FUNda MENTALs Topic 5

User Manual: FUNdaMENTALs-Topic-5 Fundamentals Of Design

Open the PDF directly: View PDF .
Page Count: 62

Download
Open PDF In Browser	View PDF

FUNdaMENTALS of Design
Topic 5
Power Transmission Elements I

5-0

1/1/2008

Power Transmission Elements I
There are many ways to store or generate power,
but all are useless without a transmission to transmit the
power from a source to the place where it is needed in a
form in which it can be used. Furthermore, the basic
physics of power generation and transmission indicate
that high speed and low torque are more efficient to create than low speed and high torque. The former is generated by high voltage and low current which is more
economical than low voltage and high current. Once
again, a transmission is often required to transform
power into a useful form. But what combination of elements provides the best price/performance? For example, how should high speed low torque rotary motion
from a motor be transformed into low speed, high force
linear motion?
By designing machines in a modular fashion, an
engineer can assume power will be provided in a certain
manner. The engineer can then proceed with the design
of the modules and elements that will transmit and utilize the power. This also allows some flexibility in optimizing the power source as the design precedes. For
example, initial calculations may have indicated that a
certain size motor was required, but in designing the
power transmission system, the motor size may
decrease/increase depending on the inertia and effi-

ciency of the power transmission system. Accordingly,
this chapter will focus on some of the elements of transmission systems, such as pulleys and wheels, belts and
couplings. Chapter 6 will discuss screws and gears.
Chapter 7 will discuss selection of optimal transmission
ratios and actuators that use the elements in Chapters 5
and 6.
There are an amazing array of transmission components, and engineers need to become familiar with
what is available, and what niche each element occupies. By applying FUNdaMENTAL principles and
Maudslay’s Maxims, you will better be able to separate
marketing hype from engineering reality when you consider:
• Kinematics: motion, accuracy, space
• Dynamics: forces, speeds, life
• Economics: design, build, maintain

By creating machines for robot design competitions, you will have to use fundemental principles to
help create the transmissions you need from very basic
elements which will greatly hone your design skills.
Remember to look at the world around you from garden
tractors to hot-rods. Keep your eyes open!

Topic 5
Power Transmission Elements I
F

Topics:
•
•
•
•
•
•
•
•
•
•

Transmissions
Pulleys
Winches
Belts & Cables
Wheels
Steering & Suspensions
Clutches & Differentials
Cams
Shafts
Couplings

5-1

1/1/2008

Pulleys
Pulleys are one of the oldest and most ubiquitous power transmission
elements, but they require careful design. If a belt or cable runs around a fixed
shaft, friction between the belt and the shaft can cause the efficiency to be low,
and the cable can rapidly wear. A pulley reduces these effects with rolling contact between the cable and the machine, but it must be of sufficient size, typically 20 times the cable diameter, to prevent fatiguing of the cable strands. In
addition, a crown on a pulley can be used to keep a flat belt centered on the
pulley even if the pulley is not perfectly aligned (see page 5-10).
Pulleys transmit power from one location to another, and they can
form a transmission ratio. The pulley equivalent of a lever and fulcrum is in
fact Atwood’s machine, which is comprised of a pulley mounted to a fixed
object, and a pulley that moves with the cable and to which a lifting hook is
attached. Since the tension in the cable is everywhere equal, one force input to
the cable becomes two forces acting on the output pulley, and the force is doubled, although the lifting speed is half that of the cable. For multiple strand
sheave system (like seen on large cranes), Workin = efficiency*Workout:

F out = η F in N number _ of _ cable _ strands
V out =

As the plot shows, if the coefficient of friction can be kept below 0.1,
very doable for a properly lubricated plain bearing, and the ratio of the pulley
to shaft diameter is above 4, then efficiencies of greater than 90% can be easily
obtained. Consider a system with D/d = 8 and 10 sheaves in the system, so the
system efficiency would be 0.975310 or about 78%. If the sheaves are supported by needle roller bearings, the coefficient of friction can be as low as
0.005, and the efficiency of 10 sheaves would be 0.988810 or about 98.76%.
Which system should be used, a simple plain bearing system, or a more expensive needle roller bearing system? It depends on the functional requirements
of the system, including cost and maintenance requirements! The analysis thus
becomes your guide to selecting the proper machine elements that are to be
used!
Although the plot shows that it is desirable to maximize the ratio of
pulley to shaft diameter, the shaft must be sized to meet several functional
requirements. The shaft must:
•
•

V in
N number _ of _ cable _ strands
•

Large cranes have many strands of cable running over many sheaves
(pulleys) in order to keep the cable tension low, which enables a smaller diameter cable and lower torque winch to be used. On such systems, are the pulleys
supported by rolling element bearings to increase system efficiency? If the belt
or cable rides on a pulley of diameter D with inside sliding-contact-bearing-onshaft diameter d, and coefficient of friction μ, the efficiency η is found from:

work out = work in − work friction _ losses

out

)θ = ( F

)

D − ( F out + F in ) μ d θ
2
2

⎛ D − dμ ⎞
η = F out = ⎜
⎟
F in ⎝ D + d μ ⎠

Withstand applied loads without yielding: Recall that for a simply supported pin of length L, the maximum moment is FL/4
Not deflect too much: If the shaft is cantilevered, the angular deflection
can cause the cable to rub on the side of the pulley and cause premature
failure. In severe cases, it can cause the belt to track off the pulley despite
the crown. Simply supported shafts have zero slope in the middle and thus
help maintain tracking.
The shaft diameter must support a sliding bearing whose maximum pressure load and maximum PV ratings1 are not exceeded.
• The bore and the faces of a plastic pulley typically act as their own
radial and thrust sliding contact bearings. A metal pulley should have
modular sliding contact bearing elements for radial and thrust loads.

Look at the mechanisms under the glass of a photocopy machine, and
observe its cable drive. How does it use pulleys? Look closely at a crane as it
operates. Look at the belts on the engine in a car. How might your machine
use pulleys and how can you mount them so that they do not tilt under load?

1.
PV = (pressure)(velocity), and is a critical parameter for sliding contact bearings, as discussed in
Chapter 10.

Mechanical
advantage = Number
of strands attached to
the load! (=2 here)

Pulley Bearing Efficiency

Pulleys
100%

•
W

Pulleys (Sheaves) are a most
fundamental power transmission
element
–
–
–
–
–

Mechanical advantage
Capstans
Efficiency
Tracking
Mounting

Efficiency

90%

80%
η,
η,
η,
η,

μ=0.05
μ=0.1
μ=0.2
μ=0.3

70%
2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

Pu l le y/B e arin g di am e te r

5-2

1/1/2008

Pulleys: Capstans
Before there were pulleys that could rotate on a shaft, a rope was
likely laid over a round bar in order to allow a person to pull down and have a
load be hoisted. Someone probably figured out that they were having to pull
down with more force than the weight of the load. They probably then made
the connection with the wheel and thus the pulley was born; however, reciprocity would indicate that there could be some use for this effect, and as every
sailer will tell you, a capstan is an indispensable element. A capstan functions
as sort of a one-way brake. It allows you to pull in a rope, yet with mild effort,
you can resist very large forces that are attempting to pull the rope back out.
The derivation of the capstan equation is very straightforward and
also is used to obtain the breaking torque that can be obtained when a capstan
is applied as a band break. A band break uses a band anchored at one end
which passes over a rotating drum and is pulled at the other end to brake the
drum. When configured as shown with the direction of rotation pulling
against the band segment that is anchored, the principle of self help also
applies. As the free-end of the band is pulled, the friction between the band
and the shaft pull harder on the fixed end. The limit of the tension in the fixed
end is a function of the wrap angle and the breaking force F that is applied. For
a capstan and a brake, the cable tension T2 that can be held by applying a tension T1 for a wrap angle θ and a coefficient of friction μ is found by considering a differential element. The output tension is equal to the input tension plus
the product of the coefficient of friction with the normal force caused by the
two tensions. This normal force is:
⎛ dθ ⎞
⎛ dθ ⎞
N = (T + dT ) sin ⎜
⎟ + T sin ⎜
⎟ ≈ Tdθ
⎝ 2 ⎠
⎝ 2 ⎠
θ
T 2 dt
μθ
T + dt = T + μ T d θ
∫T 1 T = ∫0 μ d θ T 2 = T 1e
− μθ
) D brake
T 2 (1 + e
⎛ D brake ⎞ ⎛ T 1 + T 2 ⎞
Γ brake = μ ⎜
⎟⎜
⎟⇒
2
4
⎝ 2 ⎠⎝
⎠

The brake torque is a function of the maximum normal force, which is
the vector sum of the tensions T1 and T2, the brake drum radius Dbrake/2, and
the coefficient of friction, which can be easily found using the law of cosines.
Is there thus a maximum brake torque angle and how can one find it?

N brake _ max = T 1 + T 2 − 2T 1T 2 cosθ
2

Γbrake =

μ N brake _ max Dbrake
2

μθ

2μ e − μ cosθ + sinθ

= 0 _ for _ optmal _ wrap _ angle

Enter numbers in BOLD, results in RED
Capstan
Band brake
Angle of wrap, theta (degrees)
180 Brake diameter, Dbrake (m)
0.1
Coefficient of friction, mu
0.10 Brake torque factor, BTF
2.37
Capstan ratio, CR
1.37 M aximum brake torque (N-m) 0.1185
Holding force, Fhold (N) (Fbrake)
10 Optimal wrap angle (use Goal S eek)
Resistable pulling force, Fpull (N) 13.69 Optimal angle, phi (degrees)
202
d/dθ =0
Cable pulling around a shaft
3E-03
Net force out
7.30 Capstan ratio, CRO
1.42
Efficiency
73% Brake torque factor, BTFO
2.38
"Optimal" brake torque (N-m) 0.1190

In terms of the analysis process, note that once the equation for the
tensions in the band is found, the next step was to create a hypothesis to find
the normal force on the shaft. In the case of a wrap angle of 180 degrees, the
sum of the two forces is the force the shaft must resist and so vector addition of
the tensile forces makes sense. If the wrap angle was 360 degrees, the net
force on the shaft would be less than the maximum tension force. On the other
hand, the capstan force may be much larger, so a conditional “MAX” statement
can be used in the spreadsheet to use the greater of the tension forces in the
band, or the vector sum of the forces. These relations can be used to design
brakes or to determine the efficiency of putting a cable over a round shaft to
see if it is worth it or not to make a pulley. In industrial practice one would
likely never do the latter, but in a robot contest for a quick small trigger mechanism, for example, it might be acceptable.
How can you use the mechanical advantage provided by a capstan, or
how can you tolerate the losses incurred with just running a cable over a round
shaft? Reciprocity strikes again!

T+dT
N=Tdθ

Pulleys: Capstans

Band brake

μN

dθ

•

A capstan is typically a fixed, or controlled rotation (with one-way clutch), bodyof-revolution which a cable wraps around
– A capstan can also form the basis for a band brake, where a band is anchored to a
structure, and then wraps around a shaft…

A cable wrapped around a capstan by θ radians with coefficient of friction m and
being held with a force Fhold, can resist the pull of a cable with many times higher
force Fpull
− μθ

•

F hold = F pull e

•

If a belt or cable runs around a fixed shaft, then there is a lot of friction between the
belt and the shaft, and the efficiency is low:
Angle of wrap (degrees, radians) θ
Coefficient of friction μ
Holding force Fhold
Pulling force that can be resisted Fpull

Capstan Ratio (Tout/Tin)

7
6

μ = 0.3

μ = 0.2

μ = 0.1

720
0.2
1
3.52

6.283

Force
being
held

3
2
1
0

135

180

225

270

315

Braking
force

360

Wrap angle (degrees)

5-3

1/1/2008

Winches
A winch is a device used to control the tension and position of a cable
by winding it onto a drum. Winches are one of the principle elements of
cranes, as they provide the lifting force for the hook and often the boom.
Winches are also commonly used to provide a pulling force for equipment such
as tow trucks. Until the introduction of hydraulics, winches were also one of
the principle means of actuation for construction equipment. Robots, particularly those used in competitions, may use a winch to help pull themselves up a
steep slope, or to deploy a wall. In accordance with the FRDPARC design
tables, consider first the primary function requirements for a typical winch:

The first step in winch design is to evaluate the amount of cable and
force needed and then to determine the drum size and motor/gearbox that is
required. For a simple winch to evaluate the cable force and drum wraps, the
spreadsheet WinchCable.xls can be used:

Table 1:
Functional Requirements
High cable force & speed

Design Parameters

Analysis

Motor/gearbox torque/speed
Drum diameter

Small diameter needed to minimize motor torque, or use large
gear ratio

Large cable capacity

Drum diameter, length

Beware of increasing effective
diameter on cable force, see
spreadsheets

Even cable winding

Drum length
Fairleader

Difficult to analyze, run experiments or seek similar solutions

Given these primary issues, consider the remaining corresponding
issues that are most often overlooked by designers, but which are critical to the
design process:

For the case of a winch raising a boom, the spreadsheet Winchboom.xls can be used to select the detail design parameters. In terms of
detailed design considerations, note the picture that shows what looks like a
good winch, except that the drum was connected directly to the motor. Even
for a robot design competition, haste makes waste. Typically the drum must be
supported by its own bearings because it is long with respect to the motor shaft
diameter and the motor bearings would be overloaded by a large radial load far
from the motor’s front bearing. Torque from the motor should be applied via a
flexible coupling, universal joint, or a well-aligned spline, else misalignment
from manufacture or deflection of components from the applied load could
deform the motor shaft and cause premature wear of the bearings.

Table 2:
References

Risks

Countermeasures

Motor gearbox catalogs, Commercial winch companies
(search the internet)

Static holding force could
cause electric motor to burn out

Use a brake or non-back drivable transmission (worm gear).
Cranes often use hydraulics.

Commercial winch companies

Short length, large diameter can
make the effective cable diameter increase creating need for
large motor.

Bearings and structure are
cheaper than motors, use a
longer drum

Uneven cable winding can
reduce capacity

Good cable management, do
not skimp on the fairleader

Commercial winch companies

A fairleader is a flared opening that guides the cable in and helps it
wind properly. The simplest form is just a smooth rounded opening which ideally the cable never even touches. When the cable pulls an object from the side
of the winch, a fairleader with rollers may be required as can be determined by
the capstan equations. A powered fairleader moves back and forth ensuring
that the cable winds in an orderly fashion without randomly overlapping itself.
How would you optimize motor/gearbox size with drum size to pack
your winch in as small a space as possible? Do you have enough torque so a
large diameter short length drum could be attached directly to the gearbox output shaft, or do you need a long drum supported at each end by bearings?

http://www.gowarn.com/winches.htm

Winches
•

A winch is a motorized drum that controls cable tension
– A single wrap of cable on the drum requires a long or large diameter drum
• Effective drum diameter and winch force capability remain constant
– Multiple wraps of cable on the drum allow for more cable in a smaller place
• Drum diameter and winch force capability vary
– A fairleader is a device to control the input/output of the cable so it winds on the drum in an
orderly fashion
• The simplest design just uses smooth rounded static features
• Vertical and horizontal roller designs reduce cable wear

rr
Lc

5-4

Lw
La

1/1/2008

Belts & Cables
Belts and cables are very common power transmission elements
because their elastic nature enables them to pass over round objects (pulleys)
typically with a high degree of efficiency. The term power transmission literally means the power output from the device equals the product of the efficiency and the power input to the device. Therefore, analyzing any power
transmission system really is as simple as keeping track of the products of
speeds and torques (or forces) and efficiencies.
Belts and cables are laterally compliant, which makes them forgiving
of misalignment, so they are commonly used applications from automobiles to
office equipment. In all cases, the key functional requirements of a belt or
cable are to transfer tensile loads and to pass over a pulley. The life of a belt or
cable is a function of its pre-tension, the diameter of the smallest pulley, and
the load it is expected to carry. The key efficiency issue is to minimize contact
at different diameters to prevent differential slip between the belt and the pulley. The belt cross section must be chosen to withstand the sum of the stresses
from these three sources while maintaining a desired level of stiffness.
A cable is a flexible tensile element whose elements are stranded
together such that all of the strands share the load. When the cable passes over
a pulley, the strands locally slip over each other so the bending stress in any
single strand is far lower than if a solid bar of the same diameter were bent
over the pulley. However, slip dissipates energy, so cable drives are used primarily in low speed applications. The load that cables can carry is a function
of the pre-tension, the coefficient of friction between the cable and the pulley,
and the wrap angle. Hence the pulley acts like a rotating capstan.
A belt can have many forms, but the three principle types are: flat,
toothed, and vee belts. Flat belts transfer a load by being pre-tensioned, and as
with cables, the pulleys act like capstans. Flat belts are generally very efficient, unless forced to run on crowned pulleys to accommodate misalignments
(see page 5-10). A flat-toothed belt positively engages the pulley, and requires
only enough tension such that the load acting on the tooth angle does not create
a force large enough to pop the tooth out of the pulley. The efficiency is still
high because the tooth shape on the belt and the pulley are designed to minimize differential slip1. A Vee belt uses the principle of self-help, whereby the

greater the load, the more the Vee shape wedges itself into the pulley. This
increases the torque transmission capacity but decreases the efficiency.
The lateral flexibility of belts and cables can accommodate errors in
rigid elements. The kinematics of belt and cable drives allow them to achieve
the desired motion with a great deal of accuracy in a modest space. In particular, their lateral compliance allows system components to be mounted with
greater ease than gears or screws. For example, cables can be routed around
pulleys whose axes of rotation are not necessarily parallel to each other. With
flat belts, one must use crowned surfaces or edge guides to deal with the invariable tendency for a belt to “walk off” a pulley (see page 5-10).
The dynamics of belts and cables, which often contain reinforcing elements such as steel or KevlarTM, make them efficient power transmission elements. Not only can high power often be transmitted, but high force or high
torque can also often be transmitted very repeatably. On the other hand, it is
difficult to obtain the stiffness or load capacity of gears or screws. Speeds are
limited by unsupported lengths of belts or cables vibrating. Fortunately, measuring the frequency of vibration is straightforward and it can be used as an
accurate measure of belt tension. Typical linear systems may achieve speeds
of many meters per second, and rotary systems may run at thousands of rpm.
The economics of belt and cable system design are very good precisely because of their ability to elastically average errors, and similarly, they
can be very straightforward to build and maintain. Key factors affecting cost
include not only the belt and pulleys, but also managing belt or cable stress
around pulleys, maintaining proper belt or cable tension, and achieving reasonable alignment of the pulleys.
For every linear and rotary motion you are contemplating in your
robot, sketch how you could use a belt or cable as the power transmission element. What are the limits on pulley size and location? What sort of speeds and
loads must be accommodated? Can a belt or cable enable more convenient
placement of a motor? How might you control belt or cable tension? How
might you change the belt or cable if it were to become damaged?

