Professional Documents
Culture Documents
= R
Frequency: or f in units of Hz. For rotating
machinery at 3000 rpm, f = 50 Hz. In general
only influences fatigue if there are environmental
effects present, such as humidity or elevated
temperatures.
Waveform: Is the stress history a sine wave,
5
Waveform: Is the stress history a sine wave,
square wave, or some other waveform? As with
frequency, generally only influences fatigue if
there are environmental effects.
S - N Curve
If a plot is made of the applied stress amplitude
versus the number of reversals to failure (S-N curve)
the following behavior is typically observed:
6
S - N Curve Endurance Limit
If the stress is below
e
(the endurance limit or
fatigue limit), the component has effectively
infinite life.
e
0.35
TS
- 0.50
TS
for most steels and
7
e
0.35
TS
- 0.50
TS
for most steels and
copper alloys.
If the material does not have a well defined
e
,
often
e
is arbitrarily defined as the stress that
gives N
f
=10
7
.
Fatigue Design Approaches
Stress-Life Approach
If a plot is prepared of log(
a
) versus log (2N
f
) (where
2N
f
represents the number of reversals to failure, one
cycle equals two reversals) a linear relationship is
commonly observed. The following relationship
8
commonly observed. The following relationship
between stress amplitude and lifetime (Basquin, 1910)
has been proposed:
( )
b
f f a
N 2
2
'
= =
A
In the previous expression
f
m
0?
11
( ) Soderberg
y
m
a a
m
=
=
1
0
( ) Goodman
m
a a
m
)
`
=
=
1
0
12
( ) Goodman
TS
a a
m
)
`
=
=
1
0
( ) Gerber
TS
m
a a
m
|
|
.
|
\
|
=
=
2
0
1
In the previous expressions,
a
is the stress amplitude
denoting the fatigue strength for a nonzero mean stress,
is the stress amplitude (for a fixed life) for
fully reversed loading (
m
= 0 and R = -1), and
y
and
TS
are the yield strength and the tensile strength,
respectively.
0 =
m
a
13
Soderberg gives a conservative estimate of the
fatigue life.
Gerber gives good predictions for ductile alloys for
tensile mean stresses. Note that it cannot distinguish
between tensile and compressive mean stresses.
Fatigue Design Approaches
Different Amplitudes
How do we handle situations where we have varying
amplitude loads, as depicted below?
14
The component is assumed to fail when the total
damage becomes equal to 1,or
1 =
i
i f
i
N
n
15
It is assumed that the sequence in which the loads
are applied has no influence on the lifetime of the
component. In fact, the sequence of loads can have
a large influence on the lifetime of the component.
A very common approach is the Palmgren-Miner
linear damage summation rule.
If we define 2N
f i
as the number of reversals to
failure at
ai
, then the partial damage d for each
different loading
ai
is
16
different loading
ai
is
ai
ai
fi
i
N
n
d
at failure to Reversals
at Reversals
2
2
= =
Consider a sequence of two different cyclic loads,
a1
and
a2
. Let
a1
>
a2
.
Case 1: Apply
a1
then
a2
.
In this case, can be less than 1. During the
i
i
N
n
17
In this case, can be less than 1. During the
first loading (
a1
) numerous microcracks can be
initiated, which can be further propagated by the
second loading (
a2
).
i
i f
N
Case 2: Apply
a2
then
a1
.
i
i f
i
N
n
In this case can be greater than 1. The first
loading (
a2
) is not high enough to cause any
microcracks, but it is high enough to strain
harden the material. Then in the second loading
18
harden the material. Then in the second loading
(
a1
), since the material has been hardened it is
more difficult to initiate any damage in the material.
Fatigue Design Approaches
Strain-Life Approach
The stress-life approach just described is applicable
for situations involving primarily elastic deformation.
Under these conditions the component is expected to
have a long lifetime. For situations involving high
19
have a long lifetime. For situations involving high
stresses, high temperatures, or stress concentrations
such as notches, where significant plasticity can be
involved, the approach is not appropriate. How do
we handle these situations?
