1
ME 322 - Mechanical Design 1
Partial notes ndash Part 4 (Fatigue)
2
Ni =
Np =
NT =
Metal Fatigue is a process which causes premature irreversible damage or
failure of a component subjected to repeated loading
FATIGUE - What is it
3
Metallic Fatigue
bull A sequence of several very complex phenomena encompassing several disciplines
ndash motion of dislocations
ndash surface phenomena
ndash fracture mechanics
ndash stress analysis
ndash probability and statistics
bull Begins as an consequence of reversed plastic deformation within a single crystallite but ultimately may cause the destruction of the entire component
bull Influenced by a componentrsquos environment
bull Takes many forms
ndash fatigue at notches
ndash rolling contact fatigue
ndash fretting fatigue
ndash corrosion fatigue
ndash creep-fatigue
Fatigue is not cause of failure per se but leads to the final fracture event
4
The Broad Field of Fracture Mechanics
5
Intrusions and Extrusions The Early Stages of Fatigue Crack Formation
6
Locally the crack grows in shear
macroscopically it grows in tension
Δσ
Δσ
c
Schematic of Fatigue Crack Initiation Subsequent Growth
Corresponding and Transition From Mode II to Mode I
7
The Process of Fatigue
The Materials Science Perspective bull Cyclic slip
bull Fatigue crack initiation
bull Stage I fatigue crack growth
bull Stage II fatigue crack growth
bull Brittle fracture or ductile rupture
8
Features of the Fatigue Fracture Surface of a Typical
Ductile Metal Subjected to Variable Amplitude Cyclic
Loading
(Collins ref 22 )
A ndash fatigue crack area
B ndash area of the final static
failure
9
Appearance of Failure Surfaces Caused by
Various Modes of Loading (SAE Handbook)
10
Factors Influencing Fatigue Life
Applied Stresses
bull Stress range ndash The basic cause of plastic deformation and consequently the accumulation of damage
bull Mean stress ndash Tensile mean and residual stresses aid to the formation and growth of fatigue cracks
bull Stress gradients ndash Bending is a more favorable loading mode than axial loading because in bending fatigue cracks propagate into the region of lower stresses
Materials
bull Tensile and yield strength ndash Higher strength materials resist plastic deformation and hence have a higher fatigue strength at long lives Most ductile materials perform better at short lives
bull Quality of material ndash Metallurgical defects such as inclusions seams internal tears and segregated elements can initiate fatigue cracks
bull Temperature ndash Temperature usually changes the yield and tensile strength resulting in the change of fatigue resistance (high temperature decreases fatigue resistance)
bull Frequency (rate of straining) ndash At high frequencies the metal component may be self-heated
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
2
Ni =
Np =
NT =
Metal Fatigue is a process which causes premature irreversible damage or
failure of a component subjected to repeated loading
FATIGUE - What is it
3
Metallic Fatigue
bull A sequence of several very complex phenomena encompassing several disciplines
ndash motion of dislocations
ndash surface phenomena
ndash fracture mechanics
ndash stress analysis
ndash probability and statistics
bull Begins as an consequence of reversed plastic deformation within a single crystallite but ultimately may cause the destruction of the entire component
bull Influenced by a componentrsquos environment
bull Takes many forms
ndash fatigue at notches
ndash rolling contact fatigue
ndash fretting fatigue
ndash corrosion fatigue
ndash creep-fatigue
Fatigue is not cause of failure per se but leads to the final fracture event
4
The Broad Field of Fracture Mechanics
5
Intrusions and Extrusions The Early Stages of Fatigue Crack Formation
6
Locally the crack grows in shear
macroscopically it grows in tension
Δσ
Δσ
c
Schematic of Fatigue Crack Initiation Subsequent Growth
Corresponding and Transition From Mode II to Mode I
7
The Process of Fatigue
The Materials Science Perspective bull Cyclic slip
bull Fatigue crack initiation
bull Stage I fatigue crack growth
bull Stage II fatigue crack growth
bull Brittle fracture or ductile rupture
8
Features of the Fatigue Fracture Surface of a Typical
Ductile Metal Subjected to Variable Amplitude Cyclic
Loading
(Collins ref 22 )
A ndash fatigue crack area
B ndash area of the final static
failure
9
Appearance of Failure Surfaces Caused by
Various Modes of Loading (SAE Handbook)
10
Factors Influencing Fatigue Life
Applied Stresses
bull Stress range ndash The basic cause of plastic deformation and consequently the accumulation of damage
bull Mean stress ndash Tensile mean and residual stresses aid to the formation and growth of fatigue cracks
bull Stress gradients ndash Bending is a more favorable loading mode than axial loading because in bending fatigue cracks propagate into the region of lower stresses
Materials
bull Tensile and yield strength ndash Higher strength materials resist plastic deformation and hence have a higher fatigue strength at long lives Most ductile materials perform better at short lives
bull Quality of material ndash Metallurgical defects such as inclusions seams internal tears and segregated elements can initiate fatigue cracks
