Vous êtes sur la page 1sur 15

ORIGINAL CONTRIBUTION doi: 10.1111/ffe.

12366

An improved method for estimation of Ramberg–Osgood curves of


steels from monotonic tensile properties

J. LI1,2, Z. ZHANG2 and C. LI2


1
Harbin Air Force Flight Academy, Nan Tong Street, Harbin 150001, China, 2The Science Institute, Air Force Engineering University, East Chang Le
Road, Xi’an 710051, China
Received Date: 6 August 2015; Accepted Date: 16 October 2015; Published Online: 10 November 2015

A B S T R A C T In the present study, first a method for estimating cyclic yield strength is improved and com-
pared using the experimental data of 121 steels from literature. Correlations between cyclic
deformation properties (K′ and n′) and monotonic tensile data are then investigated, and a
simple method requiring only the monotonic tensile properties is developed for estimation
of the Ramberg–Osgood curve. Prediction capability of the proposed method is not only eval-
uated using the aforementioned 121 steels, but also compared with several commonly used
methods that are available in the literature. The proposed correlations are shown to predict
the cyclic deformation properties of most of considered steels reasonable well.

Keywords cyclic yield strength; estimation method; monotonic tensile properties;


Ramberg–Osgood curve; steel.

LIST OF SYMBOLS a, p = Scale parameter and shape parameter in log-logistic distribution function,
respectively
b = Fatigue strength exponent
c = Fatigue ductility exponent
E = Young’s modulus
I = The rank number
I = The total number of considered steels
K′ = Cyclic strength coefficient
N = Monotonic strain hardening exponent
n′ = Cyclic strain hardening exponent
2Nf = Number of reversals to crack initiation
Pf = Cumulative probability
RA = Percent reduction in area
R2 = The square of the multiple correlation coefficient
Sf = True fracture strength
Su = Ultimate tensile strength
Sy = Monotonic yield strength
S′y = Cyclic yield strength
V = Coefficient of variation
α = True fracture ductility parameter
β = Adjusted coefficient
χ = Parameter which is varied with the Su/E ratio in MFPCM
Δε, Δεe, Δεp = Total, elastic and plastic strain ranges, respectively
Δε*e = The estimated elastic strain range at 104 reversals
Δσ = Axial cyclic stress range
εf = True fracture ductility
ε′f = Fatigue ductility coefficient
σ′f = Fatigue strength coefficient

Correspondence: J. Li. E-mail: lijing02010303@163.com

412 © 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
AN IMPROVED METHOD FOR ESTIMATION OF RAMBERG–OSGOOD CURVES OF STEELS 413

INTRODUCTION basis of the four Manson–Coffin’s parameters (Eq. (3))


are fairly good, despite somewhat significant scatter in
Mechanical properties determined from monotonic tensile the experimental data. That is to say, for steels, in the
tests are usually used in the design of components and absence of K′ and n′, the two material parameters can
structures. However, the cyclic loading process can signif- be estimated by using the four Manson–Coffin’s fatigue
icantly alter the stress–strain behaviour of a material. In parameters with accuracy and reliability. Unfortunately,
service, most structures experience cyclic loading, thus it should be noted that the four Manson–Coffin’s param-
detailed knowledge of the cyclic deformation behaviour eters are also fatigue performance parameters, and the
of a material is essential for reliable life estimations of fatigue test data are necessary to obtain reliable results.
cyclically loaded structures. Usually, fatigue properties Although some correlations among the four fatigue
are obtained by performing fatigue tests on companion parameters and the available monotonic tensile proper-
specimens of materials. However, experimental determina- ties have been proposed and verified,1 when they are used
tion of cyclic deformation properties requires significantly to theoretically calculate the cyclic strength coefficient
more time and cost, as compared to determination of and cyclic strain hardening exponent, the results are not
monotonic tensile properties from a simple tension test. satisfactory. This conclusion will also be verified in the
So, over the years, many researchers have attempted to following narrative. In addition, for the cyclic strength
develop correlations among the monotonic tensile data and coefficient and cyclic strain hardening exponent, direct
fatigue properties of materials.1–5 If reliable correlations correlations between the two cyclic deformation parame-
with reasonable accuracy can be established, durability ters and the available monotonic properties are limited in
performance predictions and/or optimization analyses can the literature. Systematical verification to the limited
be performed, while substantially reducing time and cost correlations is also necessary for initial engineering
associated with material fatigue testing. design studies.
As for life prediction, the classical approach to predict Recently, Li et al.4 proposed an equation to estimate
crack initiation of nucleation life is based on the strain the cyclic yield strength (S′y) of steels from two mono-
amplitude, Δε/2, referred to as the ε–N approach,6,7 tonic tensile properties, i.e. the ultimate tensile strength
which can be represented as Manson–Coffin equation (Su) and the percent reduction in area (RA). However,
′ verifications by Lopez9 show that this equation always
Δε Δεe Δεp σ f  b  c
¼ þ ¼ 2N f þ ε′f 2N f (1) underestimates the cyclic yield strength of steels when
2 2 2 E the tested cyclic yield strength is larger than 900 MPa.
where Δε, Δεe and Δεp are the total, elastic and plastic Therefore, an improved equation should be proposed to
strain ranges, respectively. σ′f and b are the fatigue estimate the cyclic yield strength better. In the present
strength coefficient and exponent, respectively. ε′f and study, the expression in Ref.4 is verified and modified
c are the fatigue ductility coefficient and exponent, first, and then correlations among cyclic strength coeffi-
respectively. E is the Young’s modulus, and 2Nf is the num- cient, cyclic strain hardening exponent and monotonic
ber of reversals to crack initiation. The classical ε–N ap- tensile properties are developed for 67 steels from
proach works with real stresses and strains, and uses the literature.9,10 Several correlations that are available in
Ramberg–Osgood description for the Δσ–Δε hysteresis the literature are also examined and compared to the
loops. The Ramberg–Osgood relationship is as follows8 proposed ones using a different set of data from 54 steels
from SAE J1099 Technical Report11 and Refs.1,12–16
 1 This data set is chosen to provide an unbiased basis of
Δε Δεe Δεp Δσ Δσ n′
¼ þ ¼ þ (2) comparison among the methods and to verify the
2 2 2 2E 2K ′
predictive accuracy of the proposed correlations.
where Δσ is the axial cyclic stress range. K′ is the cyclic
strength coefficient, and n′ is the cyclic strain hardening ex-
ponent. Assuming the Ramberg–Osgood’s elastic and plas- IMPROVED CORRELATION AMONG CYCLIC
tic strain ranges perfectly correlate with the correspondent YIELD STRENGTH AND MONOTONIC
Manson–Coffin’s ranges, then from Eqs. (1) and (2), the fol- TENSILE PROPERTIES
lowing correlations can be obtained1
According to SAE J1099 Technical Report,11 the cyclic
b σ′f yield strength is defined as the stress to cause 0.2%
n′ ¼ and K ′ ¼ n′ : (3)
c ε′f inelastic strain as measured on a cyclic stress–strain curve.
It is usually determined by constructing a line parallel to
By using the fatigue test data of 724 steels, Meggiolaro the slope of the cyclic stress–strain curve at zero stress
and Castro4 have verified that K′ and n′ estimated on the through 0.2% strain. If the Ramberg–Osgood curve

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
414 J. LI et al.