Rather than the stress amplitude
a
, the loading is
characterized by the plastic strain amplitude
Under these conditions, if a plot is made of
versus log(2N
f
), the following linear behavior is
generally observed:
. 2 /
p
e A
|
|
.
|
\
|
e A
2
log
p
20
To represent this behavior, the following relationship
(Coffin-Manson, 1955) has been proposed:
( )
c
f f
p
N 2
2
e
'
=
e A
21
2
where is the plastic strain amplitude, is the
fatigue ductility coefficient (for most metals ,the
true fracture ductility) and c is the fatigue ductility
exponent (-0.5 to -0.7 for many metals).
2 /
p
e A
f
e
'
f
e
Fatigue Design Approaches
General Approach
In general, how does one know which equation to apply
(stress-life or strain-life approach)? Consider a fully
reversed, strain-controlled loading. The total strain is
composed of an elastic and plastic part, i.e.
22
2 2 2
p
e
e A
+
e A
=
e A
The Coffin-Manson expression may used to express the
term . What about ?
2 /
p
e A
2 /
e
e A
From the Basquin Law (stress-life approach):
( )
b
f f
N 2
2
'
=
A
For 1-D, elastic loading
and thus
E E
a e
/ 2 / 2 / = A = e A
23
and thus
( )
b
f
f
e
N
E
2
2
'
=
e A
and we may write that:
( ) ( )
c
f f
b
f
f
N N
E
2 2
2
e
'
+
'
=
e A
Fatigue Design Approaches
A plot of this is depicted below:
24
Fatigue Design Approaches
HCF/LCF
If the amplitude of the total strain is such that we have
significant plasticity, the lifetime is likely to be short
(Low Cycle Fatigue or LCF; strain life approach). If
the stresses are low enough that the strains are elastic,
the lifetime is likely to be long (High Cycle Fatigue
25
the lifetime is likely to be long (High Cycle Fatigue
or HCF; stress-life approach).
The transition life (at 2N
t
) is found by setting the
plastic strain amplitude equal to the elastic strain
amplitude.
( ) ( )
c
t f
b
t
f
N N
E
2 2 e
'
=
'
c b
f
f
t
E
N
|
|
.
|
\
|
'
e
'
=
1
2
r
c
d
g
r
c
d
g
r
c
d
g
* large influence on crack growth for certain combinations of environment, load ratio and frequency.
r
c
and d
g
refer to the cyclic plastic zone size and the grain size, respectively.
Fatigue Crack Growth
Regime A
Concept of the threshold stress intensity factor K
t h
:
when Kis K
th
, where K
th
is the threshold stress
intensity factor, the rate of crack growth is so slow
that the crack is often assumed to be dormant or
11
that the crack is often assumed to be dormant or
growing at an undetectable rate.
An operational definition for K
th
often used it is
that if the rate of crack growth is 10
-8
mm/cycle or
less the conditions are assumed to be at or
below K
th
.
An important point is that these extremely slow
crack growth rates represent an average crack
advance of less than one atomic spacing per cycle.
How is this possible? What actually occurs is that
there are many cycles with no crack advance, then
the crack advances by 1 atomic spacing in a single
12
the crack advances by 1 atomic spacing in a single
cycle, which is followed again by many cycles with
no crack advance.
Fatigue Crack Growth
Regime B
When we are in regime B (Paris regime) the following
calculation can be carried out to determine the number
of cycles to failure. From the Paris Law:
( )
m
K C
dN
da
A =
13
( ) K C
dN
A =
K can be expressed in terms of
a Y K A = A
Where Y depends the specific specimen geometry.
Thus the Paris Law becomes:
( )
m
a Y C
dN
da
A =
Assume that Y is a constant. Integrate both sides:
14
Assume that Y is a constant. Integrate both sides:
( )
} }
A =
f f
N
m
m
m
a
a
m
dN CY
a
da
0
2 /
2 /
0
For m > 2:
( ) ( ) ( )
( )
( )
( )
(
(
A
2 / 2 2 / 2
0
2 /
1 1
2
2
m
f
m
m
m
m
a a CY m
N
f
=
15
For m = 2:
( )
0
2
2
1
1
a
a
n
CY
N
f
f
A
=
The constants C and m are material parameters
that must be determined experimentally. Typically
m is in the range 2-4 for metals, and 4-100 for
ceramics and polymers.
