Sunteți pe pagina 1din 20

Effects of defects, inclusions and

inhomogeneities on fatigue strength


Y. Murakami* and M. Endo*
*Department of Mechanical Science and Engineering, Faculty of Engineering,
Kyushu University, Higashi-ku, Fukuoka 812, Japan
*Department of Mechanical Engineering, Fukuoka University, Jonan-ku, Fukuoka
814-01, Japan
(Received 19 July 1993)

Models predicting the effects of defects, inclusions and inhomogeneities on the fatigue strength of
metals are reviewed. Their historical development is discussed, from the Isibasi model and the Frost
model to recent models based on fracture mechanics. The method of application of existing models
and the data necessary for prediction are explained. Fracture mechanics models mostly employ the
value of threshold stress intensity factor range AKth ~¢ for a long crack as the basic data but usually
apply only to two-dimensional cracks or two-dimensional defects. Furthermore, a careless assumption
of the value of AKth ~c will result in a large prediction error of fatigue limit trw. Models that do not
use AKth ~¢are also introduced and discussed. Several models that consider the effects of non-metallic
inclusions and inhomogeneities are also introduced, and the difficulties of prediction are indicated.
The importance of understanding the fatigue limit phenomenon for the accurate prediction and choice
of relevant models is emphasized, with regard to the existence of small non-propagating cracks and
the statistical nature of the behaviour of small cracks and microstructures.

(Keywords: fatigue thresholds; fatigue limit prediction; small cracks; small defects; non-metamc inclusions;
inhomogeneities; models; fracture mechanics; engineering applications)

The objective of this review is to present a guide for steels but rather the effect of notches as structural
company engineers who are seeking a way to predict discontinuities or cracks formed by fatigue. This is
the fatigue strength of structural components containing understandable in that the materials treated in Frost et
small defects, inclusions and inhomogeneities. Various al.'s studies were comparatively low-strength materials
models will be summarized to clarify whether or not the with ultimate tensile strength tru - 80-1000 MPa. How-
available data are sufficient for using the existing models ever, their approach was innovative at the time because,
proposed by previous researchers. (The models and in their evaluation of fatigue strength, the presence of
approaches reviewed here are limited to those predicting defects was assumed. By contrast, at the same time in
the fatigue limit stress of metals containing small defects, Japan, Isibasi ls,19 performed a series of studies on cracks
inclusions and inhomogeneities.) The summary will also and was encouraging Japanese researchers to consider
define the kinds of basic data necessary for users of fatigue strength from the viewpoint of crack propagation.
these models, the assumptions on which these models Fracture mechanics was thus born 2°. Paris 21 applied
are based, whether these models are applicable to three- it to fatigue crack propagation, and thereafter the
dimensional as well as two-dimensional defects, and progress of research leading to short cracks from long
their range of application. Depending on the situation, cracks was a natural progression, although there was of
engineers will be able to choose an appropriate model course a problem in that linear fracture mechanics
from those listed in this overview to solve problems of cannot be applied to the small crack problem as it
fatigue. stands. In recent years, the importance and difficulty of
The presence of small defects in materials attracted
the evaluation of small cracks and short cracks has
much attention in early fatigue studies, when steelmaking
attracted the attention of fatigue researchers, and many
technology was still in its infancy. At that time, many
papers and textbooks have reviewed the fatigue crack
large non-metallic inclusions or pores were commonly
growth threshold for small or short cracks 22-35. The
present in materials, which caused not only a serious
reasons for the progressive activities in this field may
reduction in fatigue strength of hard steels but also
be summarized as follows.
considerable scatter 1. Detailed studies on defects and
inclusions were conducted by many researchers 2-11 at
this time but none proposed models in an explicit form. 1. Small cracks and small defects are not detected in
A series of studies by Frost eta/. 12-17, which were design, and in addition the incubation period for the
conducted subsequently, did not consider the effects of propagation of a small crack is long.
non-metallic inclusions or defects inherently existing in 2. Even if small cracks or defects are detected and their

0142-1123/94/030163-20
(~) 1994 Butterworth-Heinemann Ltd Fatigue, 1994, Vol 16, April 163
Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

sizes and shapes are known, the conventional fracture CLASSIFICATION OF MODELS AND THEIR
mechanics approach is not applicable. APPLICATION METHODS
3. As the size of a small crack or a small defect
approaches the same order of magnitude as that Table 1 summarizes typical existing models for small
of the material microstructure, crack behaviour is cracks and small defects. The basic data required for
influenced by statistical factors of the microstructure. calculating the fatigue limit o.w with these models are
4. With the progress of steelmaking technology, quanti- classified as (a) material or (b) geometrical parameters.
tative evaluation of the effects of non-metallic In the fifth column of Table 1, the necessary data for
inclusions on fatigue strength has become important prediction are listed in the separate categories defined
below.
from the viewpoints of both material evaluation and
quality control of products 35. (a) Material parameters can be listed as follows
(numbers in brackets correspond to the models listed in
Table 1).
Summarizing existing models from a historical point
of view, they may be loosely classified into the following 1. Category M1
three groups. Basic strength properties obtained by standard
static material testing or representative sizes of
1. Frost's model and other similar models. Frost 12 microstructures:
presented an empirical equation of the form o~,,,l = C o-u Ultimate tensile strength (5,6)
in an investigation of the relationship between fatigue o.v Yield stress (4)
limit o'wand crack length l. Kobayashi and Nakazawa36 o'0.2 0.2% offset yield strength (10)
conducted a similar study and modified Frost's model Ha The Brinell hardness (5)
to oAwl= C. Murakami and Endo 37 proposed the Hv The Vickers hardness (14,16)
equation 0 ~ = C for small surface defects with d Grain size (11,17)
the projection area of defects (area), where n = 6.
Murakami and Endo 38 and Murakami and Matsuda39 2. Category M2
pointed out that the change in exponents, Basic strength properties obtained by fatigue test:
n = 3 --->4 --->6, in the models of Frost, Kobayashi AKeffth Threshold effective stress intensity
et al. and Murakami et al. is due to the difference in factor range (18)
size ranges of the defects investigated in their studies. AKi th Intrinsic threshold stress intensity factor
Frost 12 performed experiments on cracks or notches range (17)
Agth Ic Threshold stress intensity factor range
with depths l = 100-20900 I~m and Kobayashi et
al. 36 with l = 30-1100 ixm, while Murakami and for long cracks (7-11,15)
Endo 37 investigated smaller cracks. Here it should be o'~ Fatigue limit of plain (smooth)
noted that a short edge crack with depth l = 0.1 mm specimens (1,5-11,15)
(100 ~m) is equivalent (though it appears small) to o'Yc Cyclic yield stress (10)
a three-dimensional crack of Marea = 316 ixm from . Category M3
a stress intensity factor point of view37. For small Several models need other material data, which
cracks of ~10 I~m < V'a--Te-a < 1000 ~m, the value of must be obtained from data of categories M1
n tends to - 6 . 38 and M2, or must be determined by additional
2. Approaches based on fatigue notch factor. There are fatigue tests:
studies by MitchelP°,41 and Nordberg 42 in which C, n Material constants (2,3,12)
Peterson's equation43 is applied to small cracks, C' Material constant in Peterson's formula,
small defects and non-metallic inclusions, although which is approximated by
Peterson's equation was originally proposed for large C' = 0.0254(2070/o'u) 1.8
notches. = 0.0254(600/HB) 1.8
3. Fracture mechanics approaches. The fracture mech- where C' is in mm, o'u in MPa and HB
anics approaches to small crack problems began with in kgf mm -2 40--42 (5,6)
Kitagawa and Takahashi's study44. These approaches a Factor that accounts for the ease of
may be simply classified into that based on AKth38'44-57 cross-slip (17)
and that centred on the cyclic plastic zone size Co Material constant used in Isibasi's
calculation58 based on the Dugdale model59. Concern- model (Co concept) (1)
ing AKth, it has been agreed that the value of AKth sc k Material constant; rate of crack closure
for short cracks or small cracks is smaller than that development (17)
of AKth ~c for long cracks. Hence in these models lao Critical transition length of la which
AKth is generally expressed as a function of AKth ~, separates behaviour controlled by short
the fatigue limit o'~ of plain specimens, the fictitious cracks from that by long cracks (18)
crack length lo, the non-damaging crack length ll and lo Fictitious crack length calculated by
so on. In these models, researchers employ their Equation (8) below, with AKth lc and
own hypotheses based on a fracture-mechanics-like o.~o (see Figure 3) (7,15)
concept. ll Critical length of a non-damaging crack
(see Figure 3) (15)
Models on these lines cover mostly two-dimensional /2 Minimum length of crack to which
cracks or two-dimensional notches only. There are few conventional linear fracture mechanics
models in which an equation covering three-dimensional are applicable (D. Taylor and
cracks or defects is presented, and its applicability Knott65,66 showed that/2 is
verified by experiments on various materials. approximately related to grain size or

164 Fatigue, 1994, Vol 16, April


Table 1 Comparison of models predicting the fatigue strength effect of small defects, non-metallic inclusions and inhomogeneities

Performance rn

Fatigue test Prediction of


Data necessary (necessary data Stress Evaluation allowable
Model Ref. Model concept and prediction for prediction to be concentration Type of defect of mean critical defect Prediction
"'Is
no. Researcher (year) approach and category determined) factor, Kt investigated stress effect size error

1 Isibasi 60 2-D model M2: Necessary Necessary for 2-D notches No No Undefined
(1948) eo concept: trio (Crwo and Co) notches
19 At fatigue limit, the stress at a M3:
(1967) distance Co (material constant) from eo 5"
t~
the notch root or crack tip is equal GI:
to the fatigue limit t r ~ of plain 1
specimens
o~
2 Frost et al. 12-17 2-D model M3: Necessary Not necessary 2-D cracks and Yes (see Yes (see Refs +7% t~
(1959-1974) OZwI = C C (C) 2-D notches Refs 12, 16 12 and 17)
GI: and 17)
1 5"
~r

3 Kobayashi and 36 2-D model M3: Necessary Not necessary 2-D cracks and No No Undefined
Nakazawa (1967) c#~l = C C (C) 2-Dnotches tt3
GI:
l

4 de Kazinczy 61 3-D model MI: Necessary Not necessary Pores and No No Undefined
(1970) For medium-strength cast steels, try (~r~) mieroshrinkage
crw = Owo/(1 + K':V/-D) GI: cavities
where K ' = trv/Kg D
G+M:
/q

5 Mitchell 40 3-D model M1: Necessary Necessary Inclusions, No No Undefined


(1977) Prediction based on Peterson's GrU (MPa) or (trwo) graphites, gas
41 formula 43 H B (kgf mm -2) pores and
tQ (1979) tr..,, = O ' w o / [ l + ( K t - 1 ) / ( l + C ' / p ] M2: microshrinkage
t- where trw is in MPa and C' = tr.o (MPa) cavities
@ 0.0254(2070/tr.) l'a or C' =
..t 0.0254(600/HB)l.s GI:
tD K, and p (mm)