(Eq. (2)) is used to describe the cyclic stress–strain APPROXIMATIONS OF THE RAMBERG–OSGOOD
relation, the cyclic yield strength may be determined by CURVE
substituting a value of 0.2% for the plastic strain ampli-
tude in Eq. (2) as follows As mentioned above, the stress where the line paralleling
to the slope of the cyclic stress–strain curve at zero stress

S′y ¼ K ′ ð0:002Þn : (4) through 0.2% strain intercepts the cyclic stress–strain
curve is taken as the cyclic yield strength. It is expected
The cyclic yield strength can be used to determine the that the cyclic yield strength calculated from Eq. (4) will
stress amplitude at which significant cyclic plastic defor- be reasonable well. Therefore, the cyclic yield strength,
mation starts to occur during cyclic loading. Therefore, if successfully predicted from monotonic properties, can
the cyclic yield strength, if successfully predicted from be used to estimate the cyclic strength coefficient
monotonic properties, can serve as an important property and/or cyclic strain hardening exponent. As mentioned
for fatigue design. In Ref.,4 Li et al. proposed the follow- in the previous section, verification from 121 steels shows
ing expression from 27 alloy steels to relate cyclic yield that Eq. (6) gives reasonable estimation for cyclic yield
strength (S′y) to ultimate tensile strength (Su) and percent strength. Combining Eq. (4) with Eq. (6), the following
reduction in area (RA). expressions can be obtained.
 0:16 "  0:216 #
0:002 0:002
S′y ¼ ð1 þ RAÞSu  (5) ′
K ¼ 500 n′
0:089ð1 þ RAÞ 1:35
Su 1:35
 þ 120
lnð1  RAÞ lnð1  RAÞ
(7)
Tested data from Ref.9,10 listed in Tables 1 and 2
are used to verify the cyclic yield strength predicted by   0:216 
 
Eq. (5). In Fig. 1, a plot of the tested cyclic yield strength lg K ′  lg 0:089ð1 þ RAÞ1:35 Su 1:35  lnð0:002
1RAÞ þ 120
versus the predicted ones is provided. It can be seen from n′ ¼
lg 500
this figure, Eq. (5) always underestimates the cyclic yield
strength when the experimental cyclic yield strength is (8)
larger than 900 MPa. However, a correlation between
tested S′y and predicted ones from Eq. (5) can be Equations (7) and (8) show that K′ (or n′) can be
obtained from the non-linear least squares fit showing estimated if the data of n′ (or K′) is available. Using the
good agreement with the data. The non-linear expression experimental data of aforementioned 121 steels,
is as follows (unit: MPa) predicted K′ versus tested K′ and predicted n′ versus
 0:216 tested n′ are plotted in Figs. 3 and 4, respectively. It can
0:002 be seen from these figures, both the predicted K′ and n′
S′y ¼ 0:089ð1 þ RAÞ1:35 Su 1:35  (6)
lnð1  RAÞ are within the scatter band of factor ±20%. Although
 
þ120 R2 ¼ 0:961 : these results are encouraging for estimating the two
cyclic deformation parameters, it is important to keep in
mind that the tested K′ or the tested n′ is required. In
Figure 2 shows the comparison of the predicted cyclic the absence of both tested K′ and n′, Equations (7) and
yield strength (Eq. (6)) versus the experimental ones (8) will be inability. At this condition, correlations among
taken from Ref..1,9–16 It can be seen from Fig. 2, in com- K′ (or n′) and monotonic tensile properties should be de-
parison with the experimental data from Ref.,9,10 all the S veloped to obtain the two cyclic deformation parameters.
′y calculated by the modified correlation (Eq. (6)) are In a plot of K′ versus Su, a relatively clear distinction
within a scatter band of factor ±20%. Good predictions among steels of different Su/Sy material group exists on
based on the modified correlation are expected because the basis of the tested data of 67 steels.9,10 This plot is
this correlation is obtained from fittings of the data from provided in Fig. 5. In this figure, the cyclic strength coef-
Ref.9,10 in this study. Besides, experimental data from ficient data of the Su/Sy ≥ 1.4 group is greater than the
SAE J1099 Technical Report11 and Refs.1,12–16 (see cyclic strength coefficient data of the Su/Sy ≤ 1.2 group
Tables 3 and 4) which are not used for deriving the at most of ultimate tensile strength. Cyclic strength coef-
proposed correlation (Eq. (6)) are also used to evaluate ficient data of the intermediate group (1.2 < Su/Sy < 1.4)
the predictability of the proposed method. Figure 2 shows is generally in between these two groups. It is interesting
that most of these considered data1,11–16 are also within to be found that this distinction is similar to the distinc-
the scatter band of factor ±20%. In other words, in the tion of cyclic hardening/softening behaviour for metallic
absence of experimental data, the modified method can materials, where Su/Sy ≥ 1.4 the alloy behaves in a cyclic
predict the cyclic yield strength of steels reasonable well. hardening manner, while Su/Sy ≤ 1.2 the alloy behaves in

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
AN IMPROVED METHOD FOR ESTIMATION OF RAMBERG–OSGOOD CURVES OF STEELS 415

Table 1 Monotonic and cyclic deformation properties of selected steels from the AISI Bar Steel Fatigue Database9

