4 - Fatigue Chapter 4 v. 2014.03 - Lecture Notes

Muzio M.
Gola
Costruzione di macchine - Machine Design
Fatigue - Chapter 4
a
ol
Chapters
.g
1 History and problem overview
2 Stress-life: material properties
m
3 Stress-life: component - infinite life
4 Stress-life: finite life io
5 Strain-life
6 Crack propagation and fracture
uz
m
 Pierluigi Lenoci
These slides are a support to the Machine Design, and related, classes held by prof. M.M. Gola at POLITO, Technical University of Torino – For educational purposes and classroom use only
Chapter 2 - Stress-life: finite life
m.gola
2011
2014
Finite life
1. Linear Damage or Miner's Rule
a
2. Cycle counting
ol
3. Finite life and stress spectra
4. Finite life and the exceedance diagram
.g
5. Appendix: the Haibach extension
m
io
uz
m
Version Muzio M. Gola - DIMEAS 2

March 2014 Politecnico di Torino - Italia
Sections 1, 2 - Finite life
m.gola
2011
2012
These sections refer to one-dimensional tensile stresses (mainly

tensile, but compressive stress could be included as well without
difficulty). They deal with variable stresses that sooner or later will
a
produce a failure, i.e. life is finite.
ol
Section 1 defines the Miner’s rule for cumulative damage at multiple
.g
stress levels applied during the component life. A numerical example
is worked out for better clarity.
m
First the concept of “equivalent stress amplitude” is introduced.
Then it is shown how the simplified interpolation formulas
io
introduced in Ch. 2 Sect. 8 can extend the power of this concept.
uz
Section 2 shows one “cycle counting” method (the bath-tub version
of the rain-flow), and its practical application when a stress
m
spectrum is to be obtained from a time signal.

1. Linear Damage or Miner's Rule (1/11)
m.gola
2011
2008
The “linear damage rule” was first proposed by Palmgren* in 1924

and was further developed by Miner** in 1945. Today the method is
commonly known as Miner's Rule.
a
The linear damage rule is based on the concept of cumulative
ol
damage.
.g
* A. Palmgren, “Die Dauerhaftigkeit der Kugellager,” Z. Ver. Deutsch. Ing.,
68 (1924)
** M. A. Miner, “Cumulative damage in fatigue,” J. Appl. Mech., 12 (1945)
m
Read also: io
M. A. Miner, “Estimation of fatigue life with Particular Emphasis on
uz
Cumulative Damage , ch. 12 in ” Metal Fatigue”, G. Sines and J.L.
Waisman editors, Mc Graw Hill, USA, (1959)
m

m.gola
2011
2008
The cumulative damage concept requires a “damage fraction” wi as

the fraction of the total damage W necessary to produce a crack.
Failure is predicted to occur when
a
the sum of the wi ’s reaches the limit W: ∑ wi = W
ol
i
i : index of each set of applied load cycles at constant stress
.g
level σi
wi : damage fraction accumulated during the load cycles of
m
interval i at constant stress level σi
W : total damage constant
io
uz
The linear damage rule states that the damage fraction at a given
constant stress level σi is equal to a mathematical fraction:
m
the number ni of applied cycles at stress level σi

n
divided by the fatigue life Ni at stress level σi , i.e. : w i = i
Ni
m.gola
2011
2008
The most natural assumption seems to be:

ni
∑N = 1
a
i i
ol
When there is only one level σi , i=1, final failure occurs for ni = Ni
(of course...what else?): then, according to Miner, the damage
.g
constant W is assumed to be equal to 1.
m
The results of Miner's original tests showed that the damage
constant W corresponding to failure ranged from 0.61 to 1.45,
io
other researchers have shown variations as large as 0.18 to 23.0,
uz
with most results tending to fall between 0.5 and 2.0,
it seems to be true that the average value is close to Miner's
m
proposed value of 1; however, the dispersion is rather discouraging.