< 6 Nordberg 42 3-D model MI: Necessary Necessary Non-metallic No No Undefined


O (1981) Prediction based on Peterson's trU (MPa) (try) inclusions
f°rmula43 M2:
ta
o'w = t r ~ l [ l + ( K , - 1 ) / ( l + C ' / p ) ] trio (MPa)
where Crw is in MPa and C' =
0.0254(2070/~ru)l.s Ol:
"1. K, and p (mm)

Continued over
i
"11 Table 1 Continued

tt~
t- Performance

Fatigue test Prediction of t~


tD
Data necessary (necessary data Stress Evaluation allowable
Model Ref. Model concept and prediction for prediction to be concentration Type of defect of mean critical defect Prediction
no. Researcher (year) approach and category determined) factor, K, investigated stress effect size error
0D
7 Kitagawa and 44 2-D model M2: Necessary Not necessary 3-D cracks No No Undefined ta
Takahashi (1976) The Kitagawa-Takahashi diagram 0"*0 and AKth lc (0".,o, AKth |¢
45 (AKin VS a relation on a log-log GI: and AKth)
"O (1979) graph)
--I. aorl
G+M:
AK~h
~r
R.A. Smith 46 2-D model M2: Necessary (0"*0 Not necessary Cracks No No Undefined
(1977) Smith's model: 0".0 and AKth lc and AK,h l~) 0
Agth lc constant line and 0"*0 GI: ¢b
constant line on a log-log graph l ¢b

El Haddad, 47 2-D model M2: Necessary (0"*0 Not necessary 2-D cracks and No No Undefined t~
Topper and (1979) 0"w = 0 " . 0 ~ ) 0"wo and AKth lc and nK~h I~) 2-D notches O
Smith where GI:
lo = (AKth lo/0".,o)2/z" l

10 Usami and 58 3-D model MI: Necessary (0-~, Not necessary 2-D and 3-D Yes Yes Undefined
0"0. 2
Shida (1979) Threshold condition: at fatigue Agth I~ and cracks, surface o~
limit, cyclic plastic zone size at a M2" rpc(w)) roughness,
0"*0"and A K t h lc
crack tip r ~ = material constant microshrinkage
M3: cavities and
rp,:(w) inclusions
For instance, when R I> 0
r~t~) = /oq[Sec(IrA0"J40"vo)-l] G+M:
where 0"v¢ = 0-70"o.2 and loq
l~q = (K/0"o)2/qr

11 Tanaka, Nakai 62 3-D model MI: Necessary (0-*0, Not necessary 2-D and 3-D Yes Yes Undefined
and Yamashita (1981) 0-. = o".0 { 1 + [ ( 2 / ~ ' ) ~ x d 0-'~. K~ and cracks
cos-1//Q~)- 11/(1 +.L.~K m/ M2: AKth It)
o1,,Vm,-,o)} x V ~ l b 0-*0 and A g t h lc
where
too = d / 2 a n d b = l + too M3:
o~fr and K~
GI:
a
G+M:
Ioq
Table 1 Continued

Performance

Fatigue test Prediction of


Data necessary (necessary data Stress Evaluation allowable
Model Ref. Model concept and prediction for prediction to he concentration Type of defect of mean critical defect Prediction
no. Researcher (year) approach and category determined) factor, K, investigated stress effect size error

12 Murakami and 37 3-D model M2: Necessary Not necessary 3-D holes No Yes Undefined
Endo (1983) The ~ parameter model n and C (n and C)
O~w~ = C GI:
where (rw = rotating bending or
tension-compression fatigue fimit

13 D. Taylor 63 3-D model GI: Necessary Not necessary 3-Dcracks No No Undefined


(1984) The P - a plots approach: Prediction a ( P-a plots)
64 of the Kitagawa-Takahashi diagram G+M:
(1992) with statistical distribution P - a plots
O

14 Murakami and 38 3-D model MI: Not necessary Not necessary 2-D and 3-D No Yes -+10% for
Hv (kgf mm -e) cracks, 2-D and Hv =
¢b
Endo (1986) The ~ parameter model
AK,~ = 3.3 x 1 0 - 3 X ( H v + 1 2 0 ) 3-D notches, 3- 70-720
GI:
(Varea) 1/3, ~w = 1.43(Hv + 120)/ aX/'~rea (~m) D holes,
( V area) 1/6 Vickers
Upper hound of o,,: indentations and
~,~ ~ 1.6 Hv non-metallic
where AKth is in MPa m ~/: and ~.. inclusions
~r..~ are in MPa

15 L u k e , Kunz, 48 2-D model M2: Necessary (cry. Necessary for Semicircular No Yes Undefined
Weiss and (1986) For c r a c k s ~ c e , cr~ and AKth Ic AKth ,c and ll) notches cracks and 2-D
-rl
Stickler 49 O'w = O'.o ~lo/(l-l~ +lo ) GI: notches
(io (1989) where l, Kt and F °.
e- lo = (AK~h Jcr~o)z/w
For notches with a semicircular G+M:
crack at the notch root, ll
¢.D
¢.D ~rw = Or~o~/1+4.5(/~/p)/K,
where 1~ must be determined by
< curve-fitting
O
m
Continued over

"o

._L

o)
g
t~
"TI

Table 1 Continued 0
e=
Performance

~D Fatigue test Prediction of


p Data necessary (necessary data Stress Evaluation allowable
< Model Ref. Model conceptand prediction for prediction to be concentration Type of defect of mean critical defect Prediction
0 nO. Researcher (year) approach and category determined) factor, Kt investigated stress effect size error
t~
16 Murakami, 50 3-D model MI: Not necessary Not necessary 3-D cracks, Yes Yes -+15% for
Uemura, (1990) The ~ parameter model Hv (kgf mm -2) non-metallic H,,=
"O Natsume and For surface defects: GI: inclusions, 100-740
Miyakawa ~r = 1.43(Hv+120)/( aV'a-~)~:6 x (~m) carbides and
[(1-R)/2] ~ inhomogeneities
For internal defects:
~ = 1.56(Hv+120)/(aV~-~) a/6 x O
[(1-R)/2] °
where ~r is in MPa and a = 0.226 O
+ Hv x 10-4 t~
or. and stress ratio R are predicted
by trial-and-error method

17 Abdel-Raouf et 51-55 2-D model MI: Necessary Not necessary Surface cracks Yes No Undefined
al. (1991-1993) For smooth specimens: d (AK~h, AKi th, a
A~,h = Ho, AKi ,~/(FQ~X/--~) M2: and k)
For notched specimens: AK~ ,h
AO'th = HclAgi th/(FgtQe~/Tra)
In these equations, M3: cb
Q;" = Ae/Ae = l + q e x p ( - a a / d ) , a and k
and H d = A K t h / / ~ l d ~ i t h = AKth/ GI:
[AK,h + (AKth - AKi th) X a, K~ and F
exp(-kao)] G+M:
where ae = a - 0.4d .°
AK,~

18 Melander et al. 56 3-D model M2: Necessary Not necessary Inclusions No Yes Undefined
(1992) A~w = AKcff th/[ U~/ pF(~o) ] AKeff th (AKeff ~h and
57 where M3: basic data for
(1992) U = ( U ' - U o ) e -k'~ + Uo, l.o calculating U'
Uo = 1 / ( l - R ) and fl = la/p and k' by FEM
In the above equations, U',k': GI: analysis)
fitting parameters and F(flo): shape la and p
factor for la = l~o (critical transition (radius of a
length of l~) pore or an
inclusion)
G+M: rn
U
(or U' and k')
Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

cyclic plastic zone size) (Figure 3)


rpc(w) Cyclic plastic zone size at fatigue limit I

(material constant) (10)


oh) Slip band zone size, which is assumed
to be d/2 (11)
Frictional stress of dislocation motion
n
(11)
Critical value of microscopic stress
intensity factor (11)
I 21 I
(b) Geometrical parameters and other parameters
can be listed as follows.

.
(a)
Category G1
Geometrical parameters:
a Depth or half of surface length of a
three-dimensional surface crack
(7,11,13,17)
D Defect size defined by the diameter of
the smallest circle enclosing a defect
(4)
Kt Elastic stress concentration factor
(5,6,15,17)
l Half of the length of a two-dimensional
crack in a wide plate, or the depth of a
two-dimensional surface crack or notch
(Figure 1) (1-3,7-9,15)
la Length of an annular crack surrounding
a spherical pore (18)
X/area Square root of projection area of a
(b) (c)
defect (see Figure 6) (12,14,16) Figure 1 Definition of dimensions of two-dimensional crack or
p Notch root radius of a notch (see notch: (a) crack in a wide plate; (b) surface crack; (c) surface
notch
Figure 1), pore or inclusion (5,6,18)
F Shape factor of a crack (15,17)
dimensional cracks in a wide plate, Isibasi's model is
. Category G+ M reduced to the form*
Data which must be calculated from a
combination of data of Categories G1, M1 and V2/eo + e2o (1)
M2, or other data which must be determined by trw = trwo 1 + eo
experiment:
AKth Threshold stress intensity factor range Though simple in concept, this model allows company
(7,17) engineers to determine trw from the evaluation of eo
/eq Equivalent crack length = (1(/O'o)2/I1"; from only a few specimens.
K = stress intensity factor at the tip of The value of eo decreases with increasing material
a three-dimensional crack; tro= strength. Figure 2 demonstrates Equation (1) by the
applied nominal stress (10,11) relationship between trw/O'woand crack half-length l for
Kg Constant, dependent on defect the cases of eo = 10, 20, 40 and 80 Ixm. It is interesting
geometry (4) that this model gives, in spite of its simplicity, essentially
P-a Probability of crack growth P as a the same result as that derived from modem fracture
plots function of crack length a (13) mechanics. For instance, E1 Haddad et al.'s model 47 (9),
Usami and Shida's modeP 8 (10) and Tanaka et a/.'s 62
U Crack-opening ratio (18)
(11) must eventually employ a constant similar to Co,
and therefore their models essentially belong to the
Isibasi's model 19,6° (1) same group as Isibasi's model.

Isibasi thought that the fatigue limit O'w of a notched Frost's model 12-17 (2)
or cracked specimen is determined under the condition Frost's model is expressed by the empirical formula
where the stress at a distance eo from the notch root or
crack tip is equal to the fatigue limit t r ~ of plain trawl= C (2)
specimens (Co concept), where eo is a material constant. where: trw = fatigue limit; and l = crack depth (Figure
Isibasi's model ~ was proposed in 1948 and it has been 1). This model is as simple as Isibasi's model t9,6° (1),
used in Japan to evaluate notch and size effects in
fatigue. Thereafter, many models and associated exten-
sive theories were proposed. However, the authors are *Although Isibasi's idea was too early for Equation (1) t o be
derived by fracture mechanics, we can derive it considering the
not aware whether such models were used by company meaning of Isibasi's eo concept explained for a notch in the
engineers in place of Isibasi's simple model. For two- literature19,6o.

