3 Fatigue

You might also like

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 32

3

Fatigue

3.1 Fatigue
Fatigue is a time-dependent fracture failure in ductile material. For fatigue to take place, a crack has to be
present either on surface or in the bulk. When a tensile load is applied, the crack tip will have larger stress than the
nominal stress. If stress concentration factor is large enough, the crack tip stress may exceed yield strength.
Ensuing localized yielding will prevent a crack from growing catastrophically; it slows down crack growth. But when
another tensile load reappears, it could give impetus for the crack to grow in a different direction. But localized
yielding at crack tip will again prevent the crack from growing further. This punctuated growth cycle is repeated
many times and will make an incipient crack take a long time to grow and to reach a critical size for brittle fracture
to take place. It is possible also that the crack propagation significantly reduces the nominal cross sectional area so
that a component fails by a tensile rupture. Picture below shows the progression of crack growth across a
component cross-sectional area over two decades in a component before finally fast fracture occured. 1

Fatigue thus requires a periodic loading with stress magnitude lower than S y . Because cracks are
produced by manufacturing and handling processes, it is impossible to prevent any crack from forming. Fatigue is
therefore almost certain to occur for mechanical components, and in this sense fatigue is much more important
failure mechanism than plastic yielding and rupture.

3.2 S-N Curve

1
S. Lynch, “Some fractographic contributions to understanding fatigue crack growth,” International Journal of
Fatigue Vo. 104, 12-26 (2017).
Bending stress has one important advantage for fatigue testing. Around the circumference of a rod,
bending stress varies from tensile to compressive when a constant bending moment is applied. Under such
loading, rotating a rod (test specimen) using a motor will therefore produce a periodic stress from tensile to
compressive on any surface point targeted for fatigue testing. The test setup below adds one more advantage:
constant bending moment between the second and third bearings from the left. This constant bending moment is
a result of a unique shear force diagram: (i) positive and constant between the first and second bearings from the
left; (ii) zero shear force between the second and third bearings; (iii) negative and constant (and equal in
magnitude with (i)) between the third and fourth bearings. This rotating beam fatigue testing machine (R. R.
Moore fatigue test, ASTM Standards E466-E468) is used to establish what is called an S-N curve based on
alternating bending stress (with zero shear stress).

S-N curve of steel is plotted in a log-log scale and is composed of two segments: (i) straight line with
negative slope between 103 and 106 cycles for steel and (i) a horizontal line beyond 10 6 cycles. This S-N curve
demarcates the "no fatigue" region and another region where fatigue failure data reside. Below is an example of S-
N curve for wrought steels. The horizontal line beyond 10 6 cycles implies that if maximum bending stress (on
surface) is lower than 0.5 Su , where Su is ultimate tensile strength, then a steel component will not experience
fatigue at all. The 0.5 Su stress value for rotating bending is called endurance limit Sn . In order to generate an S-N
'

curve for steel we need only two points: 103 and 106 fatigue strengths for steels.

Not all metallic alloys have endurance limit. Below are several S-N curves for unnotched 2014-T6 wrought
Al alloy under axial loading. Each curve doesn't saturate and keeps declining beyond 10 6 cycles. Al alloys do not
have endurance limit, and to replace the 10 6-cycles fatigue strength we typically use 108-cycles fatigue strength
instead.
S-N curve can be described by a power law
β
S=α N ,
where α and β are two parameters obtained by curve fitting the power law to the two data: (i) fatigue strength at
10 cycles, i.e., S ( 10 ) , and (ii) endurance limit, i.e., fatigue strength at 10 cycles, i.e., S ( 10 ) , or its equivalent
3 3 6 6

at around 108 cycles for metals that do not have true endurance limits. From these two data, it can be shown that
if S ( 103 )=f S u and S ( 106 ) =S n, where f =0.9 typically, then

2
( f Su ) 1 Sn
α= , β = log ,
Sn 3 f Su
where the log function is of basis 10.

Below is a list of endurance limit or its equivalent from rotating bending S-N curves for different alloys. For
all these alloys: S ( 103 )=0.9 S u. Axial load can produce slightly lower values of S ( 103 ), which is around 0.75 Su,
but we shall use 0.9 Su . Note also that S ( 108 ) is actually not an endurance limit because the fatigue strength
keeps decreasing with increased number of cycles, but the 108-cycles fatigue strength is conventionally used for
metals that do not have true endurance limits.

1. Steel alloys. Endurance limit Sn' =S ( ≥ 106 )=0.5 S u .


' 6
2. Cast irons. Endurance limit Sn=S ⁡(≥ 10 )=0.4 Su.

3. Al alloys. S ( 5 ⨯ 10 )=
8
{
0.4 Su Su <50 ksi
19 ksi S u ≥ 50 ksi
. ¿

4. Mg alloys. S ( 108 ) =0.35 S u .


5. Cu alloys. S ( 108 ) =( 0.25 to 0.5 ) Su .
8
6. Ni alloys. S(10 )=(0.35 to 0.5)Su .
7. Ti alloys. Endurance limit S'n=S ( 10 6 )=( 0.45¿ 0.65 ) Su .

Different loadings have to be considered as well, even though we started with rotating bending loading. Below are
endurance limit values for steel alloys for different loading types.

 Reversed bending is regarded same as rotating bending. When we bend up and down a paper clip, the
resulting loading is reversed bending.
 Reversed axial produces a uniform stress across cross-sectional area, unlike bending that is linearly varying
radially. Therefore, it is more likely for cracks to grow and fatigue to occur when loading is axial when
3
compared with bending loading. As a result: S ⁡( 10 )=0.75 S u and
6 '
S(10 )=(0.7−0.9)S =( 0.35−0.45) S u. Quite often though we take endurance limit for reverse axial
n
'
identical to rotating bending: Sn=0.5 S u .
 Reverse torsion. For pure torsional loading, maximum distortion energy theory yields equivalent stress
σ =√3 τ , so that τ =0.58 σ . Equivalent stress σ of fatigue strength for reverse torsion is taken to be equal
' ' '

to rotating bending endurance limit in order to predict endurance limit. Therefore,


6
S(10 )=0.58⨯0.5 Su =0.29 S u, where 0.5 Su is rotating bending endurance limit. In addition,

S ( 10 )=
3 0.72 Su
{ steels
0.63 Su other ductile alloys
.

