Fatigue and Fracture Mechanics

Design for fatigue strength

Objective of fatigue calculations

The reoccurrence of loads in machine components produces stresses, which are

often called repeated, alternating or fluctuating stresses.
The objective of modern fatigue design is to ensure safe operation of a component
during its entire life. This is achieved by using a combination of analytical methods,
numerical simulations and data from experimental programmes.

• High cycle fatigue (HCF)

The number of load repetitions is large, typically over 104 cycles and it is assumed
that no initial defect exists in the structure. The cyclic stress level is generally below
material yield. This approach is frequently called:

• Low cycle fatigue (LCF)

The number of load repetitions is small (less than 103 cycles) and generally
accompanied by a significant amount of plastic deformation. Cracks may initiate
early in the life. The fatigue process is controlled by the strain deformation. This
approach is frequently called or the strain-life approach or the local approach to
fatigue.

• Fatigue crack growth (FCG)

A known defect assumed to exist in the structure and crack growth from the defect
is related to the number of cycles using fracture analysis. It is usually suitable only
for an elastic dominated stress and hence for HCF. This area is undergoing a rapid
development during the last 15 years and is commonly referred to as.

The basics of high cycle fatigue (HCF) are given in the following sections while the
area of FCG is covered during the introduction to fracture mechanics. Low cycle
fatigue is beyond the scope of this course and is less used in the aerospace industry.

14-1
 Fatigue and Fracture Mechanics

Fatigue tests

Fatigue analysis is based on data obtained in experiments. The basic data commonly
used in fatigue strength design includes correlation between the cyclic stress
amplitude and the number of cycles to failure of simple coupons.

However, likely deviations occur between the service components and the test piece
due to many factors such as:

geometrical differences
loading conditions and
environmental effects, for example corrosion.

The simplest fatigue test takes a smooth piece of material and applies a fluctuating
load, which varies between fixed upper and lower limits. Loading is continued until
the specimen is separated (broken!) into two pieces. These tests are commonly
called: uniaxial, load control, constant amplitude tests, and considered the base
for the fatigue life estimation in the HCF region.

An example of such a test is shown below:

Constant Fixed End

Specimen
amplitude load
Max Min

Constant amplitude test

This illustrates the simplest fatigue test. Most experimental programmes include test
coupons that have notches, grooves or holes. For example, it is common practice in
the aerospace industry to cycle to fracture a typical structural joint using two plates
and a fastener.

14-2
 Fatigue and Fracture Mechanics

Definitions

Fatigue under constant amplitude of applied stress, S (MPa)

Three typical examples:

S Smax Smax
Sa
Sa
time S ∆S
0
Sa
0
Smin Smin
one cycle Sm
a. Completely reversed, R= -1, Sm = 0 b. Non-zero positive mean stress, Sm >0

S Smax=∆S

Sm
Sa
Smin
0

c. Tension only and Smin = 0

Assuming a constant cross section area the following cyclic stress definitions are
used:

♦ The stress range, ∆S is the difference between the maximum and the minimum
stress values.
♦ Averaging the maximum and the minimum stress values gives the mean stress,
Sm.
♦ The mean stress Sm, may be zero, as in (a), but often it is not, as in (b).
♦ Half the range is called the stress amplitude, Sa, so that this is the variation
about the mean.

Stress range ∆S = S max − S min

∆σ S max − S min
Stress amplitude S a = =
2 2
S + S min
mean stress S m = max
2

14-3
 Fatigue and Fracture Mechanics

The term alternating stress has the same meaning as stress amplitude. It is also
useful to note that:

Smax = Sm + Sa and Smin = Sm − Sa

The signs of Sa and ∆S are always positive, since Smax > Smin, where tension is
considered positive. The quantities Smax, Smin, and Sm can be either positive or
negative. The ratio between minimum and maximum stress, R, is often used:

Smin
R= and therefore :
Smax

Smax
Sm = (1 + R ) and ∆S = Smax (1 − R )
2

where R is called the stress ratio or R-ratio.

♦ Cyclic stress with zero mean can be specified by giving the amplitude Sa, or by
giving the numerically equal maximum stress, Smax.

♦ The term completely reversed stress is used to describe a situation of Sm = 0,

hence R = -1, as in figure (a) on previous page. Zero-to-tension stressing refers to
cases of Smin = 0, hence R = 0, as in figure (c) on previous page.

♦ If the mean stress is not zero, two independent values are needed to specify the
loading. Some combinations that may be used are: Sa and Sm, Smax and R and ∆S
and R, Smax and Smin.

The same system of subscripts and the prefix ∆ are used in an analogous manner
with other variables, such as reversed bending moments M, cyclic strains ε,
fluctuating load (force) P.
For example, Pmax and Pmin are maximum and minimum loads, ∆P is load range, Pm
is mean load, and Pa is load amplitude. If there is any possibility of confusion as to
which variable is used with the ratios R, a subscript should be used, such as Rε for
strain ratio.

Note: FOR LOCAL STRESS CALCULATION S IS COMMONLY REPLACED BY σ

14-4
 Fatigue and Fracture Mechanics

Example – the Comet jet aeroplane fatigue failure

The Comet jet history demonstrates the importance of design for fatigue in the
commercial jet transportation.

Background

♦ It was the first modern commercial jet design after WW2.

♦ The first commercial plane designed to cruise at high altitude of 12000m.
♦ Therefore the first plane where the fuselage was pressurised and depressurised
during every flight to maintain passengers’ comfort.
♦ Fuselage had to be capable of withstanding the fatigue stress cycles.
♦ It is widely believed that the Comet fatigue failure set back the British industry
and assisted in the emergence of Boeing as a commercial leader.

Sequence of disasters

♦ 2 May, 1953 – Comet plane disintegrates in mid-air soon after take off in India
during a heavy tropical thunderstorm. Investigation concludes the accident is due
to extreme stormy weather and crew control. It is not viewed as a cause for
concern.
♦ 10 January, 1954 – A comet explodes over the Mediterranean Sea in good
weather. No flaws in design discovered and the plane remain in service.
♦ 8 April, 1954 – third crash above Rome investigation begins and the plane is
grounded but no real reason has been found.

Investigation

♦ A testing set up is constructed using a Comet plane on the ground: the plane
fuselage is subjected to repeated pressurised cycles and the wings to the cyclic
flight loads.
♦ After about 3000 pressurisation cycles fatigue cracks initiate and grow from a
corner of a window and advanced until…..
♦ Secondary cracking take place by bending and shearing off the center over the
outer portion.
♦ Complete piercing and catastrophic failure.

14-5
 Fatigue and Fracture Mechanics

Conclusions

The failures of the three Comet jet planes were due to fatigue cracking that was not
accounted for in the initial plane design. The design was modified and in the
subsequent planes the windows sections were replaced with a new reinforced panel.

14-6
 Fatigue and Fracture Mechanics

Some basic considerations in fatigue calculations:

Aspect Testing of In -
Laboratory specimens Service components

Loading Uniaixial, constant Irregular, multiaxial, not-

amplitude completely known
Complex with notches,
Geometry Small, smooth specimen grooves fillets etc.
Corrosion, temperature,
Environment Well known and controlled fretting
Good surface finish, As received, surface
Manufacturing homogeneous, heat treated treatment etc.

Scatter Scatter usually small Scatter usually large

Material Inspection is carried out Only snap-shots may be

microstructure during the tests possible

Fatigue lecture objectives:

1. Understand basics fatigue failure and the fatigue terminology.

2. Prediction of fatigue strength and fatigue life using the S-N
diagram for both, finite and infinite life.
3. Be able to account for stress concentration, mean stress and simple
multiaxial load.
4. Apply the stress-life analysis to design.
5. Calculate service component fatigue damage using damage
summation.

14-7
 Fatigue and Fracture Mechanics

Fatigue using Stress-life approach (High Cycle Fatigue)

The stress-life approach is the traditional and still the most widely used method of
estimating the life of components. It has a long history, initially developed during
the industrial revolution of the 19th century when metal structures became widely
used in engineering design.

1. Fatigue limit and the S-N diagram

The first concept that emerged from a stress-life analysis of several metals was that
of ‘endurance limit’, which characterises an applied stress amplitude below which
the material is expected to have an infinite life. Using this concept for design, it is
therefore only necessary to ensure that the stress in the structure is below the
material fatigue limit. However, this simplistic approach may be inappropriate in
many fatigue cases.

To establish the strength of materials under fatigue loads, a series of tests are
conducted on similar specimens subjected to fluctuating, completely reversed
stress, R=-1 and mean stress is zero!. The number of cycles to failure of the
specimens is recorded and the stress amplitude is plotted on a log paper against the
number of cycles to failure. A typical S-N diagram of steel alloy is shown below.

Low cycle High cycle

Reversed Stress amplitude Sar

Finite life Infinite

life

800

600

400
Chromium-
300
molybdenum
200 steel Se
Fatigue
100 limit

100 101 102 103 104 105 106 107 108

Number of fatigue cycles to failure , Nf

14-8
 Fatigue and Fracture Mechanics

The S-N diagram has two distinct regions:

Finite life Number of cycles 107 or less Specimens failed

Infinite life Number of cycles 107 or over Specimen do not fail

It is customary in design applications and standards to assume that no fatigue failure

may occur in a component if the stress amplitude, called the fatigue strength,
corresponds to a number of cycles above 107 cycles. This stress is often referred to
as the material ‘endurance limit’ or ‘fatigue limit’ explained above.

However, some structural steels such as non-ferrous metals and alloys do not have
an endurance limit!

Within the S-N diagram, the data in the region of life to failure between 103 and 107
cycles is found to approximate a straight line on a log-log plot. The corresponding
relation is illustrated below.

A
Reversed
Stress B Sar = A Nf B
amplitude
Sar
103 < Nf < 107
3
1 10 107
Cycles to failure

Sar is the reversed stress amplitude or the cyclic stress at

R = -1 (zero mean stress) and is the stress value shown on
the material S-N diagram.
The finite life region of the S-N diagram could also be split into the high cycle
fatigue life and the low cycle fatigue life. In the low cycle fatigue we assume
material yield at the critical area and therefore a stress based approach may not be
appropriate.

