FFT_5

Fatigue failure theories and Design against fluctuating load (Contd.)

Category II – Designing For Fluctuating Uniaxial Stresses

This category adds the complication of mean stresses, and we will employ the modified-
Goodman diagram in addition to the (simpler) S-N diagram. Repeating or fluctuating
stresses as shown in Fig. 1 (FFT_1), have nonzero mean components and these must be taken
into account when determining the safety factor. Figures 8, 9 and 10 all show experimental
evidence of the effect of mean-stress components on failure when present in combination with
alternating stresses. This situation is quite common in machinery of all types.

Effect of Combined Mean and Alternating Stress

The presence of a mean-stress component has a significant effect on failure. When a tensile
mean component of stress is added to the alternating component the material fails at lower
alternating stresses than it does under fully reversed loading. Fig. 8 shows the results of tests
made on steels at ≈ 107 to 108 cycles (a) and aluminum alloys at ≈ 5 X 108 cycles (b) for
various levels of mean and alternating stresses in combination.

(a)

(b)

Fig. 8: Effects of Mean Stress on Alternating Fatigue Strength at Long Life (a) Steels based
on 107 to 108 Cycles (b) Aluminum Alloys Based on 5 x 108 Cycles.

1
FFT_5

The plots are normalized by dividing the alternating stress σa by the fatigue strength Sf of the
material under fully reversed stress (at the same number of cycles), and dividing the mean
stress σm by the ultimate tensile strength Sut of the material. There is a great deal of scatter in
the data, but a parabola that intercepts 1 on each axis, called the Gerber line, can be fitted to
the data with reasonable accuracy. A straight line connecting the fatigue strength (1 on the y
axis) with the ultimate strength (1 on the x axis), called the Goodman line, is a reasonable fit
to the lower envelope of the data.

The Gerber line is a measure of the average behavior of these parameters (for ductile
materials) and the Goodman line is a measure of their minimum behavior. The Goodman line
is often used as a design criterion, since it is safer than the Gerber line.

Fig. 9 shows the effects of mean stresses (ranging from the compressive regime to the tensile
regime) on failure when combined with alternating tensile stress for both aluminum and steel.
It is clear from these data that compressive mean stresses have a beneficial effect and tensile
mean stresses are detrimental. This fact provides an opportunity to mitigate the effects of
alternating tensile stresses by the deliberate introduction of mean compressive stresses.

Fig. 9: Compressive and Tensile Mean Stress Effect on Fatigue Life

Fig. 10 shows another view of this phenomenon by plotting the S-N curve (on semi-log axes)
for a hypothetical material with compressive mean stress, no mean stress, and tensile mean
stress added. The fatigue strength or endurance limit of the material is effectively increased
by the introduction of a compressive mean stress, whether applied or residual.

Fig. 10: Effect of Mean Stress on Fatigue Life

2
FFT_5

Various Failure Lines for Fluctuating Stresses

Fig. 11 shows the modified-Goodman line, Gerber parabola, Soderberg line, and
the yield line plotted on σm-σa axes. The Gerber parabola best fits the experimental
failure data and the modified-Goodman line fits beneath the scatter in the data as
shown in Fig. 8, which superimposes these lines on the experimental failure points.
Both of these lines intersect the corrected endurance limit Se or fatigue strength Sf on
the σa axis with Sut on the σm axis. A yield line connecting Sy on both axes is also
shown to serve as a limit on the first cycle of stress. (If the part yields, it has failed,
regardless of its safety in fatigue.) The Soderberg line connects Se or Sf to the yield
strength Sy and is thus a more conservative failure criterion. These failure lines are
defined by

 m 2 
Gerber parabola: a  S e 1  2  (15a)
 Sut 
  
Modified – Goodman line a  Se 1  m  (15b)
 Sut 
  
Soderberg line  a  S e 1  m  (15c)
 S y 


While the Gerber parabola is a good fit to experimental data, making it useful for the
analysis of failed parts, the modified-Goodman line is a more conservative and
commonly used failure criterion when designing parts subjected to mean plus
alternating stresses. The Soderberg line is less often used, as it is overly conservative.