' 6
 Reversed biaxial. Endurance limit corresponds to rotating bending value: Sn=S ⁡(≥ 10 )=0.5 S u, and
3
S ⁡( 10 )=0.9 S u. The biaxial stress state is converted into equivalent stress of maximum distortion energy
' 2 2 2 1 /2
theory: σ =(σ x +σ y −σ x σ y + 3 τ ) . Note that each stress of the reverse biaxial case here has to be
alternating with a zero mean. Equivalent stress is placed in a rotating bending S-N curve in order to determine
fatigue number of cycles N for the obtained σ '.
For alloys other than steel, these steel alloys’ rules for different loading types apply equally by using correct Su .
Figure below summarizes three different loadings (bending, axial, torsion) for steel alloys.

3.3 Endurance Limit Modifying Factors


Effects of different loading types and materials on S-N curve are summarized below with additional factors
considered: load, gradient, surface, temperature, reliability.
'
Endurance Limit. 106 -cycles fatigue strength (endurance limit) Sn is a modified version of Sn due to surface factor
C S, gradient factor C G, load factor C L, temperature factor C T , reliability factor C R :

Sn=C L CG C S CT C R Sn' ,
' '
where Sn =0.5 Su is the rotating bending endurance limit for steels if there is no other specific data. Sn =0.4 S u
for cast iron if there is no other specific data. Each of these factors represents a penalty to the initial benchmark
'
bending endurance limit Sn . Sn ≡ S n (5 ⨯ 10 )=
' 8
{0.4 Su S u <50 ksi
19 ksi Su ≥ 50 ksi
for Al alloys if there is no specific data.

Sn ≡ S (10 )=0.35 Su for Mg alloys if there is no specific data. Sn' ≡ S ( 108 )= ( 0.25 to 0.5 ) S u for Cu alloys if
' 8

' 8
there is no specific data. Sn ≡ S (10 )=(0.35 to 0.5) Su for Ni alloys if there is no specific data.
S'n=S ( 10 6 )=( 0.45¿ 0.65 ) Su for Ti alloys if there is no specific data.

Numerically it does not matter much to the endurance limit Sn if two values of a factor differ by about 0.01,
although a difference of around 0.1 is noticeable if all other factors do not change. Percentage change of the
overall factor product C L C G CS C T C R is approximately equal to the sum of change percentages from each factor

d ( C L CG C S CT C R ) d C L d C G d C S d CT d C R
≅ + + + +
CLCG CSCT CR CL CG CS CT CR
so that we can now assert that a 3% change for each factor is noticeable if all five factors change simultaneously.
Otherwise, a 5% change will be noticeable if two or three factors change simultaneously. We conclude that a 0.05
factor value resolution is roughly what we need for most calculations, i.e., 0.85 and 0.83 read off a chart are
effectively equal, while 0.85 and 0.79 will be noticeable in the resulting endurance limit.
Surface Factor. C S is correlated with surface roughness, thus surface polishing (manufacturing) methods,
and tensile strength Su . Effect of environment on surface is effectively included, although we have to keep in mind
that an environmental effect is time-dependent, causing C S to change over time. Trend captured by surface factor
C S is as follows: higher tensile strength Su increases brittleness, increases a possibility of surface crack presence,
thus should reduce C S. Surface polishing methods, including corrosion effect, can affect C S irrespective of the
tensile strength magnitude. A surface treatment method that results in a smoother surface will have a higher C S.
Environmental effects has a similar trend, although their mechanisms are chemical in nature. All bending, axial,
torsional, shear loadings obey the same surface factor chart below. 2

Forging is most often conducted as a hot forging at a temperature that allows for recrystallization of the material
being forged. It is typically around 1200 ℃ for steels. Hot forging accommodates a more complex design
geometry than cold or warm forging. Examples of hot forged components are crankshaft, connecting rod, and gear.

Gradient Factor. C G is influenced by hardenability, where large specimen is more likely to have poorer
hardenability than small specimen. Large diameter specimen has less uniform ultimate tensile strength across
cross section because of its poorer hardenability, implying that the reliability of reported Su for a large diameter
specimen will be lower than that of a small diameter specimen. Thus, C G of a large diameter specimen is less than
C G of a small diameter specimen.

Assuming a round bar of diameter d , gradient factor C G changes according to these formulas

{
−0.107
0.879 d for 0.3 ≤ d ≤ 2 in.
C G= −0.157
0.91 d for 2<d ≤10 in.
or in SI unit

{
−0.107
1.24 d for 7.62 ≤ d ≤ 51 mm
C G= −0.157 .
1.51 d for 51< d ≤ 254 mm

Plot of the two SI-unit formulas is shown. Notice that C G cannot be larger than 1, so the curve with a value of 1 at
d=7.62 mm is the valid curve for the first interval 7.62 ≤d ≤ 51 mm, and the other curve is 1.51 d−0.157 , valid
for the second interval. It is therefore clear that C G ≡ 1 for d <7.62 mm. The decreasing C G with increased d
captures the hardenability effect.