The exact borderline between the LCF and the HCF regions is not always very clear
but is most often assumed to be at fatigue strength corresponding to a life of
between 103 to 104 cycles.

14-9
 Fatigue and Fracture Mechanics

Past design books provided factors to modify the endurance limit for practical
application by, for example, introducing factors that account for each single effect.
These factors were used to modify (usually reduce) the endurance limit of the
laboratory controlled test components. For instance:

Se = KaKbKcKdKeKf……S’e

Where Se is the modified endurance limit. Ka,b,c,d,e,f…are the modifying parameters

for different effects, of for example; manufacturing, geometry, reliability,
temperature etc., taken from tables and graphs. However, this simplified approach
could lead to errors and is not widely used anymore.

2. Mean stress effect

So far we have only mentioned fully reversed cyclic stress where the mean stress of
the fatigue cycle is zero, Sm=0. In many applications this is not representative and
the mean stress has shown to influence the fatigue behaviour of metals.

There are several ways to include the mean stress effect in fatigue life procedures.
One popular method is the constant life diagram. This involves plotting values of
several combinations of mean stress and stress amplitude that produce the same life
to failure, each curve indicating a constant life. Another procedure often used is to
plot curves of the mean ratio (R = Smin/ Smax ) for the maximum stress versus life to
failure.

Nf3 Constant life curves

Stress Amplitude

Maximum stress

Nf2 Nf1>Nf2>Nf3
R<0

Nf1 R=0

R>0
Constant R-ratio curves
σu
Mean stress Cycles to failure

14-10
 Fatigue and Fracture Mechanics

Mean stress correction models

Several models were developed to compare life plots of tests using different mean
stress. These models are primarily used to estimate the life of a component with a
particular mean stress to the fully reversed (zero mean stress) material cyclic test
data.
A frequently used relationship is:

n Where: Sa = Applied stress amplitude (with mean stress)

s a +  s m  =1
Sm = Mean stress
s ar  s u 
Su = Ultimate tensile stress
Sar = Reversed stress amplitude (no mean stress)

In the equation above, ‘n’ value refers historically to the following:

1. n = 1 The ‘modified Goodman equation’

2. n = 2 The ‘Gerber parabola’

3. n = 1, Su = Sy The ‘Soderberg relation’

Su is the material ultimate strength and Sy is the yield stress.

The above three models could be presented graphically on the constant life diagram
by plotting the mean stress Vs the stress amplitude, where σar is the stress amplitude
in the case of zero mean stress (R = -1):

Constant life curves

Stress Amplitude Sar

Gerber

Modified Goodman

Soderberg

Mean stress Sy Suts

14-11
 Fatigue and Fracture Mechanics

Other relations also exist between the fully reversed data and the mean stress such as
the ‘Walker equation’ and corrections that account for negative mean stress.

The following observations are usually made about the foregoing mean stress
models:

1. Soderberg provides a conservative estimate of fatigue life.

2. Goodman is adequate for brittle materials but conservative for ductile behaviour
if tension dominate the fatigue life.

3. Gerber is in general good for ductile materials such as steel alloys but it does not
distinguish between life in tensile or compression mean stress.

If not mentioned otherwise, the modified Goodman is considered for mean

stress correction in these lectures.

Example (mean stress)

Calculate the fatigue life of a smooth bar subject to a constant stress range of 1110.6
MPa and mean stress of 200 MPa. Materials properties are: Su = 1400, Sy = 1170, A
= 1240 all in MPa and B = -0.06.

Solution:
∆S 1110.6
Sa = = = 555.3 MPa
2 2

Using modified Goodman: 1

s a  s m 
where: + =1
Sa 555.3 s ar  s u 
Sar = = = 647.85 MPa
Sm 200
1− 1−
Su 1400 where: Sar = ANf
B

1  1 
  − 
 Sar  B   647.85  0.06 
Nf =   = = 50x103 Cycles to failure
A   1240 

(Hand calculators hint: invert the fraction to use a positive exponent)

Repeat this example using Soderberg and compare the results

14-12
 Fatigue and Fracture Mechanics

3. Stress concentration effect

One of the major factors of reduction in life of a service part in comparison to tests
data is the change in geometry that locally raises the cyclic stress. These
unavoidable geometric discontinuities also affect the local stress field and the shape
and form of the fatigue cracks. Much work was done in the past to compare the so
called notched parts to smooth data. We have here restricted our study to the
classical approach to fatigue.
In static design the effect of notches is considered by using engineering handbooks
that contain data of different notch geometry and the relevant stress concentration
factor (SCF) and applying:

σ = Kt S

Where σ is the local stress in the vicinity of the notch; S the global or NOMINAL
stress and Kt is the stress concentration factors taken from data books.

In fatigue the SCF is replaced by the so-called Fatigue notch factor, Kf (sometimes
also called fatigue strength reduction factor):

Allowable cyclic stress without the notch

Kf =
Allowable cyclic stress with notch

It was shown that in most fatigue cases that reduction in fatigue strength is less than
the increase in the local static stress, which means that Kf < Kt in most cases.

A concept was introduced to the fatigue of notched components using a notch

sensitivity factor, q. This factor is used to correlate between Kt and Kf and is
mainly a material parameter.
Various relationships were developed between these three parameters (Kt, Kf and q).
Most of them use empirical (experimental) relation. The most widely used ones take
the form:
−1
q = Kf
Kt − 1

14-13
 Fatigue and Fracture Mechanics

Since Kt is widely available for many notch geometries and q is a given material
data, it is possible to calculate Kf, the fatigue reduction factor.

An example of a notch sensitivity graph for various notch radii is given below
(Shigley, Mechanical Engineering Design, 1977) for four steels and an aluminium
alloy:
Notch Sensitivity, q

Steels
Aluminum

Notch radius, r, mm

In engineering literature, the value of q is sometimes represented using the

following empirical formula:
1
q=
α
1+
r

Where r is the notch radius (in mm) and α is a material constant.

14-14
 Fatigue and Fracture Mechanics

Example: Fatigue life estimates - stress concentration and mean stress

Calculation of fatigue life is required for the top skin joint on the wing of a large
commercial jet. The joint skin plate is made of standard aluminium Al 2024-T3.
From a separate load analysis it was found that a remote cyclic reversed load of
about 20 kN and a local constant stress of approximately 70 MPa is applied to the
structure. Estimate the fatigue life of the plate in the vicinity of the bolt’s hole as
shown below. The 2024-T3 properties are: S-N curve constants, A=839 MPa, B=-
0.102, σu=476 MPa and α=0.51. Assume factor of safety of 1.8.

Scrap top view

Section view

6mm Pa 1/2”
50mm
σm

Solution procedure:
1. It is first necessary to obtain the notch concentration factor, Kt, and the notch
sensitivity, q using the available data.
2. From those, the fatigue notch factor, Kf is calculated.
3. Next, the cyclic and mean stress components’ are evaluated.
4. One of the mean stress correction methods is used to obtain the equivalent
reversed stress amplitude, σar and this is multiplied by Kf to obtain the local
fatigue stress amplitude.
5. This reversed stress is used to estimate the life from the material S-N curve.
Solution:
1. Using the figure shown below for stress concentration factors (Peterson 1974):

W = 50mm
d = 12.7mm
r = 6.35mm
d/W = 0.254

Kt = 2.4

14-15
 Fatigue and Fracture Mechanics

The notch sensitivity is: 1

q= = 0.926
0.51
1+
6.35

2. And hence the fatigue notch factor:

K f = 1 + q(K t − 1) = 1 + 0.926(2.4 − 1) = 2.296

3. We now obtain the applied cyclic stresses:

 P   20x 103 
∆σlocal cyclic range =  (w − d )t  =  (50 − 12.7 )6  = 89.4 MPa
   

σconstant = σmean = 70 MPa

max
Cyclic
reversed Constant stress ∆σ
mean
stress
∆σ
min

∆σ 89.4
σa = 2
=
2
= 44.7 MPa

σmean = σcons tan t = 70 MPa

4. The completely reversed stress using the modified Goodman theory (FOS=1.8):
(Sar) (FOS) (Kf)
σa 44.7
σar = = = 52.4 And: 52.4 x 1.8 x 2.296 = 216.6 MPa
σ
1− m 1−
70
σu 476
1

[ ]
   1 
5. Finally using the material  σ ar   B 216.6  − 0.102 
Nf = =
S-N curve equation:  A  839

Nf = 583,407 cycles

14-16
 Fatigue and Fracture Mechanics

Example (in class exercise – stress concentration)

A shaft with a step-down in diameter has dimensions, as defined in Figure 1 below,

of d1 = 50, d2 = 55, and ρ = 1.3mm. The shaft is subjected to reversed bending and is
made of a quenched and tempered steel. The 106 cycles fatigue strength in bending
of this steel using smooth specimens is σar = 500 MPa.
What completely reversed bending moment Ma can be applied to the notched shaft
for 106 cycles before fatigue failure expected? (use Figures 1 and 2).
Notch radius, ρ, in mm

Kt q

ρ/d1 Notch radius, ρ, in inches

Figure 1 Figure 2

Solution procedure: obtain Kt and q, calculate Kf, calculate allowed fatigue strength,
and calculate Ma using bending theory.

14-17
 Fatigue and Fracture Mechanics

Example (in class exercise – mean stress)

A shaft of circular cross section is subjected to a steady bending moment of 1500

Nm and simultaneously to an alternating bending moment of 1000 Nm in the same
place. Calculate the necessary diameter of the shaft to withstand fatigue failure if a
factor of safety of 2.5 is to be used. The material yield stress is 210 MPa (use
Soderberg rule) and the fatigue limit in reversed bending is 170 MPa.

Solution procedure:
a. Determine the mean and amplitude moments.
b. Determine the mean and amplitude stresses using bending theory, diameter
unknown.
c. Set up the Soderberg mean stress correction equation, include the FOS.
Resolve for the diameter.

14-18
 Fatigue and Fracture Mechanics

4. Irregular load history

In practice, engineering components are rarely subjected to constant load amplitude

and airframes are no exception. Methods and standards have been developed that
compare between the irregular load cycles fatigue damage and the constant
amplitude laboratory coupon results. This allows an estimate of the service
component damage providing that some form of the load history is available.