bull Temperature ndash Temperature usually changes the yield and tensile strength resulting in the change of fatigue resistance (high temperature decreases fatigue resistance)
bull Frequency (rate of straining) ndash At high frequencies the metal component may be self-heated
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
3
Metallic Fatigue
bull A sequence of several very complex phenomena encompassing several disciplines
ndash motion of dislocations
ndash surface phenomena
ndash fracture mechanics
ndash stress analysis
ndash probability and statistics
bull Begins as an consequence of reversed plastic deformation within a single crystallite but ultimately may cause the destruction of the entire component
bull Influenced by a componentrsquos environment
bull Takes many forms
ndash fatigue at notches
ndash rolling contact fatigue
ndash fretting fatigue
ndash corrosion fatigue
ndash creep-fatigue
Fatigue is not cause of failure per se but leads to the final fracture event
4
The Broad Field of Fracture Mechanics
5
Intrusions and Extrusions The Early Stages of Fatigue Crack Formation
6
Locally the crack grows in shear
macroscopically it grows in tension
Δσ
Δσ
c
Schematic of Fatigue Crack Initiation Subsequent Growth
Corresponding and Transition From Mode II to Mode I
7
The Process of Fatigue
The Materials Science Perspective bull Cyclic slip
bull Fatigue crack initiation
bull Stage I fatigue crack growth
bull Stage II fatigue crack growth
bull Brittle fracture or ductile rupture
8
Features of the Fatigue Fracture Surface of a Typical
Ductile Metal Subjected to Variable Amplitude Cyclic
Loading
(Collins ref 22 )
A ndash fatigue crack area
B ndash area of the final static
failure
9
Appearance of Failure Surfaces Caused by
Various Modes of Loading (SAE Handbook)
10
Factors Influencing Fatigue Life
Applied Stresses
bull Stress range ndash The basic cause of plastic deformation and consequently the accumulation of damage
bull Mean stress ndash Tensile mean and residual stresses aid to the formation and growth of fatigue cracks
bull Stress gradients ndash Bending is a more favorable loading mode than axial loading because in bending fatigue cracks propagate into the region of lower stresses
Materials
bull Tensile and yield strength ndash Higher strength materials resist plastic deformation and hence have a higher fatigue strength at long lives Most ductile materials perform better at short lives
bull Quality of material ndash Metallurgical defects such as inclusions seams internal tears and segregated elements can initiate fatigue cracks
bull Temperature ndash Temperature usually changes the yield and tensile strength resulting in the change of fatigue resistance (high temperature decreases fatigue resistance)
bull Frequency (rate of straining) ndash At high frequencies the metal component may be self-heated
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
4
The Broad Field of Fracture Mechanics
5
Intrusions and Extrusions The Early Stages of Fatigue Crack Formation
6
Locally the crack grows in shear
macroscopically it grows in tension
Δσ
Δσ
c
Schematic of Fatigue Crack Initiation Subsequent Growth
Corresponding and Transition From Mode II to Mode I
7
The Process of Fatigue
The Materials Science Perspective bull Cyclic slip
bull Fatigue crack initiation
bull Stage I fatigue crack growth
bull Stage II fatigue crack growth
bull Brittle fracture or ductile rupture
8
Features of the Fatigue Fracture Surface of a Typical
Ductile Metal Subjected to Variable Amplitude Cyclic
Loading
(Collins ref 22 )
A ndash fatigue crack area
B ndash area of the final static
failure
9
Appearance of Failure Surfaces Caused by
Various Modes of Loading (SAE Handbook)
10
Factors Influencing Fatigue Life
Applied Stresses
bull Stress range ndash The basic cause of plastic deformation and consequently the accumulation of damage
bull Mean stress ndash Tensile mean and residual stresses aid to the formation and growth of fatigue cracks
bull Stress gradients ndash Bending is a more favorable loading mode than axial loading because in bending fatigue cracks propagate into the region of lower stresses
Materials
bull Tensile and yield strength ndash Higher strength materials resist plastic deformation and hence have a higher fatigue strength at long lives Most ductile materials perform better at short lives
bull Quality of material ndash Metallurgical defects such as inclusions seams internal tears and segregated elements can initiate fatigue cracks
bull Temperature ndash Temperature usually changes the yield and tensile strength resulting in the change of fatigue resistance (high temperature decreases fatigue resistance)
bull Frequency (rate of straining) ndash At high frequencies the metal component may be self-heated
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
5
Intrusions and Extrusions The Early Stages of Fatigue Crack Formation
6
Locally the crack grows in shear
macroscopically it grows in tension
Δσ
Δσ
c
Schematic of Fatigue Crack Initiation Subsequent Growth
Corresponding and Transition From Mode II to Mode I
7
The Process of Fatigue
The Materials Science Perspective bull Cyclic slip
bull Fatigue crack initiation
