You are on page 1of 16

International Journal of Fatigue 23 (2001) 751–766

www.elsevier.com/locate/ijfatigue

On some basic problems of fatigue research in engineering


Xiulin Zheng
Department of Materials Science and Engineering, Northwestern Polytechnical University, Xi’an, Shaanxi 710072, People’s Republic of China

Received 1 March 2001; received in revised form 19 April 2001; accepted 26 April 2001

Abstract

Research on the basic fatigue formula, the criterion for omitting small loads, the rule for cumulative fatigue damage, and models
for predicting the fatigue crack initiation (FCI) life and probability distribution under variable-amplitude loading are comprehensively
studied and summarized in the present paper. First, the formula for FCI life is given as a function of the equivalent stress amplitude,
and the procedures for investigating overload effects are introduced. Using the above formula to analyze test results of the overload
effect on FCI life yields the formula for FCI life containing the overload effect factor, z, which can be adopted to characterize the
load interaction effect when Miner’s rule is used to compute the cumulative fatigue damage. Fatigue test results and analysis of
the underload effect show that the FCI threshold in the FCI life formula can be taken as the criterion for omitting small loads in
life prediction and in compilation of the load spectrum. Then, the fatigue damage function is defined, the cumulative fatigue damage
can be computed and the FCI life of structure members under variable-amplitude loading can be obtained by using Miner’s rule.
Procedures for predicting the probability distribution of FCI life or fatigue life are described and examples are introduced. Further-
more, factors affecting the value of z and the applicability of Miner’s rule are briefly discussed. Finally, further areas for fatigue
research are tentatively proposed. Most interest should be focused on research into generally applicable life prediction models and
the approach for fatigue reliability assessment of structure details made of the metals with continuous strain-hardening characteristics,
which have been widely adopted in aeronautical engineering, under service load and environments.  2001 Elsevier Science Ltd.
All rights reserved.

Keywords: Fatigue formula; Fatigue crack initiation; Overload effect; Criterion for omitting small load; Strain-hardening characteristics; Fatigue
damage function; Miner’s rule; Life prediction; Joints

1. Introduction ured by using notched specimens to simulate the


geometry discontinuity in structural members, where the
From the viewpoint of engineering applications, the stress concentration occurs [1,3–5,9], and the test results
purpose of fatigue research consists of: (1) predicting the analyzed by one of the existing formulae, either empiri-
fatigue life of structures, (2) increasing fatigue life and cal or analytical [1,3,4,12–14]. The FCP life of structural
(3) simplifying fatigue tests, especially fatigue tests of members is analyzed by using the fracture mechanics
full-scale structures under a random load spectrum [1]. approach [3–5,8], and will not be discussed in the
The fatigue life of an engineering structure principally present paper.
depends upon that of its critical structure members. Aircraft structures usually operate under randomly
The fatigue life of an aircraft structure member can variable loading [15]. To date, formulating models for
be divided into two phases, the fatigue crack initiation predicting the fatigue life of elements under variable-
(FCI) life and the fatigue crack propagation (FCP) life, amplitude loading (VAL) is still the active field in
to be experimentally investigated and analyzed [1–10]. fatigue research [7,9,10,16–18]. It is well known that
Even for butt welds containing uncrack-like defects, the there are three governing factors in the life prediction of
FCI life occupies a great portion of the total fatigue life structural members under VAL: the fatigue life curve
and cannot be neglected [11]. FCI life has been meas- and the corresponding expression under constant-ampli-
tude loading (CAL), the load spectrum sustained by the
structure elements, and the rule for computing cumulat-
E-mail address: xiulinzheng@yahoo.com (X. Zheng). ive fatigue damage. Actually, the cumulative fatigue

0142-1123/01/$ - see front matter  2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 2 - 1 1 2 3 ( 0 1 ) 0 0 0 4 0 - 8
752 X. Zheng / International Journal of Fatigue 23 (2001) 751–766

damage rule is used to convert the fatigue damage under work has contributed to research on the overload effects
VAL into that under CAL. Until now, only fatigue test on FCI life and fatigue life of notched elements and
data or/and the fatigue curves under CAL have been structures [29–33], which has been summarized in part
given in many handbooks of the fatigue and mechanical elsewhere [34]. It was shown [34,35] that not only single
properties of metals [19–21], and empirical fatigue for- tension overloading increases FCI life and fatigue life,
mulas were applied in the fatigue design of metallic but also periodic overloading more efficiently increases
structures [22–25]. the FCI life and fatigue life of notched elements and
In the real service loads of engineering structures, structures as well. As a result, tension overloading was
small loads occupy the greatest portion of the load spec- taken as an approach for increasing fatigue life
trum [15,24,25]. Therefore, how to distinguish the small [34,36,37]. Most past research on overload effects on
load cycles that do not result in fatigue damage in metals FCI life or fatigue life was not correlated with life pre-
and then omit them is very important in load spectrum diction. However, attempts were made to apply the
compilation and life prediction of structure members, fatigue curves obtained after overloading to predict the
and in fatigue tests of full-scale structures under random fatigue life under VAL, and the prediction accuracy of
VAL, which are the most expensive type of test [26]. the fatigue life of notched elements was increased to
However, the criterion for omitting small loads in the some extent [33]. On the other hand, the overload effect
load spectrum has not been clearly expounded [2], or an on the fatigue life could not be expressed by a good
empirical criterion was adopted [15,22–25]. fatigue formula at that time [33]. Therefore, a generally
There exist more than 50 models for cumulative accepted life prediction model could not be obtained.
fatigue damage and life prediction, which were reviewed Quantitative analysis of the overload effect on FCI life
recently [27]. However, none of above-mentioned mod- of notched elements could only be done after a good
els can be generally accepted, and more efforts are formula for FCI life was put forward [38–40].
needed in order to provide general and reliable cumulat-
ive fatigue damage analysis and life prediction models 2.1. FCI life formula
to structure designers [27]. In past fatigue research, most
attention was focused on searching for damage functions In most engineering applications, the nominal stress
but further study on the basic formula for fatigue life applied to the structural elements is kept below the yield
has been more or less ignored [27]. This may be the stress of the metal. Consequently, the structural member,
main reason why universally accepted cumulative as a whole, is elastic under service loads but plastic
fatigue damage rule and life prediction models could not deformation will occur to form a small plastic zone at
be obtained [28]. the notch root because of stress concentration, as shown
Research on the basic fatigue formula, the criterion in Fig. 1 [38].
for omitting small loads, the rule for cumulative fatigue Fatigue crack initiation at the notch root may be
damage and models for predicting the FCI life and its assumed to occur due to fracture of a hypothetical
probability distribution under VAL are comprehensively fatigue element located at the notch root, as shown in
studied and summarized in the present paper. Otherwise, Fig. 1. Thus, the fatigue life of a smooth test specimen
models for general and reliable cumulative fatigue dam- can be taken as the FCI life of a notched element if the
age analysis and life prediction could be obtained. Spe- fatigue element at the notch root undergoes the same
cial emphasis will be placed on taking the materials stress–strain history as the metal at the notch tip. This
characteristics, structure details and real loads applied hypothesis was expounded by many investigators [41–
to structures into consideration in the fatigue research. 43] and developed into the local strain approach [43],
Experimental procedures based on the theoretical analy- which is still used for predicting FCI life of notched
sis are also required. Then, a discussion is made on the elements [2,7,8].
applicability of Miner’s rule for cumulative fatigue dam- From the above model, it can be seen that the govern-
age and the conditions under which Miner’s rule can be ing parameters for fatigue crack initiation at the notch
used. Finally, topics for further fatigue research are ten-
tatively proposed.