2
S. A. McKelvey, A. Fatemi, “Surface finish effect on fatigue behavior of forged steel,” International Journal of Fatigue Vol. 36
Issue 1, 130-145 (2012).
Bending and torsional stresses share a common linear dependence on radial coordinate. They each have
the largest stress at the surface. If there is an incipient crack at the centre of a bent or twisted rod, then this crack
is not expected to cause a fatigue concern because bending and torsional stresses at the centre is zero. The trend
of decreasing C G with diameter d therefore affects bending and torsional stresses equally. Hence, bending and
torsional stresses use the same C G formulas above. Note that the Juvinall textbook uses discrete values of C G for
bending and torsion as shown in the table below (Table 8.1 in the Juvinall 7 th Edition.)

It is recommended to use the C G formulas instead of the table.

Axial load, however, creates a uniform tensile stress across the cross section of a rod. An incipient crack at
the centre of a rod under tension is expected to cause a fatigue concern. Axial load therefore creates a higher
probability of applying its load to a crack than bending or torsional loading. Thus, in principle C G of axial load is
lower than C G of bending or torsional load.

The table above provides value ranges for axial C G, but these table values are vague. We shall therefore
use the C G formulas for bending and torsional cases:

{
−0.107
1.24 d for 7.62 ≤ d ≤ 51 mm
C G= −0.157
1.51 d for 51< d ≤ 254 mm

for a given d and then subtract the result with 0.1 to give the axial C G value.

The formulas above are applicable equally for rotating and non-rotating torsional bars. They are also
applicable for rotating bars in bending. For non-rotating bars in bending, the diameter d in the formulas is replaced
by an effective diamater d e that depends on the bar’s cross section and shape. The following table lists how d e is
computed.
Load Factor. Load factor C L is assigned 1 for axial and bending loading.

For torsional loading, which is shear, C L accounts for the numerical factor produced by the maximum
distortion energy theory:

[ ]
1 /2
σ ' = σ xx2 +σ yy2+ σ zz2 −σ xx ( σ yy +σ zz ) −σ yy σ zz +3 ( σ xy 2 +σ yz2 +σ xz 2) .

For pure torsional load, we have σ ' =√ 3 τ ; thus shear stress is √ 3 times more potent to causing a fatigue than a
normal stress. It is equivalent to τ =0.58 σ ' so that the fatigue strength is reduced by 0.58 when a shear
(torsional) stress is applied.

They lead to the following values for load factor:

{
1 bending
C L = 1axial .
0.58 torsion

If a torsional load in a general stress state is present, then C L =1. In this case, we’ll have to obtain equivalent
stresses and compare them with the bending’s endurance limit.

Temperature Factor. C T =1 for temperatures below 30-40% of melting point in Kelvin (absolute)
temperature scale. Above the 40% treshold, a metal develops creep and quite possibly becomes reactive to
ambient as well. At that point, our fatigue theory practically is not valid anymore. For steels, the 40% threshold
temperature is around 400 ℃ . Even before reaching this temperature, elastic properties of steels will weaken so
that they will affect tensile strength and thus endurance limit. In our course though we will always assume C T =1
unless otherwise stated.
Reliability Factor. C R is correlated with reliability when more than one fatigue test is carried out in order
to produce the data for the other factors: surface, size, load, temperature. We want to specify reliability factor
more carefully when we use more than one component and want all of them to not fatigue.

Reliability Reliability Factor


0.50 1.00
0.90 0.90
0.95 0.87
0.99 0.81

We also need the 103 -cycles fatigue strength. Only temperature factor C T influences Sf =S ( 103 )
because time to complete 103 is 3 orders of magnitude shorter than the time to complete the 106 cycles. Sf is
given by the following values:

 Bending: Sf =0.9 Su CT
 Axial: Sf =0.75 Su C T

 Torsional: Sf =0.9 Sus CT =


{ 0.72 S u CT steel alloys
0.63 S u C T other ductile alloys

3.4 Effect of Mean Stress on Fatigue Strength


So far we have considered only alternating load for bending, axial, torsional stresses or their combination
in biaxial stress state.

We now want to consider a more general case, where mean load or mean stress σ m is no longer zero. The
stress amplitude of the alternating part is σ a.

From the diagram above, σ max=σ m +σ a and σ min=σ m −σ a . They give

σ max + σ min σ max −σ min


σ m= , σ a= .
2 2
Notice that σ a ≥ 0 , while the other three are arbitrary. If σ a=0 , then the loading is static. Fatigue failure under
static stress in a ductile material is equivalent to plastic failure. Therefore, fatigue strengths for different N ’s
(numbers of cycle) will converge to Su as σ a → 0. In such case, meaning of N , number of cycles, is irrelevant as
the number of cycles required to induce a plastic failure is zero when σ a=0 . As a result, incorporating mean stress
would change the S-N curve into a σ a−σ m curve below. When the curve is linear toward Su in the σ m axis, it is
called Goodman curve. We shall not use the two other curves (Gerber, Soderberg) and others available in
literature.

We recover S-N curve by setting σ m=0 ; therefore, fatigue strengths for different N ’s from an S-N curve
are points in the σ a axis. For each N , a Goodman curve is described by

σ a=S ( N ) 1−( σm
Su
, )
where S(N ) is obtained from an S-N curve. In particular, the infinite-lifetime Goodman curve is

σ a=Sn 1−
( σm
Su),

'
where Sn=C L CG C S CT C R Sn . If σ m=Su (i.e., a static load leading to a plastic failure), then σ a=0 as expected.
Thus, Goodman curves for different N ’s look like those in the diagram below. The dotted dashed lines correspond
to the yield-limit envelope. The infinite-lifetime envelope is located below the 106 line. Thus, it is possible for a
component to have an infinite lifetime and to experience plastic deformation, which will occur in triangle GAB.

Compressive mean stress is located to the left of the σ a axis. Such compressive mean stress does not
guarantee a compressive alternating stress, and the maximum stress could end up positive (tensile). Zero σ max, i.e.,
no tensile load, corresponds to the line σ a=−σ m. Because a fully compressive load will not cause fatigue, the
region below and to the left of this line will not experience fatigue. Hence, the infinite-lifetime envelope is (i) the
prism VHCGA0 and also (ii) the region below and to the left of the line σ a=−σ m. The prism captures a presence of
tensile load but not causing yielding and fatigue, while the second region captures a fully compressive load and not
causing fatigue.

