19

Aspects of fatigue crack growth

~ ~

M W Brown, MA, PhD

Department of Mechanical Engineering, University of Sheffield

The propagation of fatigue cracks under low levels of cyclic stress can be modelled by linear elasticfiacture mechanics. Elastic-plastic
parameters must be used for the higher stress amplitudes of conventional S-N curves in both the low and the high cycle fatigue regimes.
The occurrence of different modes of crack growth is discussed and the use of a propagation law to predict component lifetime is
demonstrated.
NOTATION propagation may be assisted by the joining of a number
of cracks to form the critical crack at failure. Fatigue
crack length cracks may nucleate at a variety of sites such as notches
constant or regions of stress concentration produced by the
Young modulus
geometry of a component under load, small defects in
modulus of rigidity
the material, larger defects produced in manufacture or
path independent J integral
in welds, inclusions or microstructural discontinuities
cyclic strength coefficient formed by plasticity in a persistent slip band. By what-
stress intensity factor ever process a particular fatigue crack may be formed,
strain intensity factor detailed studies of crack development in laboratory
constant specimens have shown that under constant amplitude
strain-hardening exponent
loading minute propagating cracks are present after
number of cycles only 1 or 2 per cent of life has expired (1). Extrapolation
applied tensile load back to the start of the test indicates that microscopic
load at first yield cracks initiate in the first few cycles.
crack closure ratio This paper is concerned with the growth of fatigue
stress ratio, t ~ , , , ~ , , / ~ ~ ~ cracks in metals once they have passed out of the first
maximum plastic zone size
stage, in which propagation is dominated by micro-
plastic zone size ahead of crack structural inhomogeneities and cracks are generally
geometry factor contained within a single grain (2). Such fatigue cracks
plastic zone constant, equation (4) may be large, that is several centimetres long, or in
maximum engineering shear strain
regions of high stress they can be small, that is down to
crack tip opening displacement
0.1 mm. Many models for the growth of such cracks
range of stress or strain
have been proposed (3, 4), but the most useful single
normal strain
development has been the application of fracture
normal strain on maximum shear strain mechanics to the problem of fatigue.
elastic strain
For long cracks, that is those experiencing low levels
plastic strain
of stress where linear elastic fracture mechanics (LEFM)
constant of proportionality
Poisson ratio can be used, predictions of crack extension in engineer-
ing structures can be made from laboratory crack
stress
growth data (5) by using handbooks of stress intensity
von Mises equivalent stress
factors for different component geometries (6-8). There
cohesive stress in yield zone
are many useful books and reviews of LEFM behaviour
tensile strength
yield stress (5, %13), and therefore only a brief outline of the behav-
iour of long cracks is given below.
For small cracks, stresses must be higher if the cracks
are to be able to grow above the threshold level, and
1 INTRODUCI'ION therefore the more elaborate methods of elasto-plastic
Dhenomenon of fatigue has been studied for 150 fracture mechanics (EPFM) must be used. This remains
years: and for the majohty of that time it has been an area of current research and debate, but it is an
widely accepted that fatigue fracture is the final result of important area that cnables fatigue cracks to be
the growth of cracks, More recently it has become clear analysed beyond the limitations of LEFM (14). The tra-
that the whole of the fatigue life of a structure is con- ditional approach to fatigue design involves much
cerned with the propagation of a crack which for much laboratory testing to generate both S-N curves and
of its lifetime may be very small and unobservable when strain-life curves in order to characterize a material.
using standard crack detection methods. Alternatively, However, it has been demonstrated that the S-N curve
since a number of small cracks are often initiated, crack (l),the strain-life curve (15, 16) and even the fatigue
limit (2) may be derived by integration of EPFM crack
Thu paper was presented at an internatioMI cotference on 'Fatigue ofengineerin# growth laws to calculate fatigue endurance. Therefore
materials and structures' held in S W u U on 15-19 September 1986. The MS was
received on I 0 February 1987 and was accepted for publication on 25 August some of the more successful methods for EPFM
1987. analysis are outlined below, together with a discussion
9/88 6IMechE 1988 0263-7154/88 $200 +M Roc Imtr Mech Engrs Vol202 No Cl

Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016

20 M W BROWN

!/I
on their limitations. Finally an example of low-cycle
fatigue life prediction is given.