Damage summation

A simple and widely used criterion for predicting the extent of fatigue damage
induced by a block of cycles having the same stress amplitude and mean is called
the Miner’s rule. It implies that the fraction of life used by a certain block of cycles
is equal to the ratio between the number of cycles used in the block, ni, and the
number of cycles to failure, Nf, for the same stress amplitude, on the S-N curve.
Failure is expected when the total sum of all such fractions is equal to unity.

i=m
n1 n n n
+ 2 + 3 + ...... = ∑ i = 1
Nf 1 Nf 2 Nf 3 i =1 N fi

This is illustrated below:

σa3 σa
σ σa1
σa2
σa3
time σa1
σa2
n1 n2 n3 Nf3 Nf1 Nf2
Number of cycles in a block Nf Cycles to Failure

14-19
 Fatigue and Fracture Mechanics

It is often customary for practical applications to estimate the damage for a

particular service history, for example an aeroplane’s particular flight pattern that
includes taxiing – take off – cruising - landing. The sequence is repeated until the
cumulative damage reaches unity and the number of repetitions to failure is
calculated:

1
Number of repetitions = i = m
∑ ni
i =1 N fi

Although the Miner’s rule is widely used it is essential to understand its limitations,
in particular:

1. It is a LINEAR summation model where cycles at different stress amplitude

have the same weight on the damage and total failure.
2. The ORDER in which the stress blocks are applied has no significance.
3. Fatigue failure occurs when the sum of the fractions reaches a particular value
– usually one.

It has been shown and is widely accepted that these three assumptions are often
inaccurate.

Example - irregular load 1

The stress history shown below was measured near a titanium Ti-6Al-4V landing
gear component during normal operational conditions. Estimate the number of
repetitions to failure of this stress history if it has the following fatigue properties:
σu = 1233 MPa, A = 1889 MPa and B = -0.104.

800
σ, Stress, MPa

220

0
time

14-20
 Fatigue and Fracture Mechanics

Solution:

800
σ, Stress, MPa

220
1 2 3 4 5 6 7 8 9 10
0
time
One repetition

In one repetition of the history we count one large cycle and 10 smaller ones.

For the large cycle the stress amplitude and mean are:

σa = (800 – 0)/2 = 400 MPa and σm = (800+0)/2 = 400 MPa

Using the modified Goodman correction to obtain the fully reversed stress
amplitude:

σa ,effective 400
σar = = = 592.1 MPa
σm ,effective 400
1− 1−
σu 1233

and from the material S-N curve the number of cycles to failure is:

1  1 
  − 
 σar   B   592.1  0.104 
Nf =   = 
A  1889 

Nf = 69917 cycles

14-21
 Fatigue and Fracture Mechanics

For the smaller stress cycles similar analysis:

σa = (800 – 220)/2 = 290 MPa and σm = (800+220)/2 = 510 MPa

σa ,effective 290
σar = = = 494.6 MPa
σm,effective 510
1− 1−
σu 1233

1  1 
  − 
 σar   B   494.6   0.104 
Nf =   =
A  1889 

N f = 394677 cycles

Summation for one repetition using Miner’s rule:

1 1
Number of repetitions = i=m = = 25227 (Until failure)
ni  1   10 
∑  + 
i =1 Nfi  69917   394677 

25227 repetition of this load history to failure assuming Miner’s rule damage
summation.

(Or fraction of fatigue damage for each repetition is 3.964x10-5)

14-22
 Fatigue and Fracture Mechanics

Example - irregular load 2

Two smooth steel bar specimens were subjected to fatigue tests under cyclic axial
stress of ±400 MPa and ±250 MPa. Specimens failure occurred after 2x104 and
1.2x106 cycles respectively, at these two tress levels. Estimate the fatigue life of a
similar specimen subjected to a stress level of ±300 MPa if it has already undergone
2.5x104 cycles at ±350MPa.

Solution:

First we calculate the S-N curve constants A and B from the two tests

We have:
Sa1 = 400 MPa Nf1 = 2x104
Sa2 = 250 MPa Nf2 = 1.2x106

New test:
Sa3 = 350 MPa n3 = 2.5x104 (already used)
Sa4 = 300 MPa n4 = ? (left until failure)

Sa = A Nf B

Since A1 = A 2

B B B
Sa1 =  Nf 1  400  2x10 
4
=  1
1. 6 =  
Sa 2  Nf 2  250  1.2x106 
   60 

B=
log1.6
= −0.1148 A= Sa1 = 400
= 1247 MPa
 1
log 
 60 
B
(
Nf 1 2x104 )
− 0.1148

14-23
 Fatigue and Fracture Mechanics

The expected failure from the first part of the test at stress amplitude of 350 MPa is

1  1 
   − 
 Sa 3  B   350  0.1148 
Nf 3 =   =  = 6.4 x 104 cycles
A  1247 

n3 2.5x 104
and the part damage is = = 0.39
Nf 3 6.4 x 104

The expected failure from the second part of the test at stress of 350 MPa is

1  1 
  − 
 Sa 4  B   300  0.1148 
Nf 4 =   = = 24.5 x 104 cycles
 A  1247 

And finally applying Miner’s rule

n3 n4
+ =1 N 4 = (1 − 0.39 )24.5x 104 = 1.49 x 105 cycles
Nf 3 Nf 4

until failure is expected in the 4th incomplete test.

14-24
 Fatigue and Fracture Mechanics

Tutorials STRESS - LIFE

1. The analysis of the cyclic stresses on steel part of an undercarriage aircraft

component shows that during each flight it is subjected to the following stress
history: N1 = 100 cycles at +/- 400 MPa and N2 = 10 cycles at +/- 500 MPa. If A
and B are 1600 MPa and –0.1 respectively, estimate the component life before
failure assuming total flights = 1/(N1/N1f+ N2/N2f). (5429 flights).

2. A stepped shaft with a smallest diameter D = 15mm is subjected to the load

history shown below. Assume that the S-N curve with a fatigue notch effect of
Kf = 1.92 included is:
S(R= -1) = 2250 Nf –0.172 MPa

A factor of safety of 10 on the life (repetitions) is required and 50 load history

repetitions are anticipated and the material has ultimate stress σu = 1000 MPa.

Estimate the number of repetitions to failure of the load history. (314).

3. A steel bar is subjected to a fluctuating axial load that varies from a maximum of
330 kN in tension to a minimum of 100 kN in compression. The steel properties
are: σu = 1090 MPa, σy = 1010 MPa and σe = 510 MPa (fatigue limit).
Determine the bar diameter to give infinite life based on factor of safety of 2.5.
(Answer: 38.7mm).

4. The stepped shaft shown in the figure below is subjected to a steady axial pull of
50 kN and a uniform bending moment M. If the yield strength of the shaft
material is 300 MPa and the fatigue limit in reversed bending is 200 MPa,
calculate the maximum value of M to avoid fatigue failure in the shaft. Kt for the
fillet radius is 1.55 and the notch sensitivity factor q=0.9. (135.1 Nm)

37.5 mm
M 25 mm M

50 kN 50 kN

3.75 mm

14-25
 Fatigue and Fracture Mechanics

5. For an S-N curve of the form σa = ANfB, two points (N1, σ1) and (N2, σ2) are
known. Develop equations for the constants A and B as a function of these
values. Considering the S-N plot on a log-log coordinates, what are the
significance of A and B?

6. A circular rod made of aircraft quality steel (σu = 1172 MPa, A=1643,
B= -0.0977 and q = 0.9) is loaded axially and has a step change in diameter.
The dimensions (see the figure overleaf (Peterson, 1980) are, d1 = 15, d2 = 18
and ρ = 1 mm.

a. Calculate safety factors in both, stress and life if during service 30000 cycles
are applied at zero-to-tension load, where the peak load is 70 kN (1.165,
4.74).

b. If a stress safety factor of 1.7 is required. Is the design adequate? If not, try to
find a suitable fillet radius.

14-26
 Fatigue and Fracture Mechanics

Fatigue Advanced Topics

Multiaxial stresses effect

So far, our analysis has been restricted to the uniaxial cyclic stress that is in the
same direction as the applied cyclic load. However, most engineering components
are subjected to complex loading and the local stress is not uniaxial, but multiaxial.
It involves more than one stress component. Many theories were developed and are
still under research to try to predict parts’ lives due to multiaxial loading.

Most of the multiaxial fatigue approaches introduce effective cyclic and mean
stresses that could be compared with the uniaxial S-N curve. One such approach is
to assume that the effective stress is related to the octahedral shear stress yield
criteria:

σa ,efective =
1
(σxa −σya ) + (σya −σza ) + (σza −σxa ) + (τ
2 2 2 2
xya + τ2 yza + τ2 zxa )
2

Where σa,effective is the effective stress amplitude and σxa,ya,za are the cyclic stress
amplitude components.

Using this criteria the effective mean stress, σm,effective is the sum of the three mean
stress components:

σm,efective = σ xm + σ ym + σzm

To obtain a completely reversed (R=-1) cyclic stress, a mean stress theory can be
used, for example the modified Goodman correction:

σa ,effective
σar =
σ
1 − m ,effective
σu

In the special case of torsion fatigue the octahedral shear stress is:

14-27
 Fatigue and Fracture Mechanics

σa ,efective = 3 τxya σm ,efective = 0

Example (Multiaxial fatigue including mean stress)

The fuselage of a medium commercial jet is pressurised and depressurised upon

every takeoff and landing. It has a life expectancy of 25 years with an average of 3
takeoffs a day. The fuselage is made of aluminium 2024-T3 and has an approximate
diameter of 4m and a wall thickness of 1”. The maximum internal pressure during
flight is about 3.9 MPa. Due to the combined fuselage and tail weights, part of the
fuselage is also subjected to a maximum constant bending moment of 3 MNm.

What is the expected life in flight cycles of the plane fuselage and how is this
compared with an estimated life, using the material S-N curve and assuming a
minimum fatigue safety factor of 2? (The 2024-T3 properties are: S-N curve
constants, A=839 MPa, B=-0.102, σu=476 MPa and α=0.51).