Goodman curve strikes a reasonable compromise between experimental agreement and safety. Most
fatigue data lie between Gerber and Goodman curves. Gerber curve would be not sufficiently careful because it
takes the most optimistic interpretation to the fatigue data. Soderberg curve, however, would be too conservative.
In addition, linear Goodman curve is more convenient for design analysis than parabolic Gerber curve and others
available in the literature.

For a pure torsional loading, the σ m-intercept is set at 0.8 Su for steels and 0.7 Su for other ductile
metals. We then have to get fatigue strength values for different N 's from torsional S-N curve and identify them
along the σ a axis of Goodman diagram.

When we have a general biaxial (2D) stress state these are the steps for using a Goodman diagram:

1. Goodman diagram is constructed using rotating bending S-N curve.


2. Equivalent mean bending stress uses the larger principal stress, combining all mean stresses:
σm
2
2


σm 2
σ em= + τ m +( ) .
2
3. Equivalent alternating bending stress uses the maximum distortion energy theory, combining all alternating
stresses:
2 2 1 /2
σ ea=(σ a +3 τ a ) .

4. Place (σ em , σ ea ) in Goodman diagram to determine the fatigue behavior.


These steps are a simplification of a more complete method that relies on the 2D equivalent stress of maximum
distortion energy theory
2 2 2 1/ 2
σ '=( σ x +σ y −σ x σ y +3 τ ) ,

where in general each stress has both mean and alternating components. Under this general case, σ ' will thus have
mean and alternating components, which can then be placed in a Goodman diagram. This general method,
however, is often impractical for design analysis. Fatigue analysis involving a 3D stress state with mean and
alternating components is beyond the scope of the course.

3.5 Effect of Stress Concentration on Fatigue Strength


When a notch is present, stresses around the notch will be higher due to stress concentration factor K t .
Fatigue strength should therefore decrease for the region around the notch. We can therefore define fatigue
stress concentration factor K f

Unnotched Sn
Notched S n= ,
Kf
and given that the unnotched fatigue strength depends on surface, size, load, temperature, and reliability factors,
we must have
'
C L CG C S C T C R S n
Notched S n= .
Kf

Fatigue stress concentration factor combines two effects: (i) stress concentration factor and (ii) crack-
assisted fracture. K t amplifies nominal stress and depends on the size and shape of a geometrical discontinuity
such as notch. Crack-assisted fracture depends on likelihood of cracks, which can be thought of as microscopic
notches. This likelihood depends on hardness, thus Su . To express the dependence of fatigue stress concentration
factor on the two variables, we use notch sensitivity factor q so that
K f =q K t + ( 1−q ) .

q varies from 0 to 1 and can be thought of as a linear weight factor for stress concentration factor K t . If q=0 ,
then K f =1, meaning that there is no notch and endurance limit reverts back to the original expression:
'
Sn=C L CG C S CT C R Sn . At the q=0 limit, microscopic notches (cracks) determine the fatigue behavior
without any interaction with a macroscopic notch. If q=1, then K f =K t so that the macroscopic notch amplifies
the stresses that determine fatigue behavior. Thus, 1 ≤ K f ≤ K t .

We see from the figure above that in the limit of large notch radius, we have q → 1 so that K f → K t for
any Su . This is the q=1 limit, where the notch “dominates” fatigue behavior. For small notch radius, the regular
fatigue behavior shows up, which depends on Su . Combined low Su and small notch radius should produce
improved fatigue strength than combined high Su and small notch radius. This is because lower Su implies better
ductility. Thus, low tensile strength and small notch radius would make K f approach 1; this is the q=0 limit.

q plot is less readily available; thus if no plot is available and a notch has a diameter around 1 cm, it is safe
to assume that q=0.8 .

Stress concentration factor effect is ignored for the 103-cycles fatigue strength because the shorter
fatigue lifetime.

3.6 Effect of Residual Stress on Fatigue Strength


When calculated tensile stress exceed yield strength S y , compressive residual stress is produced. If we
assume an ideal plastic material, where plastic stress reaches a plateau at yield strength, then residual stress is
given by

σ residual=S y −σ calculated .
When the calculated stress is a function of time, residual stress preserves the difference between maximum and
minimum stresses:
σ max−σ min =2 σ a =constant ,

because residual stress is a static stress. Hence, residual stress has no effect on alternating stress σ a. Figure below
shows an example of having calculated stress exceeding yield strength of 300 MPa. Maximum stress remains at
300 MPa as a result of the ideal plastic material assumption. Resulting residual stress shows up as a negative stress
and as a result preserves the difference between the maximum and minimum stresses.

Mean stress σ m is reduced, however, by residual stress, as can be observed in Figure above: mean stress
is trending downward. The stress pinning at yield strength means we have a constraint

σ m+ σ a =S y
when residual stress is at work. This constraint is shown as a dashed-dotted straight line in the diagram below,
connecting the point ( σ m , σ a )= ( 0 , S y ) on the vertical axis and ( σ m , σ a )= ( S y ,0 ) on the horizontal axis. Because
residual stress preserves alternating stress, a calculated stress point b is moved horizontally to another stress point
b ' that intersects with the σ m+ σ a =S y line. The stress point b ' is the actual stress point. The move b → b' means
that residual stress improves fatigue life. Same conclusion is arrived at for the other calculated stress points c , d .
Problem 7
A straight 19-mm-diameter steel shaft is subjected to a cyclical torsional loading. Given its 600 MPa ultimate
tensile strength and C S=0.9, determine the maximum torsion so that the shaft has an infinite lifetime. Use
C T =1 and C R =1.