2 LINEAR ELASTIC FRACTURE MECHANICS I

I
Conventional fracture mechanics describes the stresses 1 ,
and strains around the tip of a crack in terms of the
stress intensity factor, K. Paris and Erdogan (17) Region 1 I Region2
showed that the rate of fatigue crack growth could also
Near threshold
I Power law
be related to the range of stress intensity factor, AK,
through a power law (Fig. la): growth I growth

da I
-= C(AK)"' I Region 3
dN
where C and m are material constants which are either
Microstructure
jependant mode
IlII cracks
I I Rapid growth
obtained by experiment or found in data tables (5), and
I
AK = K,, - Kmjn
I/
Here K,,, and Kminare the maximum and minimum
values of the stress intensity factor during a fatigue
cycle, having the general form
K = YaJ(na) (3)
where (T is the applied stress and a is the crack length. 'I
I
I
I
The dimensionless geometry factor Y is of the order of I
unity for cracks which are small compared to the width I
of a component, and actual values are recorded in
Log stress intensity factor range
various handbooks (a) For. most engineering com-
ponents, Y may be estimated by finding a similar
geometry in the handbook for which the stress intensity (a) Propagation rate
has been calculated previously.
Alternatively, a K calibration can be obtained by the
finite element method for the actual geometry of the
component and an assumed crack path (18). For much stages
of their lives, cracks grow in mode I (as defined in Fig.
lb), which corresponds to stage I1 crack growth along a
plane normal to the maximum principal stress (19). t Ao t Ao
Here the crack is opened cyclically by the stress applied
normal to the crack plane and growth rate is given by
the Paris law, being the linear region 2 shown in Fig.
la. Thus in the complex stress field of a cracked com-
ponent a crack will follow the path denoted by
maximum K , (18) if it is a stage I1 crack.
The reason that AK provides a unique parameter for
determining crack growth is that K characterizes the
crack tip stress field. An example is shown in Fig. 2a
where the effective stress ahead of a crack has been
found by finite element analysis for various. applied
loads P (20). Since the stress variation close to the crack
tip is given by a straight line relationship on these Modes
logarithmic axes, the various curves for each stress level
may be normalized by plastic zone size ahead of the
crack tip, r,, as shown in Fig. 2b, in order to give a @
single line representative of stress near the crack, where
the fatigue damage processes take place. So the plastic
zone size characterizes the stress field intensity and this
in turn is related to K, where for a yield stress cry and
small-scale yielding conditions

r, = y;( (4) 7

with /? = 2 for plane stress or 6 for plane strain. Figure (b) Stages and modes
2c shows that the crack growth data of Solomon (21)
for a mild steel depends on applied plastic strain range Fig. 1 Fatigue crack growth behaviour
Roc lnstn Mcch Engrs VolZO2 No C1 Q IMechE 1988

Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016

 21
b
Q
c
0
Q IMechE 1988 Proc Instn Mech Engrs Vol 202 No C1
Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016
22 M W BROWN

loo I , , , , ,, , 1
(2) by the value of K at which the crack opens during
Aa = 64 MPa the loading cycle.
R = 0 20°C Figure l b shows that the early part of fatigue crack
growth (stage I) is in fact a mixed mode problem, com-
bining modes I and 11. This may also be related to
Key region 1 in Fig. la, where cracking is observed to be of
v Shear a crystallographic nature close to threshold, and there-
0 Uniaxial fore highly dependent on microstructure. Locally the
0 Equi-axial
crystallographic crack experiences a mixed mode
loading, depending on the orientation of the relevant
0
slip system for propagation. Figure 3 shows an example
of both near-threshold growth and the power law
0 region for AISI 316 stainless steel (23), measured under
three different types of applied biaxial stress field.
V Clearly the parameter K is well able to describe fatigue
00
V
crack growth in the complex stress fields usually found
in components as well as the simple uniaxial stress situ-
V %O ation employed in the laboratory.
VQ Crack growth curves for a range of materials are
V& available in the literature (lo), as are tables of the con-
stants C and rn in equation (1) and values for the thresh-
old stress intensity factor, AK,,, , below which no LEFM
crack growth can take place (11). A useful observation
(9, 11) is that if AK is divided by the Young modulus
then the power law regions for many different metals
come together within a single band. Thus the strain
parameter AK JE appears to determine crack growth
Q rate, irrespective of other material properties, and may
0
0 therefore be considered as a fundamental crack growth
vo parameter.
00