Solution:

We consider the fuselage to be a thin walled cylinder under cyclic internal pressure
and constant bending moment. In this case the principal stresses coincide with the
hoop and longitudinal stress directions and hence the shear stress components are
zero. We also consider the radial, through the wall thickness stress to be zero (plane
stress).

The stress due to the internal pressure fluctuates between 0 - maximum while the
stress due to the bending is a constant tension. The pressure contributes to
longitudinal and axial stress components, mean and alternating, and an additional
longitudinal mean component exists due to the bending moment.

14-28
 Fatigue and Fracture Mechanics

Cyclic stress due to pressure

Longitudinal and hoop directions
σmax

σmean

Hoop stress time

σ1a σ1m
Longitudinal Constant stress due to bending moment
stress σ2a σ2m M σmean
P
Longitudinal direction

The problem is solved by using:

♦ Multiaxial stresses – an effective octahedral shear stress.

♦ Modified Goodman for mean stress.
♦ Material S-N curve.
♦ Note that fatigue and not strength safety factor is used.

The number of cycles during the plane operation is:

365 x 3 x 25 = 27,375 flight cycles

The stress components are:

 Pmax 
 r
3.9 x 2 x 103
 2 =
σ1a = σ1m = = 153.54 MPa
 t  2 x 25.4
 
 
( Hoop due to pressure)

 Pmax 
 r
2
σ2 a =   = 76.77 MPa
 2t 
 
 
(Longitudinal due to pressure)

14-29
 Fatigue and Fracture Mechanics

 P max 
 r 6
=  2  + M max r = 76.77 + M max = 76.77 + 3000 x 10 =
σ 2m
 2t 




I πr 2 t π 2 x10 3 2
25 (
. 4 )
76.77 + 9.4 = 86.17 MPa
(Longitudinal due to pressure and moment)

σ 3a = σ 3 m = 0 (plane stress)

The effective octahedral shear amplitude is:

1
σa ,efective = (σ1a − σ 2a )2 + (σ 2a − σ3a )2 + (σ3a − σ1a )2
2

1
σa ,efective = (153.54 − 76.77 )2 + (76.77 − 0 )2 + (0 −153.54)2 = 132.97 MPA
2

The effective mean stress is:

σm,efective = σ1m + σ2 m + σ3m = 153.54 + 86.17 + 0 = 239.71 MPa

And to obtain the completely reversible effective stress:

σa ,effective 132.97
σar = = = 267.86 MPa
σm,effective 1 − 239.71
1−
σu 476

From the material S-N curve we find the corresponded life to this stress:

14-30
 Fatigue and Fracture Mechanics

1  1 
  − 
 σar   B   267.86   0.102 
Nf =   =
 A   839 

N f = 72651 (cycles to failure)

Calculated fatigue safety factor:

72651
= 2.65
27375

2.65 > 2 (as requested)

Would the design still be safe if the internal fuselage pressure was to
increase to 4.5 MPa?

Example (in class exercise)

A solid shaft with 20mm diameter is subjected simultaneously to a constant (steady)

bending of 200 Nm and alternating torsion of 220 Nm. Calculate the effective stress
amplitude for this case if σu=450 MPa.
Hints: Sbending = 32M/πd3 and τtorsion = 16T/πd3 and note the torsion case.
(Answer: 560 MPa)

14-31
 Fatigue and Fracture Mechanics

♦ The critical plane approach

The majority of the current and most accepted methods of fatigue assessments today
apply the so-called critical plane approach. Some of these applications are very
different and we only intend to introduce a few, well known, theories. The main
difference between these methods and the classical approaches is in the
understanding that the fatigue damage is related to the material micromechanical
physical process.

The fatigue process is described in two stages: In the first stage slip bands’
decohesion is primarily caused by shear and in the second stage a normal stress to
the shear planes opens and augments the shear cracks. This process can be described
schematically as shown below:

(a) (b)

γac
γac
σmaxc

In figure (a) the crack is initiated on the plane of the maximum shear stress– fatigue
stage 1. In Figure (b) the crack is open and grows due the normal stress to the plane
of the maximum shear stress – fatigue stage 2. For high-strain fatigue case this
process is modelled in terms of the maximum shear strain and the normal strain.

One of the main advantages of this approach is that any loading mode is already
considered since the fatigue damage is accumulated directly on a critical plane in the
component. However, the computation is fairly complex and only very recently
have these models been introduced commercially.

14-32
 Fatigue and Fracture Mechanics

The McDiarmid model (stress approach)

The McDiarmid parameter is based on finding a critical plane on which fatigue

failure will occur. Fatigue strength is a function of the maximum shear stress and the
normal stress occurring on the plane of the maximum shear stress. The critical plane
for failure is that subjected to the greatest value of the parameter:

σequivalent = τamp + C1σn,max

where:
τamp = maximum shear stress amplitude
σn,max = maximum normal stress occurring on the plane
of maximum shear stress amplitude,
C1 is a material constant

For cyclic conditions the McDiarmid parameter is therefore calculated using:

σequivalent = τamp + C1 (σamp + σmean )

where:
σamp = Stress amplitude normal to the plane
σmean = Mean stress normal to the plane

The maximum value of shear stress τmax is always at an angle 45° to the maximum
principal stresses, and is calculated using:

1
τmax = (σ1 − σ3 ) where : σ1 〉 σ 2 〉 σ3
2

On the plane of the maximum shear stress, the normal maximum stress is given by:

1
σ n ,max = (σ1 + σ 3 )
2

14-33
 Fatigue and Fracture Mechanics

The Brown-Miller model (strain approach)

In this theory it is assumed that failure is dominated by the maximum shear strain
γmax and the strain εn that acting normal to the maximum shear plane. The
contribution of the normal strain is assumed to be governed by a material parameter
(C) as follows:

γ max
+ C ε n = Const.
2

γ max ε1 − ε3 ε1 + ε3
where: = and εn =
2 2 2

However, this theory does not correlate well all cases of multiaxial fatigue. It may
be shown that under pure shear conditions the maximum shear strain plane is on the
surface – this results with surface fatigue cracks. Under equibiaxial strain
conditions the maximum shear strain plane is at 450 through the thickness – results
with through thickness fatigue cracks.

To improve the Brown-Miller fatigue model two cases of fatigue cracks are
considered:

TYPE A CRACKS – SUFACE CRACK (max shear strain plane on surface)

TYPE B CRACKS – THROUGH THICKNESS CRACK (max shear plane

inclined 45 degree to surface).

14-34
 Fatigue and Fracture Mechanics

Cycle Counting

Service components in many cases experience variation of load with time, results
with irregular stress history. It is therefore necessary to count individual events and
define cycles that can be used for damage summation. Many debates concerning the
summation of such events emerged in the past. Today, by far the most widely used
procedure is called the rainflow counting method. Several standards have
developed computer numerical algorithms for use of this method and an example of
small sequence counting using the rainflow counting method is illustrated below:

Definitions for irregular loading:

A I Rainflow
σ example:
G
C
0 time
E H
F J
B
D 6
Load,
2 8 stress,strain
Peaks: A, C 4
Valleys: B, D Time
7
Simple ranges:A-B, B-C 3
Overall ranges: A-D, D-G 1 5
9
a. Initial history and possible cycle.
6
2 8
4

7
3 3’
5 9
b. First cycle extraction and other possible cycle.
6 8 2 2’

7 7’

5
9
c. Second cycle extraction and last possible cycle before
the large cycle merge.

14-35
 Fatigue and Fracture Mechanics

Fatigue life prediction using finite element analysis (FEA)

The use of elastic, static, finite element analysis (FEA) to estimate the critical cyclic
stresses of service components is well established and has been widely used for the
last 30 years. However, in many cases, at the critical location, the local stress is
above the elastic limit and in order to establish the cyclic stress or strain or the
critical hysteresis loop, a non-linear elastic-plastic FEA is required. The principles
of fatigue estimate using an elastic-plastic FEA are briefly described and an example
is shown for estimation of life of a component

Typical elastic-plastic analysis required the following material input data:

1. Elastic-plastic cyclic stress-strain response.

2. Hardening low or the elastic-plastic material response upon
unloading.
3. Fatigue strain-life low such as the Manson-Coffin Basquin
relationship.
4. Multiaxial fatigue criteria such as the Brown-Miller parameter.

For the elastic-plastic FEA a non-linear analysis is performed and iterative solution
procedure is required using convergence criteria. Most modern FEA codes have
internal convergence criteria and the analysis iterates automatically until the
convergence criterion is satisfied.

Usually the loading history is sub-divided into a certain number of load increments
to improve the convergence of the non-linear iterative procedure.

A typical elastic-plastic fatigue FEA procedure is illustrated overleaf, the material

cyclic stress-strain response is shown in (1), and the stress-stain cyclic loop resulting
from loading O-A-B-A’ in (3) is shown in (2). The simulated strain results are used
as input for a strain based durability analysis that assumes that the life (Nf) of the
component is dependent on the strain range (∆ε) in a critical location (4). It is
normally sufficient to model one half of the strain range cycle by using the segment
O-A in (3).

14-36
 Fatigue and Fracture Mechanics

FEA durability (fatigue) procedure – constant amplitude load:

σ
4 5
3
2

(1) The material cyclic response is entered as a stepwise value of stress-strain or as a

non-linear equation.
σ 5 A A’
4
4 5
3 3
2 1

2 1

ε
O
2
1
3
4
5 ∆ε

(2) The material hardening model, for example, kinematics or isotropic behaviour
is selected to describe the elastic-plastic unloading part of the hysteresis loop.
Force
A A’
F
O

0.25 0.50 0.75 1 Time

-F
B

(3) Incremental loading is entered for a complete cycle.

Material
∆ε Strain-life
curve

Nf
(4) Estimation of the component life using the FEA results and strain-life curve.

14-37
 Fatigue and Fracture Mechanics

Example 2

Several years ago Rover design engineers investigated premature failure of the
suspension arm in some of the Metro cars front wheels (fiction only). The load
pattern applied to this component is complex and includes 3 directions of direct load
and at least one component of torsional load – see illustration. An experimental
programme was carried out using the service attachments with a simplified load –
only in the vertical direction. The component was strain gauged and fatigue cycled
to failure.
The laboratory load was higher than that used in service and elastic-plastic
durability FEA analysis was requested to confirm the experiments.