bull Stage I fatigue crack growth
bull Stage II fatigue crack growth
bull Brittle fracture or ductile rupture
8
Features of the Fatigue Fracture Surface of a Typical
Ductile Metal Subjected to Variable Amplitude Cyclic
Loading
(Collins ref 22 )
A ndash fatigue crack area
B ndash area of the final static
failure
9
Appearance of Failure Surfaces Caused by
Various Modes of Loading (SAE Handbook)
10
Factors Influencing Fatigue Life
Applied Stresses
bull Stress range ndash The basic cause of plastic deformation and consequently the accumulation of damage
bull Mean stress ndash Tensile mean and residual stresses aid to the formation and growth of fatigue cracks
bull Stress gradients ndash Bending is a more favorable loading mode than axial loading because in bending fatigue cracks propagate into the region of lower stresses
Materials
bull Tensile and yield strength ndash Higher strength materials resist plastic deformation and hence have a higher fatigue strength at long lives Most ductile materials perform better at short lives
bull Quality of material ndash Metallurgical defects such as inclusions seams internal tears and segregated elements can initiate fatigue cracks
bull Temperature ndash Temperature usually changes the yield and tensile strength resulting in the change of fatigue resistance (high temperature decreases fatigue resistance)
bull Frequency (rate of straining) ndash At high frequencies the metal component may be self-heated
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
6
Locally the crack grows in shear
macroscopically it grows in tension
Δσ
Δσ
c
Schematic of Fatigue Crack Initiation Subsequent Growth
Corresponding and Transition From Mode II to Mode I
7
The Process of Fatigue
The Materials Science Perspective bull Cyclic slip
bull Fatigue crack initiation
bull Stage I fatigue crack growth
bull Stage II fatigue crack growth
bull Brittle fracture or ductile rupture
8
Features of the Fatigue Fracture Surface of a Typical
Ductile Metal Subjected to Variable Amplitude Cyclic
Loading
(Collins ref 22 )
A ndash fatigue crack area
B ndash area of the final static
failure
9
Appearance of Failure Surfaces Caused by
Various Modes of Loading (SAE Handbook)
10
Factors Influencing Fatigue Life
Applied Stresses
bull Stress range ndash The basic cause of plastic deformation and consequently the accumulation of damage
bull Mean stress ndash Tensile mean and residual stresses aid to the formation and growth of fatigue cracks
bull Stress gradients ndash Bending is a more favorable loading mode than axial loading because in bending fatigue cracks propagate into the region of lower stresses
Materials
bull Tensile and yield strength ndash Higher strength materials resist plastic deformation and hence have a higher fatigue strength at long lives Most ductile materials perform better at short lives
bull Quality of material ndash Metallurgical defects such as inclusions seams internal tears and segregated elements can initiate fatigue cracks
bull Temperature ndash Temperature usually changes the yield and tensile strength resulting in the change of fatigue resistance (high temperature decreases fatigue resistance)
bull Frequency (rate of straining) ndash At high frequencies the metal component may be self-heated
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
7
The Process of Fatigue
The Materials Science Perspective bull Cyclic slip
bull Fatigue crack initiation
bull Stage I fatigue crack growth
bull Stage II fatigue crack growth
bull Brittle fracture or ductile rupture
8
Features of the Fatigue Fracture Surface of a Typical
Ductile Metal Subjected to Variable Amplitude Cyclic
Loading
(Collins ref 22 )
A ndash fatigue crack area
B ndash area of the final static
failure
9
Appearance of Failure Surfaces Caused by
Various Modes of Loading (SAE Handbook)
10
Factors Influencing Fatigue Life
Applied Stresses
bull Stress range ndash The basic cause of plastic deformation and consequently the accumulation of damage
bull Mean stress ndash Tensile mean and residual stresses aid to the formation and growth of fatigue cracks
bull Stress gradients ndash Bending is a more favorable loading mode than axial loading because in bending fatigue cracks propagate into the region of lower stresses
Materials
bull Tensile and yield strength ndash Higher strength materials resist plastic deformation and hence have a higher fatigue strength at long lives Most ductile materials perform better at short lives
bull Quality of material ndash Metallurgical defects such as inclusions seams internal tears and segregated elements can initiate fatigue cracks
bull Temperature ndash Temperature usually changes the yield and tensile strength resulting in the change of fatigue resistance (high temperature decreases fatigue resistance)
bull Frequency (rate of straining) ndash At high frequencies the metal component may be self-heated
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
8
Features of the Fatigue Fracture Surface of a Typical
Ductile Metal Subjected to Variable Amplitude Cyclic
Loading
(Collins ref 22 )
A ndash fatigue crack area
B ndash area of the final static
failure
9
Appearance of Failure Surfaces Caused by
