You are on page 1of 65

Fatigue for Engineers

Prepared by

A. F. Grandt, Jr.
Professor of Aeronautics and Astronautics
Purdue University
W. Lafayette, IN 47907

June 2001

1
Eastern Regional Office Southern Regional Office
8996 Burke Lake Road - Suite L102 1950 Stemmons Freeway - Suite 5068
Burke, VA 22015-1607 Dallas, TX 75207-3109
703-978-5000 214-800-4900
800-221-5536 800-445-2388
703-978-1157 (FAX) 214-800-4902 (FAX)

Midwest Regional Office Western Regional Office
1117 S. Milwaukee Avenue, Bldg. B, Suite 13 119-C Paul Drive
Libertyville, IL 60048-5258 San Rafael, CA 94903-2022
847-680-5493 415-499-1148
800-628-6237 800-624-9002
847-680-6012 (FAX) 415-499-1338 (FAX)

Northeast Regional Office International Regional Office
362 Clock Tower Commons 1-800-THE-ASME
Route 22
Brewster, NY 10509-9241 You can also find information on these courses and all of ASME, including
914-279-6200 ASME Professional Development, the Vice President of Professional
800-628-5981 Development, and other contacts at the ASME Website:
914-279-7765 http://www.asme
http://www.asme.org
.org

2
Objective

• Overview nature/consequences of the
fatigue failure mechanism
• Determine number of cycles required to
– develop a fatigue crack
– propagate a fatigue crack
• Discuss implications of fatigue on
design and maintenance operations

3
Structural Failure Modes
Force

• Excessive Deformation
displacement
– Elastic
– Plastic

Force
• Buckling Yield

• Fracture Permanent
displacement

• Creep Displacement

• Corrosion
• Fatigue
4
Fatigue Failure Mechanism
• Caused by repeated (cyclic) loading
• Involves crack formation, growth, and final
fracture
Fracture

Crack Length (a)
Stress

Crack Crack Nucleation
a Crack Growth

Time
Elapsed Cycles N

• Fatigue life depends on initial quality, load, . . .
5
Paper Clip Experiment
• Bend wire repeatedly until fracture

• Note:
– life (number of applied load cycles)
depends on:
• applied stress amplitude
• component “quality” (notches, scratches, etc.)
– heat emitted >> plastic deformation
6
Characteristics of Fatigue
• “Brittle” fracture surface appearance
• Cracks often form at free surface
• Macro/micro “beach marks”/ “striations”
Beach marks Striations

20 µ m
0.3 in
7
Fatigue is problem for many
types of structures

8
Exercise
Describe fatigue failures from your
personal experience
– What was cause of fatigue failure?
– What was nature of cyclic load?
– Was initial quality an issue?
– How was failure detected?
– How was problem solved?

9
Exercise
Estimate the fatigue lifetime needed for:
– Automobile axle
– Railroad rail
– Commercial aircraft components
• landing gear
• lower wing skin
– Highway drawbridge mechanism
– Space shuttle solid propellant rocket motor
cases

10
Exercise
• Give an example of a High Cycle
Fatigue (HCF) application.
– What is the required lifetime?
– What are consequences of failure?
• Given an example of a Low Cycle
Fatigue (LCF) application.
– What is the required lifetime?
– What are consequences of failure?

11
Fatigue Crack Formation

12
Fatigue Crack Formation
Objective
– Characterize resistance to fatigue crack formation
– Predict number of cycles to “initiate” small* fatigue crack
in component
*crack size ~ 0.03 inch
= “committee” crack

Crack Length (a)
Fracture

Approach
Crack Formation
– Stress-life concepts
Crack Growth
(S-N curves)
– Strain-life concepts Elapsed Cycles N
13
Stress-life (S-N) Approach
Concept: Stress range controls fatigue life
S ∆S/2
S
∆S

time
S
Log cycles N
Note:
• Life increases as load amplitude decreases
• Considerable scatter in data
• “Run-outs” suggest “infinite life” possible
• Life N usually total cycles to failure
14
Model Stress-life (S-N) Curve

