Session F3H

978-1-61284-469-5/11/$26.00 2011 IEEE October 12 - 15, 2011, Rapid City, SD

41
st
ASEE/IEEE Frontiers in Education Conference
F3H-1
Teaching Deflection of Stepped Shafts:
Castigliano's Theorem, Dummy Loads, Heaviside
Step Functions and Numerical Integration

E. M. Odom
Professor of Mechanical Engineering, University of Idaho, eodom@uidaho.edu

C. J. Egelhoff
Professor of Mechanical Engineering, United States Coast Guard Academy, Carla.J.Egelhoff@uscga.edu


Abstract - The need for finding the deflections of shafts,
many of which are stepped or varying cross-sectional
areas is timeless. Each generation of engineers has used
that part of mechanics of materials theory that fit the
calculating capability available to them. The method
presented here is offered in that vein. The method uses
an engineer's ability to construct free body diagrams,
derive moment equations, and knowledge of energy
methods. The problem solution is kept general until the
last step which is a digital numerical integration. The
digital numerical integration can be performed on a
wide variety of software to include TKSolver,
MatLab

, MathCad

, EES

and spread sheets. This

method keeps the section properties independent of the
moment equations making it straightforward to include
scaling and shape factors on the cross-sectional
dimensions. This allows an engineer to run any number
of "what if" scenarios during a design process.
Additionally, this method provides intermediate
opportunities to validate the solution path by a) plotting
the moment equation and comparing it against shear
and moment diagram developed by hand, or b) plotting
the cross section and comparing it against the drawings.
Thus far, this approach to solving for the deflection of
stepped shafts has been presented to nearly 300 junior
Mechanical Engineering students.

Index Terms Deflection, Equation-solving software, Non-
uniform diameter, Scaling factor, Shaft design.
INTRODUCTION
At the present time, there are wind energy farms being
proposed and built, products are shipped around the world
via ships, trucks and airplanes, geared turbofan engines are
being developed and automotive hybrid power trains are
being designed. Deep inside of these macro technologies,
there resides a shaft doing the job as it was designed. In the
design of these shafts there is typically a strength and a
deflection/slope design specification. This paper presents
an approach to calculating shaft deflection that is accessible
to undergraduate engineering students.
A literature review of methods for calculating shaft
deflections indicates numerous methods have been
developed. For example, in 1960 Bert listed eight different
ways, ranging from graphical to Laplace transforms. [1]
Scholars and practitioners reflect in their writings the
evolution of shaft deflection calculations.[2]-[5] We have
previously taught a structured approach. [6] Textbooks
have taken several approaches to account for non-uniform
diameters, from using an equivalent diameter [7],
advocating commercial software [8], or avoiding the topic
altogether [8]-[10]. More recently, textbooks [11]-[13]
seem to be gravitating towards integrating the moment
curvature relationship

with the moment equation expressed using discontinuity
functions. While this method is quite correct, completing
the calculus and then accurately computing the integration
constants using the boundary conditions can be algebraically
tedious, time-consuming and fraught with opportunities for
error. The method proposed in this paper applies a
numerical integration step to the approach presented by
Professor Ju in his course notes and in an article [14].
Professor Jus approach is based on Castiglianos Theorem
and the use of Heaviside step functions to write the moment
equation. If performed by hand, this approach is
algebraically intense. However, the boundary conditions are
embedded in the formulation, and once created, the
formulation is ready for numerical integration.[15]
Here, we present a brief description, an example
solution augmented by the introduction of size and shape
factors to improve the design.
DESCRIPTION OF APPROACH
The method uses an engineer's ability to construct free body
diagrams, develop moment equations, and knowledge of
energy methods. Following the theory of Castigliano, the
internal energy due to bending is

(1)
Session F3H
978-1-61284-469-5/11/$26.00 2011 IEEE October 12 - 15, 2011, Rapid City, SD
41
st
ASEE/IEEE Frontiers in Education Conference
F3H-2
where M is the moment along the length of the beam as a
function of x, E is Youngs modulus, and I is the area
moment of inertia (which for a stepped shaft is also a
function of x). For this type of problem, Castiglianos
second theorem indicates that the deflection (or slope) at a
point is equal to the partial derivative of the strain energy
with respect to a load (or moment) applied at that point. If a
load (or moment) is not applied at the point of interest, then
a dummy-load (or dummy-moment) can be applied at that
point. After the partial derivative with respect to the
dummy-load (or dummy-moment) is performed, the dummy
load (or moment) is set to zero, i.e.