Fig. 11: Various Failure Lines for Fluctuating Stresses

Creating the Modified-Goodman Diagram

Fig. 12 shows a schematic plot of the three-dimensional surface formed by the

alternating stress component σa, the mean stress component σm and the number of
cycles N for a material possessing an endurance-limit knee at 106 cycles. If we look in
at the σa-N plane as shown in Fig. 12b, we see projections of lines on the surface that
are S-N diagrams for various levels of mean stress. When σm = 0, the S-N diagram is
the topmost line, connecting Sut to Se. As σm increases, the σa intercept at N = 1 cycle
reduces, becoming zero when σm = Sut.

3
FFT_5

Fig. 12c shows projections on the σa–σm plane for various values of N. This is called a
constant-life diagram, as each line on it shows the relationship between mean and
alternating stress at a particular cycle life. When N = 1, the plot is a 45° line
connecting Sut on both axes. This is a static failure line. The σa-intercept decreases as
N increases, becoming equal to the endurance limit Se beyond about 106 cycles. The
line connecting Se on the σa axis and Sut on the σm axis in Fig. 12c is the modified-
Goodman line, taken at section G-G as shown in Fig. 12a.

Fig. 12: Effect of a Combination of Mean and Alternating Stresses

Fig. 13 shows an “augmented” modified-Goodman diagram. This is an

embellishment of the modified-Goodman line shown in Fig. 8 and 11. The yield lines
and the compressive-stress region are included. Note that this diagram usually
represents a section such as G-G from the three dimensional surface in Fig. 12. That
is, the modified-Goodman diagram is usually drawn for the infinite-life or very high-
cycle case (N > 106). But, it can be drawn for any section along the N axis in Fig. 12,
representing a shorter finite-life situation.

Fig. 13: An "Augmented" Modified-Goodman Diagram

4
FFT_5

The line CF is the Modified-Goodman line and can be extended into the compressive
region (shown dotted) based on empirical data such as those shown in Fig. 9.
However, it is conventional to draw the more conservative horizontal line CB to
represent a failure line in the compressive region. This in effect ignores the beneficial
effects of compressive mean stress.

In the tensile region, the line GE defines static yielding and the failure envelope is
defined as the lines CD and DE to account for the possibility of either fatigue or yield
failure. If the mean component of stress were very large, and the alternating
component very small, their combination could define a point in the region DEF that
would be safely within the Modified-Goodman line but would yield on the first cycle.
The failure envelope is defined by the lines surrounding the shaded area labeled
ABCDEA. Any combination of alternating and mean stress that falls within that
envelope (i.e., within the shaded area) will be safe. Combinations landing on those
lines are at failure and if outside the envelope will have failed.

In order to determine the safety factor of any fluctuating-stress state, we will need
expressions for the lines that form the failure envelope shown in Fig. 13. The line AG
defines yielding in compression and is

m a
  1 (16a)
S yc S yc

Line BC defines fatigue failure in combination with compressive mean stress and is

a  S f (16b)

Line CF defines fatigue failure in combination with tensile mean stress and is:

m a
 1 (16c)
S ut S f

Line GE defines yielding in tension and is

m a
 1 (16d)
Sy Sy

These equations are shown on Fig. 13.

Applying Stress-Concentration Effects with Fluctuating Stresses

The alternating component of stress is treated the same way as it was for the case of
fully reversed stress (see design steps of Category I). That is, the geometric stress-
concentration factor Kt is found, the material’s notch sensitivity q is determined, and
these are used in Eqn. 9b (FFT_3) to find a fatigue stress-concentration factor Kf. The
local value of σa is then found from Eqn. 10 (FFT_3) for use in the modified-
Goodman diagram.

5
FFT_5

The mean component of stress σm is treated differently depending on the ductility or

brittleness of the material and, if ductile, on the amount of yielding possible at the
notch. If the material is brittle, then the full value of the geometric stress
concentration Kt is usually applied to the nominal mean stress  mnom to obtain the
local mean stress σm at the notch using Eqn. 7 (FFT_3). If the material is ductile,
Dowling[1993] suggests one of three approaches based on Juvinall[1967] depending
on the relationship of the maximum local stresses to the yield strength of the ductile
material.