1.
Chains operate on a similar kinematic principle, and the use of high strength steel enable them to
carry much higher loads; however, chains require lubrication, but other than that, can be designed into systems with essentially the same methods.

Belts & Cables
Applications and Engineering
–
–
–
–

Linear motion
Rotary motion
Crawler tracks
Manufacturing & Assembly

Thanks to Prof. John Lienhard for this photo!

http://www.genielift.com/ml-series/ml-1-5.html

Powerout = Powerin
Vbelt = Routωout

speedoutTorqueout = speedinTorquein
Γout = ωinΓin/ωout

ωin, Γin

Rin

ωout, Γout

Vbelt = Rinωin

Bishop Brady High School FIRST robotics team
2005 robot (Prof. Slocum was the coach)

•

5-5

1/1/2008

Belts & Cables: Stress, Tension, & Center Distance
To determine the belt cross section required to carry a load F, pretension T and stresses induced by passing around the pulley of diameter D, start
with determining the diameters of the pulleys needed. The pulleys should meet
the optimum transmission ratio requirements, and be of a minimum size for the
cable diameter or belt thickness. The minimum pulley size depends on the
type of cable or belt used, and is best determined using design charts provided
by the manufacturer, or from a design handbook. In general, the stress is
inversely proportional to the pulley diameter. A thin flat belt of modulus E,
poisson ratio η, conforming to a pulley of diameter D, undergoes plain strain1,
and the resulting stresses in the belt are determined in bandstress.xls:

σ load =

F
wt

σ pre − tension =

T
wt

σ pulley =

(

D 1 −η 2

)

A key factor in manufacturing systems with belts or cables is the
space constraint imposed by pulley size and center distance. As a primary factor, the need for a reasonable pulley diameter can have a huge effect on the
space required for the design. For a flat metal belt, proper pulley size is based
on the stress it produces in the belt. For a molded rubber (including reinforced)
belt or cable, the ratio of pulley diameter to cable diameter depends on how the
belt is reinforced or the number of strands in the cable. The pulley diameter
should be 30 to 50 times the belt thickness or cable diameter. For string and
hemp rope, 10:1 is the minimum ratio. The larger the ratio, the greater the life.
Once the pulley diameter is selected, the center distance can be determined based on the total amount of space available, the length of available
belts, and the need for a tensioning system. The law of cosines provides the
basis for finding belt length as a function of pulley diameter and center distance, but finding the angle at which the belt is tangent to the pulley results in
non linear equations best solved iteratively. The spreadsheet pulleycenterdistance.xls can easily accomplish this calculation using the “Goal Seek” tool in
ExcelTM to find a center distance that drives the error value to zero. For example:

The strain is all in a plane (the cross section remains rectangular)

With respect to tension in the belt or cable, the capstan effect can be
used to determine the required tension needed to transmit a desired torque or
force, such as done in the spreadsheet capstan.xls. How should the tension be
determined? An estimate can be made by measuring the force F needed to
deflect the belt sideways a small amount δ. A simple more accurate means is
to pluck the belt and measure its frequency of vibration ω (Hz). Given its
length L and mass per unit length q, the pre-tension T is determined in the
spreadsheet bandstress.xls from (units of N, m, kg, s or N, mm, g, s):

T = FL
2δ

= (ω L ) q
2

Nmks

= (ω L ) q 10 6
2

Nmmgs

Consider the belts or cables you are planning on using, and determine
the minimum pulley size to give you the life needed for your robot. In general,
the pulley size shown in handbooks is for millions of cycles, so you may be
able to use a smaller pulley (run an experiment!). What stresses and pre-tensions exist? Will you need to use a toothed belt? Will your belted module still
meet the functional requirements of your concept?

Belts & Cables: Stress, Tension, & Center Distance
•

Belts and cables are very robust elements,
but they require engineering of 3 basic
details:
–

Stress from wrapping a belt or cable around a
pulley
Tension in the belt or cable
Center distance between pulleys

–
–

θ
R2

pulley_center_distance.xls
By Alex Slocum, last modified 2/19/04 by Alex Slocum
Pulley Center Distance Calculation
Enters numbers in BOLD, Results in RED
Belt pitch, P
Number of teeth on belt N
Start with a guess for C, and then use Goal Seek
Desired center distance, C
Large pulley pitch radius, R2
Small pulley pitch radius, R1
Length of belt, Lt, Lf
Tangent segment, A
gamma, g
Phi, f
Theta, t
error (use "Tools/goal seek" to set value of B17 to zero by changing B9)

5-6

Toothed belt
0.2
100 Enter the number of teeth on the belt,
if center distance ends up being too
small, increase number of teeth
8.24
0.5595
0.5595
20
8.2423
0.0000
1.5708
1.5708
0.000

Bandstress.xls
By Alex S locum
Last modified 4/22/02by Alex S locum
S tress in a flat belt wrapped around a pulley
Enters numbers in BOLD, Results in RED
Belt parameters

Thickness, t (mm)
0.10
Width, w (mm)
5
Modulus, E (N/mm^2)
2.00E+05
Poisson ratio, n
0.29
Forces
Load to be carried, F (N)
10
Belt stress, sigF (N/mm^2)
20.0
Pulley wrap stresses
Pulley diameter, D (mm)
50
stress
437
Motor torque required (N-mm)
250
Capstan effect
coefficient of friction, mu
0.2
Wrap angle, q (degrees)
180
required pre-tenion, pT (N)
5.3
Belt stress, sigT (N/mm^2)
10.6
Total stress
467
Total strain
0.23%
Check: Tension in the belt (pluck it like a guitar string)
Measured frequency of lateral vibration (hz)
150
Free-length (mm)
300
density (g/mm^3)
7
mass per unit length (g/mm)
0.004
Tension (N)
7.1

1/1/2008

Belts & Cables: Linear Motion
There are many examples of belts and cables being used to convert
motor torque and rotation to linear force and displacement, where generally
linear motion bearings provide guidance for the load being moved. The belt or
cable wraps around two pulleys, and the load is attached to the belt or cable in
between the pulleys. Because the attachment zone is usually not intended to
travel around the pulley, a spring or screw actuated tensioning mechanism can
be used to maintain or create belt tension. When there is a need for the attachment zone to pass around the pulleys, special belts can be purchased, or a chain
can be used. An example is the cable on a ski lift.
A load is often attached to the belt at a zone between two pulleys, and
the belt or cable wraps half way around the drive pulley. As shown on page 57, this limits the capstan effect that can be achieved. For a given tension, if
more force is required, multiple wraps around the drive pulley are needed. In
this case, as the cable or belt winds in on one side of the drum and lets out from
the other side, the helix angle formed by the cable diameter divided by the
drum diameter implies that the cable will translate along the drum. Thus the
drum needs to be long enough to give the cable enough room to translate. In
addition, for precision motion and to prevent slip, there needs to be a portion of
the cable that never leaves the drum. Otherwise there is slip caused by a portion of the stretched cable accumulating on the drum and then being let off the
other side. Furthermore, tension needs to be maintained in the band. Maintaining tension is important if precision linear position is to be determined
using a rotary encoder to measure drum revolutions. If unsure, run a benchlevel experiment!
Because a cable’s diameter is typically smaller than a belt’s width,
cables are often used for linear motion systems. In addition, the cable can be
wrapped around a drive drum. This takes advantage of the capstan effect to
minimize the required pre-tension and minimizes radial loads on the pulleys’
bearings. A very common example of a linear motion system driven by a cable
is the scanning shuttle in a photocopier. Open the lid and watch! If the total
travel distance is L (+/- L/2), and the drum diameter is D, then with a safety
factor of 1.2 to prevent slip, the number of wraps N on the drum should be:
N = 1.2

L
πD

Consider a semiconductor test equipment application. The task was to
index a tray of packaged semiconductor Devices Under Test (DUTs) under a
test head and then plunge them into the test head’s electrical contactors. As the
picture shows, a steel belt can be used instead of a cable because a thin flat
steel belt can conform to a smaller diameter pulley while having a much higher
axial stiffness. The difficulty was that the devices had to be kept in an oven so
they could be tested at 100 oC! It was desired to keep the servo motor and sensors outside of the oven. Note the two wraps of the steel belt around the capstan drive. Why is the bench level prototype system at an angle? The system is
at an angle because of a software error that caused the axis to plunge before it
was properly positioned. The design engineer working with the software engineer was assured that this could never ever happen; however, the mechanical
engineer decided to mitigate possible software risk with the countermeasure of
breakaway points!
As another example, consider the top view picture of the hockey puck
dispensing module designed for the 2.007 robot competition MechEverest.
The module moves a line of hockey pucks forward so they fall into a slot to
score. Given space constraints, the force cannot be applied through the pucks’
center of mass. The Functional Requirements of the module are to provide linear force and motion from a distance. This requires two design parameters: a
mechanism to create the force, and a guidance mechanism to withstand the
moment caused by the force. The force is created by a string that wraps around
the capstan attached to the motor shaft and a pulley and is attached to a carriage. The carriage is made from a plastic block with two holes which rides on
two welding rod shafts that provide guidance. Any load applied to the carriage
creates a moment that is withstood by the welding rods, so no moment is
applied to the cable.
For each linear motion module for which you have created concepts,
investigate the feasibility of using a belt or cable drive. Sketch concepts of the
modules and use the spreadsheet Bandstress.xls to perform the design calculations. When is a capstan drive, where the cable makes several wraps around
the drum attached to the motor, most effective, verses a simple belt that passes
around two pulleys, one of which is driven?

Belts & Cables: Linear Motion
•

Belts & Cables are a effective way to convert rotary to linear motion
load

•

The force F in a belt with tension T on a pulley of diameter D that can be
generated by the torque Γ can be conservatively estimated by:
• F = 2Γ/D for toothed belts, and F = 2Γ/D for flat belts but is limited by the
capstan effect
– Play with bandstress.xls
– The speed is simply
• Vlinear=ωmotor*D/2

•

Belts run on pulleys
– For flat belts, pulleys must be crowned to accommodate misalignment
– Timing belt pulleys must be aligned to prevent premature failure
Should this metal belt driven axis be at
this angle? NO, it was the software
(honest) that made the vertical axis
plunge while the metal belt driven
horizontal axis was still extended…

5-7

1/1/2008

Belts & Cables: Crawler Tracks
The wheel is undoubtedly one of the greatest inventions in history. It
enabled people to spread far and wide, and to travel and trade. An interesting
historical artifact of wheeled vehicles are the tracks (ruts) they have left in
many trails. In fact, there are places in the original Oregon trail where you can
still see the ruts left by wheels. This also likely means that there was mud, and
who has not been stuck in the mud or snow or tried to ride a bike in soft sand?
With the advent of the steam engine came the tractor to replace the horse, and
the weight of the vehicle and its payload increased dramatically; and with it, an
even greater tendency to sink in the mud. At what point did someone think
“there has to be a better way!”? It turns out that in the late 18th century someone thought to spread the load out onto a moving track. Was this inspired by a
belt on a pulley? Later in the 19th century with the advent of steam power,
who first thought that a steam powered locomotive could economically run on
steel tracks that sat unused most of the time? We may never know.
As tractors grew larger and heavier, the size of wheels required to
keep them from sinking became ridiculous. Then on November 24, 1904, Benjamin Holt replaced the rear drive wheels of a 40 horsepower Holt Junior Road
Engine Number 77 with a pair of tracks nine feet long and 2 feet wide. The use
of tracks not only helped keep tractors from sinking into the soft earth, reciprocally, they also helped keep the tractor from compacting the soft earth making
fields easier to plow.1
Are crawler tracks merely used to keep tractors from sinking into the
mud? Instinctively it seems that they should also have better traction; but why
then do we even bother with wheels? Given the functional requirement to not
slip (spin your wheels), there are two possible design parameters: 1) frictional
contact between the surfaces, 2) interlocking contact between the surfaces.
If the strategy selected is to rely on friction contact, in general, the
coefficient of friction, on a macro scale, is independent of the surface area of
contact. The tractive effort is a product of the weight of the vehicle and the
coefficient of friction between the wheels and the surface. Therefore, when
creating detailed concepts for a frictional drive interface, selecting the best
materials is the primary design parameter. In robot competitions, it is impor1.

From www.caterpiller.com Also search on www.Google.com for caterpillar history tractor

tant to do simple tilting plane friction tests to determine which materials in
your kit have the highest coefficients of friction with respect to the surfaces on
which they will be driving. When doing these tests, you can experiment with
the area of contact and the type of surface (smooth or carpeted) to see what is
best. The secondary design parameter is the tread or surface features on the
wheels. In real-world vehicles, the treads act to channel water out of the interface, so the rubber can make good contact with the road. The tread is also
important when driving on softer surfaces, or on snow, where the tread helps to
maintain traction by interlocking contact with the surface.
If the primary strategy is to utilize interlocking contact, then the principle design parameter is no longer the coefficient of friction, it is the shear
strength of the contact region. In this case, the amount of tractive effort is a
function of the area and geometry of the contact points and the shear strength
of the materials in contact. Given the width w of each track, the number n of
track elements on each track that are in contact with the ground, the length L of
each track element (the pitch), and the shear strength τ of the ground (which is
often locally higher because the weight of the tractor does some compaction!),
the tractive effort of the machine will be F = 2wLτ.
Tracked vehicles can use their incredible tractive effort to enable them
to turn by driving one track forward while driving the other backwards so they
can turn in place. This is called skid steering, but this results in large lateral
forces on the tracks. Simple rubber belts on crowned rollers used as tracks on
robots can come off when skid steering. Real tracked vehicles have a center
guide feature that engages grooves in the guidance sprockets to resist lateral
forces. Can you make center guiding ribbed features on a rubber belt?
Note in the above, it is important to think of the process by which
frictional versus interlocking contact means are considered. How then can you
decide which system to use for a robot contest? What are the functional
requirements for tractive force? If the competition surfaces are smooth,
crawler tracks provide no direct tractive advantage. But carpet can be engaged
by sandpaper belts... What are the functional requirements for mobility?
Crawler tracks can also allow a robot to go over obstacles that may otherwise
cause a wheeled vehicle to hang up in the middle. Do tracks need a suspension
system like wheeled vehicles?