Various Modes of Loading (SAE Handbook)
10
Factors Influencing Fatigue Life
Applied Stresses
bull Stress range ndash The basic cause of plastic deformation and consequently the accumulation of damage
bull Mean stress ndash Tensile mean and residual stresses aid to the formation and growth of fatigue cracks
bull Stress gradients ndash Bending is a more favorable loading mode than axial loading because in bending fatigue cracks propagate into the region of lower stresses
Materials
bull Tensile and yield strength ndash Higher strength materials resist plastic deformation and hence have a higher fatigue strength at long lives Most ductile materials perform better at short lives
bull Quality of material ndash Metallurgical defects such as inclusions seams internal tears and segregated elements can initiate fatigue cracks
bull Temperature ndash Temperature usually changes the yield and tensile strength resulting in the change of fatigue resistance (high temperature decreases fatigue resistance)
bull Frequency (rate of straining) ndash At high frequencies the metal component may be self-heated
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
9
Appearance of Failure Surfaces Caused by
Various Modes of Loading (SAE Handbook)
10
Factors Influencing Fatigue Life
Applied Stresses
bull Stress range ndash The basic cause of plastic deformation and consequently the accumulation of damage
bull Mean stress ndash Tensile mean and residual stresses aid to the formation and growth of fatigue cracks
bull Stress gradients ndash Bending is a more favorable loading mode than axial loading because in bending fatigue cracks propagate into the region of lower stresses
Materials
bull Tensile and yield strength ndash Higher strength materials resist plastic deformation and hence have a higher fatigue strength at long lives Most ductile materials perform better at short lives
bull Quality of material ndash Metallurgical defects such as inclusions seams internal tears and segregated elements can initiate fatigue cracks
bull Temperature ndash Temperature usually changes the yield and tensile strength resulting in the change of fatigue resistance (high temperature decreases fatigue resistance)
bull Frequency (rate of straining) ndash At high frequencies the metal component may be self-heated
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
10
Factors Influencing Fatigue Life
Applied Stresses
bull Stress range ndash The basic cause of plastic deformation and consequently the accumulation of damage
bull Mean stress ndash Tensile mean and residual stresses aid to the formation and growth of fatigue cracks
bull Stress gradients ndash Bending is a more favorable loading mode than axial loading because in bending fatigue cracks propagate into the region of lower stresses
Materials
bull Tensile and yield strength ndash Higher strength materials resist plastic deformation and hence have a higher fatigue strength at long lives Most ductile materials perform better at short lives
bull Quality of material ndash Metallurgical defects such as inclusions seams internal tears and segregated elements can initiate fatigue cracks
bull Temperature ndash Temperature usually changes the yield and tensile strength resulting in the change of fatigue resistance (high temperature decreases fatigue resistance)
bull Frequency (rate of straining) ndash At high frequencies the metal component may be self-heated
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
11
Component
Geometry
Loading
History
Stress-Strain
Analysis
Damage
Analysis
Allowable Load - Fatigue Life
Material
Properties
Information path in strength and fatigue life prediction
procedures
Strength-Fatigue Analysis Procedure
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
12
Stress Parameters Used in Static Strength and
Fatigue Analyses
peak
t
n
peak
K
S
n
y
dn
0
T
r
peak
Str
ess
M
b) a)
n
y
dn
0
T
r
peak
Str
ess
S
S
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
13
Constant and Variable Amplitude Stress Histories
Definition of a Stress Cycle amp Stress Reversal
One
cycle
m
max
min
Str
es
s
Time 0
Constant amplitude stress history a)
Str
es
s
Time
0
Variable amplitude stress history
One
reversal
b)
max min
max mi
max min
mi
max
n
n
2 2
2
peak peak peak
peak peak
peak peak
peak
m
peak
pe
peakpe
ak
ak
a
R
In the case of the peak stress history
the important parameters are
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
14
Str
ess
Time
Stress history Rainflow counted cycles
i-1
i-2
i+1
i
i+2
Stress History and the ldquoRainflowrdquo Counted Cycles
1 1i i i iABS ABS
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
15
A rainflow counted cycle is identified when any two adjacent reversals in thee
stress history satisfy the following relation
1 1i i i iABS ABS
The stress amplitude of such a cycle is
1
2
i i
a
ABS
The stress range of such a cycle is
1i iABS
The mean stress of such a cycle is
1
2
i im
The Mathematics of the Cycle Rainflow Counting Method
for Fatigue Analysis of StressLoad Histories
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
16
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
17
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
18