Fatigue, 1994, Vol 16, April 169


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

Mitchell's model 4°,41 (5) and Nordberg' s model 42 (6)


Mitchell and Nordberg applied the following Peterson's
formula43 to the fatigue strength prediction of metals
o lem containing non-metallic inclusions or small defects:
orwO
' ' 1\ ~w -- 1 + ( g t - 1 ) / ( 1 + C'/p) (5)
0.1 Peterson's formula employs the elastic stress concen-
tration factor Kt, the tip radius p of a geometric
discontinuity, the fatigue limit trwo of a plain specimen
and a material constant C', which for ferrous-based
metals can be approximated in millimetres by
¢ ¢ ¢
0.01 t t IIItl I I I I I ,,,I . . . . . . i,I C' =0.0254(2070) 1"8
0.01 0.1 1 \oruJ
Crack length, l mm or
Figure 2 Relationship between ~r.hr.o and 1 estimated by Isibasi's / 6 0 0 \ l"s
model C' = 0.0254 [ ~ ) (6)

though it was proposed before the development of where oru is in MPa.


fracture mechanics. His early paper ~2 shows that for In general, non-metallic inclusions have various irregu-
mild steels the value of C is almost independent of the lar shapes. The value of Kt for such inclusions changes
material. For a greater variety of materials, however, significantly owing to a slight deviation from simple
the value of C is material-dependent13-17 and must be geometry such as a sphere or ellipsoid, and clearly
determined by experiment. For 12 materials with tensile therefore the difficulty in obtaining the exact value of
strength o'o ranging from 77 to 925 MPa, the value of Kt is a disadvantage for practical use of this model. In
orw can be estimated to an accuracy of -+7% 17. order to avoid this problem, Mitchell41 assumed all
It is clear from the form of Equation (2) that Frost defects to be equivalent to a hemispherical pit with
et al. did not think that Equation (2) was valid for a Kt = 2.5. Table 2 shows the comparison between the
very small value of crack length l. prediction by Mitchell's model and experiments.

Kobayashi and Nakazawa' s model36 (3) Kitagawa and Takahashi' s model 44.45 (7)
This is a modification of Frost's m o d e l 12-17 (2) and Kitagawa and Takahashi were the first to characterize
is expressed as quantitatively the fatigue threshold behaviour of small
cracks, namely that AKth decreases with decreasing
o~wl= C (3)
crack size. Their study contributed much towards the
As already mentioned in the introduction, the discrepancy development of many subsequent studies on small cracks
between the exponents n = 3 (Frost et al. 12-]7) and 4 and small defects. Figure 3 shows typical short crack
(Kobayashi eta/. 36) is due to the range of crack length data in the form of a plot of Atrth as a function of crack
examined in their experiments. This will be discussed in length l, which is widely known as the
more detail later because this probelm is related to the Kitagawa-Takahashi diagram or the Kitagawa plot. In
validity of the models: that is, to the crack length range order to combine the data for several materials, they
to which Equations (2) and (3) are applicable. proposed to summarize the data ranging from long to
short cracks with the normalized stress intensity factor
De Kazinczy's modeP 1 (4) AKth/AKth lc and normalized crack size 1/lo4s. Here,
De Kazinczy proposed the following equation: AKth ,¢ is the threshold stress intensity factor for a long
crack and 1o is the fictitious crack length calculated by
{Two Equation (8) below. However, they did not explicitly
or. = 1 + K ' ~ - D (4) give an equation of AKth as a function of crack size.
where: D = defect size defined as the diameter of the Various modifications of their model have followed their
study38,46-58,62-64.
smallest circle enclosing a defect; and K' = constant.
The equation is based on the fracture mechanics concept
that strength reduction is proportional to the stress R.A. Smith's model 46 (8)
intensity factor at the defect. This means that a defect This model, known as Smith's model, is a simplified
is regarded as a crack. However, the dependence of model of Kitagawa and Takahashi's model~,45 (7), in
AKth on defect size was not considered in de Kazinczy's which AKth or trw is approximately expressed by the
study because the anomalous behaviour of short or small two straight lines of constant AKth ic and constant orwo
cracks had yet to be realized. shown in Figure 3. Obtaining the values of AKth ~c and
It is shown in this work that the value of K' in o-wo in experiments based on conventional fracture
Equation (4) is expressed as a function of both the yield mechanics allows a prediction of the fatigue strength,
stress try and a constant K s dependent on the defect which in practice is a little non-conservative over the
geometry. For natural defects, such as inclusions, the transition from long to short cracks, though the difference
values of K s and D must be assumed before prediction between the prediction error looks small on a log-log
because they can be known only after fatigue tests. graph (Figure 3). Taking this into consideration, D.

17(1 Fatigue, 1994, Vol 16, April


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

Table 2 Comparison of experimental results37 with the prediction of Mitchell's modeP'. Material: annealed 0.13% carbon steel with tru =
353 MPa, Hv = 120 and trio = 181 MPa

Experimental result for specimen with a Prediction using accurate values of p and Approximate prediction assuming a
drilled hole K, hemispherical pit

Diameter, d Depth, h Error Error


(mm) (mm) ~r,~ (MVa) p (mm) K~ tr,~ (MPa) (%) p (mm) K~ tr,, (Maa) (%)

0.1 0.05 181 0.05 2.00 168 -7.2 0.05 2.5 163 -9.9
0.1 0.1 172 0.05 2.35 164 -4.7 0.1 2.5 150 -13.0
0.1 0.2 157 0.05 2.50 163 3.8 0.2 2.5 132 -15.9
0.2 0.1 157 0.1 2.00 158 0.6 0.1 2.5 150 -4.5
0.2 0.2 147 0.1 2.35 152 3.4 0.2 2.5 132 -10.2
0.2 0.4 137 0.1 2.50 150 9.5 0.4 2.5 114 -16.8
0.5 0.25 142 0.25 2.00 140 -1.4 0.25 2.5 126 -11.3
0.5 0.5 128 0.25 2.35 130 1.6 0.5 2.5 108 -15.6
0.5 1 118 0.25 2.50 126 6.8 1 2.5 94 -20.5

Approximate prediction is done for a hemispherical pit with a radius which is the greater dimension of d/2 and h. Mitchell assumed K~ =
2.5 for a hemispherical pit

03 1000 Smith's model[8] specimens. Replacing AOrth by 2tr, and AtrwO by 2~wot,
we obtain from Equations (7) and (8) the relationship
@
J ( /o ~1/2
trw = trio \ / ~ o ) (9)
d ,
e-
Although lo is termed the fictitious crack length and
~ 100 is considered a material constant, the value of 1o itself
does not correspond to any physical dimension within
~o / o Typicalexperiment
~ adata
lL the material 31.
The advantage of this approach is that the fatigue
t" strength of a component containing a short crack can
11 1o 12 be evaluated from material constants tr~ and AKth ~¢
t--

10 t-- I I determined by conventional fracture mechanics fatigue


0.01 0.1 1 experiments. However, no estimation method of lo for
Crack length or depth, l mm three-dimensional cracks has been shown.
Figure 3 Typical data plotted on the Kitagawa-Takahashi diagram
Usami and Shida's model 5s (1(9)
Usami and Shida assumed that, at the fatigue limit,
the size of the cyclic plastic zone rp¢, at a crack tip, is
equal to a material constant rp<w). The value of rp¢ is
Taylor and Knott 65,66 defined a critical crack length l2 related to the stress intensity factor for a defect, K, and
above which linear elastic fracture mechanics can be cyclic yield stress trv¢ in terms of the two-dimensional
applied (Figure 3), and proposed a simple conservative Dugdale modeP 9. The physical reality of the cyclic
approach, in which an inherent crack length is assumed plastic zone rp~(w) in the model is not examined by
to be 12 whenever the defect size is smaller than/2. experiments but it should, instead, be considered to be
a material parameter determined by fatigue data.
El Haddad et al.'s model 47 (9) Depending on the values of stress ratio R, the fatigue
In order to formulate the Kitagawa-Takahashi diagram limit is expressed by the following equations:
shown in Figure 3, El Haddad et al. proposed an
equation of the form rpc(w) =/eqLSec[ 4~-yc ) - 1 R>~0 (10)
AKth 1~
Atrth = V'rr(l + lo) (7) rpc(w) =/eq[SeC ( 7r ) ] R~<0
L \4trw/trmaxw + 2R - 1
where:
AOrth = threshold nominal stress range; (11)
AKth lc = threshold stress intensity factor range for a where: Ao'w = stress range; and O'max w : maximum
long crack; stress at fatigue limit. When the compressive minimum
l = half the crack length; and 46,47 stress is so large that the compressive plastic zone size
to = (AKth ,c/Ao'wo) 2 (8)
"/T t i n this paper, the stress range is defined as A~ = ~m.~ - O'min =
2¢r,(cra = alternating stress amplitude), i.e. Ao'th = A0rw = 20rw and
where Ao'~ is the fatigue limit stress range for plain A~r~ = 2try.

Fatigue, 1994, Vol 16, April 171


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

at a crack tip may be larger than the tensile plastic zone


size, the following equation is used: I I I

• [ //TO'max w ~ ] ~" ] - ~ ~ ........ ii"~'"". Tanaka et al.'s m o d e l -


rpc(w, = / , q L S e c ~ ) - 1d (12) ~ o ~ _ ~ = "St.,. -.......... Smith's model

In the above Equations (10)-(12), leq is half the length ~ o5 -1.0 Ov(M~] a"%u',,
of an equivalent two-dimensional crack assumed in an o
C-
s 2 oc
S20C
(468)
(tg&) Tanaka eta/" ~ ' ~
o'~'-~
. .
infinite plate and is defined by the following equation: -6
t-
A Mild steel (289) Fto~t (1959) ~B~J I
t/) G/,0 II Haddad (ISTg) " ~
t3 SM4I (2St) OhucN da (t$75)
leq - - (K/tr°)~2 (13) ~ o.~ •