Log ∆S/2
• Se = endurance limit ∆S/2 = σf ’ (2N)b
for steels
– Se ~ 0.5 ultimate stress Sult Se
– Se ~ 100 ksi if Sult 〈 200 ksi
Log reversals 2N
• σf ’ = fatigue strength coefficient
• b = fatigue strength exponent
typically -0.12 < b < -0.05

Note: Measure life in terms of reversals 2N
(1 cycle = 2 reversals)
15
S-N Curve: Mean Stress
S Smax
Mean stress effects life
stress ratio R = Smin / Smax ∆S = 2Sa
Smean = 0.5(Smin + Smax)
Sa = 0.5(Smax - Smin) = ∆S/2 time
Smin
Mean stress models
“Haigh” constant life diagram
Sa/Se + Sm /Sult = 1 Stress Amplitude N = 103

∆S/2 = (σf ’ - Smean) (2N)b N = 106

Mean Stress
16
S-N Curve: Other Factors
• S-N curves are very sensitive to
– surface finish, coatings, notches
– prior loading, residual stresses
– specimen size effects, etc.
• Many empirical “knock-down” factors
• S-N approach best suited for HCF (High
Cycle Fatigue) applications
– limited by local plastic deformation
– strain-life approach better for LCF (Low
17
Cycle Fatigue)
Strain-life (ε - N) Approach
Concept: Strain range ∆ε controls life
Experiment ∆σ
ε
• Control ∆ε
∆ε
• Measure ∆ε
– “Reversals” (2Nf) time
to failure (1 cycle
σ
= 2 reversals)
– Stable stress range ∆σ ∆σ
needed to maintain ∆ε
Note: “stable” ∆σ usually occurs
time
18
by mid-life (2Nf /2)
Cyclic Stress-Strain Curve
Relate stable cyclic stress and strain ranges
∆σ ∆ε
σ
ε σ
∆ε ∆σ
∆ε
time time
∆σ
ε

“Hystersis” loop
Cyclic stress-strain curve
∆σ/2
E = elastic modulus
∆ε/2 = ∆σ/2E + (∆σ/2K ’)1/n’ K’ = cyclic strength coefficient
n’ = strain hardening exponent

19 ∆ε/2
Plastic Strain-Life Curve
Relate “plastic” strain amplitude ∆εp/2
with reversals to failure 2Nf
Compute ∆εp/2 = ∆ε/2 - ∆σ/2E = total - “elastic” strain amplitudes

∆εp/2 = εf ’ (2Nf)c
Log ∆εp/2

εf ’ = fatigue ductility coefficient
c = fatigue ductility exponent
typically -0.7 < c < -0.5
Log 2Nf
20
Total Strain-Life Curve
Plot total strain amplitudes versus life 2Nf
∆εtotal /2 = ∆ε/2 = 0.5 ∆εelastic +0.5 ∆εplastic = ∆σ/2E + 0.5 ∆εplastic

∆ε/2 = {(σf ’ - Smean)/E} (2N)b + εf ’ (2Nf)c
Log strain amplitude

∆εp /2 = εf ’ (2Nf)c

∆σ/2E = {(σ f ’ - Smean)/E} (2N f)b

2Nt = “transition” life Log 2Nf
21
Total Strain-Life
Log strain amplitude
∆ε/2 = {(σf ’ - Smean )/E} (2N)b + εf ’ (2Nf)c

∆εp =εf ’(2Nf)c
HCF
LCF
∆σ/2E = {(σ f ’ - Smean)/E} (2N f)b

2N t = “transition” life Log 2N f

Note:
– Plastic strain dominates for LCF
– Elastic strain dominates for HCF
– Transition life 2Nt separates LCF/HCF
22
Variable Amplitude Loading
• Load amplitude varies in many applications
• Use of constant amplitude S - N or ε - N
data requires “damage model”
• Miner’s rule* 2S ai
S
Σ(Ni/Nf) = 1 Ni

time
*Use with caution!
Ni = number of applied cycles of stress amplitude Sai
Nf = fatigue life for Sai cycling only
23
Example Problem
Assume:
– σf’ = 220 ksi, b = - 0.1
– stress history shown (1 block of loading)
Find: number of blocks to failure
+ 100 ksi
S S + 80 ksi
2N = 1000

time
- 80 ksi 2N = 1000
S 2N = 100
- 100 ksi
24
Solution
Σ(Ni/Nf) = 1 2Nf = {(∆S/2) / (σf ’ - Smean)}1/b

