29 1

Twenty-seventh John Player Lecture

Metal fatigue-past, current and future

K J Miller, PhD, ScD, CEng, FIMechE, FIM, FIProdE, FCGI
Director of SIRIUS, Structural Integrity Research Institute, University of Sheftield

Metal fatigue has been a problem for more than 150 years, but because of rapid developments infracture mechanics analyses, possibly
at the expense of the traditional approach based on cyclic deformation processes, a far better understanding of fatigue failure behaviour
has recently been achieved. Consequently the engineer now has the basic tools at hislher disposal to make good assessments of the
numerous factors that control the fatigue lifetime of engineering materials, components and structures.
Additionally, more intensive interdisciplinary research studies involving chemists, materials scientists, mathematicians and physicists-
but engineering led-have generated both greater insights into long-known industrial problems and routes to required solutions.
This paper traces the growth of recent developments in understanding metal fatiguefrom the days of our mentors to the present day,
and concludes with a brief review of some future research areas that ore now available for exploitation.

1 INTRODUCTION of a total of 80 were concerned with the growth rate

behaviour of a crack.
When attempting to survey the current scene in metal Briefly, my paper will present our current under-
fatigue and to predict future directions of research it is standing of fatigue cracks in terms of their length, rate
frequently wise to look back at the work of past tea- of growth, shape, orientation and coalescence or inter-
chers who, in my undergraduate years, included such ference, since it is these factors which dominate the
figures as Benham, Ford and Turner, while from teach-
failure process. This paper is not intended to be a rigor-
ings of the literature, Crossland, Forsyth, Forrest and ous analytical treatment of this exceedingly wide and
Frost will always be remembered. interdisciplinary subject, but I hope it will provide a
Further important insights were to come later from useful introduction for the non-specialist engineer.
the United States via the works of McClintock, Argon, Ample references will, however, be provided for the frac-
Pelloux, Coffin and Manson. In more recent years, the ture specialist and teacher.
United Kingdom research of Brown, Dugdale, Smith It will be appreciated by all that the progress that
and Tomkins has opened new chapters on the subject of
has been achieved since reference (1) was published has
metal fatigue which have an important bearing on the
been due t o three factors:
understanding required by designers of engineering
plant. Finally, of course come the teachings of so many The advanced nature of experimental facilities, par-
international stars too numerous to mention, although ticularly servo-hydraulic machines including biaxial
reference will be made to several in the course of this and multiaxial test facilities, electron microscopes
Lecture. and now the acoustic microscope, and two- and
Today I give thanks to all my teachers at this, the three-dimensional elastic-plastic finite element solu-
twenty-seventh John Player Lecture, not only for their tions to cracked body problems via supercomputers.
encouragement, but also for helping me to acquire and The willingness of scientists of all related disciplines
maintain a healthy scepticism about what we consider to collaborate to solve real engineering problems.
to be the fundamentals of the fatigue process. In this The need of industry to acquire solutions to more
respect, my Lecture will attempt to illustrate the truth complex problems as plant becomes more efficient
of the statements made by Gough, a Past President of due to increasing levels of applied loads and oper-
the Institution, who remarked in an opening address (1) ating temperatures.
on the changing emphasis and pattern of today’s
subject :
2 FATIGUE DAMAGE
there will naturally persist, for long periods, certain threads
representing problems o f . . . long range character, including The quantification of metal fatigue damage has long
the continuing quest for a clearer understanding of the been the subject of much research and intense dis-
basic mechanisms of deformation and fracture, but the main cussion by materials scientists and mechanical engi-
pattern will change sometimes in a surprisingly sudden and neers. The former concentrated on microstructural
decisive manner aspects, such as the effects of inclusion shape and size,
In the past decade a major shift of emphasis has slip bands and surface features, while the latter empha-
occurred, and this Lecture will underline some of the sized bulk material behaviour, for example Basquin and
recent changes. In particular we now have a far better Coffin-Manson endurance curves. The linking of the
understanding of the fatigue failure process and can cyclic stress-strain response of a metal with the behav-
confidently move from a qualitative understanding to iour of micro and macro defects has now provided a
quantitative analyses. In this respect it should be noted better understanding of fatigue damage than hitherto.
that in reference (l),itself a classic, only two papers out Despite the fact that cyclic loading can change the
deformation response of a metal and its microstructure,
This kcrurr wus presented ut un Ordinary Meeting held in London on 27 March metal fatigue damage can now be equated to crack
1991. The MS wus rtwiurd on I S Fehruury 1991. length and the rate of damage accumulation to the rate
CO1791 0 IMechE 1991 0954-4062/91 $2.00 + .05 Proc Insln Mech Engrs Vol 205
292 K J MILLER

AF2 is attained. In the latter example, because of the

higher cyclic force levels, a greater extent of plasticity
occurs and I EFM is no longer adequate to quantify
precisely fatigue damage and elastic-plastic fracture
LEFM regime mechanics (EPFM) analyses have to be invoked. In this
10-2 - respect it can be seen in Fig. l a that at A F l the fatigue
crack may be growing within one single favourably
c oriented grain of diameter d for a large fraction X of
lifetime. It is for this reason that the microstructure in
EPFM regime terms of inclusions, precipitates and grain size has such
E
- an important effect on fatigue behaviour. Large struc-
tures, particularly welded structures, inherently contain
defects larger than 0.5 mm in size and these may propa-
gate at very low stress levels, permitting analysis by
10-6 - LEFM. Components with smooth surfaces however,
require a more detailed study, involving EPFM and, for
the exceedingly long-life regime, microstructure-based
micromechanics.
Figure 1b illustrates how damage accumulates from
finely polished engineering surfaces. This figure is
n/Nc important because it provides the link with the tradi-
(a) Long (low stress) and short (high stress) cracks and their tional cyclic stress versus number of cycles (S-N)
associated zones of characterization by linear elastic and approach of yesteryear and future investigations. In
elastic-plastic fracture mechanics particular it shows that our previous understanding of
fatigue crack initiation requires modification. In the
past, initiation has been interpreted by engineers as the
establishment of a crack of a given length, for example
1 mm, but as more research was gathered, this was
shown to be inadequate and smaller lengths of 0.5 mm
and 0.2 mm have been cited in the literature. All such
definitions are inadequate as Fig. 1 shows, since they
take no account of the applied stress level, the rate of
crack propagation, the phase of crack extension, or
microstructural features. To resolve this dilemma it is
best to consider first the meaning of the fatigue limit
itself.