SM41
SM90
(251} Uslm~i
(373) IUtagava
(1975)
(1978)
-
a~
a) • H T 80 (728) K i l a ~ a (19"/9) = r ~
> • 13Cr¢=tst steel(r89) Us4m~l (1979) ~ .
tl:l COlOpef FrOlt (IgS30
where: K = stress intensity factor of a crack; and O'o = ~- o.i Aluminum (30/=) Frolt (196/~)

applied nominal stress. Regarding an irregularly shaped I , I , I r


0. I I I0 I00
defect as an equivalent elliptical crack, this model is o.m
extended to apply to three-dimensional defects. Figure
Relative crack length, leq I Io
4 shows an example of the comparison of the experimen- Figure 5 Comparison of prediction by Tanaka et al.'s model and
tally obtained fatigue limits with the predictions. Claims experimental data 62
of good agreement between predicted values and exper-
iments must be carefully considered because even a 50%
error appears small on a log-log graph, especially for length/eq calculated by Equation (13). Figure 5 shows
small defects. a comparison of Equation (14) with various experimental
data.
It is very interesting to notice that El Haddad's
Tanaka et al. "s model 62 (11) model 47 (9), Usami and Shida's modeP s (10) and
Tanaka et ai. assumed that the threshold condition is Tanaka et al.'s model 62 (11) based on fracture mechanics
determined by whether or not the crack-tip slip band are eventually very similar to Isibasi's simple modeP 9,6°
blocked by the grain boundary propagates into an (1).
adjacent grain, and derived the following equation. This
model is more comprehensive than the previous two The aN/-a~parameter model 37"38,5° (12,14,16)
models and includes El Haddad's model 47 (9) and Usami In 1983, Murakami and Endo 37 proposed a new
and Shida's model 58 (10) as a special case. geometrical parameter aV"~rea for two-dimensional and
three-dimensional defects on the basis of both microscopic
trw = o'~ [1 + (2/w) bxf~ c°s-~(l/b) - l ] observation on cracking from small surface defects and
1 q- gno/o~fr~Tro) 0 three-dimensional numerical stress analysis37,72-74 of
(14) cracks with various shapes. From their experimental
results they derived the following empirical formula:
where: O~r = frictional stress of dislocation motion;
K m = critical value of microscopic stress intensity factor; O"wa ~ = C n ~6 (15)
and b = half crack length l plus slip band zone size too.
where: trw is r otat~_g bending or tension-compression
The values of tr,,,o, O~frand K m must be experimentally fatigue limit; v area = geometrical parameter of a
obtained for direct use of this model. The value of OJo defect, which is defined as the square root of the area
is assumed to be equal to half the grain size, i.e. too = d~ obtained by projecting a defect or a crack onto the
2. This model is two-dimensional but can also be applied
plane perpendicular to the maximum tensile stress
to three-dimensional defects by using an equivalent crack (Figure 6); and n, C = constant. The negative aspect

I000 = . I= -,3/2
500
...........ii i
E

IO0

50

,<
................ ..N,
O.OOOf O.OOI 0.01 O.I I IO IO0
Surfoce roughness, mm
Maximum tensile stress direction
Figure 4 Comparison of prediction by Usami and Shida's model
and experimental datass Figure 6 Definition of

172 Fatigue, 1994, Vol 16, April


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

of this equation is that we need fatigue tests to determine model is a three-dimensional model. Note that the
n and C. 67 choice of aX/'~rea as a geometrical parameter comes from
In 1986, Murakami and Endo 38 revised the model stress intensity factor consideration 37,72-74. Thus it may
and proposed equations as be said that, starting from a fracture mechanics approach,
Murakami and Endo have returned to a Frost-type
Arm = 3.3 × 10 -3 (Hv + 1 2 0 ) ( a ~ ) 1/3 (16) model.
The applications of the aX/-d'mareaparameter model have
1.43 (Hv + 120) (17)
= (
been widespread, not only to the fatigue problems of
small holes37-39,50,68,75-81,83 and small
cracks 38'39,68'76-81'83, but also to those of surface
where: AKth = threshold stress intensity factor range scratches 38,68, surface roughness 71,85, non-metallic
under the stress ratio R = - 1 (MPa ml/2); crw = rotating i n c l u s i o n s 5°'69-71'77-84'86'88'13°, c o r r o s i o n pits 71,85,86, car-
bending or tension-compression fatigue limitL.(._MPa); bides in tool steels 82, second phase in Al-Si eutectic
and Hv = Vickers hardness (kgf mm-2). V a r e a is in alloys87 and spheroidal graphites in cast irons 76,89-95.
p.m. The above equations enable one to predict AKth
These are summarized in detail in Refs 35 and 96.
and crw without a fatigue test. They concluded from It must be remembered that Equations (16)-(19) are
more than 100 experimental data that the prediction
valid over a a~/-~rea range which is dependent on
error is mostly less than - 1 0 % for notched and cracked material 37-39,5°'68"83,96. The valid upper limit of a ~
specimens having a ~ less than 1000 p~m and for H v is considered to be -1000 p~m. If the predicted value of
ranging from 70 to 720 (Figure 7) 38-6s.
Crwfor a ~ exceeds the ideal upper bound of fatigue
Equations (16) and (17) are for surface defects, while limit O'wu ( ~ 1.6Hv), the defect having such a small
for internal three-dimensional defects the following should be regarded as non-
equation is proposed by Murakami et al. 69 and Murakami detrimental37,38,68,78--83,86,96.
and UsukiT°:
1.56 (Hv + 120) (18) D. Tailor's m o d e P 3,64 (13)
=
D. Taylor pointed out that large scatter in fatigue test
results is an essential feature of short cracks and it arises
Murakami et al. 5° further extended Equation (17) to from effects of inhomogeneities of the microstructure
predict the fatigue limit for various values of stress ratio and residual stress in the vicinity of the crack. This
R. The prediction equation is expressed as argument will become more important in the future. In
order to model quantitatively the fatigue threshold
1.43 (Hv + 1 2 0 ) x ( ~ ) ~ phenomena with scatter, Taylor used the P - a plots, in
orw = ( a~)l/6
which the probability P of crack growth is expressed as
a function of the crack depth a of a semi-elliptical crack.
a = 0.226 + H v x 10 -4 (19) Although the P - a plots may not be easily obtained, this
In predicting trw by Equation (19), no fatigue test is approach is useful to designers who need to consider
necessary. It is sufficient to measure the values of the probability of fracture because it provides the fatigue
and H v and estimate R by trial and error at the limit expressed as a probabilistic variable.
place where the defect exists5°,71. They reported that
the prediction error was less than - 1 5 % for Hv = Luk6A et al.' s model 48,49 (15)
100-740. 30 In this model, two different equations are used to
Murakami and Endo's model, ~ w ~ -- C and trw predict the fatigue limit of cracked specimens and
= 1.43 (Hv + 120)/(aW~-~) ~/6, is similar in form to notched specimens. One is the same as the one proposed
Frost's model, oawl = C. The difference is the adoption originally by Tanaka et al.62
of X/area instead of l; that is, Murakami and Endo's For a crack that exists on a specimen surface, the
following E1 Haddad-type equation 62, cf. Equation (9),
is employed:
/o / "2
O A (120) m F (650) o K (114) trw = trwo l - ll + 1o] (20)
• B (153) o(3 (520) • L (355)
0.05 & C (160) • H (319) • M (244)
&D-1 (180) <>1-1 (378) a N (720) 10%error where 1 = crack depth; and 1o = the fictitious crack
O-2 (170) I-2 (375) ~ ..
o4° [] E (177) * O (70) . . . ~ length defined by Kitagawa and Takahashi 4~, Smith 46
& and El Haddad et at. 47, calculated by Equation (8). lz
.
is the maximum crack size below which the fatigue limit
® _.....~.-o 10% error is equal to that of unnotched specimens. Murakami and
T. Endo 97-99 showed the existence and definition of
~0.01 " "'" such cracks for a few materials (Figure 11).
For notches, they derived the following equation:
" AKIn = 3.3 x 10- (Hv + 120) (a~'~rea)
0.005 I i i i i = = tl i i , , , ~ t iI trwO[1 + 4.5(lllp)] 1:2
10 100 1000 O'w = Kt (21)
a~-~rea /am
on the assumption that non-propagating cracks at a
Figure 7 Comparison of prediction by Murakami and Endo's model
and experimental data3s (letters correspond to the materials given notch root are of semi-elliptical shape and the crack
in Ref. 38. Numbers in parenthesis indicate the Vickers hardness, depth is the same as ll for unnotched specimens.
Hv, of materials) Equation (21) can be used for notches whose stress

Fatigue, 1994, Vol 16, April 173


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

concentration factor Kt does not exceed 4. 49 For defects where:


other than notches, the estimation methods of the notch AKth = threshold stress intensity factor (SIF) range for
root radius p and the value of Kt are not shown. Figure a given loading condition;
8 shows a comparison of Equation (21) with the A g i th = intrinsic threshold SIF range (the minimum
experimental data. SIF range that will cause a fully open crack to grow),
Equations (20) and (21) require t r y , AKth 1¢ and ll which is measured at a very high stress ratio or at a
to be obtained by experiments, and specifically 11 in high stress ratio with intermittent underloads to eliminate
Equation (21) must be determined by curve fitting of crack closure;
fatigue test data 48,49, obtained by using several specimens k = a material constant that reflects the rate of closure
containing a very small crack or notch. Thus it seems development; and
that l~ has a physical meaning in Equation (20) but a~ = a - rp/2 (rp is a plastic zone radius).
loses that meaning in Equation (21), because it must Abdel-Raouf et al. 52-55 suggested that rp = 0.8d for a
be determined by curve fitting of data points. smooth specimen.
Using Equations (22) and (23), the threshold stress
A b d e l - R a o u f a al. 's model sl-55 (17) range Atrth is expresssed by the following equations. For
Abdel-Raouf et al. modelled the fatigue limit and smooth specimens with a crack:
short crack growth behaviour considering effects of
Hcl AKi th
surface strain concentration and crack closure. The Atr,h -- - - (24)
surface strain concentration is given by the following
equation:
and for notched specimens with a crack:
Ac
Qe -- A~ = 1 + q e x p ( - a a / d ) (22) H¢IAKi th
atr, h - (25)
where: FKt Qe
Ae = local strain range at crack depth of a; where: F = the shape factor of a crack; and Kt = the
Ae = nominal strain range; elastic stress concentration factor. Equations (24) and
d = average grain size in the direction of crack growth; (25) give the variation of AO'th with crack depth a and
= a factor that accounts for the ease of cross-slip; account for a non-propagating crack below the fatigue
and limit Atrw, which is given as the maximum value of Atrth.
q = a constant equal to 5.3 for typical polycrystalline The predicted values of Atrw were in good agreement
metals cycled near the fatigue limit. with experimental values for various materials, specimen
The crack-closure effect is expressed by a parameter geometries and loading conditionss3-55.
Hal, which is a function to model the build-up of closure In this model, AKi th, A g t h and k are the data to be
with crack depth and 1 ~< H¢~. For the user to calculate determined by experiments, d and a are the quantities
H¢l the following equation is helpful: to be measured or estimated and, for pores and defects,
AKth AKth F and Kt must be calculated by stress analysis considering
the shape of the crack front.
Hc~ - AKi th -- AKth + ( A K t h -- AKi t h ) e x p ( - - k a e )
(23)
Melander et al. 's model 56,57 (18)
In this model, the effective threshold stress intensity
factor range AKaf th is employed: that is, the equation
is expressed as

25° I Atrw --
A g e f f th
U V ppF(13o)
(26)