Fatigue strength corresponding to the infinite lifetime is given by the formula

Sn=C L CG C S CT C R Sn' ,
'
where Sn is the 106-cycles fatigue strength value based on the reference rotating bending test. Because generic
'
steel material is stated without further specification, we use Sn =0.5 Su =300 MPa.

Three of the five fatigue factors are given: surface factor C S=0.9; temperature factor C T =1 ; reliability factor
C R =1. Gradient factor C G is related to the specimen size represented by the d=19-mm diameter. We use the
following formula valid for 7.62 ≤d ≤ 51 mm:

C G=1.24 d−0.107=1.24 ×19−0.107=0.90 ,

while the load factor C L is 0.58 for the pure torsional loading. Hence the infinite-lifetime fatigue strength is

Sn=0.58× 0.90 ×0.90 ×1 ×1 ×300=140.94 MPa .


Notice that the 140.94 MPa fatigue strength that gives an infinite lifetime is 23.5% of the 600 MPa tensile strength.

Maximum torsion T (i.e., torque) is obtained by equating the maximum torsional shear stress
16 T
τ max= 3
,
πd
that occurs at the shaft surface. With d=0.019 m, we obtain τ max=0.7425T MPa so that equating it with the
140.94 MPa fatigue strength we obtain T =189.82 N-m torque (torsion). It is noted that a safety factor of 1 is
used here as it is not specified.
Problem 8
A 40-mm diameter, commercially polished steel shaft is subjected to a torsional cyclical loading. Its ultimate tensile
strength is 900 MPa. Determine the 103-cycles and the 106-cycles fatigue strengths (endurance limits) with a 99%
reliability.

Let us first determine the 106-cycles fatigue strength. Generic steel grade is implied in the problem
'
statement; therefore the reference bending fatigue strength (endurance limit) is Sn =0.5 Su =450 MPa. The load
factor is C L =0.58 due to the pure torsional loading. The gradient factor is
−0.107 −0.107
C G=1.24 d =1.24 × 40 =0.84 .
The 40-mm diameter has a smaller load factor than that of the 19-mm diameter in Problem 13. The smaller load
factor is caused by a larger diameter that has a larger variation in ultimate tensile strength. Steel alloys undergo
heat treatment to harden (strengthen) its surface which is achieved by fast quenching. Such heat treatment
involves a series of furnace heating and oil or water quenching. For a larger-diameter specimen slower quenching
prevails in the bulk, further from the surface, due to slower heat conduction. It is much easier for a smaller-
diameter specimen to achieve a uniform cooling rate throughout its cross sectional area; therefore, a smaller-
diameter has a more uniform tensile strength throughout its cross sectional area.

Surface factor is C S=0.9 as determined from the chart below for Su=0.9 GPa and commercially
polished surface.

Temperature factor is C T =1 , while reliability factor is C R =0.81 for the 99% reliability requirement
from the table below. Reliability factor decreases with increased reliability because a smaller reliability factor
implies a lower fatigue strength and thus a lower load to be applied to a specimen in order to achieve the same
lifetime. The lower load ensures a lower number of specimen failures, thus a lower failure probability.
Therefore, the 106-cycles fatigue strength is

Sn=0.58× 0.84 × 0.9× 1× 0.81× 450=159.83 MPa .


It is noted that 159.83 MPa represents about 18% of the 900 MPa tensile strength. Problem 13 gives a 23.5%
percentage due to its lower reliability.

We now determine the 103-cycles fatigue strength using the formula below for steel alloys

Sf =0.9 Sus CT ,

where Sus is ultimate shear strength, which is equal to 0.8 Su . Therefore, we obtain Sf =S ( 103 ) =648 MPa . A
thousand cycles for a rotating equipment occurs usually under a minute so that other fatigue factors such as load,
gradient, and surface are not relevant because these factors operate over a much longer period. For pure torsional
loading, a fatigue failure under a thousand cycles must occur at the specimen’s surface.
Problem 9
A 19-mm diameter, mirror-polished steel shaft is subjected to a torsional cyclical loading of 75 N-m in amplitude.
Its ultimate tensile strength is 840 MPa. Determine the fatigue stress at 10 5-cycles with a 99% reliability. What
should be the shaft radius if a safety factor of 1.5 is applied?

Obtaining the fatigue strength at the 105-cycles requires knowing S-N curve S=α N β, where α and β are
two parameters of the curve, which are determined from curve fitting using the 10 3- and 106-cycles fatigue
strengths: S ( 103 )=f S u and S ( 106 ) =C L CG C S C T C R S n' . The curve fitting will yield

2
( f Su ) 1 Sn
α= , β = log 10 .
Sn 3 f Su

The 103-cycles fatigue strength is calculated below. With C T =1 and Su=840 MPa, we obtain

S ( 10 )=0.9 S us C T =604.8 MPa


3

as Sus =0.8 Su .
'
We next compute the 106-cycles fatigue strength. Sn =0.5 Su =420 MPa. C L =0.5 for the pure
−0.107
torsional loading. C G=1.24 × 19 =0.90 for the 19-mm diameter. The mirror-polished surface and
Su=840 MPa give C S=1 . C T =1 due to the no mention of specific temperature requirement. C R =0.81 for the
99% reliability. Thus,

Sn=0.58× 0.90 ×1 ×1 ×0.81 × 420=177.58 MPa .


We can now compute the two parameters:

604.8 2
α= =2059.82 MPa ,
177.58
1 177.58
β= log 10 =−0.1774 .
3 604.8
Fatigue strength at 105 cycles can be computed:
−0.1774
S ( 105 )=2059.82× ( 105 ) =267.2 MPa .
Equating the maximum torsional shear stress at the shaft surface

16 T
τ max= ,
π d3
where T =75 N-m, with the 105-cycles we have

16 ×75 267.2× 106


3
= ,
πd 1.5
where the 1.5 safety factor was applied. Hence, d=0.0129 m or 1.29 cm. Hence, the 19-mm diameter of the
shaft already satisfies the 1.5 safety factor even though the 19-mm diameter is definitely an oversized (thus
corresponding to a higher safety factor).
Problem 10
A cantilever beam serves as a spring for a latching mechanism. When assembled, the free end is deflected 0.075
in., which corresponds to a force F of 8.65 lb. When the latch operates, the end deflects an additional 0.15 in.
Would you expect eventual fatigue failure?