(2)

(3)
The problem solution is kept general until the last step
which is numerical integration. The numerical integration
can be performed using a wide variety of software products
including TKSolver, MatLab

, MathCad

, EES

and
spread sheets.
Additionally, this method provides intermediate
opportunities to validate the solution path by (a) plotting the
moment equation and comparing it against the shear and
moment diagrams developed by hand, (b) plotting the cross
section and comparing it against the drawings, and (c)
changing the stepped shaft into one with a convenient
constant diameter along the length and then comparing the
results to a closed-form analytical solution. The simple
solution steps are listed below:
- Draw a FBD and apply a dummy-load (and moment)
at an arbitrary location, . Use statics to solve for
reactions.
- Write a moment equation for the entire length using
Heaviside functions as needed to serve the purpose of
discontinuity function brackets
- Take the partial derivative of the Moment equation
with-respect-to the dummy-load (and dummy-
moment).
- Set dummy-load, Q=0, the dummy-moment m=0 and
re-write M(x)
Q,m=0
.
- Write Castiglianos integral for deflection, o
Q
,
inserting M(x) and cM/cQ. Input this equation into an
equation-solver and calculate o at any number of user-
selected locations. Similarly, write Castiglianos
integral for slope, u
Q
, inserting M(x) and cM/cm.
Input this equation into an equation solver and
calculate u at any number of user-selected locations.
EXAMPLE SOLUTION
The example problem shaft (Figure 1) is supported by
bearings near the ends and loaded transversely with a 7-kN
concentrated load positioned at 140-mm from center of the
left bearing support. The material is steel and the initial
shaft geometry is provided. The analyst is to examine the
shaft for deflection [16]. The largest allowable slope at the
bearings is specified as 0.001 radians.


FIGURE 1
SHAFT WITH MULTIPLE STEP CHANGES IN DIAMETER (TOP). FREE BODY
DIAGRAM SHOWING THE EXTERNAL LOAD (P=7-KN), DUMMY LOAD (Q),
DUMMY MOMENT (), REACTIONS (RL, RR) AND DIMENSIONS.

We begin by drawing the free body diagram as shown,
which has been constructed to include a dummy-load, Q,
and dummy-moment, m, at a distance, , from the left
support.
Using statics, solve for the reactions. These reactions
include an expression which contains the dummy-load, Q,
the dummy-moment, m, the beam length, L, and the point of
interest coordinate position, . The other variable terms
include the distance, , of the load, P, from the left bearing
support R
L
.

(4)

(5)
It may be useful to point out to students how the loads,
P and Q, are apportioned to each reaction, and that the m/L
terms comprise a couple whose influence on each reaction is
independent of the location of m.
Before writing the moment equation, the nomenclature
and definition of a Heaviside step function is introduced.
H(a,b)=0 if a<b
H(a,b)=1 if a>b
(6)

Session F3H
978-1-61284-469-5/11/$26.00 2011 IEEE October 12 - 15, 2011, Rapid City, SD
41
st
ASEE/IEEE Frontiers in Education Conference
F3H-3
We write the moment equation as a function of x for the
entire beam. Include dummy-loads and dummy-moments.

(7)

Depending on the instructor, various portions of this
equation can be commented upon. In this form, (7), a brief
review of chain-rule differentiation for the R
L
x term, or the
expression for R
L
may be helpful to students. The second
term, Q(x-), may need explanation to help students
remember there are two, complementary axes at work here,
that , is user-selected first, then x- is used for integration.
Next, write the partial derivative with respect to the
dummy load, Q, and with respect to the dummy-moment, m.

(8)

(9)
Write the Moment Equation setting the dummy term(s)
to zero.

(10)

Using Castiglianos Theorem, substitute (8) or (9) along
with (10) into (2) or (3) to construct the correct integral to
be solved as detailed in (11)and (12) below.

(11)

(12)
Solve. There are a couple of items to code into the
software.
First, the diameter as a function of shaft position will
need to be defined to calculate the area moment of inertia
(13) during integration.

(13)
This can be accomplished by using a look-up table or a
series of if-then-else statements in a user-defined sub-
routine, depending on the equation-solving software options
and the users programming skills [17]. Figure 2A shows
the list function used by TKSolver which serves as a look-
up table and Figure 2B shows EES

code for the user-

defined function which produces the same result.
We are aware that we are integrating across a
discontinuity. However, we have found numerical
convergence to be very rapid compared to closed-form
solutions.

