Você está na página 1de 12

MODULE B51EM ADVANCED MECHANICS OF MATERIALS

Topic 1.3: Fatigue

1.3.1: The fatigue phenomenon


Fatigue is a type of fracture, which takes place gradually over a period of time under
cyclic stress. The mechanism (Figure 1i) essentially involves the shear stresses ahead
of a crack tip and the crack grows by one increment in each cycle of loading.

Figure 1

A fatigue test normally consists of applying a cyclic stress, and measuring number of
cycles to failure, N (Figure 2). It is normal to quantify the cyclic stress as the
difference between maximum and minimum stress, the stress range, S, twice the stress
amplitude σa. An S-N curve is usually plotted on a log-log scale, since numbers of
cycles to failure usually reach millions or hundreds of millions. Some metals (notably
steels) show an endurance limit, which is the stress range below which fatigue will
not occur no matter how many cycles are applied. In others (e.g. aluminium alloys), it
is normal to define an endurance limit, which is the stress range at which the number
of cycles to failure is acceptably high (typically 108).

Stress logS
σa Endurance limit

σm
2σa,e
Fatigue limit
2σa,l

Time logN
Figure 2
Tests in manufactured components generally produce very scattered results, mainly
because fatigue cracks grow from the most highly loaded stress concentrator on the
structure, which is often associated with a manufactured discontinuity, such as a weld
toe. Figure 3 shows the results from some tests on butt welds, which indicate the kind
of scatter involved.

Figure 3

A lot of the variation in fatigue results is associated with crack initiation, where a
crack is formed from a concentrator by a process of slip on relatively few atomic
planes. Figure 4ii shows a crack being formed on a highly polished piece of nickel.

Figure 4

S-N curves are normally specific not only to the material, but also to the particular
geometry of the structure, and even the environment. In some cases, a set of curves
are produced for different probabilities of failure.
1.3.2: Fatigue limit design
Many applications (such as machinery) require designers to use an infinite (or near-
infinite) fatigue life. When dealing with such fatigue limit design it is normal to use
stress amplitude (σa) and it is necessary to consider the mean stress (σm). A fatigue
diagram is constructed using the above quantities and also the yield stress in tension
(σy,t )and in compression (σy,c) and the ultimate tensile strength, (σu), Figure 5, a
fatigue-safe region being identified within this diagram.
Stress
amplitude
σy,t

σa,e

Fatigue-safe region

σy,c σy,t σu
Compression Mean stress Tension

Figure 5

The Goodman relationship is often simplified to a formula which relates the


amplitude at the fatigue (or endurance) limit σ a for a mean stress of σ m to that at a
⎛ σ ⎞
mean stress of zero: σ a = σ a ,e ⎜1 − m ⎟
⎝ σu ⎠

1.3.3: Using S-N Curves


Fatigue is a type of fracture which takes place gradually over a period of time under
cyclic stress. It is normal to quantify the cyclic stress as the difference between
maximum and minimum stress, the stress range, S. Tests in manufactured
components generally produce very scattered results, mainly because fatigue cracks
grow from the most highly loaded stress concentrator on the structure, which is often
associated with a manufactured discontinuity, such as a weld toe. A lot of the
variation in fatigue results is therefore associated with crack initiation.

The traditional way of dealing with this has been to plot the stress range S, versus the
number of cycles to failure at that stress range, N, on a log-log scale. S-N curves are
normally specific not only to the material, but also to the particular geometry of the
structure, and even the environment. In some cases, a set of curves are produced for
different probabilities of failure, e.g Figure 6iii.
Figure 6

Most practical fatigue loading situations will involve a spectrum of stress range and,
for design purposes, it is often possible to assume that the fatigue loading is
represented by several stress ranges, S1, S2,…. Si acting simultaneously. In such cases,
Miner’s Rule of cumulative fatigue damage can be used:

ni
D=∑
i Ni

where ni is the number of cycles experienced in a given time, t, at Si and Ni is the


number of cycles to failure at that stress range. A fatigue damage of unity corresponds
to failure in the time, t, so that the fatigue life can be calculated as t/D.

