Vous êtes sur la page 1sur 18

Department of Mechanical, Materials & Manufacturing Engineering

Material Models and Modes of Failure MM4MMM


Convenor: Dr W Sun (Coates Building B68, w.sun@nottingham.ac.uk)

FATIGUE
Fatigue is the failure which occurs in metals subjected to repeated or otherwise
varying load which never reaches a level sufficient to cause failure in a single
application (L P Pook, The role of crack growth in metal fatigue, J W Arrowsmith Ltd 1983)
The first systematic investigations of fatigue were carried out by Whler in the 1850s
and 60s. The investigations were associated with the failure of railway axles.
Investigations of other failures (e.g. the Comet jet propelled passenger aeroplane etc.)
have resulted in improvements in the understanding of the causes of fatigue failures.
Major contributors to the understanding of fatigue failure since Whler have included
Goodman, Griffith, Neuber and Irwin. In particular, recent contributions to the
literature on fracture mechanics have had a profound influence on the development of
methods of predicting fatigue failures.
Basic phenomena
The failure mechanism for an initially uncracked component with a smooth (polished)
surface can be split into four parts, i.e.
(i)
Initiation of a crack;
(ii)
Stage I crack growth (on planes of high shear stress);
(iii) Stage II crack growth (on planes of maximum principal stress); and
(iv) Final fracture, which occurs when the crack reaches a critical length
corresponding to ductile tearing (plastic collapse) at one extreme, or cleavage
(brittle fracture) at the other extreme.
As we have seen, most metals are made up of a large number of grains (crystals).
Within each grain the atoms are arranged in an ordered manner and the mechanical
properties depend on the orientation of the grain relative to the applied stress system.
Slip can occur in metals by the movement of dislocations along crystallographic planes
due to the shear stress on the plane. In metals, some surface grains will be
orientated so that their slip planes are in the direction of the maximum applied shear
stress. These are the grains in which cracks are likely to initiate.

20

Originally smooth
surface with slip
lines

Sliding due to
shears

Stress
concentration

Stage I and Stage II


growths leading to
final fracture

Stage I crack growth (or micro-crack growth) is on planes of maximum shear stress.
This growth is quite small, usually only a few grains. After this, fatigue cracks tend to
join up and grow along planes on which the maximum tensile stress exists; this is
stage II crack growth (or macro-crack growth).
Finally, the crack will become so large that on applying the maximum load (during the
cyclic load) either fast fracture or plastic collapse or a combination of them will occur.
21

The stage II fatigue surface appears to be flat at low magnification, whereas the final
fracture surface (i.e. the surface created during the last loading application), often has
a rougher appearance, indicating that a lot of plastic deformation has occurred.
A threshold for the speed of fatigue crack propagation exists because it is not
physically possible for a crack to grow by less than about one lattice spacing per cycle.
However, because crack growth isnt necessarily continuous along the whole crack
front, somewhat lower average rates can be observed.
A typical example of macro-crack behaviour (note other mechanisms are also
possible).

The fracture surface can thus contain striations. Not all metals show striations, and if
they do they may not cover the whole area. The spacing between each striation is the
growth per cycle.
22

Since engineering alloys have inclusions present, these can act as sites for the growth
of voids in the very high strain regions near crack tips. Fatigue crack growth can be
increased because these voids can coalesce and join up with the crack front during
cyclic loading.
If the load range is changed during a fatigue test, this causes the surface appearance
to change. These changes in surface appearance are called beachmarks (note: these
are not the same as the striations which are seen at higher magnification).
Fatigue design philosophies
(i)
Safe Life: The structure will not fail within the design life (may be indefinite,
eg. valve springs etc). When the design life is reached, the component is scrapped
(life may be extended after this by periodic inspection for cracks).
(ii)
Fail Safe: It is recognised that fatigue cracks may occur. The structure is
therefore designed so that cracks will not lead to failure before they are detected (and
repaired). This requires a specified inspection period at the design stage.
(iii) Damage Tolerant: Assumes crack exists and uses fracture mechanics to predict
the growth to see if they can possibly cause failure between inspection periods; needs
slow fatigue crack growth rates and high fracture toughness.
Effect of stress regimes