1
U = (U' - Uo)e -~''~ + Uo, Uo-l_ R ,
12.
200 fl=--
/a
(27)
p

E where:
Airw = fatigue limit stress range;
U = crack-opening ratio;
~ 150 p = radius of an inclusion (assumed as a spherical
pore);
la = length of an annular crack surrounding the pore;
Equation (21)for l1 = 100~m and ~ = 2 2 0 MPa /7(/30) = the shape factor for la = lao (critical transition
length, which separates behaviour controlled by short
100 I t t ¢ cracks from that by long cracks and is thought to
0 0.2 0.4 0.6 0.8 correspond to a microstructural barrier such as a grain
Notch radius, p rrrn boundary); and
Figure 8 Comparison of prediction by Luk~l~ et al.'s model and
R = stress ratio.
experimental data. Material: 2.25Cr/1Mo steel with Hv = 165, o'v The value of AK=~ th must be experimentally measured
= 380 MPa and cru = 530 MPa 'm for a given material. U' and k' are fitting parameters,

174 Fatigue, 1994, Vol 16, April


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

which are determined by FEM analysis of fatigue crack


closure.
From a fracture mechanics point of view, this model
(which could be called 'elastic-plastic fatigue crack
propagation analysis') may be perfect. However, to
determine the threshold condition of a crack emanating
from small defects or non-metallic inclusions, a fatigue
crack closure analysis by a method such as FEM must
be done. In this kind of FEM analyis, we need data for
the cyclic stress-strain properties or the constitutive 5 4140 13 o 23 ¢I00 , 26.
equation of the microstructures in question. , i ii
PERFORMANCE OF MODELS AND CHOICE OF
PREDICTION METHOD
The existence of many models reflects the importance
and difficulty of fatigue problems related to small cracks
and small defects. However, it is not easy for company !
engineers to use complicated models that include many
parameters to be determined by difficult experiments d= 40ran d = 100]~m
and lengthy fatigue tests. Simple models based on many
assumptions or hypotheses and few materials data are o'. = 1 8 1 M P a o'w= 1 7 2 M P a
also unsuitable for practical use. In this section, we shall Figure 10 Interior shapes and configurations of non-propagating
discuss the uses and functions of the prediction models cracks observed at a small hole in an annealed 0.13% carbon steel 99
and describe how to choose a relevant model.

Importance of understanding the fatigue limit a small-crack problem. Consequently, the fatigue limit
phenomenon of steel specimens containing small defects is not
Correct understanding of the fatigue limit phenomenon determined by crack initiation conditions but by threshold
is necessary to judge whether a model can predict conditions for propagation of a crack emanating from a
accurately the fatigue threshold of small cracks or defect.
defects. However, models must be carefully used if the size
Figure 9 shows a crack emanating from an artificial of the non-propagating crack observed at the fatigue
hole, observed at the fatigue limit 5°. This crack is a so- limit of a plain specimen or the grain size is employed
called non-propagating crack, which started from the for modelling. Nisitani and Murakami 1°' observed 26
hole edge in the early stages of the stress cycle and non-propagating cracks at the rotating bending fatigue
afterwards ceased to propagate under the same stress limit and 354 cracks at the reversed torsional fatigue
level. Figure 10 illustrates a sectional view of such non- limit in a 0.13% carbon steel smooth specimen 5 mm
propagating cracks 99. Many similar observations have in diameter. The length of non-propagating cracks was
been given in the literature 37'9''92'95"97'98'101-103, so that not constant (ranging from - 0 I~m to 100 I~m in rotating
some models (10-18) are established under the assump- bending). Moreover, it was noted that in one specimen
tion that the fatigue limit problem of steels is essentially a few grains having non-propagating cracks coexisted
with many grains that contained only slip bands and
also many grains that had neither crack nor slip bands.
Therefore, if AKth for each non-propagating crack is
defined by a different crack length and the same fatigue
limit stress, it follows that we have different values of
AKth for a single specimen. This is related to the
statistical properties of a microstructure.
The same situation can also be seen for the case of
non-propagating cracks emanating from extremely small
defects. Murakami and T. E n d o 97, performing rotating
bending fatigue tests of 0.13% carbon steel specimens
containing 12 artificial small holes of either 40 or 50 p.m
in diameter, pointed out that miscellaneous phenomena
were observed at the fatigue limit of a specimen, where
some holes had non-propagating cracks on both sides
of the hole edge, some holes had a crack on one side
of the edge and other holes had no cracks. It must be
realized that there appear to be three distinctly different
states on the surface of a specimen at the fatigue limit:
(1) non-propagation of cracks; (2) slip band formation
but without cracking; and (3) neither slip band formation
Pigm-e 9 Non-propagating crack observed at fatigue limit of marag-
ing steel ( H v = 740). Fatigue limit stress amplitude crw = 666 MPa; nor cracking. As defect size decreases, the length of
m e a n stress crm = - 2 5 5 MPa; hole diameter d = 50 I~m; hole depth non-propagating crack emanating from the defect tends
h = 701xm 5° to have a larger scatter. Although the length of non-

Fatigue, 1994, Vol 16, April 175


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

propagating cracks for a large defect tends to have a trusions could greatly affect the stress concentration
reproducible value, we should not assume this nature factors. The use of stress concentration factors for
of a large defect for a small defect; otherwise, our estimating the fatigue strength of steels is therefore not
models will be ignoring real observational evidence. only unreasonable but also impractical, as the inclusions
Thus we can understand that we must be careful in or other small natural defects have various shapes and
modelling to add to the initial defect size the length of some are far from spheroidal or ellipsoidal4-6,1°6-113. In
a non-propagating crack that happened to initiate at the order to study the effects of small surface defects,
edge of the small defect in question. The large scatter Murakami and his c o l l e a g u e s 37-39,5°,68,75,76,97-1°°,114,115
of phenomena on a specimen under a constant stress have conducted systematic experiments using specimens
amplitude is peculiar to fatigue related to small cracks containing artificial small holes with diameters ranging
and defects. Concerning this problem, D. Taylor's point from 40 Ixm to 500 p.m. Figure 11 shows some results
is i m p o r t a n t 31"63-66. He considered the statistical nature of their experiments 97,99, which indicate an increase in
of small-crack growth in model 1363,64. Miller 29, Abdel- fatigue limit with decreasing size of artificial surface
Raouf et aL 51'52 and Blom 1°4 also considered the holes and the existence of a critical size of a hole
importance of this statistical nature of small cracks. exerting no influence on fatigue strength, though the
As mentioned previously, if we evaluate the fatigue critical size is dependent on materials. Since the sizes
limit of a plain specimen or a specimen containing a of the artificial holes on the surfaces of these specimens
small defect by AKth, it follows that we define many are very small in comparison with specimen size, and
different value of AKth for a single specimen. The values the hole shapes are approximately geometrically similar,
of AKth evaluated in this way will be scattered not their stress concentration factors Kt are considered
within - 1 0 % but almost 100%. On the other hand, if almost equal. Nevertheless, Figure lI shows that the
we pay attention to the applied nominal stress instead fatigue strength varies distinctly depending on hole size.
of AKth, the fatigue limit stress orw will stay within a This clearly indicates the unreasonableness of fatigue
very narrow band. A specimen will fail at the stress strength evaluation based on the stress concentration
10% higher than orw and will have no non-propagating factor alone.
crack at the stress 10% lower than orw. It may be said To avoid this difficulty of models using Kt, some
that the reproducibility of the fatigue limit or, of current attempts 4°-42 have been made to evaluate the notch
commercial metals in fatigue tests is very good~t. This effect of small defects and inclusions by using the
experimental fact is favourable for the prediction of conventional fatigue notch factor approach. However,
fatigue strength, because a good model may be obtained it does not seem that these approaches could overcome
only from simple material factors and geometrical factors the difficulty in assuming correct Kt and notch root
of defects. radius p for natural small defects and inclusions.

Importance of use of valid experimental data for the Dependence of AKth or o% on size of crack or defect
approval of models The threshold stress intensity factor range, AKth, is
Care must be exercised when the value of the fatigue useful for determining the maximum allowable stress
limit of a plain (smooth) specimen orwois employed as when cracks or defects are detected in machine parts
one of the basic data of a model, because there is a and structures under service loading. However, recent
possibility that orwOitself may include the influence of experimental studies 24,38,44,45,62.116-119 s h o w that the
small defects. The data for tension-compression fatigue
of high-strength steels must also be carefully checked
before use, because misalignment of specimens in fatigue
tests and bending of specimen shape due to heat 260
treatment always tend to lower the fatigue strength. • 0.13% C steel
Unusually low fatigue strength of data 1°5 must therefore 240' " 1 ( o 0.46% C steel
be questioned. Company engineers do not usually have n 2017-T4 AI alloy
Q_ 220
the fatigue data requested in the models proposed by ~- ~"~"~,o A 70•30 bras~
many researchers. Therefore, models that do not need ~= 200
fatigue tests will be welcome by them, or some suggestions
for obtaining data equivalent to fatigue data should be - 180q
presented. They are naturally very careful to use models E
that do not have a declared prediction error. "~ 160 I
-I

Questions about use of stress concentration factor Kt _ 140


Many efforts have been made to evaluate the stress 120'
concentration factors Kt of inclusions by assuming that
Z~
their shapes are spherical or ellipsoidal cavities. These 100
assumptions are, however, rough estimations, in that
any slight deviation from the geometry or small pro-
0 , I , I , I , I , I
0 100 200 300 400 500
CThis point should not be misunderstood with the scatter of fatigue Diameter of hole, d /~m
strength of high-strength steels, whose fatigue strength is strongly
influenced by the existence of non-metallic inclusions. If the size Figure 11 Effect of a small drilled hole on fatigue strength 97'99.
of most detrimental non-metallic inclusions is definitely known, the Regardless of equal stress concentration factor, fatigue strength
inclusion can be treated with the same model for small defects35 decreases with increasing size of hole

176 Fatigue, 1994, Vol 16, April


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

100-
[~ Frost
n Mild steels

and Nakazawa -'-=H


"e
13.
~i 0.47% C steel (annealed) I
Eutectoid steel [

i o 0.45-.0.46%
M u T i C steel (annealed)
and Endo - - ~1 v o
DO []
10 D
oo0,O o~ -m" o+ []
==
o* ~-----'~11
oo° gooo ~ 1 6
oO
4

