Dr Andrei Lozzi

Machine Design & CAD

Mech Design II

School of Aerospace
Mechanical & Mechatronic
Engineering

Shafts and Axles

Shaft lecture AL 2011

Refs: R L Norton, Machine Design , Prentice Hall.

N Reshetov, Machine Design, MIR Pub, Moscow.
J E Shigley et al, Mechanical Machine Design, ed 4 to 7, McGraw Hill.

1 Introduction Shafts and axles have very wide range of applications, consequently although there are
specific similarities, they can also vary widely in appearances. Figure 1 gives an idea of what they can
look like. There is a categorical difference between shafts and axles, it is that shafts rotate about their
long axis, whereas axle remain stationary and provide a pivot for other components to rotate on .
At times axles have been considered as just structural members and the methods of analysis that has
been developed for shafts have not been applied to them. Yet the effective difference between typical
examples of each, is that axles may be just subjected to fewer load cycles than shafts. The analysis then
can be the same for both, and in these notes the name shaft will stand for both items, unless for some
particular reason the need arises to differentiate between them.

Fig 1. a) Plain constant diameter transmission shaft, with or without keyways. b) & c) Machined
stepped shafts for machine tools or gear boxes, with screw threads, keyways and ground lands to
possibly carry bearings or gears. d) A large steam turbine shaft, e) crankshaft, f) rotating railways shaft,
often referred to as an axle. g) Fixed axle as used on heavy trucks, trailers and aircraft landing gear.

2 An Overview of the analysis. There are two objectives to be kept in mind in the analysis of most
shafts, The stress levels have to kept sufficiently low to provide the required service life, and the
various deflections have to be controlled to ensure that the shaft can function as expected. The loads
that usually produce the greatest stresses and strains in shafts are moments, caused by transverse forces
and reactions on the shaft, and the torque transmitted by the shaft. Secondary loads may be longitudinal
trust forces down the along axis of a shaft and contact stresses caused possibly by interference fits. Both
the maximum allowable stresses and maximum allowable deflections are used separately and
independently to determine shaft diameters. It is common that deflections requirements dictated larger
shaft diameters than the diameters required to give adequate fatigue life.

R2

R3

R1

F3
F2

F1

a b
c
d
Fig 2 Above is almost a caricature of a loaded shaft. At station a there is a sprocket carrying the
tension F1 from a chain. Bearings at b and d locate the shaft but they fix the shaft at only those stations.
The helical gear at c transmits a transverse force F2 and an axial force F3 to the shaft. The shaft is of
greater diameter between bearings to reduce deflections there. The bending moments nearly completely
determine the lateral deflections at the gear and angular deflection at the bearing d. Note that single
bearings should only be subjected to combinations of transverse and axial forces, no bending moment.
3 Stresses and strains. Stress analysis determines shaft diameters at one location at the time. It uses
moments, torques, surface finish, stress concentration and others factors, all of which must apply to the
location of interest, to determine the diameter of the shaft only at that location. In contrast deflections
analysis requires the overall shape and size of the shaft, as well as all the loads on it, be known
before the deflections of the shaft can be calculated. Hence the normal procedure is to use stress analysis
to determine the minimum diameter of the shaft at critical locations, followed by some projections as to a
reasonable shape everywhere else. The deflections of the proposed shaft are then calculated, again at one
location at the time. Finally and obviously, the size of the shaft everywhere along its length has to be at
least equal to the larger of the two diameters dictated by the stress and deflection analysis.
4 An example. In Fig 2, deflection analysis determined the minimum shaft diameter between b and
d to control the lateral and angular c (shown but not labeled) deflection at the gear wheel c. Gear
teeth have to make precise contact with each other or much of their capacity is wasted. Stress analysis
determined the minimum shaft diameters at the bearings, since the extra cost of larger bearings can
be significant, the shaft is shown machined down at the ends. Angular deflection at the bearings had to
be checked, because many bearings will tolerate very little misalignment before failing.

Shown in Fig 3 (and 4) is a short section of a shaft where the stresses at station x=0 are considered from
a force F and a torque T applied at x=x1. The force causes in part a moment about the Z axis, with its
associated normal stress x. We well know that x varies linearly between maxima at y=D/2 and 0 at
the Z axis. The torque T causes a shear stress xy(T) which is a maximum on the surface and points
tangentially to the circumference. We know that:

xy(T)

Fx1 D

xy (T )

O
T

I
TD

Eq 1

Eq 2

X
Fig 3. At left, the stresses at x=0
are shown from loads applied to a
shaft at x=x1. One stress is
omitted from Fig3 for clarity, but
is shown by itself in Fig 4.