3 THE FATIGUE LIMIT

10-71
t Previous reports have directly or indirectly related the
fatigue limit to the inability of a material to initiate a
0 I crack. Figure l b shows that at cyclic stress levels
n/N, Aa, > Aa, > Aa5 failure does not occur since the crack
(b) Short crack growth behaviour as influenced by stops at barriers indicated by b, , b, and b, respectively.
microstructural barriers of different strengths However at A a , , which is slightly greater than the
and spacing fatigue limit, these barriers are not sufficiently strong as
Fig. 1 Crack growth behaviour as a linear function of the to arrest the crack and so failure results.
applied number of cycles Accordingly the fatigue limit is a limit on the ability
of a crack, whatever its length, to propagate to failure.
This definition is reinforced by the knowledge that
of fatigue crack growth. This simple statement is best cracks have often been found to occur in plain and
appreciated with reference to Fig. 1, which has, as its notched specimens at stress levels below the fatigue
foundation, many studies on various metals. limit. Frost and Phillips (3) made many pertinent
In Fig. la, the growth of the crack is shown for remarks about non-propagating cracks at notches and
two typical conditions. Should a defect be 1 mm deep the creation of cracks below the fatigue limit, for
or larger, it can grow with a minimal amount of cyclic example :
plasticity at its tip and hence be analysed by linear A crack can form at a relatively early stage, grow for a
elastic fracture mechanics (LEFM) (2). If the applied short period and then remain dormant.
cyclic force is reduced from AF3 to AF, the defect will Thompson (4) similarly stated :
not grow, that is a fatigue limit is attained. Similarly if
the surface of a specimen or component is finely ground The test ... was stopped at .. . 4 per cent of life. A very few
and polished, then a much larger cyclic force A F l is slip lines were found, which usually went completely across
the grain in which they occurred. Electro-polishing caused
required to cause crack propagation although if this most of them to turn into broad black lines, suggesting that
cyclic force is reduced, then again a threshold condition they were already cracks.
Part C: Journal of Mechanical Engineering Science @ IMechE 1991
 METAL FATIGUE-PAST, CURRENT A N D FUTURE 293

In Fig. lb, the barriers to early propagation, b, , b,, tance. Figure 1 indicates that fatigue failure cracks start
b,, are of increasing strength requiring a greater stress propagating from the very first cycle of loading, but
for them to be overcome. As examples, the weakest may become quiescent for a large part of the lifetime
barrier may be a twin boundary, the intermediate while they have difficulties crossing the barriers. This is
barrier may be a grain boundary, and the strongest
barrier may be a pearlite zone in a ferrite-pearlite
microstructure [see Fig. 2 taken from references (5) to
(911.
Obviously the fatigue limit refers to the stress level
required to overcome the strongest barrier to propaga-
tion which will be represented by a microstructural dis-

(d) Crack growth affected by several grain boundaries in

Nimonic 90 (8)

(a) A crack arrested at a twin boundary in brass (5)

(e) Crack growth affected by the internal grain structure of a

nickel base superalloy (9) [courtesy of Rolls-Royce plc]
(b) A crack arrested at a grain boundary in Waspaloy (6)

(c) A crack arrested at a pearlite zone in a medium carbon (f) Fracture surface facets within grains of a powder Astroloy
steel (7) (9) [courtesy of Rolls-Royce plc]
Fig. 2 Microstructure-influenced fatigue crack growth
0 IMechE 1991 Proc Instn Mech Engrs Vol 205
294 K J MILLER

consistent with fracture mechanics philosophy, which diately develop at the crack tip. Certainly if any defect
tells us that associated with an applied stress or strain exists, irrespective of its size, then that is where one
range is a distance a. In simple LEFM terms, for should expect cyclic plasticity to occur. Even on
example, this is given by a stress intensity factor smooth, highly polished specimens of polycrystalline
metals, defects in the form of surface scratches - 2 pm
AK = YAoJ(.na) (1) deep can be found, as well as precipitates, inclusions
where and triple-point boundaries, all of which can act as
AK = stress intensity factor range stress concentrations and plasticity initiation zones. The
Y = specimen geometry and loading system factor real problem here is one of size identification. Slip
A o = cyclic stress range bands can vary substantially in terms of width and
a = crack length length, while crack-like surface defects can be as small
as fractions of a micrometre and be growing as slow as
and the fatigue limit within the LEFM regime is given l o p 3 pm per cycle. Cyclic softening and cyclic hard-
by the threshold condition A&,, at which, for a given ening will complicate this picture, but certainly it is
cyclic stress level the crack will not propagate, or for a more rational to assume that the fatal crack and its
given crack depth the stress range is insufficient to cause associated plasticity are generated simultaneously in the
propagation. Thus the fatigue limit stress level has to be first cycle of loading.
associated with a given crack size and this is shown in Finally, it should be noted that during cyclic loading,
Fig. 3. many hundreds of individual slip bands may be gener-
The variable fatigue limit stress value for the entire ated in a polycrystalline metal even at stress levels well
range of crack depths of importance to engineers (say below the fatigue limit, but only one very small and
to lo-' m) is given by the condition that the crack extremely difficult to locate crack, and its associated
propagation rate da/dN is zero. Microstructural bar- plasticity, is required to eventually cause failure.
riers of increasing strength, b,, b, and b,, are indicated
on Fig. 3 which match the conditions shown in Fig. lb.
Three different zones may be seen in Fig. 3 which will 4 RATE OF FATIGUE DAMAGE ACCUMULATION
vary in extent dependent on the material under investi- The variation of fatigue crack growth rate associated
gation (10). with Figs 1 and 3 is shown in Fig. 4.
Zone A-B is referred to as the microstructural short At the lowest stress range, Ao, , the short microstruc-
crack growth regime, while zone B-C is referred to as tural crack grows but eventually stops at the fatigue
the physically small crack growth regime. Only in zone limit denoted by barrier b,. It could continue to grow
C-D is it permissible to use LEFM to assess the fatigue at the same stress range, but only if it could be extended
limit condition. Later discussions will indicate that to to a length a , which is the fatigue limit threshold condi-
improve the fatigue resistance of a metal, that is tion for a physically small crack. Even if the stress level
decrease fatigue crack growth rates, a decrease in grain is increased to a level AD,, sufficient to overcome the
size is required in zone A-B while for zone C-D an first barrier, a fatigue limit results due to the separation .
increase in grain size, particularly at low mean stress distance b, to a,. At the stress range level ACT* fatigue
levels, can increase the value of AKt,,. Zone B-C is not failure results due to the overlap of the two curves rep-
significantly affected by grain size, although it is affected resenting the microstructural and physically small crack
by the decreasing effect of microstructural barriers as growth phases.
the crack length increases. The crack growth equations, equivalent to the curves
It is worth making a few comments regarding the of Fig. 4, may be expressed in terms of the shear plastic
question : 'which comes first, the persistent slip band or
the fatigue crack?, since as Fig. l b indicates, the fatigue
crack is seen to be growing from the very first cycle. Microstructural
Obviously if a long crack-like defect exists prior to short crack Physically small
crack growth
loading and eventual failure, cyclic plasticity will imme-
Equation (3)
A B C D