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
19
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
20
Stress History
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Str
es
s (
MP
a)x
10
2
-3
-2
-1
0
1
2
3
4
5
6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Reversal point No
Str
es
s (
MP
a)x
10
2
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
21
Number of Cycles
According to the
Rainflow Counting
Procedure (N Dowling ref 2)
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
22
The Fatigue S-N method (Nominal Stress Approach)
bull The principles of the S-N approach (the nominal stress method)
bull Fatigue damage accumulation
bull Significance of geometry (notches) and stress analysis in fatigue
evaluations of engineering structures
bull
bull Fatigue life prediction in the design process
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
23
A B
Note In the case of smooth
components such as the
railway axle the nominal stress
and the local peak stress are
the same
peakS
Smin
Smax
Woumlhlerrsquos Fatigue Test
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
24
Characteristic parameters of the S - N curve are
Se - fatigue limit corresponding to N = 1 or 2106 cycles for
steels and N = 108 cycles for aluminum alloys
S103 - fully reversed stress amplitude corresponding to N = 103
cycles
m - slope of the high cycle regime curve (Part 2)
Fully reversed axial S-N curve for AISI 4130 steel Note the break at the LCFHCF transition and
the endurance limit
Number of cycles N
Str
ess a
mp
litu
de
S
a (
ksi)
Sy
Se
Part 1 Part 2 Part 3
Infinite life
Su
S103
10m A m
aS C N N
Fatigue S-N curve
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
25
Most of available S - N fatigue data has been obtained from fully reversed rotational bending tests
However material behavior and the resultant S - N curves are different for different types of loading
It concerns in particular the fatigue limit Se
The stress endurance limit Se of steels (at 106 cycles) and the fatigue strength S103 corresponding
to 103 cycles for three types of loading can be approximated as (ref 1 23 24)
S103 = 090Su and Se = S106 = 05 Su - bending
S103 = 075Su and Se = S106 = 035 - 045Su - axial
S103 = 072Su and Se = S106 = 029 Su - torsion
103 104
105 106 107
01
03
05
10
S103
Se
Number of cycles Log(N)
Re
lative
str
ess a
mp
litu
de
S
aS
u
Torsion
Axial Bending
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
26
Approximate endurance limit for various materials
Magnesium alloys (at 108 cycles) Se = 035Su
Copper alloys (at 108 cycles) 025Sult Se lt050Su
Nickel alloys (at 108 cycles) 035Su ltSe lt 050Su
Titanium alloys (at 107 cycles) 045Su ltSelt 065Su
Al alloys (at 5x108 cycles) Se = 045Su (if Su le 48 ksi) or Se = 19 ksi (if Sugt 48 ksi)
Steels (at 106 cycles) Se = 05Su (if Su le 200 ksi) or Se = 100 ksi (if Sugt200 ksi)
Irons (at 106 cycles) Se = 04Su (if Su le 60 ksi) or Se = 24 ksi (if Sugt 60 ksi)
S ndash N curve
33
1 1 1
2
10101log lo
0
g
1
3
A
m mm ma
m m
e e
A
a aS C N or N C S C S
SSm and A
S S
N
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
27
Fatigue Limit ndash Modifying Factors
For many years the emphasis of most fatigue testing was to gain
an empirical understanding of the effects of various factors on
the base-line S-N curves for ferrous alloys in the intermediate
to long life ranges The variables investigated include
- Rotational bending fatigue limit Sersquo
- Surface conditions ka
- Size kb
- Mode of loading kc Se = ka kb kc kd ke kfmiddotSersquo
- Temperature kd
- Reliability factor ke
- Miscellaneous effects (notch) kf
Fatigue limit of a machine
part Se
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
28
Ca
Effect of various surface finishes
on the fatigue limit of steel
Shown are values of the ka the
ratio of the fatigue limit to that
for polished specimens
Below a generalized empirical graph
is shown which can be used to
estimate the effect of surface finish
in comparison with mirror-polished
specimens [Shigley (23) Juvinal
(24) Bannantine (1) and other
textbooks]
Surface Finish Effects on Fatigue Endurance Limit
The scratches pits and machining marks on the surface of a material add stress concentrations to the
ones already present due to component geometry The correction factor for surface finish is sometimes
presented on graphs that use a qualitative description of surface finish such as ldquopolishedrdquo or ldquomachinedrdquo
(from J Bannantine ref1)
ka
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
29
Size Effects on Endurance Limit Fatigue is controlled by the weakest link of the material with the probability of existence (or density) of a
weak link increasing with material volume The size effect has been correlated with the thin layer of
surface material subjected to 95 or more of the maximum surface stress
There are many empirical fits to the size effect data A fairly conservative one is
bull The size effect is seen mainly at very long lives
bull The effect is small in diameters up to 20 in (even in bending and torsion)
Stress effects in non-circular cross section members
In the case of non-circular members the approach is based on so called effective diameter de