∆S/2 Smean 2Nf 2Ni Ni/Nf
(ksi) (ksi)
Σ(Ni/Nf) = 1

80 0 24,735 100 0.0040 When:
1/0.0089
50 +50 206,437 1000 0.0048 = 112.5
6 -6
50 -50 21 E 1000 4.74 E Answer
0.0089 112 blocks
25
Load Sequence Effects
• Hi-lo strain ε sequence
ε results in compressive
mean stress σ when
last large ε peak is
tension
t • → increases life
σ σ • If last ε peak had been
ε t compression, would
result in tensile mean
stress
Mean stress
• → decreases life
Load sequence important!
26
Notch Fatigue
• Notches can reduce life
• Define Fatigue Notch Factor
Kf
Kf = Smooth/notch fatigue
∆S/2 strength at 106 cycles
Smooth 106 = ∆Ss /∆Sn
Notch 1 < Kf < Kt
∆Ss /2
(Kt = elastic stress
∆Sn /2
concentration factor)
Log cycles N Kf = 1 → no notch effect
Kf = Kt → full notch effect
27
Neuber’s Rule
Kf 2∆s∆e = ∆σ∆ε Kf = fatigue notch concentration factor
(∆s,∆e) = nominal stress/strain ranges
(away from notch)
(∆σ,∆ε) = notch stress/strain ranges
(∆s,∆e) Neuber’s rule relates notch and
nominal stress/strain behavior
(∆σ,∆ε) Solve with:

∆ε/2 = ∆σ/2E + (∆σ/2K ’)1/n’

∆ε/2 = {(σf ‘ - Smean)}(2Nf)b + εf ‘ (2Nf)c
28
Summary “Initiation” Methods
• Total strain-life approach combines:
– original S-N curve (best suited for HCF) and
– plastic strain-life method developed for LCF
problems
• S-N and strain-life often viewed as crack
“initiation” approaches
– actually deal with life to form “small” crack
– crack size implicit in specimen/test procedure
– typically assume “committee crack” ~ 0.03 in.

29
Initiation Summary Cont’
• Notches increase local stress/strain and
often are source for crack formation
– complex problem leads to local plasticity
– characterize by fatigue notch concentration
factor Kf,, Neuber’s rule
• Load interaction effects result in local
mean stress
– can increase/decrease life
– invalidate Miner’s rule
30
Fatigue Crack Growth

31
Crack Growth Approach
• Assumes entire life

Crack Length (a)
Fracture

fatigue crack growth Initial crack

– ignores “initiation” Crack Growth

– assumes component Elapsed Cycles N
cracked before cycling begins
• Used with “damage tolerant design”
– protects from pre-existent (or service) damage
– based on linear elastic fracture mechanics
32
Damage Tolerance
The ability of a structure to resist prior
damage for a specified period of time
Initial damage
Desired Life
– material

Crack size
– manufacturing
– service induced
– size based on
inspection capability,
time
experience, . . .
33
Fatigue Crack Growth
Objective
– Characterize material resistance to fatigue crack growth
– Predict catastrophic fracture and “subcritical” crack
growth
Approach

Crack Length (a)
Fracture
– Assume crack growth
Initial crack
controlled by stress
Crack Growth
intensity factor K
• fracture Elapsed Cycles N
• growth rate da/dN
34
Stress Intensity Factor K I
KI is key linear elastic fracture mechanics
parameter that relates:
– applied stress: σ σ
– crack length: a
– component geometry: β(a) Crack

a
(β(a) is dimensionless)