m.gola
2011
2008
By far the largest effort in this respect, however, was spent by Kotte
and Eulitz of the Technical University of Dresden … who collected
hundreds of test programmes, i.e. many thousands of variable-
amplitude and corresponding constant-amplitude tests and carried
a
out new Miner calculations.
ol
One outstanding result of all these checks was that Miner's rule
.g
generally proved wildly unconservative; that is, the specimens -- in
the case of the IABG programme, the components -- failed long
m
before their predicted lives; factors of 10 and more on the
unconservative side were not uncommon.
io
from: Walter Schütz, A History of Fatigue, Engineering Fracture
Mechanics Vol. 54, No. 2, pp. 263-300, 1996
uz
Despite these limitations, the linear damage rule is still widely

m
used. However, for conservative estimates of the life of a structure a

damage constant less than 1 prescribed.
m.gola
2011
2008
Data on the right show a simple s a,i ni Ni

case for σm=0 and eight blocks 1 61 1,0E+04 4,8E+04
of stresses. 2 58 1,0E+04 7,5E+04
3 55 1,5E+04 1,3E+05
a
4 50 1,7E+05 3,5E+05
We shall employ non dimensional
ol
5 43 2,0E+05
values sa=σa/Rm*100 for stresses. 6 38 1,6E+06
.g
7 29 8,0E+06
100
sa 8 25 2,0E+07
90
m
80
70
60
io sa,3 The cycles to
50 be taken into
account are
uz
40
30
20
only the blue
ones, i=1 to 4
m
10
n3 N3 Log N
0 for which Ni
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08
exists.
m.gola
2011
2008
Then the damage fractions wi s a,i ni Ni n i/ N i

can be calculated and summed. 1 61 1,0E+04 4,8E+04 0,21
2 58 1,0E+04 7,5E+04 0,13
In this case the sum is 0,93. 3 55 1,5E+04 1,3E+05 0,12
a
4 50 1,7E+05 3,5E+05 0,47
Then no failure should occur.
ol
Σ=0,93
.g
100
sa
90
m
80
70
60
50
io
uz
40
30
20
m
10
Log N
0
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08

m.gola
2011
2012
On the basis on the total number s a,i ni αi

of valid cycles, nTOT, the utilisation 1 61 1,0E+04 0,049
fraction αi is now introduced: 2 58 1,0E+04 0,049
3 55 1,5E+04 0,073
a
ni 4 50 1,7E+05 0,829
αi =
ol
n TOT nTOT= 2,05E+05 Σ=1,000
.g
100
ssaa,i
90
m
80
The aim is to
70
60
50
io obtain an
“equivalent stress
uz
40
30
amplitude” acting
for the total
20
m
10
Log N
0 number of cycles
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08
nTOT.
m.gola
2011
2012
The “equivalent stress amplitude” σ a,eq when applied for the total
number of nTOT (valid) cycles would produce the same damage of all
the single stress σ a levels applied each for its own n cycle number.
a
k
n ≤ σDk −1 106 is here written in the
Wöhler equation σ a N
ol
alternative form s ka N k
n ≤ sD −1 10
6
with: sa=σa/Rm*100 .
.g
100
sa σ a,eq The developments
90
N in the example of
m
80
next slide are valid
70
io only for the
60
simplest case,
50
where σm is the
uz
40
same for all σa
30
20 (in the example
m
10 σm=0; it could be
0 σm=any but equal
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08 for all )

m.gola
2011
2008
k
In cases when σm is the same for all σa,i , 6 s 
formula (6 for the “equivalent stress” Ni = 10  D − 1  (1

 s a ,i 
s a,eq holds; it is the stress amplitude
ni
which would produce the same damage ∑ N ≡ ∑ wi ≤ 1 (2
a
if applied for all nTOT cycles, as stated in i i i
ol
formula (5 written for the case σm=0 ∑ ni = n TOT (3
.g
i
100
sa σ a,eq = 0,52 ⋅ R m
ni = αi n TOT (4
N = 2,19 E + 5
90
m
80
70 Put 1) and 4) in 2):
60
50
io s ka,eq n TOT ≤ sDk −1 106 (5
uz
40
30
with: 1/k
20
 k 
m
10 s a,eq =  ∑ αi s a,i 
 (6
0  i 
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08

m.gola
2011
2012
When, instead, each σa,i or sa,i has a different σm,i or sm,i , then:
[s a ,i (s m ,i )] Ni = [s D (s m ,i )]
k k
106 (1
a
[ (sm,i )] 106 (1'
ol
k
s
ni N = D
∑ N ≡ ∑ w i ≤ 1 (2 [s a ,i (s m ,i )]
.g
i k
i i i
m
∑ ni = n TOT (3 In this case the formulation of an
i io equivalent stress amplitude
ni = αi n TOT (4 acting for the total number of
cycles nTOT is not possible
uz
1’) and 4) in 2): because a term s ka,eq in:
[ (sm,i )]
k
m
s 1 s ka,eq n TOT ≤ const