Z
x=0

Fig 4. At right is the same piece of

shaft as shown above, but here the
transverse shear stress associated
with force F, xy(F) is shown.
This stress reaches a maximum
on the neutral axis (Z axis) and
zero at y=D/2, in opposition to
the variation in xy.

x=x1

xy(F)

5 Transverse Shear stress. This last stress xy(F), is nearly always omitted from any consideration, but
it is worthwhile to see why this is so, and perhaps when it may not be so. We can gauge the relative
significance of this stress by comparing its magnitude to the magnitude of x. The transverse shear stress
is a maximum on the neutral axis. For a circular section this is given as:
2
4F
xy
,
(where A D
)
Eq 3
4
3A
Since the bending moment, and its associated normal stress increases with x, and the transverse shear
stress remains constant, we can ask for what value of x1 will the two stresses, xy(F) and x be of equal
magnitude, ie:

 xy xy (F )

Fx1 D

D
that is where:

64

Eq 4

4F
2
3 D

D
x1
,
6

Eq5

4
ie x1 is 0.17 of the shaft diameter!

Eq 6

x is 10 times larger than xy(F) where x1 = 1.7 D, and 100 time where x1 = 17 D. Not only do
transverse shear stresses go from their maximum to 0 exactly where normal stresses due to bending go
from 0 to their maximum, but also these shear stresses are very small except for shafts with exceedingly
short separation between loads and bearings. Proportionally transverse shear stress contribute more to
total deflection than they do to total stress, but usually by less than a few % points.
Stresses in Shafts. Given that the highest coexisting biaxial stresses in Fig 3 are: x and
xy(T).
These stresses are shown on a Mohrs circle in Fig 5.
6

xy
2

Fig 5. Mohrs circle of the significant stresses

in a shaft caused by the shear force and torque
shown on Fig 3.

Giving the biaxial principal stresses:

xy

1, 2

x x

2
2

xy

Eq 7

These are substituted in the von Mises biaxial expression:

2

'2 1 1 2 2
giving:

'2 x 3 xy

Eq 8
Eq 9

Since the force and torque may in general be made up of steady and alternating components, (Fm, Fa) and
(Tm, Ta), we can calculate the mean and alternating distortion energy stresses starting with Eq 1 & 2 and
getting:
'2
2
2
m xm 3 xym
Eq 10

a ' 2 xa 2 3 xya 2

Eq 11
If there are mean and alternating axial forces on the shaft, the stresses resulting from those forces can be
added to the normal stresses in Eq 10 & 11.

7 The Goodman option. Substituting Eqs 1 & 2, with the appropriate expansions for I & J and using
the variable M= Fx1 for the moment, instead of F & x1 separately, into Eq 10 & 11. Then, substituting
these expanded von Mises stresses into the Goodman equation:
1 a' m'

Eq 12
FS S f S u
Rearranging the resulting expanded Goodman relation, to give diameter D for a required factor of safety
FS, after some time gives:

32 FS

Where:

Ma/m Ta/m kf/,fm kfs/,sfm Sf, Su -

Ma

3
2
k fs Ta
4

Sf

fm

Mm

3
2
k fsmTm
4

S ut

Eq 13

moment, alternating or mean

torque, alternating or mean
stress concentration factor for alternating or mean moment
stress concentration factor for alternating or mean torque
fatigue and ultimate tensile strength

If the average stress, or nominal stress, is below yield condition, then a stress concentration factor applies
to the mean stress (see Norton p385)
if: av kf < Sy, then: kfm = kf
Eq 14
This is done on the grounds that in the stress concentration zone the stress cannot exceed the yield stress.
Thus a stress concentration factor can only apply if yielding does not occur. If yielding occurs in the
whole load cycle:
kfm = 1
Eq 15
Then the stress concentration factor can have no effect as at no time in the cycle will the stress be below
yield level.
The advantage of the Goodman condition is that substituting equations 1, 2 , 8, 9, 10, 11 into the linear
equation 12, gives an expression that can be rearranged to give shaft diameters directly, ie Eq 13.
8 The Gerber option. The principal drawback to the using of the Goodman line is that it does not
follow the experimental results at all well, for fatigue failures under combined mean and alternating
stresses. Some experimental data is shown in Fig 6, note that this straight line is not a good fit to the
distortion energy stresses at failure, ie the curved line above it. The Goodman line understates the
available strength by a variable amount, from about 20% to 50%. This may be too much to give
away to competitors whom are prepared to go the extra flops (floating point operations).