Belts: Crawler Tracks
•

Tracks primarily help when there is loose media or a surface into which they can dig
– Track treads can be made by gluing rubber strips onto a belt’s surface

•

Tracks (and treads) usually do not help on smooth surfaces unless:
– A “sticky” elastomer is used where apparent coefficient of friction also depends on contact area
• Its apparent coefficient may actually be higher at lower contact pressure
– Dust or dirt on the surface needs to be penetrated/pushed aside by tread teeth

•

Tracked vehicles skid steer, and as a result there are often large lateral forces on the
tracks which try to pull them off the drive and idler sprockets
– A central ridge can resist these forces

n ...

www.caterpillar.com

5-8

1/16/2008

Belts & Cables: Rotary Motion
Belts and cables were first created as rotary transmission elements. In
the early years, a water wheel turned a large shaft that ran the length of a factory. Numerous pulleys located along the shaft transferred power from the
main shaft to smaller shafts, which in turn often transferred the power to other
machine elements. This system was used until electric motors were invented
to deliver power where it was needed. Belts and cables are now used with
electric motors to efficiently transfer power from one axis to another while also
achieving modest transmission ratios. Belts and cables also do not require
lubrication and they are generally tolerant of misalignment.
The torque capacity of a smooth belt transmission is largely affected
by the capstan effect. This also effects the center distance between pulleys.
The greater the diameter ratio of the pulleys, the farther apart they need to be in
order to get a reasonable wrap angle around the smaller pulley. Unless the pulley wrap angle is reasonable (90 degrees or more), high tension will be
required to transfer significant torque. Furthermore, the coefficient of friction
also has a great effect, and thus the system must be kept very clean. If excessive belt tension is required, it can lead to shaft overloading. Given these
issues, for high torque applications, toothed belts are often used.
The accuracy of motion, or how well the output shaft follows the
motion of the input shaft, depends on backlash and belt stiffness. The former
almost by definition does not occur in a belt transmission else there would be
no torque transmitted. Problems can occur if a toothed belt is used and its tension is too low. When torque is reversed, lost motion (backlash) will occur as
the tension in the other side builds up to handle the reversal of load. In addition to the accuracy of motion transferred, the pulleys must be accurately
aligned or else the belts will either wander off, or if edge guides are used,
excessive edge wear can occur. The primary means of preventing pulley misalignment is good rigid shafts and bearing supports to prevent angular deflection of components subject to belt tension.
A principle feature of a rotary motion cable driven system is that a
cable can bend in three dimensions. Hence the axis of rotation of the motor
does not have to be precisely aligned with the axis of rotation of the output
shaft. When more torque is needed than a cable wrapped around a pulley (capstan) can supply, a cable with beads molded onto it can be used. A bead belt,

also enables power transmission to occur in three dimensions. Another example is the use of rubber bands riding in grooved wheels to make a two-wheel
drive vehicle into a four-wheel drive vehicle as shown on page 5-14. The rubber bands’ stretch provides enough tension to transmit the torque, without
requiring the wheels’ center distance to be adjustable.
When only partial rotation of a pulley is required to transmit high
torque, such as to raise a boom, the ends of the belt (or cable) can simply be
anchored on the pulleys. If very limited bidirectional motion is required, the
midpoints of the belt can be secured to the pulleys to prevent slip. Look
closely at the pictures that show a bucket on the front of a mobile robot. A
string is tied to a pulley attached to the bucket, and the other end is tied to a
shaft extending from a motor. The string is used as a transmission to increase
the motor torque to allow the bucket to be raised. Gravity (it’s free!) is used to
lower the bucket, and the overall mechanism is greatly simplified.
Multiple wraps around the pulleys can enable motion with almost a
full revolution and with high torque, accuracy, and stiffness. Commercially,
this type of system is available as the Roto-LokTM drive from Sagebrush Technology. This type of cable transmission is used for many types of precision
systems where only partial rotary motion is required with high stiffness and no
backlash. The key to any cable drive is maintaining cable tension. If the cable
is anchored with a clamp, only steel cables will not stretch with time. A spring
can be used to maintain constant tension on one end of the cable. The drums
often have helical grooves on them in which the cables lay, but for robot contests, smooth surface cylinders typically suffice.
Think of all the rotary motions that occur in your machine (or strategy) and ask yourself how a belt, which would allow a motor to be located
away from the actuated component, might make the design more robust or easier to build? Where are you planning on using gears, and how might belts
replace them? What would be the physics, cost, risks, and countermeasures?
Is the capstan effect (how many wraps do you need?) sufficient for your belts
to transfer torque, or do they need to be physically anchored?

Belts & Cables:
Rotary Motion
•

Flat belts and cables (string drives) require higher tension to transmit torque
–

•

For first order analysis, the belt tension needs equal the desired torque divided by the coefficient of
friction and the small pinion radius, BUT CHECK WITH THE CAPSTAN EFFECT EQUATIONS!

Vee-belts use the the principle of self help:
– Increased tension caused by power being transmitted, wedges the belt in a Vee-shaped
pulley groove, so it can transmit more torque…
Synchronous Drives (timing, gear, ladder belts) can transmit torque between shafts and also
achieve a transmission ratio
–

They combine the positive timing action of gears with the flexibility, speed and low noise level of belts
•

For an in-depth discussion on synchronous drive design, see Stock Drive Products on-line tech library:
http://www.sdp-si.com/Sdptech_lib.htm
T+dT
N=Tdθ
dθ

μN
T

Ken Stone, Director of the
MIT Hobby Shop, and his
string-powered lathe
© 2008 Alexander Slocum
web.mit.edu/hobbyshop

Roto-Lok drive from Sagebrush Technology Inc.
–http://www.sdp-si.com/Sdptech_lib.htm

5-9

1/1/2008

Belts & Cables: Manufacturing & Assembly
There are two primary manufacturing and assembly issues with belts
and cables: manufacturing the belt, and assembling the belt and its associated
components. Belts are often continuous, although they can be spliced to obtain
a desired length. Belts are also often fiber reinforced to give them more
strength and to prevent them from stretching with time. The primary function
of a belt’s rubber or polymer casing is thus to form the gripping surfaces. In
robot competitions, however, you may sometimes need to make your own belts
from rubber sheet or O-ring material. This can be accomplished by overlapping the material and then cutting through it with a sharp knife. The resulting
cuts will match and a suitable adhesive, such as cyanoacrylate (Super GlueTM),
can be applied. Because these belts will not have a fiber core, they can creep;
therefore do not leave tensioned rubber belts on your machine when it is being
stored. Nylon string belts can be made by holding both ends to a soldering iron
and then forcing the melted ends together.
When assembling a belt system, the pulleys’ axes of rotation can
never be perfectly parallel, so a flat belt will want to drift off the pulley. This is
called tracking. to keep a belt from moving off the pulleys, they must be
crowned (rounded profile) or one of the axis be made so it can be oriented with
respect to the other. The crown forces the belt material on either side to want
to climb towards the middle. Neither side can win this tug-of-war, so the system is stable. A flat pulley is at best neutrally stable. On a concave surface,
the side with more belt in contact will cause the belt to drift further to that side
until it falls off.1 The crown need not be accurate, and it is easily created on
the lathe; however, merely forming a raised central portion on a flat wheel, for
example by placing an O-ring on the wheel, does not work. The crown must
be formed as a continuous radius.
Cables need a pulley with a groove in its circumference to keep the
cable on the pulley. The larger the radius of the groove, the more forgiving of
misalignment and the less wear of the pulley caused by the cable as it enters
the groove. On the other hand, the closer the radius of the groove is to the
cable, the less the cable is compressed under heavy loads. For robot design
contests, pulley misalignment is often a dominant manufacturing issue. The
1.
Schwamb, P., Merrill, A.L., James, W.H., Elements of Mechanism, Third Edition, John Wiley &
Sons, 1921.

figure shows how the Abbe effect dictates the groove radius needed to accommodate pulley misalignment. Conservatively, the groove radius R required to
accommodate a cable of diameter d on a pulley of diameter D misaligned by an
amount δ is:
R=

Dδ d
+
2
L

Tension in the belt or cable can be maintained by stretching the belt
by displacing one of the pulleys, using a spring-loaded idler pulley (which also
should be crowned), or connecting the ends of the belt or cable with a tensioned spring element. Regardless of the method used, being able to consistently set the belt tension to the desired value has a great effect on the
repeatability, and hence quality, of a machine. In fact, as shown in the spreadsheet bandstress.xls on page 5-6, the tension T in a belt or string can be determined from the measured frequency of lateral vibration ω (Hz), the free length
L, the cross sectional area A, and the density ρ:

T = ( ω L ) Aρ
2

A common example of an idler pulley used to maintain tension is the
derailleur on a bike with gears. The tension in the chain from the rider pedaling occurs directly between the sprockets, and one function of the springloaded derailleur is to maintain constant tension in the chain even as the ideal
length of the chain changes as the rider changes gears.
When timing belts are used, they are typically tensioned by moving
the bearing assembly that holds one of the sprockets, and then locking it in
place. However, the tensile forces from the belt could cause the pulley assembly to become misaligned. In addition, pulling the belt tight with one hand
while tightening the holding screws with the other can be difficult. It is better
to build the system with a cable tensioning bolt to pull (or push) the pulley
assembly so the belt reaches the desired tension.
The bane of many a machine with belts is pulley misalignment and
improper belt tension. How will your machine establish and maintain belt tension and alignment?

Belts & Cables: Manufacturing & Assembly
•
•

Misaligned pulleys increase wear and decrease life!
Two pulleys’ axes of rotation can never be perfectly parallel, so a flat belt will want to
drift off (tracking)
– Pulleys must be crowned (round profile) to keep a belt from walking off
• On a concave surface, the side with more belt in contact will cause the belt to drift
further to that side until it falls off
• A flat pulley is at best neutrally stable
• Great, OK, bad, & horrid pulleys:
– Tension must be maintained either by proper pulley center distance and belt elasticity, or a
mechanism to tension the pulleys or a pulley that pushes sideways on the belt

R
Ød

5-10

1/1/2008

Chains
Chains are often used to transmit very large forces and torques relative to their size. There are many different types of chains, and one of the most
common types is roller chain which uses metal links connected together by
pins and spaced apart by bushings. Sprockets, toothed wheels which are a special form of gear, positively mechanically engage the chain so there can be no
slip. Accordingly, chains are extremely versatile. They can be used to transmit
power between two rotating shafts, and they can also be used to convert rotary
to linear motion or to enable linear to linear motion. Just take a close look at a
forklift, where a hydraulic cylinder provides the force for the first stage to
move. How are chains used to move the other stages? Take a careful look at
the photos of the FIRST robot whose telescoping stages use chains and idler
sprockets to extend them.
As you read the following, try to construct a picture of the design in
your mind: An angle plate attached to two bearing blocks on a first bearing
rail is moved vertically by a chain that makes a loop around a drive sprocket
on the motor/gearbox and an idler pulley on the top of a first mast. (This first
mast is attached to a horizontal motion carriage). A linear bearing rail is
bolted to the first mast, linear motion blocks ride on the rail1, and the angle
plate is bolted to the bearing blocks. A chain is attached to the angle plate by
bolts that pass through the links. A second mast is also bolted to the angle
plate so it can move up and down with the bearing blocks. A second linear
bearing rail is attached to the second mast. A linear motion bearing block
rides on the second rail and the robot’s lifting arm is attached to this bearing
block. A chain is attached to the lifting arm and runs over an idler pulley at
the top of the second mast and is connected to the top of the first mast. Now
play the movie in your head: what happens as the drive sprocket begins to
turn?
Another versatile aspect of chains is that they can be laid onto a surface, and anchored at the ends to act as a gear rack (page 6-23); in addition, a
chain can be held to a curved surface to act as a gear. As shown in the photo, a
chain was used as a rack for a FIRST robot. Note how the sprocket was
welded to a clamp-type collar and then mounted on a motor/gearbox shaft so
the force generated at the sprocket teeth by the motor/gearbox torque creates a
1.

See page 10-12

reaction force on the shaft that passes directly through the shaft-support bearing and thus will not harm the motor/gearbox: Motor gearbox shafts are typically only meant to transmit torque and have very limited moment capacity.
Hence this design eliminates the need for a separate support shaft, bearings,
and coupling.
Sprocket center distance and tension in the chain can be determined
using the same calculations as used for belts. With regard to acceptable loads
in chains, manufacturers provide recommendations for different sizes and
numbers of strands. Like belts and cables, tension must be set or maintained;
however, because the teeth on the sprockets are relatively large and protrude
nearly all the way through the chain, chain drives rarely skip and fail to transmit the desired torque. For bi-directional motion, lack of proper chain tension
will result in backlash (lost motion or hesitation) when the motion is reversed.
Chain tension can be set using an idler pulley on a spring loaded arm, as is
done for belts, or by making one sprocket movable so the tension can be set
and then the sprocket locked into place. Idler pulleys are often mounted at one
point and are used to direct the motion of the chain, such as for extending linear axes. Idler pulleys can be purchased that have integral ball bearings with
their hubs, and then a bolt, e.g., a shoulder bolt, can be used as a non-rotating
axle to support them.
Sprockets are readily available in many different sizes. A chain
breaker tool is used to press out pins from bushings so the chain can be made
the proper length and then rejoined using master links. Various special types of
links can also be obtained so other elements can be more readily attached to the
chain. Cutting teeth can also be attached as in a chain saw. All are readily
available from catalogs.
Look at the chain, sprockets, and shifting system on a bicycle. How
does the system accommodate the different length of chain required to go
around different sets of sprockets? Look at a forklift, how are chains used?
What other devices use chains? Can a chain, driven by one sprocket, be used
to drive multiple sprockets? Could a chain be used as a crawler track on a
robot?

Chains
•

Chains are a very robust means to transmit very large forces and torques
– Rotary-to-rotary motion is very common (look at your bicycle!)
– Rotary-to-linear motion can be obtained by laying a chain on a surface, and
anchoring each end with bolts through the links
• A sprocket mounted to the motor "crawls" along the chain
• The sprocket can be welded or bolted to a squeeze-type collar…
– Linear-to-linear motion
for extending axes
• Idler pulleys are
invaluable
Bishop Brady High School's rookie FIRST robotics
team used chain drives to realize a very simple and
effective robot that earned the Rookie All Star Award
and Top Rookie Seed awards at the 2005 Manchester
regional competition and a trip to the finals in Atlanta.

5-11

1/1/2008

Chains: Engineering
It is a simple task to design a system that uses roller chains to ensure
that they can carry intended loads for the life of the chain. Roller chains’ bearing contact interfaces between the rollers and pins act as mini shock absorbers,
but their load capacity needs to be decreased by a kservice factor that corre-

The size of the roller chain, in the USA, is referred to as the Chain
The pitch, distance between the rollers’ centers, increases with
chain number, as do other characteristic dimensions and load capacity:

Number2.

ANSI Steel Roller Chain (units of inches and pounds)
Roller
Link plate Breaking
Chain # Pitch
Pin Diam.
thickness
Load
Diam. Width
25
1/4
0.130 0.125
0.091
0.030
1050
35
3/8
0.200 0.188
0.141
0.050
2400
41
1/2
0.306 0.250
0.141
0.050
2600
40
1/2
0.313 0.313
0.156
0.060
4300
50
5/8
0.400 0.375
0.200
0.080
7200
60
3/4
0.469 0.500
0.234
0.094
10000
80
1
0.625 0.625
0.312
0.125
17700

sponds to the type of system and load they experience1:
kservice factor
Loading
Smooth
Moderate shock
Heavy shock

Input power
Hydraulic Drive

Electric Motor

1.0
1.2
1.4

1.0
1.3
1.5

I.C. Engine &
mechanical drive
1.2
1.4
1.7

Single row chain is very robust to misalignment of sprockets, because
the sprocket teeth protrude into the chain it cannot wander off the sprocket the
way a belt can wander off a misaligned pulley. Multiple strand (row) chain is
used to transmit very large forces with a low profile, or to provide redundancy
should one link fail. With each added row, misalignment capability decreases.
Merely adding a second row to a chain does not double its load capacity:
Number of Strands
1
2
3
4

Multiple Strand Factor
1
1.7
2.5
3.3

Kstrand factor
1.00
1.18
1.20
1.21

Given an applied chain tension T, be it from a linear or rotary motion
application, an equivalent tension has to be determined to reflect the type of
load and number of strands of chain required. An equivalent tension must be
below the acceptable limits for the particular chain used, as specified by the
manufacturer in order to obtain the desired life:

T equivalent = T applied × K service factor × K strand factor

ASME Standard B29.1M-1993

Working
Load
140
480
500
810
1400
1950
3300

Some practical use guidelines include:
•
•
•

•
•
•
•
•

•

The total length of chain loop should not exceed 100 pitches
Number of sprocket teeth: <17 low speed, 17-21 moderate speed, >21 high
speed.
Due to the manner in which the rollers engage the sprocket teeth, assuming the driver sprocket is rotating at a constant speed, the driven sprocket
speed will vary by cos(2π/Nteeth on the driven sprocket)
Number teeth large sprocket/Number teeth small sprocket < 7
Noise increases with increasing Load/Maximum Load
At least 1/3rd of the teeth on a sprocket should engage the chain
Idler sprockets can be used to maintain tension
Lubricated chain and sprocket life can be long because the teeth nominally
have rolling contact with the bushings, although there is enough slip to
require lubrication.
Use protective covers to keep stuff out of the chain drive!