2. FCI life formula considering overload effects

The maximum load applied to a structure may be con-


sidered as the overload by comparison with smaller loads
in the load spectrum [15,24,25]. Therefore, it is neces-
sary and important to investigate the overload effects on Fig. 1. Schematic illustration of the plastic zone and the hypothetical
FCI life and fatigue life of structural members. A lot of fatigue element at the notch root [38].
X. Zheng / International Journal of Fatigue 23 (2001) 751–766 753

root are the fatigue life of smooth specimens under cyc- tension overload applied to notched elements before
lic strain and the local strain range at the notch root. The fatigue testing is shown in Fig. 2 [40,54], where the
local strain range should be correlated with the nominal nominal stress applied to the notched elements during
cyclic stress applied to the notched elements for con- overloading, SOL, is higher than the maximum cyclic
venient applications in engineering. Based on the above- nominal stress, Smax. In any case, the value of SOL is
mentioned hypothesis, the modified Manson–Coffin for- kept below the yield stress of the metal, and the notched
mula for fatigue life under cyclic strain [44] and the element, as a whole, remains elastic, a situation encoun-
approximate formula for local strain range at the notch tered in most engineering applications.
tip [45], the formula for FCI life of notched elements Overloading before fatigue testing (see Fig. 2) will
was developed as follows [38]: result in greater local strain and compressive residual
2/1+n −2
stress, and a larger plastic zone at the notch root, than
Ni⫽C[⌬s2/1+n
eqv ⫺(⌬seqv)th ] (1) does fatigue loading (see Fig. 1). The overload effect on
and the FCI life and then the fatigue life should be attributed
to the strain hardening of the metal and the compressive

冪2(1−R)K ⌬S,
1 residual stress in the surface layer at the notch root
⌬seqv⫽ t (2)
induced by overloading [33,34]. Therefore, it may be
deduced that the overload effect would be related to the
where C and (⌬seqv)th are, respectively, the FCI resist- local strain induced by overloading. The local strain at
ance coefficient and the FCI threshold, and n is the the notch tip induced by overloading can be calculated
strain-hardening exponent. In Eq. (2), ⌬S and R are the by the following equation [40,45]:

冋 册
nominal stress range and stress ratio, respectively, and Kt
1/1+n
is the stress concentration factor of the notched elements. 1
eOL⫽2 (⌬seqv)2OL , (3)
When R=⫺1, ⌬seqv=1/2Kt⌬S=KtSa; Sa is nominal stress ESk
amplitude. As a result, ⌬seqv may be referred to as the
equivalent stress amplitude. It may be seen from Eqs. where E and Sk are, respectively, the Young’s modulus
(1) and (2) that Eq. (1) shows the effect of the stress ratio and the strength coefficient in Hollomon’s equation [55].
on the FCI life and can thus be conveniently applied to (⌬seqv)OL is the equivalent stress amplitude applied to
predict the FCI life of structures under variable-ampli- the notched elements during overloading, and can be cal-
tude loading, where not only the stress range but also culated by Eq. (2) taking the value of R to be zero,
the stress ratio is varied. 1
(⌬seqv)OL⫽
冑2
It was shown that Eq. (1) can be applied and fits well KtSOL. (4)
the test results of FCI life obtained by testing notched
elements of various geometries under various stress
ratios [38,46–52], and test results for the FCI life of In a set of fatigue tests, (⌬seqv)OL and thus eOL are
notched elements experiencing overloading [39,40]. kept constant. Then, the effects of eOL and (⌬seqv)OL on
It may be worthwhile to note that C, the FCI resistance FCI life can be quantitatively analyzed by using Eq.
coefficient, may be considered to be the equivalent stress (1) [39].
amplitude when Ni=1/4 cycle, equivalent to the crack
initiation at the notch root during a tension test. The FCI 2.3. Overload effect factor
threshold, (⌬seqv)th, is the upper limit of the equivalent
stress amplitude, below or equal to which no fatigue Analysis of the results obtained by using Eq. (1) to
damage will occur in the metal at the notch root and the fit test results of the overload effect on FCI life, as shown
FCI life tends to be infinite [38]. Both C and (⌬seqv)th in Fig. 3, reveals that tension overload has no significant
are material constants, and can be predicted from the effect on the FCI resistance coefficient C in Eq. (1)
tensile properties [1,53]. If the values of C and (⌬seqv)th (which could also be deduced from the physical signifi-
are obtained and substituted into Eq. (1), the expression cance of C given in Section 2.1), but increases the FCI
for FCI life can be given without the need for fatigue
tests. Some examples of predicting FCI life from tensile
properties have been given in [50,51,53].

2.2. Experimental procedures for investigating


overload effect

Experimental procedures for investigating the over-


load effect on FCI life should be designed on the basis
of theoretical analysis and practical applications. The Fig. 2. Schematic illustration of tension overloading [40,54].
754 X. Zheng / International Journal of Fatigue 23 (2001) 751–766

Fig. 3. Effect of tension overloading on the FCI life of aluminum alloys: (a) LY12CZ alloy, (b) LC4CS alloy [40].

threshold [40]. The higher the overloading equivalent


stress amplitude (⌬seqv)OL, the longer the FCI life and
the higher the FCI threshold (see Fig. 3) [40]. The general expression for overload effect on FCI thres-
Fatigue test results of notched elements made of hold can be written as follows:
aluminum alloys, titanium and ultrahigh strength steels
(⌬seqv)th⫽(⌬seqv)th0⫹aeOL, (6)
[33] also show that the higher the overload, the higher
the increase of the fatigue limit and fatigue life, which where (⌬seqv)th and (⌬seqv)th0 are, respectively, the FCI
show the same pattern as those in Fig. 3. At that time, threshold with and without experiencing tension over-
however, quantitative analysis of the overload effects on loading, and a is a constant.
the fatigue limit and fatigue life was not given [33]. This Substituting Eq. (4) into Eq. (6), we can obtain the
pattern was also observed in experiments for the over- correlation between FCI threshold and overloading equi-
load effects on the corrosion FCI life of LY12CZ alloy valent stress amplitude as follows:
sheets [56].
(⌬seqv)th⫽(⌬seqv)th0⫹z(⌬seqv)2/1+n
OL , (7)
Using Eq. (3) to calculate eOL and plotting the data of
eOL and (⌬seqv)th as in Fig. 4, one can obtain a straight where z may be defined as the overload effect factor.
line between eOL and (⌬seqv)th. Regression analysis gives z=6.57×10-4 for LY12CZ alloy sheets and z=2.98×10⫺4
the following equations [40]: for LC4CS alloy sheets.

앫 for LY12CZ alloy sheets 2.4. FCI life formula considering the overload effect
(⌬seqv)th⫽176⫹2260eOL; (5a)
Substituting Eq. (7) into Eq. (1), one can obtain the
formula for FCI life with considering the overload effect
앫 for LC4CS alloy sheets as follows [40]:
(⌬seqv)th⫽174⫹2875eOL. (5b) Ni⫽C{⌬s2/1+n 2/1+n 2/1+n −2
eqv ⫺[(⌬seqv)th0⫹z(⌬seqv) OL ] } . (8)
Thus,

앫 for LY12CZ alloy sheets


Ni⫽1.58⫻1013[⌬s1.78
eqv ⫺(176⫹6.57 (9)
⫻10−4(⌬seqv)1.78
OL )
1.78 −2
] ;

앫 for LC4CS alloy sheets


Ni⫽5.70⫻1013[⌬s1.89
eqv ⫺(174⫹2.98 (10)
⫻10−4(⌬seqv)1.89
OL )
1.89 −2
] .

Fig. 4. Correlation between FCI threshold and the local strain


induced by tension overloading: (a) LY12CZ alloy, (b) LC4CS alloy The above FCI life formula [Eq. (8)] reveals the corre-
[40]. lation of the FCI life with the geometry of notched
X. Zheng / International Journal of Fatigue 23 (2001) 751–766 755

elements (Kt), the cyclic loading conditions (⌬S, R), the Test results also show that tension overloading has no
tensile properties, the FCI threshold and the overload remarkable effect on the FCI life or fatigue life of the
effect factor, which characterize the interaction effect medium-carbon steel 45 (similar to 1045), butt welds of
between the higher and lower loads in the load spectrum 16Mn steel and friction-welded joints of 45 steel [57–
[40,54], and may be thought a good one. Eq. (8) combin- 62,65,66]. In other word, the overload effect factor of
ing Miner’s rule for cumulative fatigue damage can be the metals with discontinuous strain-hardening charac-
conveniently applied to predict the FCI life of metallic teristics is approximately equal to zero; i.e., z=0. There-
notched elements under VAL [40,54,56–62]. fore, Eq. (8) can be simplified into Eq. (1), and Eq. (1)
can thus be used to predict the FCI life or fatigue life
of notched elements of metals with discontinuous strain-
3. Dependence of overload effect on strain hardening characteristics under VAL [57–62,65,66].
hardening characteristics of metals Based on the above test results and analysis, it may
be thought that the value of z, the overload effect factor
3.1. Two kinds of metal of metallic notched elements, depends on the strain-
hardening characteristics of the metal, and the overload
According to their strain-hardening characteristics, effect factor z was thought to be a material constant.
metals can be divided into two kinds [55]: (1) metals In [40,46], an empirical expression was put forward to
with discontinuous strain-hardening characteristics and estimate the value of z as follows:

冉冊冉 冊
(2) metals with continuous strain-hardening character- 1/2
istics; the former with a yielding plateau on their tension E n s0.2 2
z⫽ (E·Sk)−1/1+n. (13)
curve and the latter without. Low-carbon steels and 5 ef sb
welded joints belong to the first kind, and aluminum Eq. (13) could be applicable to the notched elements
alloys, titanium alloys and ultrahigh strength steels of aluminum alloys [46,52]; however, it needs to be
belong to the second kind. checked further.
3.2. Dependence of z value on strain-hardening
characteristics
4. Criterion for omitting small loads
For the metals with continuous strain-hardening
characteristics, such as LY12CZ and LC4CS aluminum In past, the underload (or subload) was defined as the
alloys, tension overloading increases the FCI life and nominal stress amplitude lower than or equal to the
threshold of notched elements, as shown in Fig. 3. endurance limit, and the underload effect on fatigue life
Fatigue test results on notched elements of aluminum was investigated experimentally by using smooth speci-
alloys, titanium, ultrahigh strength steels also show mens. Test results showed that underload cycles applied
[33,63,64] that tension overloading increases FCI life to smooth specimens increased the fatigue life or the
and thus the fatigue life and endurance limit. In other endurance limit of low-carbon steel [67] and cast iron
words, the overload effect factor of notched elements of [68], which was called “coaxing”. However, past
metals with continuous strain-hardening characteristics research on the underload effect was not associated with
is higher than zero; i.e., z⬎0. the omission of small load cycles in life prediction
For metals with discontinuous strain-hardening [67,68].
characteristics, such as low-carbon high-strength steels The omission of small load cycles is necessary and
16Mn and 15MnVN, tension overloading has no important in compilation of the load spectrum [15,26],
appreciable effect on the FCI life and threshold of in prediction of the fatigue life and in assessment of the
notched elements, i.e., (⌬seqv)th=(⌬seqv)th0, as shown in fatigue reliability of structures [22,24], and is most cost-
Fig. 5. In notched elements with and without experienc- effective in fatigue tests of components and structures
ing overloading, the FCI life can be expressed approxi- under long-term variable-amplitude or random loading
mately as follows [58]: histories [26]. To date, some empirical criteria have been
proposed and used [22–24,26,66]. How to omit the small
앫 for 16Mn steel loads in life prediction by using the local strain approach
1.79 −2
was not clearly expounded [2,6,7].
Ni⫽3.33⫻1014[⌬s1.79
eqv ⫺459 ] ; (11)
4.1. Theoretical consideration
앫 for 15MnVN steel
It was pointed out [26] that the main problem associa-
Ni⫽5.60⫻1014[⌬s1.83
eqv ⫺361
1.83 −2
] . (12)
ted with omission of small loads is to identify the “dam-
aging” and “non-damaging” cycles. In solving this prob-
756 X. Zheng / International Journal of Fatigue 23 (2001) 751–766

Fig. 5. Effect of tension overloading on the FCI life of low-carbon high-strength steels: (a) 16Mn hot-rolled plates, (b) 15MnVN normalized
plates [58].

lem, a good formula for FCI life should first be given effect on FCI life of 15MnVN steel notched elements at
[28]. ⌬seqv⬎(⌬seqv)th, as shown in Fig. 7 [48].
Eq. (1) shows that there exists an FCI threshold Test results of the underload effect in [38,40,58] show
expressed by equivalent stress amplitude, i.e., (⌬seqv)th. again that the underload cycles have no appreciable
Theoretically, when ⌬seqvⱕ(⌬seqv)th, fatigue damage effect on the FCI life at equivalent stress amplitude
will not occur in the metal at the notch root and the FCI higher than the FCI threshold. Therefore, the FCI thres-
life tends to be infinite; otherwise, fatigue damage will hold, (⌬seqv)th in Eq. (1), could be preliminarily taken
occur and the FCI life will be limited. If the cyclic loads as the threshold to distinguish the damaging and non-
expressed by equivalent stress amplitude are lower than damaging cycles, and as the criterion for omitting small
or equal to the FCI threshold, i.e., ⌬seqvⱕ(⌬seqv)th, these load cycles in life prediction and in compilation of the
loads can be defined as the underload. If the underload load spectrum.
cycles have no appreciable effect on the FCI life at cyc- Further examination of the above conclusion was car-
lic loads higher than the FCI threshold, i.e., ried out by fatigue tests of notched friction-welded joints
⌬seqv⬎(⌬seqv)th, the FCI threshold can thus be taken as of 45 steel under more complex load spectra as shown
the criterion for omitting small loads. in Figs. 8 and 9, which were denoted as Spectrum 1 and
Based on the above considerations, experimental pro- Spectrum 2, respectively [66]. The difference between
cedures could be proposed for investigating the Spectrum 1 and Spectrum 2 is that 107 cycles of small
underload effect and the criterion for omitting small load or underload in Spectrum 1 are applied to the
loads. notched friction-welded joints before the fatigue tests
under block loading shown in Fig. 9. Spectrum 2 was
4.2. Criterion for omitting small loads for metals with used for the comparative study to determine whether the
discontinuous strain-hardening characteristics small load cycles could be omitted in the life prediction.
In the design of the load spectra, the fatigue threshold
In order to investigate the underload effect, fatigue needs to be known in advance. It was reported [60] that
tests were carried out by a two-level load spectrum the fatigue threshold is 409.2 MPa [also see Eq. (17a)]
shown in Fig. 6: the notched specimens were loaded at and the corresponding nominal stress amplitude is 204.6
⌬seqvⱕ(⌬seqv)th for 1×107 or at least 5×106 cycles first, MPa for notched friction-welded joints of 45 steel with
and then cyclically loaded at ⌬seqv⬎(⌬seqv)th, to see if Kt=2.0 and at R=⫺1. In Spectrum 1, 107 load cycles at
the pre-underload cycles have an effect on FCI life or
not [38,40,54].
The underload up to 107 cycles has no remarkable

Fig. 6. Schematic illustration of fatigue test program to investigate Fig. 7. Test results of underload effect on FCI life of 15MnVN steel
underload effect on FCI life. notched specimens expressed by Eq. (12) [48].
X. Zheng / International Journal of Fatigue 23 (2001) 751–766 757

Spectrum 2 come from the same population. In other


words, the first continuous 107 cycles under
⌬seqvⱕ(⌬seqv)th (=409.2 MPa) do not result in fatigue
damage in the notched friction-welded joints. This con-
firms again that the FCI threshold in Eq. (1) can be taken
as the criterion for omitting small loads for the notched
elements of metals with discontinuous strain-hardening
characteristics.

4.3. Criterion for omitting small loads for metals with


continuous strain-hardening characteristics
Fig. 8. Schematic illustration of the block loading used for investigat-
ing the underload effect on the fatigue life of notched friction-welded The 107 cycles under ⌬seqvⱕ(⌬seqv)th0 have no
joints [66]. appreciable effect on the FCI life of notched elements
of LY12CZ alloy sheets [38], which is a metal with con-
tinuous strain-hardening characteristics. The same results
were also observed for notched specimens of 30CrMnSi-
Ni2A ultrahigh strength steel and its welded joints [49].
However, tension overloading increases the FCI thres-
hold of notched elements of metals with continuous
strain-hardening characteristics [see Fig. 4 and Eq. (8)].
Whether the FCI threshold increased by tension over-
loading, i.e., (⌬seqv)th in Eq. (8), could be taken as the
criterion for omitting small loads needed to be studied
further.
For LY12CZ alloy sheets, tension overloading of
Fig. 9. Schematic illustration of the block loading for fatigue tests of
notched friction-welded joints of 45 steel for comparative study [66]. (⌬seqv)OL=620 MPa increases the FCI threshold from
(⌬seqv)th0=180 MPa to (⌬seqv)th=237 MPa [see Fig. 4
and Eq. (4) or Eq. (9)]. Thus, fatigue tests were perfor-
the nominal stress amplitude S0=195 MPa, a little lower med by the following procedures: (1) LY12CZ alloy
than the nominal fatigue stress amplitude of 204.6 MPa, notched specimens were overloaded at (⌬seqv)OL=620
are applied to the notched specimens first. Thereafter, MPa, (2) specimens were underloaded at ⌬seqvⱕ237
the five levels of nominal stress through S1 to S5 in one MPa for 5×106 cycles and (3) specimens were cyclically
load block shown in Fig. 9 are higher than 204.6 MPa loaded at ⌬seqv⬎237 MPa to determine the FCI life;
and repeated until failure of the notched friction-welded which may be illustrated schematically in Fig. 10. If the
joints. In this case, the fatigue life is defined as the load underload or small load cycles of 5×106 at ⌬seqvⱕ237
blocks applied to the notched test specimen until its MPa after overloading of (⌬seqv)OL=620 MPa had no
final fracture. remarkable effect on the FCI life of LY12CZ notched
The fatigue test results of notched friction-welded elements expressed by Eq. (9), the FCI threshold
joints of 45 steel under Spectrum 1 shown in Fig. 8 and increased by overloading of (⌬seqv)OL=620 MPa could
Spectrum 2 shown in Fig. 9 are listed in Table 1 accord- be taken as the criterion for omitting small loads for
ing to the order of increasing fatigue life. LY12CZ notched elements experiencing the same over-
It is well known that fatigue test results of elements loading.
under both CAL and VAL are dispersed in a wide scatter Results obtained from the above-mentioned fatigue
band, and it is usual to employ a log-normal distribution test are shown in Fig. 11 [54]. It can be seen from Fig.
to represent the statistical variability of fatigue life 11 that the underload or small load cycles of 5×106 at
[8,69]. It was also shown that the fatigue lives under ⌬seqvⱕ(⌬seqv)th=237 MPa, the equivalent stress ampli-
both Spectrum 1 and Spectrum 2 follow the log-normal tude below or equal to the FCI threshold increased by
distribution and the logarithmic mean values of fatigue overloading of (⌬seqv)OL=620 MPa, have no appreciable
life agree well with each other [60]. The results obtained effect on the FCI life of LY12CZ notched elements,
from the further statistical test suggest that there is no which experienced the same magnitude of overloading
significant difference between the two logarithmic mean of (⌬seqv)OL=620 MPa.
fatigue lives, and between the two logarithmic standard Based on the above test results and analysis, it may
deviations under the significance level of 5% [66]. It can be thought that the FCI threshold increased by over-
thus be thought that the fatigue test results of the notched loading could be taken as the criterion for omitting small
friction-welded joints of 45 steel under Spectrum 1 and load cycles in the life prediction of notched elements
758 X. Zheng / International Journal of Fatigue 23 (2001) 751–766

