Darrin Willis Masters Thesis Corrected Final

DESIGN AND DEVELOPMENT OF A NOVEL
LIGHTWEIGHT LONG-REACH COMPOSITE

ROBOTIC ARM
by
Darrin Willis
Bachelor of Engineering, University of Ontario Institute of Technology, 2007
A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE
REQUIREMENTS FOR THE DEGREE OF
Masters of Applied Science
in
The Faculty of Engineering and Applied Science
Mechanical Engineering
Supervisor(s):
Scott Nokleby, Faculty of Engineering & Applied Science

Remon Pop-Iliev, Faculty of Engineering & Applied Science
Examining Board:
Ghaus Rizvi, Faculty of Engineering & Applied Science
External Examiner: Matthew Kaye, Faculty of Energy Systems & Nuclear Science
THE UNIVERSITY OF ONTARIO INSTITUTE OF TECHNOLOGY

June 2009
c
Darrin
Willis 2009
DESIGN AND DEVELOPMENT OF A NOVEL LIGHTWEIGHT

LONG-REACH COMPOSITE ROBOTIC ARM
Darrin Willis
Degree of Master of Applied Science, 2009
Faculty of Engineering and Applied Science
University of Ontario Institute of Technology
ABSTRACT
Metallic robotic arms, or manipulators, currently dominate automated industrial operations, but
due to their intrinsic weight, have limited usefulness for large-scale applications in terms of
precision, speed, and repeatability. This thesis focuses on exploring the feasibility of using
polymeric composite materials for the construction of long-reach robotic arms.
Different
manipulator layouts were investigated and an ideal design was selected for a robotic arm that has
a 5 [m] reach, 50 [kg] payload, and is intended to operate on large objects with complex
curvature.
The cross-sectional geometry of the links of the arm were analyzed for optimal stiffness- and
strength-to-weight ratios that are capable of preserving high precision and repeatability under
time-dependent external excitations. The results lead to a novel multi-segment link design and
method of production.
A proof-of-concept prototype of a two degrees-of-freedom (2-DOF) robotic arm with a reach of
1.75 [m] was developed. Both static and repeatability testing were performed for verification.
The results indicated that the prototype robot main-arm constructed of carbon fiber-epoxy
composite material provides good stiffness-to-weight and strength-to-weight ratios.
element analysis (FEA) was performed on a 3-D computer model of the arm.
Finite
Successful
verification led to the use of the 3-D model to define the dimensions of an industrial-sized robotic
arm. The results obtained indicate high stiffness and minimal deflection while achieving a
significant weight reduction when compared to commercial arms of the same size and capability.
ii
Dedication
To my father Garth, my mother Lynne, and all my friends and family who always
supported my efforts.
iii
Acknowledgements
First and foremost I would like to thank my advisors, Dr. Scott Nokleby and Dr.
Remon Pop-Iliev, who have provided me with continued support throughout my graduate career. In addition to this, I am grateful for their unending guidance throughout
all aspects of my thesis work. Their advice will prove to be valuable in my future
endeavours and for this their efforts will not be forgotten.
I would like to thank Mr. James Beltrano, Mr. Steven Bemis, Mr. Mike MacLeod,
and Mr. Garth Willis. Without their help I would not have successfully constructed
my prototype.
I also feel the need to acknowledge John Damiani, Jeff Martin, Kyle Mullin, Brad
Timson, and David Tripp. Their assistance throughout varying aspects of this project
as part of their undergraduate theses is greatly appreciated. Additionally, I would
also like to thank fellow graduate students Emad Abdalla, Michael Frejek, Sasha
Ginzburg, and Brian Riess who were always willing to offer their assistance.
I would like to thank BRIC Engineered Systems, the Ontario Centres of Excellence
(OCE), and the Natural Sciences and Engineering Research Council (NSERC) of
Canada whose financial support has made my research possible.
iv
Table of Contents
Abstract
ii
Dedication
iii
Acknowledgements
iv
Table of Contents
ix
List of Tables
List of Figures
xiv
1 Introduction
1.1
Problem Definition and Goal of Research . . . . . . . . . . . . . . . .
1.2
Customer Requirements . . . . . . . . . . . . . . . . . . . . . . . . .
1.3
Typical Properties of Traditional Commercial Robotic Arms . . . . .
1.4
Review of Polymer Composites . . . . . . . . . . . . . . . . . . . . .
1.4.1
Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.2
Particle-Reinforced Composites . . . . . . . . . . . . . . . . .
1.4.3
Fiber-Reinforced Composites . . . . . . . . . . . . . . . . . . .
1.4.3.1
Mechanical Properties . . . . . . . . . . . . . . . . .
1.4.3.2
Fiber Materials . . . . . . . . . . . . . . . . . . . . .
10
Polymeric Nanocomposites . . . . . . . . . . . . . . . . . . . .
12
1.4.4
1.5
1.6
1.4.5
Fiber-Polymer Composite Production Processes . . . . . . . .
13
1.4.6
Structural Composites . . . . . . . . . . . . . . . . . . . . . .
16
1.4.7
Methods of Strengthening Hollow Composites . . . . . . . . .
18
Review of Literature on Composite Robots . . . . . . . . . . . . . . .
20
1.5.1
Production Techniques . . . . . . . . . . . . . . . . . . . . . .
20
1.5.2
Connection Techniques . . . . . . . . . . . . . . . . . . . . . .
22
1.5.3
Results Achieved . . . . . . . . . . . . . . . . . . . . . . . . .
23
1.5.4
Composite Arms Used in Industry . . . . . . . . . . . . . . . .
28
1.5.5
Opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
Summary of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
2 Link Design
2.1
2.2
Manipulator Layout Selection . . . . . . . . . . . . . . . . . . . . . .
31
2.1.1
Comparison of Various Types of Robotic Arms . . . . . . . . .
31
2.1.2
Serial Arm Configuration . . . . . . . . . . . . . . . . . . . . .
34
Analysis of the Cross-Section of the Two Main Links . . . . . . . . .
36
2.2.1
Cross-Section Testing Procedure . . . . . . . . . . . . . . . . .
36
2.2.2
Results of Cross-Section Analysis for the Main Links . . . . .
39
2.2.2.1
Basic Shaped Hollow Tubes . . . . . . . . . . . . . .
39
2.2.2.2
Circular Tubes with Support Ribs . . . . . . . . . .
41
2.2.2.3
Circular Tube with Different Shaped Inner Tubes . .
42
2.2.2.4
Annulus with Different Sized Inner Tubes . . . . . .
44
2.2.2.5
Rectangular Tubes with Support Ribs . . . . . . . .
45
Analysis of Tapering the Main Links . . . . . . . . . . . . . .
47
Optimization of the Cross-Section . . . . . . . . . . . . . . . . . . . .
49
2.3.1
Determination of the Objective Function . . . . . . . . . . . .
50
2.3.2
Design of the Proof-of-Concept Prototype . . . . . . . . . . .
55
2.3.3
Formulation of the Equations for the Cross-Sectional Area . .
57
2.2.3
2.3
31
vi
2.4
2.3.4
Formulation of the Equations for Moment of Inertia . . . . . .
60
2.3.5
Formulation of the Equations for Angle of Twist . . . . . . . .
63
2.3.6
Determination of Constraints . . . . . . . . . . . . . . . . . .
65
2.3.7
Determination of Bounds for the Design Variables . . . . . . .
68
2.3.8
Material Specifications . . . . . . . . . . . . . . . . . . . . . .
68
2.3.9
Solutions Found from Optimization of Cross-Section . . . . . .
69
2.3.9.1
I-beam Supported Rectangular Link . . . . . . . . .
69
2.3.9.2
X-Supported Rectangular Link . . . . . . . . . . . .
70
2.3.9.3
Study Without Upper Bounds . . . . . . . . . . . . .
70
2.3.9.4
Recommendations from Optimization Results . . . .
71
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
72
3 Prototype Link Construction
73
3.1
Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
3.2
Modular Construction Process . . . . . . . . . . . . . . . . . . . . . .
74
3.3
Mold Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
75
3.3.1
Phase 1: Wooden Mold . . . . . . . . . . . . . . . . . . . . . .
75
3.3.2
Phase 2: Four Piece Fully-Detachable Aluminum Mold . . . .
76
3.3.3
Phase 3: Two Piece Fully-Detachable Aluminum Mold . . . .
78
Prototype Link Construction . . . . . . . . . . . . . . . . . . . . . . .
78
3.4.1
Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
78
3.4.2
Materials Used for Prototype . . . . . . . . . . . . . . . . . .
79
3.4.3
Ply Lay-up Pattern for Prototype . . . . . . . . . . . . . . . .
80
3.4.4
Manufacturing Procedure . . . . . . . . . . . . . . . . . . . .
81
3.5
Link Attachment Bracket . . . . . . . . . . . . . . . . . . . . . . . . .
82
3.6
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
86
3.4
vii
4 Prototype Link Testing