A mean stress fatigue-concentration factor Kfm is defined based on the level of local
mean stress σm at the stress concentration versus the yield strength. Figure 14a shows
a generalized fluctuating-stress situation. Figure 14b depicts localized yielding that
may occur around a stress concentration. For this analysis an elastic-perfectly plastic
stress-strain relationship is assumed as shown in part (c). Three possibilities exist
based on the relationship between σmax and the material’s yield strength Sy.

Fig. 14: Variation of Mean Stress-Concentration Factor with Maximum Stress in Ductile
Materials with Possibility of Local Yielding

6
FFT_5

Possibility I: If σmax < Sy no yielding occurs (see Fig. 14d) and the full value
of Kf is used for Kfm.

Possibility II: If σmax > Sy but |σmin| < Sy, local yielding occurs on the first
cycle (Fig. 14e), after which the maximum stress cannot exceed Sy. The local
stress at the concentration is relieved and a lower value of Kfm can be used as
defined in Fig. 14g, which plots the relationship between Kfm and σmax.

Possibility III: If the stress range Δσ exceeds 2Sy causing reversed yielding as
shown in Figure 14f. The maximum and minimum stresses now equal ±Sy and
the mean stress becomes zero, making Kfm = 0.

These relationships can be summarized as follows:

If K f  max nom  S y then: K fm  K f (17a)

S y  K f  anom
If K f  max nom  S y then: K fm  (17b)
 mnom

If K f  max nom   min nom  2 S y then: K fm  0 (17c)

The absolute values are used to account for either compressive or tensile situations.
The value of the local mean stress σm for use in the modified-Goodman diagram is
then found from Eqn. 10 (FFT_3) with Kfm substituted for Kf. Note that the stress-
concentration factors should be applied to the nominal applied stresses, be they
normal or shear stress.

The local applied stresses (with their fatigue stress-concentration effects included) are
used to calculate the alternating and mean von Mises stresses. This calculation is done
separately for the alternating and mean components σ'a and σ'm. (See equations 20a
and 20b). We will use these von Mises components to find the safety factor.

Determining the Safety Factor with Fluctuating Stresses

Fig. 15 shows four views of the tension side of the augmented modified-Goodman
diagram and also shows a combination of mean and alternating von Mises stresses at
point Z representing a part subjected to fluctuating stresses. The safety factor for any
fluctuating-stress state depends on the manner in which the mean and alternating
components can vary with respect to one another in service. There are four possible
cases to consider, as shown in Fig. 15.

1. The alternating stress will remain essentially constant over the life of the part
but the mean stress can increase under service conditions. (Line YQ in Fig. 15a)
2. The mean stress will remain essentially constant over the life of the part but the
alternating stress can increase under service conditions. (Line XP in Fig. 15b)
3. Both alternating and mean stress components can increase under service
conditions but their ratio will remain constant. (Line OR in Fig. 15c)

7
FFT_5

4. Both alternating and mean stress components can increase under service
conditions but there is no known relationship between their amounts of increase.
(Line ZS in Fig. 15d)
The safety factor for each of these cases is calculated differently. Note that Sf will be
used in the following expressions to represent either the corrected fatigue strength at
some defined number of cycles or the corrected endurance limit. So, Se can be
substituted for Sf in any of these expressions if appropriate to the material used.

Fig. 15: Safety Factors from the Modified-Goodman Diagram for Four Possible Load-
Variation Scenarios

8
FFT_5

For Case 1 Failure occurs at point Q and the safety factor is the ratio of the
lines YQ/YZ. To express this mathematically, we can solve Eqn. 16d for the value of
m@Q and divide that by m@Z .
  
 m @ Q  1  a  S y , m @ Z  m
 S 
 y 

 S   
N f  m @ Q  y 1  a  (18a)
 m @ Z m  S y 

If 'a were so large and 'm so small that point Q was on line CD instead of DE, then
Eqn. 16c should be used instead to determine the value of m@Q.

For Case 2 Failure occurs at point P and the safety factor is the ratio of the
lines XP/XZ. To express this mathematically, we can solve Eqn. 16c for the value of
a@P and divide that by a@Z .
  
 a @ P  1  m  S f , a @ Z  a
 S ut 
a @ P S f  m 
Nf   1   (18b)
a @ Z a  Sut 

If 'm were so large and 'a so small that point P was on line DE instead of CD, then
Eqn. 16d should be used instead to determine the value of a@P .

