Sei sulla pagina 1di 11

International Journal of Fatigue 24 (2002) 783793

www.elsevier.com/locate/ijfatigue

Estimation methods for fatigue properties of steels under axial and


torsional loading
K.S. Kim
a

a,*

, X. Chen b, C. Han a, H.W. Lee

Department of Mechanical Engineering, Pohang University of Science and Technology, South Korea 790784
b
School of Chemical Engineering, Tianjin University, Peoples Republic of China, 300072
c
Department of Mechanical Engineering, Pusan National University, Pusan, South Korea
Received 7 June 2001; received in revised form 6 August 2001; accepted 27 September 2001

Abstract
Uniaxial and torsional fatigue tests have been conducted on eight steels. The cyclic equivalent stress and strain amplitudes can
be fitted by the Ramberg-Osgood relationship. Fatigue lives are found correlated with the equivalent strain amplitude. Seven methods
for estimating uniaxial fatigue properties from tensile properties or hardness have been evaluated. The modified universal slopes
method by Muralidharan and Manson, the uniform material law by Baumel and Seeger and the hardness method by Roessle and
Fatemi predicted over 93% of test cases within the factor of 3 compared with observed lives. These methods are also found
applicable to torsional fatigue with fatigue properties estimated from uniaxial fatigue properties based on the equivalent strain
criterion. 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Uniaxial fatigue; Torsional fatigue; Fatigue properties

1. Introduction
The strain-based fatigue life analysis is routinely performed to assess the fatigue resistance of structural
components [1]. The characterization of fatigue endurance of engineering materials is usually made through
uniaxial fatigue tests, and fatigue properties under such
loading are available for a large body of materials. Often,
however, circumstances are encountered in service
where the fatigue resistance of a component needs to be
answered within a short timeframe but fatigue data for
the material are not available. The situation becomes
even more difficult if the loading condition is multiaxial.
Some multiaxial fatigue criteria [2,3] require torsional
fatigue data to determine all material constants. However, torsional fatigue data can be found only for a limited number of materials. Therefore, it would be desirable to have an estimation scheme for torsional fatigue
properties from readily available material properties.

* Corresponding author. Tel.: +82-54-279-2182; fax: +82-54-2795899.


E-mail address: illini@postech.ac.kr (K.S. Kim).

Over the last few decades, many researchers have


attempted to develop relations between monotonic tensile properties and uniaxial fatigue properties of engineering materials. If reliable relations with reasonable
accuracy can be established, they can serve to provide
fast solutions to fatigue problems without time and cost
involved in fatigue testing. Manson [4] first proposed
two methods; the four-point correlation method and the
universal slopes method, to estimate the strain-life curve
using monotonic tensile data. Mitchell [5] proposed a
method suitable for steels. Baumel and Seeger [6] proposed the uniform material law for metals. Muralidharan
and Manson [7] proposed a modified universal slopes
method, and Ong [8] suggested a modified four-point
correlation method. Roessle and Fatemi [9] recently proposed a method requiring only hardness and the modulus
of elasticity to estimate uniaxial fatigue properties of
steels. In a study conducted by Park and Song [10], six
methods [48] were evaluated for a total of 138 ferrous
and nonferrous metallic materials. They reported that
those proposed by Baumel and Seeger [6], Muralidharan
and Manson [7], and Ong [8] yielded better predictions
over others. Another assessment of different methods
was carried by Ong [11], in which the ASM data for 49

0142-1123/02/$ - see front matter 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 2 - 1 1 2 3 ( 0 1 ) 0 0 1 9 0 - 6

784

K.S. Kim et al. / International Journal of Fatigue 24 (2002) 783793

Nomenclature
e
s
t
g
Nf
sf
ef
b
c
tf
gf
bo
co
K
n
K0
n0
s
e
E
G
n
ef
su
sy
RA
EL
HB