Table 1
The tested fatigue life (Nbe) and cumulative fatigue damage (Dce) of 45 steel friction-welded joints under Spectrum 1 and Spectrum 2 [66]

No. Spectrum 1 Spectrum 2

Nbe (block) Dce log Nbe log Dce Nbe (block) Dce log Nbe log Dce

1 3.57 0.4907 0.5525 ⫺0.3092 3.72 0.5111 0.5702 ⫺0.2915


2 3.85 0.5289 0.5850 ⫺0.2767 5.19 0.7136 0.7151 ⫺0.1466
3 5.11 0.7032 0.7087 ⫺0.1530 6.73 0.9254 0.8280 ⫺0.0337
4 5.32 0.7316 0.7259 ⫺0.1358 7.45 1.0248 0.8732 0.0106
5 8.92 1.2264 0.9503 0.0886 7.94 1.0923 0.9000 0.0383
6 8.92 1.2264 0.9503 0.0886 12.15 1.6709 1.0846 0.2230
7 11.43 1.5720 1.0581 0.1964
8 13.22 1.8178 1.1212 0.2595
9 14.81 2.0359 1.1704 0.3087
Xa 8.35 – 0.8692 0.0075 7.20 0.8284 ⫺0.0333
Sb – – 0.2320 0.2320 0.1745 0.1745

a
The mean value.
b
The standard deviation.

5. Cumulative fatigue damage calculation and life


prediction

5.1. Fatigue damage function

From fatigue test results in Section 2, we have defined


the overload effect factor z, which can be used to charac-
terize the load interaction effect [1,40,54,56–62].
Besides, the FCI threshold in Eq. (1) or Eq. (8) can be
taken as the criterion for omitting small loads in the life
prediction and the compilation of load spectrum. Conse-
Fig. 10. Schematic illustration of the procedures for fatigue tests to quently, the FCI life of metallic notched elements could
investigate the underload effect on the FCI life of LY12CZ notched be predicted by using Miner’s rule and Eq. (8) with or
elements: (1) overloaded at (⌬seqv)OL=620 MPa, (2) underloaded at
⌬seqvⱕ237 MPa for 5×106 cycles and (3) cyclically loaded at
without taking the overload effect into consideration,
⌬seqv⬎237 MPa to determine the FCI life [54]. depending on the strain-hardening characteristics of the
metal from which the notched structural elements are
made.
In order to predict the FCI life of notched elements
under VAL, the expression for calculating the fatigue
damage — i.e., the fatigue damage function, Dj — can
be defined according to Miner’s rule as follows:
1
Dj ⫽ (14)
Ni,j
1
⫽ .
C{⌬s 2/1+n
eqv,j −[(⌬seqv)th0+z(⌬seqv)2/1+n
OL ]
2/1+n −2
}
Fig. 11. Test results of FCI life (indicated by circles) of LY12CZ
notched specimens obtained from the fatigue test carried out according And the cumulative fatigue damage can be computed as:
to the procedure shown in Fig. 10, and the solid curve drawn according
to Eq. (9) [54].

k
nj
Dc ⫽ (15)
N
j⫽1 i,j
of metals of continuous strain-hardening characteristics,


k
when the notched elements experience overloading of nj
⫽ .
the same direction and magnitude. This conclusion was
j⫽1
C{⌬s 2/1+n
eqv,j −[(⌬seqv)th0+z(⌬seqv)2/1+n
OL ]
2/1+n −2
}
further verified by the FCI life prediction and test results
of LY12CZ notched elements with a stress concentration In Eq. (15), k is the number of levels of the equivalent
factor of 4.0 under a flight-to-flight spectrum [40]. stress amplitude. It can be seen from Eqs. (14) and (15)
X. Zheng / International Journal of Fatigue 23 (2001) 751–766 759

that a good formula for the FCI life or fatigue life con- Then, computing the cumulative fatigue damage
sidering the stress ratio effect is the key point for accur- according to Eq. (15) and the equivalent stress ampli-
ate computation of the cumulative fatigue damage in the tude spectrum of 45 steel notched elements.
life prediction of structure members under VAL, where 3. Obtaining the fatigue life of 45 steel notched elements
the stress ratios are varied. Without a good formula for under the load spectrum shown in Fig. 12 when the
the FCI life or the fatigue life, the accurate life prediction cumulative fatigue damage reaches critical value,
under VAL would be impossible. i.e., Dc=1.0.

5.2. Life prediction of notched elements of metals with The predicted fatigue life of notched elements of 45
discontinuous strain-hardening characteristics steel under the load spectrum shown in Fig. 12 is 11.2
blocks and the logarithmic mean value of 9 test data
45 carbon steel has discontinuous strain-hardening is 10.6 blocks [57]. Good agreement exists between the
characteristics. Test results show that tension over- predicted life of the 45 steel notched elements and test
loading has no appreciable effect on the fatigue life of results.
notched elements of 45 steel (i.e., z=0) [57]. Therefore, For notched friction-welded joints of 45 steel under
the fatigue life of 45 steel notched elements under VAL Spectrum 1 in Fig. 8 and Spectrum 2 in Fig. 9, the
can be predicted by using Eqs. (14) and (15), taking z=0 fatigue life predicted by the above procedures and Eq.
and Dc=1.0. (17a) is 7.3 blocks, which agrees well with the test
Since the fatigue life was measured by using small results listed in Table 1 [60].
size notched specimens, the fatigue life is mostly The procedures mentioned above were also used for
expended in the crack initiation stage and can be accurately predicting the FCI life or fatigue life of
approximately taken as the FCI life, particularly in the notched elements of metals with discontinuous strain-
long life range [42]. The expression for the fatigue life hardening characteristics, such as 15MnVN steel
of 45 steel notched elements was obtained from test [57,62,70], old bridge steel [65] and 16Mn steel butt
results as follows [57]: welds [58,59].
1.736 −2
Nf⫽2.39⫻1014(⌬s1.736
eqv ⫺303 ) . (16)
5.3. Life prediction of notched elements of metals with
continuous strain-hardening characteristics
The load spectrum used for predicting and testing the
fatigue life of the 45 steel notched elements is shown in
Fig. 12. For 45 steel notched elements, the fatigue life LY12CZ aluminum alloy belongs to the kind of met-
can be predicted by following procedures. als with continuous strain-hardening characteristics, and
has the overload effect factor greater than zero (i.e.,
1. Substituting all of the nominal stress ranges ⌬S and z⬎0). Therefore, the load interaction effect needs to be
the stress ratios R (=⫺1) in the load spectrum and the taken into account in life prediction. In this case, Eq. (9)
stress concentration factor Kt (=2.0) of the notched should be substituted into Eq. (14) to define the fatigue
elements into Eq. (2) to transform the load spectrum damage function of the notched elements of LY12CZ
in Fig. 12 into the equivalent stress amplitude spec- alloy. The procedures for predicting the FCI life of
trum, and omit all of the equivalent stress amplitudes LY12CZ notched elements was given as follows [40].
lower than or equal to the FCI threshold.
2. Substituting Eq. (16) into Eq. (14) to define the 1. Calculating the value of (⌬seqv)OL from the maximum
fatigue damage function of 45 steel notched elements. nominal stress taken from the load spectrum and the
Kt value (=4.0) of the notched elements by using
Eq. (2).
2. Substituting (⌬seqv)OL into Eq. (9) to obtain the
expression for the FCI life of notched specimens
experiencing overloading.
3. Substituting all of the nominal stress ranges and the
stress ratios in the load spectrum and the Kt value
of the notched element into Eq. (2) to compute the
equivalent stress amplitude spectrum and omit all of
the equivalent stress amplitudes lower than or equal
Fig. 12. Load spectrum for predicting and testing the fatigue life of to the FCI threshold increased by overloading.
45 steel notched elements [57]: (1) S1=240 MPa, n1=105 cycles; (2)
S2=350 MPa, n2=8×104 cycles; (3) S3=400 MPa, n3=2.5×104 cycles; 4. Substituting the expression for the FCI life of
(4) S4=500 MPa, n4=104 cycles; (5) S5=S3, n5=n3; (6) S6=S2, n6=n2; (7) LY12CZ notched elements experiencing overloading
S7=S1, n7=n1. into Eq. (14). Then, computing the cumulative fatigue
760 X. Zheng / International Journal of Fatigue 23 (2001) 751–766