Note the factor of 7 difference between the working load and the
breaking load for a chain. Determine the size of the chain on your bicycle and
based on the crank length, determine how hard you could push on a pedal
before you snapped the chain on your bike. How strong are you?
2.

Size 25 and 35 chains’ rollers do not roll on the bushings.

Chains: Engineering
•
•

There are many different components that designers can use to create systems
A chain breaker (a press) pushes out the roller chain's pin
– Master links are used to splice chain ends together to form desired length loops
• Master links are available with fittings to allow things to be attached to the chain

•

Idler pulleys allow for routing of the chain
– Idler pulleys can have integral ball bearings
– A bolt through the center supports and anchors the idler pulley

5-12

1/1/2008

Wheels
We can only guess the first ancient human thought who perhaps saw a
rock rolling down a hill. Did they think that since a heavy rock could roll,
maybe other heavy things they were carrying could rest on top of the rock? Or
maybe Og stepped on a log which rolled out from under him? Or maybe
Ogette was stacking firewood and the stack became too high and started to roll
apart? Maybe she observed a tree falling on top of two other tress and then the
assembly rolled away? Whatever the origins, there are several important lessons: We are a race of builders who takes advantage of what others have done;
and we are a race of observers who learns by observation and experimentation.
Consider some of the typical functional requirements for a wheel, and think
about the design parameters and physics that come to mind.
Table 3:
Functional Requirements

Design Parameters

High speed

D diameter (large)

Climb over obstacles

D diameter (large)

Analysis

Vvehicle= 2ωmotor/Dwheel

F ≥ FM

Dh − h 2
D −h
2

Controllability

Wheel size, steering linkage

Larger wheels go faster, may be
more difficult to control

Traction

Materials in contact, vehicle
weight distribution

Ftraction= μFnormal
Ftraction= 2Γtorque/Dwheel

Easy attachment

Flange, spline

Torques and moments

What other functions do wheels have? Why are bike tires not hard
rubber so you would not have to worry about them going flat? Why are they
not filled with foam rubber? Why is it easier to ride a bike with big wheels
than with little wheels? And why are little kids bikes not designed with big
wheels? In addition to bicycle wheels, what are the kinematics, dynamics, and
economics of wheel-based vehicles? What are the risks and countermeasures
associated with each design choice?

The kinematics of a wheel can be determined by realizing that the
instant center is the contact point between the wheel and the ground. In the
case of a wheel encountering an object, the instant center is at the contact point
with the object (e.g., a curb). When a vehicle is supported by many wheels, as
is typically the case, the primary issue is the alignment, or in some cases, purposeful slight misalignment so as to help bias a high speed vehicle against spin
outs. When a car turns, the linkage must adjust the angles between the wheels
and the car body, but also the relative angles between the wheels may be
adjusted because each wheel takes a separate path through the turn. When the
wheels on your car are misaligned, the result is typically uneven wear and
higher fuel consumption as well as the potential for poorer handling. In a robot
contest, misalignment can result in lost tractive capability and poor controllability.
The dynamics of wheels are principally a result of their angular
momentum. Large wheels’ moments of inertia increase with the 4th power of
their diameter (3rd power for a thin rim spoked wheel), and this directly affects
the torque required to get a wheel up to speed. Often the reflected inertia of a
large wheel (see page 7-4) can be greater than the inertia of a vehicle itself!
What about angular momentum? Can it help or hurt the performance of a vehicle? What about the speed of a wheel on a surface and the potential for entrapment of air or liquid (hydroplaning)? How can it be prevented?
The economics of wheels are terrific in that they are simple to design,
manufacture, and implement because they have so few parts compared to, for
example, a track system. Their simplicity is perhaps how/why they were discovered ages ago and have been in use ever since. However, for specialized
applications, even though they may be fundamentally simple, the details are
what matters. For example, adhering a layer of sandpaper or Velcro to a plastic
wheel to greatly increase its tractive effort on a soft surface such as carpet.
Wheeled vehicles are very simple to design and build and they work
well on even surfaces, so make a simple vehicle and test it on the contest table
as soon as possible. Would you be better off with a tracked vehicle? Can a
wheeled vehicle be made to skid-steer so it can turn in place? What makes
street sweepers so agile? What benefits can your robot obtain if it is highly
agile? What are the trade-offs to using one motor and control channel to control the steering system?

ØD

Wheels

FS
h

Home-grown Omnidirectional wheel

Jwheel 75 mm diameter = 7.1e-5 kg*m^2

•
•
•

Traction & Controllability
Size, Torque & Contact Pressure
Manufacturing & Mounting

Photo by Rick Slocum www.100jpegs.com

What does SaintVenant say about
this overhang?

Wheel fixed to chain driven
shaft supported by pillowblock
bearings
© 2008
Alexander Slocum

5-13

1/1/2008

Wheels: Traction & Controllability
Traction is the ability of a wheel to turn without slipping and this
directly affects the force that the vehicle can exert on its own mass and on other
objects. Controllability is the ability of the wheel to steer the vehicle. Steering
can be done using a linkage and is affected by how the wheel makes contact
with the road as it is turned. Steering can also be in the form of pure traction if
skid steering (differential speed between the wheels) is used. Without good
traction a vehicle will have poor controllability1. Consider front wheel drive
cars, which have better traction than rear wheel drive because more of the car’s
weight is over the front wheels (unless the trunk is filled with Scuba gear!) If
tractive forces are being applied that are nearly all of the total tractive effort,
will there be enough tractive effort left for the steering effort?
Steering a fast moving vehicle, such as a bicycle going down a steep
hill, can be easier than steering a slow moving vehicle. The angular momentum principle requires that the torque vector in a system be perpendicular to the
time rate of change of angular momentum. If a wheel is spinning about an axle
at a high rate of speed, and you attempt to rotate the wheel about another axis,
a torque on the system will be produced about the third perpendicular axis. If
you were to sit in office chair that can swivel while holding horizontal a spinning disk, when you tilt the spinning disk, your chair will start to swivel.
When the spinning disk is supported in a gimble, a frame that allows the disk to
pivot about two axis orthogonal to the spin axis, it forms a gyroscope, or gyro.
From this basic principle, rate gyros and control gyros evolved: A rate gyro’s
support axes are free to spin, and sensor’s measure their orientation. By
observing the orientation of the support axes, one can deduce the rotation rate
(angular velocity) of the system that holds the gyro. Conversely, a control gyro
uses torque motors to apply torques to the gyro’s support axes, and the resulting reaction torques on the structure cause it to move with respect to the spinning gyro. How might robots use these principles?
Traction is important for steering and for moving forward, and a primary design parameter for enhancing traction is the normal force between the
wheel and the surface. How does driving up an incline2 affect the traction of a
1.
In addition, in robot contests in particular where humans often control the robots, if the wheels are
too big, then a small signal from the controller will send the robot zooming off into a barrier.
2.
Look ahead to Chapter 8 for a detailed discussion of system stability

two-wheel-drive vehicle verses a four-wheel-drive vehicle? Does front-wheeldrive hurt or help? How else can the normal force be increased? Another primary design parameter for enhancing traction is the effective coefficient of
friction between the wheel and the surface. In many engineering applications,
surface finish does not affect friction, but in the case of a wheel running on a
rough surface in the presence of dirt, treads can help to push aside debris and
help prevent hydroplaning. In the case of a soft surface, treads or other features can actually engage features on the surface to create an effective coefficient of friction greater than 1.
Wheels can also be used to drive components, and in this application,
they are commonly known as friction drives. Perhaps the most common friction drive is the use of wheels to control the motion of paper in ink jet printers.
Are the wheels smooth rubber, or do they have tiny surface features that
engage the micro rough surface of the paper, thus driving it like a gear rack?
Coordinate Measuring Machines (CMMs account for 10% of all machine tool
sales) often use friction drives to provide an extra margin of safety. If a person
gets in the way, the friction drive wheels will slip before can hurt the user.
The primary design challenge with friction drives is that the preload
force must be greater than the desired drive force/coefficient of friction! In
addition to preloading one roller against another, magnets can be used to preload the wheels against the ferritic steel surfaces. This idea has been invented
many times by designers who have created inspection robots to drive along
metal pipes. If magnets can preload the wheels to the surface, then the same
magnets can preload a small wheel attached to a motor to the magnet wheel,
thus creating a friction drive transmission between the motor output and the
wheel. This is the concept for the Magnabot robots for hospital automation
designed at MIT3. The goal in this case was to create robots that could drive
on ceilings in halls and stairwells to automate delivery functions in hospitals.
Plan ahead for steering and traction. When you get your kit of parts,
play with the kit materials to evaluate their relative friction coefficients, and
their friction coefficients on the contest table surface! What wheels are in the
kit, and what other materials do you have that might enable you to make even
better wheels?

http://pergatory.mit.edu/magnabots

Magnetic
Wheels

Wheels: Traction & Controllability
Drive
Motor

Nick Powley obtained
4WD by cutting grooves
in his wheels and using
rubber bands as drive
belts…

Timing-belt
Pulleys

Idler
Wheels

•

Friction is independent of surface finish, unless there is physical engagement
– Like gear teeth on a rack, sandpaper grips carpet, plastic…

•

Two-Wheel Drive vehicles are simple to build
– Against a 4-Wheel Drive or a Tracked vehicle, they lose

•

The smaller the wheel:
– The greater the traction force the motors can cause
– The slower the vehicle per unit input to the controller!
Tim Zue’s sandpaper covered platform
enabled him to get to the other side every time!

Torque-Speed for Bosch Motor
(MS-L 2/98)

1.8
1.6
Torque (Nm)

1.4
1.2
1
0.8
0.6
0.4

Tom Slowe made a very successful 2WD vehicle with
sandpaper covered rear driven wheels

0.2
0
0

Speed (RPM)
© 2008 Alexander
Slocum

100

5-14

1/1/2008

Wheels: Coefficient of Friction
A very common question faced by design students building contest
machines is how to obtain a high coefficient of friction between the wheels and
the surface? Various methods tried over the years including gluing rubber
bands to plastic wheel surfaces, and screwing sharp objects into the wheels to
create spiked wheels. The former is not too effective, and the latter is extraordinarily dangerous. In fact, sharp-spiked wheels are like spinning saw blades
and can shred skin. Never make sharp-spiked wheels!
What about treads and how to make them on the surface of plastic
wheels? Treads are generally useful on soft ground where they can dig in and
form interlocking features which then shear before slipping. Plastic wheels
interestingly enough can have treads put on them by mounting them in a lathe
and using a knurling tool to form the tread pattern on the surface; although a
knurling tool is usually used with metal objects, it also works on plastic
wheels!
A compromise on the spiked surface is to glue coarse sandpaper onto
the surface. The prime challenge then is what adhesive to use, and what type
of joint should be used between the ends of the sandpaper. The wheel surface
should be scuffed up with sandpaper, and cleaned with alcohol before the
epoxy is applied. The sandpaper can be cut to “exact” length to form a buttjoint, and the joint filled with epoxy. Rubber bands or tape can be used to keep
radial pressure applied until the epoxy sets. As long as the sandpaper does not
overhang the edges of the wheels, there should not be free edges to catch and
peel.
So which type of tractive surface should be used? What are the risks
and countermeasures associated with each? The simplest solution for the problem at hand is the best, including predicting what happens when you get into a
pushing contest with the other contestant. Do you have enough torque and
friction to climb over obstacles, as discussed on the next page?
The pictures show different wheel surfaces, and a picture is also
shown of a simple bench level experiment to determine the coefficient of friction for each surface on different surfaces. Each of the tests was performed 11
times and the high and low values were discarded. The average of the 9 readings then meant that random errors would be reduced by a factor of 3. The cart

was placed on different positions on the test surfaces, and the variance was less
than 10%. An interesting observation is that the 80 grit sandpaper grits seemed
to interlock with the carpet fibers, so the test cart never slid, but merely tipped
over. hence the value for the coefficient of friction of sandpaper on carpet is
not accurate.
Designers need to be aware that if they are creating a four-wheeldrive machine that steers by driving wheels on one side of the machine forward
and wheels on the other side backwards, a skid-steered machine, that the
wheels have to slide sideways. Hence the coefficient of friction which helps
when driving forward hurts when trying to turn. If the machine has enough
torque, it can overcome the resistance, but control can be difficult. A bench
level experiment is warranted early on!
If designers are creating a two-wheel-drive skid-steered machine,
then the only detail to remember is that the non-driving wheels should have a
very low coefficient of friction between them and the surface. This enables
them to slide sideways easily when the machine is steering, and it has no effect
on the tractive effort of the machine.
How would you obtain an “accurate” coefficient of friction” test for
the sandpaper covered wheels on a carpeted surface? What is the coefficient of
friction of a rubber band covered wheel on a plywood surface? On a carpeted
surface? Would making many small radial slits on the wheels be better than
the knurled surface? To minimize cost (time) and risk, should you just go
ahead and cover your motor-driven wheels with sandpaper? If you have a
four-wheel-drive car, make sure to do a quick bench level experiment to make
sure your motors have enough torque to actually skid steer the car. You may
have too much friction to steer!

Wheels: Coefficient of Friction
•
•
•

Simple inclined plane tests were done
Hypotenuse length = 400 mm
Raise end of ramp till test car slips: Coefficient of friction

μ = tan θ =

h
2
2
L −h

Wheel surface
Plain
Knurled
Sandpaper covered

Surface
Plywood
along grain
across grain
0.45
0.47
0.47
0.48
0.91
0.92

Carpet
0.31
0.64
1.81

– The sandpaper covered wheel tester on carpet did not actually slip, it tipped!

5-15

1/1/2008

Wheels: Size, Torque, and Contact Pressure
The primary design parameters for a wheel are its size and the tread
material. The former affects the overall transmission ratio of the machine and
the size of the object that the wheel can climb over, and the latter affects the
coefficient of friction and hence the achievable traction force that can be
obtained. The width and the shape of the contact surface, may also be of
importance depending on the application. For example, when driving on soft
surfaces, a wide compliant wheel may be desirable to minimize sinking into
the surface. When driving in a hard surface, a narrower wheel has lower frictional losses caused by the fact that the contact area effectively means that the
wheel is rolling simultaneously with two different radii in contact with the
ground.
In order to determine how big a wheel must be in order to overcome
an obstacle, the sum of the moments about the contact point must equal zero:
the tractive effort F moving the vehicle forward must create a moment to overcome the force FM acting downward through the center of the wheel:

(

)

∑ M = 0 = F D 2 − h − F M Dh − h 2
F ≥ FM

Dh − h 2
D −h
2

The tractive force F that is applied depends on whether the vehicle is
a two or four wheel drive vehicle. In the former case, the back wheels are pushing the vehicle forward, and the maximum tractive force that can be obtained is
a function of the motor torque Γ, wheel diameter D, normal force FN, and coefficient of friction μ:

2 μ F NΓ
D

What happens when a four wheel drive vehicle is used? The front
wheels engage the obstacle and quickly lose contact with the ground. Their
tractive effort is dependent on the motor torque, wheel diameter, coefficient of
friction and the tractive force from the rear wheels pushing them against the

obstacle. Unless the coefficient of friction is near 1, which is rare, having four
wheel drive does not necessarily enable the vehicle to twice as easily climb
over obstacles. However, once the front wheels make it over the object, what
happens to the rear wheels when they engage the object? In general, the efficiency of a four wheel drive vehicle in climbing over obstacles is equal to the
coefficient of friction between the wheels and the road, such that the effective
obstacle climbing tractive force a 4WD vehicle with equal weight distribution
on its wheels can achieve is typically (1 + μ) times the tractive effort of a 2WD
vehicle.
The analysis shows that typically the force required to climb over an
obstacle drops in proportion to the ratio of the height of the obstacle to the
wheel diameter. Thus it is no big surprise that bigger wheels can more easily
climb over objects. But how does the use of a bigger wheel affect the required
motor torque?
What about contact pressure and the coefficient of friction between
the wheel and the ground? If the contact pressure is too high, the wheel or the
ground may permanently deform. This is very unlikely in any robot contest
unless the table purposefully has a very soft surface. But does it help the tractive effort to have a high contact pressure? In general, unless there are tread
features that physical deform and engage the surface, the effective coefficient
of friction will not be affected by the surface contact pressure.
For hard wheels on a hard surface, such as nylon wheels running on
metal that act as a linear motion bearing, it is necessary to determine the contact stress. This can be done using Hertz Contact Stress theory as described in
detail in Topic 9. Unless a very compliant wheel is used, it is very difficult to
obtain contact across all the widths of rectangular profile wheels that support a
vehicle. To avoid loading and wearing the edges of the wheels, a slight crown
is typically used.
Experiment with WheelClimb.xls and you might be surprised at how
low an object a 2WD vehicle with modest coefficient of friction can climb!
Nylon wheels on plywood may only have a coefficient of friction of 0.1. No
wonder many a student who did not bother to do the calculations was stunned
to see their great machine spin its wheels when faced with low barriers! What
can you do to increase the coefficient of friction between the wheel and the
table? How can you easily achieve 4WD or increase μ so 2WD is sufficient?