n TOT ∑α a ,i
= 1 (5
i[s i
D (s m ,i )]
k
10 6
cannot be isolated.
m.gola
2011
2012
However, an equivalent stress amplitude

sa
100
can be found under certain assumptions. 90

~
sa
If Haigh master curves are obtained 80
according to an interpolation done as in

70
sD −1
a
60
Sect. 8 of Ch. 2, the formula of sl. 9 of N
that Section is now written:

50
sa 10^6
ol
40
~
30 sD
sa s (s )
.g
= a m which for the level i : 20
sD −1 sD (s m ) 10
sm
~ s (s )
0
sa,i 0 10 20
sm
30 40 50 60 70 80 90 100
m
= a,i m,i (6
sD −1 sD (s m,i )
io ~
sak,i 1
Eq. (6 put in eq. (5 gives: n TOT ∑ αi k 6
= 1 (7
sD −1 10
uz
i
Now the denominator is no longer dependent on sm,i then:
α ~
m
n TOT ∑ s k = s k 106 (8
i a,i D −1
i
~
sak,eq
2. Cycle counting (1/8)
m.gola
2011
2008
The number of cycles at each stress amplitude needs to be known,

or “counted”. The constant amplitude blocks used by whatever
cumulative damage theory are extracted from a non-uniform time
history, as it will be shown in this section.
a
ol
50
40 σ2 The figure to the left
.g
30 σ3 illustrates a very simple
σ4 case of a “block”
20
σ1 σ5
m
10 spectrum, which is of
stress (ksi)
0
io course a gross
-10 simplification of reality
-20 for educational
uz
-30 purposes.
-40
m
number of cycles
from: http://www.engrasp.com/doc/etb/mod/fm1/miner/miner_help.html

m.gola
2011
2008
The figure below shows one possible case of (experimentally

detected) time history of stresses. It is necessary to find a method
to extract cycles which must be used for damage calculations.
a
ol
60
50
.g
40
30
stress (ksi)
20
m
10
0
-10
-20
io
uz
-30
-40
-50
m
0 5 10 15 20 25 t (s)
from: http://www.engrasp.com/doc/etb/mod/fm1/miner/miner_help.html

m.gola
2011
2008
We describe here the “rainflow” counting method in the so called

“bathtub” version. They extract constant amplitude cycles from a
non-uniform time history.
a
A section of the time history must be isolated so that it begins and
ends at the maximum value, points • .
ol
300 • •
.g
250 σ (MPa)
200
m
150
100 io
50
uz
0
-50
m
-100
-150
t
-200

m.gola
2011
2008
The diagram is interpreted as if it were a basin full of liquid.

The bottom point 1 is where liquid is spilled, and the basin partially
emptied…
a
ol
300
.g
σσ(MPa)
(MPa)
250
200
m
150
100 io
50
uz
0
-50
m
-100
-150
tt
-200 •1
m.gola
2011
2008
… the total level descent at point 1 (from +300 to –200 MPa , i.e.
σa=250, σm=50 MPa) is the cycle n. 1 detected.
This leaves a number of secondary basins to empty by spilling at the
a
bottom point of each: 2 through 12 .
ol
300
.g
250 σB (MPa)
200
m
150
100 io
50
9
uz
0
2 4 8 12
-50
10
m
-100
5 7 11
-150
3 6 tt
-200

m.gola
2011
2008
… further levels are detected, 13 through 23 …
a
ol
300
.g
250 σB (MPa)
200
m
150
100 io 18
50
19 21
uz
0
17 20 22 23
-50
13
m
-100
14 15
-150
16 tt
-200

m.gola
2011
2008
… and the last cycles 24 to 27.
a
ol
300
.g
250 σB (MPa)
200
m
150
100 io
50
27
uz
0
-50
24 25 26
m
-100
-150
tt
-200

m.gola
2011
2014
The table below summarises the results:

stresses (MPa)
σmin σmax ∆σ σa σm n σmin σmax ∆σ σa σm
n
1 -200 300 500 250 50 22 0 100 100 50 50
a
2 0 200 200 100 100 23 0 200 200 100 100
24 -50 50 100 50 0
ol
3 -150 200 350 175 25
25 -50 0 50 25 -25
4 0 100 100 50 50
26 -50 50 100 50 0
5 -100 100 200 100 0
.g
27 50 150 100 50 100
6 -150 100 250 125 -25
7 -100 100 200 100 0 These cycle counts can be grouped in
8 0 150 150 75 75 blocks. See colours: in this example
m
9 50 200 150 75 125
10 -50 250 300 150 100 they are not numerous, but in the real
11 -100 250 350 175 75 cases and with digit rounding they may
12 100 250 150 75
io 175 number up to thousands. One group is
13 -50 100 150 75 25
14 -100 100 200 100 0 missing: which one?
uz
15 -100 50 150 75 -25
16 -150 100 250 125 -25
To each block a “total occurrence”
17 0 50 50 25 25 number will be associated:
m
18 100 150 50 25 125

19 50 200 150 75 125 α=2/27 α=2/27
20 0 150 150 75 75 α=3/27 α=2/27
21 50 150 100 50 100
α=2/27 α=2/27 others α=1/27

Sections 3, 4 – Applications of finite life concepts
m.gola
2011
2012
Also these sections refer to one-dimensional tensile (mainly tensile,

but compressive can be included) stresses,
Section 3 shows how a stress spectrum can be used to assess finite
a
life at fatigue, deepening the tool presented in Section 1.
ol
Section 4 introduces advanced matters on the exceedance diagram.
.g
m
io
uz
m

3. Finite life and stress spectra (1/9)
m.gola
2011
2008
As an application of ideas developed so far, the

one shown on the right is a very simplified sm sa ni
example of spectrum of stresses (expressed as 0 50 50
1,1E+05
a percent of Rm) applied at a design point. 15 47 55
a
1,5E+04
18 50 61
ol
6,7E+03
We shall not use the results from the cycle 40 35 58
6,7E+03
counting of the preceding section, in order to
.g
30 30 43
2,3E+05
have a smaller and more treatable, yet
20 32 1,2E+06
representative of the concept, number of
m
0 25 25
1,3E+07
points.
io 30 20 29
5,3E+07
We shall also make use of the Wöhler curves in

sx = σx / Rm
the extended Haibach form explained in Sect. 5,
uz
Appendix to this chapter, so to become familiar
with it.
m

m.gola
2011
2008
80
sm sa ni
70
0 50 50
1,1E+05
a
60
15 47 55
1,5E+04
ol
50 10^4 18 50 61
6,7E+03
10^5
.g
40 35 58
6,7E+03
40 10^6
10^7 30 30 43
2,3E+05
30 10^8
m
20 32 1,2E+06
20
0 25 25
1,3E+07
10
io 30 20 29
5,3E+07
uz
0
0 10 20 30 40 50 60 70 80 90 100 sx = σx / Rm
m
Each (sm ; sa) point can be represented on the Haigh diagram.

m.gola
2011
2008
80 ni
70 0 50 50
15 47 55
60
0 18 50 61
15 50 40 35 58
18 30 30 43
40
20
30 30 20 32
40 0 25 25
20
30 20 29
10
0
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08
Wöhler curves are traced for each sm, and ni’s represented for
each sa . Mind the colour code coupling each sa to its sm
m.gola
2011
2008
1,3 E+5
80 sm sa ni
sa = σa / Rm
70 0 50 50
1,1E+05
15 47 55
1,5E+04
60
18 50 61
6,7E+03
sm 50 40 35 58
6,7E+03
15 40 30 30 43
2,3E+05
30 20 32 1,2E+06
0 25 25
1,3E+07
20
30 20 29
5,3E+07
10
0
1,E+03
ni,15
1,E+04 Ni,15
1,E+05 1,E+06 1,E+07 1,E+08
Then the “damage fraction” wi for each level can be calculated

(here shown for wi,15 , i.e. wi,15=ni,15/Ni,15= 1,5E+4 /1,29E+5 )
m.gola
2011
2008
The foregoing procedure is quite clear in

principle, but has the disadvantage that Wöhler sm sa ni
curves for each sm must be calculated. Then, 0 50 50
1,1E+05
stress levels sa must be coupled with the 15 47 55