damage according to Eq. (15) and the equivalent expressions for the curves of fatigue life with given sur-
stress amplitude spectrum. vivability and the probability distribution of fatigue
5. Obtaining the FCI life of the notched elements under strength [65,72,75].
VAL when the value of the cumulative fatigue dam-
age reaches 1.0. 6.2. Prediction of probability distribution of fatigue
life of notched elements of metals with discontinuous
The FCI life of LY12CZ alloy notched elements with strain-hardening characteristics
Kt=4.0 under a flight-by-flight random load spectrum
predicted by using the above procedures is 1441 flight Based on the above assumption and the procedures in
hours while the mean value of 4 test data is 1848 flight Section 5.2, the probability distribution of the fatigue life
hours; the predicted FCI life is about 0.78 of test results of notched friction-welded joints of 45 steel under VAL
and shorter than the test results but on the safe side [40]. shown in Figs. 8 and 9 could be predicted by using the
The above procedures were also adopted to predict the expressions of fatigue life with given survivability (Sv),
FCI life of LY12CZ notched elements under VAL in a which are rewritten as follows [66]:
corrosive environment of 3.5% NaCl solution. The pre-
dicted corrosion FCI life of LY12CZ notched elements 앫 Sv=50%,
under VAL agrees well with test results [52,56]. 1.764 −2
Nf⫽2.181⫻1014[⌬s1.764
eqv ⫺409.2 ] ; (17a)

6. Prediction of probability distribution of fatigue 앫 Sv=99.9%,


life under VAL 1.764 −2
Nf⫽1.098⫻1014[⌬s1.764
eqv ⫺348.6 ] ; (17b)
6.1. Basic assumptions
앫 Sv=95%,
It is well known that FCI life and fatigue life test Nf⫽1.514⫻1014[⌬s1.764 1.764 −2
eqv ⫺375.7 ] ; (17c)
results under both CAL and VAL have great dispersion
[8,20,52,69], especially for brittle ceramic materials
[71,72] and some composites [73]. No matter whether 앫 Sv=5%,
the predicted mean fatigue life agrees well with that of Nf⫽3.143⫻1014[⌬s1.764 1.764 −2
eqv ⫺445.6 ] ; (17d)
test results under VAL, it is still hard, however, to draw
a conclusion about the accuracy of the life prediction
model. Usually, the log-normal distribution is used to 앫 Sv=0.1%,
represent statistical variability of fatigue life and fatigue Nf⫽4.333⫻1014[⌬s1.764 1.764 −2
eqv ⫺480.2 ] . (17e)
test results under VAL [8,20,69,74]. If the predicted
probability distribution of fatigue life under VAL agreed
with that of fatigue test results under the same VAL
spectrum, the life prediction model could be thought The predicted probability distribution of fatigue life
reliable [1,2,18]. On the other hand, a new model could and the test results of 45 steel notched friction-welded
be provided for the fatigue reliability assessment of joints under VAL shown in Figs. 8 and 9 are shown in
the structures. Fig. 13(a), where the mean rank was taken to be the
It was assumed that the fatigue life curve with given
survivability could be used to predict the fatigue life of
the notched elements under VAL with corresponding
survivability. If the fatigue life with different surviv-
ability under VAL is obtained based on the above
assumption, the probability distribution of the fatigue life
under VAL can be easily determined. The above
assumption has been checked by lots of test results
[57,60–62,65,66,70].
In order to meet the requirement of the life prediction
with given survivability and the fatigue reliability assess-
ment of structures, fatigue test results under CAL should
be analyzed statistically. Then, the fatigue life curves
with given survivability — the so-called P–S–N Fig. 13. Predicted probability distribution of fatigue life (a) and the
curves — and the corresponding expressions could be critical cumulative damage (b) of 45 steel notched friction-welded
given. Procedures were developed to determine joints under VAL shown in Figs. 8 and 9 and the test results [66].
X. Zheng / International Journal of Fatigue 23 (2001) 751–766 761

estimated value of the failure probability, Pf [66]. It


should be pointed out that the test results shown in Fig.
13 include all of the ones obtained under Spectrum 1 in
Fig. 8 and Spectrum 2 in Fig. 9 because they are thought
to be from the same population [66], as mentioned in
Section 4.2.
Dividing the predicted fatigue life with given surviv-
ability by the predicted life (in blocks) with 50% Sv, one
can easily obtain the critical value of the cumulated
fatigue damage with corresponding survivability, Dc,
which is also plotted on normal probability paper as
Fig. 14. Illustration of the load spectrum for predicting and testing
shown in Fig. 13(b). As may be seen, the predicted the FCI life of notched specimens of LY12CZ aluminum alloy sheets,
fatigue life and values of the cumulated fatigue damage, which was converted into an equivalent stress amplitude spectrum as
Dc, of 45 steel notched friction-welded joints follow the follows [57]: (1) ⌬seqv=400 MPa, n=30 cycles; (2) ⌬seqv=350 MPa,
log-normal distribution, which agrees well with test n=60 cycles; (3) ⌬seqv=300 MPa, n=480 cycles; (4) ⌬seqv=280 MPa,
n=2000 cycles; (5) ⌬seqv=250 MPa, n=3000 cycles; (6) ⌬seqv=180
results [66].
MPa, n=3000 cycles; (7) ⌬seqv=350 MPa, n=60 cycles; (8) ⌬seqv=300
Similar results were obtained for the notched elements MPa, n=480 cycles.
and welded joints of low-carbon steels with the discon-
tinuous strain-hardening characteristics. The predicted
probability distributions of the fatigue life and the cumu- with survivability were given in [57] and rewritten as
lative fatigue damage of low-carbon steel notched follows:
elements and welded joints follow the log-normal distri-
bution, and agree well with the test results [57,60– 앫 for 50% Sv,
62,65,66,70]. It may be thought that the fatigue formula 1.78 −2
Ni⫽3.71⫻1013[⌬s1.78
eqv ⫺223.7 ] ; (18a)
[i.e., Eq. (1)], the criterion for omitting the small loads
and Miner’s rule can be successfully used to predict the
fatigue life and its probability distribution of notched 앫 for 95% Sv,
elements and welded joints of metals with discontinuous
Ni⫽1.72⫻1013[⌬s1.78
eqv ⫺196.1
1.78 −2
] ; (18b)
strain-hardening characteristics under VAL. In other
words, the generally applied life prediction model is
obtained for notched elements and welded joints of met- 앫 for 84.1% Sv,
als with discontinuous strain-hardening characteristics 1.78 −2
Ni⫽2.28⫻1013[⌬s1.78
eqv ⫺209.4 ] ; (18c)
under VAL through our recent research.

6.3. Prediction of probability distribution of fatigue 앫 for 15.9% Sv,


life of notched elements of metals with continuous 1.78 −2
Ni⫽6.35⫻1013[⌬s1.78
eqv ⫺233.5 ] ; (18d)
strain-hardening characteristics

For the notched elements of metals with continuous 앫 for 5% Sv,


strain-hardening characteristics, the probability distri- 1.78 −2
Ni⫽9.18⫻1013[⌬s1.78
eqv ⫺238.2 ] . (18e)
bution of fatigue life under VAL could be predicted, in
principle, from the above-mentioned assumption and the
procedures in Section 5.3. Preliminary attempts were
made to predict the FCI life and the probability distri- Using Eqs. (18) and the procedures in Section 5.3,
bution of notched elements of LY12CZ aluminum alloy the FCI lives of LY12CZ notched elements with given
under block loading, as shown in Fig. 14 [57]. survivability under the load spectrum shown in Fig. 14
Before the fatigue tests under VAL shown in Fig. 14, can be predicted and are listed in Table 2, where the test
a single overloading of (⌬seqv)OL=620 MPa was applied results are also presented. Dividing the FCI lives with
to the LY12CZ notched elements. Hence, the same over- given survivability by the FCI life with 50% Sv, the
loading of (⌬seqv)OL=620 MPa should be also applied to values of the cumulative fatigue damage can be obtained
the LY12CZ notched specimens before the fatigue tests and are also listed in Table 2. Plotting the logarithmic
to determine the FCI life curves with given survivability values of the predicted fatigue life and the cumulative
under CAL, as shown in Fig. 2. Then, the test results fatigue damage with given survivability on normal prob-
for the FCI life of LY12CZ sheets under CAL were ana- ability paper, straight lines can be obtained as shown in
lyzed by using the procedures given in [61]. Fig. 15, where the test results are also shown.
The expressions for the FCI life of LY12CZ sheets The predicted FCI life of LY12CZ notched elements
762 X. Zheng / International Journal of Fatigue 23 (2001) 751–766