SAE spec. E (GPa) Su (MPa) Sy (MPa) RA (%) S′y (MPa) K′ (MPa) n′ Su/Sy
1020 207 502 295 64 317 1171 0.210 1.70
1038 201 582 331 54 342 1340 0.220 1.76
1541 197 783 471 55 421 1400 0.193 1.66
1038 219 652 359 53 358 1420 0.222 1.82
1141Nb 220 695 418 53 405 1448 0.205 1.66
1050M 211 821 450 50 413 1987 0.253 1.82
1151V 206 761 452 51 456 1473 0.189 1.68
10V45 213 765 465 48 436 1642 0.213 1.65
1141V 214 725 450 49 447 1467 0.191 1.61
1045 209 747 510 62 440 1302 0.174 1.46
1141AL 216 771 457 57 424 1515 0.205 1.69
1141V 220 789 494 47 481 1441 0.177 1.60
5150 210 867 472 56 497 1485 0.176 1.84
1090 219 1090 729 14 535 1877 0.202 1.50
15V41 193 1071 666 33 712 1575 0.128 1.61
10V45 216 909 606 42 616 1691 0.162 1.50
1090 214 1124 760 38 602 1964 0.190 1.48
1090M 212 1251 751 14 637 1943 0.179 1.67
4620 208 998 688 58 603 1824 0.178 1.45
1090 213 1147 678 22 618 2163 0.202 1.69
8620 212 991 694 54 601 1872 0.183 1.43
8620 210 1145 796 51 705 2078 0.174 1.44
4320 202 994 920 63 652 799 0.032 1.08
9310 197 902 804 71 616 1035 0.083 1.12
1141AL 227 925 814 59 593 1260 0.121 1.14
41B17M 213 872 827 68 611 1029 0.084 1.05
15B35 219 940 866 65 645 1094 0.085 1.09
4620 211 964 892 62 667 1213 0.096 1.08
8822 212 946 884 67 644 1154 0.094 1.07
5140 218 1039 957 53 702 1329 0.103 1.09
10B21 212 1105 1062 71 756 1109 0.062 1.04
4140 207 1043 929 61 670 1460 0.125 1.12
86B20 205 1034 989 64 705 1268 0.094 1.05
4140 201 1514 1363 48 911 1614 0.092 1.11
8822 208 1723 1528 50 1095 2055 0.101 1.13
4140 200 1401 1306 48 852 1306 0.069 1.07
4140 204 1537 1330 42 895 1591 0.093 1.16
4140 203 1248 1158 48 856 1696 0.110 1.08
4140 207 1240 1167 47 851 1617 0.103 1.06
4130AL 213 1483 1285 44 915 2151 0.138 1.15
51B60 200 1970 1830 22 1272 2490 0.108 1.08
9254 206 2020 1841 51 1381 2316 0.083 1.10
9254AL 205 2450 2270 4 1922 3322 0.088 1.08
1022 200 604 457 69 348 1180 0.197 1.32
C-70 201 964 765 17 572 1528 0.158 1.26
1141Nb 217 802 602 54 481 1254 0.154 1.33
15V24 207 878 646 61 615 1453 0.138 1.36
1038 218 743 560 69 450 1242 0.163 1.33
1141V 215 797 610 58 487 1280 0.154 1.31
5120 214 1008 780 58 628 1784 0.168 1.29
20MnCr5 194 960 695 51 567 1930 0.197 1.38
20MnCr5 194 1053 852 57 613 2062 0.195 1.24
9310 195 1201 990 57 798 2098 0.156 1.21
1090M 217 1388 1065 25 730 2005 0.163 1.30
8620 214 1311 990 52 927 2137 0.134 1.32
20MnCr5 194 1337 1071 52 976 2231 0.133 1.25
86B20 206 1502 1198 60 1075 2118 0.109 1.25
1050M 203 829 440 34 523 1292 0.146 1.88
1050M 203 2360 2000 15 1796 3538 0.109 1.18

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
416 J. LI et al.

Table 2 Monotonic and cyclic deformation properties of selected steels10

Materials E (GPa) Su (MPa) Sy (MPa) RA (%) S′y (MPa) K′ (MPa) n′ Su/Sy


SNCM630 196 1100 951 49 758 1060 0.054 1.16
SNCM439 208 1050 950 37 664 1000 0.066 1.11
SCM440 204 1000 846 36 580 1040 0.094 1.18
SCM435 210 951 795 66 615 1070 0.089 1.20
SFNCM85S 201 825 565 66 431 1320 0.180 1.46
SF60 208 820 580 53 425 1350 0.186 1.41
S45C 206 798 590 39 447 1150 0.152 1.35
S25C 209 508 280 52 309 1140 0.210 1.81

For Su/Sy ≥ 1.4


 
K ′ ¼ 1:21Su þ 555 R2 ¼ 0:712 : (9)

For 1.2 < Su/Sy < 1.4


 2 
K ′ ¼ 3:63104 ðSu Þ2 þ 0:68Su þ 570 R ¼ 0:813 : (10)

For Su/Sy ≤ 1.2


 
K ′ ¼ 2:16104 ðSu Þ2:1 þ 738 R2 ¼ 0:921 : (11)

Based on these correlations, in the absence of K′ and n′,


Fig. 1 Experimental S′y versus predicted S′y by Eq. (5).
the cyclic stress–strain curves can be approximated for
steels. In Figs. 6 and 7, the K′ and n′ estimated by Eqs. (8)
through (11) are shown using the monotonic tensile proper-
ties of steels from Refs..9,10 It can be seen from the two
figures, most of the estimated K′ and n′ are with a factor
of ±20% scatter band. That is to say, in the absence of K′
and n′, the develop equations are reasonable reliability to es-
timate these two parameters for steel.

EVALUATION OF AVAILABLE METHODS FOR


ESTIMATING K′ AND N′ FROM MONOTONIC
TENSILE PROPERTIES

It can be seen from Eq. (3) that the Ramberg–Osgood’s


parameters (K′ and n′) can be estimated from the
Manson–Coffin’s parameters (ε′f, σ′f, b and c). It should
be kept in mind that Eq. (3) is derived on the basis of
Fig. 2 Experimental S′y versus predicted S′y by Eq. (6) using the
the assumption that the Ramberg–Osgood’s elastic
tested data from Refs.1,9–16.
and plastic strain ranges perfectly correlate with the
correspondent Manson–Coffin’s ranges (compatibility
a cyclic softening manner. When Su/Sy is between 1.2 assumption). However, the incompatibility may exist
and 1.4, alloy behaves either hardening or softening because of the fact that different pieces of data are
manners. Within the three material groups, a relatively used to determine the Manson–Coffin’s parameters and
good agreement exists between the data of steels and Ramberg–Osgood’s parameters. Extensive verifications
the linear and/or non-linear least squares fit (see Fig. 5). are necessary to determine whether the compatibility as-
The correlations are given by (units: MPa) sumption is reasonable or not. In Ref.,4 the correlations