Zero mean stress and zero mean strain testing


In practice there are two different types of fatigue testing:
- zero mean stress (under load or stress control)
- zero mean strain (under deformation or strain control)
23

Low stress or strain range


Stress control

y
m = 0

High stress or strain range


Strain control (usually servo-hydraulic loading frame)

5
1

m = 0
5

3
2

As stress increases, they may exceed yield point and get some reversible plastic
deformation each cycle (hysteresis).

24

Many materials under elastic-plastic cycling show initial cyclic hardening or softening.

Eventually a stable stress-strain loop results. If a set of stabilized hysteresis loops are
plotted (obtained by loading material at different strain levels) and the tips of the
hysteresis loops are connected, then the locus of the extremities of the stable stressstrain loops for different applied stress (or strain) ranges forms the cyclic stress-strain
curve (back-bone curve) for a material.

25

The shape of the stable stress-strain hysteresis loop is practically the same as double
the back-bone cyclic stress-strain curve useful for calculations.
S-N and -N curves
Usually plot linear Sa vs log Nf , but sometimes use log-log.
S-N curve
S

~ 107

Log Nf

N curve
The transition between mainly plastic and mainly elastic regions occurs at about 10 4
cycles. The two regions can be treated differently for design purposes:
Log a

LSF

HCF
Log Nf
4

~ 10

High cycle, low-stress fatigue


primarily elastic deformation;
use stress range to quantify and S-N data
Basquins equation
where

Sa = B ( Nf ) b (high-cycle)

b = fatigue strength exponent 0.1 typically


26

Low cycle, high stress fatigue


Stresses are generally high enough to give appreciable plastic deformation prior to
failure; use strain range to characterise
Manson-Coffin
(or Coffin-Manson)
where

ap = C ( Nf ) c (low-cycle)

c = fatigue ductility exponent 0.6 typically


C = true fracture ductility in monotonic tension

Log a

Plastic dominant
ea ap
e
p

a
a
Combined a
2
2

Elastic dominant

y
E
Log Nf
4

~ 10

Variable loading
So far, we have considered situations where the component experiences repetitions of
a constant loading cycle. In practice, many applications see a variety of loading
conditions: for example, aircraft components may experience relatively severe loading
conditions during take-off and/or landing as well as a less severe fatigue loading
throughout each flight.
The Palmgren-Miner rule is used to calculate total damage accumulation, assuming
linear damage contribution from each fatigue element:

N1
N
N
2 ... ... i 1
Nf1 Nf 2
N fi
i

where Ni is the number of cycles with a given mean stress m and alternating stress
ai , and Nfi would be the number of cycles to cause failure at mi , ai .
If the summation is < 1, Miners rule implies that failure does not occur.
27

Effect of Stress Concentrations


Predominantly Elastic Behaviour

i.e. Kt nom < y

If the material behaves elastically at the stress concentration, then initiation will occur
in the same number of cycles as a test carried out on a plane specimen at
( K t nom ) . However, when the crack begins to grow it moves into a stress field
which approaches

nom .

Therefore the crack growth rate drops by comparison with a

. In some cases, crack arrest may occur after


the initiation and a small amount of growth, if nom is below the threshold level.
plane specimen test carried out at

If experimental tests (with zero mean load) are carried out on the actual components,
then the fatigue strength reduction factor Kf can be determined.

28

Goodman diagrams are often used for design.

If a safety factor, F, is also introduced, this is applied to both the alternating and mean
stresses.

When small plastic zones occur, a suitable experimentally determined value of Kf may
be used to estimate the fatigue behaviour. Also, if Kf has not been determined then Kt
can be used; this will usually (but not always) result in conservative design.