The effective diameter de for non-circular cross sections is obtained by equating the volume of material
stressed at and above 95 of the maximum stress to the same volume in the rotating-bending
specimen
0097
0097
1 03
0869 03 100
10 8
1189 8 250
b
b
e
e
e
e
if d inS
S d if in d in
or
if d mmS
S d if d mm
k
k
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
30
+
-
095max
max
005d2
The material volume subjected to stresses
095max is concentrated in the ring of
005d2 thick
The surface area of such a ring is
max
22 2
095 095 007664
A d d d
The effective diameter de for members
with non-circular cross sections
rectangular cross section under bending
F
09
5t
t
Equivalent diameter
max095 095 005
A Ft
A Ft t Ft
200766 005
0808
e
e
d Ft
d Ft
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
31
Loading Effects on Endurance Limit
The ratio of endurance limits for a material found using axial and rotating
bending tests ranges from 06 to 09
The ratio of endurance limits found using torsion and rotating bending tests
ranges from 05 to 06 A theoretical value obtained from von Mises-Huber-
Hencky failure criterion is been used as the most popular estimate
( ) ( )(07 09)
07 09 ( 085)c
e axial e bendi
c
ngS S
suggested by Shigl y kek
( ) ( )0577
057( 059)
e torsion e bendi
c
ng
c
S
suggested by Shigl kek y
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
32
From Shigley and Mischke Mechanical Engineering Design 2001
u T u T
e RT e RT
u RT u R
d
T
e T dS
S SS S
Sk
Sk
Temperature Effect
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
33
Reliability factor ke
The reliability factor accounts for the scatter of reference data such
as the rotational bending fatigue limit Sersquo
The estimation of the reliability factor is based on the assumption that
the scatter can be approximated by the normal statistical probability
density distribution
1 008ae
zk
The values of parameter za associated with various levels of
reliability can be found in Table 7-7 in the textbook by Shigley etal
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
34
(source S Nishijima ref 39)
S-N curves for assigned probability of failure P - S - N curves
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
35
Stress concentration factor Kt and the
notch factor effect kf
Fatigue notch factor effect kf depends on the stress
concentration factor Kt (geometry) scale and material
properties and it is expressed in terms of the Fatigue Notch
Factor Kf
1f
f
kK
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
36
1
peak
n
2
2
11
3
3
3
2
A D B
C
F
F
2
2
2
2
2
2
3
3
3
3
11
C
A B
D
Stresses in axisymmetric
notched body
n
peak
t n
FS
A
and
K
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
37
Stresses in prismatic notched body
n
peak
t n
t
FS
A
and
K
K S
A
B C
2
2
2
2
2
2
3
3
D
E
11
peak
n
2
2
11
3
3 3
2
A D B
C
F
F
1
E
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
38
Stress concentration factors used in fatigue
analysis
peak
t
n
peak
K
S
n S
x
dn
0
W
r
peak
Str
ess
S
n
x
dn
0
W
r
Str
ess
M
peak
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
39
Stress concentration
factors Kt in shafts
Bending load
Axial load
S =
S =
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
40
S =
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
41
Similarities and differences between the stress field near the notch
and in a smooth specimen
x
dn
0
W
r
peak S
tres
s
S
n=S peak= n=S
P
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
42
The Notch Effect in Terms of the Nominal Stress
np
n1
S2
cycles N0 N2
Sesmooth
N (S)m = C
S
max
Str
ess r
an
ge
S
n2
n3
n4 notched
smooth
e
e
f
SS
K
S
f t
K Kf
K Fatigue notch factor
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
43
1
2
3
Nsm
ooth
22
peak
smooth
e
peak
Str
ess
1
2
Nnotched
notched
e
M
notched
f notched
smooth
smoo he
e
tfo NNrK
Definition of the fatigue notch factor Kf
1
22= peak
Str
ess
2
Nnotched
notched
e
S
Se
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
44
1
1 1 11
tf t
KK q K
a r
1
1q
a r
PETERSONs approach
a ndash constant
r ndash notch tip radius
11
1
tf
KK
rr
NEUBERrsquos approach
ρ ndash constant
r ndash notch tip radius
18
330010
u
u
a inS
for S in in
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
45
The Neuber constant lsquoρrsquo for steels and aluminium alloys
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
46
Curves of notch sensitivity index lsquoqrsquo versus notch radius
(McGraw Hill Book Co from ref 1)
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
47
Plate 1 W1 = 50 in d1 = 05 in Su = 100 ksi Kt = 27 q = 097 Kf1 = 265
Illustration of the notchscale effect
Plate 2
W2= 05 in
d2 = 05 i
Su = 100 ksi
Kt = 27
q = 078
Kf1 = 232
pea
k
m1
r d1
W1
pea
k
r
d2
W2
m2
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
48
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Nf (logartmic)
JuvinalShigley method
Nf (logartmic)
Collins method
Sar ar ndash nominallocal
stress amplitude at zero
mean stress m=0 (fully
reversed cycle)
frsquo
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
49
Procedures for construction of approximate fully reversed
S-N curves for smooth and notched components
Sar ar ndash nominallocal stress amplitude at zero mean stress m=0