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
AN IMPROVED METHOD FOR ESTIMATION OF RAMBERG–OSGOOD CURVES OF STEELS 417

Table 3 Monotonic and cyclic deformation properties of selected steels from SAE J109911

SAE spec. E (GPa) Su (MPa) Sy (MPa) RA (%) S′y (MPa) K′ (MPa) n′ Su/Sy
1008 207 331 234 77.5 200 1443 0.318 1.41
1008 203 365 255 77.9 204 1234 0.290 1.43
1010 203 331 200 80.4 190 867 0.244 1.66
1015 207 414 228 68 252 945 0.213 1.82
1020 203 441 262 61.8 267 1962 0.321 1.68
1020 186 393 255 64 208 1233 0.286 1.54
1025 207 547 306 62.8 288 1042 0.207 1.79
1045 – 752 517 44 397 1022 0.152 1.45
1045 207 1827 1689 51 1369 3371 0.145 1.08
1045 207 2241 1862 41 1872 3947 0.120 1.20
1045 – 718 424 48 375 1401 0.212 1.69
1045 – 942 620 39 540 1770 0.191 1.52
1045 – 1322 787 21 741 2066 0.165 1.68
1045 – 1516 865 6 837 3048 0.208 1.75
1045 – 2297 1636 18 1934 3083 0.075 1.40
1045 207 1956 1729 38.3 1269 3366 0.157 1.13
1045 207 1344 1275 59 917 1751 0.104 1.05
10B21 197 1048 999 67.6 679 1089 0.076 1.05
15B27 203 847 772 69 577 903 0.072 1.10
15B27 203 916 854 66.5 644 1026 0.075 1.07
4130 200 1427 1358 54.7 835 1696 0.114 1.05
4130 221 896 778 67.3 603 1264 0.119 1.15
4140 200 2033 1895 20 1193 1974 0.081 1.07
4142 200 1551 1447 47 955 1756 0.098 1.07
4142 200 1929 1860 37 1105 1910 0.088 1.04
4142 207 1413 1378 48 1049 2266 0.124 1.03
4142 207 1757 1584 42 1156 1997 0.088 1.11
4142 207 1929 1722 35 1338 2399 0.094 1.12
4340 200 1468 1371 38.1 863 1996 0.135 1.07
4340 193 827 634 43.4 471 1337 0.168 1.30
4340 207 1171 1102 56 675 1037 0.069 1.06
5160 203 1584 1488 39.7 1062 1964 0.099 1.06
0030 207 496 303 46 317 738 0.136 1.64
0050A 209 787 415 19 403 1165 0.171 1.90
0050A 209 583 402 26 373 896 0.141 1.45
8630 207 1144 985 29 704 1502 0.122 1.16

Table 4 Monotonic and cyclic deformation properties of selected steels1,12–16

Materials E (GPa) Su (MPa) Sy (MPa) RA (%) S′y (MPa) K′ (MPa) n′ Su/Sy Refs.
API5DS-135 200 1175 1033 60 800 1910 0.14 1.14 1
API5LGr.B 208 423 294 60 277 1229 0.24 1.44 1
API5LX-60 198 533 457 46 409 813 0.12 1.17 1
SAE1020 205 491 285 54 270 941 0.18 1.72 1
SAE4340 200 1250 1060 36 700 1890 0.16 1.18 1
SAR60 205 620 540 40 500 1122 0.13 1.15 1
30CrMnSiNi2A 200 1655 1308 52.3 1180 2648 0.130 1.27 12
30CrMnSiA 203 1177 1105 53.6 805 1772 0.127 1.07 12
GC-4 200 1875 1513 43.7 1429 3411.4 0.14 1.24 12
1045HR 205 621 380 50 345 1258 0.208 1.63 13
347S 200 588 251 52 292 1329 0.244 2.34 14
S460N 208 643 500 26 410 1115 0.161 1.29 14
SAE4320 201 1705 1345 55.5 1220 2636 0.124 1.27 15
SAE86201 180 1683 1420 36.2 1308 2881 0.127 1.19 15
SAE86202 203 1764 1357 13.9 1252 2811 0.130 1.30 15
SAE86203 207 1677 1402 14.2 1266 2886 0.133 1.20 15
1141AF 200 875 524 40.2 564 1205 0.122 1.67 16
1141QT 212 777 670 61.8 493 1133 0.134 1.16 16

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
418 J. LI et al.

Fig. 3 Experimental K′ versus predicted K′ by Eq. (7) using the Fig. 6 Experimental K′ versus predicted K′ by Eqs. (9–11) using the
tested data from Refs.1,9–16. tested data from Refs.9,10.

Fig. 7 Experimental n′ versus predicted n′ by Eqs. (8–11) using the


Fig. 4 Experimental n′ versus predicted n′ by Eq. (8) using the tested tested data from Refs.9,10.
data from Refs.1,9–16.
presented in Eq. (3) are evaluated using 724 steels to
verify the coherence between Manson–Coffin’s and
Ramberg–Osgood’s elastic and plastic strain ranges. After
the statistical evaluation of the considered parameters
using the log-logistic distribution, Meggiolaro and
Castro4 found that there is a reasonable (but not exact)
correlation between the cyclic strain hardening exponent,
n′, and the b/c ratio, with a coefficient of variation of
V = 15%. Here the coefficient of variation, V, is defined
as the ratio between the standard deviation and the mean.
The cyclic strength coefficient, K′, estimate based on n′
and on Manson–Coffin’s parameters is also fairly good
for steels, despite the somewhat significant scatter in the
experimental data, with V = 15% as well.4 The similar
results are also obtained by Nieslony et al..17 In the pres-
ent study, in order to verify the reasonability of the com-
Fig. 5 Cyclic strength coefficient versus ultimate tensile strength patibility assumption for steels, the Ramberg–Osgood’s
(data from Refs.9,10). elastic and plastic strain ranges versus the correspondent

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
AN IMPROVED METHOD FOR ESTIMATION OF RAMBERG–OSGOOD CURVES OF STEELS 419