I I EPFM (high stress) and

Microstructural 1 LEFM (low stress) cracks

/
,,

b3 (13 (15

Crack length
Fig. 4 Effect of cyclic stress level and barrier strength on the
propagation or non-propagation of a fatigue crack.
Equations (2) and (3) can be translated to stress range
Crack length
from the plastic strain range via knowledge of the
Fig. 3 Fatigue limit dependency on crack length cyclic stress-strain behaviour of the material
Part C: Journal of Mechanical Engineering Science @ IMechE 1991
 METAL FATIGUE-PAST, CURRENT AND FUTURE 295

strain range since, as will be seen later, this parameter is Navarro and de 10s Rios (16), which is theoretical and
related to the crack growth direction, and furthermore based on dislocation mobility. It is sufficient to report
it provides the necessary link between the Coffin- here that it is now becoming possible to derive S-N
Manson, Morrow and Benham studies of the late 1950s curves and to calculate fatigue limits from a knowledge
and early 1960s to fatigue fracture via the cyclic defor- of the constituents of metals and their microstructure,
mation approach to fatigue endurance. Thus together with knowledge of the cyclic stress-strain
behaviour of the material.
da
- = AAyi(d - a)
dN
5 CONSEQUENCES OF SHORT CRACK
for the microstructural short crack growth regime; here GROWTH BEHAVIOUR
A and tl are material constants, Ayp is the plastic shear The implications and consequences surrounding an
strain range, d is the distance to the strongest barrier, understanding of microstructural and physically small
and a is crack length. For the physically small crack crack growth behaviour are extremely wide-ranging and
growth regime will dominate fatigue research work for the next two
da decades at least. Simply put, all the classical work of the
- = BA$U - C (3) previous 100 years, which investigated long-life fatigue
dN behaviour and fatigue limits, can now be reanalysed in
where B and are material constants and C is the terms of short cracks. For example, those effects which
longer crack threshold condition. More information on were previously known to affect the fatigue limit and
the derivation and formulation of these two simple fatigue life, namely,
equations is given in references (11) and (12). Corrosion environments
It will be noted that for LEFM type cracks in the Grain size
so-called long crack, low-stress regime Mean stress levels
Random loading
da Multiaxial stress states
- = DAK” (4)
dN Fretting
Surface treatments
where AK is the stress (or strain) intensity factor, and D
and n are material constants. Thus fracture mechanics and which previously were studied only from know-
based equations are now available to cover the total ledge of monotonic or cyclic stress-strain behaviour,
spectrum of crack lengths of importance to engineers, hardness, yield and tensile strength, stress-strain cycle
and that lifetime can be assessed by the integration and counting, strain hardening and ageing behaviour, clas-
coupling of the appropriate crack propagation equa- sical yield theories, residual stresses, etc., should now be
tions. Figure 3 also indicates that it is insufficient to analysed from the viewpoint of how they affect short
describe cracks as ‘short’ or ‘long’ insofar that length crack growth characteristics.
alone is not an adequate description. For example an An example of this new thinking is that previously
LEFM-type crack will be a crack that can be ade- the fatigue limit was considered to be identified only by
quately quantified at cyclic stress levels less than 0.7 of a stress value. From previous arguments listed above,
the cyclic yield stress while a ‘short’ crack, quantified by however, it should be clear that it is also linked to a
EPFM, is one driven by cyclic stress levels above the distance, for example the distance to the strongest
LEFM limiting condition. barrier in a microstructural system.
The important zone of Fig. 4 is at very short crack Space and time d o not permit all the developments
lengths and very slow crack growth rates because this is currently under way to be discussed here, but a few
where most of the fatigue life of a specimen, component illustrative examples will now be presented.
or structure is spent. Any technique to separate the two
curves, represented by equations (2) and (3) is therefore 6 GRAIN SIZE EFFECTS
beneficial to fatigue resistance. Figure 4 also shows why
surface finish is important to engineers. Should surface It has long been known that reducing the grain size of a
scratches be deeper than b, , then the fatigue strength metal increases the plain specimen fatigue limit of a
will reduce dramatically. Conversely, if the surface finish metal (17-19). Conversely, in the last decade research
is less than b,, that is lower than the distance between has shown (2&22) that increasing grain size can
microstructural barriers, then improved fatigue strength increase the LEFM threshold value, thereby increasing
is achieved, although it is not worth while to improve the fatigue limit for structures containing long cracks.
further the surface finish because fatigue cracks will still The latter effect is explained by a mismatch of fatigue
grow quite fast initially and will only slow down to crack faces as they close (a mismatch which increases as
prolong lifetime when the crack reaches the major grain size increases), thereby reducing the cyclic crack
microstructural barrier. opening distance and crack propagation rate. This effect
Further reading to appreciate the short crack growth is reduced and eliminated the higher the applied mean
behaviour of metals, both in terms of micromechanics stress (see Fig. 5), simply because the crack faces still
analyses and an understanding of the microprocesses remain open at the minimum level of the applied cyclic
involved can be obtained via references (13) and (14). In stress.
particular, two models are now being widely developed, Paradoxically the reverse effect in plain specimens
one due to Brown and Hobson (15), which is empirical can also be explained by a crack growth analysis, From
being based on experimental results, and one due to equation (2) it is seen that if the grain size d represents
0 IMechE 1991 Proc lnstn Mech Engrs Vol 205
 K J MILLER

a Grain size
0 153pm
A 40pm
0 19pm
Of interest to historians of metal fatigue research is
the excellent work of Forrest and Tate (17) who studied
the effect of grain size in a 70/30 brass and made many
pertinent remarks regarding the initiation of small
cracks. Indeed, they presented an early model of the
Kitagawa-Takahashi diagram (24) in their paper to
describe physically small fatigue crack growth behav-
iour.