I , , , . , ,I i , i , i , ,,I I I I I ' ' ''1 ' ' ~ ' I II[ I I

0.01 0.1 l mm 1 10
i i I t i I i t I I t i t t i i t I

10 100 1000
aq rea am
Figure 12 Dependence of AKth on crack or defect size from data by Frost TM, Kobayashi and Nakazawa ~ , and Murakami and Endo 38

value of AKth is dependent on crack size; that is, the constant value depending on the specific material, and
value of A g t h decreases with decreasing crack size. this indicates that the slope of the relationship between
Figure 12 gives the relationship between AKth and AKth and ~ (or l) on a log-log scale varies from
defect or crack size, in which data for fatigue limits of one-third to zero with increasing crack length or depth.
specimens containing either a small defect or a small Figure 12 shows that the slopes of
crack have been collected from the literature 12'36'3s. The 1/4 and 1/6 correspond to a transient value from one-
values of AKth are calculated by assuming the defects third to zero. This means that Frost-type equations -
as the equivalent cracks. For circumferential cracks or Equations (2), (3), (15)-(19) - must be used according
notches (two-dimensional defects) the crack or notch to each application range of defect or crack size.
depth I is employed as defect size on the abscissa. For If AKth |c is employed in a model and an engineer
three-dimensional defects such as holes (Figure 6), who wants to use the model does not have the
Murakami and Endo's parameter ~/area is employed. experimental data for AKth~c, he or she will simply
These are related b y 37'38'68'96 assume or estimate it from data for similar materials. If
the engineer assumes it to be 3 MPa m 1/2 instead of the
(unknown) true value of 2 MPa m 1:2, the prediction of
a~ = X~I or l- a~ (28)
fatigue limit trw will result in -50% error for a small
crack, say 300 MPa instead of the true fatigue limit of
In the region of ~ ~ - 1000 ~m, the relationship 200 MPa, although the difference of AKth~¢ values
between AKth and a ~ is approximately linear with between 3 MPa m 1/2 and 2 MPa m 1/2 seems very small**.
the slope of 1/3 on a log-log scale, and the following A fatigue limit prediction of 300 MPa instead of 200 MPa
equation holds regardless of materials3s. could have severe practical consequences. We also need
to check whether the value of AKth i¢ is measured for a
AKth ~ (a~) 1/3 or Agth 0c p/3 (29) sufficiently long crack; that is, whether AKth ~¢ is the
If we convert the empirical formulae obtained by correct value independent of crack size.
Kobayashi and Nakazawa36 (#wl = C) and by Frost et
d. ~2-~7 (trawl = C) to the same form as Equation (29), Applicability of models and types of defect
we have In general, a model is applicable to certain kinds of
defect. In this section, various models will be examined
Art h 0c (~)1/4 or AKth OC /1/4
with the results of critical experiments. Table 3 shows
(equivalent to o'~l = C) (30) the sources of data and experimental fatigue limits.
Circumferential cracks and notches are typical two-
and dimensional defects, semi-elliptical surface cracks and
AKI h oc ( a ~ ) 1/6 or A g t h oc/1/6
(equivalent to oawl= C) (31)
**For a long two-dimensional crack of 200 mm in a wide plate, the
respectively. prediction error may also be - 5 0 % but the absolute value of or,, is
Contradiction between Equations (29)-(31) was dis- very small because ~r. is --3 MPa or - 2 MPa. Therefore, the
cussed in detail by Murakami and M a t s u d a 39, who prediction error due to the incorrect assumption of dJfth ~ may not
be so serious. However, for a short two-dimensional crack of 40 Van
attributed the difference in the exponents 1/3, 1/4 and in a plate, not only the error ( - 5 0 % ) but also the difference in the
1/6 to the difference in the size ranges investigated: with absolute value of the predicted fatigue limit Crw is large. Therefore,
increasing crack length or depth, AKth approaches a the incorrect assumption of AKth ic results in a serious error.

Fatigue, 1994, Vol 16, April 177


Effects o f defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

Table 3 Experimental results for various kinds of defect

Shape and dimensions of specimen and Fatigue limit trw or


defect Material applied stress ~r Researcher Ref.

(a) Circumferential crack Annealed 0.56% carbon steel trw = 136 MPa Isibasi 18
tru = 602 MPa (rotating bending)

(
14 ~)m dla. ~ ~~)25--0/~m

(b) Circumferential notch 2.25Cr/1Mo steel ~ . = 165 MPa Luk~i~ et al. 48


tru = 530 MPa (tension-compression)
Hv = 165

5 ~ l--"
70/=m

(c) Semi-elliptical surface crack Annealed 0.46% carbon steel tr. = 201 MPa Murakami and Endo 96
tru = 543 MPa (rotating bending)
,_,2.a= 213/~r~, Hv = 170
7///,,~, I==[ t'/~/ b/a = 0.7

,3//~se sedtioh'//~" Crack starter


///./////..///.///~/~//' (small hole)
bpecimen oiameter = 10 mm"
(d) Drilled hole Annealed 0.46% carbon steel trw = 201 MPa Murakami and Endo 96
~ru = 543 MPa
i(Dia.=200 l'EI,F I Hv = 170
(rotating bending)

~/'//~ ~ , ~pth =

Specimen diameter = 10 mm
(e) Non-metallic inclusion Quenched and tempered 0.35% tr = 724 MPa Murakami et al. 69
carbon steel (rotating bending)
Hv = 570 Stress at the centre of
inclusion is 716 MPa
Nf = 4.02 x 106

#m

Specimen diameter = 8 mm

drilled holes a r e typical artificial t h r e e - d i m e n s i o n a l Evaluation of non-metallic inclusions


defects, a n d non-metallic inclusions are typical natural Non-metallic inclusions considered to be mechanically
t h r e e - d i m e n s i o n a l defects. equivalent to small defects. W h e n non-metallic inclusions
M o d e l s 1 - 3 , 7 - 1 4 , 16 and 17 are available for cause fatigue failure, it is v e r y difficult to predict the
circumferential cracks. M o d e l s 1 - 3 , 5, 6, 10-12, and fatigue strength in advance o f a fatigue test. This is
14-17 are available for circumferential notches. M o d e l s because very little w o r k has b e e n d o n e o n t h e effect
4, 5, 7, a n d 10-17 a r e available for semi-elliptical surface of non-metallic inclusions. Several o b s e r v a t i o n s 12°-126
cracks. M o d e l s 4 - 6 , 10, 12, 14 a n d 16 a r e available for s u p p o r t the r e a s o n a b l e n e s s o f the c o n c e p t that a non-
drilled holes. M o d e l s 4 - 6 , 10, 12, 14, 16 a n d 18 are metallic inclusion m a y b e r e g a r d e d as mechanically
available for non-metallic inclusions. equivalent to a void o r defect with t h e s a m e s h a p e a n d
size. T h e size a n d location o f t h e fatal inclusion a p p e a r i n g
on the fracture surface of a s p e c i m e n after fatigue

178 Fatigue, 1994, Vol 16, April


Effects of defects, inclusions and inhomogeneifies on fatigue strength: Y. Murakami and M. Endo

fracture can often be identified, and in such cases the


stress applied at the fatal inclusion should be larger than
the value calculated by the fatigue strength prediction
Equation (17) or (18). In this manner, we can examine
the applicability of models to inclusion problems.
Murakami and his c o l l e a g u e s 69-71,77-84 investigated
many 'fish eye' fatigue failures and indirectly verified
the validity of their prediction model, the
parameter model. Table 4 shows an example of the
comparison of such investigations 69. We can see that
the stresses (oe) applied at the fatal inclusions are
actually larger than the estimated fatigue limits (tr'w).
The concept of the X/-dT~ parameter model (12, 14,
16) suggests that there should be no difference in the i - ......
influence of inclusion shape if the values of ~ of
two inclusions are identical. This prediction is in complete
agreement with the results of the fatigue tests conducted
by Duckworth and Ineson 1' who artificially introduced
spherical and angular alumina particles into ingots
(Figure 13) and found no definite difference of effect
of inclusion shapes. Their ~ r i m e n t a l data can be
arranged clearly by the Varea parameter model as
shown in Figure 14 '27.

Prediction of fatigue strength scatter caused by non-


metallic inclusions. Few models are able to predict the
effect of non-metallic inclusions on fatigue strength. This
may be because adequate reliable quantitative data on
non-metallic inclusions are hard to obtain and accordingly b
the statistical approach inevitably has to be used. Figure 13 Typical example of alumina particles artificially added
Mitchell 4°m proposed the prediction method of fatigue into ingots11,~2~: (a) typical spherical particles; (b) typical angular
strength scatter with the combination of his model (5) particles
and the statistical distribution of defects contained in
cast steels. the scatter of fatigue strength of high-strength steels.
Nordberg 42 (model 6) showed the prediction method of The upper bound of fatigue strength, ~w~, is obtained
fatigue strength scatter caused by non-metallic inclusions when fatigue fracture is not affected by defects or
with essentially the same method as Mitchell. Nordberg inclusions, and the value can be estimated by the
carded out complicated mathematical calculations of empirical formula tTwu - 1.6Hv, where Orw~is in MPa
possible size distributions of non-metallic inclusions and H v is in kgf mm-2.
contained in specimens. Microscopic inspection, to The lower bound of fatigue strength, ~,~, can be
establish the statistical frequency of size distribution, obtained when the largest inclusion is located just in
must be performed for this kind of prediction. contact with the surface of a specimen 7°. The prediction
Larsson et aL 128 combined fracture mechanics with equation for trwl is given by
a statistical size distribution of inclusions and made
predictions of the fatigue lives of shot-peened specimens. 1.41 (Hv + 120) (32)
In this study, an inclusion is assumed to behave as a
notch or a penny-shaped crack.
Murakami et al. 69 and Murakami and Usuki 7° proposed where V'-~eam= is the expected maximum size of
a method for predicting the upper and lower bounds of inclusions contained in a definite volume. In order to

Table 4 Size and location of inclusions and fatigue limit predicted by Equation (18): bearing steel (Hv = 758) 69

Fatigue fimit
Nominal stress Cycles to Inelusionsize, Distance from Nominal stress predicted by
at surface, ~r failure, Nf aV~'~rea surface, h Shape of at inclusion, ~' Eq. (18) Or'w
(MPa) ( x 104) (p,m) (l~m) inclusion (MPa) (MPa) ~/o,',,

1030 125.25 24.0 310 954 806 1.18


1030 556.77 15.9 140 995 863 1.15
981 422.95 15.2 28 973 870 1.12
981 898.01 9.9 74 963 934 1.03
981 175.51 12.5 10 978 899 1.09
1030 224.64 18.6 74 1011 841 1.20
1030 30.52 22.2 24 1023 817 1.25
1030 26.48 28.4 110 1003 784 1.28
971 735.45 28.9 350 886 782 1.14

Fatigue, 1994, Vol 16, April 179


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

Return period,
1.5
T
• • 13 _99. 9998f5x
0 o
OlO00 Q 12 _99. 99951-2x
99.998 I- l
- 99.999--5xi
°5I
0 s -