(fully reversed cycle)
Sar (l
og
art
mic
)
Nf (logartmic)
Manson method 09Su
100 101 102 103 104 105 106 2106
Sersquokakckbkdke
Sersquokakckbkdkekf
Se
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
50
NOTE
bull The empirical relationships concerning the S ndashN curve data are
only estimates Depending on the acceptable level of uncertainty
in the fatigue design actual test data may be necessary
bull The most useful concept of the S - N method is the endurance
limit which is used in ldquoinfinite-liferdquo or ldquosafe stressrdquo design
philosophy
bull In general the S ndash N approach should not be used to estimate
lives below 1000 cycles (N lt 1000)
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
51
Constant amplitude cyclic stress histories
Fully reversed
m = 0 R = -1
Pulsating
m = a R = 0
Cyclic
m gt 0 R gt 0
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
52
Mean Stress Effect
smlt 0
sm= 0
smgt 0
No of cycles logN 2106
Str
ess a
mplit
ude lo
gS
a
time
smlt 0
sm= 0
smgt 0
Str
ess a
mplit
ude S
a
0
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
53
The tensile mean stress is in general detrimental while the compressive mean stress is beneficial or
has negligible effect on the fatigue durability
Because most of the S ndash N data used in analyses was produced under zero mean stress (R = -1)
therefore it is necessary to translate cycles with non- zero mean stress into equivalent cycles with
zero mean stress producing the same fatigue life
There are several empirical methods used in practice
The Hiagh diagram was one of the first concepts where the mean stress effect could be accounted
for The procedure is based on a family of Sa ndash Sm curves obtained for various fatigue lives
Steel AISI 4340
Sy = 147 ksi (Collins)
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
54
Mean Stress Correction for Endurance Limit
Gereber (1874) S
S
S
S
a
e
m
u
2
1
Goodman (1899)
S
S
S
S
a
e
m
u
1
Soderberg (1930) S
S
S
S
a
e
m
y
1
Morrow (1960)
S
S
Sa
e
m
f
1
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and fatigue
life N = 1-2x106cycles
Sm- mean stress
Se- fatigue limit at Sm=0
Su- ultimate strength
frsquo- true stress at fracture
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
55
Mean stress correction for arbitrary stress
amplitude applied at non-zero mean stress
Gereber (1874)
2
1a m
ar u
S S
S S
Goodman (1899) 1a m
ar u
S S
S S
Soderberg (1930) 1a m
ar y
S S
S S
Morrow (1960)
1a m
ar f
S S
S
Sa ndash stress amplitude
applied at the mean
stress Smne 0 and
resulting in fatigue life of
N cycles
Sm- mean stress
Sar- fully reversed stress
amplitude applied at
mean stress Sm=0 and
resulting in the same
fatigue life of N cycles
Su- ultimate strength
frsquo- true stress at fracture
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
56
Comparison of various
methods of accounting
for the mean stress effect
Most of the experimental data lies between the Goodman and the yield line
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
57
Approximate Goodmanrsquos diagrams for ductile
and brittle materials
Kf
Kf
Kf
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
58
The following generalisations can be made when discussing
mean stress effects
1 The Soumlderberg method is very conservative and seldom used
3 Actual test data tend to fall between the Goodman and Gerber curves
3 For hard steels (ie brittle) where the ultimate strength approaches the
true fracture stress the Morrow and Goodman lines are essentially the
same For ductile steels (of gt S) the Morrow line predicts less
sensitivity to mean stress
4 For most fatigue design situations R lt 1 (ie small mean stress in
relation to alternating stress) there is little difference in the theories
5 In the range where the theories show a large difference (ie R values
approaching 1) there is little experimental data In this region the yield
criterion may set design limits
6 The mean stress correction methods have been developed mainly for the
cases of tensile mean stress
For finite-life calculations the endurance limit in any of the equations can be
replaced with a fully reversed alternating stress level corresponding to that
finite-life value
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
59
Procedure for Fatigue Damage Calculation
NT
n 1
n 2
n 3
n 4
c y c l e s
e
N o N 2
e
Ni(i)m = a
2
N1 N3
1
3
Str
ess r
an
ge
1 1 2 2
1 1 2 2
1 21 2
11 2
i i
i i
i
i in n ni
i i
n cycles applied atn cycles applied at n cycles applied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
1 1 2 2
1 1
R
i i
LD n N n N n N
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
60
n1 - number of cycles of stress range 1
n2 - number of cycles of stress range 2
ni - number of cycles of stress range i
1
1
1D
N - damage induced by one cycle of stress range 1
11
1
n
nD
N - damage induced by n1 cycles of stress range 1
2
2
1D
N - damage induced by one cycle of stress range 2
22
2
n
nD
N - damage induced by n2 cycles of stress range 2
1i
i
DN
- damage induced by one cycle of stress range i
ini
i
nD
N - damage induced by ni cycles of stress range i
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
61
Total Damage Induced by the Stress History
1 1 2 2
1 1 2 2
1 2