Strain range in axial fatigue test


Stress range in axial fatigue test
Shear stress range in torsional fatigue test
Shear strain range in torsional fatigue test
Cycles to failure
Axial fatigue strength coefficient
Axial fatigue ductility coefficient
Axial fatigue strength exponent
Axial fatigue ductility exponent
Shear fatigue strength coefficient
Shear fatigue ductility coefficient
Shear fatigue strength exponent
Shear fatigue ductility exponent
Cyclic strain hardening coefficient
Cyclic strain hardening exponent
Cyclic shear strain hardening coefficient
Cyclic shear strain hardening exponent
Equivalent stress
Equivalent strain
Youngs modulus
Shear modulus
Poisson ratio
True fracture ductility
Ultimate tensile strength
Yield strength (0.2%)
Reduction in area (%)
Elongation
Brinell hardness

steels were used with the modified and original versions


of the four-point correlation method [4,8] and the universal slopes method [4,7] and Mitchells method [5]. It was
concluded in [11] that the modified four-point correlation method and the modified universal slopes method
gave a satisfactory agreement between predicted and
experimental fatigue lives.
In this paper, the results of strain controlled axial
fatigue tests and torsional fatigue tests on eight steels
are presented. Seven aforementioned methods are evaluated for estimating uniaxial fatigue properties from tensile properties or hardness. Also, the approaches based
on the equivalent strain criterion and the maximum shear
strain criterion will be investigated for estimating torsional fatigue properties from uniaxial fatigue properties.

2. Experiment
The test materials used in this investigation are eight
steels purchased in the form of wrought bars. The chemi-

cal compositions of the materials are given in Table 1.


The monotonic tensile properties obtained from solid
specimens with a diameter of 6 mm are listed in Table 2.
The geometry of the fatigue specimen is shown in
Fig. 1. The specimen was gun-drilled and honed through
the center. The gage section of the outside contour was
machined, ground and polished with alumina powder
(0.3m). The effect of residual stresses that could have
been introduced in the machining process has been ignored
in this study.
Fatigue tests were conducted on a servo-hydraulic
MTS axial-torsional materials testing system. All tests
were carried out under strain control. A tension-torsion
extensometer with an axial gage length of 20 mm and
a diameter of 12.5 mm was used to control the strain.
Triangular waveforms were used for both axial and torsional fatigue. Tests were carried out under fully
reversed strain cycling with frequencies in the range of
0.51Hz. The lower frequency was applied to higher
amplitude tests. Failure was defined as a drop of 10%
in load for axial tests, and a drop of 10% in torque for

K.S. Kim et al. / International Journal of Fatigue 24 (2002) 783793

785

Table 1
Chemical composition of test materials (wt.%)
Material

Si

Mn

Cu

Ni

Cr

Mo

SNCM630
SNCM439
SCM440
SCM435
SFNCM85S
SF60
S45C
S25C

0.32
0.39
0.42
0.38
0.2
0.43
0.43
0.27

0.25
0.24
0.22
0.21
0.25
0.18
0.18
0.24

0.45
0.72
0.71
0.7
0.8
0.69
0.69
0.53

0.013
0.09
0.01
0.015
0.017
0.023
0.023
0.019

0.017
0.018
0.01
0.018
0.008
0.007
0.007
0.002

0.18
0.13
0.13
0.11

2.52
1.65
0.08
0.09
0.49

2.54
0.67
1.01
0.83
0.55

0.51
0.16
0.22
0.16
0.19

Table 2
Mechanical properties of test materials.
Material

E (GPa)

G (GPa)

sy (MPa)

ty (MPa)

su (MPa)

EL (%)

RA (%)

HB

SNCM630
SNCM439
SCM440
SCM435
SFNCM85S
SF60
S45C
S25C

196
208
204
210
201
208
206
209

77
80
80
81
80
79
79
80

0.273
0.296
0.283
0.3
0.26
0.311
0.298
0.29

951
950
846
795
565
580
590
280

581
560
440
460
340
274
341
182

1100
1050
1000
951
825
820
798
508

19
13
13
18
21
19
17
19

49
37
36
66
66
53
39
52

327
323
319
300
241
167
234
153

Fig. 1. Specimen geometry (unit: mm).