Ds -
D4 -

- 99.995-2xl
_~99.99 - l )~F T: 3180 for
- ~ A n \ o n ^ ~~ z ~ ~- 9
1.0 o 99.98-5xi )3F 100 specimens
O A ,L.
i
O
7 "- _ 3 i i
.-x ~-~ 99.91- ]0 FT:318,O r
61-. 99 81-5×10~-1~]..~ec'mens
b
>
u..
_=- 99.sl-2×1o2L- A
O
::} ~ 99.0-
Rotating bending n- ,'I - ~ 98 -
o
0.5-- z~ Angular alumina particles
O Spherical alumina particles 3 ~o 95-
2 ~ 90
Tension compression 8O
Spherical alumina particles 1--~
50
0-0
10 28.7 ~tin
]
0 I I I I ] I I
O.1 o lO 20 30 40 50 6O
10 4 10 5 10 6 10 7 10 8
al/fa'~max #m
Nf
Figure 14 M o d i f i e d S - N c u r v e ( o " / o dw v s N t ) z27. T h e r e is n o Figure 15 Statistical distribution of the maximum size a ~ , ~ , =
substantial difference in the effects of shape of inclusions, angular of inclusions for a roll steel (Hv = -560, <Tu = 1868 MPa) 13°
or spherical (see Figure 13)

estimate the expected maximm size of inclusions


contained in a definite number of specimens, Murakami
and his c o l l e a g u e s 70,71,76,78-84,86,130 applied the statistics
of extreme v a l u e s 129 to the distribution of inclusions.
~ 15 shows the distribution of the maximum size
o.o,o j_.C-
1000 - - T h e upper bound : ~

max of inclusions observed in the unit area 8O0


So(=254 mm 2) of a roll steel 13°. Estimating the return
period T for N S_Sl~cimens, the expected maximum size
Q. 2-----'-" o%
of inclusion Vaream~x can be determined by the pro-
600
cedure indicated in the fiug_u~.
Introducing Hv and Vaream~x to Equation (32), the ¢., _ ~ ~ . . ~ - ~ O The lower bound of
m fatigue strengthfor
lower bound or,= of fatigue strength for N specimens is N specimens (Prediction)
predicted as in Figure 16 z3°. The lower bound of ",1 4OO
experimental data is well predicted by the equation. Experiment ( O notfish'eyebrokenfracture
Similar applications to various materials have been U_
reported in Refs 70, 71, 76, 78-84 and 86. The accurate 2OO
prediction of the lower bound of fatigue strength will Prediction equation of lower bounds :
be very useful not only for the design of machine Owl= 1.41 ( HV+ 120 )/(arv~-eamax)lm
elements or structures but also for the improvement of
the grade and quality of materials. The line of the lower o i i i i
450 500 550 600 650
bound of fatigue strength in Figure 16 is more helpful
for the determination of allowable stress than taking an Vickers hardness, Hv
ambiguous safety factor. Figure 15 can provide a Figure 16 Comparison of rotating bending fatigue test data with
guide for the control of non-metallic inclusions in the predicted upper and lower bounds of fatigue strength. Material:
steelmaking process. roll steel (Hv = ~560, ~% = 1868 MPa) 13°

180 Fatigue, 1994, Vol 16, April


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

Inhomogeneities Fundamentals and Engineering Applications', Stockholm,


1981
A few models have been proposed that consider the 23 Miller, K.J. Fatigue Eng. Mater. Struct. 1982, 5, 223
influence of inhomogeneities such as the Si phase 24 Suresh, S. and Ritchie, R.O. 'The propagation of short
in aluminium alloys aT, carbides a2, alumina particles 127, fatigue cracks', Report No UCB/PP/83/1014, Dept of
graphites in cast irons76,a9-9s and cast steels4°,4L In Materials Science and Mineral Engineering, University of
California, Berkeley, 1983, pp. 1-70
these models, the inhomogeneities are regarded as 25 Davidson, D.L. and Suresh, S. (Eds) 'Proceedings of
mechanically equivalent to small defects or cracks. More International Conference on Fatigue Crack Growth Threshold
detailed observations 131-133 and discussion from the Concepts', Philadelphia, USA, 1983
viewpoint of fracture mechanics will be necessary in the 26 Miller, K.J. and de los Rios, E.R. (Eds) 'The Behaviour of
Short Fatigue Cracks', EGF Publ. 1, Mechanical Engineering
future to develop new improved composite materials. Publications, London, 1986
27 Lies, B.N., Hopper, A.T., Ahmad, J., Brock, D. and
Kanninen, M.F. Eng. Fract. Mech. 1986, 23, 883
CONCLUSION 28 Miller, K.J. Fatigue Fract. Eng. Mater. Struct. 1987, 10, 75
29 Miller, K.J. Fatigue Fract. Eng. Mater. Struct. 1987, 10, 93
Fatigue strength prediction models for small defects, 30 Taylor, D. Int. J. Fatigue 1988, 10, 67
non-metallic inclusions and inhomogeneities have been 31 Taylor, D. 'Fatigue Thresholds', Butterworth, London, 1989
reviewed. First the historical development of models 32 Miller, K.J. Prec. Inst. Mech. Eng. 1991, 205, 1
was discussed. They can be loosely classified into three 33 Suresh, S. 'Fatigue of Metals', Cambridge University Press,
Cambridge, 1991
categories: (1) empirical models such as Isibasi's and 34 Klesnil, M. and Luke, P. 'Fatigue of Metallic Materials',
Frost's models; (2) models based on the so-called fatigue Materials Science Monographs 71, Elsevier, Amsterdam, 1992
notch factor approach; and (3) fracture mechanics 35 Murakami, Y. 'Metal Fatigue: Effects of Small Defects and
models. It is very interesting to notice that complicated Nonmetallic Inclusions', Yokendo, Tokyo, 1993
36 Kobayashi, H. and Nakazawa, H. Trans. Jpn Soc. Mech.
fracture mechanics models eventually give results that Eng. 1969, 36, 1856
are very similar to those of Isibasi's simple model. 37 Murakami, Y. and Endo, M. Eng. Fract. Mech. 1983, 17, 1
Fracture mechanics models mostly employ the value 38 Murakami, Y. and Endo, M. In 'The Behaviour of Short
of AKth lc for a long crack as the basic data and are Fatigue Cracks', (Ed. K.J. Miller and E.R. de los Rios),
mostly available only for two-dimensional cracks or EGF Publication 1, Mechanical Engineering Publications,
London, 1986, pp. 275-293
defects. A careless assumption of the value of AKth tc 39 Murakami, Y. and Matsuda, K. Trans. Jpn Soc. Mech. Eng.
may result in a large prediction error of the fatigue limit Set. A 1986, 52, 1492
Or w . 40 Mitchell, M.R. Trans. ASME, J. Eng. Mater. Technol. Oct.
The application of the various existing models has 1977, 329
41 Mitchell, M.R. 'A unified predictive technique for the fatigue
been summarized (in Table 1) and the basic data resistance of cast ferrous-based metals and high hardness
necessary for prediction have been indicated. wrought steels', SAE/SP-79/448, 1979, pp. 31-66
42 Nordberg, H. In 'Proceedings Swedish Symposium on Non-
Metallic Inclusions in Steel' (Ed. H. Nordberg), 1981,
REFERENCES pp. 395-428
43 Peterson, R.E. In 'Metal Fatigue,' McGraw-Hill, New York,
1 Garwood, M.F., Zurburg, H.H. and Erickson, M.A. In 1959, pp. 193-306
'Interpretation of tests and correlation with service', American 44 Kitagawa, H. and Takahashi, S. In 'Proceedings 2nd Inter-
Society for Metals, 1951, pp. 1-77 national Conference on Mechanical Behavior of Materials',
2 Isibasi, T. and Uryu, T. Rep. Res. Inst. Appl. Mech., Kyushu Boston, 1976, pp. 627-631
University 1952, 1, pp. 107-115 45 Kitagawa, H. and Takahashi, S. Trans. Jpn Soc. Mech. Eng.
3 Ransom, J.T. Trans. ASM 1954, 46, 1254 1979, 45, 1289
4 Frith, P.H.J. Iron Steel Inst. 1955, 180, 26 46 Smith, R.A. Int. J. Fract. 1977, 13, 717
5 Ramsey, P.W. and Kedzie, D.P. Trans. AIME, J. Met. 47 El Haddad, M.H., Topper, T.H. and Smith, K.N. Eng.
1957,9, 401 Fract. Mech. 1979, 11, 573
6 Cummings, H.N., Stulen, F.B. and Schulte, W.C. Prec. 48 Luke, P., Kunz, L., Weiss, B. and Stickler, R. Fatigue
ASTM 1958, 58, 505 Fract. Eng. Mater. Struct. 1986, 9, 195
7 Ineson, E., Clayton-Cave, J. and Taylor, R.J.J. Iron Steel 49 L,.tkM, P., Kunz, L., Weiss, B. and Stickler, R. Fatigue
Inst. 1956, 184, 178 Fract. Eng. Mater. Struct. 1989, 12, 175
8 Ineson, E., Clayton-Cave, J. and Taylor, R.J.J. Iron Steel 50 Murakami, Y., Uemura, Y., Natsume, Y. and Miyakawa, S.
Inst. 1958, 190, 277 Trans. Jpn Soc. Mech. Eng. 1990, 56, 1074
9 Atkinson, M.J.J. Iron Steel Inst. 1960, 195, 64 51 Abdel-Raouf, H., Topper, T.H. and Phimtree, A. Scr. Metall.
10 Uhrus, L.O. Iron Steel Inst. Sp. Rep. 1963, 77, 104 Mater. 1991, 25, 597
11 Duckworth, W.E. and Ineson, E. Clean Steel, Iron Steel Inst. 52 Abdel-Raouf, H., Topper. T.H. and Plumtree, A. Fatigue
Sp. Rep. 1963, 77, 87 Fract. Eng. Mater. Struct. 1992, 15, 895
12 Frost, N.E. Prec. Inst. Mech. Eng. 1959, 173, 811
13 Frost, N.E.J. Mech. Eng. Sci. 1960, 2, 109 53 Abdel-Raouf, H., DuQuesnay, D.L., Topper, T.H. and
14 Frost, N.E.J. Mech. Eng. Sci. 1963, 5, 15 Plumtree, A. Int. J. Fatigue 1992, 14, 57
15 Frost, N.E. and Greenan, A.F. J. Mech. Eng. Sci. 1964, 6, 54 DuQuesnay, D.L., Abdel-Raouf, H., Topper, T.H. and
203 Plumtree, A. Fatigue Fract. Eng. Mater. Struct. 1992, 15, 979
16 Frost, N.E. and Greenan, A.F. Z Mech. Eng. Sci. 1967, 9, 55 DuQuesnay, D.L., AbdeI-Raouf, H., Topper, T.H. and
234 Plumtree, A. In 'Proceedings Fatigue '93', Montreal, Quebec,
17 Frost, N.E., Marsh, K.J. and Peek, L.P. 'Metal Fatigue', Canada, 1993, vol. I pp. 239-244
Oxford University Press, Oxford, 1974 56 Gustavsson, A.I. and Melander, A. Fatigue Fract. Eng.
18 Isibasi, T. and Uryu, T. Rep. Res. Inst. Appl. Mech. Kyushu Mater. Struct. 1992, 15, 881
University 1953, !I, 65-74 57 Larsson, M., Melander, A. and Nordgren, A. 'The effect of
19 Isibasi, T. 'Fatigue of Metals and Prevention of Fracture', inclusions on the fatigue behaviour of a hardened spring steel
Yokendo, Tokyo, 1967 - An experimental and theoretical study', Report IM-2808,
20 Irwin, G.R. Trans. ASME, J. Appl. Mech. 1957, 24, 361 Swedish Institute for Metals Research, Stockholm, 1992
21 Pads, P.C. and Erdogan, F. Trans. ASME. J. Basic Eng. 58 Usami, S. and Shida, S. Fatigue Eng. Mater. Struct. 1979, 1,
1963, ~ , 528 471
22 Biicklund, J., Blom, A.F. and Beevers, C.J. (Eds) 'Proceed- 59 Dugdale, D.S.J. Mech. Phys. Solids 1960, 8, 100
ings of International Conference on Fatigue Thresholds: 60 Isibasi, T. Memo Fac. Eng. Kyushu University 1948, 11, 1