Manson–Coffin’s ranges are plotted in Fig. 8 based on compatibility equation (Eq. (3)) are not recommended
the tested data of the 59 AISI steels.9 In Fig. 8, the for the other kind of materials such as the aluminium
subscript MC and RO represent the Manson–Coffin’s and alloys because the compatibility cannot be guaranteed
Ramberg–Osgood’s strain ranges, respectively. That is (see the verifications in Refs.4,17). Based on the indirect
method, the four fatigue parameters are predicted by
2σ ′f  b Δσ seven existing estimation methods, and then the K′ and
ðΔεe ÞMC ¼ 2N f and ðΔεe ÞRO ¼ ; (12)
E E n′ are calculated using the compatibility equation. These
estimation methods listed in Table 5 are available in liter-
 1 ature.1,2,18–22 For each method, plots of predicted versus
   c   Δσ n′
Δεp ¼ 2ε′f 2N f and Δεp ¼2 : (13) experimental cyclic strength coefficient and predicted
MC RO
2K ′ versus experimental cyclic strain hardening exponent
are made, using the data listed in Tables 2 and 4.1,11–16
It can be seen from Fig. 8, for most of considered This data set is chosen to provide an unbiased basis of
steels and loading conditions, the Ramberg–Osgood’s comparison among the methods because these data are
and Manson–Coffin’s strain ranges fairly well correlate not used for deriving the proposed correlations (Eqs.
with each other, especially for the elastic strain ranges. (9–11)). The predicted cyclic strength coefficient, (K′)
Therefore, the compatibility assumption is reasonable Pre, and cyclic strain hardening exponent, (n′)Pre, are
for most of considered steels. It is important to point compared with the tested ones, (K′)Test and (n′)Test, in
out here that, the Ramberg–Osgood and Manson–Coffin Figs. 9 through 15 where the dotted lines, solid lines
equations are not physical laws although they describe and dashed-dotted lines represent factors of ±10%, ±
well the cyclic response of many steels. Instead, the 20% and ±30% scatter band, respectively. It should be
compatibility equation (Eq. (3)) must be regarded as a pointed out that the true fracture strength (Sf) and true
measure of the coherence between those equations.4 fracture ductility (εf) are calculated by (1 + RA)Su and
Therefore, such estimates should not be used to replace ln(1  RA), respectively,23 because the two tensile prop-
experiments. Whenever possible, all six material parame- erties are unavailable in SAE J1099 Technical Report.11
ters should be independently obtained from experimental It can be seen from Figs. 9–14, the four-point correla-
data. These estimates, even the compatibility equation tion method (FPCM), universal slopes method (USM)
and proposed method, are only admissible during the and Baumel–Seeger method (BSM) tend to overestimate
first stages of design. the cyclic strength coefficient, while the estimated cyclic
In the absence of the four Manson–Coffin’s parame- strength coefficient by modified FPCM (MFPCM),
ters, some methods have been proposed for estimation modified USM (MUSM) and Medians method are all
of the fatigue properties (ε′f, σ′f, b and c) from simple scattered in a relatively wide range. Although the
tensile data.1,2,18–22 Therefore, K′ and n′ can also be Mitchell’s recommendation slightly underestimates the
estimated indirectly by substituting the predicted values cyclic strength coefficient (see Fig. 15), this method
of ε′f, σ′f, b and c into Eq. (3). This method will be may be said to provide the best estimation results among
hereafter referred to as indirect method. It should be the seven estimation methods. As for the cyclic strain
highlighted that the following evaluations and compari- hardening exponent, it can be seen from Table 5 and
sons of the indirect methods are based on extensive verifi- Eq. (3), all the estimated n′ data by USM, MUSM,
cations that the compatibility assumption is reasonable BSM and Medians method are constants. However, the
for most of steels. These indirect methods and even the experimental n′ are ranging from 0.069 to 0.321 (see

Fig. 8 Comparison of the elastic and plastic strain ranges calculated using the Manson–Coffin’s and Ramberg–Osgood’s parameters.

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
420 J. LI et al.


Tables 2 and 4). Therefore, the four estimation methods

 S 0:179
cannot predict the cyclic strain hardening exponent well

0:52  14 lg εf þ 13 lg 1  81:8 SEu Sfu


(see Figs. 10 and 12–14). For the other three estimation
methods (FPCM, MFPCM and Mitchell’s method), the
estimated n′ are all scattered in a wide range as shown
in Figs. 9, 11 and 15. This is indicating that all the seven
estimation methods are failed to estimate the cyclic strain

 lg εf

c

hardening exponent from monotonic tensile properties,


for steels.

 

0:00737Δεe =2 Except the indirect methods, several approaches have


2:074
also been proposed to predict cyclic deformation behav-
0.56 iour by direct estimating the cyclic strength coefficient
0.59

0.58
and cyclic strain hardening exponent from the monotonic
0.6

0.6

lg

tensile properties. In a study conducted by Zhang et al.,3 a


1
4

method for estimating the cyclic strength coefficient


o

and cyclic strain hardening exponent from monotonic


E
Sf



tensile properties is proposed. To estimate the cyclic


lg
0:0792  0:179 lg Sfu
S

strain hardening exponent, three characteristics based


Su 0:81 i

on the monotonic strain hardening exponent, n, true


i
b

fracture ductility parameter, α (α = RAεf), true fracture


 16 lg 2ðSuSþ345
E

strength, Sf, monotonic yield strength, Sy, and ultimate


0:16

0.087

tensile strength, Su, are determined empirically. These


h
0.12

0.09
0.09
n h
lg

parameters are shown to provide the following three


characteristics.3
1
6
 S 0:179 1=3

(1) n′ > n for α < 20% and for Sf/Sy < 1.6.


(2) n′ < n for α < 20% and for Sf/Sy > 1.6.
(3) SS  SS ≈nn for α > 20%.
f u

0:413εf 1  81:8 SEu Sfu

y y

Based on these three characteristics, new relationships


Su 0:53
ε′f

for estimating the cyclic strain hardening exponent are


derived by Zhang et al.3 These relationships are listed
E

in Table 6, where β is the adjusted coefficient (β = 1 for


0:0196ε0:155
0.7579ε0.6


Sf/Sy < 1.6, while β = 1 for Sf/Sy > 1.6). In a similar


f
f

0.59χ

manner, equations relating the cyclic strength coefficient


0.45

to the monotonic tensile properties are derived in Ref.,3


εf

εf
Su 0:832

which is also listed in Table 6. It should be pointed out


 0:893
Table 5 Estimation methods for Manson–Coffin’s parameters

σ′f (MPa)

that the true fracture strength (Sf) and the true fracture
E

ductility (εf) are also determined by (1 + RA)Su and ln


If Su/E ≤ 0.003, χ = 1; if Su/E > 0.003, χ = 1.376–125.4Su/E.
Su
Sf

1.9018Su

E0:623
Su(1 + εf)

Su + 345

(1  RA), respectively,23 in the present study.