torsional tests. The stress amplitude was measured in a


preset interval of cycles. The stress amplitude at approximately half-life, where the stress-strain response was
stable, was used for the cyclic stress-strain curve. At
least seven tests were conducted for each type of axial
and torsional loading at strain amplitudes ranging from
0.2% to 2%.

e sf
(2Nf)bef(2Nf)c for uniaxial fatigue,
2 E

(1)

g tf
(2Nf)b0gf(2Nf)c0 for torsional fatigue,
2 G

(2)

The material constants sf, ef, b and c are the uniaxial


fatigue properties, and tf, gf, bo, and co are the torsional
fatigue properties. The strain amplitudes of Eq. (1) and
Eq. (2) can be split into elastic and plastic components,
and they can be individually related to life by equating
to the first and second terms, respectively, on the right.
An outline is given in the following section for estimating uniaxial fatigue properties that are examined in
this study. The equations in the original papers have
been rewritten in the nomenclature of the present paper.
The details how these equations were obtained can be
found not only in the original paper for each method but
also in review papers [10,11]. It is also noted that the
true fracture ductility ef in some of the equations may
be obtained from efln[100/(100RA)].

3. Methods for estimating uniaxial fatigue


properties

3.1. Four-point correlation method

The relationship between the applied strain amplitude


and fatigue life under uniaxial loading and torsional
loading can be expressed by Basquin-Coffin-Manson
equations:

Manson [4] proposed the four-point correlation


method to estimate the strain-life curve using monotonic
tensile properties. The four points here include two
points on the elastic strain-life curve and two points on

786

K.S. Kim et al. / International Journal of Fatigue 24 (2002) 783793

the plastic strain-life curve. The fatigue properties are


related to monotonic tensile properties as follows:

E
sf 10b log 2+ log
2

2.5su(1+ef)
E

(3a)

1
1
ef 10c log20+ log
2

1 3/4
e
4 f

1
0.0132e 1
1
c log
log e3/4
,
3
1.91
3
4 f

efef,

log

su
,
E

4)+ log

2.5su(1+ef)
E

(6b)
(6c)
(6d)

(3b)

where ee is the elastic strain range at 104 cycles and


is given by

(3c)

ee sf 2log[0.16(s /E)0.81] log(s /E)


u
f
],
[103
2 E

(3d)

where sf sf.

where e is the elastic strain range at 104 cycles and


is estimated by
e10b log(410

0.81

1
0.00737ee/2
c log
logef,
4
2.074

2.5(1+ef)
0.9
,
b
log[1/(4105)]
log

1
su
b log 0.16
6
E

(3e)

3.5. Modified universal slopes method


The universal slopes method [4] was modified by
Muralidharan and Manson [7], and the following equations were proposed:
sfE0.623

3.2. Universal slopes method


An alternative approach was also proposed by Manson
[4], in which it was assumed that the slopes of the elastic
strain-life and plastic strain-life curves do not vary with
materials. The fatigue properties are estimated by
sf1.9018su,

(4a)

b0.12,

(4b)

ef0.7579e0.6
f ,

(4c)

c0.6.

(4d)

(6e)

su
E

b0.09,
ef0.0196e0.155
f

0.832

su
E

(7a)
(7b)

0.53

c0.56.

(7c)
(7d)

3.6. Uniform material law

Mitchell [5] suggested the following equations for


steels with hardness below 500 HB:

This method, proposed by Baumel and Seeger [9],


may be considered as a universal slopes method. However, it assigns different slopes to the unalloyed and lowalloy steels, and to aluminium and titanium alloys. Only
the elastic modulus and tensile strength of the material
are needed for estimation of fatigue properties. For unalloyed and low-alloy steels, the equations are given by

sfsu345(MPa),

sf1.5su,

(8a)

b0.087,

(8b)

ef0.59y,

(8c)

3.3. Mitchells method

(5a)

2(su+345)
1
,
b log
6
su

(5b)

efef,

(5c)

c0.58.

c0.6.