29

Elastic-plastic behaviour at stress concentrations


The problem can be solved by plasticity theory for some components or finite element
analysis. However, an approximate solution is often good enough.

y K t nom 2 y

(1st load application)

For plane strain (or constrained) situations the linear theory gives a reasonable
estimate of strain (at A) in the stress concentration region.
For plane stress,
Neubers theory is better (ie. at A)
Removal of load If Kt nom < 2y, then unloading will be elastic. Subsequent loading
and unloading will occur elastically with m mean stress) and a (alternating stress).

30

For Kt nom > 2y, first load application is the same as for Kt nom < 2y. However, on
unloading, reverse yielding occurs (to position B).

Note: If the material cyclically hardens or softens, the second load application will not
finish up at A (or A). However, after several cycles a steady - loop (with constant
P) will be achieved. This can be estimated by using the cyclic-stress strain curve
rather than the monotonic stress-strain curve (as shown below):

31

The plastic strain ranges or the stress ranges can now be used, via
curves, to estimate the fatigue life.

a N or Sa N

Multiaxial fatigue
The most common multiaxial stress criterion used is the maximum shear stress
criterion.
32

The uniaxial S-N curve is used, with the effective S given by

S = 1 2 = 2
where 1 2 3 are the principal stresses. This approach is sensible for initiation
(which is controlled by shear stress).
For low cycle fatigue, usually strain range rather than stress range is used.
multiaxial conditions, the effective strain range is taken as

Under

eff = 1 - 3
where 1 2 3 are principal strains
Influence of Manufacture and Service Conditions
Large scale residual stresses are often present as a result of the method of
manufacture:Casting non-uniform solidification and cooling leads to differential thermal
contraction;
Welding weld metal contracts during cooling to give tensile stress in the weld (see
later);
Machining mechanical and thermal stresses combine to create complex residual
stresses
These residual stresses have little effect on static strength in a ductile material, but
contribute to fatigue strength considerably because of the mean stress effect: tensile
stresses reduce fatigue strength; compressive stresses are beneficial particularly at
the surface where most fatigue cracks start. Stress relieving heat treatments of large
cast components and welds are commonplace.
Surface roughness and metallurgical condition also influence fatigue strength.
Surface treatment (high cycle fatigue)
Surface treatment (mechanical and/or metallurgical) can be used to induce
compressive residual stresses near the surface. This has the effect of reducing the
mean stress near the surface (where fatigue cracks initiate) under cyclic loading
conditions. Hence the cyclic stress can be increased by comparison with a component
which does not have compressive residual surface stress.
(i)
Mechanical Methods
Pre-loading

(ii)
Metallurgical Methods
flame-hardening
33

Shot-peening
Surface rolling
Autofrettage

carburising
nitriding

Welds
When two components are welded together, residual stresses are created. This is
because the weld metal solidifies and cools and hence tries to contract, whereas the
components (which are relatively cool) do not contract as much.
Hence the
contraction of the weld metal is resisted by the rest of the component, thus inducing
tensile residual stresses in the weld region and compressive residual stresses
elsewhere.
The residual stresses are often as high as the yield stress of the material. As well as
the residual stresses, weld regions inevitably contain flaws. Also, the metallurgical
structures, grain size etc. in the weld metal (WM), heat-affected-zone (HAZ) and
parent metal (PM) are all different. Stress concentrations occur in weld regions (e.g.
at weld toes).
The toe and root regions often contain cracks after welding. Hence under cyclic
loading conditions these cracks can grow and lead to fatigue failure. Fatigue failures
may be in any of the WM, HAZ or PM regions.
All of the above features result in reductions in the fatigue strength of welds as
compared with the parent metal.
In some cases it is possible to induce a more favourable microstructure by heat
treatment.
Stress concentrations and weld toe cracks etc. can be reduced or
eliminated by grinding or other machining processes. The tensile residual stresses can
be reduced or even converted to beneficial compressive residual stresses by surface
treatment (e.g. peening).
BS153 contains welding standards and KG Richards
(Fatigue Strength of Welded Structures, W I Abington Hall, England) has classified
them.
Because of the larger number of variables, the scatter in fatigue data for welds is
much greater than that for plain specimens.