1:12Su

1.5Su

1.5Su

Figure 16 shows the predicted results using Zhang’s esti-


mation method on 54 steels from SAE J1099 Technical
Modified four-point correlation methoda

Report11 and Refs..1,12–16 It can be seen from this figure,


all the estimated K′ and n′ data by Zhang’s estimation
103flg½0:16ðSu =EÞ lgðSf =EÞg .
Universal slopes method (USM) 18

21

i
Modified universal slopes method

method are scattered in a relatively wide range although this


Baumel–Seeger methodb (BSM)
Four-point correlation method

method is somewhat complicated to be used to estimate the


cyclic deformation properties.
0:81

For comparisons, experimental data from SAE J1099


Technical Report11 and Refs.1,12–16 are also used to eval-
Mitchell’s method 22
Evaluation methods

Medians method 1

uate the predictive accuracy of proposed correlations


(Eqs. (8–11)). The predicted results are shown in Fig. 17.
(MFPCM) 19

h 2
(MUSM) 20

For the cyclic strain hardening exponent, as can be found


(FPCM) 18

2Sf

from Fig. 17, the proposed correlations give good estima-


E
Δεe ¼

tion results for steels based on the number of points in-


cluded in ±30% scatter band. For the cyclic strength
b
a

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
AN IMPROVED METHOD FOR ESTIMATION OF RAMBERG–OSGOOD CURVES OF STEELS 421

Fig. 9 Comparison of the FPCM estimated and experimental K′ and n′ using the tested data from Refs.1,11–16.

Fig. 10 Comparison of the USM estimated and experimental K′ and n′ using the tested data from Refs.1,11–16.

Fig. 11 Comparison of the MFPCM estimated and experimental K′ and n′ using the tested data from Refs.1,11–16.

Fig. 12 Comparison of the MUSM estimated and experimental K′ and n′ using the tested data from Refs.1,11–16.

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
422 J. LI et al.

Fig. 13 Comparison of the Medians method estimated and experimental K′ and n′ using the tested data from Refs.1,11–16.

Fig. 14 Comparison of the BSM estimated and experimental K′ and n′ using the tested data from Refs.1,11–16.

Fig. 15 Comparison of the Mitchell’s method estimated and experimental K′ and n′ using the tested data from Refs.1,11–16.

Table 6 Zhang’s method3 for estimation of Ramberg–Osgood’s parameters

α K′ (MPa) n′ a
 
S2
u Sf
3

 0:545 lg 


S5

K ′ ¼ 57 Sf εn n′ ¼ 1:06 3 lg 500ε 1 þ β


1  SSuy

y
α < 5% or 10% ≤ α < 20%  1220
f ð fÞ
 S2
 f 


Sf Sy n 0:545
lg Sy Su

5% < α < 10% K ′ ¼ 57 Su εf  1220 n′ ¼ 1:06 2 lg 500ε 1 þ β


1  SSuy

ð fÞ
 S2
 lg
f
Sf Sy n 0:545
K ′ ¼ 57
S Sy Su
α > 20% Su εf  1220 n′ ¼ Sf S
y
u 2 lg ð500εf Þ

a
β = 1 for Sf/Sy < 1.6, while β = 1 for Sf/Sy > 1.6.

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
AN IMPROVED METHOD FOR ESTIMATION OF RAMBERG–OSGOOD CURVES OF STEELS 423

Fig. 16 Comparison of the Zhang’s method estimated and experimental K′ and n′ using the tested data from Refs.1,11–16.

Fig. 17 Comparison of the proposed method estimated and experimental K′ and n′ using the tested data from Refs.1,11–16.

coefficient, the predicted results of the proposed correla- Factors of ±10%, ±20% and ±30% deviation from exper-
tions are slightly scattered. However, the proposed imental values are included in the tables, along with the
method still be said to provide the best estimation results percentages of the data within each factor. It can be seen
among these considered estimation methods. from Tables 7 and 8 that method proposed in this study
The cyclic strength coefficient (K′) and cyclic strain had the most data with all deviation factors, either for K
hardening exponent (n′) predicted from the proposed ′ or for n′. For example, for K′ and n′, 80.2% and
method (Eqs. (8–11)) as well as the aforementioned eight 89.2% of the predicted values are within a factor of ±
methods are compared in Tables 7 and 8, respectively. 20% of the experimental values based on the proposed
For each method, data from Refs.1,9–16 are included. method, respectively. In contrast, the best results

Table 7 Prediction of cyclic strength coefficient (K′) and percent deviation from experimental data

Data from Refs.9,10 Data from Refs.1,11–16 All

0.9–1.1 0.8–1.2 0.7–1.3 0.9–1.1 0.8–1.2 0.7–1.3 0.9–1.1 0.8–1.2 0.7–1.3


(±10%) (±20%) (±30%) (±10%) (±20%) (±30%) (±10%) (±20%) (±30%)
(FPCM) 0.239 0.433 0.522 0.296 0.407 0.537 0.264 0.421 0.529
(USM) 0.164 0.343 0.478 0.093 0.167 0.278 0.132 0.264 0.389
(MFPCM) 0.299 0.552 0.731 0.333 0.611 0.722 0.314 0.578 0.727
(MUSM) 0.299 0.522 0.642 0.315 0.463 0.574 0.306 0.496 0.612
Medians method 0.328 0.537 0.716 0.296 0.519 0.667 0.314 0.529 0.694
(BSM) 0.298 0.478 0.642 0.259 0.481 0.611 0.281 0.479 0.628
Mitchell’s method 0.324 0.574 0.897 0.296 0.556 0.815 0.312 0.566 0.860
Zhang’s method 0.164 0.299 0.433 0.204 0.352 0.537 0.182 0.323 0.479
Proposed method 0.618 0.927 0.985 0.407 0.648 0.852 0.524 0.802 0.926

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
424 J. LI et al.