(5d)

where
y1 for

3.4. Modified four-point correlation method


Ong [8] proposed a modified four-point correlation
method. The estimation equations are given as
sfsu(1ef),

(6a)

su
0.003,
E

su
su
y1.375125.0 for 0.003.
E
E

(8e)
(8f)

Notice that b and c are very close to those of the modified universal slopes method.

K.S. Kim et al. / International Journal of Fatigue 24 (2002) 783793

e s s

2 2E 2K

3.7. Hardness method


Roessle and Fatemi [9] proposed a simple method to
estimate fatigue properties from Brinnell hardness and
elastic modulus for steels with hardness between 150 HB
and 700 HB. This method will be referred to as the
hardness method in this paper. This method has an
advantage in that it only requires hardness and the modulus of elasticity, whereas all other methods described use
tensile properties. Hardness can be obtained from nondestructive tests, and the modulus of elasticity is essentially
constant for each class of materials. This makes the hardness method particularly attractive where tensile data are
either not available or destructive tests cannot be done.
The exponents b and c are the same as those of the
modified universal slopes method, i.e. 0.09 and 0.56,
respectively. The remaining properties are
sf4.25(HB)225 (MPa),

(9a)

1
ef [0.32(HB)2487(HB)191000].
E

(9b)

787

1/n

g t
t

2 2G 2K0

for uniaxial loading,

(10)

1/n0

for torsional loading,

(11)

The cyclic strain hardening coefficients (K,K0) and


exponents (n,n0) obtained by fitting the stress-strain
data are given in Table 4. The cyclic hardening coefficients and exponents can be estimated from their
relations to fatigue properties; n=b/c, K=sf/en
f ,
0
n0=b0/c0, K0=tf/gn
f . The results are also given in Table
4. The agreement with direct regression is found to be
good only for four materials (SCM435, SFNCM85S,
SF60, S25C). Substantial differences are found for other
materials (SNCM630, SNCM439, SCM440, S45C). The
stress amplitudes computed from two sets of K and n
for the latter four materials showed small differences at
low strain amplitudes, but up to 20% differences were
observed at high strain amplitudes. This discrepancy
would be attributed to the scatter in fatigue data, which
will influence the values of fatigue coefficients and
exponents, and subsequently cyclic hardening coefficients and exponents. Thus, one might say that estimating cyclic stress-strain curves based on fatigue properties
could lead to considerable error in certain situations.
The equivalent stress (s) and strain (e) in this study
are defined by

4. Results and discussion


4.1. Results of fatigue tests
The fatigue properties of the materials determined
from test results are listed in Table 3. All materials exhibited cyclic softening except S25C, which hardened cyclically. The failure crack size was in the range of 5 mm
to 15 mm for most tests. The crack path, under visual
inspection, was normal to the specimen axis (mostly
stage II growth) in uniaxial fatigue, and in torsional
fatigue the crack grew in the axial direction (stage I
growth) until it branched off at the onset of stage II
growth before the test was stopped. The midlife hysteresis loops from fatigue tests were used to determine the
stable cyclic stress-strain properties. The cyclic stressstrain data can be fitted by the Ramberg-Osgood
relationship:

s(3/2sij sij )1/2,

(12)

e(2/3e e ) (2/3e e ) ,
e e 1/2
ij ij

p p 1/2
ij ij

(13)

where sij , e , and e are the deviatoric stress (sij =sij


skkdij /3), elastic deviatoric strain (eeij =eeij eekkdij /3), plastic strain, respectively, and repeated indices imply summation
over
1-3.
It
is
obtained
that
s=s, e=2(1+n)ee/3+ep for uniaxial loading (ee: elastic
strain, ep: plastic strain), and s=t3, e=g/3 for torsional
loading. Fig. 2 gives the equivalent stress-strain data for
all the materials. The solid lines in these figures represent the stress-strain curves fitted by the RambergOsgood relationship.
It should be mentioned that the shear stress in torsion
e
ij

p
ij

Table 3
Fatigue properties of test materials obtained under uniaxial loading and torsional loading.
Material

sf (MPa)

ef

tf (MPa)