7 CORROSIVE ENVIRONMENTS
Again it is well known that fatigue testing in a corrosive
environment can eliminate the in-air fatigue limit of
plain samples. To understand the cause of this, Fig. 7
shows the crack growth behaviour in air (solid lines) of
a crack at a stress level below the in-air fatigue limit.
From A to B the short crack initially grows quickly and
arrests at the grain boundary barrier d. It cannot
continue to grow at the same stress level which explains
Stress ratio, ~ , , , , / o , , ~ ~ the fatigue limit. Should a similar sample have an initial
Fig. 5 Effect of grain size and stress ratio R on the LEFM crack length slightly longer than a t h , this would grow
stress intensity threshold [after (ZO)] following the characteristic C to D.
Recently (25, 26) tests have been carried out to study
the strongest barrier in the system, then the smaller the short cracks growing in air and also with an aggressive
grains, the smaller the value d, the slower the crack environment introduced for a period so as to cause the
growth rate and the greater the fatigue resistance. crack to elongate from d to ath(B to C on Fig. 7). This
In summary form, what is required for maximum elongation occurs by the conjoint action of chemical
fatigue resistance is to have small grains on the surface and mechanical processes to overcome the strength of
of a specimen or component, and large grains in the the microstructural barrier. Obviously the speed at
interior. Effectively this will separate the two phases of which this is done will be a function of the chemistry of
crack growth rate depicted in Fig. 4, which can only be the environment (fluid composition, pH value, electro-
bridged by increasing the applied stress level, leading to potential) and the applied stress level, the crack length
both an increased fatigue endurance and a higher itself and the material microstructure. In our work, this
fatigue limit. low crack extension rate is associated with the metal
Figure 6 shows schematically the behaviour of short dissolution rate; this equates well with the experimental
cracks in samples of different grain size. Recent experi- data; see also reference (27).
ments (23)on an aluminium alloy having different grain The broken lines in Fig. 7 illustrate that the corrosive
sizes of 30,120 and 250 pm confirm this behaviour. environment may (but does not always) cause both
short and long cracks to propagate faster, however, the

t major consideration has to be the bridging of the ‘in-air

fatigue’ gap between a non-propagating microstruc-

h Cyclic stress range = constant turally short fatigue crack and a propagating physically
small crack.
More recent studies have involved the periodic intro-
Large grain size duction and removal of a corrosive environment at
\
B several crack lengths which correspond to various
Intermediate grain size
E
0
Small grain
I size Crack growth behavioui
In air
_ _ _ _In an aggressive
U
environment
\
”
Crack length
Fig. 6 Microstructural short fatigue crack growth behaviour , /

for a fine-grained aluminium alloy at a low cyclic

stress level (solid line). Increasing the grain size to
intermediate and large increasingly reduces fatigue
endurance for the same cyclic stress level. Note that
even for the smallest grain size, the barriers to growth
Crack length
are extremely weak in the easily deformed aluminium
alloy and the crack passes easily from grain to grain, Fig. 7 Microstructurally short and physically small fatigue
albeit at an exceedingly low average speed shown as crack growth rates at a cyclic stress level below the
the broken part of the curve ‘in-air’ fatigue limit
Part C: Journal of Mechanical Engineering Science @ IMechE 1991
 METAL FATIGUE-PAST, CURRENT AND FUTURE 291

barrier distances in order to just overcome each suc-

cessive barrier. These tests are a little difficult to
perform because of the statistical variation of grain size,
t D , = small notch depth (laboratory sample)
the variable strength of individual barriers and, last but D2 = large notch depth (large structure)
not least, the various mechanisms of crack development a
including pitting, crack coalescence and mixed mode
growth. Nevertheless, the fundamental issue here is that,
for the fatigue limit to be removed, the aggressive
environment needs only to be introduced for a short
period of time in order to overcome the in-air fatigue
limiting barrier.

L I c
8 NOTCHES 1 .o
Stress concentration factor, K,
Two quite distinct approaches prevail in the literature Fig. 8 Relationship of the fatigue limit to notch depth,
regarding notch fatigue analysis. That this should be so threshold stress intensity factor and stress concentra-
is indicative of the complexity of the problem requir- tion factor. Note that the fatigue limit is independent
ing stress analysts to study notch stress-strain fields, of the stress concentration factor at K, = Z and is
fracture analysts to study elastic-plastic mechanics, and dependent only on AK,b and the notch depth
metallurgists to study material considerations (for
example notch sensitive materials). Seldom are all disci- physics of the notch problem was finally made when a
plines simultaneously involved in deriving solutions. study of the propagation behaviour of a fatigue crack in
Studies of the effect of notches are important, since and beyond the notch root stress-strain field was
fatigue cracks invariably start from some form of stress published (31). In Fig. 9, the master curve (thick black
concentration. Such localized features may be of metal- line) shows the material property characteristic, that is
lurgical significance, such as precipitates and inclusions, the fatigue limit defined as da/dN = 0 for a material
or a manufactured geometrical discontinuity, such as a having cracks of various lengths. The other curves
drilled hole, keyway, etc. show, for various cyclic stress levels, the stress intensity
The classical approach via an understanding of the characteristic. Should the two curves representing the
net-section cyclic stress levels, stress concentration material behaviour and the stress intensity behaviour
factors and notch strength reduction factors, has now intersect, then a non-propagating crack will result. Two
been extended to include Neuber stress-strain analyses. features are worth noting: first that the convergence
These have culminated into the local strain approach and divergence of the two curves represent deceleration
whereby the behaviour of a plain specimen can be and acceleration of a notch root crack respectively and,
related to the notch root material behaviour. Simple secondly, that the length of a non-propagating crack
elastic-plastic finite element analyses have supported increases as the cyclic stress level increases (32);see also
these developments. In several cases such analyses may reference (3).
be adequate but ultimately they are unsatisfactory since To conclude this section on notches, a brief analysis
they cannot cover all forms and sizes of notch profiles, will now be made of a very common industrial problem,
applied mean stress levels and complex loading condi- namely the failure of threaded joints. Bolts and set-
tions. This is because a fundamental assumption is screws invariably suffer high mean stress levels, indeed a
made that failure in both plain and notched specimens permanent clamping action depends on the magnitude
is related to crack initiation, itself an imprecisely defined
instant. Additionally, such approaches do not include
the effect the crack itself has on the stress-strain field
Aol > Aa2 = AaFL> Aa3
and therefore cannot explain the fatigue limit due to the ,A 01
non-propagation of a notch crack. /
,- A02
The fracture mechanics approach to understanding
notches began with Frost and Phillips, see reference (3), ___-
,/

