fatigue Strength and
Elasto-Plastic Behavior of Materials (Stuttgart, West Germany,
1979)
12. Smith, K. N., Watson, P. and Topper, T. H. 'A stress-strain
function for the fatigue of metals' J o f M a t e r , JMLSA 5 No 4
(December 1970) pp 7 6 7 - 7 7 8
AUTHORS
The authors are with the Design Division of Illinois
University's Department of Mechanical and Industrial
Engineering. Inquiries should be addressed to: Dr J. Fash,
Design Division, Department of Mechanical and Industrial
Engineering, University of Illinois at Urbana-Champaign,
Urbana, IL 61801, USA.