For cases where Y depends on crack length, these
integrations generally will be performed numerically.
16
integrations generally will be performed numerically.
Important: Note that in the Paris regime the rate
of crack growth is weakly sensitive to the load
ratio R. The key parameter governing crack
growth is K .
In these expressions, we need to determine the initial
crack length a
0
and the final crack length a
f
(sometimes called the critical crack length).
How do we determine the initial crack length a
0
?
Cracks can be detected using a variety of techniques,
17
Cracks can be detected using a variety of techniques,
ranging from simple visual inspection to more
sophisticated techniques based ultrasonics or x-rays.
If no cracks are detectable by our inspection, we
must assume that a crack just at the resolution of
our detection system exists.
How do we determine the final crack length a
f
?
We know that eventually the crack can grow to a
length at which the material fails immediately, i.e.
c
K K
max
18
or
c f
K a Y
max
Thus we may solve for a
f
as follows:
2
max
2
2
1
Y
K
a
c
f
=
A very important idea that comes from this analysis
is the following: even if a component has a detectable
19
is the following: even if a component has a detectable
crack, it need not be removed from service! Using
this framework, the remaining life can be assessed.
The component can remain in service provided it is
inspected periodically. This is the crack-tolerant or
damage tolerant design approach.
Fatigue Striations
An advancing fatigue crack leaves characteristic
markings called striations in its wake. These can
provide evidence that a given failure was caused by
fatigue. The striations on the fracture surface are
20
fatigue. The striations on the fracture surface are
produced as the crack advances over one cycle, i.e.
each striation corresponds to da.
21
Fatigue striations on the fracture surface (2024-T3
aluminum). Crack advance was from the lower left to
the upper right.
Models for Paris Regime
1. Geometric (or) CTOD models
( )
E
K
dN
da
y
t
' '
A
= A ~
2
22
striation spacing
Important: This implies that the Paris Exponent m is
equal to 2.
Advantages of CTOD-based models:
Physically appealing
Usefulness in multiaxial fatigue
Link to microstructural size
2. Damage Accumulation Models
23
2. Damage Accumulation Models
Basic Idea: Accumulation of local plastic strain over
some microstructually significant distance at crack tip
leads to fracture.
( )dx x u
u dN
da
rc
y
y
0 ,
4
0
* }
A =
In the previous expression u
*
represents the critical
hysteresis energy, or the work of fracture and r
c
represents some microstructurally significant distance.
( )
* 2
4
u
K
dN
da
24
* 2
u dN
y
x x
+
y y
y y
+ etc). The only
nonzero term in this expansion will be
1/2(
y y
y y
). We know that
y y
= /H, so
we need to solve for
y y
. We know that
( ) ( )
zz xx yy yy
v
E
+ = e
1
We also know that
z z
= 0 (plane stress). We have
also assumed that
x x
= 0, so that
( ) v = = e
1
0
15
( )
yy xx xx
v
E
= = e
1
0
yy xx
v =
and thus
( )
2
1
1
v
E
yy yy
= e
E E
yy yy
= e =
so
and finally
16
H v v
yy yy
2 2
1 1
= e
=
( )
2
2
1 2 2
1
|
.
|
\
|
= = u
}
H v
BbHE
d
v
so
If the crack grows all the way across the plate and
the plate separates into two pieces, the system will
have zero strain energy. So the strain energy
decreases as the crack grows. Calculate the change
in strain energy d of the plate if the crack grows
by some amount da. Use the same assumptions used
17
by some amount da. Use the same assumptions used
previously to calculate the total strain energy. Then
use the result that
fixed u
a B
(
c
u c
=
1
( )
da
H v
BHE
d
2
2
1 2
|
.
|
\
|
= u
u c
and thus
19
( )
2
2
) (
1 2
1
H v
HE
a B
fixed u
=
(
c
u c
=
Use the relationship between energy release rate
and stress intensity factor for plane stress:
E
K
2
=
( )
2
2
1 2
1
,
v H
E K E K
= =
So
20
( )
2
1 2 v H
Assume that this material has a know fracture
toughness (plane stress) K
Ic
. Calculate the value of
that must be imposed to make the crack advance.