Wheels: Size, Torque and Contact Pressure
•

How big must a wheel be to make sure it can overcome obstacles?
– The smaller the wheel diameter, the lower the torque to turn it
• Ideally, the wheel just slips when maximum torque is applied
– This keeps you from stalling the motor and potentially burning it out

– The smaller the diameter, the higher the contact pressure, and the greater the wear
• Hertz Contact stress theory can determine if wheels are too heavily loaded!
• The bigger the wheel, the easier it is to drive over an obstacle
• See Wheelclimb.xls
Relative tractive effort

3.00
2.50
2.00
1.50

ØD

1.00
0.50
3

Wheel height / obstacle height

FS
h

5-16

1/1/2008

Wheels: Manufacturing & Mounting
There are four basic steps in making and mounting wheels to a robot
vehicle for a design contest: 1) selecting the material of the wheel for use with
the contact surface, 2) making (or obtaining) the wheel, 3) mounting the wheel
to a shaft such that it either spins freely on the shaft, or torque is transmitted
without slip between the shaft and the wheel, and 4) making sure that all the
wheels are properly aligned so that the vehicle efficiency is maximized.
Selecting the proper wheel material is very important. Some wheels
are made of plastic because they are designed for use on carts that are pushed,
so their coefficient of friction with the floor material is not important. However, for a robot contest, a low coefficient of friction is bad; thus if you are
given plastic wheels, or plan on making a plastic wheel, you may want to
adhere a layer of rubber or sandpaper (if allowed) to the outside rim of the
wheel to increase its coefficient of friction with the ground. If the ground is
soft, you may even want to consider making the wheel have teeth, much like
that of a fine pitch gear, so as to help it obtain better traction.
Assume the wheel diameter has been chosen, for example as discussed to ensure the vehicle can overcome an obstacle. If a suitable wheel is
available as an off-the-shelf component, use it! Otherwise you can easily make
a wheel by cutting material out of a sheet of flat stock. Remember, in most circumstances for robot design contests, the width of the wheel does not matter
when driving on a hard or a carpeted table. Beware, that if you are skid steering your vehicle on a carpeted surface and the wheel is too thin, the wheel can
get stuck in its own groove and the vehicle will not be able to turn.
The width of a wheel must also be sufficient to prevent it from folding
in half. In the limit, a piece of paper will fold over sideways if you try to use it
as a wheel. Simple beam bending calculations can be used to estimate the axial
strength of a wheel, or you could do an FEA, or a simple test. What is more
likely to be an issue is the effect of wheel width on its connection to a shaft.
If the wheel is to run true and not wobble as it rolls, then the required
width of the mounting region can be calculated. Assuming the mounting tolerances and the allowable axial wobble at the wheel rim, the wobble is simply a
sine error. For a high speed system, an engineer would create a spreadsheet or
program to calculate the wobble. They would then determine the moment put

on the shaft that would be caused by the changing angular momentum vector
of the wheel. In this manner, the imbalance forces, and hence the vibration in
the system, can be determined. Given an allowable level of vibration, the engineer can determine how accurately or over what size region the wheel needs to
be mounted.
For a typical robot design contest, where dynamic balance of the
wheel is probably not an issue, Saint Venant’s principle can be used as a design
guideline. Assume that if the wheel is of radius R, then if the length of engagement on the shaft is less than R/5, the wheel is likely to dominate the shaft and
we are likely to get excessive wobble, even for a robot contest. On the other
hand, an engagement ratio of 1:1 is probably unrealistic. If we try to engage
the wheel on the shaft with a characteristic dimension of about R/3 or less,
acceptable performance will be obtained.
If a wheel is to be driven by a motor, it must be securely attached to
the shaft. The best way for this to be done is by using a shaft with an integral
hub onto which the wheel is bolted, just like on a car. The next best thing is a
spline connection that can be nearly achieved with a hexagon shaft and a wheel
that has been broached to have a hexagon hole in its center. If this is not possible, then the wheel should be keyed to the shaft. If this is not possible, then the
wheel can be pinned to the shaft; however, take care that the pin and hole are
located on the outside of the wheel such that wheel loads that cause bending of
the shaft do not place the region of the shaft with the hole in bending. As a last
resort, press fit the wheel onto the shaft. These same mounting ideas apply to
gears or pulleys on shafts.
The last item is alignment of the wheels, and this typically involves
making sure the axis of rotation of the front and rear wheels respectively are
parallel. This is easily accomplished by means of a fixture, which can be a
straightedge that is placed along the wheels’ faces. However, this does require
that the bearings that support the axles to be mounted in an angularly adjustable fashion. Many students are indeed surprised when they put together their
robots and they have trouble controlling them because the wheels are so poorly
aligned. In the case of tracked vehicles, poor misalignment can cause belts to
come off even highly crowned pulley wheels.
Have you carefully designed your wheels, connections and alignment
capabilities? Now would be a good time to check!

Wheels: Manufacturing & Mounting
•

A positive engagement is required to effectively transmit torque from a shaft to a wheel
–
–
–

•

A wheel-on-shaft subassembly should overhang the gearbox as little as possible!
–

•

A hexagon hole can be broached into pulleys or wheels and square or hexagon shafts can be
threaded!
Press-fits, keys, or pins also transfer torque, and E-clips are a great way to retain thrust washers
In production, a three lobe polygon is the best interface (see DIN standards)
If the overhung shaft has a lot of compliance, an “outrigger” bearing can be added without over
constraining the shaft, if it confines the shaft to its nominal central position

Beware of system deformations and the fact that they can take up all the clearance in a
system and cause it to bind

Too much overhang (remember Saint-Venant!). Consider the use of
an outrigger bearing, or mount the wheel on a properly supported
shaft and drive the shaft with the motor using a coupling

Differential

5-17

1/1/2008

Wheels: Steering & Suspensions
There are a number of strategies for steering a mobile robot. Each of
which has unique kinematic and dynamic properties. The easier the implementation method, the more modest the performance:
•

•

Differentially drive the back wheels while the front wheels are on fixed
axles: This requires good control of the differential speed of the rear
wheels, and the front wheels must be able to skid across the ground. Thus
the rear wheels should have a high friction surface (e.g., rubber or sandpaper glued to the wheel surface) while the front wheels should have a low
friction surface (e.g., plastic).
Differentially drive the back wheels while the front wheels are mounted on
casters: This requires good control of the differential speed of the rear
wheels.
Use 4WD: The wheels on each side (or use treads) of the vehicle are torsionally connected together, and skid steering (differential speed between
the two sides of the vehicle) is used to turn the vehicle.
Drive the rear wheels with a differential and steer the front wheels with a
linkage: This uses one actuator for steering and thus reduces the total tractive effort of the vehicle. If the contest requires more maneuverability
than tractive effort or speed, this may be the most desirable strategy. One
must make sure to check the required and achievable tractive effort!
Drive the rear wheels with a differential and steer a single front wheel
using a gear that is driven by a pinion attached to a motor: This is the
most maneuverable system, and the 3-point support of the vehicle eliminates the potential need for a suspension, but the vehicle may be less stable.

For each of these strategies, there may be many different concepts
with which to achieve the strategy. As always, cost verses performance must
drive the design decision. In general, the best all-around performance system
is the rear wheel drive with castored front wheels. The best brute-force
machine with good maneuverability is typically the 4WD or tracked vehicle
with skid-steering. The most maneuverable, and strangely not often seen, is
the three wheeled vehicle where the two rear wheels are driven with a differential and the vehicle is steered using a single front wheel attached to a rotating
platform that is steered by a motor.

The vehicle suspension also greatly affects controllability and performance. Most robots built for contests do not have suspension elements, and
instead rely on the compliance of the support structure and prayer to ensure
that their wheels always contact the ground. When shock loads are to be
incurred by driving fast over rough terrain, however, independent suspension is
a must to maintain contact with the ground and to maintain control. In addition, a suspension can soften impacts and help prevent the breaking of components. Accordingly, there are various forms of suspensions. Once gain, the
cost increases with the performance:
•
•
•

No suspension and just rely on the compliance of the support structure
Use a pivot shaft to support one set of wheels so they can move out of
plane with the rear wheels (like the front axle of a small tractor)
Independent suspension for at least two if not four of the wheels

In order to determine which suspension is best, consider the functional requirements for your overall strategy, and assess your available time.
In general, for most robot contests, unless the contest requires a great deal of
speed over a rough terrain, a pivot shaft suspension is more than adequate.
Alternatively, a three wheeled robot contest vehicle probably does not require a
suspension.
Assess the planarity of the contest tables and judge the risk that your
vehicle may end up being supported by three wheels only. What countermeasures can you implement to ensure that, no matter what, the driven wheels
always remain in contact with the table? Does this adversely affect the stability of your vehicle? Would you be better off adding a suspension? What is the
simplest suspension that you require? If you were to make a 3-wheeled vehicle, what countermeasures might your propose to minimize the risk of tipping?
Would outrigger skids or wheels be effective? How can you easily mount the
third wheel and attach a gear to it so it can be steered? Can you attach a gear to
a castored wheel? Where should the steered wheel be located, in the front or
the rear of the vehicle?

Wheels: Steering & Suspensions
•

A steering linkage ensures that all the vehicle’s wheels
axes of rotation meet at a single point, which is the
instant center (center of the turning radius)
–

•

A suspension linkage ensures that all wheels keep
contact with uneven ground
–

–

•

The Ackerman linkage is a 6-bar linkage developed for
automotive steering

The linkage allows vertical motion of the wheel, but
maintains wheel alignment
• Advanced suspensions may cant (tilt) wheels to help
with high speed cornering
Recall page 3-26 and the concept of elastically averaged
Instant center
design

10o

20o

Vehicles with crawler tracks (e.g., bulldozers) typically
allow at least one of the Crawler tracks to pivot to help
ensure the treads to maintain ground contact

Special thanks for the
truck front suspension
images to F. Zafar
Shaikh, Senior
Technical Specialist,
Ford Motor Company

5-18

1/1/2008

Wheels: Steering & Suspensions
Have you ever taken the time to carefully examine the steering and
suspension on a real car? It can be difficult to pick it up and turn it upside
down (physically, but mentally its a great exercise!). Have you ever carefully
examined a radio controlled model car? High performance radio controlled
cars are amazing little machines with full suspensions, differentials, and steering mechanisms. Their critical performance features are often scaled models
of full-sized high performance machines. What to scale and what to redesign
for is governed by consideration of:
•
•
•

Kinematics: motion, accuracy, space
Dynamics: forces, speeds, life
Economics: design, build, maintain

For example, the pictures show close up photographs of a Traxxas RC
4WD off-road truck. The left side sequence shows the front suspension in the
relaxed, half-loaded, and fully loaded positions. Sketch the linkage and label
the critical elements. Notice that the suspension links themselves are essentially a 4-bar linkage. The shock absorber and suspension spring are colinear,
and are anchored to the middle of the bottom link. Would this be done in a
full-scale truck? Why or why not? Do all things automatically scale from
large to small systems? The answer is “no”; however, we can always learn
from one scale as we design on another!
Consider nature, where a cat can jump off a table that is five times its
height. What would happen if an elephant jumped off a 20 meter high ledge?
Often, the smaller something is, the stronger it is in proportion to its size., not
just by having less chance of defects initiating a failure, but from a mass/
strength perspective. In the case of the model truck, the spring/shock absorber
can be more conveniently attached to the middle of the lower link. In a full
size truck, it would be more likely attached to near the wheel. It is a combination of the kinematics, dynamics, and economics that drives this difference
between the full-size and scaled trucks.
The suspension allows the body to move with respect to the wheels,
so the driveshaft must be able to move up and down, and also accommodate
the turning of the wheels. How many degrees of freedom are required from

this joint? Does the driveshaft change length? How does it also accommodate
misalignment? Look at the drive shafts emanating from the black box in the
center of the underside of the truck. Why does there seem to be a tube within a
tube? What kind of joint would allow the smaller tube to slide in and out of the
larger tube while still transmitting torque? Why is the author asking so many
questions instead of just providing a figure with lots of labeled items?
Now look at the steering linkage shown in the figures on the right side
of the page. Sketch the primary elements. How does the link that connects the
actuator which creates the force to cause the wheel to turn, connect to the system? How does the link that connects to the wheels themselves accommodate
the up and down motion of the suspension? How many degrees of freedom are
required at each joint? Why does it seem that there is a rotary motion link
attached to a four bar linkage that then pushes links attached to the wheels?
Would a full-size truck or car use such a system? Why can this system get
away with this type of design? Does it have anything to do with the mass of
the truck relative to its size?
In addition to shafts that must accommodate large motions, there are
many precision shafts that support gears which must remain in precise alignment. These shafts must be supported by precision bearings (see Chapter 10!)
which get hot when they spin at high speed. What about thermal expansion?
Even very small displacements can cause HUGE forces, which readily destroy
bearings when they are applied to very stiff systems. Hence the importance of
deterministic design and the identification of both the degrees of freedom
required and the sensitive directions of the system!
All those fundamentals of Chapter 3 apply over and over! Wire the
fundamentals into your bio neural net so whenever you are faced with a problem (opportunity!), immediately dissect it in terms of the fundamentals!
Always ask “why” (because if you understand the fundamentals, you can solve
anything, and it makes getting a date a lot easier). This is why is so much FUN
to do the MENTAL exercise of taking things apart and analyzing them. As you
do these exercises, you train your brain, and you will more readily notice
details and impress your friends (and make a better living!).

Wheels: Steering &
Suspensions
•

•

How to transmit power
to the wheels & enable
them to move with
respect to the vehicle to
absorb shocks, move
across uneven surfaces?
How to transmit power
to the wheels and enable
them to steer?
How does a tractor’s
suspension work ?
–

Do you really need 4
wheel independent
suspension?

5-19

1/1/2008

Wheels: Clutches & Differentials1
Clutches and differentials are very important transmission components whose operation is normally transparent to the user until they are needed.
A clutch normally behaves as a rigid element until a certain torque is exceeded,
and then it slips. A standard or “open” differential equalizes the torque on two
output shafts while constraining the sum of their rotary displacements to be
equal to the input rotary displacement.
In the case of an automotive clutch, the clutch allows slip between the
engine and the transmission so that the engine will not stall at low speeds. In
the case of a robot or manufacturing equipment, a clutch is typically used to
prevent overloading of components. For example, a clutch can prevent gear
teeth from being sheared off when a driven member encounters a rigid object.
Clutches can also be configured to resist torque in only one direction. A simple form of this type of clutch can be made by pressing a spring over the ends
of two shafts, and anchoring it only to the driving shaft. For torque applied in
one direction, the spring winds tighter and transmits torque. For torque applied
in the opposite direction, the spring unwinds and torque is not transmitted. The
kinematics of a clutch are transparent when it is engaged, transmitting torque
without slip, and it functions as a revolute joint when it is disengaged. The
dynamics of a clutch are more complex, because they depend on the friction
coefficient; however, because of the requirement for high friction to enable the
clutch to transmit high torque, they act as either rigid elements or very highly
damped elements.
The most common form of clutch is a simple mechanical clutch where
one plate is forced against a second plate through the use of a spring. By
adjusting a screw to change the pressure on the spring, the slip torque can be
adjusted. The design parameters affecting a clutch’s torque capacity are: the
diameter of the plates, the coefficient of friction, and the force between the
plates, A subtle design parameter, however, is the means by which the clutch
plate is mounted that enables it to move forward and engage its mate. If it
slides on a spline shaft, then there will be backlash (see page 3-23), and the
clutch is not suitable for use in devices where a high degree of precision is
required in either direction of motion. For precision applications, the clutch
plate should be mounted on a diaphragm.
1.

Special thanks to Greg bean for his most useful edits and comments!

The force applied to the clutch plate can be applied mechanically or
electromagnetically, and the clutch can either be normally engaged or normally
disengaged. In the former case, springs maintain pressure between the elements, and a linkage or electromagnetic field is used to release the pressure.
Once the clutch engages and power makes its way to the driveshaft, a
differential is used to divide the power between the wheels. A differential has
a single input shaft and two output shafts, and it has two functions. The first
function is to divide the input torque equally between two output shafts, so
each shaft in effect sees a torque source and can thus rotate at a different rate
than the other if needed. The classic use is in the rear end of a car which allows
the outside wheel to go faster than the inside wheel when the car is going
around a curve. However, when one wheel is on ice and the other wheel is on
pavement, the differential can only apply to the wheel on pavement the same
torque that is applied to the wheel on ice, not very much, since the wheel on ice
easily slips. The vehicle may thus get stuck, with the total torque on the output
shafts insufficient to move the vehicle. The second function is to act as the
final gear ratio between the input shaft and the output shafts. Differential gear
ratios of up to 4:1 are not uncommon.
Why not just drive one wheel and avoid the need for a differential? If
only one wheel in a vehicle were driven, then the maximum tractive force
would be less than half that if two wheels were driven with a differential. Only
driving one of the wheels would also cause the vehicle path to curve, because
the driving force is not applied through the vehicle’s center of friction. For a
robot contest vehicle, if a single motor has all the power that is required to
move the vehicle, and the mass distribution is such that the full torque of the
motor can be applied to the driven wheel without causing it to slip, then a differential is not needed. If the wheel is likely to slip, then a differential can be
used to enable the single motor to drive two wheels. Of course, it also depends
on what type of steering strategy you are using.
Can you make a clutch from elements in your kit to prevent critical
gears from being damaged? Is it safe to press a gear onto a shaft and rely on
the interference-fit to act as a clutch? What design parameters affect the interference-fit force? Can you machine flexural elements into the gear’s hub so as
to ensure proper preload between the gear and the hub to maintain a desired
slip torque? Can you make a differential with the gears you have, or do you
need to use a commercial differential?