Table 8 Prediction of cyclic strain hardening exponent (n′) and percent deviation from experimental data

Data from Refs.9,10 Data from Refs.1,11–16 All

0.9–1.1 0.8–1.2 0.7–1.3 0.9–1.1 0.8–1.2 0.7–1.3 0.9–1.1 0.8–1.2 0.7–1.3


(±10%) (±20%) (±30%) (±10%) (±20%) (±30%) (±10%) (±20%) (±30%)
(FPCM) 0.269 0.433 0.493 0.241 0.333 0.426 0.257 0.388 0.463
(USM) 0.269 0.388 0.507 0.111 0.222 0.296 0.198 0.314 0.413
(MFPCM) 0.119 0.298 0.567 0.333 0.574 0.611 0.215 0.421 0.587
(MUSM) 0.209 0.433 0.612 0.13 0.296 0.537 0.174 0.372 0.579
Medians method 0.149 0.388 0.597 0.185 0.352 0.63 0.165 0.372 0.612
(BSM) 0.164 0.373 0.597 0.167 0.389 0.63 0.165 0.380 0.612
Mitchell’s method 0.147 0.309 0.529 0.204 0.5 0.648 0.172 0.394 0.582
Zhang’s method 0.164 0.284 0.403 0.222 0.315 0.426 0.190 0.298 0.413
Proposed method 0.574 0.956 0.985 0.574 0.812 0.963 0.574 0.892 0.975

predicted by MFPCM among the methods from the litera- xp


F ðxÞ ¼ ; x > 0; a > 0; p ≥ 1 (14)
ture are 57.8% and 42.1% of the predicted K′ and n′ within ap þ xp
the same factor of experimental values, respectively.
where p is a shape parameter, and a is a scale parameter.
In order to have a better quantitative comparison
The corresponding PDF of the log-logistic distribution
among different estimation methods, the comparisons
is given by24,25
for different methods are given in the format of fitted
probability density functions (PDF) for the prediction pap xp1
errors. For the statistical study, each data set is sorted in f ðx Þ ¼ ; x > 0; a > 0; p ≥ 1: (15)
ðap þ xp Þ2
ascending order, and then each data point is associated
to its mean rank. After that, each data set is fitted using The cumulative probability Pf of estimated data is
the selected probability distribution. In Ref.,4 Meggiolaro evaluated by the mean rank method as
and Castro evaluated 12 continuous probability distribu-
i
tions (i.e. beta, Birnbaum–Saunders, gamma, inverse Pf ¼ (16)
I þ1
Gauss, logistic, log-logistic, normal, log-normal, Pearson,
Gumbel (extreme value) and Weibull) using the Manson– where i is the rank number, and I is the total number of
Coffin’s and Ramberg–Osgood’s parameters and their considered steels. Symbols in Fig. 18 (a) and (b) show
estimates of 724 steels. The chi-square and Anderson- the (K′)Test/(K′)Pre and (n′)Test/(n′)Pre ratios of the
Darling tests are used to evaluate the goodness-of-fit of proposed method, respectively. Optimum values of the
each of the considered distributions.4 It is shown that parameters a and p are obtained by applying the least
the log-logistic distribution is the one that best fits the squares method.26 The curvilinear lines in Fig. 18 indi-
Ramberg–Osgood parameters.4 Therefore, the log- cate the results of fitting. The optimum values of a and
logistic distribution is selected in the present study. The p are listed in Tables 9 and 10. Then the best-fitted
cumulative distribution function of the log-logistic distributions of different estimation methods are shown
distribution function is given by24,25 in Fig. 19. It can be seen from this figure, in contrast to

Fig. 18 The estimates of the proposed method (the curvilinear lines show the fit of log-logistic distribution): (a) K′, and (b) n′.

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
AN IMPROVED METHOD FOR ESTIMATION OF RAMBERG–OSGOOD CURVES OF STEELS 425

Table 9 The optimum values of a and p for n′

Proposed Medians USM MUSM BSM Mitchell Zhang MFPCM FPCM


a 0.987 0.921 0.703 0.874 0.937 1.181 1.166 1.006 0.739
p 12.213 4.055 4.056 4.056 4.056 4.965 3.143 4.034 4.854
R2 0.991 0.993 0.993 0.993 0.993 0.986 0.994 0.996 0.992

Table 10 The optimum values of a and p for K′

Proposed Medians USM MUSM BSM Mitchell Zhang MFPCM FPCM


a 1.001 0.922 0.723 0.845 0.895 1.084 1.021 0.992 0.792
p 10.952 6.134 5.758 6.320 5.200 6.767 3.423 5.942 6.197
R2 0.996 0.997 0.997 0.997 0.998 0.993 0.987 0.992 0.998

reasonable well, either for cyclic strength coefficient or


for cyclic strain hardening exponent.

CONCLUSIONS

The present study investigated correlations among the


monotonic tensile properties, cyclic yield strength, cyclic
strength coefficient and cyclic strain hardening exponent
for steel materials. Experimental data for 67 steel mate-
rials from literature9,10 are used to obtain the correla-
tions. For comparisons without bias, 54 additional steel
materials from SAE J1099 report11 and Refs.1,12–16 are
also used to confirm the predictive accuracy of the
proposed correlations in comparison to the correlations
that is available in the literature. Based on the data
analysis and discussions presented in the preceding
sections, the following conclusions can be drawn:

1 A modified correlation among the cyclic yield


strength, ultimate tensile strength and percent reduc-
tion in area is developed by a non-linear least squares
fit. When the cyclic yield strength of steel is unavail-
able, this method may be said to be promising for
estimation of the cyclic yield strength.
2 Based on the modified correlation of cyclic yield
strength, a new method is proposed to estimate
the cyclic deformation properties. The proposed
method can estimate the cyclic strength coefficient
and/or cyclic strain hardening exponent very well
for steel materials when one of the cyclic strain
hardening exponent and cyclic strength coefficient
is available.
Fig. 19 Probability density functions of Ramberg–Osgood parame- 3 In the absence of any fatigue data, correlations
ters: (a) K′, and (b) n′. between the cyclic strength coefficient and ultimate
tensile strength are found when the data are separated
the available estimation methods, the proposed correla- into three material groups based on the Su/Sy ratio.
tions on the basis of the monotonic tensile data predict Then the cyclic strain hardening exponent can be
the cyclic deformation properties of most of steels estimated by using the proposed method.

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426
426 J. LI et al.