Solution
To find the critical value of for crack advance,
simply set K (determine above) equal to K
Ic
and
solve for :
21
( )
2
1 2 v H
E
K
Ic
=
Blister Test
Evaluate K
The interface toughness of an adhesive is to be
estimated on the basis of a proposed method which
entails the bonding of a thin elastic disk with the
adhesive on a rigid substrate. Pressure is applied
22
adhesive on a rigid substrate. Pressure is applied
to the bonded side of the disk through a tiny hole
in the substrate, by pumping a fluid. This
pressurization results in partial debonding of the
disk and the formation of a bulge or blister in
the disk. This method is hence referred to as the
pressurized bulge test or blister test.
Consider a two-dimensional version of this test
which is shown in the figure. A thin beam is
adhesively bonded to a rigid surface through which
an incompressible fluid is forced resulting in a
blister of half-width l. Assuming that all concepts
of linear elastic fracture mechanics may be extended
23
of linear elastic fracture mechanics may be extended
to this case, apply the compliance method to
determine the tensile crack-tip stress intensity
factor in terms of the blister dimension l and
the piston displacement q.
24
Assume that q = 0 when the beam is undeflected,
and that all length dimensions normal to the plane
of the figure are unity. Let the thickness of the beam
and its elastic modulus be h and E, respectively.
You are given the following information:
1. The transverse deflection of an elastic double
25
1. The transverse deflection of an elastic double
cantilever beam of length 2l subjected to a
uniform pressure is
( )
( )
3
2
2 2
2Eh
x l p
x
=
2. A measure of the compliance c(l) may be defined
as q = c(l) Q, where Q is the force acting on the
piston per unit thickness as shown in the figure.
( ) qH dx x w
l
l
=
}
3.
26
4. From the compliance interpretation of the
energy release rate, , it is known that
2
2
1
Q
dl
dc
=
Solution
The compliance of the system is defined as
Observe that (i) the pressurized fluid is
( )Q l c q =
27
Observe that (i) the pressurized fluid is
incompressible, (ii) the pressure under the blister
is equal to the pressure in the piston, and (iii) the
pressure in the piston is directly related to Q by
H
Q
p =
Substitute the expression for the displacement w
into the incompressibility condition to obtain:
5
3
8
15
l
EqHh
p =
Use these two equations to eliminate p
28
Use these two equations to eliminate p
Q
H h
l
E
q
2 3
5
1
15
8
=
This is our compliance relationship.
2 3
5
1
15
8
) (
H h
l
E
l c =
Using the compliance interpretation of , (note:
there are two crack tips at the blister) we find that
4
29
2
2 3
4
2
1
6
4
2
1
Q
H h
l
E
Q
dl
dc
tip per
=
|
.
|
\
|
= =
But q = c (l)Q. Using this result
2
6
2 3
32
75
q
l
H h
E =
For plane stress, we take this equation with the
link between K
I
and to get
( ) ,
32
75
2 1
2
6
2 3
2
2 1
|
|
.
|
\
|
= = q
l
H h
E E K
I
30
32
|
.
\
l
qE
l
h Hh
K
I
3
2
3
4
5
=
or
Aircraft turbine engines
HCF
Youngs modulus E appears in this equation
because K
I
has been written in terms of the
displacement q. If we make use of the known
relationship between q and Q, the stress
intensity factor becomes
31
intensity factor becomes
Q
h Hh
l
K
I
2
3
2
=
which is the desired result.
Stability of Crack Growth
In this worked example, we discuss the driving force
for fracture in brittle solids and the issues of stability
of subcritical crack growth by recourse to the concepts
of strain energy release rate which were discussed in
lecture. Consider a cracked specimen which is loaded
under compliant conditions in Mode I, as shown in the
32
under compliant conditions in Mode I, as shown in the
following figure. The specimen has unit thickness
(B =1). C is the compliance (a function of crack
length) of the specimen and C
M
is the compliance of
the testing machine (which may be regarded as a
spring) in which the specimen is loaded in tension.