(A)
Define the diameter as a function of x
function dia(x)
if(x<15) then dia:=30 else;
if(x<40) then dia:=35 else;
if(x<100) then dia:=40 else;
if(x<110) then dia:=55 else
if(x<210) then dia:=45 else;
if(x<275) then dia:=40 else;
if(x<300) then dia:=35 else;
if(x<315) then dia:=30 else
endif;
endif;
endif;
endif;
endif;
endif;
endif;
endif
end
(B)
FIGURE 2
(A) USER-CREATED TKSOLVER LIST FUNCTION DEFINES THE SHAFT
DIAMETER ALONG THE LENGTH. (B) USER-CREATED EES

CODE ALSO
DEFINES SHAFT DIAMETER. BOTH SERVE AS LOOK-UP TABLES,
DETERMINING DIAMETER FOR ANY LOCATION, X.

Second, the geometry and moment are checked by
creating plots such as the one shown in Figure 3.


FIGURE 3
EQUATION-SOLVER GRAPH USED TO ENSURE THE GEOMETRY FUNCTION
WORKS AS INTENDED. SHAFT MOMENT DIAGRAM SERVES AS A CHECK
BEFORE PROCEEDING.

This prompt check of the shaft profile ensures that the
problem is proceeding as expected; similarly, a check of the
moment diagram would support confidence in the solution if
done prior to attempting the repeated integrations.
Session F3H
978-1-61284-469-5/11/$26.00 2011 IEEE October 12 - 15, 2011, Rapid City, SD
41
st
ASEE/IEEE Frontiers in Education Conference
F3H-4
Next, the crucial action is to integrate with respect to x
but then create a series of solutions (such as in a table) using
as the indexing variable. Selecting 316 positions of for
this 315-mm shaft is sufficient to obtain a smooth deflection
curve.


FIGURE 4
INITIAL SOLUTION FOR SHAFT DEFLECTION AND SLOPE.

With the equation-solver solution for deflection and
slope at every 1-mm along the beam (Figure 4) we find that
the loaded shaft does not yet meet the slope limitation of
0.001 rad at the bearing supports (read left bearing slope
0.0014 rad on graph). Therefore, we elect to apply a scaling
factor to adjust the shaft geometry (diameter) in order to
mitigate the excessive bearing slopes.
SCALING FACTOR
If either deflection or slope is deemed unsatisfactory, it is
simple to modify the I(x) equation by inserting a scaling
factor, sc, as shown in (14). Then one can solve a new list
of values to obtain a new deflection curve.

(14)

In the case of this textbook problem, a scaling factor of
about sc=1.1, when applied to the given shaft profile, results
in acceptable slope conditions at the bearings (<0.001
radians) and deflection much less than the solution shown
initially (deflection itself was not specifically limited in the
original problem statement).
With sc=1.1, both the left support slope and the
maximum deflection are lessened. The determination of the
precise scaling factor, in this case sc=1.1058, can be
accomplished using built-in optimization or goal-seek
functions available within individual equation-solvers. We
have found, however, that a guess-and-check approach will
result in finding the scaling factor as close as 99% in only 3-
6 guesses, quickly enough for even a novice user.
SHAPE FACTOR
The method as presented keeps the section properties
independent of the moment equation, making it
straightforward to include scaling and shape factors on the
cross-sectional dimensions. This allows an engineer to run
any number of "what if" scenarios during the design
process.
It is easily seen (Figure 4) that initially, the maximum
deflection occurs at about 155-mm from the left bearing
support. Since the loading of 7-kN is located at 140-mm
from the left bearing, and the 7-kN load may represent a
gear meshing at that location, it may be desirable to design
the shaft such that the maximum deflection occurs more
precisely at the location of the 7-kN load. This will help
ensure that the gear teeth will mesh properly and that no
unintended loading results from a shaft which is sloped at
that important location.
To shift the location of maximum deflection, we
introduce a shape factor, sh(x), and we use experience to
implement this factor. By inspecting the shaft geometry and
the initial deflection curve, we imagine that by judiciously
adding more material to the shaft, we may impact the
deflection in a favorable way. We dont need additional
material everywhere (e.g. not at the left end). We select a
special shape point at x
sh
=110-mm, just to the right of the
largest-diameter section at the collar. We want to increase
material to the right of x
sh
and decrease material to the left
of x
sh
. Accordingly, we define sh(x) as in (15).