3 ELASTIC-PLASTIC FRACTURE MECHANICS

The applicability of LEFM is limited by the applied
0
stress level since, for valid calculations, only small-scale
yielding can be permitted at the tip of a crack. Equation
0 (4), taken with the generally accepted small-scale yield-
ing requirement r, < a/50, provides the restriction that
the stress should not exceed one-third of the yield stress
(2). This is a strict requirement since for many metals
AK the fatigue limit is of the order of yield strength. For
MPa \/m
engineering components therefore it is important to
Fig. 3 Fatigue crack growth at low stress (23) develop EPFM methods for crack propagation
analyses.
Figure 4a shows the effect on crack growth of varying
in exactly the same way as mJa, where K is a fitted the stress range in stainless steel (by = 395 MPa), which
constant of proportionality to correlate with crack illustrates the generally observed fact that EPFM
growth rate. cracks grow faster than the LEFM predictions as is wit-
In equation (2), the maximum value of K clearly cor- nessed at the two highest stress ranges shown. In this
responds with the instant of peak tensile stress in a figure, AK is plotted as the effective stress intensity
cycle. However, K cannot take a negative value as com- factor range,
pressive loads are carried by the faces of a crack, avoid-
ing the intensification of stress around the crack tip. It AK = K,, - KO, (5)
is therefore conventional to take Kminequal to zero for
cyclic loading when R < 0, where R = t~,$u,,,, on the where K O , corresponds to the crack opening load in a
assumption that the crack will close up at zero load. fully reversed load cycle under plane strain conditions
Although crack closure has attracted much attention (24). Another example of the breakdown of LEFM
recently, there is much to commend this simple LEFM characterization of fatigue crack growth is shown in
rule (22), at least until clear models for closure under Fig. 4b, where crack growth under three different
both plane stress and plane strain conditions become biaxial stress states is unable to be correlated by the AK
established. Even with such models, the effects of frac- parameter (23). Note that here the applied stress ampli-
ture surface roughness and the formation of oxides on tude is just half of the monotonic yield stress.
crack closure loads will be difficult to quantify. One EPFM approach to this problem is to replace
However, one should strictly replace Kminin equation AK by a strain intensity factor (25) in equation (l), since
Proc Insln Mech Engrs Vol 202 N o C1 Q IMechE 1988

Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016

 ASPECTS OF FATIGUE CRACK GROWTH 23

I
c L
316 stainless steel
0
Ao = 386 MPa V

lo4
R = -1 20°C

0 Equi-biaxial V
0
0 Uniaxial
X 191 V Shear 8

El 560 V 0 0
lo3 0 on
v V

V
0
lo00 - 0 0
o n
v o
v o
v o

i
v O

$ vv 0
i 0
o
10
$ 0 0"
$ 0 8"
0
h = PJP,
100 - 0 0
0
00,"
1 .o
0 1 p,

0.1
1 10
AK
MPa\/m

(a) Due to high stress levels

10
10 100
AK
MPa\/m
(b) Due to stress biaxially

Fig. 4 Breakdown of LEFM behaviour

fatigue is generally believed to be a strain controlled gave a fundamental crack growth correlation. Thus AK&
process, such that, for uniaxial loading, is applicable to both LEFM and EPFM regimes, and
equation (6), an empirically based expression, can be
+
AK, = (A&, qA&/(na) (6) used directly for EPFM if the Paris law constants are
where q is the proportion of the stress range for which known. Just as AK represents the crack tip stress field
the crack is open (14). Typically q = 0.5 for fully in LEFM, A& bears some relationship to the crack tip
reversed loading, and E AKEis equal to the stress inten- strains, although this has not been rigorously defined.
sity factor range for linear elastic conditions with The strain intensity factor is related to the cyclic J
A&, = 0. integral (26) a concept developed by Dowling (27) that
This is a useful extension to the Paris law, since for has enjoyed wide application in representing high strain
elastic loading AKe and AK are directly proportional. fatigue crack growth (1S16, 26), although it is not
The concept of AKE for EPFM situations is based on always successful in correlating data (15, 16, 24). The J
fatigue being a strain controlled process, as was also integral, defined by Rice (a), is a parameter that char-
noted above for LEFM where the parameter AKJE acterizes the stresses at the tip of a crack in a non-linear
0 IMechE 1988 Proc Instn Mech Engrs Vol 202 No C1

Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016

24 M W BROWN

elastic material. As such, no actual unloading of stress plane strain, as defined in reference (24). Equation (10)
can be allowed, although this occurs in real elastic- has been extended to conditions beyond general yield to
plastic materials in every load cycle. The concept of the give (15)
parameter AJ in cyclic loading must therefore be
divorced from the original concept of the J integral, as
it cannot represent actual crack tip stress values in the
same way that AK can in the elastic case. So A J is a
6= 5
nE'
In( 1 + 2) + AaAe { a'
2a0(l + n)
+ ...I (12)