1,5E+04
corresponding curve having the same sm. 18 50 61

6,7E+03
40 35 58
6,7E+03
This is not terribly inconvenient, but there is a
30 30 43
better way, in that it simplifies the graphics, and 2,3E+05
makes use of the only Wöhler curve for the life 20 32 1,2E+06
106 cycles. 0 25 25
1,3E+07
30 20 29
5,3E+07
Let us take for instance point sm=18, sa=50 … …

m.gola
2011
2008
sm sa ~
sa
80
sa ~
sa sa
70 0 50 50
sD −1 sD 15 47 55
60 *
18 50 61
61
50 10^4
10^5 40 35 58
40 10^6 30 30 43
10^7
30 20 30 38
10^8
0 25 25
20
30 20 29
10
0 sm sx = σx / Rm
0 10 20 30 40 50 60 70 80 90 100
~ 6
As shown in sl.7-Sect.8-Ch.2, s a / sD = sa / sD −1 = 10 / N
1/k
( )
so for the same life N we can use ~ sa at sm=0 instead of s a .
m.gola
2011
2008
80 sm sa ~
sa *
sa
70 0 50 50
15 47 55
60 *
*
* 18 50 61
50 10^4
10^5 40 35 58
40 *
* 10^6 30 30 43
10^7
30 20 32 40
* 10^8
0 25 25
20
30 20 29
10
0 sm sx = σx / Rm
0 10 20 30 40 50 60 70 80 90 100
So all points • can be transformed into equal life * points at sm=0.

Let us now remark that for life calculation s a / sD ≡ σ a / σD , then
Haigh and Wöhler diagrams in non-dimensional form are useful.
m.gola
2011
2008
80 s m* ~
sa * ni
sa
70 0 50 50
1,1E+05
0 55 55
1,5E+04
60 *
*
*
50 * 10^4
0 61 61
6,7E+03
10^5 0 58 58
6,7E+03
40 *
* 10^6 0 43 43
2,3E+05
10^7
30 0 40 40
1,2E+06
* 10^8
20
* 0 25 25
1,3E+07
0 29 29
5,3E+07
10
0 sm sx = σx / Rm
0 10 20 30 40 50 60 70 80 90 100
All points to consider are now those listed and marked with *

m.gola
2011
2008
sm ~
sa ni
80
~
sa
70 0 50 50
1,1E+05
0 55 55
1,5E+04
60
0 61 61
6,7E+03
~
sa
sm 50 0 58 58
6,7E+03
0 40 0 43 43
2,3E+05
0 40 40
1,2E+06
30
0 25 25
1,3E+07
20 0 29 29
5,3E+07
10
0
1,E+03 1,E+04 ni,50
1,E+05 Ni,50
1,E+06 1,E+07 1,E+08
Now all “damage fractions” wi will be calculated as ni/Ni , where

the only Woehler curve used is the one for σm=0 .
4. Finite life and the exceedance diagram (1/12)
m.gola
2011
2008
80
~
s a,i ni Σ ni
~
sa 61 2,0E+04
1,4E+04 2,0E+04
1,4E+04
70 58 2,0E+04
1,4E+04 4,0E+04
2,8E+04
0
55 3,0E+04
2,1E+04 7,0E+04
4,9E+04
60
50 2,3E+05
3,3E+05 2,8E+05
4,0E+05
43 2,8E+05
4,0E+05 5,6E+05
8,0E+05
50
38 2,2E+06
3,2E+06 2,8E+06
4,0E+06
40 29 1,1E+07
1,6E+07 1,4E+07
2,0E+07
25 2,8E+07
4,0E+07 4,2E+07
6,0E+07
30
3,2E+06
3,3E+05
2,0E+04
3,0E+04
2,0E+04
4,0E+05
1,6E+07
4,0E+07
20
10
Log N
0
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08
Each value ~ sa,i is represented with its life ni in the “exceedance