The 0.075 in. is the initial constant deflection of the cantilever beam, which produces the force F
indicated in the diagram above. We thus expect constant bending stress to be produced by the 0.075 in.
deflection. When the additional 0.15 in. deflection occurs, there will be alternating bending stress on top of the
constant bending stress.

Because the force F is linearly proportional to the deflection, the force produced by the additional 0.15 is

F 0.075+0.15
= ,
8.65 0.075
giving F=25.95 lb.

Another way to find F=25.95 lb. is much longer and requires us to determine the beam’s spring constant. For
completeness, this second way is detailed here. The spring constant is found first using the beam deflection
formula
3
PL
δ= ,
3 EI
where δ is the beam tip deflection. This formula is available in Appendix D of the Juvinall textbook. P is the force,
which is denoted as F ; E is Young’s modulus; I is bending area moment

1 3 −5 4
I= ×0.75 × 0.1094 =8.18 × 10 ¿ ;
12
and L=4 in. The spring constant is thus

3 EI −6
k= 3
=3.834 ×10 E ,
L
where the Young’s modulus E is of high carbon steel with 490 BHN (Brinell hardness number). Various steel alloys
have an almost identical Young’s modulus of E=30 Mpsi. From materials science perspectives, the constant
Young’s modulus despite changing alloy compositions and heat treatments for different steel alloys comes from
the fact that Young’s modulus depends on average atomic-scale deflections which are dominated by the iron
atoms. Hence, k =115.02 lb/in. Therefore, the total deformation of 0.075+0.15=0.225 in. will produce 25.9
lb force, which is practically identical to the first method using the linear proportionality.

The maximum bending stress occurs at the beam surface and it is equal to

M ( h /2 )
σ= Kf ,
I
where K f is fatigue stress concentration factor and M is bending moment. The bending moment is
8.65 × 4=34.6 lb-in at the lowest force and is 25.95 × 4=103.8 lb-in at its largest. The average bending
moment is thus

1
( 34.6+103.8 )=69.2lb ∙∈,
2
while the amplitude of the alternating bending moment is

1
( 103.8−34.6 )=34.6 lb ∙∈.
2
Fatigue stress concentration factor requires knowing notch sensitivity factor q :

K f =1+ ( K t −1 ) q ,

where K t is static stress concentration factor. K t =1.7 is given already, so we need only to determine the notch
sensitivity factor q that depends on fillet radius and tensile strength. The fillet radius of 0.125 in. is regarded as the
notch radius. Using the chart below we find q ≈ 0.97 (shown as the black circle) given that the 490 BHN gives

Su=500 × 490=245 ksi .


Thus,

K f =1+ ( 1.7−1 ) 0.97=1.68 ≈ 1.7 .

So there is virtually no change from K t .

The average bending stress from the average bending moment M m=69.2 lb-in is

0.1094
69.2 ×( )
M ( h /2 ) 2
σ m= m Kf = 1.7=78.67 ksi ,
I 8.18× 10−5
while the alternating bending stress from the alternating bending moment M m=34.6 lb-in is

0.1094
34.6 ×( )
M a ( h /2 ) 2
σ a= Kf = 1.7=39.33 ksi .
I 8.18× 10
−5

We now need to determine the 106-cycles fatigue strength. The 490 BHN predicts an ultimate tensile
strength of 245 ksi so that Sn =0.5 Su =122.5 ksi. Load factor is C L =1 for bending. Gradient factor is
'

1/ 2
determined by obtaining first the equivalent diameter d e =0.808 ( 0.1094 × 0.75 ) =0.231 in., which is less
than the minimum diameter of 0.3 in. required to use the gradient factor formulas

C G= {
0.879 d−0.107 for 0.3 ≤ d ≤ 2 in.
0.91 d
−0.157
for 2<d ≤10 in.

Thus, we set C G=1. Surface factor is C S=0.55 for the machined-finish surface and 245 ksi ultimate tensile
strength.
Taking the temperature and reliability factors as 1, the 10 6-cycles fatigue strength is thus
'
Sn=S n C L C G C S C T C R=122.5 ×1 ×1 ×0.55 × 1×1=67.4 ksi,
which defines the limit for the alternating stress, while the limit for the mean stress is the ultimate tensile strength
of 245 psi. The straight-line equation connecting these two limits is

σ a=67.4 1−( σm
245
.)
For the operating mean stress of 78.67 ksi, the alternating stress value at the Goodman line is

σ a=67.4 1−( 78.67


245 )
=45.8 ksi ,

which is larger than the operating alternating stress of 39.33 ksi. Thus, we don’t expect eventual fatigue failure.
The Goodman diagram below shows the load point in relation to the Goodman line.
Problem 11
A bevel-gear shaft mounted on two 40-mm 02-series ball bearings is driven at 1720 rev/min by a motor connected
through a flexible coupling. The figure shows the shaft, the gear, and the bearings. The shaft has been giving
trouble—in fact, two of them have already failed—and the downtime on the machine is so expensive that you
have decided to redesign the shaft yourself rather than order replacements. A hardness check of the two shafts in
the vicinity of the fracture of the two shafts showed an average of 198 Bhn for one and 204 Bhn of the other. As
closely as you can estimate the two shafts failed at a life measure between 600,000 and 1,200,000 cycles of
operation. The surfaces of the shaft were machined, but not ground. The fillet sizes were not measured, but they
correspond with the recommendations for the ball bearings used. You know that the load is a pulsating or shock-
type load, but you have no idea of the magnitude, because the shaft drives an indexing mechanism, and the forces
are inertial. The keyways are in. wide by in. deep. The straight-toothed bevel pinion drives a 48-tooth bevel
gear. Specify a new shaft in sufficient detail to ensure a long and trouble-free life.