Worked Example on Fatigue by the S-N Method

Question: An aircraft fuselage made from 7075-T6 aluminium alloy is subject to the
following alternating stresses; S = 200 MPa, 100 times per day and S = 300 MPa
twice daily. Calculate the fatigue life of the fuselage for failure probabilities of 0.01
and 0.99. (Use Figure 6)

Answer: Be careful about interpolation when reading logarithmic scales. The


following table uses the P = 0.01 and the P = 0.99 curves in Figure 6.

S (MPa) No. of N0.01 N0.99 n/N0.01 n/N0.99


cycles in 1
Day, n
200 100 2.5×106 1×109 4×10-5 1×10-7
300 2 4.5×104 3×105 4.44×10-5 6.67×10-6

ni
So, for the P = 0.01 case, ∑N = 8.44 ×10−5 , and hence the fatigue life is 1 day/D
i
which is just under 32½ years. The corresponding value for P = 0.99 is about 405
years. So, we can conclude that there is a 99% chance that the fuselage will fail within
405 years and there is a 1% chance that it will fail within 32½ years. If you were
responsible for the maintenance of a fleet of 100 ‘planes, you would plan your
inspection intervals on this basis.

Note that the 300 MPa cycles take as much out of the fatigue life (n/N values
approximately the same) as do the 200 MPa cycles even though there are a lot less of
them. This shows the critical importance of stress range on fatigue life and is a
consequence of the logarithmic relationship between S and N.

1.3.4: The Fracture Mechanics Approach to Fatigue


The fracture mechanics approach is based on crack propagation and therefore assumes
that a fatigue crack is present at the beginning of the fatigue life. This is rarely the
case, and the approach is therefore confined to cases where a crack has been
identified and it is necessary to know the time it will take to propagate to a critical
size. Alternatively, it may be used to determine inspection intervals where the initial
crack size is assumed to the smallest crack which can be found by inspectors.

The fundamental law of fatigue crack propagation is the Paris Law:

da
= C ∆K m
dn

and, taking ∆K = Y ∆σ π a = YS π a , this can be rewritten:


da
= CY m S mπ 2 a 2
m m

dn

Separating variables and integrating:


N af
da
∫ dn = ∫ CY
0 ai
m
S mπ 2 a 2
m m

i.e.
af
1 da
N=
CS π m m
2 ∫Y
ai
m m
a 2

Provided that ai and af are finite, this equation can be integrated to give the number of
cycles a crack takes to propagate from one size to another. Special fatigue tests need
to be carried out to determine the material constants m and C .

Note that taking the logarithm of both sides of the above equation gives:

log N = log M + log G − m log S

where M is material-dependent and G is geometry-dependent, which says that one


would expect S-N curves for a given material all to have the same slope, but a
different intercept.
Worked Example on Fatigue by the F-M Methodiv

Question: A large steel sheet is subject to alternating stress between 100MPa tension
and 50MPa compression and has been found to contain and edge crack of depth
2mm. If the plane strain fracture toughness is 25 MPa m , m is 3.0 and C is1×10-12
(both in MPa and m units), find the fatigue life of the plate in number of cycles.

Answer: First find the critical crack length at maximum stress:


2
1⎛ K ⎞
2
1⎛ 25 ⎞
acrit = ⎜ Ic ⎟ = ⎜ = 1.58 × 10−2 m
π ⎝ Y σ max ⎠ π ⎝ 1.12 ×100 ⎟⎠

To find the number of cycles to failure, calculate N for the crack to grow from an
initial size of 2×10-3m to the final size equal to the critical:

af 1.58×10−2
1 da 1 da
N=
CS π m m
2 ∫a Y m a m 2 = 1×10−12 ×1503 × π 1.5 ×1.123 ∫ a1.5
i 2×10−3