gf

b0

c0

SNCM630
SNCM439
SCM440
SCM435
SFNCM85S
SF60
S45C
S25C

1270
1380
1400
1100
1040
978
1400
821

1.54
1.89
0.675
0.996
0.316
0.187
0.449
0.216

0.0732
0.0722
0.0879
0.067
0.0924
0.082
0.107
0.0961

0.823
0.801
0.650
0.708
0.522
0.439
0.564
0.458

858
969
754
512
533
504
630
426

1.51
3.68
0.315
0.360
0.251
0.286
1.22
0.249

0.0606
0.0855
0.0814
0.0454
0.071
0.0668
0.0802
0.0741

0.706
0.765
0.54
0.519
0.406
0.417
0.564
0.376

788

K.S. Kim et al. / International Journal of Fatigue 24 (2002) 783793

Table 4
Cyclic stable stress-strain properties of test materials

Material

Regression
K

K0

n0

Estimation from fatigue properties


K
n
K0

n0

SNCM630
SNCM439
SCM440
SCM435
SFNCM85S
SF60
S45C
S25C

1,060
1,000
1,040
1,070
1,320
1,350
1,150
1,140

0.054
0.066
0.094
0.089
0.180
0.186
0.152
0.210

592
601
643
553
676
609
552
565

0.050
0.072
0.108
0.085
0.173
0.156
0.119
0.199

1,220
1,300
1,480
1,100
1,280
1,340
1,640
1,140

0.086
0.112
0.151
0.088
0.175
0.160
0.142
0.197

0.089
0.090
0.135
0.095
0.177
0.187
0.190
0.210

828
837
898
560
679
616
612
560

are all based on the measured strain, and the error in


stress has no effect on the predicted lives.
The uniaxial and torsional fatigue lives were correlated with equivalent strain amplitudes in Fig. 3. Good
correlations were found for all materials.
4.2. Uniaxial fatigue properties and life prediction
Life prediction was performed with the fatigue properties estimated by seven afore-mentioned methods on the

Fig. 2.

The equivalent cyclic stress-strain curves of test materials.

was computed with the assumption of uniform distribution through the wall thickness of the specimen. This
assumption may lead to inaccuracy in the stress (9.3%
underestimation if purely elastic). However, in the presence of substantial plasticity, which is the case for most
test conditions, the error becomes considerably smaller.
It is also noted that the fatigue criteria used in this study

Fig. 3. The equivalent strain amplitudelife relationship of test


materials.

K.S. Kim et al. / International Journal of Fatigue 24 (2002) 783793

789

Fig. 4. Comparison of fatigue lives for uniaxial fatigueexperimental and predicted by the four-point correlation method.

Fig. 6. Comparison of fatigue lives for uniaxial fatigueexperimental and predicted by Mitchells method.

eight materials tested under uniaxial loading. Fig. 4


through 10 give the results for all the methods. The fourpoint correlation method by Manson [4] overestimated
life up to the factor of 10 (Fig. 4), and not many points
fell within the bound of factor 3. The universal slopes
method of Manson [4] provided good statistics with 91%
of predicted lives within the factor of 3 (Fig. 5). Mitchells method [5] yielded non-conservative lives for the

majority of test cases. Only 66% of predicted lives based


on this method were within the factor of 3 (Fig. 6). The
modified four-point correlation method by Ong [8] gave
acceptable predictions for short lives, but overly nonconservative predictions were obtained for lives greater
than 104 reversals (Fig. 7). As much as 24% of predicted
lives were outside of the factor 10 bound for this method.
The modified universal slopes method by Muralidharan

Fig. 5. Comparison of fatigue lives for uniaxial fatigueexperimental and predicted by the universal slopes method.

Fig. 7. Comparison of fatigue lives for uniaxial fatigueexperimental and predicted by the modified four-point method.