In a problem like this it’s important to identify loads acting on a component first. The 16-teeth bevel gear are of
straight-toothed type so that it produces a radial load to the shaft (i.e., tangential load with respect to the bevel
gear circle). The product of the radial load and the bevel gear radius gives the torque that produces a torsional
shear stress to the fillet of the right ball bearing where the failure occurs. The radial load also produces a bending
stress at the fillet.

The fillet bending stress is time-dependent because the shaft rotates while the gear-to-gear contact line is fixed in
space. In contrast torsional shear stress is constant. Both stresses are evaluated at the shaft surface where the fillet
is located. Because there is practically zero axial force acting on the shaft, the shaft tensile stress is negligible. The
time-dependent bending stress means that the shaft failure at the fillet is fatigue failure. Because the mean stress
is not zero, then we have to deal with Goodman diagram.

The shaft drives an indexing mechanism, which is a quick, precise rotational motion control. The radial load from
2 v0
the bevel gear is thus pulsating and inertial. The impact force formula F= √ mk cannot be used because the
π
shaft is viewed as a torsional spring. The impact force formula comes from a mass-spring model which doesn’t
correspond to a torsional spring model.

We have to use conservation of energy to determine the radial load:


2
1 2 τ V
Iω = ,
2 2G
where I , ω , τ ,V , G are respectively, moment of inertia, angular velocity, torsional shear stress, shaft volume,
and shear modulus. The indexing mechanism produces a high angular acceleration for the shaft to arrive at some
angular velocity, and when the shaft stops rotating due to a high angular deceleration it will convert its kinetic
energy into a torsional shear stress.

Total shaft and components weight is approximately

0.282 lbf/in ×
3
((
π 40 mm 2
4 25.4 mm
π 2
)
∙ 14 .5+ ( 4 −1.5 ) ∙ ( 2+ 1 ) =17.1 lbf,
4
2
)
where 0.282 lbf/ in. 3 is weight density of steel. The first term in the brackets corresponds to a straight shaft of
length 4 +6.5+2+2=14 .5 in. and a 1.57-in. diameter; this straight shaft approximates the shouldered shaft and
leaves out the bevel gear which is represented by an annulus of a 1.5-in inside diameter, 4-in. outside diameter,
and 3-in. long. Thus,

(( ) )
2
1.57
+2
2
mR 1 2
I= ≈ ( 17.1 lbm ) =16.6 lbm- in 2 ,
2 2 2

where the bevel gear has a diameter of 16 teeth/ ( 4 teeth/in. ) = 4 in. so that its radius is 2 in. The radius for
computing I is the average between the 40-mm diameter (i.e., 1.57-in. diameter) and the 4 in. diameter.

Because 1720 rpm is equal to 180 rad/s, the shaft kinetic energy is therefore 269,273 lbm- in 2 /s2 or 58.13 lbf-ft.
With the shaft volume of 60.64 in.3 from the weight calculation above, and G=11,500 ksi, the torsional shear
stress is

τ =√ 2 KG / V =
√ 2∙ 58.13 ∙12 in/ft ∙ 11,500,000
60.64
=14,848.6 psi ,

1 2 1 lbf
where K= I ω is the kinetic energy. Recall that 1 lbm= 2 so that
2 32.17 ft/ s
269,273 lbm−¿2 /s 2
=58.13 lbf ∙ ft .
2 ft
( 12∈¿ ft ) ×32.17 lbm ∙ 2 /lbf
s
The torque required to produce such stress is


T= =11878.9 lbf-in ,
R
4
1.57
where J=π ∙ =0.6 in. 4 and R=0.75 in. The radial load acting on the shaft is therefore
32
11878.9 lbf-in
F= =5939.5 lbf,
2 in.
where the 2 in. is bevel gear radius. This radial load has a large magnitude even though it produces a relatively low
kinetic energy. This is due to the pulsating character of the shaft rotation.

Nominal bending stress is


(5939.5 × 3 ) 0.75
σ= 4
=53777.1 psi,
π ∙0.75 /4
where the 3 in. is the moment arm which is the distance between the fillet and the bevel gear.

Standard fillet radius for 40-mm bore single-row 02-series deep-groove ball bearing is 1.0 mm. Thus,
r 1 D 40
= =0.026 , = =1.05 , where 38.1 mm is the 1.5-in. shaft diameter. Static stress concentration
d 38.1 d 38.1
factor for bending is thus 2.1, while for torsion 1.5. Hence if we simply use the static stress concentration factors,
without converting them into the fatigue stress concentration factors, the alternating stress is

σ a=53.8 ×2.1=113 ksi,


while the mean stress is

τ m=14.8 ×1.5=22.2 ksi.


The mean stress is converted into an equivalent mean stress using the maximum principal stress

σm
2
2
σ em= + τ m +
2√
σm 2
, ( )
which yields σ em=τ m=22.2 ksi. The alternating stress is converted into an equivalent alternating stress using the
maximum distortion energy equivalent stress
2 1/ 2
σ ea=( σ a +3 τ a )
2
,

yielding σ ea=σ a=113 ksi.


'
Our next step is to determine the endurance limit Sn=C L CG C S CT C R Sn . With an average hardness of about
200 BHN, the tensile strength is 0.5 ×200=100 ksi. The various factors are

C S ≈ 0.8 ;
−0.107
C G=0.879 ∙ 1.5 =0.84 ;
C L =C T =1 ;

and C R =0.814 for 99% reliability. Reliability factor C R needs to reflect an almost 100% reliability because two
'
shafts had failed. With Sn =0.5 Su =50 ksi, we obtain Sn=1× 0.84 ×0.8 ×1 ×0.814 ×50=27.4 ksi.