4.1
87
Physical Prototype Testing . . . . . . . . . . . . . . . . . . . . . . . .
87
4.1.1
Single Link Static Testing . . . . . . . . . . . . . . . . . . . .
87
4.1.1.1
Weight Support Component . . . . . . . . . . . . . .
88
4.1.1.2
Apparatus . . . . . . . . . . . . . . . . . . . . . . . .
89
4.1.1.3
Results . . . . . . . . . . . . . . . . . . . . . . . . .
90
4.1.1.4
Analysis . . . . . . . . . . . . . . . . . . . . . . . . .
91
Repeatability Testing of the 2-DOF Prototype . . . . . . . . .
91
4.1.2.1
Motor Control . . . . . . . . . . . . . . . . . . . . .
92
4.1.2.2
Results . . . . . . . . . . . . . . . . . . . . . . . . .
92
Destructive Tests . . . . . . . . . . . . . . . . . . . . . . . . .
94
4.1.3.1
Compression Test . . . . . . . . . . . . . . . . . . . .
94
4.1.3.2
Three Point Bending Test . . . . . . . . . . . . . . .
98
Finite Element Analysis . . . . . . . . . . . . . . . . . . . . . . . . .
100
4.2.1
Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
100
4.2.2
Material Properties . . . . . . . . . . . . . . . . . . . . . . . .
101
4.2.3
Modeling
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
102
4.2.4
Loading Conditions . . . . . . . . . . . . . . . . . . . . . . . .
103
4.2.5
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
104
4.3
Comparison of Physical and Virtual Prototype . . . . . . . . . . . . .
106
4.4
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
106
4.1.2
4.1.3
4.2
5 Design of the Full-Scale Robotic Arm
108
5.1
Motor Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
108
5.2
Dimensional Analysis of Links . . . . . . . . . . . . . . . . . . . . . .
111
5.2.1
Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
111
5.2.2
Testing Procedure . . . . . . . . . . . . . . . . . . . . . . . . .
112
5.2.3
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
113
viii
5.2.3.1
Links with Height Twice the Width . . . . . . . . . .
113
5.2.3.2
Relationship of Thickness and Static Deflection . . .
116
5.2.3.3
Links with Optimized Height-to-Width Ratio . . . .
117
5.2.3.4
Different Layering Patterns . . . . . . . . . . . . . .
118
5.3
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
121
5.4
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
123
6 Conclusions and Recommendations for Future Work
124
6.1
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
124
6.2
Recommendations for Future Work . . . . . . . . . . . . . . . . . . .
127
A MATLAB Code
135
A.1 Declaration of Constants . . . . . . . . . . . . . . . . . . . . . . . . .
135
A.2 Objective Function . . . . . . . . . . . . . . . . . . . . . . . . . . . .
136
A.3 Non Linear Constraint . . . . . . . . . . . . . . . . . . . . . . . . . .
137
A.4 Area of I-beam Supported Rectangular Link . . . . . . . . . . . . . .
138
A.5 Inertia of I-beam Supported Rectangular Link . . . . . . . . . . . . .
138
A.6 Area of X-Supported Rectangular Link . . . . . . . . . . . . . . . . .
139
A.7 Inertia of X-Supported Rectangular Link . . . . . . . . . . . . . . . .
139
A.8 Linear Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . .
140
A.9 Output for I-beam Supported Rectangular Link . . . . . . . . . . . .
140
A.10 Output for X-Supported Rectangular Link . . . . . . . . . . . . . . .
142
A.11 Output for I-beam Supported Rectangular Link Without Upper Bounds143
A.12 Output for X-Supported Rectangular Link Without Upper Bounds
B Danaher Data Sheets
144
147
B.1 Motors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
147
B.2 Gearheads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
150
ix
List of Tables
1.1
Commercially Available Robotic Arms [14] . . . . . . . . . . . . . .
1.2
Properties of Three Fiber-Epoxy Composites [5] . . . . . . . . . . . .
12
2.1
Effect of Taper Angle on the Displacement and Stress . . . . . . . . .
48
2.2
Comparison of Length of Two Main Links to Prototype Payload . . .
57
4.1
Vertical Displacement of the Link as Measured Using a Laser . . . . .
90
4.2
Laser Measurements For Repeatability Test . . . . . . . . . . . . . .
94
4.3
Comparison of Prototype Link Constructed of Different Materials . .
106
5.1
Static Flexibility of Four Commercial Robotic Arms . . . . . . . . . .
112
5.2
Stress and Deflection in Link for Different Layering Patterns . . . . .
119
List of Figures
1.1
Characteristics of Dispersed Phase Geometry [5] . . . . . . . . . . . .
1.2
Upper and Lower Bounds for Composites Mechanical Properties [5] .
1.3
Fiber Stress-Position Profiles, Polymer Composite Stress-Strain Curve

[5] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4
a) Laminar Composite Example - Skies b) Sandwich Panels [5] . . . .
17
1.5
Series of Elongated Members for Strengthening Support Systems [20]
19
1.6
Carbon-Epoxy Composite SCARA robot [27] . . . . . . . . . . . . . .
21
1.7
Schematic of Attachment of Steel Flange to Composite Arm [26] . . .
23
1.8
Stacking Angle for Max Fundamental Frequency [31] . . . . . . . . .
26
1.9
R R k R Robotic Arm [32] . . . . . . . . . . . . . . . . . . . . . . .
27
1.10 Kuka Robotic Arm Featuring Composite Third Link [1] . . . . . . . .
29
2.1
Schematics of Different Types of Main Configurations for Robotic Arms 32
2.2
6-DOF Serial Configuration of the Arm . . . . . . . . . . . . . . . . .
2.3
Force Applied on Single Straight Member Representing Two Main Links 38
2.4
Cross-Sections of Family of Basic Shaped Hollow Tubes . . . . . . . .
40
2.5
Cross-Sections of Family of Circular Tubes with Support Ribs . . . .
41
2.6
Cross-Sections of Family of Circular Tube with Different Shaped Inner
36
Tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
43
2.7
Cross-Sections of Family of Annulus with Different Sized Inner Tubes
44
2.8
Cross-Sections of Family of Rectangular Tubes with Support Ribs . .
46
xi
2.9
Displacement Versus Mass for Different Cross-Sections . . . . . . . .
47
2.10 Tapered Member Representing the Two Main Links . . . . . . . . . .
48
2.11 End-Effector Displacement Versus Mass for Various Taper Angles . .
49
2.12 Configuration of Robotic Arm to be Analyzed for Cross-Section Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.13 Side View of Free Body Diagram of the Analyzed Robotic Arm
50
. . .
51
2.14 Free Body Diagram for Bending of the Two Main Links . . . . . . . .
52
2.15 Free Body Diagram for Bending of the Wrist Configuration . . . . . .
53
2.16 Free Body Diagram for Torque Applied on the Main Links . . . . . .
54
2.17 Resulting Deflection of the Wrist Configuration due to Twisting of the

Main Links . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
2.18 Geometric Representation of Deflection due to Twisting of the Main

Links . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
2.19 Free Body Diagram for Maximum Output Torque of PR110 Motor
Acting on Main Links . . . . . . . . . . . . . . . . . . . . . . . . . . .
56
2.20 Cross-Sectional Area of the I-Beam Supported Rectangular Tube . . .
58
2.21 Cross-Section of the X-Supported Rectangular Tube . . . . . . . . . .
59
2.22 Middle and Corner of Cross-Section for the X-Supported Rectangular