Vertical
Fore-Aft

Torsional

In service Lateral

Gauge 1 Load -ve

Gauge 3 Gauge 2
Bearing Pin

Experimental set up

The solution stages used are shown below:

Two FEA models were used; assuming plane stress condition or assuming plane
strain condition. The stages described previously were applied using a commercial
FEA code.

Initially, strain gauge results obtained from static elastic tests were compared to the
FEA analysis. The results are shown below for the plane-stress conditions.

14-38
 Fatigue and Fracture Mechanics

An example of the FEA analysis results:

1. Results from a simplified model using static plane stress analysis.

2. Comparison between FEA results and the experimental results using several
fatigue life prediction models (Shatil et al, 1994).

14-39
 Fatigue and Fracture Mechanics

Advanced Fatigue Tutorials

1. A cylindrical pressure vessel with closed ends is made from the alloy Ti-6Al-4V
(A=1889, B=-0.104, σu = 1233 MPa) and has a diameter of 250mm and a wall
thickness of 2.5mm.
a. What applied pressure will cause fatigue failure at 105 cycles? (14.6)
b. What is the maximum pressure applied if the vessel is designed for endurance
(assume factor of safety of 2)? (6.5)
c. Calculate the minimum cylinder thickness, designed for endurance as in b.
and using the pressure obtained at a.

2. In a paper by E K Walker, it is proposed that mean stress effects can be

estimated by using: σequivalent = σmax (1-R)m where σequivalent is an equivalent stress
range at R=0 that causes the same fatigue failure as the actual σmax for any given
mean stress, and m is a material constant.

a. Modify the above equations so it is expressed in terms of equivalent

completely reversed (R= -1) stress amplitude, σar.
b. Based on the above develop an equation for the particular case when m=0.5.

3. a) Develop expression for the elastic-plastic uniaxial case of the Brown-Miller

model if ε1, the maximum principal strain is known and assuming the following
material properties: C = 0.2, νelastic = 0.3 and νplastic = 0.5. Note also that for the
uniaxial case ε3 = ε2 = -νε1.
b) Modify the strain life equation to consider the Brown-Miller model in the
uniaxial case.

14-40
 Fatigue and Fracture Mechanics

Introduction to LEFM

Design for damage tolerance

It is generally accepted that most service components or structures contain some

form of defect or flaw by the time manufacturing is completed. Fracture control of a
structure is the concerted effort by the design engineers to guarantee safe operation
by ensuring that any crack developed from the flaws is ‘sustained’ during the
service load history.
Fracture control is the combination of measures to prevent in service fracture and
includes the damage tolerance analysis, material selection, design improvements,
experimental programme and maintenance inspection and replacement schedule.

The Damage Tolerance analysis has two main objectives:

1. Predict the effect of cracks on the structural strength (fracture margins).
2. Predict crack growth during service history (fracture control).

Practical fracture design commonly distinguishes between two types of fracture

mechanisms:
1. Brittle fracture is that associated with cleavage fracture and results with
‘sparkling’ differently orientated fractured surfaces.
2. Ductile fracture is that associated with the initiation of voids from inclusions
and results with dull looking fracture surfaces.
During ductile fracture some local plastic deformation always occurs next to the
crack tip.

In fracture analysis the applied stress (far field stress) is related to the processes
taking place at the crack tip. Three modes of fracture are distinguished as shown
below. However, in practical applications Mode 1 fracture is usually the most
important and our analysis relates to Mode 1 fracture.

y
x

Opening mode (I) Shearing mode (II) Tearing mode (III)

Modes of fracture

14-41
 Fatigue and Fracture Mechanics

The stress intensity factor K

The principles of linear elastic fracture mechanics (LEFM) are based on work by
Griffith and Irwin. Griffith has demonstrated that for a narrow ellipse in a wide
plate, where the stress concentration is define as:

Kt = 1+(2a/2b)

as the ellipse becomes narrower, or a narrow slot (resembling a crack) b Æ 0 the

normal stress increases to infinity or Kt Æ ∞. This is shown schematically below.

σ Tip singularity

2a
Normal stress
2b
Crack Crack front

This approach was further developed to the practical design of fracture in metals by
Irvin using the equations for ealstic stress distribution at the crack tip for arbitrary
stress and fracture modes as follows:

σy
y

σx
τxy

θ
crack
x

14-42
 Fatigue and Fracture Mechanics

And for Mode 1 only the crak tip stress field is a dependent on r and θ as follows:

K1 cos θ 1 − sin θ sin 3θ 

σx =
2πr 2  2 2 

K1 cos θ 1 + sin θ sin 3θ 

σy =
2πr 2  2 2 

σz = 0 (plane stress)
(
σz = υ σ x + σ y ) (plane strain )

K1 cos θ sin θ cos 3θ

τ xy =
2πr 2 2 2

For the particular case of a centre crack in infinite plate below, it can be shown that
the stress normal to the crack tip, σy, is defined by the following:

Remote field stress

Centre crack in
infinite plate
2a

Cσ a
σy =
2πx

And, for a crack through a plane with θ = 0 the stress at the crack tip is reduced to:

K
σy =
2πx

14-43
 Fatigue and Fracture Mechanics

Comparing the two equations and assuming that: C = π

(refer to literature for full solution)

We can define now the basic equation for the stress intensity factor, K:

Where a is half crack length as defined above, σ is the remote field stress and β is a
geometrical factor.

For infinite plate above: β=1

πa
For finite plate (or use the Table below) : β = sec
W

Since the equations developed above relate to Mode 1 fracture the stress intensity
factor is denoted as:

K = K1

β solutions for three common crack geometries are given below:

Centre crack in finite width plate: σ

2 3
a  a   a  2a
β = 1 + 0.256 − 1.152   + 12.20  
W W W
or W

πa
β = sec
W

14-44
 Fatigue and Fracture Mechanics

Single edge crack in finite width plate: σ

L/W = 2 only:

2 3 a
a  a   a 
β = 1.12 + 0.23 + 10.56   − 21.74   + L
W W W W
4
 a 
30.42  
W

Double edge crack in finite width plate: σ

2 3
a  a   a 
β = 1.12 + 0.43 − 4.79   + 15.46   a a
W W W
W

Notes:

The stress intensity factor K is a means of characterising the elastic stress

distribution near the crack tip but it has no physical meaning!

The stress intensity factor, K, has the units of:

MNm-3/2 or MPa m or N/mm3/2 (all same SI units!).

The fracture toughness K1c

One of the most important design implications of LEFM is the use of the fracture
toughness, K1c. Since K defines the entire stress field near the crack tip, fracture
occurs when it reaches a critical value. The value of the critical K is a material
dependent property and is determined from established experimental procedures.

14-45
 Fatigue and Fracture Mechanics

The fracture toughness is always determined for PLANE-STRAIN condition only.

If an assumed or detected defect exists in a structure, the designed stress must not
exceed a critical fracture stress for a given material. For example, for a crack at an
infinite plate:

K1c
σfracture =
πa

Typical fracture toughness values for some common materials:

Material K1C ( MPa m ) G1C (kJ/m2)

Mild steel 150 70
HSS 90 58
Cast iron 13 1.6
Titanium alloys 75 65
Aluminium alloys 34 19
Glass 0.5 0.02
Epoxy 0.6 0.2

Trends of K1c

♦ Metals typically have toughness values in the range of 20 to 200 MPa m .

♦ Polymers toughness values are in the range of 1 to 5 MPa m
♦ Ceramics toughness values are about 1 to 10 MPa m .
♦ Increase in metal strength DECREASES the fracture toughness.
♦ Heat treatment of metals DECREASES the fracture toughness.
♦ Loading rate INCREASES toughness of steel

IN GENERAL: INCREASE DUCTILITY = INCREASE in K1c

Another Note on K1c

The material K1c value is obtained under carefully controlled plane-strain test
condition and therefore it is the most conservative value of the critical fracture for a
particular material. This value is used for design against catastrophic failure and is
considered to be dependent on the material only.

14-46
 Fatigue and Fracture Mechanics

Example

Determine the maximum load that the component shown below can withstand if it
is made of Ti 6Al 4V (σy = 900 MPa, K1c = 100 MPa m1/2).

Solution:

a = 10mm W = 100mm a/W = 0.1

For a single edge notch using the geometrical solutions above:

β = 1.12 + 0.23x 0.1 + 10.56 (0.1)2 − 21.74 (0.1)3 + 30.42 (0.1)4 = 1.23

Using the fracture toughness to obtain the critical stress:

K1c = 100
σ max imum = = 458.7 MPa
β πa 1.23 π 0.01

Since 458.7 < 900 the component fracture before yield occurs.

Pmax = σA = [458.7x106] x [0.1 x 0.03] = 1.376 MN

10mm

30mm
100m

14-47
 Fatigue and Fracture Mechanics

Energy methods for fracture

Historically, the energy methods superseded the development of the intensity factor,
in particular, Griffith and Irwin models that are commonly called:

The Strain Energy Release rate, G

In the case of a simple tensile bar the stress is uniform and the total strain energy of
a member with a cross section area A and length L is simply the area under the
elastic stress-strain diagram as shown below, summed up over the entire member
volume:

P, σ

σ
P
2
L U=1/2(σ /E)A U=1/2Pδ
A
=>

ε δ

Since: δ = εL = σL/E, and P = σA for the elastic case shown.