who showed the possible irrelevance of stress concen-

tration factors. The work of Smith and Miller (28, 29)
indicated two limiting conditions for notches, and in
particular the separation between failure and non-
failure of a notched component on the basis of whether
or not a crack would continue propagation; one fatigue
limit, AoFL, boundary condition being given by the
simple formula :
Non-propagating crack length at Aa3
OSAK,,,
AaFL= ___ (5)
JD log (crack length)
where D is the depth of the notch (see Fig. 8). Fig. 9 The variation of the crack tip stress intensity factor
It was this study that led to the realization that with crack length for (a) a given material (thick solid
cracks, in general, initiated very early in life and propa- line), and (b) various cyclic stress levels. Note that the
gated at a decreasing rate (30). A total analysis of the fatigue limit AoFL = Aa2
0 IMechE 1991 Proc lnstn Mech Engrs Vol 205
298 K J MILLER

of the mean stress. Designers invariably choose high- 3. The orientation of the surface plane of the engineer-
strength materials for such fasteners in order to in- ing component, structure or specimen, with respect
crease clamping forces despite the fact that increasing to the crack growth direction, is also predetermined
the R ratio (B,,,~,,/B,,,~~)while keeping the stress range in uniaxial fatigue.
(cT,,,,, - omin)constant, decreases the threshold stress 4. The effect of material texture (for example
intensity factor range AK,,, . anisotropy), itself a three-dimensional feature, is fixed
Recourse to the classical modified Goodman diagram and cannot be quantified by uniaxial testing alone.
with further modifications to account for the appropri- 5. The shape of the growing crack is predetermined in
ate stress concentration factor is not particularly helpful uniaxial fatigue.
since :
To appreciate these important features, and their
1. The contours of lifetime are very closely spaced at relationship with the multiaxial stress-strain state,
high mean stresses. attention is directed to Figs 10 and 11.
2. Microscopic and macroscopic yielding conditions Shear strain
give rise to sharp discontinuities in the lifetime con- L
tours.
3. No assistance is provided to understanding the c - P
t
~ I
sequence and duration of failure mechanisms at the
thread notch root.
Recent work (33) has shown that the mechanical
fatigue behaviour of a notch subjected to a high mean
stress can be calculated from a fracture mechanics
analysis. Such an analysis indicates that the fatigue limit
in a high-strength steel is associated with a non-
propagating fatigue crack in the notch root which is
determined by the microstructural texture and can be as
small as 12 pm. The next problem to solve is how
quickly such a small crack can develop in a corrosive
environment via interactions of stress corrosion, corro-
sion fatigue and mechanical fatigue processes.
It follows that the Goodman diagram now requires
detailed interpretation and modification via an analysis
of short fatigue crack behaviour. This is particularly
true for fasteners which are periodically subjected to Fig. 10 Mohr’s circles of strain
corrosive conditions.
a 0
9 MULTIAXIAL FATIGUE
The first major breakthrough in the study of complex
loading came in 1973-75 (34,35).This work showed the
irrelevance of classical theories of yield deformation
(Von Mises, Tresca, et al.) when applied to fatigue frac-
ture. This had been hinted at in the classical and careful
work of Crossland (36)and his colleagues when testing
pressurized tubes. Had knowledge of fracture mechanics
been available at that time, the two recent books on
multiaxial fatigue (37,38) would have been published Laboratory test Critical location in a
much earlier. structure or component
Most research on fatigue is performed by subjecting
simple specimens to uniaxial loading and yet most
engineering components are subjected to complex
loading involving interactions between bending, torsion
and axial forces. In uniaxial fatigue five important
factors can be, and frequently are, ignored in a physical
appreciation of the fatigue process, namely :
1. The strains in the two minor orthogonal directions
are fixed and uncontrolled, and hence may be
deemed, incorrectly, to have no importance, that is in
uniaxial fatigue the principal strains t2 and t3 are the
same and negative being equal to t 1 / 2 if constancy of
volume is assumed. (b) Identical stress-strain fields but different crack growth
2. The orientation of the planes of crack growth in rela- directions with respect to the surface (shown shaded)
tion to the applied uniaxial strain is predetermined Fig. 11 The importance of the orientation of the applied
and follows the stage I and then the stage I1 direc- stress-strain field with respect to (a) the defect plane
tions first distinguished by Forsyth (39). and (b) the surface plane
Part C : Journal of Mechanical Engineering Science @ IMechE 1991
 METAL FATIGUE-PAST, CURRENT AND FUTURE 299

In Fig. 10, the maximum principal strains in a cycle are loaded via the individually controlled loading
are depicted in the form of Mohr's circles of strain. In modes of clockwise/anti-clockwise torsion, axial tension/
fatigue fracture it is clear that two (not one) strain compression and internal/external pressurization of
parameters are important; firstly, namely the size of the the tubes.
largest circle created in a cycle, and which is equal to To conclude this section of the paper, readers are
el - e3 (that is the maximum engineering shear strain, again reminded that our current understanding was
y,,J and, secondly, the location of this circle in strain achieved by examining the progression of cracks and
space, that is the displacement of the largest circle from relating this experimentally determined feature to the
the origin which is equal to (el + eJ2. Consequently it applied stress-strain field. Such studies have now per-
is possible to d o fatigue tests in which one or the other mitted us to appreciate not only the sometimes unusual
of these parameters, or both, are varied during the orientation of cracks, but also the development of their
course of a fatigue test. shape (40); an important characteristic, especially in
It follows that it is now possible to d o fatigue tests relation to the leak-before-break design philosophy for
(40) in which : pressure vessels and piping. Indeed, following the classi-
fication of Forsyth, we are now able to sustain shear
1. The magnitudes of the principal stresses in a cycle (stage I) crack growth where previously it was not
are constant (that is fatigue tests in which the stress thought possible and, similarly, to get stage I1 cracks to
amplitudes are zero). change to stage I, that is a reversed event to that wit-
2. The cycle stress-strain characteristic can be a circu- nessed in uniaxial push-pull fatigue.
lar hysteresis loop.
3. The effect of anisotropy can be quantified. 10 CUMULATIVE DAMAGE AND MIXED
4. The shape of a fatigue crack can be changed. MODE LOADING
5. The relationship between cyclic stress strain behav-
iour (the classical approach) and fatigue fracture As an amusing finale, Fig. 12 shows that it is possible to
mechanics (the modern approach) can be quantified double the fatigue life of a material subjected to a stress
for all forms of loading modes via, for example, the state that would lead to failure in 400000 cycles by pre-
Tomkins model (41). viously subjecting the material to another stress state
6. The complex loading history of an engineering com- that would also, if applied separately, cause failure in
ponent can be simulated in a laboratory. 400 OOO cycles.
In terms of the Palmgren-Miner cumulative damage
To illustrate both the necessity to link cyclic stress- law,
strain behaviour to crack mechanics and to show that n
the classical approach to rationalize different loading C-=1
modes via the Tresca or Von Mises hypotheses is not Nf
sufficient, attention is drawn to Fig. 11. In the two-
dimensional configuration of Fig. 1 1 a, the stress system I .8
Experiments
is identical in both the critical zone of the component
and the laboratory specimen simulation. Thus, the
equivalent stress or strain, whichever way it is
calculated, is identical. The laboratory specimen,
however, will not fail but the component will fail if a
critical but small and unobserved central crack exists,
having an orientation that is horizontal in the
laboratory specimen but vertical in the component. A
similar situation exists in Fig. l l b where the
three-dimensional stress-strain response is identical in C
both cases, as is the orientation of the crack growth .-0
e
Ln