Tube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
59
2.23 Cross-Sectional Area of the X-Supported Rectangular Tube . . . . . .
60
2.24 Neutral Axis for the Cross-Sectional Geometries of Concern
61
. . . . .
2.25 Moment of Inertia for the I-beam Supported Rectangle Cross-Sectional

Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
61
2.26 Moment of Inertia for the X-Supported Rectangle Cross-Sectional Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
63
2.27 Equivalent Hollow Rectangle for the Ideal Cross-Sectional Geometries
64
xii
2.28 Mean Area Enclosed within Boundary of Centerline for Equivalent Hollow Rectangle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
3.1
Schematic of Modular Construction Process . . . . . . . . . . . . . .
75
3.2
One Piece Solid Wooden Mold . . . . . . . . . . . . . . . . . . . . . .
76
3.3
Four Piece Fully-Detachable Mold . . . . . . . . . . . . . . . . . . . .
77
3.4
Two L-Shaped Piece Fully-Detachable Mold . . . . . . . . . . . . . .
78
3.5
Vacuum Apparatus Used to Produce the Composite Link . . . . . . .
82
3.6
Bracket for Attaching Composite Link to the Joint Motor . . . . . . .
83
3.7
Composite Link Connected to the Motor with Attachment Bracket .
84
3.8
2-DOF Main-Arm Proof-of-Concept Prototype Mounted to the Floor
84
3.9
2-DOF Main-Arm Proof-of-Concept Prototype . . . . . . . . . . . . .
85
4.1
Device for Supporting Weight on the Composite Link . . . . . . . . .
88
4.2
Link Mounted to Wall . . . . . . . . . . . . . . . . . . . . . . . . . .
89
4.3
Schematic of Experimental Setup . . . . . . . . . . . . . . . . . . . .
90
4.4
First Repeatability Test Position (45 , 45 ) . . . . . . . . . . . . . . .
93
4.5
Second Repeatability Test Position (55 , 35 )
. . . . . . . . . . . . .
93
4.6
Adapter Plate Used for Compression Test . . . . . . . . . . . . . . .
95
4.7
Apparatus for Compressive Test . . . . . . . . . . . . . . . . . . . . .
96
4.8
Compressive Force Versus Displacement of Composite Link . . . . . .
97
4.9
Section of Link After Compressive Test . . . . . . . . . . . . . . . . .
97
4.10 Apparatus for Three Point Bend Test . . . . . . . . . . . . . . . . . .
98
4.11 Bending Force Versus Displacement of Composite Link . . . . . . . .
99
4.12 Section of Link After Three Point Bend Test . . . . . . . . . . . . . .
100
4.13 Modular Base Unit for Virtual Prototype . . . . . . . . . . . . . . . .
101
4.14 Exterior Profile for Virtual Prototype . . . . . . . . . . . . . . . . . .
101
4.15 Bounded End of Virtual Prototype Composite Link . . . . . . . . . .
103
xiii
4.16 Surface Traction on Virtual Prototype Composite Link . . . . . . . .
104
4.17 Deflection of Virtual Prototype Link . . . . . . . . . . . . . . . . . .
105
5.1
Scenario for Motion of Full-Scale Robotic Arm . . . . . . . . . . . . .
109
5.2
Free Body Diagram of Second Main Link . . . . . . . . . . . . . . . .
110
5.3
Loading Scenario for Model of Full-Scale Main Links . . . . . . . . .
113
5.4
Diagram of Largest Link That Fits Through 0.76 [m] Diameter Opening115
5.5
Average Cross-Sectional Area Versus Static Deflection for Links with

Width-to-Height Ratio of 2.0 . . . . . . . . . . . . . . . . . . . . . . .
5.6
5.7
116
Average Cross-Sectional Area Versus Static Deflection for Links with

Width-to-Height Ratio of 3.63 . . . . . . . . . . . . . . . . . . . . . .
120
Free Body Diagram of Second Main Link . . . . . . . . . . . . . . . .
122
xiv
Nomenclature
CAN computer area network

CFRP carbon fiber reinforced plastic
DOF degrees-of-freedom
FEA finite element analysis
MDF medium density fiberboard
MEKP methyl ethyl ketone peroxide
PAN polyacrylonitrile
PD proportional and derivative
R revolute joint
SCARA selectively compliant assembly robot arm
US United States
xv
Chapter 1
Introduction
1.1
Problem Definition and Goal of Research
Currently, robotic arms that are used for large-scale applications are constructed of
either steel or aluminum. These materials, however, have a low specific stiffness
(stiffness/density) and specific strength (strength/density) and thus the respective
robotic arms feature a poor payload-to-weight ratio. Therefore, the robotic arms
are typically very large and heavy in order to provide high precision and repeatability
while being rated for the required payload. Consequently, very powerful and expensive
motors are necessary to provide movement for such heavy robotic arms.
Due to the intrinsic weight of steel and aluminum arms, there is a trade-off between
providing high stiffness and lightweight. Therefore, a material with higher specific
stiffness and strength needs to be implemented into the construction of robotic arms.
The goal of this research is to investigate the use of composite materials in robot
arm construction. In addition to this, in order to offer further improvements over
current industrial-sized robotic arms in terms of both the stiffness and load carrying
capacity of the arm while decreasing the weight, the design of lightweight robotic links
is investigated thoroughly in this research work.
1.2
Customer Requirements
The objective of this thesis was to design a 6-DOF robotic arm capable of large-scale
applications which met or exceeded the following functional requirements:
As light as possible.
Payload of 50 [kg].
Reach of 5 [m].
Repeatability equivalent to industry standard.
Able to fit through a 0.76 [m] (30 [in]) diameter opening.
Tool linear speed of 0.15 [m/s] (6 [in/s]).
Capable of performing operations on large structures with complex curvature.
One such application of the robotic arm is performing operations on airplanes. Consequently, the robotic arm must be designed for high maneuverability in order to navigate the geometry of the plane. Additionally, since airplanes are very large structures,
the entire robotic arm must be able to move around the periphery of the airplane. As
a result, the emphasis on creating a lightweight robotic arm is of even greater concern.
1.3
Typical Properties of Traditional Commercial

Robotic Arms
Currently, there are no commercially available robotic arms that can satisfy the robot
specifications. As far as 6-DOF commercial robotic arms that are rated for a payload
of approximately 50 [kg] are concerned, they have a maximum reach of approximately
2 [m]. Therefore, to have a better idea of the specifications of competitor robotic
2
arms, longer reach arms (that subsequently have a higher rated payload) were also
investigated. Table 1.1 outlines some of the primary specifications for five commercial
robotic arms.
Table 1.1: Commercially Available Robotic
Kuka
Fanuc
ABB
PROPERTY
KR 60-3 M-710iC/50 IRB 4600-60
Mass
635 kg
560 kg
435 kg
Payload
60 kg
50 kg
60 kg
Reach
2033 mm
2050 mm
2050 mm
Repeatability 0.20 mm
0.07
0.19 mm
Axis 1 Speed
128 /s
175 /s
175 /s
Axis 2 Speed
102 /s
175 /s
175 /s
Axis 3 Speed
128 /s
175 /s
175 /s
Arms [14]
Kuka
KR 100-2P
1465 kg
100 kg
3500 mm
0.20 mm
102 /s
96 /s
95 /s
Motoman
EPH100
1520 kg
100 kg
3010 mm
0.30 mm
110 /s
110 /s
110 /s
The first three robotic arms that have a payload and reach of approximately 50 [kg]
and 2 [m], respectively, have a mass ranging from 435 [kg] to 635 [kg]. Consequently,
their payload-to-weight ratio is approximately 0.10. These robotic arms only have a
reach of 2 [m], so in order to provide a reach of 5 [m] the payload-to-weight ratio would
be considerably worse. The larger arms that have a payload of 100 [kg] and a reach
of 3.0 [m] and 3.5 [m], have an even lower payload-to-weight ratio of approximately
0.066. Therefore, it can be assumed that a similar robot that has a payload of 50
[kg] and a reach of 5 [m] would feature a payload-to-weight ratio of approximately
0.05. Therefore, the total weight of the robotic arm would be in the range of 1,000
[kg]. Additionally, the repeatability quoted in Table 1.1 is generally around 0.20
mm, therefore the robotic arm is designed to this specification.
All the robotic arms seem to have very high motor speeds. Motor speeds this high are
required for pick-and-place operations but for the intended application these speeds
are unnecessary. Thus, less powerful, and consequently lighter, motors can be used
to operate at lower speeds in order to reduce the overall weight of the robotic arm.
The problem with current large-scale application robotic arms is that they are very
3
heavy. Consequently, the motors and drive systems that are required for the robot
to operate at the desirable levels of speed and accuracy are very powerful and thus
expensive. In order to alleviate this problem, a lighter alternative to the heavy steel
and aluminum arms that are currently used must be realized. Polymer composites are
potentially a feasible material choice for constructing a lightweight large-scale robotic
arm that still offers long reach, high payload, and good repeatability.
1.4
1.4.1
Review of Polymer Composites