Table 2
The predicted FCI lives with given survivability and the test results of LY12CZ notched elements under block loading shown in Fig. 14 [56]
Predicted FCI lives with given survivability

Sv (%) 95 84.1 50 15.9 5


Ni (blocks) 27.9 49.6 118 275 467
Dc 0.236 0.42 1.0 2.33 3.95

Test results

Specimen no. 1 2 3 4 5 6
Sv (%) 87.5 75 62.5 50 37.5 25
Ni (blocks) 57.1 81.1 94.1 103.8 106.1 112.1
Dc 0.483 0.686 0.796 0.878 0.898 0.948

It can be seen that there are four materials constants


in Eq. (8): C, (⌬seqv)th0, n and z, which must each follow
a certain respective distribution. If the probability distri-
bution of these material constants can be predicted or
experimentally determined, the FCI life curves with
given survivability at various amplitudes of (⌬seqv)OL
could be obtained from the probability distribution of the
above-mentioned four constants, and used for predicting
the probability distribution of fatigue life under VAL.

Fig. 15. Prediction of the probability distribution of FCI life and test
results of notched elements of LY12CZ sheets under VAL shown in
7. Factors affecting overload effect and
Fig. 14 [56]. applicability of Miner’s rule

It can be seen from Eqs. (14) and (15) that whether


with 5O% Sv is 118 blocks, a little longer than 94 blocks, or not Miner’s rule could be adopted in the accurate life
the mean value of test results (see Table 2). From Table prediction of metallic notched elements depends upon
2 and Fig. 15, it can be seen that the predicted results the fatigue formula selected and the value of the over-
of the probability distribution of fatigue life under VAL load effect factor, z. Apart from considering the strain-
could be thought to agree with the test results [57] and hardening characteristics of metals, however, the follow-
the test results are almost within the range of the pre- ing factors have effects on the value of z and should
dicted FCI lives for respectively 15.9% and 84.1% Sv, be taken into account in life prediction using Miner’s
i.e., within ±1.0s scatter band. rule [76].
However, the test results of the FCI life at higher sur-
vivability are a little longer but those at lower surviv- 7.1. Effect of specimen geometry on overload effect
ability are shorter than those predicted. The reasons for factor
this discrepancy may be that the sample size of the
fatigue tests is too small [57] and the mechanical proper- The overload equal to the yield stress of the metal
ties of the LY12CZ sheets varied in a large range applied to 16Mn and 15MnVN steel smooth specimens
[40,54,57]. Besides, the strain-hardening exponent n in decreases the fatigue life and fatigue limit; i.e., the over-
Eqs. (18) may be a random variant because it is a load effect factor for the smooth specimen of low-carbon
material constant. The preliminary results mentioned high-strength steels has negative value, z⬍0
above, however, show that the procedures introduced in [1,48,58,76]. On the other hand, underload cycles
this section would be feasible, and need to be further applied to smooth specimens will result in the increase
verified by more test results. of the fatigue life [67,68]. As a result, the fatigue life
In engineering practice, the maximum nominal stress of smooth specimens of low-carbon steels, which was
applied to the various structure elements is different and predicted by using Eqs. (14) and (15) and taking z=0,
could not be determined in advance. Consequently, it is will be longer than test results when the greater stress
necessary to determine the FCI curves with given surviv- (i.e., overload) is applied to smooth specimens first; or
ability after overloading of different amplitudes of the predicted fatigue life will be shorter than test results
(⌬seqv)OL by fatigue tests, which are too expensive and when the smaller stress is applied to smooth specimens
time-consuming. This problem needs to be solved in a first. The test results given in [77] show agreement with
new way. the above argument.
X. Zheng / International Journal of Fatigue 23 (2001) 751–766 763

However, the fatigue life of smooth specimens of tion should be paid to this point in the life prediction of
cold-worked nickel–silver wire under two-level loading riveted aircraft structures.
could be accurately predicted by using Miner’s rule
[57,69]. 7.4. Effect of surface strain hardening

7.2. Overload effect factor of joints Surface strain hardening is an effective technical mea-
sure to increase FCI life and fatigue life
Welded, bolted and riveted joints are widely applied [36,37,67,68,79–82]. Test results [61,82] show that over-
in various structures, and often are the critical location load has no remarkable effect on the FCI life and fatigue
for FCI. Therefore, it is necessary to investigate the life of elements with a cold-worked hole and hammer-
overload effect and the life prediction model for such penned welds when the strain induced by cold work or
joints. Generally, joints made of metal have the same by hammer penning is greater than that induced by over-
characteristics of overload effect as the corresponding loading. In this case, the FCI life or fatigue life could
metal with specific strain-hardening characteristics be predicted by using Eqs. (14) and (15) and taking z=0
[1,33,40,58–60]. Test results [78] show that welded butt [40,61,82], irrespective of the strain-hardening charac-
joints of 16Mn steel have the same overload effect as teristics of the metal.
that of the base metal, 16Mn steel with discontinuous
strain-hardening characteristics (i.e., z=0). 7.5. Effect of load pattern
For bolted joints of titanium alloy, a metal with con-
tinuous strain-hardening characteristics, the overload Test results [32,34,35] show that the load pattern —
increases the fatigue life and fatigue limit, which means in particular, the regular periodic overload — has a much
that the value of z is greater than zero (i.e., z⬎0) [33]. greater effect on the FCI life and fatigue life of notched
elements of metals with continuous strain-hardening
7.3. Effect of overload magnitude on fatigue life of characteristics than has the single overloading. In the
aluminum riveted joints condition of periodic overloading, Eqs. (14) and (15)
could hardly be used to predict the FCI life accurately
Test results of the overload magnitude on the fatigue and a conservative life prediction would given [35,56].
life of aluminum alloy riveted joints [31] are shown in However, preliminary attempts were made to accurately
Fig. 16, where the pre-stress is higher than the maximum predict the FCI life of LY12CZ alloy notched elements
cyclic stress in fatigue tests, and could be taken as an in a corrosive environment taking the secondary over-
overload. load effect into account, and the predicted FCI life
It can be seen from Fig. 16 that the smaller pre-stress agrees well with test results [52,56]. More details on how
results in a decrease of the fatigue life and fatigue limit to take the secondary overload effect into account are
(i.e., z⬍0), while the greater pre-stress increases the given elsewhere [56]. However, the load pattern effect
fatigue life and fatigue limit (i.e., z⬎0). This may imply on FCI life and fatigue life, the applicability of Miner’s
that the same riveted joint of aluminum alloy would have rule and the life prediction model under VAL (in parti-
a shorter fatigue life in civil aircraft structures since civil cular, under random variable-amplitude loads such as
aircraft experience smaller overloads, while it would FALLSTAF) need to be investigated further.
have a longer fatigue life in a fighter aircraft since a
fighter aircraft experiences greater overloads [15]. Atten-
8. Concluding remarks and further research

In engineering applications, most elements in metallic


structures and machines contain a geometric disconti-
nuity — i.e., a notch in the broad sense — where the
stress concentration occurs and the fatigue crack
initiates. As a result, it is important to investigate the
fatigue behavior and life prediction model of notched
elements of metals.
From the test results and analysis in foregoing sec-
tions, it can be seen that accurate prediction of the FCI
life or fatigue life under VAL requires a perfect formula
for FCI life or fatigue life, and the rule for cumulative
Fig. 16. Overload effect on the fatigue life of riveted joints of alumi-
num alloy; (1) prestress of 240 MPa (35 ksi) applied to riveted joints, fatigue damage and the criterion for omitting small
(2) without applying prestress and (3) prestress of 123 MPa (18 ksi) loads.
applied to riveted joints [31,34]. Eq. (8) reveals the correlation between FCI life and
764 X. Zheng / International Journal of Fatigue 23 (2001) 751–766