Wheels: Clutches & Differentials
•

Clutches and differentials allow for differential velocities between rotating shafts
–

–

A classic example is when a car makes a turn, all the wheels have the same instant center, but the radius from
each wheel to the instant center is different
• Thus each wheel travels a different distance along an arc in the same time, and a differential is needed
to enable the speed difference to occur
– However, if one wheel is on ice, and another wheel is on dry pavement, the wheel on ice will spin
• Limited slip differentials typically use spring-loaded clutch plates to ensure some torque is always
transmitted to both wheels
• Other designs use a centrifugal force to engage locking features
• New designs use electromagnetic clutches to optimize torque transfer to the wheels
Clutches can be activated mechanically or electrically to control the torque transmitted between shafts, while
allowing differential velocity to occur (see page 6-17)
Electric clutch allows limited slip across the differential, or the
Spur and bevel gear sets
mated together and free
to spin on input shaft

Input

Input
shaft

Oldham
coupling

differential to lock forcing both shafts to be driven. See:
http://automotive.eaton.com/product/traction_stability/intellitraxx.html

Spur gear attached to
shaft with squeeze collar

Flange to which input
gear is attached

Output
shaft 1

Output 1

Right output shaft
slides in and
engages spline on
bevel gear

Electrically
activated
clutch

Output 2
Output
shaft 2
Support bearings

See page 6-19 for a differential a
© 2008
Alexander
Slocum
student
built
for his robot

Bevel gear mounted on
bearings on shaft attached
perpendicular to input shaft

5-20

Left output shaft
slides in and
engages spline on
bevel gear

Internally
splined bevel
gears (1 of 2)

1/1/2008

Cams1
A cam is a machine element whose specially shaped lobes are followed by a cam follower, which causes the cam profile to be imparted on
another object. A common application is in an internal combustion engine
where the cam is driven by a chain or belt connected to the crankshaft. The
cam lobes are thus synchronized with the crankshaft rotation to open and close
intake and exhaust valves as required. The shape of the cam lobe determines
not only when a valve is opened, but how fast, and how long it stays open,
which is called the dwell time. In a modern overhead cam engine, the cam
pushes on a cam follower, which pushes on the valve stem. The valve is held in
a normally closed position by valve springs, which, through the chain of elements cause the cam follower to remain in contact with the cam and keep the
valve normally closed. This may seem like an archaic system ripe for replacement with modern electromagnetic actuators; however, for high reliability, low
cost, and compact size in a hostile environment, cams still rule.2
When the cam profile exists on the surface of a cylinder, it is called a
barrel cam, and two degrees of freedom can be defined. Barrel cams control
screw machines, which are multi spindle lathes that make small round parts.
Each spindle holds the same type part between stations. At each station, a dedicated tool performs one operation on the part. Barrel cams typically have several tracks for controlling several tool motions. Screw machines make all the
zillions of small round parts of which some are used in virtually everything. In
fact, one of the common products made cheaply on screw machines are
screws! Most new screw machines are computer controlled, but countless old
ones keep on camming!

The system can be made to translate and rotate (move and dump) using a single
rotary motor and a barrel cam.
Cams can also be used in situations where they are stationary and the
follower moves along their contour. An early example, which helped give
birth to precision machine tools which in turn made our modern world possible, is the corrector cam. To obtain linear motion, machine tools use leadscrews which convert rotary to linear motion, as discussed in detail in Chapter
6. The lead on the screw, can only be made so accurate. A corrector cam
placed alongside the leadscrew with a cam follower connected to an arm projecting out from the nut will cause the nut to rotate forward or backward relative to the leadscrew in order to compensate for small lead errors. Today,
repeatable errors are mapped and the error correction is done digitally.
There are many types of cam followers. Radial followers are used
with the heaviest loads to convert rotary motion into linear motion, but the
range of motion of the follower is modest. Offset followers enable greater
motion of the follower, but require greater torque. Swinging followers convert
cam rotation to oscillating rotary motion. A roller tip minimizes friction, but
the roller has less load capacity than a flat follower. A compromise is the
spherical follower. All these followers, however, are limited in how rapidly the
lobe can change shape. When extremes in lobe shape occur, a knife edge follower may be used, but the load capacity will be more limited. Good lubrication is key, and Hertz contact stresses must be controlled (see Chapter 9).

Another example of a barrel cam is the YankeeTM screwdriver which
converts downward motion of a handle held by the user into rotary motion of a
screwdriver tip. In a robot for a design contest, for example, a leadscrew nut
may be prevented from rotating by trunnions (ears) that stick out its side and
slide in a axial grooves in a follow barrel cam (the nut moves inside the cam).
As the screw shaft rotates, the nut will translate until the axial groove transitions to a circumferential groove and instead of translating, the nut now rotates.

Designing a cam is a straightforward inverse problem. Given the
desired output motion as a function of input angle, the radius of the lobe at the
point of contact between the cam follower and the lobe can be determined.
One must not only evaluate the kinematics of motion, but also the dynamics of
motion including accelerations of the follower to determine the spring force
necessary to ensure that it remains in contact with the cam. If the spring force
is too high, the system will wear out too soon, and if it is too low, the follower
will lose contact with the cam surface. Fortunately, just as with linkage design
programs, there are a plethora of cam design programs, many of them available
for free on the internet!

1.
Cams might have been discussed in the context of linkages, but since they are often wheel-like and
directly connected to power sources, it is equally valid to discuss them here.
2.
M. Fischetti, “Why Not a 40-MPG SUV”, Technology Review, November, 2002 PP 40-46

Where might a cam enable your robot to achieve non-simple motion,
but without the space requirements of a linkage? Can a barrel cam be used to
enable one actuator to perform a complex motion?

Cams
•

–
–

•
•
Special thanks to Scott Hover from
Len Industries in Leslie Michigan for
the sketch, pictures, and explanation!

Rolling elements provide the highest degree of efficiency, but they take up
more space
Sliding elements are very compact, and can be efficient of the speed is high
enough to maintain oil film lubrication

A cam follower is a rolling or sliding contact element that follows the
contour of a surface and transmits the motion to a mechanism
A cam profile can be designed to create corresponding acceleration,
velocity, position, and dwell profiles in a mechanism (e.g., an engine
valve)
–
–

Stanley Tool Works’ Yankee™ screwdriver

A cam is a rotating shape whose angular motion is converted into output
motion by a cam follower which rides on the cam surface

Cam design software can create virtually any type
of cam and provide all required engineering data
See for example http://www.camnetics.com/

5-21

1/1/2008

Shafts
Shafts may seem like relatively benign machine elements, but they
are often subject to large cyclic bending and torsional loads. There are also
many different components which might be attached to them in a myriad of
different ways. In addition, to reduce bearing friction losses, it is desirable to
minimize shaft diameter. All these facts combine to make shaft design one of
the more challenging aspects of machine design. The first step in shaft design
is to assess the bending and torsional loads on the shaft. The bending loads can
be determined from a free-body diagram of the proposed system, taking care to
note where the shaft is supported in bearings and weather the support points act
as simply supported constraints, or moment supporting constraints. A bending
moment diagram can then be made to predict the bending moment as a function of position along the shaft. The torsional load Γ (N-m) can either be
known by the machine specifications directly or determined from the power P
(Watts) and speed ω (radians/second):

Γ=

Given an understanding of the bending and torsional loads on a shaft,
the next step is to calculate the stresses and deflections in the shaft, especially
given the presence of large stress concentration factors that will exist wherever
components are attached to the shaft. In addition, wherever the shaft receives
moment support from a set of bearings, the interface between the shaft and the
bearing acts as a discontinuity and the appropriate stress concentration factor
should be applied. The basic formulas for determining bending, torsional, and
shear (from radial loads) stresses in shafts are well known:
•
•
•

Bending: σ = Mc/I where I/c = πD3/32
Torsion: τ = Γr/I where I/r = πD3/16; φ = ΓL/GI where I = πD4/32
Shear: τ = F/A where A = πD2/4

Because shafts typically rotate, they can experience very high cyclic
loading, and thus they are especially prone to fatigue. In general, engineers are
taught in their basic mechanics course that steel alloys have an infinite fatigue
life if the stress is kept below one-half of the yield stress. However, this is just

a rule of thumb for initial layout of a design. When it is time to select the
actual alloy and its heat treatment, one must be very careful to check the endurance limit specifications for the chosen material. Different alloys behave differently at the upper limits of their strengths. Furthermore, what may seem
like a properly designed shaft may fail in fatigue due to an overlooked stress
concentration, which is why the next three pages are devoted to highlighting
some of the more common loading and geometry configurations. Typically a
design engineer will use a spreadsheet or program they create to do a parametric study of how the shaft stress will vary as a function of varying geometric
parameters and loads, and then check the final design using finite element analysis.
Its not just the stress that is important in shaft design, deflections of
the shaft can cause bearings to overload or seals to fail. If a bearing is designed
to primarily withstand radial loads, and the shaft which it supports is too long,
too large a moment may be created. Unless the bearings can accommodate the
resulting angular deflection, the bearing may fail prematurely. Roller and needle roller bearings are particularly sensitive to angular deflections. If a sliding
contact bearing (bushing) is used, clearance must exist between the shaft and
the bushing, excessive bending of the shaft may cause the clearance to close
causing the bearing to seize.
When shafts are coupled to other elements, such as a leadscrew driving a nut attached to a carriage, there can never be perfect alignment between
the shaft and the other element. As the system moves, there has to be sufficient
compliance somewhere in the structural loop to enable the misalignment to be
accommodated without overloading the system. Typically the misalignment is
accommodated either by clearance between members, compliance in the shaft,
or with a flexible coupling. In all cases, however, one must determine the
accuracy and repeatability requirements of the system, and make sure the
method used provides adequate performance.
Shaft alignment in design contest robots is one of the most overlooked aspects of their design. Carefully review your design and see how features can be machined in place or made easily adjustable so that shafts can be
easily aligned. The next biggest problem area is shafts that deflect too much
and cause components to bind. Do a careful review of shaft loading and component clearances to guard against this failure mode.

Shafts
•

Shafts transmit rotary and linear power via motors,
leadscrews, pistons…
– Shafts are one of the most common machine elements

•

The primary design issues are:
– Loading conditions and stress concentrations
– Mounting
– Component Attachment
Shafts

Carriage
Rotary Encoder

AC Brushless Motor
Flexible Coupling
Bearing Housing

pneumatic or hydraulic cylinder

Ballscrew

Ballnut

Support Bearings

Leadscrew driven carriage

5-22

1/1/2008

Shafts: Axial Loading
Axial loading of shafts may be tensile or compressive, and the stress
concentration factors shown in the table apply in either case. When loads are
compressive, one must also make sure that the shaft does not suffer from a
buckling instability, as discussed on page 6-6. Pneumatic and hydraulic cylinders’ shafts often are subject to very high axial loads, and in addition, they
often have end attachments that can be subject to large stress concentrations.
Their shafts typically undergo higher compressive than axial loads, but they
may also be subject to substantial tensile loads when double acting cylinders
are used. In this case, since the loads alternate between tension and compression, fatigue cracks are more likely to grow than if the system were just subject
to compressive loads.
Leadscrew shafts also typically experience very high alternating axial
tension and compression loads. Leadscrews are also often placed in tension so
that as the nut travels and heats up the shaft, thermal strains are replace the tensile preload strain and the lead (mm/rev) remains constant, as discussed in
Chapter 6. Leadscrews usually have very simple end connections, so the primary place for fatigue is in the root of the threads. The stress concentration
factor can be estimated using the third case illustrated in the table where the
thread root diameter would be r.
Case 1 shows a typical situation where a shoulder in a shaft transfers
axial loads from the shaft to the inner ring of a ball bearing. The inner ring of
the ball bearing will be chamfered, which allows for a small radius to relieve
the stress concentration. Note if the radius is not made correctly, e.g., it is
made too big, then the bearing would not seat against the shoulder and the
design would fail. For greater stress relief or more manufacturing robustness,
Case 2 shows how the shaft itself can be grooved to yield in effect a sharp corner against which the bearing inner ring can seat., However, if the groove is
made too large, it reduces the shaft diameter and the stress is too large. In
deciding which design to use, the designer can use the equations for both cases
to select the best design for their application. The spreadsheet Shaft_axial.xls
compares these two cases. Sample output is shown below. Within the spreadsheet you can also parametrically explore the trends that occur as the corner or
undercut radius is varied. When is it better to use a radiused corner and when
is it better to use an undercut?

shaft_axial.xls Radius transition or undercut between diameters?
Large diameter, D
14
Small diameter, d
8
Axial load (N)
20
M ajor shaft diameter (mm)
14
transition radius (mm)
0.5
Stress concentration factor fillet shaft, Kt
2.39
Stress concentration factor undercut shaft, Kt
2.14
Shaft length (mm)
50
Shaft modulus, E (GPa)
2.00E+11
Area at fillet shaft diameter (mm^2)
50
Area at fillet shaft diameter (mm^2)
38
Fillet design stress, sigma (M Pa)
0.950
Undercut design stress, sigma (M Pa)
1.111
Shaft deflection along minor diameter (micrometers)
0.099

Case 3 and Case 4 show two alternate low cost methods of affixing a
component to a shaft using a pin through the pulley hub and the shaft, or a key
in a keyway cut into the shaft and the pulley inside radius respectively. The
former is fast and simple, but results in large stress concentrations. As can be
seen on the following pages, this greatly reduces the torsional and bending
strength of the shaft. A pin through a shaft should only be used for modest
loads or where there are no bending or torsional loads from the hole onward.
This would be the case for a cantilevered shaft with a pulley or a wheel on its
end: the pin should go on the outside towards the end of the cantilevered shaft.
A variation on the radiused corner uses a narrow straight mini-flange
near the base of the radiused corner. This small section is large enough to provide a square shoulder against which a component can be axially located. The
comparison table shows it is too small to create a large stress concentration.
What shafts in your design might be overstressed or become fatigued?
What are the stress concentration implications for bolt design? Are you using
a long small diameter leadscrew (e.g., a piece of threaded rod) that may buckle
under high leads? Do you have linkages with thin round shafts as members?