For Case 3 Failure occurs at point R and the safety factor is the ratio of the
lines OR/OZ or by similar triangles, either of the ratios m@R / m@Z or a@R /
a@Z . To express this mathematically, we can solve Eqn. 16c and the equation of
line OR simultaneously for the value of m@R and divide that by m@Z .
 m@ R 
from Eqn. 16c:  a @ R  1  S f
 S ut 
    
from line OR: a @ R   a @ Z m @ R   a  m @ R (18c)
 m @ Z 
   m 
The simultaneous solution of these gives
Sf
m@ R  (18d)
a S f

m S ut
which after substitution and some manipulation yields
m @ R S f Sut
Nf   (18e)
m @ Z a Sut  m S f
There is also the possibility that point R may lie on line DE instead of CD, in which
case Eqn. 16d should be substituted for 16c in the above solution.

9
FFT_5

For Case 4 In which the future relationship between the mean and
alternating stress components is either random or unknown, the point S on the failure
line closest to the stress state at Z can be taken as a conservative estimate of the
failure point. Line ZS is normal to CD, so its equation can be written and solved
simultaneously with that of the line CD (Eqn. 16c) to find the coordinates of point S
and the length ZS, which are


 Sut S f 2  S f a  Sut m
m @ S  

 2
S f  Sut
2 
 
S
a @ S   f m @ S   S f
Sut
ZS  m@ Z   m @ S    a @ Z   a @ S 
2 2

 m   m @ S   a   a @ S 
2 2
(18f)

To establish a ratio for the safety factor, swing point S about point Z to be coincident
with line OZS  at point S  . The safety factor is the ratio O S  / OZ.

OZ  a 2  m 2
OZ  ZS (18g)
Nf 
OZ
There is also the possibility that point S may lie on line DE instead of CD, in which
case Eqn. 16d should be substituted for 16c in the above solution. Case 4 gives a more
conservative safety factor than Case 3. The same approach can be used to obtain
safety-factor expressions for stress-component combinations in the left half-plane of
the modified-Goodman diagram.

Design Steps for Category II

A set of design steps similar to those listed for the fully reversed case can be defined
for the case of fluctuating stresses:

1. Determine the number of cycles of loading N that the part will experience
over its expected service life.
2. Determine the amplitude of the applied alternating loads (mean to peak)
and of the mean load.
3. Determine any geometric stress-concentration factors Kt at notches in the
part’s geometry.
4. Convert the geometric stress-concentration factors Kt to fatigue-
concentration factors Kf using the material’s q.
5. Calculate the nominal, alternating tensile-stress amplitudes σa at critical
locations in the part due to the alternating service loads based on standard
stress-analysis techniques and increase them as necessary with the
appropriate fatigue stress-concentration factors from Eqn 9 (FFT_3).
Calculate the nominal mean stress amplitudes at the same critical locations

10
FFT_5

and increase them as necessary with the appropriate mean fatigue stress-
concentration factors Kfm from Eqn 17 (FFT_5).
6. Calculate the von Mises stress amplitudes for the critical locations based on
their states of applied stress. Do this separately for the mean and alternating
components. (See equations 20)
7. Determine its properties of material such as Sut, Sy, Se (or S f at the life
required), and notch sensitivity q.
8. Determine appropriate fatigue strength modification factors for the type of
loading, size of part, surface, etc., as discussed in earlier lectures. Note that
the loading factor Cload will differ based on whether there are axial or
bending loads. If the loading is pure torsion, then the von Mises effective-
stress calculation will convert it to a pseudo-tensile stress and Cload should
then be set to 1.
9. Define the corrected fatigue strength Sf at the requisite cycle life N (or the
corrected endurance limit Se for infinite life if appropriate). Create a
modified-Goodman diagram as shown in Fig. 13 using the material’s
corrected fatigue strength Sf taken from the S-N curve at the desired number
of cycles N. (Note that for infinite life situations in which the material has
an S-N knee, Sf = Se). Write equations 16 (FFT_5) for the Goodman and
yield lines.
10. Plot the mean and alternating von Mises stresses (for the most highly
stressed location) on the modified-Goodman diagram and calculate a safety
factor for the design from one of the relationships shown in equations 18.

Finally we will investigate the general categories of multiaxial stresses in both fully
reversed (III) and fluctuating (IV) load cases.