KI = σ πa β σ
β = 1.12
Note units: stress-length1/2
35
Stress Intensity Factors
 
1 K = σ π aβ  a 
 πa  W
K = σ π a Sec  
2

  W  β  a  = 112 − 0.231 a  + 10. 55 a  − 21.73 a  + 30.39 a 
2 3 4

 W . W W W W
σ = Remote Stress
 a  h
For   ≤ 0.6 and   ≥ 1.0
2a W W
≤ 0.95
W σ σ

W
Many KI W
solutions h
2a a
available

σ σ
36
Crack tip Stress Fields
Theory of elasticity gives elastic stresses near crack tip in
terms of stress intensity factor KI

All crack configurations have same singular stress field at tip
(are similar results for other modes of loading, i.e., modes II and III)
σy KI θ θ 3θ  
σx = cos 1 − sin sin  
σxy 2πr 2 2 2

y KI θ θ 3θ  
σx σy = cos 1 + sin sin 
2πr 2 2 2 
KI θ θ 3θ 
σ xy = sin cos cos
r 2πr 2 2 2 
σ xz = σ yz = 0 
θ 
x plane stress → σ z = 0 
plane strain → σ z = ν (σ x + σ y ) 
Crack
37
Kc Fracture Criterion
σ
• Fracture occurs when
K > constant = Kc
• Kc = material property 2a

= fracture toughness
• Criterion relates: σult σ
– crack size: a K c = σ πa β ( a )
Fracture Stress σ
– stress: σ
– geometry: β(a)
– material: Kc
• Plasticity limits small
crack applications Crack Size a
38
Fracture Toughness Kc
Typical Kc values (thick plate)
Ματεριαλ 2024−Τ351 7075−Τ651 Τι − 6 Αλ−4ς 300 Μ στεελ 18 Νιχκελ
(τηιχκ πλατε ) Αλυµ ινυµ Αλυµ ινυµ Τιτανιυµ (235 κσι ψιελδ) (200 κσι ψιε λδ)

Κχ 31 26 112 47 100
(κσι−ιν 1/2 )

Note Kc depends on:
– specimen thickness -- Kc decreases as
thickness increases until reaching minimum -
KIc = plane strain toughness
– crack direction (material anisotropy)
39
Fracture Example
5 ksi
σ=?

Member A fractures when
crack length a = 2.0 inch
4.0 in 8 in
and remote stress = 5 ksi
5 in
What stress will fracture 2.0 in

member B (assume same
material)?
σ=?
5 ksi

A B
40
Fracture Example Solution
5 ksi
Edge crack
4.0 in
K = σ(πa)1/2 β(a) = Kc at fracture
2 3 4
 a   a   a   a   a 
β   = 1.12 − 0 . 231   + 10. 55   − 21 . 73   + 30 . 39  
2.0 in          
W W W W W

a/w = 2/4 σ=5 a=2 → β = 2.83
5 ksi

Kc = 35.5 ksi-in1/2 = constant
σ=?
Center Crack
K = σ (π a)1/2β(a) β(a) = [Sec (π a/W)]1/2
8 in
a = 2.5 W=8 → β = 1.34
5 in
K = Kc at fracture = 35.5

41 σ=?
→ σf = 9.5 ksi
Fatigue Crack Growth
Goal: show cyclic stress intensity factor ∆K
controls crack growth rate da/dN
∆σ
Same material ∆P

Different loadings
2a
2a

Crack Face Load
Remote Load

σ P

∆σ = constant ∆P = constant

time time
42
Measure Crack Growth
∆σ
∆ K = ∆σ√π a ∆P

2a
∆K = ∆P 2a

B√πa Crack Face Load
Remote Load

da
Crack Length (a)

Crack Length (a)
dN

a*
da
dN

Number of Cycles (N) Number of Cycles (N)
43
Correlate Rate da/dN vs ∆K
2a
∆K = ∆σ π a
Crack Length (a)

a*
da

Log da/dN
dN

Number of Cycles (N)

da
dN
Crack Length (a)