Fig 6. Experimental data for steel

specimen showing the Gerber
parabola and the Goodman line,
for tests between 106 to 108 cycles.
The Gerber curve and the ASME
ellipse both represent the average
values reasonably well,

An impediment to the use of the Gerber equation is that substituting equations 1, 2 , 8, 9, 10, 11 into the
Gerber parabola Eq 16, raises the expressions of the right hand sides of Eq 10 & 11 to the 2 nd and 4th
powers before the square root sign is lost. Extracting the shaft diameter as the dependent variable
becomes a little messy but it can look like this:

letting:

m'
FS a'

FS

Se
S
u

Eq 16

a (k f M a ) 2 3 (k fsTa ) 2 S e2
4

and

2 32 FS b 2

b M m2 3 Tm2 S u2
4

Eq 17a, b

we get:

D
6

2b a 4ab a 2

Eq 18

We can also use the numerical solver to determine the diameter that for the applied mean and alternating,
moments and torques, gives a point that meets the condition defined by Eq 16. What is required are the
equations mentioned above beginning with Eqs 1 & 2, showing their relationship to functions 8, 9, 10, 11
and finally 16.
9 Shaft material. In the proportions that we usually make shafts, the requirements that they should run
adequately true (ie straight) under load, results in thicker and stiffer shafts over much of their
lengths, than the need for an adequate fatigue life. As a consequence shafts are made from material of
relatively high modulus of elasticity, that is steel. Since this modulus does not change much with alloying
elements, high quality low to medium carbon steel is preferred. Stainless steel is employed where
chemical action is significant. Cast iron is used where the shafts may be small or complex (crankshafts)
and production costs may be significant. Copper and aluminum alloys are employed in rare circumstances
where those materials have particular advantages that outweigh their low modulus.
Small mass-produced shafts are made from cold rolled stock, that comes with a smooth bright finish.
These shafts are prone to warping or distorting when their surface is machined, as residual stresses are
partly relieved, but this stock makes for cheap and rapid production. Hot rolled or hot forged billets are
more commonly used for larger industrial machines. For precision, but usually limited production work,
ground shaft stock is available. This may be obtained in a range standard diameters and in considerable
lengths. It can be in a normalized state, ready for extensive machining, or in a high temper state, if no
more than limited grinding is to be done.
10 Deflection calculations.
Beam deflection formulae due to bending, which may be found in
numerous texts on properties of solids, may be used to calculate the deflections of circular shafts. These
formulae deal with point and distributed loads, and a range of constraints at their reactions. Attached to
these notes is an example of such formulae. If a shaft has multiple loads at any location then the relevant
formulae may be used to arrive at the deflections at that location, one load at the time. The final deflection
is then the sum of the individual ones. If the loads are not coplanar it will be necessary to find the vector
sum of the deflections all points of interest along the shaft. Of course all this may be done provided that
the elastic limit is not exceeded.
These formulae typically use Macaulay functions or brackets to describe the discontinuous loads and
reactions very effectively. But there is a significant limitation in their use as they are they apply to shafts
of constant cross sections. That is, if a shaft has significant steps the application of this method requires
some careful considerations.

11 Castiglianos Energy method. There is an energy method that more readily deals with stepped
shafts and statically indeterminate designs such as the use of more than 2 bearings. The application of this
method is demonstrated with a tensile force in Figs 7a & 7b. A beam of uniform section shown on Fig 7a,
exhibits elongation when subjected to tension F. Assuming linear force/deflection response the energy
U absorbed in that deformation is contained under the response curve (straight line) shown on Fig 7b.
The absorbed energy is:
F
U
Extension
Eq 19
2
Fl
where:
Eq 20
AE

We can use Eq 20 to replace either F or

from the expression for U in Eq 19
giving:

Fig 7a above, b below.

Note that:
F

= area
Eq Energy
22
(F)/2
and

F 2l
AE 2
, or U
2 EA
2l

Eq 21a, b

U
FL

F
EA
U
AE

Eq 23

The differential of energy with respect to (wrt) force gives

deformation, and the differential of energy wrt deformation
gives the force. In principle this can be done for any linear
means of absorbing energy. A more general form of Eq 21a,
applicable where the cross sectional area A and or the force
F are not constant:
l
F2
U
dx
2 AE
0

Eq 24
Angular deflection

ds

Fig 8. at left shows a bar being bent by moment M

through an angle d, in an arc of radius , over an
arc distance ds. Over the infinitesimal distance ds
the energy absorbed is:
Md
dU
Eq 25
2
ds
EI
d
, and
Eq 26a, b

M
M2
Eq 27
dU
ds
2 EI

M2
ds
2 EI

Eq 28
For the simple case that is where M is constant, ds becomes l and: U

M 2l
2 EI

U
ML
U

and
M
M
EI

Eq 29
Eq 30a, b

Lateral deflection

Fig 9. At left is a beam deflected by a force F generating

a moment Fx. The stored energy is:
Fx
U
Eq 31
2
Fx

Eq 32
EI
Giving a stored energy in terms of the moment or
deflection:
Fx 2 or EI 2
Eq 33a, b
U
2 EI
2
And differentials wrt to force or deflection give:
U
U
, and
F
Eq 34a, b
F