4 A comparison is made of several correlations that are 4 Li, J., Sun, Q., Zhang, Z. P., Li, C. W. and Qiao, Y. J. (2009)
obtained from the literature for estimating the cyclic Theoretical estimation to the cyclic yield strength and fatigue
limit for alloy steels. Mech. Res. Commun., 36, 316–321.
strength coefficient and cyclic strain hardening 5 Li, J., Zhang, Z. P., Sun, Q., Li, C. W. and Li, R. S. (2011) A
exponent from monotonic tensile properties. It is modified method to estimate fatigue parameters of wrought
shown that all the considered indirect methods failed aluminum alloys. J. Mater. Eng. Perform., 20, 1323–1329.
to estimate the n′. Except the Mitchell’s method, all 6 Manson, S.S. (1954) Behavior of materials under conditions of
the indirect methods also failed to estimate the K′. thermal stress. NACA TR-1170, Washington, DC: National Ad-
visory Committee for Aeronautics, Washington, DC, pp. 1–34.
The Mitchell’s method provides relatively reasonable 7 Coffin, L. F. (1954) A study of the effects of cyclic thermal
estimation results for the K′ among the seven indirect stresses in a ductile metal. J. Eng. Mater. Technol., 76, 931–950.
methods. Neither K′ nor n′ can be reasonably esti- 8 Ramberg, W., Osgood, W.R. (1943) Description of stress–strain
mated by Zhang’s direct method for the considered curves by three parameters. NACA TN-902, Washington, DC:
steels. In contrast to the existed estimation methods, National Advisory Committee for Aeronautics, Washington,
DC, pp.1–28.
the proposed correlations give the best predicted re- 9 Lopez, Z. (2012) Correlations among tensile and cyclic
sults of the considered steels, either for K′ or for n′. deformation properties for steels and implications on fatigue life
5 The Ramberg–Osgood and Manson–Coffin equations predictions. MS Thesis, University of Toledo.
describe well the cyclic response of many materials, 10 Kim, K. S., Chen, X., Han, C. and Lee, H. W. (2002) Estima-
however, they are not physical laws. Instead, the tion methods for fatigue properties of steels under axial and
torsional loading. Int. J. Fatigue, 24, 783–793.
compatibility equation (Eq. (3)) must be regarded as a
11 Society of Automotive Engineers (2002). Technical report on
measure of the coherence between those equations. low cycle fatigue properties ferrous and non-ferrous materials.
Although it is found that there is a reasonable (but not SAE, J1099.
exact) correlation between the Ramberg–Osgood’s 12 Science and Technology Committee of Aeronautic Engineering
and Manson–Coffin’s parameters for most of steels, Department. (1987) Handbook of Strain Fatigue Analysis. Science
Publishing House, Beijing, China. (in Chinese)
the incompatibility may be existed because different
13 Fatemi, A. (1985) Fatigue and deformation under proportional
pieces of data are used to determine the Manson– and non-proportional biaxial loading. Ph.D. Thesis, University
Coffin’s and Ramberg–Osgood’s parameters. of Iowa.
14 Hoffmeyer, J., Doring, R., Seeger, T. and Vormwald, M. (2006)
Finally, it is important to point out that all the pre- Deformation behaviour, short crack growth and fatigue lives under
sented estimation methods can only estimate material multiaxial non-proportional loading. Int. J. Fatigue, 28, 508–520.
properties and, therefore, should not replace experimen- 15 Yin, F. (2006) Cyclic deformation and fatigue behavior of
tal testing when reliable properties and high confidence case-hardened steels. Ph.D. Thesis, University of Toledo.
16 Fatemi, A., Zeng, Z. and Plaseied, A. (2004) Fatigue behavior
are required. The use of such estimates, even the
and life predictions of notched specimens made of QT and
proposed correlations, is only admissible during the first forged microalloyed steels. Int. J. Fatigue, 26, 663–672.
stages of design; otherwise all fatigue properties should 17 Nieslony, A., Kurek, A., Dsoki, C.e. and Kaufmann, H. (2012) A
be experimentally obtained. study of compatibility between two classical fatigue curve
models based on some selected structural materials. Int. J.
Fatigue, 39, 88–94.
Acknowledgements 18 Manson, S. S. (1965) Fatigue: a complex subject—some simple
approximations. Exp. Mech., 5, 193–226.
19 Ong, J. H. (1993) An improved technique for the prediction of
The authors gratefully acknowledge the financial support axial fatigue life from tensile data. Int. J. Fatigue, 15, 213–219.
of the National Natural Science Foundation of China 20 Muralidharan, U. and Manson, S. S. (1988) A modified universal
(No. 51575524), and the Natural Science Basic Research slopes equation for estimation of fatigue characteristics of
Plan in Shaanxi Province of China (No. 2015JM5240). metals. J. Eng. Mater. Technol., 110, 55–58.
21 Baumel, A. Jr. and Seeger, T. (1990) Materials Data for Cyclic
Loading (Supplement 1), Elsevier Science Publishers: Amsterdam,
Holland.
REFERENCES 22 Mitchell, M. R. (1996) Fundamental of modern fatigue analysis
for design. ASM Handbook, Fatigue & Fracture, 19, 227–249.
1 Meggiolaro, M. A. and Castro, J. T. P. (2004) Statistical evalua- 23 Zheng, X.L. (1994) Quantitative Theory of Metal Fatigue. Publish-
tion of strain-life fatigue crack initiation predictions. Int. J. ing House of Northwestern Polytechnical University, Xi’an, China.
Fatigue, 26, 463–476. (in Chinese)
2 Roessle, M. L. and Fatemi, A. (2000) Strain-controlled fatigue 24 Ashkar, F. and Mahdi, S. (2006) Fitting the log-logistic distribu-
properties of steels and some simple approximations. Int. J. tion by generalized moments. J. Hydrol., 328, 694–703.
Fatigue, 22, 495–511. 25 Lesitha, G. and Thomas, P. Y. (2013) Estimation of the scale
3 Zhang, Z. P., Qiao, Y. J., Sun, Q., Li, C. W. and Li, J. (2009) parameter of a log-logistic distribution. Metrika, 76, 427–448.
Theoretical estimation to the cyclic strength coefficient and 26 Huy, V. L., Gaspar, J., Paul, O. and Kamiya, S. (2012) Statistical
the cyclic strain hardening exponent for metallic materials: characterization of fatigue lifetime of polysilicon thin films.
preliminary study. J. Mater. Eng. Perform., 18, 245–254. Sensor. Actuat. A-Phys., 179, 251–262.

© 2015 Wiley Publishing Ltd. Fatigue Fract Engng Mater Struct, 2016, 39, 412–426

Vous aimerez peut-être aussi