790

K.S. Kim et al. / International Journal of Fatigue 24 (2002) 783793

Fig. 8. Comparison of fatigue lives for uniaxial fatigueexperimental and predicted by the modified universal slopes method.

and Manson [7] (Fig. 8), the uniform material law by


Baumel and Seeger [6] (Fig. 9) and the hardness method
by Roessle and Fatemi [9] (Fig. 10) gave better predictions than other methods with at least 93% of predicted
lives within the factor 3 bound. Notice that two of these
methods [6,7] are the ones selected for best statistics in
the study by Park and Song [10]. The hardness method
appears to have somewhat larger scatter. The scatter in
the three methods mainly comes from SNCM630 steel.

Fig. 10. Comparison of fatigue lives for uniaxial fatigueexperimental and predicted by the hardness method.

This material had higher contents of nickel and chromium, and highest strength of all test materials. Without
SNCM630 data points, the statistics can be improved
remarkably. A notable feature is that the fatigue ductility
exponent, c, and the fatigue strength exponent, b, are
almost the same in the three methods. Thus, the estimates of b=0.09 and c=0.56 are deemed reasonable
for steels. However, sf and ef may vary somewhat
depending on the crack size used in the failure criterion
[1214]. This implies that the performance of individual
methods could vary, perhaps to a minor extent, with the
failure criterion used in the baseline fatigue data. With
the failure criterion of 10% drop in load or torque used
in this study, which produced 5 mm to 15 mm cracks,
the modified universal slopes method appears to provide
the best correlation.
4.3. Torsional fatigue properties and life prediction

Fig. 9. Comparison of fatigue lives for uniaxial fatigueexperimental and predicted by the uniform material law.

It has not received much attention how to estimate


torsional fatigue properties from readily available
material properties. Almost all investigations in this area
are concerned with estimating only uniaxial fatigue
properties. This is attributable to the fact that the availability of torsional fatigue data is limited to a relatively
small number of materials. From the perspective of
multiaxial fatigue analysis, however, the development of
a reliable estimation scheme for torsional fatigue properties is highly desirable. This is because the fatigue criteria suitable for shear fracture often require torsional
fatigue test data [2,3], while test facilities for such loading are not available as widely as the uniaxial testing

K.S. Kim et al. / International Journal of Fatigue 24 (2002) 783793

systems. The utility of such an estimation scheme would


be wide considering the fact that the majority of steels
for structural applications fail in shear mode.
The observation of crack paths under torsional loading
indicated that the materials under study are all shear
fracture materials. The torsional fatigue lives were correlated well with shear strain amplitudes, as given by Eq.
(2). However, when the uniaxial and torsional fatigue
data were plotted together, better consolidation of the
data points was found with the equivalent strain amplitude (Fig. 3), than with the maximum shear strain amplitude for all materials except SCM440. Fig. 3 shows that
the equivalent strain amplitude provides satisfactory correlation for SCM440, but the maximum shear strain
amplitude gave somewhat better results. It may be also
worth noting that the effect on fatigue life of normal
strain (or stress) on the maximum shear strain plane that
is emphasized in widely accepted shear strain criteria
[2,3] was negligibly small for SCM440 [15].
Since the modified universal slopes method [7], the
uniform material law [6] and the hardness method [9]
excelled other methods in uniaxial fatigue, they are used
to evaluate torsional fatigue data in this section. The
fatigue ductility and strength exponents do not change
significantly between axial and torsional fatigue as Leese
and Morrow [12] observed for 1045 steel. Fatemi and
Kurath [16]s data on In718 and 1045HR show essentially similar trends, although some deviation is found
in fatigue ductility exponents for In718. The regressed
data for the exponents obtained in this study for eight
materials show a trend that both b and c are almost consistently bigger in magnitudes for uniaxial fatigue than
for torsional fatigue (Table 3). These properties are usually sensitive to the scatter of data points, and the
addition of data points could alter the trend. Our experience is that the application of the exponents for uniaxial
fatigue to torsion with a proper adjustment of fatigue
coefficients does not alter predicted lives to a significant
extent. Consequently, the torsional exponents were set
identical to the uniaxial counterparts. Since the equivalent strain criterion appears more suitable for the combined axial and torsional life data than the maximum
shear strain criterion, the following estimates were
adopted (see Appendix A):
tfsf/3,

(14a)

gfef3,

(14b)

b0b,

(14c)

c0c.