planes, but, because the surface is orientated in the 0

I

position shown, one cracking system is far more

dangerous than the other, simply because it is
propagating away from the surface while the other is
propagating along the surface.
If these three-dimensional systems are thought to be
purely academic it is worth analysing the stress states
introduced into, for example, automobile axles and
driveshafts which can be subjected to two planes of
bending plus torsion, or to the case of a railway line in
which the magnitude and direction of the principal
stresses due to bending and shear across the rail are
constantly varying as the train moves along the line.
T o achieve simulated multiaxial conditions in a
laboratory however, is not a simple matter and most
tests have to be computer controlled (42). In the fatigue (dh',)push-pull
testing facility at SIRIUS (Structural Integrity Research Fig. 12 Results of fatigue tests in which the loading mode
Institute, University of Sheffield), cylindrical specimens was changed during the test
@ IMechE 1991 Proc Instn Mech Engrs Vol 205
300 K J MILLER

where n is the number of applied cycles and N , is the the effect to be seen in the short crack growth rate
number of cycles to failure. However Fig. 12 shows that characteristic; this is observed in Fig. 14. For the pure
torsional loading mode, Fig. 14a indicates a faster
n development of the fatal crack than when prior push-
c-=2 (7) pull loading is applied; see Fig. 14b for the slower
Nf
development of the fatal torsion crack.
for the particular condition indicated at X. In Fig. 14a the Navarro-de 10s Rios model can be
The answer to this conundrum cannot be solved by invoked to account for the periodic acceleration/
either an analysis of the assumptions of the Palmgren- deceleration of the fatal crack as it passes through bar-
Miner hypothesis or an interpretation of the classical riers of decreasing resistance the longer it becomes. The
stress range versus number of cycles to failure (S-N) Brown-Hobson model, however, would have to be used
curve, despite the fact that both mathematical formula- for the complicated situation of barriers being formed
tions are firmly based on the integration of a crack by a mixture of the previously formed short surface
growth law. The answer instead lies in the orientation cracks and microstructural features.
and shape aspects of cracks which differ from one stress The reverse situation occurs when torsion is applied
state to another. One other variable is the deformation prior to push-pull and a serious reduction in lifetime
orientations of a material (that is slip planes and slip results if judged from the Palmgren-Miner expression
directions) which change when the stress state is altered. (see the broken curve of Fig. 12). This is because of two
An examination of Fig. 13 shows the two stress states factors: first, the prior torsion produces a stage I
now being considered. The application of push-pull cracking orientation that is compatible to the
loading prior to torsional loading can lead to an orien- subsequent stage I1 cracking direction for the push-pull
tation and shape of push-pull stage I cracks that are loading; second, in push-pull the long duration stage I
incompatible to the direction and shape of stage I phase and stage I to stage I1 transition are overcome
cracks required to be generated in the subsequent since a push-pull stage I1 crack can develop
torsion loading. The former cracks therefore act as bar- immediately. In this latter case the slip plane
riers to the generation of the later cracks. Both forms of orientations and slip directions are not as obstructive as
these stage I cracks develop in the ferrite phase of the in the former case. A more detailed explanation of the
material being considered here, and one would expect above phenomena is given in reference (6).

Maximum shear-stress-shear-strain planes Stage I planes and cracks Transfer to stage I1 cracks

0
urface

AT ’ ~ A I J
0

Push-pull loading mode

Surface f Ar

Reversed torsion loading mode

Fig. 13 The effects of loading mode on the development of stage I and stage
I1 cracks. Here, A and B refer to case A and B types of cracks
having different orientations with respect to the surface (35), AT is
the reversed shear stress and Y, Z, M and N are stage I cracks while
S and T are stage I1 cracks
Part C: Journal of Mechanical Engineering Science @ IMechE 1991
 METAL FATIGUE-PAST, CURRENT AND FUTURE 301