diagram”, which comes from the cycle count of a typical time
history of stresses. Stresses are re-ordered in descending order,
and are represented against the total cycle number Log(∑ ni ) .
m.gola
2011
2008
80
~
sa
70
0
•
60
50
40
30
3,2E+06
3,3E+05
2,0E+04
3,0E+04
2,0E+04
4,0E+05
1,6E+07
4,0E+07
20
10
6,0E+07
7,0E+04
8,0E+05
2,0E+04
4,0E+06
2,0E+07
4,0E+04
4,0E+05
0
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08
The exceedance diagram is called so because it shows for how

many cycles n the stress amplitude ~ sa is exceeded.
The block-diagram version is a simplification of a continuos
curve, typical of a service condition (dotted grey line).
m.gola
2011
2008
80
~
s a,2 : value of stress at level 2
0 ~
sa n2 : number of cycles spent al level 2
N2 : max life at ~
70
s a,2 for σm=0
~
s a,2 60 o o o Log(n1 + n2 ): total cycles for levels 1+2
50
s D −1
40
s II
30
2,0E+04
2,0E+04
20
10
Log N
0 o o
4,0E+04
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08
This semi-log representation of the exceedance diagram can be

superimposed to the σm=0 Wöhler curve, here in the Haibach
form (see Appendix). Remark the meaning of points for the
example of level i=2, which can easily extended to any level i.
m.gola
2011
2008
80
0 ~
sa
70
2,0E+04
60 2,0E+04
3,0E+04
50 3,3E+05
s D −1 4,0E+05
40
s II 3,2E+06
30 1,6E+07
4,0E+07
2,0E+04
20 2,0E+04
10
Log N
n
0 o
4,0E+04
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08
It is perhaps a better and clearer idea to extract from the

exceedance diagram the bars representing the ni , i.e. number
of applied cycles – at each level.
m.gola
2011
2008
From: k 0
80
~
sa
6 
 sD −1 
70
Ni = 10  ~  ••
 sa,i  60
•
we can ssDD−1−1
50 •
calculate Ni •
ssIIII
40
•
and then the 30
2,0E+04
2,0E+04
ratio ni/Ni i.e.
20
the damage
10
0 Log
index. 1,E+03 1,E+04 4,0E+04
1,E+05 1,E+06 1,E+07 n
1,E+08
~
s a,i ni Σ ni Ni ni/Ni
It is easily seen that it is absolutely 61 2,0E+04 2,0E+04 4,8E+04 0,4173
worthless to include cycles with ~

58 2,0E+04 4,0E+04 7,5E+04 0,2680
sa 55 3,0E+04 7,0E+04 1,3E+05 0,2354
lower than s II (orange values); 50

43
3,3E+05
4,0E+05
4,0E+05
8,0E+05
3,5E+05
2,5E+06
0,9464
0,1583
it is then customary to truncate the 38 3,2E+06 4,0E+06 3,2E+07 0,1002
exceedance diagram excluding all 29 1,6E+07 2,0E+07 5,6E+09 0,0029
values ~
25 4,0E+07 6,0E+07 7,1E+10 0,0006
sa< s II.
m.gola
2011
2008
Both in numerical 80
~
sa
simulations and in 70
0
laboratory tests the 60
exceedance diagram
is truncated by
50
eliminating the 40
large number of 30 truncation

small cycles which
3,2E+06
3,3E+05
2,0E+04
3,0E+04
2,0E+04
4,0E+05
1,6E+07
4,0E+07
20
do not contribute to 10
damage.
0
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08
It takes as much computer time (and testing time) to deal with one small cycle as
with one large cycle. Thus the small cycles at the lower end of the exceedance
diagram consume most of the time (cost) while their effect may be very small.
Therefore, it would be advantageous if their number could be reduced. This is
called truncation.
Quoted from: David Broek, The practical use of fracture mechanics, Kluwer
Academic Publishers, 1989 (a warmly suggested reading!)
m.gola
2011
2008
0
80 The total
70
number of
•• effective cycles
•
60
s D −150 • of this load

• history is then:
s II40 •
30
nTOT=4x106.
20
10
0
1,E+03 1,E+04 1,E+05 1,E+06 1,E+07 1,E+08
The utilisation factor at ~

s a,i ni Σ ni Ni ni/Ni αi
level i is αi : 61 2,0E+04 2,0E+04 4,8E+04 0,4173 0,005
ni 58 2,0E+04 4,0E+04 7,5E+04 0,2680 0,005