purely empirical parameter, relating stress and strain where A = 2 for plastic relaxation along two 45" slip
ranges to crack length. For example, for a thumb-nail lines at the crack tip (see Fig. lb), or A = 4 for the clas-
crack (27, sical strip yield model [equation (lo)]. It follows that J
also can incorporate a plastic strain term by combining
-
A J = (3.2 Au2
2E
+ 5 AuA%) a
l+n
(7) equations (1 1) and (12). Alternative procedures for esti-
mating J are outlined by Chell(29).
Note that in equations (9) and (lo), if the applied
where n is the cyclic strain-hardening exponent. Equa- stress a is equal to the cohesive stress, the crack opening
tion (7) showed good correlation of crack growth data displacement is infinite and collapse takes place. Thus
for a variety of specimen geometries and stress levels in for realistic modelling of physical crack tip conditions,
A533B pressure vessel steel. For elastic loading, the cohesive stress should be equated with tensile
strength, or twice the tensile strength for cyclic loading.
A K ~ This is a more sensible choice than yield stress, as cohe-
AJ =-
E' sive stress represents flow stress at the crack tip. For
strain-hardening materials, flow stress will relate to
where E = E for plane stress and E' z E/(1 - v2) for tensile strength (31), and it also provides the correct
plane strain (29), and therefore EPFM analysis can upper-bound solution for collapse load.
embrace the Paris law also [equation (l)]. Thus AJ is a Figure 5 shows some crack growth data for AISI 316
useful parameter, but it must be verified as valid for any stainless steel, including the results in Figs 3 and 4, and
given material since it is purely an empirical formula- compares three candidate EPFM parameters, namely
tion that has no fundamental basis. However, there are a,, rpc and J, . Each parameter has been derived as the
arguments based on 'plastic superposition' that relate cyclic variation between maximum load and crack
AJ to the cyclic change of crack tip field variables (29). closure load. Details of the derivation of closure load
Another parameter that has received a lot of atten- and the three parameters are given in reference (24). The
tion is crack tip opening displacement 6, on the basis effect of biaxial loading is to modify the cohesive stress
that fresh surface created at the crack tip when applying value in equations (9), (10) and (1l), a compressive load
a load must provide the crack extension when the load P , (see Fig. 4b) producing a lower a. and a larger
is reversed, in order to close the crack (15, 30). Thus one plastic zone size. Various biaxial stress states and cyclic
would expect a crack extension of 612 per cycle, but in stress levels have been combined in Fig. 5 to show that
practice this value is an upper bound to measured crack for this stainless steel the most effective parameter is the
propagation rates (15), as can be seen in Fig. 2c. cyclic plastic zone size. Note also that the different
A fourth EPFM parameter is plastic zone size, which EPFM parameters are not equivalent, and selection of
also characterizes the crack tip stress strain field as pre- the optimum correlation parameter is to be recom-
viously discussed (see Fig. 2). This has an obvious physi- mended for any given material.
cal meaning, unlike the AJ parameter, and it is also In evaluating the parameters for Fig. 5, equations (9),
related to both AJ and 6. Use of the strip yield plasticity (10) and (11) were used with a replaced by stress range,
model (24) gives the plastic zone size for a mode I crack An, from maximum stress to the crack opening stress,
in a large plate of length 2a as and with no taken as 2au, where nu is the ultimate
tensile strength. This enabled the cyclic variation of 6, rp
rp = a(,,, (2) 1) - (9) and J to be derived (24), and provided a direct analogy
to AK [equation (5)] under LEFM conditions. The
tensile strength 6, was used to represent the cyclic flow
where a, is the cohesive stress within the plastic zone. stress of the material after strain hardening (31). In
For applied stresses below general yield addition, the value of u, was modified to account for
stress biaxiality effects, as described in reference (24).
For low-cycle fatigue, a particularly successful crack
growth model that addresses the contribution of plastic
strain range to crack extension is due to Tomkins (31),
and who considered the decohesion of material due to
J = Ma,6 (11) plastic flow along the 45" intense shear bands found at
the tips of cracks in strain-hardening materials (see Fig.
where M is equal to unity in the strip yield model. lb). Using this model (16),
[Other values for M of about 2 can be obtained for
different plastic zone shapes, but equation (11) is found da
to have wide validity.] = TP d&P
The cohesive stress is generally taken as twice the
yield stress for cyclic loading, but strictly it also depends (13)
on the applied stress biaxiality and the Poisson ratio for
Proc Instn Mech Engrs Vol 202 No C1 Q IMechE 1988

Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016

 ASPECTS OF FATIGUE CRACK GROWTH 25

-
-
0.1
0.01

1.o

6,
-
Irm
0.1

10 Tpc
mm
1 .O

lo3

Fig. 5 Characterization of fatigue crack growth by EPFM parameters 6, rp

and AJ (24)
u
1o4

Jc
Nlm
105

where 6, is the tensile strength or the flow stress in the dimensional analysis methods, or the self-similarity
shear band. The cyclic stress strain curve is given by the concept for growing cracks (1, 16,20). These derivations
Ramberg-Osgood formulation relate cyclic crack extension to crack length, or alterna-
tively to plastic zone size as in equation (16), and
predict a value for m of 2. This is in contrast to models
that relate crack extension to energy release at the crack
tip, which invariably predict m = 4 from an energy
and the hysteresis loop shape is assumed to be given by
balance equation (13). Many empirical results give rn
equation (14) magnified by a factor of 2. Equation (13) values between 2 and 4, but it is a consequence of curve-
may be written more concisely in terms of the cyclic fitting procedures that m is likely to be overestimated
plastic zone size to give for a sigmoidal shape of the form illustrated in Fig. la.
da Figures 3 and 4a show m = 2 at low AK values, whereas
- AS
at higher levels of K, m increases to 3.2 because the
d N = " 1 +;n
plastic zone size becomes more significant and the self-
whereas for LEFM behaviour, equations (1) and (4) similarity concept breaks down.
may be combined to give, for the centre cracked panel,
4 CRACK GROWTH MODES
The preceding sections have been concerned with mode
I cracks, on which almost all fatigue crack propagation
For a full description of both LEFM and EPFM studies have been concentrated. However, crack growth
regimes, equations (15) and (16) should be summed to due to modes I1 and I11 is found in preference to mode I
give under certain circumstances. Figure 6 shows two tradi-
tional methods of plotting crack growth rate results in
da =
- + rpA E ~ terms of the parameters AK and a respectively, together
dN 1 + 2n
with a third fatigue diagram on which the preferred
as a general crack growth equation. Note that rp must mode of cracking can be identified. The right-hand side
be evaluated from the tensile strength in this expression. of Fig. 6a shows the usual LEFM line, with parallel
Frequent use has been made of derivations based on lines for higher stresses in EPFM, as was plotted in Fig.
8 IMechE 1988 Proc Instn Mcch Engrs Vol 202 N o C1

Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016

26 M W BROWN

Increasing EPFMmodeI
stress range crack growth

+ Short cracks +Long cracks 4-

r
I
//

: Log (crack length)

Tensile strength 20,

Short cracks EPFM

Crystallographic

I
E
(C) : Log (crack length)
Fig. 6 Crack growth diagrams for R = - 1 (a) versus AK, (b) versus crack
length, (c) six regimes of propagation

4a. However, for short crack lengths less than some form (18). However, the early mode I1 growth took
critical microstructural dimension [frequently found to place below the mode I threshold (as shown in Fig. 7).
be grain size (111, very much faster propagation rates Co-planar mode I1 cracking was observed in pure shear
can be obtained from microstructural crack extension tests on stainless steel at 0.65 AK,,,. The reduction in
mechanisms (2), which are able to operate below the threshold for mode I1 growth has been predicted from
threshold level, for example lines C D and AB on plastic zone size calculations, since more extensive plas-
Fig. 6a. ticity is obtained in shear loading (33). The cracks
Figure 6b shows clearly that EPFM equations cannot branched to follow a mode I path at much higher
be used for a crack size smaller than that defined by the stresses than the threshold level, as shown by the
dominant microstructural barrier. This early regime of dashed line in Fig. 7.
propagation relates to stage I or shear cracks (Fig. lb), Mode 111 cracks can also be found in most metals at
as observed by Forsyth (19), which have a predominant high stress levels. In Fig. la, region 3 corresponds to
component of mode I1 loading since they grow on crys- accelerated mode 111 growth where cracks transfer to a
tallographic slip planes. Such small cracks are able to 45" shear or slanted plane in standard fracture mecha-
form below the fatigue limit stress (32), but it is the nics tests (3). However, mode I11 is most easily studied
mode I fatigue crack propagation threshold that defines in torsional specimens, and it is associated with gross
the actual fatigue limit value in steels (2), as plotted in yielding of the cross-section. Therefore the EPFM mode
Fig. 6c. 111 regime appears only at the highest stress levels in
Crystallographic mode I1 cracks have been observed Fig. 6c. Here crack growth is found to be rapid (M), and
in mixed mode 1/11 propagation experiments, but they once a true mode I11 crack has begun to grow, an equa-
were unable to grow for more than one or two grain tion based on a strain intensity factor similar to the
diameters without changing direction to adopt a mode I Paris law can be derived, although high values for m of
Proc lnstn Mech Engrs Vol202 No Cl @ IMechE 1988

Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016

 ASPECTS OF FATIGUE CRACK GROWTH 21

5 WORKED EXAMPLE: MULTIAXIAL FATIGUE

\ $1 AISI 3 16 stainless steel
LIFE PREDICIlON
The results of mode I biaxial fatigue crack growth
experiments on RR58 aluminium alloy (37) may be used
to predict the fatigue endurance of tubes subjected to
combined internal pressure and axial load. Such endur-
ance tests have been conducted on RR58 tubes (38), of
25.4 mm inner diameter and 1 mm wall thickness. The
honed bore for this soft material produced a 15 pm
surface finish, being the depth of the deepest scratch,
and fatigue cracks were initiated from honing marks.
Failure was defined from diametral strain measure-
ments, and shown to correspond to a crack of 0.1 mm
depth, or one-tenth of the wall thickness (39). Thus
initial and final crack lengths for the crack propagation
m analysis were taken as a, = 0.015 mm and a, = 0.10mm
respectively.
The crack growth data conformed with the Paris law
[equation (l)], with rn = 3.377 and C = 4.29 x lo-"
for AK in MPaJm and crack growth in m/cycle under
uniaxial loading. However, for shear loading
C = 6.29 x lo-" and for equi-biaxial tension
C = 3.43 x lo-" at the applied stress range of 91
MPa. Since the bore cracks were formed on polishing
marks they adopted a wide and shallow shape, to
produce edge cracks with a geometry factor Y = 1.12 in
equation (3). Thus for a bore crack under LEFM condi-
tions,
0 0.2 0.4 0.6 0.8 1.o
da
AKI"%1 -= C[I.I~AO,/(Z~Z)]~*~~~
dN
Fig. 7 Mixed mode crack growth thresholds for inclined
cracks (18) (symbols denote different specimen Since the endurance tests were conducted above yield,
geometries) an EPFM analysis is required, and the simplest
approach is to use a strain intensity factor. Equation
(18) must be changed to give crack growth rate as a
function of applied strain range, and for multiaxial
the order of 6 are obtained. Thus the mode I11 regime loading an appropriate strain is the maximum shear
should be avoided if at all possible in engineering strain range, which may be determined for any stress
design, although practical examples of mode 111 failures state. For biaxial loading, if AO corresponds to the
can be found (34, 35). maximum principal stress normal to a mode I crack,
The fatigue diagram, Fig. 6c, may be used to iden- then the maximum shear strain range in the propaga-
tify the relevant failure mechanism in fatigue. The tion experiments is
graphs in Fig. 6 were constructed for a medium carbon
steel using a strain intensity factor equation, together A0
with a second equation for short crack growth (2). A AY max =
similar diagram can be constructed for any material
once these two equations, the threshold value and the where G is the shear modulus (28.2GPa). The constant
characteristic microstructural dimension have been a = 2 for uniaxial and equi-biaxial loading, but o! = 1
established. An important region in Fig. 6c is that for shear loading from Mohr's circle of stress. Integ-
labelled 'non-propagating' cracks. It is this region that ration for crack propagation life gives, from equations
is made use of in traditional fatigue design for infinite (18) and (19),
endurance, since any inherent defects or initiated fatigue
cracks must be kept below the threshold level to avoid
failure. Should periodic overloads occur or internal
residual stresses exist the stress range may exceed the
threshold line at any time, and under these conditions
I:'+= C(1.12crCAy,J~)~.~

Equation (20) is valid for both LEFM and EPFM, and,

"
r dN (20)

crack extension can take place and failure is liable to for a given endurance, may be solved to give the fatigue
occur. strength A y,. The results are plotted in Fig. 8 on a
There are many aspects to the different types of crack r plane (M),which shows the calculated shear strain
growth represented in Fig. 6, which cannot be fully amplitude on the abscissa and the normal strain ampli-
covered in this paper. A recent publication edited by tude E, acting on the maximum shear plane as the ordi-
Smith (36)provides a modern perspective of fatigue nate. The calculated dashed lines show good agreement
crack propagation, with reference to the LEFM regime with the experimental results of Ellison and Andrews
in particular. (38) at all but the lowest endurance.
0 IMechE 1988 Proc Instn Mech Engrs Vol 202 N o C1

Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016

2a M W BROWN

0 L

0.5 Ymsx

1 o CaseA
XCaseB
---- Prediction from crack
-0.2/ propagation equation

(a) Strain amplitudes

Crack growth planes Crack shape and major growth direction

Stage I Stage I1 Stage I Stage I1

€2
Case A

--YE

Uniaxial

& 3@ & 2

Case B

(b) Crack growth planes and directions

Fig. 8 Predicted and observed fatigue strengths of biaxially stressed tubes of
RR58 aluminium alloy