34

Fretting Fatigue
If two metals are pressed together by a steady force and are also subjected to cyclic
loading, so that there is relative displacement between the two surfaces, wear occurs.
When the relative (rubbing) displacement is small ( 0.1 mm) the wear is called
fretting.
The wear is a result of the continuous welding and breaking of the
protrusions on the surfaces. The small particles of metal which are rubbed off get
trapped between the surfaces and increased the wear and pitting of the surface. If the
particles oxidise the wear rate may be very large because the metallic oxides are
harder than the metals themselves and thus cause abrasive wear.
Examples include:-

(i)
(ii)
(iii)

Fretting is increased:- (i)


(ii)
(iii)
(iv)

pinned, bolted and riveted joints


between leaf springs
press and shrink fit assemblies.
under perfectly dry conditions,
when the debris oxidises,
by increasing contact pressure and slip amplitude,
in soft metals by comparison with hard metals.

Suitable surface treatment and lubrication can reduce or even eliminate fretting.
The surface damage caused by fretting reduces the fatigue limit by comparison with
plain specimen, e.g. from 620 MPa to 193 MPa in a titanium alloy by a clamping
pressure of 28 MPa. Increasing the pressure further to 420 MPa had a relatively small
effect, reducing the fatigue limit to 123 MPa (Further details:- Metal Fatigue, Frost,
Marsh & Pook).
35

Environment, Temperature and Frequency Effects


Most fatigue tests are performed in air at high frequency (to reduce test durations)
and at ambient temperature. However, many components operate in corrosive
environments, with relatively low frequencies of cyclic loading and at a variety of
temperatures.
Frequency effects
Provided that increases of temperature which can occur due to
cyclic plasticity (hysteresis) have no effect, then loading frequency, in the range 0 to
200 Hz, seems to have relatively little effect on both the initiation and growth of
cracks. In plain bar tests, the fatigue strength increases with frequency between 200
and 2000 Hz. Above 2000 Hz the evidence is contradictory. This is possibly due to
lack of knowledge about the actual test conditions due to inertia effects etc.
The stress-strain curves for some materials are rate-dependant. Therefore, smaller
crack tip plastic strains (and zones) may occur with higher loading rates. This may
result in lower crack growth rates (with respect to cycle number).
Temperature effects When temperature is reduced below 20o C, the fatigue strength
increases.
Therefore, room-temperature data gives conservative predictions for
fatigue strength at lower temperatures. However, the effect of low temperature on
brittle (fast) fracture must be considered. At high temperatures corrosion and/or
metallurgical changes can occur; these may reduce fatigue strength. Therefore
corrosion resistant alloys have been developed, e.g. blade and disc materials for gas
turbines. Since metallurgical changes and surface chemical attack are time-dependant
processes, frequency may appear to have an effect.
Some materials have higher tensile strengths (and fatigue strengths) at elevated
temperatures that at ambient temperature. For pure metals and simple alloys, the
fatigue strength decreases with increasing temperature.
Creep at elevated temperature affects fatigue behaviour; creep-fatigue is dealt with
later in the course.
The environment can cause surface corrosion (pitting) which may reduce fatigue
strength. This behaviour is called corrosion fatigue. The presence of oxygen can often
be particularly damaging. Testing in air by dripping a corrosive material onto a fatigue
test specimen may be more damaging (because of the presence of oxygen) than
testing while completely immersed in the corrosive material.
36

Example
Material

Fatigue strength (MPa)


In air

In fresh water

In salt water

Mild steel

250

140

55

18/8 Austenitic
steel

385

355

250

Nickel

340

200

160

37

Vous aimerez peut-être aussi