Shafts: Axial Loads

Stress

Shaft loaded by
axial force F
Case 1

σ=

Case 2

Case 3

r
d

Case 4

Corner r
Width b
Depth t

πd2

4F
σ = Kt
2
π ( d − 2r )

σ≈

K t 4F

K tF
0.25π D 2 − dD

K tF
σ≈
0.25π D 2 − bt
5-23

See Shaft_axial.xls

Stress concentration factor

−0.36 − 0.2( D / d )

(r / d )
K t = 1+
5 + 0.12 /( D / d − 1)
−0.511− 0.34( d /( d − 2 r ) −1)

(r / d )
K t = 1+
0.42
3 + 0.507 / (d / d − 2r − 1)
K t ≈ 1 + 0.65 / ( d / D )

0.275

K t ≈ 1.5
B=D/4, t=D/8
1/1/2008

Shafts: Torsional Loading
Torsional loading of shafts usually results in the highest number of
cycles experienced by just about any machine element, except the individual
balls in ball bearings. In addition, thin long shafts can also suffer from buckling instability, but this rarely occurs. Torque transmitting shafts are commonly found in motors, transmissions, and machine tool spindles. In addition,
leadscrew shafts and bolts (screws) also are subject to large torques. Complicating matters are the various machine elements, such as gears, pulleys, and
attached loads, which must be attached with sufficiently robust means to transmit the torque; however, the features used to transmit the torque, such as pin
holes and keyways, can significantly weaken the shaft.
shaft_torsion.xls Radius transition or undercut between diameters?
Large diameter, D
14
Small diameter, d
8
M otor Power (W)
20
M otor speed (rpm, rad/sec)
100
10.47
Shaft torque (N-m)
1.910
M ajor shaft diameter
14
transition radius (mm)
0.5
Fillet stress concentration factor, Kt
1.45
Undercut stress concentration factor, Kt
1.52
Shaft length (mm)
50
Shaft modulus, E (GPa)
2.00E+11
Shaft Poisson ratio
0.29
Shear modulus, G (GPa)
7.75E+10
Polar I at fillet shaft diameter (mm^4)
402
Polar I at undercut shaft diameter (mm^4)
236
Fillet design torsion stress (M Pa)
27.6
Undercut design torsion stress (M Pa)
43.0
Shaft twist along minor diameter (deg, rad)
0.18 3.06E-03

Whenever a machine element is attached to a torque transmitting
shaft, at the attachment point, force must “flow” from the shaft and into the
element so it can be transmitted by the element (e.g., a gear). Thus if the interface between the shaft and the element is sharp and has a discontinuity, a stress
concentration will arise. Given that elements attached to shafts also often have

to be precisely axially aligned, it is common for the element to be registered
against a shoulder. This gives rise to the mounting conditions of Case 1 and
Case 2 as shown in the output table shown from the spreadsheet
Shaft_torsion.xls.
The bore through the machine element, such as a gear, will be chamfered, and this allows for a small radiused corner on the shaft to relieve the
stress concentration. If the radius is too big, then the element will not seat
against the shoulder and the design will fail. If a greater relief or more manufacturing robustness is required, Case 2 shows how the shaft itself can be
grooved to provide a sharp corner against which the machine element can seat.
However, if the groove is made too large, it reduces the shaft diameter and the
stress is too large. In deciding which design to use, the designer can use the
equations for the two cases to select the best design for their application. The
spreadsheet Shaft_torsion.xls compares these two cases. Sample output is
shown below, but within the spreadsheet you can also parametrically explore
the trends that occur as the corner or undercut radius is varied. When is it better to use a radiused corner and when is it better to use an undercut?
Once again, Case 3 and Case 4 show two alternate low cost methods
of affixing a component, say a pulley, to a shaft using a pin through the pulley
hub and the shaft, or a key in a keyway cut into the shaft and the pulley inside
radius respectively. The former is fast and simple to do, but results in large
stress concentrations which greatly reduces the torsional and bending strength
of the shaft. Thus it should only be used for modest loads or where there are no
torsional or bending loads from the hole onward. For the case of a cantilevered
shaft with a pulley or a wheel on its end, the pin should go towards the end of
the cantilevered shaft. This prevents bending loads from passing around the
stress concentrating pin hole. The best, but most expensive, way to attach a
torque transmitting element to a shaft is with a press or shrink fit, or a squeezetype interface.
You undoubtedly have torque transmitting shafts to which you intend
to attach torque transmitting elements such as pulleys or wheels. Run the
spreadsheet and calculate the shear stress in the shaft. More than one student
has been shocked to find that the torque output from a high reduction gearbox
can shear their shaft. Many times the break occurs at a drilled hole for a pin!

See Shaft_torsion.xls

Shafts: Torsional loads

Stress

Shaft loaded by
Torque Γ
Case 1

τ=

Case 2

D
r

Case 3

τ=

Case 4

Corner r
Width b
Depth t

π d3

16Γ
τ = Kt
3
π ( d − 2r )

d
d

K t16Γ

K tΓ
π D 3 /16 − d D 2 / 6
16Γ
τ = Kt 3
πD
5-24

Stress concentration factor

−0.3− 0.2( D / d )

(r / d )
K t = 1+
13 + 0.3 / ( D / d − 1)
−0.609 − 0.146( d /( d − 2 r ) −1)

(r / d )
K t = 1+
0.252
5 + 3.73 / (d /(d − 2r ) − 1)
K t = 1 + 1.47( d / D )

−0.197

K t ≈ 1.5
B=D/4, t=D/8

1/1/2008

Shafts: Bending
When wheels are attached to a shaft to support a vehicle, it is clear
that there will be radial loads on the shaft that will cause bending stresses and
deformations. In addition, when pulleys are attached to a shaft, it is clear that
the cable forces will also radially load the shaft and cause bending stresses and
deformations. Furthermore, when gears are attached to a shaft, there can be
significant radial loads exerted by the teeth, and these radial loads also cause
bending stresses and deformations. A quick rough estimate of the radial load
on a shaft caused by a gear of pitch radius Rpitch (m) transmitting power P
(Watts) and speed ω (rpm) is:
F radial =

30 P

πω R pitch

The stresses caused by radial loading induced bending moments can
be significant, although resulting shaft deflections are generally in a non-sensitive direction with respect to maintaining proper center distance between gears.
However, radial loads caused by gear teeth are also compounded by the pressure angle θ (degrees), or the angle at which the teeth engage, which tends to
force the gear teeth apart. For example, for simple spur gears, this radial force,
which is not significant with respect to stresses, can cause significant shaft
deflection with respect to maintaining gear center distances, is:

F radial _ pressure _ angle _ force = F radial sin θ
As before, starting with Case 1 and Case 2 type loading, where a radiused corner or an undercut is used to register the axial position of a machine
element such as a wheel, gear, or pulley, the stress concentration factor is
greatly affected by the radius of the corner or the undercut. In both cases,
increasing the radius decreases the stress concentration, but in the case of the
radiused corner, it increases the likelihood that the machine element will not
seat completely against the shoulder. In the case of an undercut, the larger the
radius, the smaller the remaining diameter is directly beneath the undercut, and
it is at this minimum diameter point that the stress needs to be determined. The
spreadsheet shaft_bending.xls can be used to explore different shaft designs
and trade-offs:

shaft_bending.xls Radius transition or undercut between diameters?
Large diameter, D
14
Small diameter, d
8
Bending moment (N-m)
10
M ajor shaft diameter
14
transition radius (mm)
0.5
Fillet stress concentration factor, Kt
2.89
Undercut stress concentration factor, Kt
1.99
Inertia I at fillet shaft diameter (mm^4)
201
Inertia I at undercut shaft diameter (mm^4)
118
Fillet design bending stress (M Pa)
575
Undercut design bending stress (M Pa)
592

The effect of a hole in a shaft for pinning a component to the shaft, or
of a key for enabling a shaft to transmit torque to a component once again also
has an important effect both in terms of the stress concentration, and the fact
that these features reduce the cross sectional area of the shaft.
Since the bending loads may be the result of loads applied to the
periphery of the machine element itself, the design engineer must also consider
the effect of the loads on the component itself, and whether or not these loads
are aligned radially, tangentially, or axially with respect to the shaft. Can these
loads cause deformations in the element that is attached to the shaft that may
weaken the attachment method? What is the level of shaft deformation, and
can the resulting radial and slope deformations cause the shaft to overload
bearings or reduce clearances to the point where the shaft rubs a bearing or a
seal and causes it to fail? Can deflections cause a gap to open between a seal
and a shaft and thereby cause the seal to leak?
Many a student has designed a robot for a contest with gears that
work fine when tested without significant loads. However, they sadly find that
the teeth skip and strip under heavy loads. The cause is usually structure or
shaft deflection caused by gear separation forces that increase the gear center
distance too much. Check your shaft stresses and deflections! Check the stiffness of your supporting structure!

See Shaft_bending.xls

Shafts: Bending Loads

Stress

Shaft loaded by
Bending moment M
Case 1

32 M
σ = Kt 3
πd

Case 2

32M
σ = Kt
3
π ( d − 2r )

Case 3

r, d
d

σ=

Case 4

Corner r
Width b
Depth t

K tΓ
π D 3 / 32 − d D 2 / 6
32 M
σ = Kt 3
πD
5-25

Stress concentration factor

−0.73− 0.42( D / d −1)

(r / d )
K t = 1+
0.16
5 + 4.38 / ( D / d − 1)
−0.59 − 0.184( d /( d − 2 r ) −1)

(r / d )
K t = 1+
5 + 0.081/(d /(d − 2r ) − 1)
K t ≈ 1 + 0.65 / ( d / D )

0.275

−0.66

(r / d )
K t = 1+
11.14
B=D/4, t=D/8

1/1/2008

Shafts: Mounting & Stability
In addition to how components are mounted on shafts and the resulting stress concentrations that can occur, particular attention has to be paid to
mounting a shaft to a structure. The three loading cases, axial, torsion, and
bending, must be considered in conjunction with different possible connections
between a shaft and a structure. Whether the shaft is fixed, or moving with
respect to the structure makes a difference in selecting the mounting method.
Shafts that do not move with respect to a structure are commonly used
as bearing rails on which linear bearings ride. They may also be fixed axles on
which components rotate, such as pulleys or wheels with integral bearings.
Fixed shafts can be held in place by a setscrew (bad), by clamping (squeezing)
the shaft which is placed in a hole (good), interference-fit (good), or held in
place in a hole with a threaded fastener (good). The method by which the shaft
is anchored must also take into account the loads placed on the shaft.
If axial loads are to be heavy, then it might be a good idea to either
pin the shaft in the hole, or to use a shaft with a shoulder so the axial load will
be transferred by contact between the shoulder and the structure. Opposite the
shoulder, the shaft would be threaded so a nut could resist tension in the shaft.
If the shaft is to resist high torsion or bending loads, then its end can be pressed
or shrunk into a hole, but then care must be taken to ensure that adequate interference is provided. An alternative method is to clamp the end of the shaft
using a split housing as is shown in the figure for two bearing rails. Recall
Saint-Venant, and seek to clamp the shaft by at least 3 shaft diameters if a true
moment connection is to truly be established.
If the shaft is to be free to move with respect to a structure, then the
bearings must be designed to accommodate bending deformations in the shaft
without losing any desired bearing clearances. For example, whenever two
shafts must operate in tandem either in the fixed or moving state, it is critical
that their horizontal and vertical parallelism be made as good as possible so as
to minimize bearing clearance that would otherwise be required to accommodate misalignment. When clearance exists in bearings, backlash can occur, and
when backlash occurs in a carriage moving on bearing rails, it can have pitch,
yaw, and roll errors, and these can be amplified by the Abbe effect to cause
large errors in your machine! Alignment can best be achieved by line-boring
mounting holes for the shafts, or their bearings, in the support plates at the

same time. In other words, clamp the two plates that will form the supports
together in the milling machine and bore the holes through both plates at the
same time. Using common reference features on the plates and boring the
holes at the same time ensures the holes will be as well-aligned as is possible.
For stability, the table shows formulas for determining static buckling
loads and shaft maximum rotation frequencies. The former is intuitively obvious to most people: push on a straw and it will buckle sideways. The latter is
less obvious, but if you rotate a shaft at its bending natural frequency, it will go
unstable and start to vibrate. This is called shaft whip, because the shaft can
literally whip about. This could damage the shaft, and most likely damage the
support bearings and any other components attached to the shaft. Note that
both are highly dependent on the mounting condition. Do not assume you have
a built-in moment resisting connection unless the shaft is held by at least 3
effective diameters. In addition, use the ROOT diameter of a leadscrew’s
thread for shaft calculations! What do you do if you are not sure if you have a
good moment supporting mount? You could run an experiment and measure
the shaft deflection under load to see if it matches the static equations.
Sometimes, as is the case with leadscrews, shafts are placed in tension, and this can increase their natural frequency significantly if the tensile
load applied is much greater than the buckling load1. In addition, some weight
limited machines, such as turbines, operate at super critical speeds, where the
machine will start up at a low frequency but then accelerate very quickly to its
operating frequency which is above the first natural frequency of the shaft! In
these cases, the damping from the support bearings plays a very important role
in helping the machine to make the transition. In robot design contests, shaft
buckling and rotation instability can be a problem because shafts are often
made from slender members. The spreadsheet shaft_stability.xls examines
various stability criteria for shafts.
Make a list of the shafts that are used, or those that may be needed, in
your machine and consider different strategies for mounting and aligning them.
What are the buckling and critical speed values for these shafts and are they an
issue at all for your design? How might you increase these values if needed?

1.
See for example Kenneth G. McDonald, Vibration Testing: Theory and Practice, 1995, John Wiley
and Sons, New York.

Shafts: Mounting & Stability
•
•

Abbe and Saint-Venant’s Principles and Exact Constraint Design must be
carefully considered when designing shaft mountings
Design the system to accommodate misalignment
–
–
–
–

Allow for clearance between bearing and shaft to accommodate misalignment
Use flexible couplings or self-aligning (spherical) bearings
Use designs that automatically accommodate misalignment
Line-bore holes at the same time through mounting plates

5-26

1/1/2008

Shafts: Component Attachment
The previous discussions focussed on stress concentrations between
components, their connections to shafts, and the conditions most likely to
result in a robust design. In general, avoid sharp transitions, even between
components mounted on the shaft and the shaft itself. When components are
attached to shafts, there are often many different types of loads applied, and in
order to account for all their effects on the total state of stress in the shaft,
(e.g., σaxial, τtorsional, σbending) use an equivalent stress criteria such as Von
Mises equivalent stress:

(σ x − σ y ) + ( σ y − σ z ) + ( σ z − σ x ) 2 + 3
2

equivelant
equivalent

τ 2xy + 3τ 2yz + 3τ 2zx

But what about practical aspects of designing such connections from
applications ranging from design contest robots that have to work for a couple
weeks to industrial robots that have to work for many years? In general, you
get what you pay for, and of course you should only pay for what you need.
Press or shrink fit connections are often most reliable for when they
are well designed, the materials behave as if they were monolithic. The biggest issue is manufacturing tolerances, which are often harder to control than
the required interference fit. The spreadsheet Joint_interference_fit.xls illustrates this quandary, and it is discussed in more detail on the following page.
Note that for high torque and high precision production systems, a design engineer may want to select mating tapered polygons1.
The next option is to use a component that has flexures cut in it and a
collar that compresses the flexural element to clamp the shaft. In the case of
using a plastic tube as a coupling between two shafts, simple hose clamps
achieve this function. A component may also be designed as shown where the
end is split and a bolt is used to squeeze the two segments together.
Many components such as gears, however, have a hub but no means
to provide a clamped-type connection, other than a shrink-fit. A spline may be
used to transfer the torque. The simplest spline-type connection to use, espe1.

See for example http://www.generalpolygon.com/index.htm

cially for a robot contest, is a hexagonal shaft that fits through a hexagonal hole
broached into the component. Other parts of the shaft may be made round on a
lathe so they can pass through bearings.
If a spline is not practical, a keyway is often the next best type of connection, and indeed it is one of the most common methods used to transmit
torque between shafts and components in industrial applications. Shown are
square cross-section keys. The components are often held axially in position
using setscrews that push against the keys. However, alternating torque loads
can loosen the setscrew by causing fatigue between its tip and the key so thread
locking compound should be used. The square keyway, however, has to be
milled and it also causes a fairly large stress concentration. A woodruff key is
a semicircular key, where the slot cut in the shaft is then cut with a simple
plunging rotary cutter2. For any key, the torque transmission capability is
equal to the product of the key cross sectional area in shear, the radius of the
shaft, and the shear strength of the key material.
After a keyway, a pinned connection is probably the most desirable.
To keep the pin from falling out, a spring pin is typically used: spring pins are
hollow cylinders with one edge slit. The torque transmission capability is then
equal to the product of the cross sectional area of the pin in shear (both sides of
the pin), the shaft radius, and the shear strength of the pin material. In order to
ensure that the hole through the component lines up with the hole in the shaft,
the component needs to be placed on the shaft and then the hole drilled through
component and shaft at the same time. Remember, the pin needs to be on a
free-end of the shaft so the shaft does not see bending loads or it is likely to
fail.
Your worst connection nightmare is just a setscrew pushing on a shaft
(either directly or on a flat on the shaft). Such a connection is virtually guaranteed to eventually work itself free, even if you use thread locking compound.
So beware, because it is amazing how many products still use this type of connection! For your robot design contest machines, which only have to work for
a period of a couple weeks, you might get away with it if you use the point of a
drill to form a conical depression in the shaft into which the setscrew seats;
thus acting kind of like a small keyway.
2.
For a good discussion on types of keyways and how to machine them, see Machinery’s Handbook,
Industrial Press, New York

Shafts: Component Attachment
•

Principal functional requirements for attaching a component to a shaft:
–

Prevent the component from slipping
• Interference-fit: Best torque transmission, but can be difficult to manufacture
– A variation is when the shaft and hub are ground to have a mating lobed fit
• Spline: Excellent torque transmission, but can be expensive
• Circumferential clamp: Very good torque transmission, low stress concentration, modest
price
• Key: Very good torque transmission, modest stress concentration, low cost
• Pinned: Good torque transmission, modest stress concentration, low cost
• Setscrew: Very low cost, but could be your worst nightmare if it loosens!
– After the first round of competition in 2007 2.007 contest: “I was wondering if I
could use a drill press in lab sometime tomorrow to drill a hole in a gear and
shaft for a pin. I realized after tonight that my set screw on my shaft still slips
even after making a flat on the shaft”

–

Minimize stress concentrations
• Use radiused corners, but be careful of having a square surface for locating components
– For larger differences in diameters, a radiused undercut can be used
– For smaller differences in diameters, a small straight flange in front of a radiused corner
can be used

5-27

1/1/2008

Shafts: Interference-Fits

strings of cell references! With reasonable descriptors, a well-written spreadsheet should need little or no additional documentation.