Designing For Multiaxial Stresses In Fatigue

The previous discussions were limited to cases in which the loading produced uniaxial
stresses in the part. It is quite usual in machinery to have combined loads that create
simultaneous time-varying biaxial or triaxial stresses at the same point. A common
example is a rotating shaft subjected to both a static bending moment and a torque.
Because the shaft is turning, the static moment creates fully reversed normal stresses
that are maximum at the shaft’s outer fiber, and the torque creates shear stresses that
are also maximum at the outer fiber. There are many possible loading combinations.
The torque might be constant, fully reversed, or fluctuating. If the torque is not
constant, it could be synchronous, asynchronous, in- or out-of-phase with the bending
moment. These factors complicate the stress calculation.

Frequency and Phase Relationships

When multiple time-varying loads are present, they may be periodic, random, or some
combination of the two. If periodic, they can be mutually synchronous or
asynchronous. If synchronous, they may have phase relationships from in-phase to
180° out-of-phase or anything in between. The possible combinations are quite varied,
and only a few such combinations have been studied to determine their effects on
fatigue failure. Collins[1993] suggests that the assumption that loads are synchronous

11
FFT_5

and in-phase is usually accurate for machine design and usually (but not always)
conservative.

Category III – Fully Reversed Simple Multiaxial Stresses

Experimental data developed for simple biaxial stresses, such as those shown in Fig.
16, indicate that for fully reversed, simple multiaxial stresses in ductile materials, the
distortion energy theory is applicable if the von Mises stress is calculated for the
alternating components using Eqn. 14 (FFT_4). This is applicable for three
dimensional case also.

After alternating components have been increased by all applicable fatigue stress
concentration factors, the effective alternating stress σ'a can then be used to enter an
S-N diagram to determine a safety factor using

Sn
Nf  (19)
a

where Sn is the fatigue strength of the material at the desired life N and σ'a is the
effective von Mises alternating stress. The set of design steps are similar to those
listed for Category I - the fully reversed case.

Fig. 16: Fully reversed combined torsional and bending biaxial stress failures plotted on σ1–σ2
axes (Reference: Torsional Fatigue Test in Chapter 6 in Machine Design by Norton)

Category IV - Fluctuating Simple Multiaxial Stresses

Von Mises Method

Shigley [1983] and Collins [1993] recommend using the von Mises effective stress for
both alternating and mean components of applied stress in simple multiaxial stress
loading. Appropriate (and possibly different) stress-concentration factors can be

12
FFT_5

applied to the alternating and mean components of the applied stresses as described
for Category I (FFT_4). Then the von Mises effective stresses for the alternating and
mean components are calculated for a triaxial stress state using

a 
 xa   ya   
2
ya   za   
2
za   xa 
2
 2 2
 6  xy a   yz a   zxa
2
 (20a)
2

m 
 xm   ym   
2
ym   zm   
2
zm   xm 
2
 2 2
 6  xy m   yz m   zx m
2
 (20b)
2

These alternating and mean von Mises effective stresses are then used in a modified-
Goodman diagram to determine a safety factor using the appropriate version of
equations 18 (FFT_5). This approach is more conservative than the Sines method (See
Chapter 6 in Machine Design by Norton) and is thus more appropriate for situations
involving stress concentrations due to notches.

Questions

1. Solve the example 6-5 and 6-6 in Norton Machine-Design.

2. A machine component is subjected to two-dimensional stresses. The tensile

stress in X-direction varies from 40 to 100 MPa while the tensile stress in the
Y-direction varies from 10 to 80 MPa. The frequency of variation of these
stresses is equal. The corrected endurance limit of the component is 270 MPa.
The ultimate tensile strength of the material of the component is 660 MPa.
Determine the factor of safety used by the designer.

3. A transmission shaft carries a pulley midway between the two bearings. The
bending moment at the pulley varies from 200 N-m to 600 N-m, as the
torsional moment in the shaft varies from 70 N-m to 200 N-m. The
frequencies of variation of bending and torsional moments are equal to the
shaft speed. The shaft is made of steel FeE 400 (Sut = 540 MPa and Syt = 400
MPa). The corrected endurance limit of the shaft is 200 MPa. Determine the
diameter of the shaft using a factor of safety of 2.

13