The energy release rate due to the extension of a crack is shown below. The total
balance to keep the energy conservation criterion can be written as the positive
change in external energy due to the cracking, dF/da that is equal to the internal or
strain energy, dU/da and the fracture energy dW/da, thus, the fracture energy is:

dW d
= (F − U )
da da

14-48
 Fatigue and Fracture Mechanics

Two cases must be considered:

1. The crack is increased by ∆a but the displacement is constant (Figure A below).

In this case the load is actually dropping so no additional external work is
required. Therefore dF/da = 0 and the total energy balance is negative:

dW d
= (− U )
da da

2. The crack is increased by ∆a but the load is constant and the change is in
displacement (δ2 - δ1), Figure B below. The load P1 is doing the work equal to
the change in the strain energy:

dW 1 dU
= P1 (δ2 − δ1) =
da 2 da

Constant
load P

P P
a
a -dP
L
A
1
δ P1 (δ2 − δ1)
a+da a+da 2
-dU
Constant δ1 δ2
Displacement
δ
A – Constant B – Constant
displacement load

14-49
 Fatigue and Fracture Mechanics

From this we conclude that the change in fracture energy is always equal to the
change in the strain energy, regardless to the loading conditions or the numerical
sign!

dW dU
OR: = G=R where:
da da
dU/da = G = Energy release rate
dW/da = R = Fracture resistance

It can be shown that fracture occurs when:

dW π σ2 a
R= =
da E

Comparing this expression with the expression for the stress intensity factor:

K = σ πa

We obtain an expression for fracture due to energy calculations:

K = ER

OR, alternatively the fracture resistance due to a critical toughness is:

K 12c The critical fracture energy

R = release rate GC = R
E

14-50
 Fatigue and Fracture Mechanics

Plane stress Vs Plane strain:

The quantities of R and K can be related to the geometry as follows:

K 12c
R =
E'

Where E’ = E (plane stress, σz =0)

E’ = E (plane strain, εz =0)

1 − υ2

14-51
 Fatigue and Fracture Mechanics

Examples

a. A steel plate 750mm wide and 5mm thick with a central crack 2a = 100mm, is
pulled to fracture. Fracture occurs at 800 kN. What is the stress intensity at fracture?

The fracture stress is:

σfr = F/A = 800000/(750x5) = 213.33 MPa

For this geometry β ≅ 1

the stress intensity is:

K = β σfr πa = 213.33 πx0.05 = 84.55 MPa m

b. Use fracture energy considerations to calculate the fracture stress of the same
plate made of the same steel as above (E = 205 GPa) but with an edge crack length
of a = 75mm.

The fracture energy for this material is:

R = K2/E = (84.55)2/205x103 = 0.03487 [MPa m] = 34870 N/m

and for the new plate at fracture:

β = 1.12

(σfrβ)2 πa
R= therefore:
E

ER 205000 x 0.03487
σfr = = = 155.5 MPa
1.12 x π x 0.075
2 2
β πa

14-52
 Fatigue and Fracture Mechanics

Crack Tip plasticity corrections (Local yield)

The main limitation of LEFM is that it assumes a very localised yield near the crack
tip. This is demonstrated in the figures below:

K-controlled
2rp
h
Plastic Far-
zone field a
b

Valid LEFM

Invalidated LEFM due to

plastic zone too large

LEFM is applicable only if the limitations on (a) (crack length) (b-a), (ligament
length) and (h) (edge distance) are satisfied. This is expressed as:

2
4 K 
a , (b − a ), h ≥   (LEFM applicable)
π  σy 

14-53
 Fatigue and Fracture Mechanics

Irwin plastic zone size

The classical LEFM approach assumes singularity at the crack tip. The normal load
is asymptotic to the y-axis and goes to infinity, as the distance is closer to the crack
tip, as shown below. However, due to the material yield this does not occurs in
ductile materials and the stress follows the material stress-strain response.

To estimate the radius of the plastic zone size from the crack tip we assume a
perfectly plastic material and the material near the crack tip follows the modified
stress distribution as shown below.

σ
Elastic
Elastic – perfectly
σy plastic
ry

rp

Plane stress correction

From the previous LEFM stress field solution it can be shown that the plastic zone
size radius, ry, for plane stress conditions is as follows:

2
1  K1 
ry =
4π  σ y 

However, this assumes elastic stress distribution and from consideration of elastic-
perfectly-plastic redistribution of the stresses it is shown that the plastic zone size is
twice as large (plane-stress):

14-54
 Fatigue and Fracture Mechanics

2
1  K1 
rp = 2 ry= plane − stress
2π  σ y 

Plane strain correction

Similar analysis is carried out for a plane strain condition. In this case the plastic
zone is constrained by the triaxial stress state. For the perfectly-plastic material this
is shown to be about one third the size of that under plane stress condition:

2
1  K1 
rp = 2 ry = plane − strain
6π  σ y 

Since the redistribution of the stresses near the crack tip is higher than that under
LEFM assumption an effective stress intensity factor is calculated using an
effective crack length:

aeff = a + rp

An effective stress intensity factor is computed using the effective crack length:

K eff = βfunction of a eff σ π a eff

Since the effective crack size, aeff, is taken into account in the geometry correction,
β, an iterative solution is usually required to solve Keff. The K is initially determined
assuming LEFM solution. A first estimate of aeff from the equation above is used to
compute a new Keff. From this new Keff a new aeff is calculated. This process is
continued until consecutive values of Keff converge. Typically, only 3-4 iterations
are required.

14-55
 Fatigue and Fracture Mechanics

Limitations of LEFM

The analysis carried out in this section assumed that the crack tip behaviour is
described by what is commonly called the ‘region of K-dominance’, even if it is
beyond the region of the classic LEFM stress field.

However, if this no longer applies, the K-analysis can no longer be used and other
methods apply using ELASTIC-PLASTIC FRACTURE MECHANICS
parameters and theory. As a ‘rule of the thumb’ the stress applied must be below
80% of the fully plastic material yield.

Example (in class exercise LEFM-1)

A high-strength steel plate with a through thickness center crack length of 2a =

20mm is subjected to tension stress of 400 MPa. If the yield strength of this steel is
1500 MPa,
a) what is the plastic zone size and the stress intensity factor for the crack? Assume
plane stress and infinitely wide plate.
b) What effect does the plastic zone has if the steel used has a yield strength of 400
MPa?
c) Assume plane-strain and yield stress of 400 MPa and recalculate.

Example (in class exercise LEFM-2)

A linkage in the landing gear of a jet plane is to be fabricated from a HSS, which
could be either:
(I) An alloy steel σy = 1900 MPa, K1c = 82 MPa m
(II) Medium carbon steel σy = 1000 MPa, K1c = 50 MPa m

Which of these two steels has the better tolerance to defects in terms of toughness
and critical crack length? (Assume factor of safety of 2). How can you achieve the
same structural damage tolerance for both materials?

14-56
 Fatigue and Fracture Mechanics

Example (in class exercise LEFM-3)

During routine inspection of commercial aircraft, a 2.5mm edge crack was detected
in a 50mm wide and 100mm long panel that is part of the wing leading edge. The
crack direction was approximately normal to the operating tension load. The panel
is made of an aircraft graded aluminium alloy 7075-T651 (σy = 505 MPa, K1c = 29
MPa m ). Assume factor of safety of 2.

1. Estimate whether the panel is safe in terms of critical fracture assuming that the
maximum allowable stress is applied.
2. What will happened if the crack grows another 1mm?
3. Estimate the critical crack length (some iteration required).

Solution procedure:

1.a Determine type of crack and calculate stress intensity factor.

1.b Compare to the material toughness property.
2. Add 1mm and repeat 1a and 1b above.
3. Compare stress intensity for several values of crack length. Try to converge so
that a particular crack length yields the required material fracture toughness.

(Solution will be given during class)

Solution (LEFM-1)

a. For plane stress infinite plate (β = 1) the plasticity zone size is:

2
1  K1  1  σ 2 a  1  400 2 x 10 
rp = =   = = 0.356 mm
2π  σ y  2  σ y 2  2  1500 2 

and therefore the effective stress intensity is:

Keff = σ π aeff where aeff = a + rp

Keff = 400 π x (10 + 0.356) x 10−3 = 72.15 MPa m

14-57
 Fatigue and Fracture Mechanics

However, rearranging the plasticity correction equations we obtain a direct

relationship based on crack length and the applied stress:

1 σ2 a σ
2
Keff = σ π aeff = σ π a + =σ πa 1 +
2π σ y2 2σ y
2

4002
Keff = 400 π x 0.01 1 + = 400 x 0.1772 x (1 + 0.0356) = 72.13 MPa m
2 x 15002

b) Answer : 86.8 MPa m c) Answer : 71 MPa m

Solution (LEFM-2)

To estimate the crack length at critical fracture we use the fracture toughness
definition to obtain critical defect for each steel:

For the alloy steel:

The maximum allowable stress is: σallowable = σy/2 = 1900/2 =950 MPa.

Using the Fracture toughness equation:

For the carbon steel:

14-58
 Fatigue and Fracture Mechanics
The critical crack length is 2a, so it is 4.8mm for the first and 6.36mm for the second
material.

It shows that the medium-carbon steel is preferred in terms of damage tolerance!

To get a toughness or stress intensity K for which the alloy steel will have the same
damage tolerance:

Alternatively we can calculate a new allowable stress:

Or a larger factor of safety of:

1900/820 = 2.3

14-59
 Fatigue and Fracture Mechanics

Tutorials
1. A large sheet of aluminium alloy is loaded in tension to a stress of 200 MPa. If
its yield stress is 400 MPa and fracture toughness is 42 MPa m1/2, what is the
maximum tolerable defect size (i) Using LEFM (ii) using plasticity zone
correction? (14mm, 12.3mm)

2. A large plate carrying a tensile load contains a hole with two cracks propagating
from the hole in a direction perpendicular to the applied load, see figure below.
(Assume that β = 1.12 for edge crack plate and β = 1 for infinite plate).

σ 2r
σ
L

(a) Set up the stress equation at which the plate will fracture by both of the
following methods.
(i) Assume the stress intensity in the cracks is the same as the stress
intensity in the edge of an infinite plate whose average stress is the
same as the peak stress round a hole in an uncracked plate.
(ii) Assume that the only effect of the hole is to make the two cracks at
opposite sides of the hole appear as one long continuous crack.
(b) At what ratio r/L do both approaches give the same estimate? Which
approach is more likely to be correct for r « L?