∆ Kth Kc
2a ∆P
∆K =
Number of Cycles (N) B πa
Log ∆ K
44
da/dN Vs ∆K
Note:
• ∆K correlates fatigue
crack growth rate da/dN
• ∆K accounts for crack
geometry

Log da/dN
• No crack growth for
da/dN < ∆Kth
Kc
• Fractures when Kmax ∆Kth
in the ∆K range à Kc
• da/dN - ∆K curve is
Log ∆K
material property
45
Sample Crack Growth Data
• da/dN - ∆K data for
7075-T6 aluminum
• Note effect of stress
ratio R = min/max
stress (da/dN ↑as R↑)
• Reference: Military
Handbook-5
• Other handbook data
are available
46
Model da/dN - ∆K Curve
Fit test data with numerical
models such as:
da da
= F (K) Paris = C∆K m
dN dN
Log da/dN

da C ∆K m
Forman =
dN (1 − R ) K c − ∆K
Kc Here C, m, Kc are
∆Kth
empirical constants
R = min/max stress
Log ∆K (are many other models)
47
Compute Fatigue Life Nf
da da
Nf = ∫
af
= F (K)
dN ao F ( K)
σ

∆σ ao, af = initial, final crack sizes
∆σ time F(K) = function of:
– cyclic stress: ∆σ, R, . . .
– crack geometry: β(a)
2a – crack length: a
– material
48
Example Life Calculation
Given: edge crack in wide plate
σ Kc= 63 ksi-in1/2
initial crack ai = 0.5 inch
Crack cyclic stress ∆σ = 10 ksi, R = 0
a (∆σ = σmax = 10 ksi)
da/dN = 10-9∆K4

Find: a) cyclic life Nf
σ b) life if initial crack size
decreased to ai = 0.1 inch
σ Note: at fracture
K = Kc = 63 = 1.12σmax (πa)1/2
∆σ = constant

time → final crack af = 10 inch
49
Solution
da
K = σ πa 112
. = C∆K m

dN
af da af da
Nf =∫ =∫
C ∆K
[ ]
m m
. ∆σ π a
ao ao
C 112

Nf =
1
[ af
1− .5m
− ao
1 − .5 m
]
( ) ( )
m
. ∆σ π
C 112 1−.5m

a) Nf = 12,234 cycles (ai = 0.5)
b) Nf = 63,747 cycles (ai = 0.1)
Note: big influence of initial crack length!
50
Fatigue Crack Retardation
Without Overload
Applied Stress (σ)

Overload

Crack Length (a)
With Overload
“Retardation”

Time
Elapsed Cycle (N)

Note “load interaction effect”
• Tensile overload can “retard” crack growth (increase life)
• Life increase due to crack tip plasticity
• Depends on magnitude/sequence of overload, material, …
• Are empirical retardation models
51
Cycle-by-Cycle Calculation
σn
Applied Stress (σ)

Variable amplitude
loading prevents
σn+1 simple life integration

Time (t)

Compute cycle-by-cycle growth in crack length a
– acurrent = aprior + da/dNcurrent
– da/dNcurrent = F(Kcurrent) * “Retardation” term
– Sum for all cycles in spectrum
Powerful technique for computer programming
52
Crack Growth Summary
• Fracture mechanics approach assumes
entire fatigue life is crack growth
• Stress intensity factor K controls fracture
and growth rate da/dN
– K = σ[πa]1/2 β(a)
– Fracture: K = Kc
– Fatigue: da/dN = F(∆K)
– Integrate da/dN for life
• Are load interaction and other effects (see
53
references)
Fatigue Design/Repair
Concepts

54
Design Philosophies
σ

Fatigue Design Criteria

Stress
Crack
• Infinite Life a

• Safe-Life
• Damage Tolerant σ Time

– Fail-safe
– Slow crack growth Crack Length (a) Fracture

• Retirement-for-cause Crack Formation

Pre-Crack
Crack Growth

Elapsed Cycles N
55
Infinite Life Criterion
Design Goal: prevent fatigue damage from ever
developing (i.e. infinite life)
• Usually based on endurance limit
• Could also employ threshold K concepts
• Leads to small design stresses/heavy members
• Limited to simple components/loading
• Often impractical/not achievable in practice
– Weight critical structure
– Complex loads