Overview
Polymer composites can be created by combining two distinct polymer materials to

obtain a material that has a combination of properties. Consequently, polymer composites can be made to offer stiffness and strength that is better than aluminum and
competitive with steel even though they are around five times lighter than steel and
half the mass of aluminum. Consequently, polymer composites are about eight to
ten times better than steel and aluminum in terms of specific stiffness and specific
strength [5]. Therefore, this research work focuses on exploring the feasibility of using
polymeric composite materials for constructing lightweight long-reach robotic arms.
1.4.2
Particle-Reinforced Composites
A particle-reinforced composite is a composite in which the dispersed material has

a sphere-like shape with similar dimensions in all directions. The matrix material is
continuous and surrounds these particles. The properties of composites are not only
a function of the properties of the materials and their relative amounts, but also a
function of the geometry of the dispersed material within the matrix material. The
characteristics of the dispersed phase geometry are the particle size, shape, distribution, and orientation as shown in Figure 1.1 [5].
4
Figure 1.1: Characteristics of Dispersed Phase Geometry [5]
Particle-reinforced composites are classified as either large particle composites or

dispersion-strengthened composites. Large particle composites are defined as composites in which the particle is so large that the particle-matrix interactions cannot
be treated on the atomic or molecular level [5]. The particles restrain movement of the
matrix material at the locations at which the particles exist. The particles bear some
of the load because the matrix transfers some of the applied stress to the particles.
Due to the size of the particles dispersed within the matrix material, the amount of
reinforcement depends on the strength of the bonds at the matrix-particle interface.
The mechanical properties of the composite are influenced by the volume fraction of
the matrix and the particle. The mechanical properties of the composite improve as
the volume fraction of the dispersed particles is increased. In fact, the mechanical
properties of the composite can be predicted within an upper and lower bound as
demonstrated in Figure 1.2 if the mechanical properties and volume fractions of the
two materials are known [5].
The particles within dispersion-strengthened composites are considerably smaller such
that the particle-matrix interactions that lead to strengthening occur on the atomic
or molecular level [5]. Consequently, the matrix bears the majority of an applied load
and the particles impede the motion of dislocations within the matrix, restricting
plastic deformation. These composites prove to offer greater improvements in the
mechanical behaviour than large particle composites but due to the small size of the
dispersed particles the fabrication of these composites is more complicated [5].
Figure 1.2: Upper and Lower Bounds for Composites Mechanical Properties [5]
1.4.3
Fiber-Reinforced Composites
1.4.3.1
Mechanical Properties
A fiber-reinforced composite is a composite in which the dispersed material is in the

form of a fiber (i.e., the length is considerably greater than the diameter). Fiberreinforced composites are categorized by the length of the fiber. Those designated
as continuous fiber-reinforced composites are composites in which the fibers are long,
whereas discontinuous fiber-reinforced composites are composites in which the fibers
are short. In fact, a continuous fiber has a length that is equal to or greater than
fifteen times that of the critical length and a discontinuous fiber has a length shorter
than this. The critical fiber length, lc , is described as being the fiber length necessary
for effective strengthening and stiffening of the composite material and can be defined
as [5]:
lc =
f d
2c
(1.1)
where f is the fiber ultimate or tensile strength, d is the fiber diameter, and c is the
smaller value of the fiber-matrix bond strength and the shear yield strength of the
matrix.
6
The mechanical characteristics of the fiber-reinforced composite depend on the ability

of the composite to transmit the applied load from the matrix material to the fibers.
The interfacial bond between the fiber and matrix material acts as the medium. The
fiber-matrix bond is eliminated at the ends of the fibers when a stress is applied and
thus there is no load transmittance there. As a result, a fiber-reinforced composite
with an equal volume fraction of the fiber material but longer fibers than a composite
of the same two materials will be able to provide more load transmittance to the fibers
since the ends of the fibers contribute to a lower percentage of the total volume of the
fiber material. This effect is illustrated in Figure 1.3a. When the length of the fiber
is equal to the critical length the maximum fiber load is achieved only at the middle
of the fiber. However, for fiber lengths that are greater than the critical length, the
maximum fiber load is achieved over a greater length of the fiber, thus increasing
the amount of load that is transmitted to the fiber. This translates to an increase in
the amount of reinforcement that is provided. However, the shorter the fiber the less
inherent flaws there will be in the material and thus the closer the actual strength
of the material will be to the theoretical value. Therefore, in order to benefit from
using continuous fibers, it is best to use a fiber with a small diameter because as the
fibers become smaller in diameter, the chances of inherent flaws in the material are
reduced [6].
The stress-strain diagram for the fiber, matrix and resulting composite material is
shown in Figure 1.3b. As it illustrates, once the applied stress has surpassed the
yield stress of the matrix material the composite material does not undergo plastic
deformation but rather the rate of elastic deformation is just increased. As the matrix
material begins to undergo plastic deformation the fibers that reinforce the composite
support more of the applied load and stretch elastically at a greater rate until the
strain has reached the tensile strain of the fiber. At this point, failure of the composite
material occurs. However, the failure is not catastrophic. After the fibers fail, the
matrix is still intact since its tensile strain is greater than that of the fibers. Consequently, the composite now contains shortened fibers that are capable of sustaining a
diminished load as the matrix continues to plastically deform [5].
Figure 1.3: Fiber Stress-Position Profiles, Polymer Composite Stress-Strain Curve [5]
Polymer composites are highly anisotropic materials in that their properties are dependent on the direction in which they are measured. When the fibers are aligned in
one direction this anisotropic property is most evident. Since continuous fibers are
always aligned, they produce polymer composites that are highly anisotropic. For
this reason it is important to determine the mechanical properties under conditions
of both longitudinal and transverse loading [5].
Under longitudinal loading the load is applied in the same direction as the fiber alignment direction. Under this condition it is assumed that the fiber-matrix interfacial
bond is strong enough to ensure that the deformation of the fibers and matrix is the
same. The modulus of elasticity of the polymer composite under longitudinal loading,
Ecl , can be determined using:
8
Ecl = Em Vm + Ef Vf
(1.2)
where Em is the modulus of elasticity of the matrix material, Ef is the modulus of