(14d)

Using these equations, the shear fatigue properties


were obtained from estimated uniaxial fatigue properties,
and the torsional fatigue lives of test materials were predicted. It was found that all three methods provide

791

Fig. 11. Comparison of fatigue lives for torsional fatigueexperimental and predicted by the modified universal slopes method.

reasonably good results. The uniform material law [6]


(Fig. 12) and the hardness method [9] (Fig. 13) yielded
better statistics than the modified universal slopes
method [7], which gave slightly larger scatter (Fig. 11).
All three methods did not perform as superbly as in axial
fatigue, but at least 85% of the predicted lives were still
within the factor 3 bound.
Leese and Morrow [12] tried approximations
based on the maximum shear strain criterion

Fig. 12. Comparison of fatigue lives for torsional fatigueexperimental and predicted by the uniform material law.

792

K.S. Kim et al. / International Journal of Fatigue 24 (2002) 783793

equivalent stress and strain data can be fitted by the


Ramberg-Osgood relationship.
2. Seven methods for estimating uniaxial fatigue properties from readily available mechanical properties were
evaluated. Three methodsthe modified universal
slopes method by Muralidharan and Manson, the uniform materials law by Baumel and Seeger, and the
hardness method by Roessle and Fatemihave been
found to provide good life predictions with at least
93% within the factor 3 bound compared with experimental lives.
3. Shear fatigue properties estimated from uniaxial
fatigue properties based on the equivalent strain criterion provided satisfactory life estimates for the
majority of specimens tested under torsion, when the
uniaxial fatigue properties were obtained by the three
methods described in 2. At least 85% of predictions
fell within the factor 3 bound compared with experimental lives.
Fig. 13. Comparison of fatigue lives for torsional fatigueexperimental and predicted by the hardness method.

Acknowledgements

(tf=sf/2, gf=1.5ef, b0=b, c0=c; see Appendix A) as


well as the equivalent strain criterion, Eqs. (14a, 14b,
14c and 14d). They used experimental data for uniaxial
fatigue properties. Their results show that both
approaches predict the torsional fatigue curve of 1045
steel closely. The maximum shear strain approach was
also tried in this study. It furnished, on the whole, more
conservative lives compared with the equivalent strain
appraoch. The correlation was found better for SCM440,
almost all data points of which fell within the bound of
factor 2 for all three methods of estimation. The overall
statistics were, however, not as good as the equivalent
strain approach: the percentage of data points within the
factor 3 bound ranged 53% to 71% for the three
methods, and the bound of factor 5 included 84%-86%
of data points.

The elastic and plastic parts of the Basquin-CoffinManson equation, Eq. (1), for uniaxial fatigue can be
written as

5. Conclusions

ee sf
(2Nf)b,
2 E

(A3)

The available estimation methods for uniaxial fatigue


properties from tensile properties or hardness were
evaluated in this study. The best methods were then
extended to estimating torsional fatigue properties. All
test materials used in this study can be regarded as ductile steels, as judged from their ductility. Therefore, the
results obtained in this study may be valid only for ductile steels. The conclusions of this study are as follows:

ep
ef(2Nf)c,
2

(A4)

1. Fatigue tests were conducted on eight steels under


axial and torsional loading. The life data correlated
well with equivalent strain amplitudes. The cyclic

The fatigue data reported in this study were obtained


in a research program of Pohang University of Science
and Technology funded by Pohang Steel Company.
Appendix A

ee sf
(2Nf)b,
2
E

(A1)

ep
ef(2Nf)c.
2

(A2)

In terms of the equivalent strain defined by Eq. (13), Eq.