The testing machine is in series with the specimen.
33
Let
T
be the total displacement which can be
regarded as prescribed.
. A
|
.
|
\
|
+ A = + A = A
C
C
P C
M
M T
The potential energy is given by
34
The potential energy is given by
2
2
2
2
2
1
2
1
2
1
A +
A
= + u =
C
C
C
P C W
M
M p
( )
(
A A
+
A
=
M
T
C C
2
2
2
1
2
1
1. Show that C
M
= 0 corresponds to fixed grip
loading (constant displacement), and that
C
M
corresponds to dead-weight loading.
2. Show that the mechanical strain energy
release rate,
35
release rate,
( )
da
dC
P
a
W
T
P 2
2
1
=
(
c
c
=
A
c
c
a
36
4. For an ideally brittle solid, show that crack
advance is stable if 0 and that it is
unstable if > 0.
T
A
(
c
c
a
T
A
(
c
c
a
Solution
1. For fixed grip loading,
T
= . From this,we note
that C
M
/C = 0 and C
M
= 0. For dead-weight
loading, the rate of change of the load P with
time is zero, i.e.
37
0 =
A
=
C
P
+
A
=
(
c
c
=
A
2
2
2
=
A
=
A
=
A
Note that
39
P
C C C
M
M
T
T
=
A
=
A
=
A
A + A = A
M T
From this equation, we note that C
M
+C = C
T
.
Thus it is seen that
da
dC P
=
2
2
\
|
+
A
+
|
.
|
\
|
+
A
=
c
c
2
2
2
2
2
3
2
2
1
da
C d
C C
da
dC
C C
a
M
T
M
T
( )
|
|
.
|
\
|
+
|
.
|
\
|
+
=
2
2 2
2
2
2 da
C d P
da
dC
C C
P
M
4. For an ideally brittle solid, d
R
/da (change
in resistance to static fracture with crack
growth) is equal to zero. Thus, it is seen that
crack advance is stable if 0 and
unstable if 0.
A
(
c
c
a
T
A
(
c
c
a
41
unstable if 0.
T
A
(
c a
Fatigue Life Calculation
Stress-Life
A circular cylindrical rod with a uniform cross-
sectional area of 20 cm
2
is subjected to a mean axial
force of 120 kN. The fatigue strength of the material,
a
=
f s
is 250 MPa after 10
6
cycles of fully reversed
loading and
TS
= 500 MPa. Using the different
procedures discussed in class, estimate the allowable
42
procedures discussed in class, estimate the allowable
amplitude of force for which the shaft should be
designed to withstand at least one million fatigue
cycles. State all your assumptions clearly.
Solution
The different expressions we have to assess the
influence of mean stresses are :
( ) Soderberg
y
m
a a
m
= =
1 0 |
( ) Goodman Modified
TS
m
a a
m
)
`
= =
1 0 |
| |
2
43
( ) Gerber
TS
m
a a
m
|
|
.
|
\
|
= =
2
1 0 |
In all cases
a
=
a
|
m = 0
=250 MPa and
m
=
(120,000 N/.0020 m
2
) = 60 MPa. The Modified
Goodman and Gerber criteria can be applied
directly to give:
220
500
60
1 250 =
)
`
=
a
|
.
|
\
|
=
a
MPa, (Gerber)
Applying the Soderberg criterion requires a bit more
44
Applying the Soderberg criterion requires a bit more
thought since it includes the yield strength (which is
not given) rather than the tensile strength. Depending
on the details of the material behavior, the yield stress
YS
could be the same as the tensile strength
TS
(for very brittle materials) or as low as 0.5
TS
(for
very ductile materials). I will assume that
YS
= 0.7
TS
= 350 MPa. Thus the Soderberg
criterion gives:
1 . 207
350
60
1 250 =
)
`
=
a
MPa, (Soderberg)
45
As you can see, you get significantly different
answers depending on the model used. The
Soderberg gives the most conservative result,
while Gerber is the least conservative.