elasticity of the fiber, Vm is the volume fraction of the matrix material, and Vf is the
volume fraction of the fiber [5].
For transverse loading, the load is applied in a direction perpendicular to the direction
of fiber alignment. In this situation, the stress that the fiber and matrix material is
exposed to is the same. As a result, the modulus of elasticity of the polymer composite
under transverse loading, Ect , can be determined using [5]:
Ect =
Em Ef
Ef Vm + Em Vf
(1.3)
Under longitudinal loading the mechanical properties are the greatest whereas they
are at their lowest while under transverse loading. Therefore, when the loading is
applied in any other direction, the mechanical properties of the polymer composite
falls somewhere between these two extremes.
Since the mechanical properties are maximized when the stress is applied in the same
direction as the alignment of the fibers, the degree of reinforcement for different scenarios is compared to that of longitudinal loading. Despite the fact that the mechanical
properties of the polymer composite are improved over that of the matrix material
while under transverse loading, the improvement is so negligible that the reinforcement efficiency is assumed to be zero. For situations where the stress direction is
changing it is better to have the fibers randomly oriented throughout the matrix material. This helps to ensure that the mechanical properties are isotropic (constant in
all directions). If the fibers are randomly and uniformly distributed within a specific
plane, then the reinforcement efficiency is 3/8 for a stress direction in any direction
in the plane of the fibers. If the fibers are randomly and uniformly distributed within
all three dimensions, then the reinforcement efficiency is 1/5 for a stress applied in
any direction [5].
1.4.3.2
Fiber Materials
There are three primary fiber materials that are used as reinforcement in polymer
composites. These are glass, aramid, and carbon. Glass is the most commonly used
fiber in polymer composites because it is relatively inexpensive and easy to produce.
It is easily drawn into fibers from a melt of glass in the molten state. The molten
glass passes through numerous platinum alloy nozzles to produce filaments of glass.
These filaments are then sprayed with an organic sizing solution that contains binders
to help form the filaments into strands, lubricants to prevent abrasion of filaments,
and coupling agents to improve adhesion between the glass fiber and the matrix
material. The filaments are then wound together to form a strand. Glass fibers
are also favoured because of their high strength. As a result, when embedded in a
plastic matrix, a composite with a very high specific strength is created. However,
despite their strength, glass fibers have a low elastic modulus. Consequently, they
produce a composite that is not very stiff. They do however produce composites that
are chemically inert making them useful in a variety of corrosive environments but
they do degrade at temperatures above 200 C. Lastly, their density is considerably
higher than the other two fiber materials, making them less desirable for the specified
application.
Aramid fibers are an aromatic organic compound made of carbon, hydrogen, oxygen,
and nitrogen. They are distinguished by their high strength and high modulus of
elasticity despite having a low density. Their strength is especially high (the highest of
the three fibers) making them good candidates for composites to be used for structural
purposes. In addition to this, they have high impact, creep, and fatigue resistance.
Furthermore they are a bit more than 10 times as expensive as glass fibers, as indicated
10
in Table 1.2, but are still relatively inexpensive. The manufacturing processes for the
fibers is relatively easy as a solution of the necessary polymers and strong acids are
extruded into hot cylinders then washed and dried on spools. Afterwards, the fibers
are stretched and drawn to improve strength and stiffness. The major drawback of
aramid fibers is that they are relatively weak in compression. For this reason, they
should be avoided for applications that apply a compressive load.
Carbon fibers have an extremely high modulus of elasticity (much higher than that of
aramid and glass fibers) and high strength. They are not really affected by temperature as they have a low thermal expansion and retain their high stiffness and strength
at elevated temperatures. Their density is still relatively low, falling between that of
aramid and glass fibers as shown in Table 1.2. Like the other two fiber materials mentioned, they are not affected by a wide variety of solvents, acids, and bases. However,
their main drawback is their cost. Carbon fibers are considerably more expensive
than glass fibers costing upwards of 80 times as much [5].
Carbon fibers are also advantageous since there is a vast diversity of different forms
of carbon which enables the user to select the required strength and stiffness by using
different organic precursor materials. In addition to this, the carbon fiber can be
totally crystalline in a form commonly known as graphite. This further enhances
the mechanical properties of the fiber. Using a polyacrylonitrile (PAN) precursor the
carbon fibers are made as follows. PAN fibers are stretched five to ten times their
length to improve their mechanical properties before being heated to 250 C in order
to stabilize their dimensions during the next heating processes. It is then heated to
between 1,000 C and 1,500 C so that the complex organic substance can decompose
to a simpler structure. The fibers are then surface treated to create carbon fibers. If
graphite fibers are to be created, then before surface treating the fibers are heated
once more to above 2,500 C enabling the microstructures to become graphitic.
11
Table 1.2: Properties of Three Fiber-Epoxy Composites [5]

Glass Vf =0.6 Aramid Vf =0.6 Carbon Vf =0.6
(E-Glass)
(Kevlar 49)
(High Strength)
3
Density
[kg/m ]
2,100
1,400
1,700
Tensile Modulus Long.
45
76
220
(GPa)
Trans.
12
5.5
6.9
Tensile Strength Long.
1,020
1,380
760
(MPa)
Trans.
40
30
20
Ultimate Tensile Long.
2.3
1.8
0.9
Strain ( )
Trans.
0.4
0.5
0.4
Cost
Fiber
1.90 - 3.30
31
175 - 225
($US/kg)
Comp.
22
55 - 62
200 - 275
1.4.4
Polymeric Nanocomposites
A nanocomposite is a composite in which the dispersed material is a particle with

at least one dimension no greater than 100 nanometers (nm). The nanoparticles are
usually in the shape of a fiber so its diameter is restricted to being smaller than 100
[nm]. Using dispersed material with dimensions on the nanometer scale rather than
micrometer scale can result in large improvements in the created composite. Since
the diameter of an average polymer coil is 40 [nm], using nanoparticles for the filler
can improve the reinforcement. By using a dispersed material that is on the same size
scale as the matrix material it ensures that they will better interact and bond with
each other [7].
Thanks in part to nanoparticles incredibly small size, they inherently have very
high surface-to-volume ratios. This means that for each nanoparticle there is a large
amount of interaction with the polymers which of course translates to a large amount
of reinforcement. Consequently, nanocomposites can provide considerably more reinforcement than an equal amount of the macrocomposite counterpart. Therefore, a
lower percentage of nanoparticles is necessary (in the range of 0.5% to 5% by weight)
to create the composite [7].
Another reason why nanoparticles prove to be an effective reinforcing material is that
12
they have a high aspect ratio (their length is considerably larger than their diameter).
As previously mentioned, this increases the amount of reinforcement that the particle
provides because the maximum load is achieved over a greater segment of the length
and thus a greater amount of the load can be transmitted from the polymer matrix
to the nanoparticles. In addition to this, since nanoparticles are so small they are less
likely to contain flaws and as a result their actual strength is relatively close to their
theoretical value [8].
Although nanocomposites appear to be very promising in terms of material properties,
they are still very much at the experimental stage. Since the long-term goal of this
work is the development of an industrial arm, fiber composites were selected since
they are a more mature technology.
1.4.5
Fiber-Polymer Composite Production Processes
The main production processes for fiber-reinforced composites are pultrusion, autoclave forming, resin transfer molding, and filament winding. All three primary fiber
materials can be produced using these techniques but there are subtle differences that
result in differences in the cost of production. Glass fibers adhere poorly to polymers
which require additional steps to be performed in order to produce the composite.
As a result, the cost of the composites using the three different fibers is much closer
than that of the fibers alone. As indicated in Table 1.2, aramid fiber composites and
carbon fiber composites are two and a half and eleven times as expensive as glass fiber
composites, respectively [5].
Prepregs are ready made tapes of continuous fiber weaves preimpregnated with a
polymer resin that is only partially cured. They enable the user to directly mold
and fully cure the product without having to add any resin. They are used in a hand
lay-up process in which layers of prepregs are stacked in a specific manner (with fibers
in certain directions for each layer) until the desired thickness of the part is achieved.
13
The temperature and pressure is then elevated to cure it and produce the polymer
composite part. Prepregs can also be utilized to simplify some of the manufacturing
techniques as the step of drawing the fibers through a resin bath is eliminated.
Pultrusion is a technique in which fibers are drawn through a resin tank to impregnate
them before pulling them through a die that preforms them to the desired shape and
establishes the resin/fiber ratio. It then passes through a heated curing die that
shapes it into the final shape and initiates curing. A pulling device is located at the
end to draw the stock through the different stages of the process [5]. Pultrusion is a
quick and cost effective technique for producing components that have a continuous
length and a constant cross-sectional shape.
The autoclave forming processing technique is able to produce parts of practically any
shape and as a result is a fairly complicated process. First, a peel ply made out of
nylon or cellophane is coated with a releasing agent like Teflon and placed on a mold.
Prepregs are used to reduce the number of steps involved and are laid up one ply at
a time by automated means or by hand. After the lay-up is sealed at the edges to
form a vacuum seal, bleeder sheets made of glass cloth are placed on the edges and
top of the lay-up in order to get rid of volatiles and excess resin. Vacuum connections
are then placed over the bleeder sheets and the lay-up is bagged. A partial vacuum is
developed and the entire assembly is placed into an autoclave and heat and pressure
are applied to cure the matrix material and form the polymer composite [6].
Resin transfer molding is a relatively simple process. A low viscosity resin is injected
under low pressure into a closed mold that contains the fiber preforms. Once all the
matrix material has been injected the part is allowed to cure. Beyond being quick,
it is also less expensive than hand lay-up and offers the possibility of automation.
However, the process is expensive since it requires two molds.
The filament winding technique begins by impregnating fibers by drawing them through
an in-line resin bath and winding them around a mandrel. Alternatively, prepregs can
14
be used instead but are often not used because they do not allow the user to control
the properties of the composites and are more expensive to purchase. Depending on
the desired properties of the product, different winding patterns like helical, polar,
and circumferential are used. After the appropriate number of layers are applied,
the part is allowed to cure in an oven or at room temperature. Once fully cured,
the mandrel is removed. This technique can be used to create open or closed-ended
hollow products. For closed-ended the mandrel needs to be made out of a material
that can be broken away in order to remove it from the part. Filament winding is
favoured for hollow parts because it is a very quick process since it is automated.
United States (US) patents 2785442, 3429758, and 4878984 [911] describe a method
or apparatus for the filament winding technique. The first two patents are fairly
similar in that they involve winding layers of fibers with opposing angles around a
mold [9, 10]. US patent 3429758 however has improvements over US patent 2785442
because it surrounds the layers of fibers with foam and concludes with a final layer
of transparent plastic for improved aesthetics and weathering characteristics [10]. US
patent 4878984 describes a filament winding machine that utilizes a parallelogram
mechanism to apply a plurality of equally spaced parallel fiber slivers. By expanding
the parallelogram, the distance between slivers and the angle can be altered which
leads to a varying wall thickness. This can be utilized to create a taper [11].
US patents 4118814, 4512835, 5609349, 5238716, and 6367225 [1216] discuss products
that are created using the filament winding technique. US patents 4118814, 4512835,
and 5609349 introduce a product that is created by cutting a part that is filament
wound [1214]. US patent 5609349 utilizes a diamond blade circular saw to cut the
part and ensure that the strength of the component is not compromised [14].
US patent 5238716 introduces a hollow composite beam with a rectangular shape [15].
It is produced by filament winding an interior layer and then placing four previously
made uni-directional fiber plates around the interior layer and then winding an exterior
15
layer. The advantage of using the pre-made plates is that it enables the beam to be
made with a non-uniform thickness in order to improve its compressive and tensile
strength.
US patent 6367225, describes a filament wound light pole that is 20 feet long and can
support a 300 pound force at its opposite end without exceeding a deflection of 20
inches [16]. In order to successfully do so, the light pole is constructed to include a
taper, the tension in the tow thread during the filament winding process is between 30
and 100 pounds, and the pole is made to include multiple layers in which the interior
and exterior circumferential layers surround a layer of helical wound fibers.
1.4.6
Structural Composites
A structural composite is typically made up of both homogeneous (a body in which