1 211 2
i i
i i
ii i
n n ni
i i
n cyclesapplied atn cyclesapplied at n cyclesapplied atD
N cycles to failure at N cycles to failure at N cycles to failure at
n nn nD D D D
N N N N
It is usually assumed that fatigue failure occurs when the cumulative damage
exceeds some critical value such as D =1
ie if D gt 1 - fatigue failure occurs
For D lt 1 we can determine the remaining fatigue life
1 1 2 2
1 1
R
i i
LD n N n N n N
LR - number of repetitions of
the stress history to failure
1 2 3
R iN L n n n n N - total number of cycles to failure
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
62
jN = a
m
j
if j gt e
It is assumed that stress cycles lower than the fatigue limit j lt e produce no damage (Nj=) in
the case of constant amplitude loading however in the case of variable amplitude loading the extension of the S-N curve with the slope lsquom+2rdquo is recommended The total damage produced by the entire stress spectrum is equal to
1
j
jj
D D
It is assumed that the component fails if the damage is equal to or exceeds unity ie when D 1
This may happen after a certain number of repetitions BL (blocks) of the stress spectrum which can be calculated as
BL = 1D
Hence the fatigue life of a component in cycles can be calculated as
N = BLNT where NT is the spectrum volume or the number of cycles extracted from given stress history
NT = (NOP - 1)2
If the record time of the stress history or the stress spectrum is equal to Tr the fatigue life can be expressed in working hours as
T = BL Tr
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
63
Main Steps in the S-N Fatigue Life Estimation Procedure
bull Analysis of external forces acting on the structure and the component in question
bull Analysis of internal loads in chosen cross section of a component
bull Selection of individual notched component in the structure
bull Selection (from ready made family of S-N curves) or construction of S-N curve adequate for given notched element (corrected for all effects)
bull Identification of the stress parameter used for the determination of the S-N curve (nominalreference stress)
bull Determination of analogous stress parameter for the actual element in the structure as described above
bull Identification of appropriate stress history
bull Extraction of stress cycles (rainflow counting) from the stress history
bull Calculation of fatigue damage
bull Fatigue damage summation (Miner- Palmgren hypothesis)
bull Determination of fatigue life in terms of number of stress history repetitions Nblck (No of blocks) or the number of cycles to failure N
bull The procedure has to be repeated several times if multiple stress concentrations or critical locations are found in a component or structure
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
64
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
65
Stress History
-100
-80
-60
-40
-20
0
20
40
60
80
100
0 5 10 15 20 25 30 35
Reversal point No
Str
ess (
ksi)
Example 2
An unnotched machine component undergoes a variable amplitude
stress history Si given below The component is made from a steel
with the ultimate strength Suts=150 ksi the endurance limit
Se=60 ksi and the fully reversed stress amplitude at N1000=1000
cycles given as S1000=110 ksi
Determine the expected fatigue life of the component
Data Kt=1 SY=100 ksi Suts=150 ksi Se=60 ksi S1000=110 ksi
The stress history
Si = 0 20 -10 50 10 60 30 100 -70 -20 -60 -40 -80 70 -30
20 -10 90 -40 10 -30 -10 -70 -40 -90 80 -20 10 -20 10 0
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
66
411108861
log60log6
log110log3
log60log10log
log110log1000log
logloglog
26
6
mC
Cm
Cm
Cm
Cm
CSmN
CSN
a
m
a
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
67
S-N Curve
10
100
1000
1000 10000 100000 1000000 10000000
No of cycles
Str
es
s a
mp
litu
de
(k
si)
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
68
Goodman Diagram
0
10
20
30
40
50
60
70
80
90
100
-100 -50 0 50 100 150
Mean stress (ksi)
Str
es
s a
mp
litu
de
(k
si)
SY
Suts
1
0
0
1 1
1
1
m
m
a m maa r at S
e uts uts
a m a
a r at Sma r uts
uts
S S Sfor fatigue endurance S S
S S S
S S Sfor any stress amplitude S
SS S
S
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
69
Calculations of Fatigue Damage
a) Cycle No11
Sar=8793 ksi
N11= C( SaN)-m= 186610268793-114=12805 cycles
D11=0000078093
b) Cycle No 14
Sar=750 ksi
N14= C( SaN)-m= 18661026750-114=78561 cycles
D14=0000012729
c) Cycle no 15
Sar=9828 ksi
N14= C( SaN)-m= 186610269828-114=3606 cycles
D14=000027732
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794
70
Results of rainflow counting D a m a g e
No S S m S a S ar (Sm=0) D i =1N i =1CS a -m
1 30 5 15 1552 20155E-13 0
2 40 30 20 2500 46303E-11 0
3 30 45 15 2143 79875E-12 0
4 20 -50 10 1000 13461E-15 0
5 50 -45 25 2500 46303E-11 0
6 30 5 15 1552 20155E-13 0
7 100 20 50 5769 63949E-07 0
8 20 -20 10 1000 13461E-15 0
9 30 -25 15 1500 13694E-13 0
10 30 -55 15 1500 13694E-13 0
11 170 5 85 8793 78039E-05 780E-05
12 30 -5 15 1500 13694E-13 0
13 30 -5 15 1500 13694E-13 0
14 140 10 70 7500 12729E-05 127E-05
15 190 5 95 9828 000027732 0000277
n 0 =15 D = 000036873 3677E-04
D=0000370 D=3677E-04
L R = 1D =271203 L R = 1D =271961
N=n 0 L R =15271203=40680 N=n 0 L R =15271961=40794