Note that two types of crack are identified, case A At higher stress levels an elasto-plastic fracture
and case B (a), as defined in Fig. 8b. It is always found mechanics parameter, such as the strain intensity
that case B gives a lower endurance than case A for a factor, the J integral, the crack tip opening displace-
given shear strain amplitude. This is because the stress ment or the plastic zone size is required to determine
acting normal to a mode I crack at a given shear strain crack growth rate.
amplitude ymax is greater for case B, giving faster mode I For type 316 stainless steel, plastic zone size is the
propagation rates. most successful parameter for the correlation of
fatigue behaviour.
The mode of fatigue crack propagation may be
assessed from a fatigue diagram to ensure that the
6 CONCLUSIONS
correct crack extension mechanism is selected.
1. For linear elastic fracture mechanics (a < ay/3), The fatigue life of an engineering component may be
fatigue crack growth rates can be characterized by accurately assessed from a crack propagation
the stress intensity factor range. analysis.
Roc Instn Mech Engrs Vol202 No C1 Q IMechE 1988

Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016

 ASPECTS OF FATIGUE CRACK GROWTH 29

REFERENCES 21 Solomon, H. D. Low cycle fatigue crack propagation in 1018 steel.

J. Muter., 1972,7,299-306.
1 Hobson, P. D., Brown, M. W.and de 10s Rim, E. R. Two phases of 22 Skelton, R. P. Crack initiation and growth in simple metal com-
short crack growth in a medium carbon steel. Symposium on ponents during thermal cycling. In Fatigue at high temperature
Behauiour of short fatigue cracks, Shetfield, 1986, pp. 4 4 4 5 9 (Ed. R. P. Skelton), 1983, pp. 1-62 (Applied Science, London).
(Mechanical Engineering Publications, London). 23 Brown, M. W. and Miller, K. J. Mode I fatigue crack growth
2 Brown, M. W.Interfaces between short, long and non-propagating under biaxial stress at room and elevated temperature. In Multi-
cracks. Symposium on Behauiour of short fatigue cracks, Shef- axial fatigue, ASTM STP853, 1985, pp. 135-152 (American Society
field, 1986, pp. 423-439 (Mechanical Engineering Publications, for Testing Materials).
London). 24 Brown, M. W, de 10s Rios, E. R. and Miller, K. J. A critical com-
3 Plumbridge, W.J. and Ryder, D. A. The metallography of fatigue. parison of proposed parameters for high strain fatigue crack
Metall. Rev., 1969, 14, 119-142. growth. Symposium on Fundamental questions and critical experi-
4 Plumbridge, W. J. Review: fatigue crack propagation in metallic ments on fatigue, Dallas, 1984 (American Society for Testing
and polymeric materials. J. Mater. Sci., 1972,7,939-962. Materials).
5 Pook, L. P. Analysis and application of fatigue crack growth data. 25 Boettner, R. C, Laird, C. and McEvily, A. J. Crack nucleation and
In A general introduction to fracture mechanics (Ed. C. E. Turner) growth in high strain-low cycle fatigue. Trans. Metall. SOC.AIME,
1979 pp. 114-135 (Mechanical Engineering Publications, London). 1965,233,379-387.
6 Rooke, D. P. and Cartwright, D. J. A compendium of stress inten- 26 Starkey, M. S. and Skelton, R P. A comparison of the strain inten-
sity factors, 1976 (HMSO, London). sity and cyclic J approaches to crack growth. Fatigue Engng
7 Tada, H., Paris, P. C. and Irwin, G. R. The stress analysis of cracks Mater. Struct., 1982,5329-341.
hundbook, 1973 (Del Research Corporation, Pennsylvania). 27 Dowling, N. E. Crack growth during low cycle fatigue of smooth
8 Murakami, Y. (Ed.) Stress intensity factors handbook, 1987 axial specimens. In Cyclic stress-strain and plastic deformation
(Pergamon Press, Oxford). aspects of fatigue crack growth, ASTM STP637, 1977, pp. 97-121
9 Frost, N. E, Mush, K. J. and Pook, L. P. Metal fatigue, 1974 (American Society for Testing Materials).
(Clarendon Press, Oxford). 28 Rice, J. R. A path independent integral and the approximate