the properties are the same at all points in the body) and composite materials. Its
properties not only depend on the properties of the materials that make up the part,
but also on the geometry of the parts [5].
A laminar polymer composite is a primary type of structural composite composed
of plies of fiber-reinforced composites stacked together. For this reason, prepregs are
often used in the construction of the laminate. The plies are stacked together such
that the high-strength direction (direction of the fibers) of each successive layer varies.
By using this stacking sequence, it ensures that the laminar composite has relatively
high strength in a number of directions in the two-dimensional plane. Consequently,
the strength in any given direction is less than the maximum that is achieved by
having the layers stacked in a uni-directional manner. Additionally, by using different
orientations of the layers, different properties can be enhanced. For example a ski, as
shown in Figure 1.4a, has the layers oriented in different sequences in order to provide
both longitudinal and torsional stiffness. Surrounding these fiber-reinforced composite
layers are the homogeneous bodies that are used to enable the ski to perform as desired
16
(for example the steel edge acts as a blade that can cut into the snow enabling the
skier to turn) [5].
Hybrid laminar composites can also be used to create a composite with a better
combination of properties. A composite is considered a hybrid if more than one fiber
or one matrix material is used in the laminate. The ply can consist of two different
fibers or matrix materials or the laminate can be made up of plies that have different
fiber or matrix materials [5].
Sandwich panels are the other primary class of structural composites. As shown in
Figure 1.4b, they consist of two outer face sheets that are adhered to and separated
by a core. The face sheets are made of a high-strength composite material such as
graphite-epoxy composite, whereas the core is constructed of a lightweight material
like foam. It is necessary to use a material that provides high strength and stiffness
for the faces because it is here that the maximum stresses occur and the majority of
the axial forces are supported. Therefore the core does not need to be as strong and
rigid so a lightweight material with lower strength and elastic modulus is chosen.
Figure 1.4: a) Laminar Composite Example - Skies b) Sandwich Panels [5]
Since the resistance to bending of a rectangular cross-sectional plate or beam is proportional to the cube of the thickness, it is optimal to use a core with a larger thickness
17
and thus using a less dense material can keep the total weight minimal. Additionally,
the cores purpose is to separate the faces and resist deformations perpendicular to
the face plane as well as providing shear rigidity along planes that are perpendicular
to the face plane. Different shapes like the honeycombs shown in Figure 1.4b can be
used to reduce the amount of material and thus the total weight [5].
1.4.7
Methods of Strengthening Hollow Composites
The following patents describe products that use an internal support structure to
improve the strength of a hollow component. US patent 1141067 [17] describes a
method for reinforcing a tube. An initially hollow tube is filled with a series of
identical hollow segments so as to create a hollow tube with an internal support
system. Additionally, it is not limited to the number of equally sized segments that
are inserted and fixed into the tube.
US patents 3339326 and 3544417 [18,19] describe support structures in which a series
of elongated members are arranged to provide support. The interlocking members
have a hollow triangular cross-section in order to create a lightweight and strong
structure. In addition to the elongated members, foam is added inside of them to
improve the strength without significantly adding to the weight. The methods for
creating the two support structures differ however in that the elongated member is
first made and then filled with foam for the first patent [18], but for the second patent
the foam core is first made and then the fiberglass fabric is wrapped around it [19].
US patents 6081955, 3779487, 4223053, and 6655633 [2023] describe support systems
that are made up of a series of elongated members that are attached together to form
a support system as illustrated in Figure 1.5. The modular units of these patents are
considerably long, in the range of 20 feet. Additionally, the cross-sectional geometries
have a variety of shapes ranging from triangular to circular and including variations
of these simple shapes. The units are sandwiched between an upper and lower layer
18
that contacts all of the elongated members. These layers transfer the applied forces to
the members that then in turn support the load. Additionally, the modular elongated
members are designed to be hollow in order to keep the weight and cost to a minimum
while still providing a considerable increase in the strength. US patent 6081955 differs
from the rest of the patents in that the modular support structure made of a composite
material is produced using hand or automatic lay-up whereas the other three patents
describe a support structure that is produced using filament winding. However, all
these patents attach the individual modular units by assembling them in the proper
orientation and applying a resin and curing it.
Figure 1.5: Series of Elongated Members for Strengthening Support Systems [20]
The modular fiber reinforced bridge described in [24] bonds the individual modular
units by first assembling them and then filament winding around the collection of the
individual units. This method is beneficial over that described in the above patents
because it provides a stronger bond.
19
1.5
Review of Literature on Composite Robots
Since composite materials are a combination of a reinforcing material and support