3. The stress intensity for a partial-through thickness flaw is given by K=

σ(πa)1/2(sec(πa/2t))1/2 where a is the depth of penetration of the flaw through a
wall thickness t. A flaw is 5mm deep in a wall 12mm thick made from 7075-T6
aluminium alloy (K1c = 24 MPa m1/2) that support a stress of 172 MPa. What will
be the critical stress level for this structure and what is the maximum crack
length this structure can withstand? (Some iterations may be required). (Answer:
σmax = 171 MPa, a = 4.9mm)

14-60
 Fatigue and Fracture Mechanics

4. A ship steel has a value of Gc = 35 kJ m-2 and E = 205 GPa.

(a) What is the fracture stress in a thin plate that is 300 mm wide and that
contains a central crack 12 mm long? (617 MPa)
(b) If the crack is 50 mm long, what is the fracture stress? (214 MPa)
(c) Increasing the plate thickness to 120 mm reduces Gc to 18 kJ m-2. What is the
fracture stress for a 12 mm-long crack? (442 MPa)

5. A cylindrical pressure vessel 8m long 4m in diameter exploded at an internal

pressure of 10 MPa. The vessel is made from 20mm thick steel with the
following properties: E = 210 GPa; σy = 1.2 GPa; Gc = 120 kJ m-2. Show that the
failure is compatible with the existence of a 16mm longitudinal crack. Assume
that 2c=2a where 2c is the surface crack length and 2a is the flaw depth.

6. A cylindrical steel pressure vessel with yield strength of 360 MPa is subjected to
a hoop stress of 140 MPa. A tensile residual stress of 80 MPa can also be
assumed to be present.
The lowest service temperature of the vessel will be -25°C, under which
conditions the K1c is 44 MPa m1/2. The vessel will be designed according to a
"leak-before-break" philosophy in which a detectable leak will occur before
brittle fracture could occur. Determine the thickness of the pressure vessel based
on fracture mechanics considerations. Assume 2a=2c as above. (10.5mm)

14-61
 Fatigue and Fracture Mechanics

Fatigue Crack Growth

Introduction

• The material in the preceding lecture has dealt mainly with static or monotonic
loading. We now consider crack growth in the presence of cyclic stresses.

• About 40 years ago a linear relationships was shown to exits for most ductile
materials between the stress intensity and the rate of crack growth. Since that
times the application of fracture become an integral part of fatigue analysis in
many application of structural design.

• In our previous fatigue lectures we have assumed that no initial defect exits in the
component and therefore the analysis was solely to establish the number of
cycles to crack initiation. However, most structure could be designed to
withstand a certain level of crack growth assuming that a manufacturing defect
exists in the part.

• The same limitations of LEFM also govern the FCG analysis, this restrict the
crack tip stress field so it is dominated by elasticity considerations. This implies
that it is mainly in the HCF regime.

• A typical modern fatigue analysis may initially consists of calculation of the

number of cycles to crack initiation using the S-N diagram and in the second
stage apply FCG analysis to evaluate the number of cycles up to a critical crack
length.
Crack Length, a

Propagation
Initiation

FCG
S-N

Initial flaw Number of cycles, N

14-62
 Fatigue and Fracture Mechanics

Principles of FCG

For constant amplitude cyclic loading the range and ratio of the nominal stresses
were previously described as:

∆S = Smax - Smin and R= Smin/Smax

where Smax and Smin are the maximum and minimum nominal applied stresses.

Using the LEFM definitions for stress intensity we obtain:

K max = β Smax πa K min = β Smin πa

∆K = β∆S πa R = K min
K max

The fatigue process at the crack tip could be illustrated as follow:

1. PSB
2. Coalescent
3. Propagation
4. Blunting
5. Closure
6. Repeat

14-63
 Fatigue and Fracture Mechanics

The Paris relationships

At the basics of FCG analysis is the fundamental relationship between the stress
intensity range, ∆K and the CRACK GROWTH RATE da/dN (or how much the
crack propagates per number of cycles).

This is plotted on a log-log diagram as shown below with the rate of crack
propagation, da/dN on the abscissa and the range of stress intensity, ∆K on the x
axes as shown. The curve below, demonstrate the FCG behaviour in metals, has
typically three distinct regions:

Fracture
 da  I
Log  II III
 dN 
m
Treshold

∆Kth ∆Kc
Log (∆K)

(I) Slow growth near the threshold value, ∆Kth.

(II) Intermediate region of stable growth follows a power law.
(III) Unstable rapid growth up to a critical value, ∆Kcr.

The equation that represent the stable growth (region B) is identified with Erdogen
and P C Paris and is written as a power law:

da
= C (∆K )m
dN

The crack growth rate per cycle is measured in a well defined experiments that are
specified in standards such as the BS and ASTM and the data obtained as shown
below.

14-64
 Fatigue and Fracture Mechanics

Effects of R-ratio on the FCG

It has been shown that the Paris FCG equation above is dependent on the R-ratio, or
the ratio between maximum and minimum applied stresses, similar to the S-N curve,
and a different curve is generated for each R ratio. In many cases the curves for
different R have similar slope (for the same material). This means that they have the
same m values but different C values in the Paris FCG equation above.

The basic material FCG relation is usually obtained for R = 0 where no compression
or crack closure involved in the fracture process. A large data base have been
accumulated for many structural materials of FCG curves at different R ratios and it
is possible, in many cases, using ESDU data for example, to interpolate from those
curves a crack growth rate for a particular stress intensity range at a particular R.

Alternatively, various empirical modifications of the FCG equation were introduced

that remove the dependency on the R ratio. Two of the most widely used are the
Walker and Forman modifications shown below.
m Where c and m are material
da  ∆K 
Walker correction: =C  constants at R =0
dN  (1− R )1− γ 

where C and m are the FCG constants for ∆K = 0 curve and γ is a material property,
typically varies from 0.3 to 1. Note also that this equation gives a family of parallel
straight lines, e.g. assume a constant slope, m.

14-65
 Fatigue and Fracture Mechanics

da C (∆K )m
Forman Correction =
dN (1 − R )(K1c − K max )

Where K1c is the material fracture toughness. The Forman correction is more
generalised and could be used, in some cases, when tests under different R values
also show FCG curves slope with various slopes, m.

Stress intensity Threshold values (∆Kth)

One of the major deficiencies of the FCG equation is its dependency on an accurate
definition of ∆Kth. It is assumed that below this value cracks would NOT propagate.
The threshold values are difficult to obtain experimentally and usually data is
extrapolated. The R-ratios have a strong effect on the threshold values. Other factors
influencing the threshold values are the crack closure and the loading rate.

For example, the fatigue ∆Kth of low strength steel can be as low as 3 MPa m1/2
For high R-ratio of 0.75 while its K1c value could be as high as 300 MPa m1/2. This
is in particular important in situations of high frequency and low load, for example
rotation of a high-speed rotor. In practice, usually a cut-off stress level (or gate) is
applied in the analysis and empirical-experimental approach must be applied.

Crack closure

Our discussion so far was referred mainly to the assumption that each cycle open the
crack-tip. However, it is generally accepted that crack closure decreases the FCG
rate.

Elber introduced one of the simplest models that account for crack closure
measuring crack closure in 2024 -T3 aluminium alloy. He made the assumption that
only the portion of ∆K in which the crack is open contributes to the FCG. An
expression for the stress intensity in which the crack just opens was introduced, Kop.

Below is a simple empirical relation to obtain an effective stress intensity range that
takes into account the closure as a function of the stress ratio R based on the Elber
model.

14-66
 Fatigue and Fracture Mechanics

 1 
∆ K eff = K max − K op = K max − ∆K − 0.5 − 0.4R  (− 0.1 ≤ R ≤ 0.7 )
1 − R 

Where ∆K is the applied stress intensity range.

However, the closure also effects the threshold values, ∆Kth, and high closure
INCREASE the ∆Kth. A simple modification, using Elber work gives:

∆ K th / eff = ∆K th (0.5 + 0.4R )

FCG constants for several materials are given below:

Life estimates using FCG - constant amplitude load

The FCG equation is based on the rate of crack propagation per cycle and the stress
intensity range, ∆K. Since ∆K increases with increase in crack length we have to
integrate between the number of cycles to failure, Nf, and the initial number of
cycles, Ni, or similarly between the respective crack lengths, af and ai.

14-67
 Fatigue and Fracture Mechanics

Nf af da
∫ dN = N f − Ni = Nif = ∫
Ni ai
f (∆K, R )

To perform the integration for a particular case a specific equation for R and ∆K is
required, while some useful closed form solutions exists it is necessary in many
cases to perform a numerical integration.

For a particular R-ratio and assuming that the geometrical factor β is independent of
the crack length (incorrect in many cases) we can integrate the equation above:

af da af da af da 1
N if = ∫ = ∫ da = ∫ da
a i C (∆K )
m
(
a i C β ∆S πa )
m
(
a i C β ∆S π )
m
a (m / 2 )

If we assume that C, β, ∆S and m are all constants, the only variable is ‘a’ the crack
length and integration is straightforward, giving:

a (f1− m / 2 ) − a i(1− m / 2 )
N if = (m ≠ 2)
(
C β ∆S π )
m
(1 − m / 2)

It is possible to show that this closed form integration is highly sensitive to the value
of the assumed initial defect, ai but less sensitive to the final one, af. This is a clear
limitation of any FCG analysis: IT IS HIGHLY SENSITIVE TO THE ASSUMED
INITIAL CRACK LENGTH, ai.

14-68
 Fatigue and Fracture Mechanics

Note about the units: To avoid difficulty with units in calculating FCG life the
units quantities should correspond to the correct ∆K units. For example if [MPa
m1/2] is used for ∆K , C should correspond to [m/cycles], stress to [MPa] and crack
length, to [m].

Example (FCG and R correction)

A centre-cracked steel plate has dimensions of W=76mm and t=6mm and it contains
initial crack of length 2a=2mm. It is subjected to tension constant cyclic loading
between load values of Pmin=80kN and Pmax= 240kN.
(Material properties: σy=1255 MPa, K1c=130 MPa m1/2, C(R=0)= 5.11x10-10
mm/cycle, m=3.24, γ(Walker)=0.42)

1. At what crack length af failure is expected and is it due to plastic yield or

fracture?