Fatigue, 1994, Vol 16, April 181


Effects of defects, inclusions and inhomogeneities on fatigue strength: Y. Murakami and M. Endo

61 de Kazinczy, F. J. Iron Steel Inst. 1970, 208, 851 97 Murakami, Y. and Endo, T. Int. J. Fatigue 1980, 2, 23
62 Tanaka, K., Nakai, Y. and Yamashita, M. Int. J. Fract. 1981, 98 Murakami, Y. and Endo, T. In 'Proceedings Fatigue '81,
17, 519 Materials Experimentation and Design', Warwick University,
63 Taylor, D. Fatigue Eng. Mater. Struct. 1984, 7, 267 1981, pp. 431-440
64 Taylor, D. Int. J. Fatigue 1992, 14, 163 99 Murakami, Y., Tazunoki, Y. and Endo, T. Metall. Trans.
65 Taylor, D. and Knott, J.F. Fatigue Eng. Mater. Struct. 1981, 1984, ISA, 2029
4, 147 100 Murakami, Y. and Matsuda, K. In 'Proceedings Fatigue
66 Taylor, D. In 'The Behaviour of Short Fatigue Cracks' (Ed. '87', University of Virginia, Virginia, USA, 1987, Vol. 1,
K.J. Miller and E.R. de los Rios), EGF Publ. 1, Mechanical pp. 333-342
Engineering Publications, London, 1986, pp. 479-490 101 Nisitani, H. and Murakami, Y. Bull. Jpn Soc. Mech. Eng.
67 Kawai, S. and Kasai, K. Fatigue Fract. Eng. Mater. Struct. 1970, 13, 325
1985, g, 115 102 Nisitani, H. In 'Proceedings 1st International Conference
68 Murakami, Y. and Endo, M. In 'Fracture Mechanics, Current Mechanical Behaviour of Materials', Kyoto, Japan, 1972,
Japanese Materials Research (CJMR)', Vol. 8, Elsevier, Voi. II, pp. 312-322
London, 1991, pp. 105-124 103 Nisitani, H. and Endo, M. Eng. Fract. Mech. 1985, 21, 215
69 Murakami, Y., Kodama, S. and Konuma, S. Int. J. Fatigue 104 Biom, A.F. In 'Basic Mechanisms in Fatigue of Metals,
1989, 11,291 Materials Science Monographs 46' (Ed. P. Luk~ig and J.
70 Murakami, Y. and Usuki, H. Int. J. Fatigue 1989, 11, 299 Polfik), Elsevier, Amsterdam, 1988, pp. 305-314
71 Kobayashi, M. et al. In 'Impact of Improved Material Quality 105 Melander, A., Rolfsson, M., Nordgren, A., Jansson, B.,
on Properties, Product Performance, and Design' (Ed. U. Hedberg, H. and Lund, T. Scand. J. Metall. 1991, 20, 229
Muralidharan), MD-Vol. 28, ASME, 1991, pp. 171-183 106 Kawada, Y., Nakazawa, H. and Kodama, S. Trans. Jpn Soc.
72 Murakami, Y. and Nemat-Nasser, S. Eng. Fract. Mech. 1983, Mech. Eng. 1963, 29, 1674
17, 193 107 Yokobori, T. and Nanbu, M. In 'Proceedings International
73 Murakami, Y. Eng. Fract. Mech. 1985, 22, 101 Conference on Fracture', Sendai, Japan, 1966, Vol. 2,
74 Murakami, Y. et al. (Eds) 'Stress Intensity Factors Handbook', pp. 1529-1542
Vol. 2, Pergamon Press, 1987, p. 822 108 Yokobori, T. and Nanbu M. Rep. Res. Inst. Strength Fract.
75 Murakami, Y. and Morinaga, H. and Endo, T. J. Soc. Mater. Mater. Tohoku University 1966, 2, pp. 29-44
Sci. Jpn 1985, 34, 1153 109 Yokobori, T., Sawaki, S., Shono, S. and Kumagai, A. Rep.
76 Toriyama, T. and Murakami, Y. J. Soc. Mater. Sci. Jpn Res. Inst. Strength Fract. Mater. Tohoku University 1976, 12,
1993, 42, 1160 pp. 29-54
77 Murakami, Y. and Shimizu, M. Trans. Jpn Soc. Mech. Eng. 110 Cowling, J.M. and Martin, J.W. Met. Technol. August 1981,
Ser. A 1988, 54, 413 289
78 Murakami, Y. In 'Basic Mechanisms in Fatigue of Metals' 111 Hyzak, J.M. and Bernstein, I.M. Metall. Trans. A 1982, 13,
(Ed. P. Luk~ and J. Pol~lk), Materials Science Monographs 33
46, Elsevier, Amsterdam, 1988, pp. 343-350 112 Hyzak, J. M. and Berustein, I.M. Metall. Trans. A. 1982,
79 Murakami, Y. JSME Int. J. Ser. 1 1989, 32, 167 13, 45
80 Murakami, Y. In 'Micromechanics and Inhomogeneity, The 113 Konuma, S. and Furukawa, T. J. Soc. Mater. Sci. Jpn 1989,
Toshio Mura 65th Anniversary Volume' (Ed. G.J. Wang, M. 38, 1128
Taya and H. Ab6), Springer-Verlag, New York, 1990, 114 Endo, M. and Murakami, Y. Trans. ASME, J. Eng. Mater.
pp. 28.3-295 Technol. 1987, 109, 124
81 Murakami, Y. In 'Impact of Improved Material Quality on 115 Murakami, Y., Kawano, H. and Endo, T. J. Soc. Mater.
Properties, Product Performance, and Design' (Ed. U. Sci. Jpn 1980, 29, 988
Muralidharan), MD-Vol. 28, ASME, 1991, pp. 89-103 116 Frost, N.E., Pook, L.P. and Denton, K. Eng. Fract. Mech.
82 Natsume, Y., Miyakawa, S., Uemura, Y. and Murakami, Y. 1971, 3, 109
In 'Proceedings Fatigue '90', Honolulu, Hawaii, USA, 1990, 117 Kobayashi, H. and Nakazawa, H. In 'Proceedings 1st
pp. 349-354 International Conference Mechanical Behaviour of Materials',
83 Murakami, Y. In 'Proceedings Fatigue '90', Honolulu, Hawaii, Kyoto, Japan, 1972, Vol. II, pp. 199-208
USA, 1990, pp. 377-382 118 Romanov, O.N., Siminkovich, V.N. and Tkack, A.N. In
84 Murakami, Y., Kawakami, K. and Saito, M. J. Spring Jpn, 'proceedings International Conference on Fatigue Thresholds:
Soc. Spring 1990, 35, 1 Fundamentals and Engineering Applications', Stockholm,
85 Nishimura, A. and Minata, O. J. Soc. Mater. Sci. Jpn 1988, Sweden, 1981, Vol. II, pp. 799-807
37, 434 119 Lies, B.N., Kanninen, M.F., Hopper, A.T., Ahmad, J. and
86 Murakami, Y., Matsuda, K., Ohkomori, Y., Kitagawa, I. Brock, D. Battelle Columbus Laboratories Rep. No. AFWAL-
and Shinozaki, K. In 'Proceedings of International Conference TR-83-4019, 1983, pp. 1-135
on Evaluation of Materials Performance in Severe Environ- 120 Lankford, J. Eng. Fract. Mech. 1977, 9, 617
ments (EVALMAT89)', Kobe, Japan, 1989, pp. 143-150 121 Lankford, J. Int. J. Fract. 1976, 12, 155
87 Murakami, Y., Ikeda, H. and Toriyama, T. In 'Proceedings 122 Lankford, J. and Kusenberger, F.N. Metall. Trans. Set. A
ICF6', 1991, pp. 433-438 1973, 4, 553
88 Emura, H. and Asami, K. Trans. Jpn Soc. Mech. Eng. Ser. 123 Lankford, J. Int. Met. Rev. 1977, 221
A 1989, 55, 45 124 Eid, N.M.A. and Thomason, P.F. Acta Metall. 1979, 27,
89 Endo, M. J. Soc. Mater. Jpn 1989, 38, 1139 1239
90 Sakai, T., Kogo, S. and Suzuld, M. J. Soc. Mater. Sci. Jpn 125 Araki, T., Shih, T. and Sagawa, R. Tetsu-to-Hagan~ Iron
1989, 38, 1268 Steel Inst Japan, 1971, 57, 2042
91 Endo, M. In 'Proceedings Fatigue '90', Honolulu, Hawaii, 126 Kunio, T., Shimizu, M., Yamada, K., Sakura, K. and
USA, 1990, pp. 1357-1362 Yamamoto, T. Int. J. Fract. 1981, 17, 111
92 Endo, M. In 'Impact of Improved Material Quality on 127 Murakami, Y., Kawakami, K. and Duckworth, W.E. Tetsu-
Properties, Product Performance, and Design', (Ed. U. to-Hagan# Iron Steel Inst Japan, 1991, 77, 163
Muralidharan), MD-Vol. 28, ASME, 1991, pp. 125-137 128 Larsson, M., Melander, A., Blom, R. and Preston, S. Mater.
93 Sugiyama, Y., Asami, K. and Matsuoka, S. Trans. Jpn Soc. Sci. Technol. 1991, 7, 998
Mech. Eng. Ser. A 1992, 58, 2287 129 Gumbel, E.J. 'Statistics of Extremes', Columbia University
94 Yano, K., Oda, S., Koide, T., Goka, M. and Ozaki, T. Press, New York, 1957
Trans. Jpn Soc. Mech. Eng. Set. C 1992, 58, 2212 130 Toriyama, T., Murakami Y. and Makino, T. J. Soc. Mater.
95 Endo, M. and Wang, X. In 'Proc. Conf. Strength of Ductile Sci. Jpn 1991, 40, 1497
Cast Iron and Other Cast Metals', Kitakyushu, Japan, 30 131 Berns, H. and Trojahn, W. VDI Z 1985, 127, 889
July-1 August 1993, pp. 51-71 132 Berns, H., Lueg, J., Trojahn, W., W~ding, R. and WiseR,
96 Murakami, Y. and Endo, M. In 'Proc. Conf. Theoretical H. Powder MetaU. Int. 1987, 19, 22
Concepts and Numerical Analysis of Fatigue', Birmingham, 133 Berns, H. In 'Proc. Fatigue '87', University of Virginia, USA,
25-27 May 1992, 1992, pp. 51-71 1987, Vol. 1, pp. 527-536

182 Fatigue, 1994, Vol 16, April

S-ar putea să vă placă și