material their mechanical characteristics can be controlled. Different reinforcing and
support materials can be combined to create an endless number of variations with
drastically different mechanical properties. Additionally, the percentages of the two
materials can be varied to offer further range of the mechanical properties. Furthermore, the way in which the two materials interact can be varied to further alter the
properties of the composite material.
The high degree of control associated with composite materials allows the user to
construct a material that is specifically designed for their application. Therefore,
composite materials are desirable for the construction of robotic arms because they
have a very high potential for improvements in stiffness, strength, and damping capability. This is why there has been a considerable amount of research performed on
the production of robotic arms out of polymer composites over the previous 20 years.
1.5.1
Production Techniques
Choi et al. [25] constructed a single link arm using an AS4/3501-5A magnamite
graphite composite. It was produced using 20 laminate plies. All the plies were
oriented in the same direction, parallel to the longitudinal axis (0 ) to increase the
bending stiffness of the link. The link was fabricated using the hand lay-up technique
using prepreg tape. Vacuuming and autoclaving were used to facilitate the production
process by shaping the surfaces of the panel and applying a pressure to evenly apply
the epoxy resin and facilitate the curing process.
In 2001, Lee et al. [26] also created a polymer composite single link arm that was
modeled as a cantilever beam. The link was created as a hollow cylindrical tube with
a length of 564 [mm] and diameter of 110 [mm]. They constructed the link using 66
20
plies of USN125 carbon prepreg to give the link a total thickness of 7 [mm]. They
used a 5 stacking angle since 0 could induce axial splitting under vertical loads
due to its low transverse stiffness and strength. The plies were laid-up on the mandrel
by hand and the link was manufactured using the autoclave vacuum bag process.
Lee et al. [27] also manufactured a SCARA (selectively compliant assembly robot arm)
type composite robot for handling large LCD glass panels as shown in Figure 1.6. The
end-effector was produced using USN125 carbon-epoxy composite. The hollow boxbeam end-effector was created using prepreg hand lay-up of 16 plies with 15 fiber
orientation. The plies were laid directly onto a Nomex honeycomb (HRH-10-1/8-4.0)
core that was used to improve the rigidity of the arm without adding any significant
increase in weight. The entire assembly was cured using the autoclave vacuum bag
degassing method.
Figure 1.6: Carbon-Epoxy Composite SCARA robot [27]
Lee et al. [28] subsequently redesigned the links of the SCARA type robot. All the
links were still designed to have a box-beam shape; however, a polyurethane foam core
was used. The prepreg plies were hand-laid at an angle of 30 on the closed mold
and the foam injected between the layers. The mold was then placed in an autoclave
to cure both the epoxy and the foam. The foam core adhered to the composite as the
excess epoxy resin penetrated the foam core.
21
Lessard et al. [29] integrated a polymer composite link into a 7-DOF manipulator.
They used LTM25 graphite with epoxy for the construction of the second link (the
upper-arm). They designed the link to be a box-beam (hollow rectangular crosssection) with a pair of holes on each end. FEA determined that the optimal laminate
lay-up is a repeating pattern of [0 /0 /45 /-45 ]. Since the force of the actuators is
exerted on the holes, two woven patches that include a 0 layer and 90 layer and
another two woven patches that include a 45 and -45 layers were added to the hole
region to improve the strength at this location.
A stainless steel end-fitting is attached to the link such that it surrounds the exterior
profile of the link. An external aluminum mold was used for the construction to control
the external dimensions of the link. The plies were cut from uni-directional prepreg
and hand-laid. The authors investigated sandwiching both an aluminum flex-core and
Nomex honeycomb between the layers but it was determined that they hindered the
compaction of the layers around the corners. Additionally, it was determined that
the cores barely improved the in-plane strength around the holes so they opted not
to include any sandwich support. Additionally, the part was cured at a maximum
temperature of 79 C using a vacuum-bagging technique.
1.5.2
Connection Techniques
As previously mentioned, the link discussed in [29] used an end-fitting that surrounded
the exterior of the link to facilitate joining. The end-fitting was bonded to the
composite link using Hysol EA 9430 rather than being bolted because it provides
a stronger connection and does not diminish the structural integrity as a result of
drilling. Graphite is quite cathodic and as a consequence can cause metals to corrode.
Additionally, it has a very low coefficient of thermal expansion. Therefore the authors
chose to construct the end-fitting out of stainless steel to help prevent corrosion and
reduce thermal stress. The end-fitting was also constructed with a 7 chamfer angle
22
to help reduce the stress-concentration at the edge of the bond.

As for the link discussed in [26], a steel flange was used to facilitate joining of the
composite arm to an electrical discharge wire cutting machine. In this case, the steel
flange was inserted into the arm as illustrated in Figure 1.7. To accommodate bonding
of the flange to the composite tube, the flange was segmented. Accordingly, it used
both an interference fit as well as an adhesive epoxy (IPCO 9923). FEA revealed that
the optimal length of the interference fitting area and adhesive bonding area of the
flange were 32 [mm] and 28 [mm], respectively. To help reduce the hoop stress in the
joint part under vertical loads, six reinforcing plies of 90 prepregs were applied on
the exterior of the composite arm at the joint locations.
Figure 1.7: Schematic of Attachment of Steel Flange to Composite Arm [26]
1.5.3
Results Achieved
Many physical and computational tests have been performed on robotic systems constructed of polymer composites and the more traditional steel and aluminum arms.
The results showed considerable improvement in both dynamic and static properties
23
of the robotic system made from composites.