Using Goodman line equation

σ a=27.4 1−( σm
100 )
ksi ,

we find σ a=21.3 ksi for σ em=22.2 ksi, showing that fatigue failure should occur since the maximum no-fatigue
alternating bending stress of 20.7 ksi is smaller than the expected 113 ksi.

Because variables in conservation of energy cannot change a lot, τ will not change significantly. So we can only
change, i.e., reduce, bending stress by two ways: (i) move bevel gear closer to the right ball bearing in order to
reduce the 3-in. moment arm, and (ii) increase fillet radius to reduce stress concentration factor, thus using a non-
standard fillet radius for ball bearing.

To ensure an infinite life, the 3-in. moment arm can be reduced to around 0.5 in. by removing the annulus
behind the bevel gear shown above. This reduction will reduce the bending stress by 6 times to 9 ksi. With the 2.1
stress concentration factor, the fillet bending stress will come out to be 18.9ksi. This gives a rather slim safety
21.3
factor of about =1.13. Obtaining a higher safety factor though would require an increase of fillet radius.
18.9
Note. A more detailed method of solution will involve the fatigue stress concentration factor K f given by

K f =1+ ( K t −1 ) q ,

where q is notch sensitivity factor, which has to be determined from the graph below. Recall that the Brinell
hardness number of the shaft is 200 BHN so that Su=500 × 200=100 ksi. With the notch radius of 1 mm, we
find from the graph below that q=0.75 so that

K f ,bending =1+ ( 2.1−1 ) 0.75=1.83


and

K f ,torsion =1+ ( 1.5−1 ) 0.75=1.38 .


The equivalent alternating bending stress is thus lower than the previously value predicted using the static stress
concentration factor K t =2.1. It is now equal to

σ ea=53.8 ×1.83=98.5 ksi,

while the equivalent mean stress is also lower than the previous value of τ m=24.5 ksi. It is now equal to

σ em=14.8 ×1.38=20.4 ksi.


The conclusion reached remains the same: fatigue failure will occur because the load point
( σ ea , σ em )=( 98.5,20 .4 ) ksi is outside the infinite-lifetime region as demarcated by the Goodman line equation

σ ea=27.4 1−( σ em
100 )
ksi .

The design change recommendations are therefore the same.


Problem 12
The pump is driven by the gear at uniform load and speed. The shaft is supported by bearings mounted in the
pump housing and has a 1000 MPa (ultimate) tensile strength and a 800 MPa yield strength. The gear has the
tangential, axial, and radial loads with their corresponding fatigue stress concentration factors. The shaft fillet is
shot-peened and is taken to have a mirror-polished surface. Estimate the safety factor with respect to eventual
fatigue failure at the fillet.

The 750 N will create a bending moment in a radial direction of the shaft. The 500 N will also create a bending
stress in the same direction. They can be summed directly to give a total radial bending moment in that direction.
The 2000 N will create a bending moment also in a radial direction but perpendicular to the first two. Because
bending moment is a vector quantity, these two perpendicular components have to be added like vectors so that
the maximum bending stress is

4 × ( √( 500 × 0.125+ 750× 0.05 ) + ( 2000 ×0.05 ) )


2 2
4M
3
K f= 3
2.0=184.4 MPa.
πR π 0.0125
Only the 2000 N force will create a torsional moment (torque) to produce a torsional shear stress to the shaft. The
maximum torsional stress is

2T 2 ×2000 ×0.125
3
K f= 1.5=122.22 MPa ,
πR π 0.01253
while the maximum axial stress is

P 500
K f= 1.8=1.84 MPa.
A π 0.0125
2

The maximum (transverse) shear stresses from the 2000 and 750 N forces is
4 V 4 √ 20002 +7502
= =5.80 MPa.
3 A 3 π 0.01252
There is no information on the fatigue stress concentration factor for the shear stress. However, even with such
factor equal to 2, the shear stress and the axial stress are about 2 orders of magnitude smaller than the torsional
stress and the bending stress. In addition, the location of maximum shear stress is different from those of
maximum bending and torsional stresses. We therefore neglect the shear and axial stresses in the subsequent
analysis.

At any surface point along the fillet's circumference, the torsional stress is the mean stress, while the bending
stress is the alternating stress. The equivalent mean stress is given by

σm
2
2
σ em= + τ m +
√σm 2
2 ( )
=122.22MPa

since σ m=0. The equivalent alternating stress is

σ ea= √σ a2+ 3 τ a2=184.4 MPa

since τ a=0.
'
The 106 -cycles fatigue line is determined by the endurance limit and the tensile strength. Sn =0.5 Su =500 MPa.
Load factor is C L =1.0 because the combined stresses are now cast in equivalent stresses. Gradient factor is
=0.88. Surface factor is C S=1 for the mirror-
−0.107
based on the 25-mm diameter information: C G=1.24 × 25
polished surface. Temperature and reliability factors are taken as one. Thus, Sn=500 MPa ×0.88=440 MPa.

Thus, the 106 -cycles fatigue line is described by

σ a=440 1− ( σm
1000
. )
The operating coordinates ( σ m , σ a )= (122.22 , 184.4 ) MPa are clearly a point within the infinite lifetime
triangle; thus, the shaft will not experience fatigue failure. Safety factor can be estimated by assuming that the
mean and alternating stresses fluctuate similarly for the risk. We first need to find the intersection point between
the fatigue line and the load line

σ a 184.4
= .
σ m 122.22

It is located at ( σ m , σ a )= ( 225.8 ,340.7 ) MPa , so that the safety factor is

√225.82 +340.7 2 =1.85 .


√122.222 +184.4 2

You might also like