10 CampbeU, J. E., Cerbericb, W.W.and Underwood, J. H. Applica- analysis of strain concentrations by notches and cracks. J. Appl.
tion of fracture mechanics for selection of metallic structural Mech., 1968,35,379-386.
materials, 1982 (American Society for Metals). 29 Chell, G. G. Some subcritical crack growth parameters and their
11 Pook, L P. The role of crack growth in metal fatigue, 1983 (Metals evaluation. In Subcritical crack growth due t o fatigue, stress corro-
Society, London). sion and creep (Ed. L. H. Larsson), 1984, pp. 239-266 (Elsevier
12 Fuchs, H. 0. and Stephens, R. I. Metal fatigue in engineering, 1980 Applied Science, London).
(Wiley Interscience, New York). 30 Pook, L. P. and Frost, N.E. A fatigue crack growth theory. Int. J.
13 Lloyd, G. J. High temperature fatigue and creep-fatigue crack pro- Fracture, 1973,9, 53-61.
pagation : mechanics, mechanisms and observed behaviour in 31 Tomkins, B. Fatigue crack propagation-an analysis. Phil. Mag.,
structural materials. In Fatigue at high temperature (Ed. R. P. 1968,18, 1041-1066.
Skelton), 1983, pp. 187-258 (Applied Science, London). 32 Miller, K. J., Mohamed, H. J. and de 10s Rios, E. R. Fatigue
14 Skelton, R. P. The growth of short cracks during high strain damage accumulation above and below the fatigue limit. Sympo-
fatigue and thermal cycling. In low cycle fatigue and lijie prediction, sium on Behauiour ofshort fatigue cracks, Shefield, 1986, pp. 491-
ASTM STP 770, 1982, pp. 337-381 (American Society for Testing 511 (Mechanical Engineering Publications, London).
Materials). 33 Gao, H, Brown, M. W. and Miller, K. J. Mixed mode fatigue
15 Tomkins, B. High strain fatigue. In Sub-critical crack growth due to thresholds. Fatigue Engng Mater. Struct., 1982,5, 1-17.
fatigue, stress corrosion and creep (Ed.L. H . Larsson), 1984, pp. 34 Brown, M. W, Hay, E. and Miller, K. J. Fatigue at notches sub-
239-266 (Elsevier Applied Science, London). jected to reversed torsion and static axial loads. Fatigue and Frac-
16 Wareing, J. Mechanisms of high temperature fatigue and c r e e p ture Engng Mater. Struct., 1985,8,243-258.
fatigue failure in engineering materials. In Fatigue at high tem- 35 Ritchie, R 0, McCliatock, F. A, Tsebegg, E. K. and Nayeb-
perature (Ed. R. P. Skelton), 1983, pp. 135-185 (Applied Science, Hashemi, H. Mode I11 fatigue crack growth under combined tor-
London). sional and axial loading. In Multiaxial fatigue, ASTM STP853,
17 Paris, P. C. and Erdogan, F. A critical analysis of crack propaga- 1985, pp. 203-227 (American Society for Testing Materials).
tion laws. Trans. ASME, 1963, Ser. D 85,528-534. 36 Smith, R. A. (Ed.) Fatigue crack growth, 30 years of progress, 1986,
18 Gao, H., Alagok, N., Brown, M. W. and Miller, K. J. Growth of (Pergamon Press, Oxford).
fatigue cracks under combined mode I and mode I1 loads. In 37 Miller, K. J. Fatigue under complex stress. Metal Sci., 1977, 11,
Multiaxinl fatigue, ASTM STP853, 1985, pp. 184-202 (American 432-438.
Society for Testing Materials). 38 Ellison, E. G. and Andrews, J. M. H. Biaxial cyclic high strain
19 Forsyth, P. J. E. A two stage process of fatigue crack growth. Pro- fatigue of aluminium alloy RR58. J. Strain Analysis, 1973, 8, 209-
ceedings of Symposium on Crack propagation, Cranfield, 1961, pp. 219.
76-94 (HMSO, London). 39 Andrews, J. M. H.and Ellison, E. G. A testing rig for cycling at
20 Brown, M. W.,Liu, H. W.,Kfouri, A. P. and Miller, K. J. An high biaxial strains. J. Strain Analysis, 1973,8, 168-175.
analysis of fatigue crack growth under yielding conditions. In 40 Brown, M. W.and Miller, K. J. A theory for fatigue failure under
Advances in fracture research (Ed. D. Francois), Vol. 2, 1980, pp. multiaxial stress-strain conditions. Proc. Instn Mech. Engrs, 1973,
891-898 (Pergamon Press, Oxford). 187,745-755.

Q IMechE 1988 Proc lnstn Mcch Engrs Vol 202 No C1

Downloaded from pic.sagepub.com at Purdue University Libraries on June 5, 2016