I 1 FUTURE DIRECTIONS
10 -
Now that we can understand the fatigue failure process
from the beginning to the end of life in terms of crack
growth using microstructural fracture mechanics,
elastic-plastic fracture mechanics and linear elastic frac-
ture mechanics, there are unlimited research and devel-
opment opportunitifs.
Future research may be classified in three distinct
areas :
10-
1. A reassessment of all fatigue failure phenomena
-
a, reported in the literature which previously used only
x
U
the stress-strain behaviour (that is the deformation
2x response) of the material as the basis of interpreta-
tion; the reassessment being based, for example, on
the behaviour of short fatigue cracks. In this cate-
gory will come the effects of surface treatments on
10 -
short crack growth and other production processes
which have long been known to affect lifetime.
2. A more complete understanding of the short crack
growth phase both in terms of complex stress loading
and the effect of microstructural variables. In this
category will come the development of models,
together with the introduction of more fatigue-
resistant alloys and an understanding of crack
10 - I
growth behaviour under random loading conditions.
10' 102 103 1
3. A study of interactive failure processes that are
known to occur under service conditions. For
Crack length
example the interaction of corrosion fatigue, stress
Pm corrosion cracking and mechanical fatigue, also the
(a) Crack growth in torsion [theoretical solid line after (16)l interaction of short cracks generated by fretting and
subsequently propagated by mechanical fatigue.
10- At the Structural Integrity Research Institute at the
University of Shefield more than 30 programmes of
/ research are devoted to crack behaviour. Most of these
d are sponsored by British industry, but underpinned by
/ grants from the Science and Engineering Research
/ Council to enable us to continue pursuing our academic
work. This ranges from the development of our acoustic
/ microscope, which should be able to see short cracks
10 -
/ that other microscopes cannot detect (see Fig. 15),
a,
I
/ O through to the development of design procedures for
E complex multiaxial fatigue loading.
-c
5
-x
a,
U
o/
I In particular a special purpose two- and three-dimen-
zw sional elastic-plastic finite element package (TOMECH)
$ I
Y
U
I has been developed and constantly updated since 1980
E
U under the direction of Bilby, Cardew, Goldthorpe and
I Howard, with the prime purpose of solving complex
defect problems, many of which are related to the afore-
10-
mentioned numerous fatigue crack growth studies.
As an example of the power of TOMECH, Fig. 16
shows predictions of growth of two regions of active
plastic deformation associated with the tip of a crack
undergoing extension in an austenitic steel used in the
nuclear engineering industry; a plastic zone ahead of
the crack tip and a compressed zone developing behind
the crack tip can be seen. Parts (a), (b) and (c) of Fig. 16
show the development of the plastic zones as the crack
grows in a large structure, while parts (d), (e) and (f)
Crack length
relate to a standard compact tension specimen. The
Pm similar results obtained provide confidence in the con-
(b) Crack growth data in torsion after prior push-pull loading servatism of using data derived from compact tension
(broken curve, assumed) specimens, in failure assessments for large-scale struc-
Fig. 14 Experimental short crack growth data tures containing long cracks (43).
0 IMechE 1991 Proc Instn Mech Engrs Vol 205
 (a) A fatigue crack is seen propagating along a twin (b) A fatigue crack changing direction after
boundary towards the grain boundary crossing a grain boundary
Fig. 15 Acoustic micrographs of Waspaloy showing grain boundaries
(slightly etched by electro-polishing prior to fatigue testing) and
persistent slip bands (PSB). Note that the acoustic image allows one
to distinguish between persistent slip bands and closed cracks; the
latter give a broader acoustic signal because of the reflective nature
of the crack faces

.+

... .*
1 1 L A & *

*
.
I
.
f .
I
f. *
1 1
.
.

*
. *
. . I
* * I .

I . . .
*f ..
. . . .
1.

..... ...... ~
. .., .,.. .*..*. ..*
,
.

,
/

.
, r ? . . . - , . T . r
/
**
. _
,

.
.

. I
.

.I
T

.
I

I ,
(e) (f)

Fig. 16 The development of two plastic zones at a n extending crack tip in (a)
to (c) a large structure, and (d) to (f) a compact tension specimen, as
detected by the TOMECH finite element program; the arrows indi-
cate the crack tip positions [after (43)]

Each of these large two-dimensional analyses invol- have an interdisciplinary approach involving materials
ving some 8000 degrees of freedom, took several hours scientists, chemists, physicists and mathematicians, all
of processing time on the Science and Engineering of whom are needed to supplement the work of engi-
Research Council's IBM 3090-600E/6VF supercom- neers attempting to solve practical engineering prob-
puter under the IBM/RAL Joint Study Agreement. lems via stress analysis, fracture mechanics analyses,
elastic-plastic (two- and three-dimensional) finite
12 CONCLUSIONS element analyses, etc; all leading to a re-evaluation of
New areas in metal fatigue research have been old design codes and the establishment of new codes of
addressed, most of which have illustrated the need to practice.
Part C: Journal of Mechanical Engineering Science 0 IMechE 1991
 METAL FATIGUE-PAST, CURRENT AND FUTURE 303