element geometry (Kt), cyclic loading conditions (⌬S, of the above-mentioned metals. Moreover, the prob-
R), material constants [i.e., the FCI threshold, (⌬seqv)th, ability distribution of the material constants in Eq. (8) —
and the FCI resistance coefficient, C] and the overload i.e., C, (⌬seqv)th, n and z — should be investigated. In
effect factor (z). It provides the theoretical basis for particular, the procedures for predicting the probability
investigating the criterion for the omission of small loads distribution of FCI life under VAL need to be studied
and for defining the fatigue damage function. further.
The FCI threshold in Eq. (1) could be taken as the Furthermore, structural elements often operate under
criterion for omitting small loads in the load spectrum multi-axial stress conditions. Therefore, it is necessary
of notched elements of metals with discontinuous strain- to do more research on the cumulative fatigue damage
hardening characteristics, while the FCI threshold in Eq. rule, the criterion for omitting small loads, the life pre-
(8) could be taken as the criterion for omitting small diction model and the fatigue reliability assessment for
loads for notched elements of metals with continuous structure elements of metals, especially with continuous
strain-hardening characteristics in compiling the load strain-hardening characteristics, under multi-axial stress
spectrum, in predicting the FCI life and in the fatigue state with variable amplitude.
tests of full-scale structure members under random VAL.
Miner’s rule can be thought to be applicable for com-
puting the cumulative fatigue damage in FCI life predic- Acknowledgements
tion if a suitable expression for FCI life or fatigue life
is adopted and the load interaction effect is taken into The author greatly appreciates the graduate students
account. The overload effect factor, z in Eq. (8), could who made many contributions to the fatigue research
be adopted to characterize the load interaction effect. summarized in the present paper.
According to Miner’s rule, the fatigue damage function
could be defined by Eqs. (14) and (15), which can be
used to compute the cumulative fatigue damage. The References
accurate life prediction of notched elements can thus be
obtained when the value of the cumulative fatigue dam- [1] Zheng XL. Quantitative theory of metal fatigue. Xi’an: Pub-
age reaches 1.0; i.e., Dc=1.0. lishing House of Northwestern Polytechnical University (NPU),
Based on the above description, the fatigue formula 1994 [in Chinese].
[2] Zhao MP. Handbook of strain fatigue analysis. Beijing: Pub-
is critical in defining the criterion for omitting small lishing House of Science, 1988 [in Chinese].
loads, and the fatigue damage function and the cumulat- [3] Kim YH, Speaker SM, Gordon DE, Manning SD, Wei RP. Devel-
ive fatigue damage computation. Of course, the above opment of fatigue and crack propagation design and analysis
conclusions need to be checked further. methodology in a corrosive environment for typical mechan-
For notched elements and welded joints of metals with ically-fastened joints. In: Report NO. NADC-83126-60, vol.I
(AD-A136414), 1983.
discontinuous strain-hardening characteristics, the gener- [4] Kim YH, Speaker SM, Gordon DE, Manning SD, Wei RP. Devel-
ally applied life prediction model under VAL was opment of fatigue and crack propagation design and analysis
obtained through our recent research, and could be pro- methodology in a corrosive environment for typical mechan-
vided to structure designers. ically-fastened joints (assessment of art state). In: Report Nol
Aluminum alloys, titanium alloys and steels with NADC-83126-60, vol.2 (AD-A136415), 1983.
[5] Wanhill RJH, De Luccia JJ, Russo MT. The fatigue in aircraft
ultrahigh strength are widely used in aeronautical engin- corrosion testing (FACT) programme. In: AGARD Report No.
eering as the critical structural elements operating under 713 (AD-A208-359), 1989.
complicated load histories. The above alloys have con- [6] Buch A. Prediction of fatigue life under aircraft loading with and
tinuous strain-hardening characteristics and a positive without use of material memory rules. Int J Fatigue
overload effect factor. Compared with metals having dis- 1989;11(2):97–106.
[7] Dupart D, Davy A, Boetsch R, Boudet R. Fatigue damage calcu-
continuous strain-hardening characteristics, structure lation in stress concentration fields under uniaxial stress. Int J
members of aluminum alloys, titanium alloys and steels Fatigue 1996;18(4):245–53.
with ultrahigh strength will have a longer fatigue life [8] Schutz W. The prediction of fatigue life in the crack initiation
under VAL because of the positive overload effect fac- and propagation stages — a state of the art survey. Eng Fract
tor. On the other hand, the procedures for predicting the Mech 1979;11(3):405–21.
[9] Laz PJ, Craig BA, Rohrbaugh SM, Hillberry BM. The develop-
fatigue life and its probability distribution under VAL ment of a total fatigue life approach accounting for nucleation
are much more complicated. Therefore, it is necessary and propagation. In: Proceedings of the 7th International Fatigue
to perform more research on the rule for computing the Congress, FATIGUE ’99, Beijing, June 1999. High Education
cumulative fatigue damage and the criterion for omitting Press, EMAS, 1999:833–8.
the small loads, as well as the procedures to determine [10] Nowack H. Development in variable amplitude prediction
methods for light weight structures. In: Proceedings of the 7th
the FCI life curves with given survivability and corre- International Fatigue Congress, FATIGUE ’99, Beijing, June
sponding expressions after overloading of different 1999. High Education Press, EMAS, 1999:991–1000.
amplitudes of (⌬seqv)OL for notched elements and joints [11] Zheng XL, Cui TX. Eng Fract Mech 1989;34(5/6):1005–11.
X. Zheng / International Journal of Fatigue 23 (2001) 751–766 765