Properly implemented, an interference-fit connection between components is perhaps the best connection means. This is due to the fact that there
is no mechanism to loosen with time; however, because there is no mechanism,
there is no means for adjustment, and hence care in manufacturing is critical.
This is especially true given that the amount of interference required to achieve
a desired interface pressure is very small, so the amount of interference
required to yield the material is also small.

Torsional and axial loads are very straightforward to model, where for
a diameter D, the stress caused by the torque G and force F are easily determined. In addition, however, the poisson contraction (or expansion) in the
radial direction from the axial load is also determined:

When designing a connection between objects, one must be careful to
consider all the forms of loading that may occur, and then find or develop simple analytical models for each. Superposition allows them to be combined
such that an estimate can be obtained for the total state of stress in the joint.
For a interference-fit connection between a gear and a shaft, loads that must be
accounted for include: torsional, axial (some forms of gears generate axial
loads), centrifugal expansion, and differential thermal expansion. In addition,
bending loads may also be present, but it is the former that most affect the joint
structure.
In order to create a spreadsheet or a program to determine the stresses,
the independent and dependant variables must be identified. In an ideal case,
all the loads are entered and the program directly or iteratively solves for the
desired parameter, such as the required (interference-fit) interference. The
analysis can also proceed from a manufacturing perspective where reasonable
dimensions are entered and the results are reviewed. In the case of a interference-fit, the latter is chosen here and is indeed a more practical method. To
start, the spreadsheet on the opposite pages has sections for entering in the
known and assumed parameters for the design. The inner and outer diameters
of the bodies are entered along with manufacturing tolerances.
The first calculation determines the minimum interface pressure
required to carry the loads so the interface does not slip. Next, the pressure
caused by the interferences must be determined. Finally, the stresses and radial
displacements caused by the various loading conditions must be determined.
All of these different calculations are performed on the spreadsheet, where
variable names are assigned to the spreadsheet cells, so the equations entered
appear as decipherable mathematical expressions and not as incomprehensible

τ=

32r Γ

πD

σz=

πD

ur =

rησ z
E

Given an interference Δ, the outer body modulus of elasticity EO and
outer diameter DO, the inner body modulus of elasticity EI and inner diameter
DI, and the nominal interface diameter Dint, the resulting interface pressure is1:
P=

Δ
2
⎞ 1 ⎛ D I2 + D int
⎞ ⎪⎫
⎪⎧ 1 ⎛ D + D
+
+
−η I ⎟⎬
η
D int ⎨
⎜
⎜
O⎟
2
2
⎪⎩ E O ⎝ D − D
⎠ E I ⎝ D int − D I
⎠ ⎪⎭
2
int
2
O

2
O
2
int

Once the amount of interference is determined, it can be determined if
the component can be heated to expand and then placed over the shaft. If a reasonable temperature can be used, then a shrink-fit is most desirable. Otherwise
a press-fit can be used where a shaft is pressed into the hole, but with care to
prevent the shaft from buckling.
It only takes a little bit of interference to create a strong joint, but it
only takes a little bit of tolerance error to make it not work. Depending on the
materials, they have to have significant ductility else they will crack when
forced together. The interference-fit can sometimes be made much larger to
ensure yielding of the materials, but make sure that the interface pressure at
yield is enough to transfer the desired loads. Look at the critical outputs highlighted in yellow on the spreadsheet Joint_interference_fit.xls as they help to
indicate if the design is good.

1.
From Slocum A.H., Precision Machine Design, Prentice Hall, Englewood Cliffs, New Jersey 1992,
pp 387-399, or for For a more detailed discussion, see S. Timoshenko, Strength of Materials, Part II, Theory and Problems, Robert Krieger Publishing Co., Melbourne, FL.

Ideal case
Diametrical Interference =
(obid-ibod)

•

Inner body outside
diameter

Outer body
inside diameter

Worst case for loose fit:
Joint may not be able to transmit
desired force or torque

•

Diametrical Interference =
[(obid+obptol-(ibod-ibmtol)]
Inner body outside
diameter

Outer body
inside diameter

Shafts: Interference-Fits
Interference-fits: most reliable connection:
– Joint_interference_fit.xls models
tolerances, differential thermal expansion,
and high speed rotation

Manufacturing issues are of greatest
concern
– Interference is typically small to prevent
too high stresses:
• Radial or circumferential stresses
• Buckling while pressing in the
component
– Tolerances may cause too high stresses, or
too low an interface pressure
– Clean, high friction interface

Worst case for tight fit:
Yield stresses may be exceeded and
outer body may rupture

Required interference parameters
Minimum required interface pressure, rPI (N/mm^2)
Add Poisson diametral interference, ddp (mm)
Diametrical Interference =
Add differential thermal expansion diameter, ted (mm)
[(obid-obmtol-(ibod+ibptol)]
Add outer body rotating radial displacement, robd (mm)
Subtract inner body rotating radial displacement, ribd (mm)
Inner body outside
Total additional interference amount to be added to ibod, addi (mm)
diameter
Interference fit calculations (assumes addi has been added to ibod)
maximum interference, maxdelta (mm)
Maximum resulting interface pressure, maxIP (N/mm^2)
minimum interference, mindelta (mm)
Minimum resulting interface pressure, minIP (N/mm^2)
Minimum safety margin (min obtained pressure/required pressure)
Outer body
inside diameter

5-28

0.8017
6.92E-06
7.50E-04
1.80E-06
4.17E-07
0.0008
0.0078
26.36
0.0038
12.77
15.9

Hexagonal shaft press fit into a plastic
wheel’s round bore for high torque
transmission
1/1/2008

Couplings
Couplings are required between rotary and linear actuators and driven
components because the actuators are intended to move in one degree of freedom (linear or rotary) but they can never be perfectly aligned. As a component
moves, it will not always be aligned with the actuator. The product of the net
difference in motion with the stiffness of the connection between the two systems yields the misalignment force on the actuator. If a rigid coupling is used,
the forces can be extremely high, and something, usually the bearings, will
have to give and they will soon fail. Furthermore, a significant portion of the
system’s power can be expended in the process. When a coupling is ideal, and
only restrains the intended motion with negligible effect (stiffness) in all other
axes, it is said to be a non-influencing coupling.
There are three strategies for coupling designs. The first is to leave
clearance between geometric features so the motion occurs across a sliding
bearing interface without causing large bearing loads. A linear example is a
clevis at the end of a hydraulic cylinder, where there is clearance between the
clevis pin and the bore into which it fits. A rotary example is an Oldham coupling which is two comprised of two cylinders, each with a slot cut in their
faces, each attached to the ends of two shafts that are facing each other. A
plastic disk with raised rectangular features on each side is placed between the
cylinders. Small motions between the shafts are accommodated by the plastic
keys in the slots. When more torque and misalignment capacity are needed
with a more linear response, four jaws on each cylinder face each other and
then an eight-armed plastic spider insert acts as the interface between the two.
The second strategy is to use a connection that is only stiff in the
desired direction. An example is a tube with circumferential slits cut in it (a
helical coupling). Merely using a piece of plastic tubing is not acceptable
because if it is short enough to provide enough torsional rigidity, it is often not
sufficiently radially compliant. Super precision instruments may go to extreme
measures and use taut wire couplings.
The third strategy is to use linkage elements (linear and or revolute
joints) to create a mechanism that transmits the appropriate force or torque. A
common example of this type of strategy is to use universal (Hooke’s) joints at
the ends of a drive shaft. Precision instruments often use wobble pins, which
are pins with conical ends that rest in a spherical cup to transmit forces.

It is important to identify the degrees of freedom that are to be constrained, and the accuracy that is to be maintained. Draw the actuator output
and the element to be moved. Attach the two with a coupling element that is
only stiff in the direction in which force is to be transmitted. For example,
when a leadscrew nut is attached to a carriage, since the leadscrew is not perfectly straight or radially stiff, it may act as its own coupling, allowing the nut
to be attached directly to the carriage. However, as the leadscrew turns, it will
impart an oscillating lateral force onto the carriage. If the nut were attached at
the end of the carriage, it would cause the carriage to pitch, which would cause
Abbe effect errors! On the other hand, if the nut is attached at near the carriage’s center of stiffness, then radial error motions of the screw will not cause
pitch errors.
With respect to rotary motion systems, it is important to note that simple pin-in-slot type couplings, such as Oldham and Spider couplings, and simple universal joints induce sinusoidal error between the actuator and driven
shaft. The maximum error εθ between the angular position of the input shaft
and the output shaft is just the eccentricity error between the shafts divided by
the coupling radius. For shafts that are angularly misaligned, the error εθ
between the angular position of the input shaft and the output shaft is about
equal to the product of the misalignment angle and the sine of the rotation
angle. At a maximum, the error is on the order of the shaft misalignment error.
Couplings that use elastic compliance between elements have negligible following error. Constant velocity universal joints are also available.
Finally, consider how the couplings are attached to the shafts. A circumferential clamping attachment is the best alternative (e.g., an Oldham coupling). Setscrews are to be avoided, but if they must be used, also use a
keyway. If a keyway cannot be used, use a thread locking adhesive and prayer.
Complete an inventory of all the interfaces between moving elements
in your machine and assess the coupling strategies you could use. What price
do you pay (materials, time) for what performance do you expect to get? Are
you sure your couplings are strong enough and will provide the desired accuracy or repeatability? Are you sure that they provide the degrees of freedom
needed to ensure that your actuators are not wasting valuable power deforming
elements that should not need to be, nor want to be, deformed!

Couplings

•

No two moving components can be perfectly aligned
–

•

To prevent over constraint from destroying components and robbing your system of power, use couplings

Identify the degrees of freedom desired, the accuracy (repeatability) needed, and the load capacity
and stiffness required in the direction of force or torque transmission
–
–

Select either a sliding, flexural, or rolling element coupling
See Page 10-21 for life calculations

http://www.sdp-si.com/Sdptech_lib.htm
http://www.heli-cal.com/

5-29

1/1/2008

Couplings: Cheap & Easy Example

inally centered? What about a wide small diameter rubber band or similar
elastic sleeve? What about bending of the pins?

Most of the couplings described above need to be purchased, but for a
robot design and build contest, this is often not possible; thus how can you
morph all the above into a low-cost effective coupling? Shafts in a simple
robot built for a contest are likely to be misaligned in 5 degrees of freedom:
hence, the coupling must allow for radial (X, Y), axial (Z), and pitch (εx) and
yaw (εy) misalignments. Before starting to sketch ideas, consider the boundary
conditions of the problem in terms of available materials and manufacturing
processes:

Bending of the pins? How can we model that? The forces on the pins
can be modelled by assuming the torque transmitted acts as a couple across the
diameter of the coupling body; thus the bending force on a pin is equal to the
torque divided by the diameter of the coupling body. The spreadsheet
Cheap_coupling_design.xls performs the calculations. Why is the torque that
can be transmitted independent of the coupling body diameter?

•
•

Materials: pins (straight or spring pins), and round stock
Manufacturing: drilling, milling, sawing, turning

Recall that there are two basic types of joints, linear and rotary, and
that the motions to occur are of one shaft relative to the other shaft. This
implies that some of the motions can be on one shaft, and some of the motions
can be on the other shaft. It is also worth noting in the thought process, that the
X and Y motions required are orthogonal to each other. Furthermore, the pitch
and yaw motions also occur about orthogonal axes. This if we develop a joint
that can allow linear motion and rotation about that axis, then we could just use
the same joint, rotated ninety degrees on the other shaft and we would have
taken care of four of the five degrees of freedom. This leaves the Z-direction,
which intuitively we can provide for with an axial slot in the coupling body.
Assuming we use a simple axial slot, how can we provide for motion
along the slot (for the Z motion) but also in and out of the slot? The answer
would seem to be a simple pin that is pressed into the shaft (or glued, just
something to keep it from falling out). If the slot is cut into a tube whose inner
diameter is larger than the shaft, then the radial clearance between the two is
the radial motion that will be accommodated by the coupling. Just make sure
to make the length of the pin equal to the outer diameter of the coupling plus
the diametrical clearance between the shaft and the coupling’s inside diameter.
As the sequence of images shows, voila! The coupling is born. However, what are the risks and appropriate countermeasures? Will the coupling
body flop down and rattle? How fast will it be turning? Can soft sponge rubber be used as a spacer between the shaft and the coupling body to keep it nom-

Inputs in BOLD, outputs in RED
Shaft diameter, ds (mm)
Coupling body outer diameter, OD (mm)
Coupling body outer diameter, ID (mm)
Pin diameter, dp (mm)
Cantilever length, L (mm)
Max. allowable stress, maxs (N/mm^2)
I/c, Ioc (m^3)
Bending moment, Mb (N-mm)
Resulting torque transfer (Mb from both sides) (N-mm)

4
8
6
1.5
2
200
0.33
66
133

What about the forces between the pin and the slot causing the pin to
dent the slot? This would be a Hertz contact pressure failure mode, and is discussed in greater detail in Chapter 9 (read forward!), but suffice to say that the
theory and a spreadsheet to predict the contact pressure are provided!
A molded plastic cheap and easy coupling should perform more than
adequately for most situations. If you determine that for high impact or high
torque applications you need a metal version, how would you manufacture it?

Couplings: Cheap & Easy Example
•

What about in a robot design contest where two shafts may be linearly misaligned
axially, vertically and horizontally, and angularly misaligned?

– How can you design a simple one-piece coupling to enable one shaft to transmit torque to the
other shaft?
• How strong must be the radial pins (coupling_low_cost.xls)
• Can the coupling be made from plastic tube to reduce shock?
• Would O-rings be useful to nominally center it?

Critical: Note the clearance between
the coupling bore and both shafts!

5-30

1/1/2008

Topic 5 Study Questions

Which suggested answers are correct (there may be more than one, or none)?
Can you suggest additional and/or better answers?
1.

Pulleys can be used to direct or amplify the force in a cable:
True
False
The amplification of force in a cable wrapped around pulleys is equal to
the number of strands of cable between the pulleys and the number is
always even:
True
False
A capstan is typically a fixed, or controlled rotation, body-of-revolution
which a cable wraps around
True
False
A capstan can be used in a band brake, where a band is anchored to a
structure, and then wraps around a shaft:
True
False
A cable wrapped around a capstan by θ radians with coefficient of friction
μ and being held with a force Fhold, can resist the pull of a cable with

10.

11.

12.

many times higher force Fpull = Fhold e μ θ

True
False
If a pulley’s outer diameter is 3-5 times larger than the shaft on which it is
supported, then a sliding contact bearing can provide 90+% efficiency:
True
False
If a belt or cable runs around a fixed shaft, then there is a lot of friction
between the belt and the shaft, and the efficiency is low:
True
False

13.

14.

15.

Belts & Cables can be used to convert rotary to linear motion, where the
force F in a belt with tension T on a pulley of diameter D that can be
generated by the torque Γ can be conservatively estimated by:
F = 2Γ/D for toothed belts
F = 2Γ/D for flat belts but is limited by the capstan effect
F = ΓD/2 for all belts
Tracks only help when there is loose media or a surface into which they
can dig and they do not help on smooth surfaces unless they are covered
with a dust layer
True
False
Tracked vehicles skid steer, and as a result there are large normal forces
on the tracks which try to pull them off the drive and idler sprockets
True
False
A central ridge formed by an idler roller in the middle of the treads can
help a tracked vehicle steer
True
False
Two pulleys’ axes of rotation can never be perfectly parallel, so a flat belt
will want to drift off (tracking) so pulleys must be crowned (rounded
profile) to keep a belt from walking off
True
False
Pulley tension must be maintained either by proper pulley center distance
and belt elasticity, or a mechanism to tension the pulleys or a pulley that
pushes sideways on the belt
True
False
Chains are a very robust means to transmit very large forces and torques
without worrying about slip between the chain and the sprocket:
True
False
Rotary-to-rotary motion is very common (look at your bicycle!) and
Rotary-to-linear motion can be obtained by laying a chain on a surface,

Source Exif Data:

File Type                       : PDF
File Type Extension             : pdf
MIME Type                       : application/pdf
PDF Version                     : 1.4
Linearized                      : No
Creator                         : pdftk 1.12 - www.pdftk.com
Producer                        : itext-paulo (lowagie.com)[JDK1.1] - build 132
Modify Date                     : 2008:01:16 00:47:15-05:00
Create Date                     : 2008:01:16 00:47:15-05:00
Page Count                      : 62

EXIF Metadata provided by EXIF.tools

FUNda MENTALs Topic 5

Navigation menu

Versions of this User Manual:

Views

Navigation