(A1) and Eq. (A2) can be written as

3E
.
where E=
2(1+n)
The addition of Eq. (A3) and Eq. (A4) provides the
equivalent strain criterion.
For torsional fatigue, the maximum shear strain criterion given by Eq. (2) can be split into
ge tf
(2Nf)b0,
2 G

(A5)

K.S. Kim et al. / International Journal of Fatigue 24 (2002) 783793

gp
gf(2Nf)c0.
2

(A6)
[3]

In terms of the effective strain amplitude Eq. (A5) and


Eq. (A6) can be written as
ee tf
(2Nf)b0,

2 3G

(A7)

ep gf
(2Nf)c0.
2 3

(A8)

[4]
[5]
[6]
[7]

[8]

In the equivalent strain approach for estimating shear


fatigue properties, Eq. (14a) through Eq. (14d) are
obtained by setting equal the right sides of Eq. (A3) and
Eq. (A7), and the right sides of Eq. (A4) to Eq. (A8).
For the maximum shear strain approach, Eq. (A5) and
Eq. (A6) for uniaxial fatigue can be written as
tf
ee

(2N )b0,
2 (1+n)G f
ep gf
(2N )c0.
2 1.5 f

(A9)

[9]

[10]

[11]

[12]

(A10)

By comparing Eq. (A9) with Eq. (A1), and Eq. (A10)


with Eq. (A2), the relation of shear fatigue properties to
uniaxial fatigue properties can be obtained
(tf=sf/2, gf=1.5ef, b0=b, c0=c).

References
[1] Morrow J, Socie DF. The evolution of fatigue crack initiation
life prediction methods. In: Sherratt F, Sturgeon JB, editors.
Materials, experimentation and design in fatigue. Westbury
House (England): Warwick, 1981:3.
[2] Kandil FA, Brown MW, Miller KJ. Biaxial low-cycle fatigue

[13]

[14]

[15]

[16]

793

fracture of 316 stainless steel at elevated temperatures. Book 280,


London: The Metals Society, 1982:203-210.
Fatemi A, Socie DF. A critical plane approach to multiaxial
fatigue damage including out-of-phase loading. Fatigue Fract Eng
Mater Struct 1988;11:14965.
Manson SS. A complex subject-some simple approximations.
Experimental Mechanics 1965;5:193226.
Mitchel MR. Fatigue and microstructure. American Society for
Metals, Metals Pack (OH), 1979:385
Baumel A Jr., Seeger T. Materials data for cyclic loading, Supplement I. Amsterdam: Elsevier Science Publishers, 1990.
Muralidharan U, Manson SS. Modified universal slopes equation
for estimation of fatigue characteristics. ASME Trans J Engng
Mater and Tech 1988;110:558.
Ong JH. An improved technique for the prediction of axial fatigue
life from tensile data. International Journal of Fatigue
1993;15:2139.
Roessle ML, Fatemi A. Strain-controlled fatigue properties of
steels and some simple approaximations. International Journal of
Fatigue 2000;22:495511.
Park J, Song J. Detailed evaluation of methods for estimation of
fatigue properties.
International Journal of
Fatigue
1995;17:36573.
Ong JH. An evaluation of existing methods for the prediction of
axial fatigue life from tensile data. International Journal of
Fatigue 1993;15:1329.
Leese GE, Morrow J. Low cycle fatigue properties of a 1045
steel in torsion. In: Miller KJ, Brown MW, editors. Multiaxial
fatigue. ASTM STP853, Philadephia: American Society for Testing and Materials, 1985:482-496.
Dowling NE. Cyclic stress-strain and plastic deformation aspects
of fatigue crack growth. ASTM STP637, Philadephia: American
Society for Testing and Materials, 1977:97-121
Socie DF, Waill LA, Dittmer D. Biaxial fatigue of Inconel 718
including mean stress effect. Multiaxial fatigue, ASTM STP 853,
Philadephia: American Society for Testing and Materials,
1985:463-481.
Han C, Nam KM, Kim KS, Cho IL. Multiaxial fatigue life prediction under irregular loading. Proceedings of the Asian Pacific
Conference on Fracture and Strength 01, Oct. 20-22, 2001
Sendai (Japan);1:400405.
Fatemi A, Kurath P. Multiaxial fatigue life predictions under the
influence of mean-stresses. Journal of Engineering Materials and
Technology. Transactions of ASME 1988;110:3808.

All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.

Potrebbero piacerti anche