[12] Forman RG. Eng Fract Mech 1972;4(2):333–45. [39] Zheng XL, Ling C. On the expression of fatigue crack initiation
[13] Barnby JT, Dinsdale K, Holder R. Fatigue crack initiation. In: life considering the factor of overloading effect. Eng Fract Mech
Proceedings of Conference on Mechanics and Physics of Frac- 1988;31(6):959–66.
ture, Cambridge, 1975, (paper 26). [40] Zheng XL. Modeling fatigue crack initiation life. Int J Fatigue
[14] Rabbe P. L’amorsage des fissures de fatigue. In: La fatigue des 1993;15(6):461–6.
materiaux et des structures. Montreal: Les Presses de l’Universite [41] Coffin LF. Fatigue in machine and structures-power generation.
de Montreal, 1981:71–105 [in French]. In: Fatigue and microstructure. Metals Park (OH): American
[15] Van Dijk GM, De Jonge JB. Introduction to a fighter aircraft Society for Metals, 1978:4–7.
loading standard for fatigue evaluation (FALLSTAFF). In: [42] Rolfe ST, Barsom JM. Fracture and fatigue control in structures.
Report NLR MP 75017U. The Netherlands: National Aerospace Englewood Cliffs (NJ): Prentice Hall, 1978:205–30.
Laboratory, 1975. [43] Standnick ST, Morrow JD. Techniques for smooth specimen’s
[16] Chen C-H, Coffin LF. Fatigue Fract Eng Mater Struct simulation of the fatigue behavior of notched members. ASTM
1998;21:1–15. STP 515, Philadelphia (PA): American Society for Testing and
[17] James MN, Peterson AE, Sutcliffe N. Int J Fatigue Materials, 1972:229–52.
1997;19:125–33. [44] Zheng XL, Zhang GZ. On the fatigue formula under strain cyc-
[18] Lazzarin P, Tovo R, Meneghetti G. Int J Fatigue 1997;19:647–57. ling. J Northwestern Polytech Univ (NPU) 1986;2(2):223–9 [in
[19] Kattus JR. Aerospace structural metals handbook, vols. 1 Chinese].
(Ferrous alloys) and 2 (Nonferrous alloys). Mechanical Data [45] Zheng XL, Ling C, Zheng ST. An approximate formula for calcu-
Center, 1970 and 1975. lating local strain range. In: Advanced computational methods,
[20] Handbook of fatigue behaviour of aeronautical metallic materials. Proceedings of the 1st International Conference on Computer
Beijing: Beijing Research Institute of Aeronautical Materials, Aided Assessment and Control of Localized Damage
1981 [in Chinese]. (CAACOLD), Portsmouth, June 1990, 1990:253–60.
[21] Conway JB, Sjodahl LH. Analysis and representation of fatigue [46] Ling C, Zheng XL. Fatigue crack initiation life of LC4CS alumi-
data. Materials Park (OH): ASM International, 1991. num alloy sheets and its overloading effect — a general formula
[22] ECCS recommendations for the fatigue design of steel structures. for FCIL of aluminum alloy. J Northwestern Polytech Univ
Institute of Metal Construction (ICOM) of Swiss Federal Institute (NPU) 1991;9(1):30–5 [in Chinese].
of Technology. Lausanne, Switzerland: ECCS, 1985. [47] Lu B, Zheng XL. Effect of microstructure on fatigue crack
[23] Zhao SB, Wang ZB. Fatigue design. Beijing: Machinery Industry initiation and propagation of 16Mn steel. Metall Trans
Press, 1982 [in Chinese]. 1989;20A:413–6.
[24] Shi YJ, Yan YM, Li ZR, Shi ZJ, Hou WW. Assessment of [48] Lu BT, Zhang JG, Ling C, Zheng XL. Experimental study on the
remaining life of steel beams of Chang-Tai-Guan bridge on the fatigue performance of 15MnVN steel. Mech Strength
line from Beijing to Guangzhou. Technical Report. Beijing: Insti- 1993;14(4):70–5 [in Chinese].
tute of Railway Engineering, Chinese Academy of Railway [49] Lu B, Zheng XL, Li D. Fatigue crack initiation and propagation
Science, 1990. in butt welds of an ultrahigh strength steel. Welding J
[25] Tomita Y, Fujimoto Y. Prediction of fatigue life of ship structural 1993;72(2):793–805.
members from developed crack length. In: Fatigue life, Proceed- [50] Lu B, Zheng XL. Predicting fatigue crack initiation life of an
ings of ASM Conference, Salt Lake City, UT, December 1985, aluminum alloy below room temperature. Fatigue Fract Eng
1985:37–46. Mater Struct 1992;15(2):1213–21.
[26] Heuler P, Seeger T. A criterion for omission of variable ampli- [51] Tay CJ, Tay TE, Shang HM, Yap M. Fatigue crack initiation
tude loading histories. Int J Fatigue 1986;8(4):225–30. studies using a speckle techniques. Int J Fatigue
[27] Fatemi A, Yang L. Int J Fatigue 1998;20:9–34. 1994;16(6):423–8.
[28] Zheng XL. On the criterion for omitting small loads in life predic- [52] Wang R. Quantitative description of corrosion fatigue behaviour
tion. Paper presented at 8th National Conference on Fatigue, of aluminum alloys. Ph.D. thesis. Xi’an: Northwestern Polytech-
Xi’an, October 1997 [in Chinese]. nical University, 1994 [in Chinese].
[29] Crews JH Jr. ASTM STP 467, Philadelphia (PA): American [53] Zheng XL. Further study on fatigue crack initiation life (II) —
Society for Testing Materials, 1967:37–52. prediction of fatigue crack initiation life. In: Goel VS, editor.
[30] Forrest G. J Inst Met (London) 1946;72:498–527. Fatigue life, Proceedings of ASM Conference, Salt Lake City,
[31] Smith CR. Report ARL-64-55, Ohio, 1964. UT, December 1985, 1985:1–4.
[32] Raithby KR, Longson J. Some fatigue characteristics of a two- [54] Zheng XL, Cheng DG, Ling C. Overloading effect on fatigue
spar light alloy structure. In: Aeronautical Research Council, Cur- crack initiation in LY12CZ alloy and a new model for life predic-
rent Paper 258. London: HMSO, 1956. tion. J Northwestern Polytech Univ (NPU) 1990;8(2):199–208
[33] Boissonat J. Experimental research on the effects of a static pre- [in Chinese].
loading on the fatigue life of structural components. In: Barrois [55] Zheng XL, editor. Mechanical properties of materials. 2nd ed.
W, Ripley EL, editors. Fatigue of aircraft structures. Oxford: Per- Xi’an: NPU Press, 2000 [in Chinese].
gamon Press, 1963:97–113. [56] Zheng XL, Wang R. Overload effect on corrosion fatigue crack
[34] Zheng XL. Overloading — an effective way to increase fatigue initiation life and life prediction of aluminum notched elements
life. Mech Strength 1982;2:15–22 [in Chinese]. under variable amplitude loading. Eng Fract Mech
[35] Schijve J. Endurance under program-fatigue testing. In: Plantma 1999;63:557–72.
FJ, Schijve J, editors. Full-scale fatigue testing of aircraft struc- [57] Wei JF. On models for predicting probability distribution of
tures. London: Pergamon Press, 1961:41–59. fatigue life under variable amplitude loading. Ph.D. thesis. Xi’an:
[36] Webber D. ASTM STP 648, Philadelphia (PA): American Northwestern Polytechnical University, 1996.
Society for Testing and Materials, 1978:78–88. [58] Zheng XL. Overload effect on fatigue behaviour and life predic-
[37] Smith IFC, Hirt MA. A review of fatigue strength improvement tion model of low carbon steels. Int J Fatigue 1995;17(5):331–7.
methods. Can J Civ Eng 1985;12:166–83. [59] Zheng XL, Cui TX, Lu B, Lu XY, Ling C. Fatigue tests and life
[38] Zheng XL. A further study on fatigue crack initiation life — prediction of butt welds without crack-like defect. Int J Fracture
mechanical model for fatigue crack initiation. Int J Fatigue 1994;69:275–85.
1986;8(1):17–21. [60] Yan JH, Zheng XL, Zhao K. Prediction of fatigue life and its
766 X. Zheng / International Journal of Fatigue 23 (2001) 751–766

probability distribution of notched friction welded joints under life of ceramic with given survivability. Eng Fract Mech
variable amplitude loading. Int J Fatigue 2000;22(6):481–94. 1996;53:49–55.
[61] Li Z. On the models for predicting fatigue life of notched [73] Zhou J, Jao SR, Si SM, Zheng XL. Fatigue behaviour of
elements and welds of low-carbon low alloy steels. Ph.D. thesis. carbon/bismilimide under tension–tension loading. Acta Aeronaut
Xi’an: Northwestern Polytechnical University, 1994. Astronaut Sinica 1990;11(5):A305–8 [in Chinese].
[62] Zheng XL, Li Z, Lu B. Prediction of probability distribution of [74] Gurney T, Maddox S. In: Hirt MA, editor. Remaining fatigue
fatigue life of 15MnVN steel notched elements. Int J Fatigue life of steel structures, IABSE Workshop Proceedings, Lausanne,
1996;18(2):81–6. 1990:189–98.
[63] Zheng XL, Jao SR, Zhang JG, Luo B, Gao PX. Plastic defor- [75] Zheng QX, Zheng XL. An expression for fatigue life and prob-
mation, microcrack formation and fatigue crack initiation life of ability distribution of fatigue strength under repeated torsion. Eng
30CrMnSiNi2A high strength martensite steel. J Northwestern Fract Mech 1993;44(4):521–8.
Polytech Univ (NPU) 1979;1:79–86 [in Chinese]. [76] Zheng XL, Lu BT, Ling C, Chen DG. On the rule for accumulat-
[64] Zheng XL, Zhang JG. Effect of overloading direction on fatigue ive fatigue damage. Acta Aeronaut Astronaut Sinica
crack initiation life of 30CrMnSiA steel. J Northwestern Polytech 1993;14(10):484–9 [in Chinese].
Univ (NPU) 1987;5(2):149–56 [in Chinese]. [77] Richardt FE, Newmark NM. A hypothesis for the determination
[65] Zheng XL, Li Z, Shi YJ, Yang YM, Shi ZJ. Fatigue performance of cumulative damage in fatigue. Proc ASTM 1948;48:767–801.
of old bridge steel and the procedures for life prediction with [78] Ling C, Zheng XL. Overloading effect upon fatigue life of 16Mn
given survivability. Eng Fract Mech 1996;53(2):251–62. steel butt welds. Hanjie Xuebao (J Welding) 1991;12(4):247–51
[66] Yan JH, Zheng XL, Zhao K. Experimental investigation on small- [in Chinese].
load-omitting criterion. Int J Fatigue 2001;23(5):403–15. [79] Dupart D, Campassens D, Balzano M, Boudet R. Fatigue life
[67] Frost NE, Marsh KJ, Pook L. In: Metal fatigue. Oxford: Claren- prediction of interference fit fastener and cold worked holes. Int
don Press, 1974:20–2. J Fatigue 1996;18(8):515–21.
[68] Cai QK. Theory of fatigue and fracture of metals. Shenyang: [80] Ling C, Zheng XL. Effects of cold expansion of hole on fatigue
Press of Northeast Institute of Technology, 1989 [in Chinese]. crack initiation location and life of LY12CZ alloy. Fatigue Fract
[69] Tanaka S, Ichikawa M, Akita S. A probabilistic investigation of Eng Mater Struct 1992;15(3):241–7.
fatigue life and cumulative cycle ratio. Eng Fract Mech [81] Thauvin G. Influence of surface roughness on fatigue and cor-
1984;20(3):501–20. rosion fatigue behaviour of 0.12C–12Cr–2.5Ni–1.8Mo–0.3V
[70] Wei JF, Zheng XL. Statistical correlation of fatigue life of steel. In: Proceedings of the 7th International Fatigue Congress,
15MnVN steel notched bar under variable amplitude tension–ten- FATIGUE ’99, Beijing, June 1999. Higher Education Press,
sion block loading. Theor Appl Fract Mech 1997;28:51–5. EMAS, 1999:2347–52.
[71] Sakai T, Fujitani K. A statistical aspect on fatigue behavior of [82] Ling C, Zheng XL. Method for predicting fatigue crack initiation
alumina ceramics in rotating bending. Eng Fract Mech life of elements subjected to CEH under variable amplitude load-
1989;32:499–508. ing. Acta Metall Sinica 1991;27(1):A75–7.
[72] Zheng QX, Zheng XL, Wang FH. On the expressions of fatigue