For fully plastic yield conditions the crack length can be estimated for this geometry
as (b = W/2):

 Pmax 
a fy = b1 − 
 2bt σy 

Insuring values we obtain:

 240000 
a fy = 381 −  = 22.1mm
 2 x38x 6 x1255 

To obtain the fracture critical crack length we use the fracture toughness equation.
The maximum far field stress in the plate is:

Pmax = 240000 = 526 MPa

Smax =
2bt 2 x38x 6
Assuming that the geometry factor, β, is constant and is equal to 1:
2 2
1  K1C  1  130 
a fc =   =   = 0.0194m = 19.4mm
π  βSmax  π  1x 526 

14-69
 Fatigue and Fracture Mechanics

We have a fracture failure since 19.4 < 22.1. However, the real value of β is crack
length dependent. This would make the estimated final crack at fracture shorter.

Make some iterations to obtain a more accurate fracture crack length.

2. Estimate the number of cycles to fracture.

We use the Walker model for the calculations at a specific R-ratio:

Smin Pmin 80
R= = = = 0.333
Smax Pmax 240

At this point it is convenient to calculate an R dependent intercept C value since it is

assumed that the exponent m is constant, after manipulation of the Walker equation,
we obtain:

C= C1
(1− R )m (1−γ )
Where C1 is the intercept for R=0

5.11x 10 −10
C= 3.24 (1− 0.42 )
= 1.094 x 10 − 9 [mm / cycle] = 1.094 x 10 −12 [m / cycle]
(1− 0.333)
The stress range is:

∆S = Smax (1 − R ) = 526(1 − 0.333) = 351 MPa

Finally, to estimate the number of cycles to failure we assume an average β value of

1.11 during the FCG and use the closed form solution for the integration from ai to
af:
a (f1− m / 2 ) − a i(1− m / 2 ) (0.0194 )−0.62 − (0.001)−0.62
Nf = =
(
C β ∆S π ) (1 − m / 2 ) 1.094 x 10 −12 (1.05x 351x π ) (− 0.62)
m 3.24

N f = 68x 103 cycles

14-70
 Fatigue and Fracture Mechanics

Where (1-m/2) = -0.62

Note that this is an initial estimate and a refinement should be carried out
numerically with geometry factor β dependency on the crack length a.

Example (in class exercise)

A service engineer has detected a 2mm long edge crack in aerospace component
made of steel alloy that has a fracture toughness K1c = 82.5 MPa m1/2 and yield stress
σy = 750 MPa. The component is subjected to a cyclic tension stress of ∆σ = 140
MPa and the material Paris-type FCG relation is:

da
dN
( [
(m / cycles) = 0.66x10 −8 ∆K MPa m 2.25 ])
Estimate how many more cycles can this component endure before failure if for this
crack geometry β = 1.12.

Solution:

1. Calculate the final crack length.

2. Calculate the number of cycles using direct integration formula.

(Answer: 3150 cycles)

Life estimates using FCG - variable amplitude load

In cases of variable load the crack growth rate, da/dN may vary from cycle to cycle,
depending upon ∆K and R of the cycle involved and also the sequence of the cycles.
In general two approaches often used:

1. Cycle by cycle analysis

Summation of crack increments for every cycle by taking into account factors such
as history effects for example overload and retardation. This could take much
computer time since the growth rate must be calculated cycle by cycle.

14-71
 Fatigue and Fracture Mechanics

2. Block cycles analysis

Accumulation of the cycles using, for example, rainflow counting and ignoring the
history effects and calculating crack growth by adding the growth of each block of
cycles (defined by stress range and stress ratio). This approach is fast and usually
applied initially to estimate the cycles damage.

Cycle by cycle procedure:

This is in fact a numerical integration of da/dN but since dN = 1 we calculate the

crack increment, ∆a, every cycle and add it to the total crack:

m
 da   da   ∆K 
a j+1 = a j + ∆ a j = a j +   for example   = C  
1− γ 
 dN  j  dN  j  (1− R ) j

Assuming Walker correction. Note that C, m and γ are material constants while ∆K
and R are cycle dependent.

The crack length after the N cycles will be:

N N  da 
=
aN a j + ∆ = =
a j ∑ ∆a j ∑  
j=1 j=1 dN  j

Block cycles analysis:

The crack growth for each block of cycles is:

 da 
∆a block = N block  
 dN  block

And the total crack lengths:

Total Total  da 
a N = ∑ ∆a block = ∑ N block  
1 1  dN  block

14-72
 Fatigue and Fracture Mechanics

In practice, the total load history is usually unknown and therefore a mixed approach
is used. Applying repetitions of predicted load history in a specific sequence.

Sequence effects (retardation)

Unlike the random loading life to crack initiation summation procedure, where the
order of the cycles was assumed unimportant (Miner’s rule) the sequence of the
applied load has a clear effect on the rate of crack growth.

For example, a high tension overload decreases the growth rate, this is because the
overload increase the crack-tip plastic zone and more cycles needed before the crack
grows beyond this region. This beneficial effect is called crack growth retardation
and is illustrated in a centre cracked plate made of 2024 T3 aluminium below:

A – No overload

B – Overload, R= -1

C – Overload, R=0

Modelling of the retardation process is complex and is not well defined (yet)
however some simple models exists for example, Wheeler introduced a retardation
factor, φr, to obtain a modified FCG rate due to overload as follows:

 da  da
  = φR
 dN  Re tard dN

δ
 ∆a + r p, current 
φR =  
 r p, overload 
 

14-73
 Fatigue and Fracture Mechanics

Where rp,current is plastic zone for the current cycle and rp,overload is the plastic zone for
the overload cycle. δ is, in general, a material property.

Units conversion

It is customary in the literature to use the empirical units for fracture measurements
and calculations.

Conversion of the stress intensity K from Ksi in1/2 to MPa m1/2 or vice versa is:

1 Ksi in1/2 = 6.89MPa (0.0254m)0.5 ≈ 1.09 MPa m1/2

and:

1 MPa m1/2 ≈ 0.92 Ksi in1/2

Example (in class exercise)

A turbine blade is fitted into aluminium alloy disc on a rotor circumference disc as
shown in the Figure overleaf.
During assembly a 0.1mm scratch was introduced into the surface of the disk. The
rotation of the turbine causes a tension cyclic stress of 350 MPa at the plane of the
scratch.

• Estimate the number of stress cycles to fracture failure of the disk.

• Calculate the maximum allowed initial crack if the rotor is to undergo 2000
cycles during its operation.
• By what factor the rotation speed has to be increased (over-speed) for
catastrophic failure to occur in service?

(Note: the stress is proportional to the square of the rotation speed).

(Material properties: K1c=35 MPa m1/2, C(R=0)= 4x10-11 m/cycle, m=3.54, assume:
R=0, β=1.12)

14-74
 Fatigue and Fracture Mechanics

Solution procedure:

1. Estimate the final crack length using fracture toughness. Answer: 2.54mm
2. Estimate the maximum number of the cycles to failure. Answer: 3117 cycles.
3. Calculate the initial crack length using the service cycles. Answer: 0.168mm.
4. Use this critical crack to obtain the fracture stress.
5. Calculate the root square of the stress ratio, maximum to in-service stress.
Answer: 2.

Example

A mild steel plate is subjected to two different cyclic fatigue load tests as follows:
1. A constant amplitude test, σmax = 250, σmin = -25 MPa.
2. Same but with one overload cycle of twice the maximum stress after 50,000
cycles.
Estimate the number of cycles to failure of the plate using FCG considerations and
assuming initial edge crack of 0.5mm in an infinite wide plate.

(Material properties: K1c=100 MPa m1/2, C(R=0)= 6.9x10-12 m/cycle, m=3, geometry
factor β=1.12, σy=500 MPa)

14-75
 Fatigue and Fracture Mechanics

Solution:

1. We assume that we can neglect the compressive stresses. Therefore R = 0 and the
stress range is 250 MPa.

The final crack length is therefore:

2 2
1  K1C  1  100 
a fc =   =   = 0.0406m = 40.6mm
π  β Smax  π  1.12 x 250 

And the total number of cycles to failure assuming β is constant:

a (f1− m / 2 ) − a i(1− m / 2 ) (0.0406 )−0.5 − (0.0005)−0.5 − 39.7584

Nf = = =
C (β∆S π )m (1 − m / 2) 6.9 x 10 −12 (1.12 x 250 x π )3 (− 0.5) − 4.2171x10 − 4

N f = 94278 cycles

2. We use the Wheeler model for crack retardation and hence need to calculate the
crack tip plastic zone before and after the large cycle was applied, taking δ = 1.8
plane stress. (Solution to be given in class).

14-76
 Fatigue and Fracture Mechanics

Tutorials

1. A support bracket is welded to a plate as shown. A cyclic tension load at the lug
causes stress cycles of +/-50 MPa at the weld. Calculate the maximum size of
defect if the stress intensity range threshold value is, ∆Kth=1.65 MPa m1/2.

(Answer: 0.27mm).
2. If da/dN = 1x10-6 m/cycles at ∆K = 10 MPa m1/2 and da/dN = 1x10-3 m/cycles at
∆K = 100 MPa m1/2 (R=0).
a. Determine Paris equation constants (Answer: 10-9, 3).
b. Convert to inch/cycles and ksi in1/2 and determine the Paris equation constants.

3. Experimental programme was carried out on a batch of aviation grade aluminium

2024-T3 Al at two different Ratios and the following Paris equations were
derived:

da
= 1.60x 10−8 (∆K )3.59 R = 0 .1
dN

da
= 3.15x 10−8 (∆K )3.59 R = 0.5
dN

Determine the value of the intercept C and the exponent γ that can describe this
behaviour by the Walker equation. (Answer: 1.42x10-8, 0.679 and Hint: only C is a
function of R)

4. Using crack length of a = 0.1 inch, σmax = 22 ksi and R=0.2 and using Paris
equation constants obtained in 2 above calculate: (K1c=100 MPa m1/2, γ =0.6)
a. Crack growth rate predicted by the Walker equation. Answer: 0.0023mm/cycles)
b. Crack growth predicted by the Forman equation. (comment on this result)
(Answer:2.127x10-5 mm/cycles)

14-77