αi = 55 3,0E+04 7,0E+04 1,3E+05 0,2354 0,0075
n TOT 50 3,3E+05 4,0E+05 3,5E+05 0,9464 0,0825
i.e. the ratio between applied 43 4,0E+05 8,0E+05 2,5E+06 0,1583 0,1
38 3,2E+06 4,0E+06 3,2E+07 0,1002 0,8
cycles ni and nTOT.
nTOT = total cycle number

m.gola
2011
2008
80 80
nTOT=4*106 nTOT=1*106
70 70
60 60
50 50
0,0825
0,0075
0,0825
0,0075
40 40
0,005
0,005
0,005
α1 = 0,005
0,8
0,1
α6 = 0,8
0,1
30 30
20 20
10 10
0 0
1,0E+03 1,0E+04 1,0E+05 1,0E+06 1,0E+07 1,0E+03 1,0E+04 1,0E+05 1,0E+06 1,0E+07
The concept of utilisation factor αi is useful because the total

cycle number nTOT may change but, as the type of service of the
mechanical part repeats in time over and over in the same way,
the distribution of the values of αi remains the same.
m.gola
2011
2008
max stress
principal stresses MPa
min stress
cumulated cycles
A practical example: the diagram above shows the load

spectra for stresses at the inner radius of a propeller shaft due
to an hydraulic device for the variation of propeller pitch-
angle during flight.
m.gola
2011
2008
~
s a,i ni αi
Σ ni Ni ni/Ni
1 61 2,0E+04 2,0E+04 4,8E+04 0,417 0,005
2 58 2,0E+04 4,0E+04 7,5E+04 0,268 0,005
3 55 3,0E+04 7,0E+04 1,3E+05 0,235 0,0075
4 50 3,3E+05 4,0E+05 3,5E+05 0,946 0,0825
5 43 4,0E+05 8,0E+05 2,5E+06 0,158 0,1
6 38 3,2E+06 4,0E+06 3,2E+07 0,100 0,8
In the case shown here the sum ∑ ni / Ni 2,13
i
is much larger than 1.
n
Since all lives ni in the sum of damage fractions: ∑ i
i Ni
appear in linear form, to get a desired damage sum W it will be
sufficient to multiply all of them by the factor φ=W/2,13 .

m.gola
2011
2008
The direct calculation of total cycles n̂ TOT can be done as follows:

k Wöhler 2k −1 Haibach
 s D −1   s D −1 
Ni = 106  ~  (1a curve: Ni = 106  ~  (1b extension:
 sa,i  i=1,j  sa,i  i=j,n
n
The sum of damage fractions: ∑ i ≤ W (2
i Ni
with the “damage criterion” W usually equal to 1.
Moreover is: ∑ ni = n TOT (3 and: ni = αi n TOT (4
i
With (1a and (1b into (2 , and then (3 and (4 , we get:
W 106
n̂ TOT ≤ ≡W
αi ~ k ~ 2k −1
∑ Ni  sa,i   sa,i 
∑ αi  sD−1  + ∑ αi  sD−1 
 
i
i=1, j   i= j,n  

m.gola
2011
2008
~
s a,i ni αi
Σ ni Ni ni/Ni
1 61 9,4E+03
2,0E+04 9,4E+03
2,0E+04 4,8E+04 0,196
0,417 0,005
2 58 9,4E+03
2,0E+04 1,9E+04
4,0E+04 7,5E+04 0,126
0,268 0,005
3 55 1,4E+04
3,0E+04 7,0E+04
3,3E+04 1,3E+05 0,111
0,235 0,0075
4 50 3,3E+05
1,6E+05 4,0E+05
1,9E+05 3,5E+05 0,946
0,445 0,0825
5 43 4,0E+05
1,9E+05 8,0E+05
3,8E+05 2,5E+06 0,158
0,074 0,1
6 38 3,2E+06
1,5E+06 1,9E+06 3,2E+07 0,100
0,047 0,8
∑ ni / Ni 1,00
i
In the example shown “total damage” 1 is found with the value

φ=0,47 . The table above shows the modified values.