The fatigue failure process has been related to the 10 Miller, K. J. and de los Rios, E. R. (Eds) The behaviour of short
growth of cracks, however small, and the stress-strain fatigue cracks, EGF publication 1, 1986, 560 pages (Mechanical
Engineering Publications, London).
field required to drive such cracks, particularly across 11 Hobson, P. D., Brown, M. W. and de 10s Rios, E. R. Two phases of
material barriers such as grain boundaries and strong short crack growth in a medium carbon steel. In (lo),pp 4 4 4 5 9 .
second-phase zones. In this respect the crack initiation 12 Miller, K. J. Initiation and growth rates of short fatigue cracks.
phase is seen not as a period during which a crack can Fundamentals of deformation and fracture, I UTAM Eshelby
nucleate, but rather a period in which a crack may pro- Memorial Symposium, (Eds B. A. Bilby, K. J. Miller and J. R.
Willis), 1985, pp 477-500 (Cambridge University Press).
pagate at a decreasing rate before developing into a 13 Miller, K. J. and de 10s Rios, E. R. (Eds) Shori ,fatigue cracks, an
dominant crack. EGF/ESIS publication. To be published, 1991 (Mechanical Engin-
Three zones of fatigue crack behaviour are identified, eering Publications, London).
each phase requiring its own branch of fracture mecha- 14 Miller, K. J. The behaviour of short fatigue cracks and their initi-
ation. Part 11-a general summary. Fatigue Fract. Engng Mater.
nics to quantify that behaviour, namely : Struct., 1987, 10(2),93-1 13.
15 Hobson, P. D. The growth of short fatigue cracks in a medium
The microstructurally short crack growth phase carbon steel. PhD thesis, University of Shefield, 1985.
The physically small crack growth phase 16 Navarro, A. and de 10s Rios, E. R. A model for short fatigue crack
The long crack growth phase propagation with an interpretation of the short-long crack tran-
sition. Fatigue Fract. Engng Mater. Struct., 1987, 10(2), 169-186.
Directions of future research have been outlined partic- 17 Forrest, P. G. and Tate, A. E. L. The influence of grain size on the
ularly with regard to long-standing engineering prob- fatigue behaviour of 70/30 brass. J. lnst. Metals, 1964-65, 93,438-
lems that can now be analysed by application of the 444.
appropriate fracture mechanics. These problems include 18 Yokobori, T. The influence of metallurgical factors on fatigue limit
and the statistical nature of fatigue fracture of plain carbon steel.
topics such as fretting, surface treatments, random Proceedings Second Japan Congress on Testing Materials, 1959, pp
loading, interactive failure processes and defect inter- 1C-13 (The Japan Society Testing Materials, Kyoto, Japan).
actions. 19 Yoshikawa, A. and Sugeno, T. Factors responsible for the sharp
fatigue limit in iron and steel. Trans. Metall. SOC., 1965, 233,
13141317 (American Institute of Mining and Metallurgical
ACKNOWLEDGEMENTS Engineers).
20 Masounave, J. and Bailon, J.-P. Effect of grain size on the
Discussions stretching over 25 years, with hundreds of threshold stress intensity factor in fatigue of a ferritic steel. Scripta
colleagues worldwide, have helped me to understand a Metallurgica, 1976, 10, 165-170.
little of the metal fatigue problem. This human inter- 21 Carlson, M. F. and Ritchie, R. 0. On the effect of prior austenite
action is acknowledged with gratitude. Just as impor- grain size on near threshold fatigue crack growth. Scripta Metal-
lurgica, 1977, 11, 1 1 13-1 1 18.
tant is the friendship and laughter that has always 22 Priddle, E. K. The influence of grain size on threshold stress inten-,
accompanied these discussions, whatever their intensity. sity for fatigue crack growth in AlSI 316 stainless steel. Scripta
Additionally, thanks are due to past and present Metallurgica, 1978, 12,49-56.
research workers whose technical writings have provid- 23 Miller, K. J., Slade, J. and Wu, X. J. The effect of grain size on
short fatigue crack growth rates. To be published, see (13).
ed stimulation and encouragement. 24 Kitagawa, H. and Takahashi, S. Applicability of fracture
Finally, grateful thanks are presented to various mechanics to very small cracks or the cracks in the early stage.
national and international funding agencies, but in par- Proceedings Second International Conference on Mechanical
ticular to the Science and Engineering Research Council behaviour of materials, (ICM2), 1976, pp 627-631 (American
of the United Kingdom, which has encouraged the Society of Metals).
25 Akid, R. The initiation and growth of short fatigue cracks in an
pursuit of knowledge pertaining to the avoidance of aqueous saline environment. PhD thesis, University of Shefield,
fracture of engineering materials ; components and 1987.
structures. 26 Akid, R. and Miller, K. J. The initiation and growth of short
To all of these, I dedicate this paper. fatigue cracks in an aqueous saline environment. In Environment
assisted fatigue, EGF publication 7 (Eds P. Scott and R. A. Cottis),
1990, pp 41 5 4 3 4 (Mechanical Engineering Publications, London).
REFERENCES 27 Akid, R. and Miller, K. J. The effect of solution pH on the initi-
ation and growth of short fatigue cracks. In Fracture behaviour
1 Proceedings of the International Conference on Fatigue of Metals, and design of materials and structures, Proceedings of ECF8,
IMechE/ASME, 1956, 961 pages (Institution of Mechanical Engi- Torino, 1990, pp 1753-1758 (EMAS).
neers, London). 28 Smith, R. A. and Miller, K. J. Fatigue cracks at notches. Int. J.
2 Journal of Strain Analysis, Special Issue, October 1975, 10(4), 193- Mech. Sci., 1977, 19, 11-22.
266. 29 Smith, R. A. and Miller, K. J. Prediction of fatigue regimes in
3 Frost, N. E. and Phillips, C. E. Studies in the formation and pro- notched components. lnt. J . Mech. Sci., 1978,20, 201-206.
pagation of cracks in fatigue specimens. In (l), pp 52G526. 30 Hammouda, M. M. and Miller, K. J. Elastic-plastic fracture
4 Thompson, N. Experiments related to the basic mechanism of mechanics analysis of notches. ASTM STP 668, 1979, pp 703-719
fatigue. In (l), pp 527-530. (American Society for Testing and Materials).
5 Jacquet, P. A. Observations on the microstructure of a brass con- 31 Yates, J. R. and Brown, M. W. Prediction of the length of non-
taining 67 per cent of copper, subjected to alternating bending propagating fatigue cracks. Fatigue Fract. Engng Mater. Struct.,
stresses. In (I), pp 506509. 1987,10(3), 187-201.
6 Zhang Wei, Short fatigue crack behaviour under different loading 32 Miller, K. J. Fundamentals of fatigue: fatigue at notches. In
systems. PhD thesis, University of Shefield, 1991. Advances in fatigue, science and technology (Eds C. Moura Branco
7 de los Rios, E. R., Mohamed, H. J. and Miller, K. J. A micro- and L. Guerra Rosa), NATO AS1 Series, 1989, pp 157-176
mechanics analysis for short fatigue crack growth. Fatigue Fract. (Kluwer Academic Publishers).
Engng Mater. Struct., 1985,8, 49-63. 33 Heyes, P. J. Fatigue cracks at notch roots under high mean stress.
8 Allen, N. P. and Forrest, P. G. The influence of temperature on the PhD thesis, University of Shefield, (to be published 1991).
fatigue of metals. In (I), pp 327-340. 34 Brown, M. W. High temperature multiaxial fatigue. PhD thesis,
9 Grabowski, L. and King, J. E. Short/long crack transition behav- University of Cambridge, 1975.
iour in nickel base superalloys. To be published as part of an ESIS 35 Brown, M. W. and Miller, K. J. A theory for fatigue failure under
publication (Eds K. J. Miller and E. R. de 10s Rios), 1991 multiaxial stress-strain conditions. Proc. lnstn Mech. Engrs, 1973,
(Mechanical Engineering Publications, London). 187(65/73)745-755.

Q IMechE 1991 Proc lnstn Mech Engrs Vol 205

304 K J MILLER

36 Crosslend, B. Effect of large hydrostatic pressures on the torsional stress corrosion and creep (Ed. L. H. Larsson), 1984, pp 215-238
fatigue strength of an alloy steel. In (I), pp 138-149. (Elsevier, London).
37 Miller, K. J. and Brown, M. W. (Eds) Multiaxial fatigue, ASTM 41 Tomkins, B. Fatigue crack propagation-an analysis. Phil. Mag.,
STP 853, 1985, 744 pages (American Society for Testing and 1968,18,pp 1041-1066.
Materials). 42 Fernando, U. S., Miller, K. J. and Brown, M. W. Computer aided
38 Brown, M. W. and Miller, K. J. (Eds) Biaxial and multiaxial multiaxial fatigue testing. Fatigue Fract. Engng Mater. Struct.,
fatigue, EGF publication 3, 1989, 686 pages (Mechanical 1990,l3(4), 387-398.
Engineering Publications, London). 43 Goldthorpe, M. R. Elastic-plastic finite element analysis of speci-
39 Forsyth, P. J. E. A two-stage process of fatigue crack growth. men size effects in the characterization of stable crack growth in
Symposium on Crack propagation, Cranfield, 1971,pp 76-94. austenitic weld steel. Shefield University, Department of Mechani-
40 Miller, K. J. and Brown, M. W. Multiaxial fatigue; an cal and Process Engineering, report 125,1991.
introductory review. In Sub-critical crack growth due to fatigue,

Part C : Journal of Mechanical Engineering Science 0 IMechE 1991