||

||

(15)
Defining the Shape Factor in this way, we then proceed
to the guess-and-check strategy, checking the location of
maximum deflection until we are satisfied that it occurs at
140-mm from the left bearing support (i.e. at the location of
the 7-kN load), which results in a Shape Factor of about
sh=0.924.
FINAL DESIGN
In the example, the slope at the ends is greater than
recommended, meaning the shaft is too compliant. A direct
method to increase the stiffness of the shaft is to increase the
diameter, dia(x), which will increase the area moment of
inertia, I(x). Also, in Figure 4 it can be observed that the
maximum deflection occurs at about 155-mm from the left
end whereas the 7-kN load is applied at 140-mm. During
redesign, it could be advantageous to not only add stiffness
to the shaft but to do so in a manner that could move the
maximum deflection (zero slope) to the location where the
7-kN load is applied.
The method of solution presented here can be modified
by changing (13) to include a scale factor (sc) and a shape
factor (sh) as shown in (14). The scale factor (sc) is
independent of the distance along the shaft; if it is greater
Session F3H
978-1-61284-469-5/11/$26.00 2011 IEEE October 12 - 15, 2011, Rapid City, SD
41
st
ASEE/IEEE Frontiers in Education Conference
F3H-5
than one it enlarges the diameter everywhere, if less than
one it shrinks the diameter everywhere.
The shape factor (sh) is dependent on the location along
the shaft. Mathematically, it is similar to the Heaviside step
function and can be explained by exploring how it is coded
as shown in (15). In use, the designer must make two
decisions. First, a location on the shaft, x, has to be selected
about which the shaping will pivot, typically at an existing
change in diameter; closer to the mid-length of the shaft is
better. For this example, x
sh
=110-mm was chosen. Second,
the amount to change (magnification) the diameter must
also be chosen. The magnification can be greater than or
less than one. If the magnification is greater than one, then
the shape factor (sh) would be greater than one for the shaft
diameters to the left of x
sh
thereby increasing those
diameters. The diameters to the right of x
sh
would be
decreased by the reciprocal of the magnification.

FIGURE 5
SUMMARY OF SHAFT GEOMETRY AT EACH STEP OF THE DESIGN PROCESS.
TOP: APPLICATION OF ONLY THE SIZE FACTOR, SC. MIDDLE: APPLICATION
OF ONLY SHAPE FACTOR, SH. BOTTOM: RESULT OF COMBINED FACTOR.
The shaft geometry at intermediate check-points is
shown in Figure 5. The top image shows before (black)
and after (red) the application of the scaling factor, sc.
The middle image shows the effects of only the shape
factor, sh, which produces the effect of decreasing the
diameter to the left of x
sh
=110 mm and increasing the
diameter to the right of x
sh
=110 mm.
When both factors are applied, as shown in the bottom
image, the combined effect is barely visible to the left of
x
sh
=110 mm but to the right the effect is greater.

FIGURE 6
SUMMARY OF SHAFT DEFLECTIONS AT EACH STEP OF THE DESIGN
PROCESS.
Finally, we summarize the shaft size and shape design
process in Figure 6 showing deflection curves for initial
calculations, with application of only the scale factor, with
application of only the shape factor, and finally with
application of both size and shape factors.
The final design shaft not only meets the slope
specified in the problem statement but also produces a
deflection nearly half that in the initial configuration.
Additionally, the location of the maximum deflection has
been adjusted using the shape factor such that the slope is
zero at the 140-mm (Figure 7).


FIGURE 7
SUMMARY OF SHAFT DEFLECTIONS DURING SHAFT DESIGN, NOTING THE
LOCATION OF MAXIMUM DEFLECTION IS SHIFTED TO 140 IN FINAL DESIGN
CONFIGURATION.

The final design using the combined effects of both
factors results in the smallest maximum deflection. The
slope at each bearing is less than the maximum allowed
0.001 radians.
Session F3H
978-1-61284-469-5/11/$26.00 2011 IEEE October 12 - 15, 2011, Rapid City, SD
41
st
ASEE/IEEE Frontiers in Education Conference
F3H-6
Listed in Table I are the salient numerical values for
initial solution, applying only the scaling factor, applying
only the shape factor and the final design which applies both
factors.
TABLE I
SHAFT SLOPE AND DEFLECTIONS BEFORE AND AFTER SCALING AND SHAPE
FACTORS ARE APPLIED FOR DESIGN

Value After
RL Slope
(rad)
RR Slope
(rad)
Maximum
deflection
(mm)
Location
Maximum
deflection
(mm)
Initial
Calculation
-0.00142 0.00133 -0.125 153
Application of
Scaling Factor
0.001 0.00093 -0.092 151
Application of
Shape Factor
-0.0011 0.0011 -0.111 140
Final Design -0.001 0.000717 -0.074 140