5. Appendix: the Haibach extension (1/2)
m.gola
2011
2008
Experimental evidence has shown that even stresses below the

fatigue limit (defined through experiments with single amplitude)
can contribute to damage provided there are also some stress
amplitudes which exceed the fatigue limit.
It is believed that the periodic overloads help overcome the grain
barrier which does not let crack propagate at stresses below the
fatigue limit, and cause them propagate the crack.
In order to allow for this, the original Miner’s rule was improved by
extending the fatigue curve beyond the fatigue limit life; following
Haibach, the fatigue curve of exponent k is extended below the
fatigue limit and towards zero stress with an exponent 2k-1 , as in
the example shown in next slide, ...

5. Appendix: the Haibach extension (2/2)
m.gola
2011
2008
… the fatigue curve of exponent k is extended below the

fatigue limit and towards zero stress with an exponent 2k-1 , as
in the example shown here below (case k=10).
80 s ka Ni = 106 sDk −1
70
0
60 s 2ak −1 Ni = 106 sD2k−−11
10 50
20
40
30
40 30
50
20
10
0
1,E+04 1,E+05 1,E+06 1,E+07 1,E+08


4 - Fatigue Chapter 4 v. 2014.03 - Lecture Notes

Uploaded by

Document Information

Original Title

Copyright

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

4 - Fatigue Chapter 4 v. 2014.03 - Lecture Notes

Uploaded by

Copyright:

Muzio M.

Version Muzio M. Gola - DIMEAS 2

These sections refer to one-dimensional tensile stresses (mainly

spectrum is to be obtained from a time signal.

Version Muzio M. Gola - DIMEAS 3

The “linear damage rule” was first proposed by Palmgren* in 1924

Version Muzio M. Gola - DIMEAS 4

The cumulative damage concept requires a “damage fraction” wi as

Failure is predicted to occur when

the number ni of applied cycles at stress level σi

The most natural assumption seems to be:

proposed value of 1; however, the dispersion is rather discouraging.

Version Muzio M. Gola - DIMEAS 6

Despite these limitations, the linear damage rule is still widely

used. However, for conservative estimates of the life of a structure a

Data on the right show a simple s a,i ni Ni

Then the damage fractions wi s a,i ni Ni n i/ N i

Version Muzio M. Gola - DIMEAS 9

On the basis on the total number s a,i ni αi

Version Muzio M. Gola - DIMEAS 11

Version Muzio M. Gola - DIMEAS 12

s 1 s ka,eq n TOT ≤ const

However, an equivalent stress amplitude

can be found under certain assumptions. 90

according to an interpolation done as in

that Section is now written:

The number of cycles at each stress amplitude needs to be known,

40 σ2 The figure to the left

Version Muzio M. Gola - DIMEAS 15

The figure below shows one possible case of (experimentally

Version Muzio M. Gola - DIMEAS 16

We describe here the “rainflow” counting method in the so called

Version Muzio M. Gola - DIMEAS 17

The diagram is interpreted as if it were a basin full of liquid.

Version Muzio M. Gola - DIMEAS 19

… further levels are detected, 13 through 23 …

Version Muzio M. Gola - DIMEAS 20

… and the last cycles 24 to 27.

Version Muzio M. Gola - DIMEAS 21

The table below summarises the results:

18 100 150 50 25 125

Version Muzio M. Gola - DIMEAS 22

Also these sections refer to one-dimensional tensile (mainly tensile,

Version Muzio M. Gola - DIMEAS 23

As an application of ideas developed so far, the

a percent of Rm) applied at a design point. 15 47 55

counting of the preceding section, in order to

We shall also make use of the Wöhler curves in

Version Muzio M. Gola - DIMEAS 24

Each (sm ; sa) point can be represented on the Haigh diagram.

Version Muzio M. Gola - DIMEAS 25

Then the “damage fraction” wi for each level can be calculated

The foregoing procedure is quite clear in

stress levels sa must be coupled with the 15 47 55

corresponding curve having the same sm. 18 50 61

Let us take for instance point sm=18, sa=50 … …

Version Muzio M. Gola - DIMEAS 28

So all points • can be transformed into equal life * points at sm=0.

Version Muzio M. Gola - DIMEAS 31

Now all “damage fractions” wi will be calculated as ni/Ni , where

Each value ~ sa,i is represented with its life ni in the “exceedance

The exceedance diagram is called so because it shows for how

This semi-log representation of the exceedance diagram can be

It is perhaps a better and clearer idea to extract from the

worthless to include cycles with ~

lower than s II (orange values); 50