Over a distance x, if forces and diameters vary:

2

Fx
Eq 35
U
dx
2 EI

12 Overview of Castiglianos methods: For each driving influence (F, M & Fx) there is a deformation
(, & ) and energy is absorbed. Taking the angular deflection as an example: the rate change of energy
wrt to the moment gives the angular deformation, at the point where the moment is applied. Thus for each
of these driving influences we can arrive at the deformation at the point were the force is applied.
The methods works if there are several forces and or moments coexisting, the differential wrt to a
particular moment or force will give the deformation due to all forces and moments, at the point of
application of that particular moment or force. This leads to the use of a dummy force or moment. If we
need to know the deformation at a particular location but there is no applied force there, we define one to
be there, carry out all the analysis, and finally we set that force or moment to 0. The final expression will
give the total deformation due to all other existing forces and moments at the required location. We will
do 2 examples to demonstrate the capacity of this method:
Example of Castiglianos method
13 Stepped conveyor drive shaft. On Fig 10 we see a simplified drawing of a drive shaft for a
conveyor drive pulley. In the development of these pulleys 2 methods have been adopted, one uses
relatively thin flexible components the other thick stiff ones. The shaft on Fig 10 represents the stiff
approach, for which the angular deflection of the shaft at the keyways is to be limited to a gradient of
1/1200 (or the same amount in radians).

Here the shaft is assumed to be symmetric about the vertical centerline. The bearings provide the
reactions R and the cheeks in the pulley transmit the load F to the shaft at the center of the keyways. In
modern drive pulleys special torque couplings are used in place of keys.
F

D1

D2

l/2

Fig 10. A conveyor drive pulley shaft. The loads F from the pulley are transmitted to the shaft near the
centers of the keyways. At these points a hypothetical moment m is imposed on the shaft. By
differentiating the total deflection energy due to all bending moments, wrt m we can arrive at the angular
deflection at b, due to all the other moments that would coexist with m, were m not =0.

U t
, given : D2 nD1
m
Total deflection energy: U t 2U x 0,a U x a ,b U x b ,l / 2
We need :

U t 2

Eq 36a, b

Rx 2 dx b Rx 2 dx 2 Rx F x b m 2 dx

Eq 37

2 Ein

2 EI

2 EIn 4

Eq 38

Note that the integral for deflection energy changes form where there is a change in cross-section or
where a load in encounter. Now there are great simplifications that are applied to the integral Eq 37: We
differentiate Ut wrt m, and then set m=0. That means that the first 2 integrals between the limits: 0 a,
and ab are =0 because they are not functions of m. In the last integral we can differentiate under the
integral sign because x is not a function of m, and finally after we differentiate, only term which are not
function of m will remain, because those which include m will go to 0 after we set m=0.

2
U t
2 Rx F x b m
2 0 0
dx , (m=0)
4
m
2
EIn

l
2 2 Rx F x b
m 2
dx
4
2 EIn
b

Fb
l 2b
EIn 4

Eq 39

Eq 40

Eq 41

If we have an indication of the ratio n of the 2 diameters (Eq 36b), we can calculate the diameter of the
smaller shaft D1 from I in Eq 41, and then the larger diameter D2. Thus as mentioned in section 3 on
page 2, we need a reasonable distribution of diameters for the whole shaft (calculated to meet stress
requirements) before we a can estimate deflections.

10

14 Three bearing shaft. The tables of deflection formulae relate to beams or shaft supported at 2
locations only, more than 2 reactions render the problem statically indeterminable. We can begin to
understand the behavior of multiple bearing shafts using Castiglianos method. By differentiating the
deflection energy wrt to R2 at the center bearing of Fig 11, we can arrive at the deflection at the center
bearing. For any particular deflection we can calculate the loads at the other 2 bearings.

R1

R2

R3
b

Fig 11. A 3 bearing shaft with a load between the first 2 bearings. If there is a lateral deflection of at
the center bearing, then the load at the other 2 bearings can be arrived at.
Total deflection energy: U t 2U x 0 ,a U x a ,b U x b ,c

Ut

2 EI

Eq 42

2
2
2
R1 x R1 x F x a R1 x F x a R2 x b dx

The deflection at b: 2

U
1

R2 2 EI

Eq 43

R x F x a R x b x b dx
1

Eq 44

After more than a few hours, rearranging this equation to give the force R2 for a given deflection 2
a
3
2 EI F c b
6
c
gives: R2
Eq 45
b
3
c b
6c
You may agree with me that the above equation indicates that if there is no deflection at the center bearing
then the reaction at the third bearing R3 will be zero. To arrive at a more balanced view we have to take
into account the stiffness of the 3 bearing mounts, because deflections will always take place.

11

12

13

14

15

16

17