CONCLUSIONS
The design of shafts of non-uniform cross-section is an
ongoing need. We have reviewed the textbook treatment of
this problem-type and presented a method of solving these
challenging problems using fundamental engineering skills
and modern engineering tools.
The method presented here relies on Castiglianos
Theorem, the Heaviside Function and Discontinuity
Equations in conjunction with equation-solving software.
This method helps to mitigate the difficulties experienced
using other methods of finding shaft deflection
everywhere along the length, such as writing sets of
governing equations for numerous sections, applying
multiple continuity relationships to those sets of governing
equations, and solving accurately for multiple constants of
integration. The method presented keeps the process
general until the equation-solver is introduced.
By way of example, we show how the shaft geometry
and moment diagrams are easily checked early in the
equation-solving process, thus keeping the process
grounded in fundamentals and accessible to undergraduate
students. The shaft slope and deflection are numerically
calculated using any of several equation-solving tools. We
show how a Scaling Factor, sc, and Shape Factor , sh, can
be applied directly to the shaft diameter to meet deflection
and slope design criteria. We note that historically finding
slope and deflection for the initial problem would have been
difficult. Using this method we not only find slope and
deflection, but we competently design and refine quickly.
Using equation-solving software we show how to distribute
the shaft material from end to end by designing the
complete shape of the shaft. Thus far, this approach to
solving for the deflection of stepped shafts has been
presented to nearly 300 junior Mechanical Engineering
students.
ACKNOWLEDGMENTS
The authors gratefully acknowledge Mitchell Odom for
providing high-quality graphics for the many figures. We
are also very appreciative to Alexander Odom for
contributing the beautiful shaft image rendered using
SolidWorks and Rhinoceros

. We also thank those who

initially solved the problem with alternate equation-solving
software: Ensign Benjamin Wiest using Mathematica

and
Lieutenant Kara Burns using Matlab

.
REFERENCES
1 Bert, Charles W., November 24, 1960. Deflections in Stepped
Shafts, Machine Design, p. 128-133
2 Shroeder, Walter, January 1947. Beam Deflections, Machine
Design, p. 85-90
3 Richmond, E. L. and R.H. Feng, September 1953. Graphical
Methods for Determining Beam Deflections, Machine Design, , p.
177-183
4 Kitching, R., N. Sanderson and S. Hinduja, June 1975. Flexibility of
rectangular beams with abrupt changes of section, International
Journal of Mechanical Sciences, Volume 17, Issue 6, Pages 403-410
5 Cueva-Zepeda, Alfredo. 1996. Deflection of Stepped Shafts Using
Macaulay Functions, Computer Applications in Engineering
Education, Vol. 4(2), p. 109-115
6 Egelhoff, C.J., E.M. Odom and B.J. Wiest, 2010. Application of
Modern Engineering Tools in the Analysis of the Stepped Shaft:
Teaching a Structured Problem-Solving Approach Using Energy
Techniques, Proceedings of the 2010 Frontiers in Education
Conference, Washington, D.C.
7 Collins, Jack A., Henry Busby and George Staab. 2010. Mechanical
Design of Machine Elements and Machines A Failure Prevention
Perspective, Second Edition, John Wiley & Sons, Inc., Hoboken, NJ,
pp. 354-357
8 Mott, Robert L. 2005. Machine Elements in Mechanical Design, 4th
Edition, Pearson Prentice Hall, Upper Saddle River, NJ
9 Hamrock, Bernard J, Steven R. Schmid and Bo Jacobson. 2005.
Fundamentals of Machine Elements, Second Edition, McGraw-Hill,
New York, NY
10 Ugural, Ansel C. 2004. Mechanical Design: An Integrated Approach,
McGraw-Hill, New York, NY
11 Norton, Robert L. 1996. Machine Design: An Integrated Approach,
Prentice Hall, Inc., New Jersey
12 Budynas, Richard G. and J. Keith Nisbett. 2008. Shigleys
Mechanical Engineering Design, 8th Edition, McGraw-Hill, New
York, NY
13 Juvinall, Robert C. and Kurt M. Marshek. 2006. Fundamentals of
Machine Component Design, 4th Edition,John Wiley and Sons, Inc,
Hoboken, NJ
14 Ju, F.D. 1971. On the constraints for Castigliano's theorem, Journal
of the Franklin Institute, Volume 292, No. 4, Pages 257-264
15 Odom, Edwin M. and Carla Egelhoff. June 10, 2010. Stepping
Through Shaft Deflection Calculations, Machine Design, pp. 58-60
16 Shigley, J. E. and Mischke C.R. 1989. Mechanical Engineering
Design, Fifth Edition, McGraw Hill, Inc. New York New York.
Problem 18-20.
17 We have successfully used TK Solver, EES

, Mathematica

,
Matlab

and MathCad using the approach described in this paper