56
Safe-Life Criterion
Design goal: component is to remain crack free for
finite service life
• Assumes initial crack-free structure
• Establish “mean life” by test/analysis
• Safety factors account for “scatter”
Problems: Desired life = mean/S.F.
Occurrence
• large safety factor predicted
mean
Failure

• no protection from
initial damage
1 2 3 4 Design Life
57
Fail-Safe Criterion
Design goal: contain single component failure
without losing entire structure
• Assumes crack is present
• Provide alternate load paths, redundant structure, crack
stoppers, etc.
• Requires detection of 1st failure
Crack size

Crack arrest 2nd member

1st member

58 Time
Slow Crack Growth Criterion
Design goal: prevent initial crack from growing to
fracture during life of structure
• Pre-existent crack size specified by inspection
limits, experience
Desired Life Fracture

Crack size
• Crack growth life
> service life x S.F.
• Based on fatigue
crack growth
resistance time

• Emphasizes nondestructive inspection
59
Retirement-for-Cause
Design goal: Use periodic inspection/repair
to achieve desired fatigue lives
Limited by repeated maintenance economics
Failure size
Crack Length

inspect/repair

60 Time
Life Extension Concepts
Component
Inspection

No Cracks Found Cracks Found
(assume small cracks)

Introduce Beneficial Reduce Stress Reduce Operating Repair Cracked
Residual Stresses via Reinforcement Loads Structure

Shot peen HCF damping materials Weight limits Replace component
Hole coldwork Flight restrictions Stop drill cracks
Interference fasteners etc. Welding
Overstress, etc.

Doublers
Patches

Metal Mechanical Fasten Metal
Composite Bond Composite Mechanical Fasten
Bond

61
Summary
• Fatigue is complex problem that
involves many disciplines
• Fatigue affects design and operation of
many types of structures
• Fatigue may be treated by several
methods/philosophies
– Assume component cracked
– Assume component uncracked
– Probabilistic methods
62
63
Your Path to Lifelong Learning

ASME offers you exciting, rewarding ways to sharpen your technical skills, enhance personal development and prepare for
advancement.
Short Courses – More than 200 short courses offered each you keep you up to speed in the technology fast lane—or, help you fill in
any gaps in your technical background.
Customized Training at your organization’s site – Do you have ten or more people at your site who could benefit from an ASME
course? Most of our courses can be offered in- house and tailored to your latest engineering project. Bring course to your company
too.
Self-study materials meet the needs of individuals who demand substantive, practical information, yet require flexibility, quality
and convenience. Return to each program again and again, as a refresher or as an invaluable addition to your reference library.
FE Exam Review – A panel of seasoned educators outline a wide range of required topics to provide a thorough review to help
practicing engineers as well as engineering students prepare for this challenging examination. Videotape Review
PE Exam Review – A comprehensive review of all the major exam topics that demons trates the necessary math, logic and theory.
Videotape, Online, or Online Live Review available.

64
FOR MORE INFORMATION CALL 1-800-THE-ASME
__________________________________________________________________________
INFORMATION REQUEST FORM

Please mail to ASME at 22 Law Drive, P. O. Box 2900, Fairfield, NJ 07007-2900, or fax to 973-882-1717, call 1-800-THE-ASME, or
email infocentral@asme.org.
Send me information on the following:
____ Short Courses ____ In-House Training ____ Self-Study Programs
____ FE Exam Review ____ PE Exam Review (videotape) ____ PE Exam Review (Online)
____ PE Exam Review (Online Live)
Name: ______________________________________________
Title: _______________________________________________
Organization: _________________________________________
Business Address: _____________________________________
City: _________________ State: __ Zip Code: _____________
Business Phone: _________________ Fax: ________________
Email: ______________________________________________

65