1.58×10−2 2×10−3
⎡ −1 −0.5 ⎤ ⎡ 1 ⎤
= 3.787 × 10 ⎢ a ⎥
4
= 7.575 ×10 ⎢ 4

⎣ 0.5 ⎦ 2×10−3 ⎣ a ⎦1.58×10−2
= 1.09 × 106

1.3.5: Fatigue Analysis for Distributed Loadingv


Most cases of fatigue loading do not involve sinusoidal loading, and we have seen
how to cope with this in cases where the loading can be split up into blocks of
different amplitude and frequency (using the Miner Sum). Here, we extend the Miner
Sum to an integral for case where we can write a mathematical probability
distribution function for the loading. This type of approach is important for cases of
natural loading (wind, waves etc.). It is fairly common to describe wave (and other
natural types of) loading as a Rayleigh distribution, which is a probability density
function. This function can be described as a straight line on semi-log axes (see
opposite), which shows the number of expected exceedances for each wave height.
The particular line in Figure 7 shows exceedances expected in 100 years from a NW
direction in the North Sea. As can be seen, Hm is the “100 year wave” of which there
is expected to be one every 100 years. Also nm is the number of waves expected in
100 years (i.e. number of waves of height greater than zero). The wave exceedance
data can therefore be described algebraically as:
H = H m + m2 log n

and, since H = 0 when log n = log nm

−Hm
m2 =
log nm
−2.303H m ⎞
nm = exp ⎜⎛
m2 ⎟⎠
i.e.

H − Hm H
Since log n = = + log nm
m2 m2

then n = nm exp ⎛⎜ 2.303H ⎞⎟


⎝ m2 ⎠

Figure 7

Differentiating, the number of waves in dn in a height range dH is given by:

exp ⎛⎜ 2.303H ⎞⎟ dH .................................[1]


2.303
dn = − nm
m2 ⎝ m2 ⎠

This equation can hold for any period of time. In the particular case here, we have
used 100 year wave data, and so dn is the number of waves expected in 100 years in
the height range dH.

The above equation can therefore be used to determine the number of cycles to be
expected at a given stress range, provided that we can determine relationship between
the wave height, H, and the stress range S. Williams and Rinne used some empirical
data for offshore platforms which indicated that, to an acceptable level of
approximation, the stress range (the so-called “hot-spot” stress range) at the critical
tubular joints was given by an expression:

S = aH b .................................[2]

where a and b are constants related to the design of the platform and the configuration
of the critical joints.

Finally, it is possible to write an algebraic equation representing the S-N curve for the
critical tubular joints, Figure 8vi.

Figure 8

i.e. S = S1 N m1 .................................[3]

Using equations 2 and 3, we can express the number of cycles to failure, N, as a


function of H, N(H), rather than as a function of S (as is implicit in equation [3]):
1
⎛ aH b ⎞
m1

i.e. N (H ) = ⎜ ⎟ .................................[4]
⎝ S1 ⎠

The next stage involves an extension of Miner’s Rule, which, so far, we have
regarded as summation of a discrete series of blocks of stress range, Si, with their
associated numbers of cycles in some reference period, ni. The Si are used to obtain
the relevant values of Ni. Instead of doing this, we now have a probability distribution
of numbers of waves of height H, and we have the corresponding relationship (S-N
data) expressed in terms of wave height:

n dn
i.e. Dref =∑ i becomes ∫
i Ni H =0
N (H )

The process, compared with Miner’s Rule is summarised in Figure 9.


Substituting from equations [1] and [4] into the damage integral above, we can obtain
an integral for the damage in 100 years (since equation [1] relates to a reference

period of 100 years):


2.303nm −b ⎛ 2.303H ⎞
D100 yr = ∫ − H ⎟ dH ....................................[5]
m1
exp ⎜
⎝ m2
1

m2 ⎛⎜ a ⎞⎟ ⎠
m1
H =0
⎝ S1 ⎠

H
log(H)

n(H)

p(H) logN(H)


dn
Damage = ∫
H =0 N
Figure 9

Since we are integrating w.r.t. H, it is important that all the dependence on H is


explicit (for example, if the expression contained S, which varies with H, we would
not be able to perform the integration). It can be seen that the integration is of the
type:

∫k H e − k3h
k2
1

which is a recursive integral which can be solved with the aid of the Gamma
Function:

Γ ( z ) = ∫ e − t t z −1dt
0

whose values can be found tabulated as a function of the “argument” z. Comparing


equation [5] with the definition of the Gamma Function, we can set:

2.303H 2.303
t=− dt = − dH and z = 1− b
m2 m2 m1

Therefore:
−b

⎛ 2.303H ⎞ ⎡ 2.303H ⎤ ⎛ −2.303 ⎞
m1

Γ ⎛⎜1 − b ⎞⎟ = ∫ exp ⎜ ⎟ ⎢− ⎥ ⎜ ⎟ dH ....[6]


⎝ m1⎠
0 ⎝ m2 ⎠ ⎣ m2 ⎦ ⎝ m2 ⎠

Dividing [5] by [6], we can write the damage integral in terms of a Gamma Function
with an argument of (1 – b/m1):

⎡ ⎤
⎢ −2.303n ⎥ ⎛ −2.303 ⎞ m1 ⎛ − m ⎞
b

D100 yr ⎛
= Γ ⎜1 − b ⎞ ⎢ m
⎥⎜
m1 ⎟⎠ ⎢ ⎟ ⎜
2

⎝ m ⎠ ⎝ ⎠
1
⎛ a ⎞ ⎥⎝ 2.303
m1

m
⎢⎣ 2 ⎜⎝ S1 ⎟⎠ ⎥⎦
2

which simplifies to:

−1 b
⎛ −2.303 ⎞
m1
a
Γ ⎛⎜1 − b ⎞⎟ ⎜
m1

D100 yr = nm ⎟
S1
−1
m1
⎝ m1 ⎠⎝ m2 ⎠

from which the fatigue life (in years) can be obtained as 100/D100yr

2.3.6: Diagnosing fatigue

Fatigue failures (Figures 10 and 11) have a characteristic microscopic and


macroscopic appearance and are not difficult to diagnose, except in brittle materials.
Figure 11

Figure 10

At the macroscopic level the fatigue crack propagation


surface is generally very flat with characteristic
“clamshell” markings which indicate the progress of the
fatigue crack with time. The clamshell markings are
essentially due to differential corrosion of the exposed
surface and therefore become more metallic in
appearance the further from the origin. The point at
which the stable part of the crack propagation becomes Figure 12
unstable is usually also relatively simple to establish.
For example, Figure 12 shows that the final ductile tear under bending stresses
occurred once the fatigue crack had crossed about half of the section, whereas the
example in Figure 13 shows the fatigue crack to have almost entirely crossed the
section. On the opposite extreme, Figure 14 shows a brittle fracture propagating from
a very shallow fatigue crack. The extent to which a fatigue crack will propagate prior
to section failure will depend on the maximum stress in the cycle and also on the
fracture resistance of the material and geometry. For spectral loading (for example
that due to wave forces) final fatigue fractures often take place during periods of high
maximum stress (e.g. storms) but this does not mean that the fatigue failure is caused
by the storm itself.

Figure 14

Figure 13

Fatigue fracture surfaces are also relatively easy to characterise at the microscopic
level, provided that the surface remains undamaged either by mechanical action or by
corrosion. Figure 15 shows the point of ductile rupture following fatigue loading in a
laboratory specimen of stainless steel, and the higher magnification (Figure 16) shows
clearly the individual fatigue striations, each of which corresponds to an individual
crack growth increment (da/dn). It is sometimes possible (although this would have to
be done carefully) to determine roughly the loading conditions i.e. ∆K from
observations of a fatigue fracture surface, provided that the local crack length (a) is
known and the material constants C and m are known, Figure 17vii.

Figure 15 Figure 16

Figure 17

i
Diagram from Reuben R L Materials in Marine Technology, Springer-Verlag, 1994.
ii
Photograph from Engel L and Kilingele H An Atlas of Metal Damage, Hanser Verlag 1981.
iii
Diagram from Callister W D Materials Science and Engineering, Wiley, 2000.
iv
Example from Callister W D Materials Science and Engineering, Wiley, 2000.
v
Williams and Rinne, Proc. Inst. Civ. Engrs. 60(1), pp635-654, 1976
vi
Diagram from Reuben R L Materials in Marine Technology, Springer-Verlag, 1994.
vii
Diagram from Callister W D Materials Science and Engineering, Wiley, 2000.

Você também pode gostar