The composite arm that is described in [25] was designed to have a flexural rigidity
equal to that of an aluminum counterpart. In doing so, it too was designed as a
solid rectangular beam with a length of 812 [mm] and a width of 19.02 [mm], but
offered a reduction in the thickness from 3.14 [mm] to 2.50 [mm]. This equates to a
reduction in the weight of over 50%, from 0.133 [kg] to 0.060 [kg]. With the use of
a proportional and derivative (PD) controller, both arms were tested. It was found
that the composite arm reduced the maximum overshoot from 1.02 to 0.3 , reduced
the energy consumed by the motor used to run the experiment by 40%, and decreased
the maximum deflection at the free end of the link from 3.4 [mm] to 1.7 [mm], all the
while reducing the settling time.
Krishnamurthy et al. [30] performed computational tests on a similar link using an
extended Hamiltons principle to derive the equation of motion and studied the unconstrained and constrained mode of vibration. They determined that the motioninduced tip displacement as well as the torque required for the manoeuvre was considerably less for the links constructed of graphite-epoxy and boron-epoxy composite
than that of aluminum and steel counterparts. They also determined that composite
links are preferred because the magnitude of the control spillover effects is very small,
which makes it easier for designing control systems.
As for the single link carbon-epoxy composite arm created by Lee et al. [26], it reduced
the weight of its cast iron predecessor by 58%, from 26.5 [kg] to 11.2 [kg]. The equallysized arms had close to the same deflection when a 10.2 [kg] payload was applied at the
end of the link. The cast iron arm deflected 38 [m] and the composite arm deflected
40 [m]. While essentially offering the same stiffness with an impressive reduction in
weight, the composite arm also offered improvements in the dynamic characteristics.
Under fixed-free conditions, the composite arm had a fundamental natural frequency
that is 1.41 times higher and a damping ratio that is 1.20 times greater than the cast
24
iron arm. Under free-free conditions, the fundamental natural frequency and damping
ratio of the composite arm was 1.27 and 5.00 times greater than that of the cast iron
arm.
The graphite-epoxy composite link that was created by Lessard et al. [29] also showed
improvements in the natural frequency. The link in question of the 7-DOF manipulator was analyzed by itself as a cantilever arm. With maximum torque applied by
the actuators, a vertical force of 3,790 [N] acting downwards and a horizontal force of
1,037 [N] acting to the right, was applied at the end of the link. This resulted in a
deflection of 0.38 [mm] for the composite link compared to 0.35 [mm] and 0.11 [mm]
for its aluminum and steel counterparts. Although the composite arm had the largest
amount of deflection, it weighed only 418 [g], considerably less than the 729 [g] for the
aluminum arm and the 2,133 [g] for the steel arm of identical dimensions. The vertical
natural frequency of the composite link was around three times greater than that of
the aluminum and steel links. The improvements in the horizontal and extensional
natural frequency were even better, being approximately four times greater.
Caprino and Langella [31] also investigated the fundamental frequency of composite
arms. They created equations for a high modulus graphite-epoxy composite hollow
cylindrical link modeled as a cantilever beam with a concentrated load acting on the
end. They determined that the ratio of the length to the radius, l/R, for the link
had a dramatic effect on the fundamental frequency. For a layering sequence,
they determined that the optimal fiber angle was decreased as the l/R ratio was
increased, indicating that longer beams require a fiber angle close to 0 . For shorter
links the fundamental frequency was maximized with a fiber angle closer to 45 since
shear is the primary force acting on the link. Similarly, for a (0/ 45n ) repeating
pattern, where n is a percentage of 0 layers, a higher l/R ratio requires a lower
percentage of 45 layers for maximum fundamental frequency as shown in Figure
1.8.
25
Figure 1.8: Stacking Angle for Max Fundamental Frequency [31]
Oral and Ider [32] created dynamic equations that were capable of determining the
deviations from the desired path for a three revolute joint robotic arm with filamentwound tapered box-beam links as illustrated in Figure 1.9. A robot with aluminum
links and one with composite links of equal length and weight were tested. The results
indicated that for a translation in one direction requiring rotation of both links, the
deviation in all directions was less for the robot with composite links than it was for
the robot with aluminum links. In fact, the total accuracy was improved by about
four times.
Ghazavi and Gordaninejad [33] also performed computational analysis on a R R k R
robot where R denotes a revolute joint, denotes two joints being perpendicular to
each other, and k denotes two joints being parallel to each other. They studied a
solid graphite-epoxy composite third link that consists of 8 plies with a repeating
layering sequence of [0 /-45 /45 /90 ]. This pattern was chosen to maintain a good
balance between flexural and shear stiffness. The composite arm was approximately
half the weight of the aluminum link of equal dimensions. This led to a 15%, 27%,
and 28% peak torque reduction for the first, second, and third joints, respectively.
26
The composite link also had improved steady-state error of the tip manipulator and
a faster and less oscillatory response. For similar reasons, the composite arm offered
reductions in the forces over its aluminum counterpart. Additionally, the axial and
out-of-plane transverse deflections of the composite link were 47% and 25% less than
that for the aluminum arm.
Figure 1.9: R R k R Robotic Arm [32]
The carbon-epoxy composite end-effector for a SCARA type robot that handles LCD
glass panels as described in Section 1.5.1 showed considerable improvements over its
aluminum predecessor [27]. The composite part weighed only 2.19 [kg], less than half
of the 4.44 [kg] aluminum model. Consequently, the total deflection of the composite
end-effector when loaded with an LCD screen was only 2.52 [mm] whereas the aluminum end effector deflected a total of 4.87 [mm]. In addition to this, the composite
end-effector had better dynamic characteristics than the aluminum counterpart. The
fundamental natural frequency and damping ratio of the composite end-effector were
1.5 and 3.6 times greater than those of the aluminum end-effector.
27
1.5.4
Composite Arms Used in Industry
German company Schubert [34] redesigned their two main-arm pick-and-place robots
to have hollow carbon fiber reinforced plastic (CFRP) links. These links were implemented to replace the old aluminum links in order to meet rising performance
requirements. The links were designed to include a taper such that the cross-section
of the link was larger on the end closer to the base in order to reduce the bending
moment. Laminate inserts and radial stiffeners were also included at the joint location. The link was constructed with a combination of prepregs with fiber directions
of 0 , 45 , and 90 to provide bending, shear, and torsional strength. The new CFRP
links provided a total weight reduction of 25% over the aluminum predecessor. Additionally, the maximum distortion inside the hollow link was measured at 0.25 [mm].
In addition to improving the stiffness of the robot, an improvement in the dynamic
behaviour of 50% was achieved. These improvements lead to the replacement of the
aluminum model in the beginning of 2007.
CBW Automation [35] currently offers a variety of robots that are constructed of
carbon fiber composites. They, however, are quite different compared to the robots
mentioned in the previous sections. Their robots all operate via a gantry system.
Additionally, the robots have very limited mobility, with the majority of them only
having 1-DOF. These robots are used for transporting large heavy parts. The use of
carbon fiber composite materials over traditional materials has lead to significantly
shorter cycle times.
In 2005, Kuka [36] began selling the first commercially available large-scale robotic arm
that features the use of composite materials. The third link of the arm is constructed
of carbon fiber composite. The arm is capable of stacking loads weighing up to 100
[kg] to heights of up to 3 [m], at rates as high as 600 palletizing cycles per hour.
The robotic arm weighs 1,200 [kg] compared to a similar traditional Kuka robotic
arm that weighs 1,500 [kg]. This arm however has a reach of only 3.2 [m] whereas the
28
traditional arm has a reach of 3.5 [m], however, the weight reduction indicates good
improvement over its traditional counterpart [1].
Figure 1.10: Kuka Robotic Arm Featuring Composite Third Link [1]
1.5.5
Opportunities
Although much research has been performed on composite robotic arms and a few
commercial arms featuring the use of composite materials are available, an opportunity exists to investigate novel cross-sectional geometries to improve the performance of composite robotic arms. Most composite robot arms feature hollow, tubular
structures. Although these structures are strong, their strength could potentially be
improved by adding an internal support structure.
1.6
Summary of Contents
The contents of the remainder of this thesis are as follows. In Chapter 2, the configuration of the industrial-sized 6-DOF robotic arm is selected. The cross-section
of the link is studied using both FEA and optimization. Using the information and
29
results obtained from these preliminary investigations, the links for a main-arm proofof-concept prototype are designed and developed as outlined in Chapter 3. Chapter
4 discusses the different tests that were performed on the proof-of-concept prototype
and outlines its performance. A FEA technique for modeling the proof-of-concept
prototype link as an anisotropic material is also presented. The results from the tests
of the physical and virtual prototype are also compared. The results of Chapter 4
are used to design the full-scale industrial-sized robotic arm in Chapter 5. The thesis is concluded with Chapter 6 and suggestions for future work required before the
industrial-sized arm can be manufactured.
30
Chapter 2
Link Design
This chapter outlines the preliminary design decisions for the construction of the
polymer composite robotic arm. The main configuration for the 6-DOF robotic arm
was selected. An optimal layout for the robotic arm was also selected. FEA was
performed on a variety of links with various cross-sections. An ideal cross-section was
selected based on minimizing the deflection relative to the weight of the link 1 . Finally,
the dimensions of the link for a 2-DOF proof-of-concept prototype were optimized for
minimal deflection with the use of MATLAB.
2.1
2.1.1
Manipulator Layout Selection

Comparison of Various Types of Robotic Arms
Before beginning to design the robotic arm to be used for large-scale applications,
the type synthesis of the robotic arm must be performed. The robotic arm can
either be serial, parallel, or a hybrid of the two. As illustrated in Figure 2.1, a serial
configuration is one in which the links are connected to each other in series, one after
another, forming an open loop mechanism. A robotic arm with a parallel configuration
1
Please see [37] for paper on this topic
31
consists of a mobile platform that is connected to the base by at least two independent
kinematic chains that form a closed loop. A hybrid robot is a combination of the two
configurations. Consequently, a hybrid robot can either have a serial configuration at
its base and a parallel base connected afterwards, or have a parallel configuration at
its base and conclude with a serial configuration.
Figure 2.1: Schematics of Different Types of Main Configurations for Robotic Arms
32
Serial and parallel configurations result in robotic arms that have many varying characteristics. A serial robotic arm is distinguished by having a large workspace. A
robotic arm with a serial configuration can reach over a significantly larger area and
or volume than a robotic arm that has the same size links using a parallel configuration. In addition to this, parallel robotic arms generally cannot reach around obstacles
like their serial counterparts can.
A robotic arm with a serial configuration is considerably simpler than one with a
parallel configuration. Since one link provides one motion rather than a group of
links it is less complex. Consequently, a robotic arm with a serial configuration is
relatively easy to design and manufacture.
For parallel robots, since multiple limbs are required, they are significantly more
complex than serial robotic arms. As a result, a parallel robotic arm costs a lot more
to produce than a serial one. Another result of the increased complexity of a parallel
robotic arm is that in comparison to a serial robotic arm it is difficult to control and
operate. The kinematics of a parallel robotic arm are more complex than that of
a serial arm which makes the algorithms used to provide motion for parallel arms
considerably more complicated than those needed for serial robotic arms.
Robotic arms with a parallel configuration have a high stiffness since the end-effector is
supported at several locations (whereas for a serial robotic arm it is only supported at
one location). Furthermore, since many links are connected in a closed chain manner,
the errors and displacements are averaged over the number of links. This is not the
case for a serial robotic arm, where the errors are additive being the sum of all axis
errors of the links. Consequently, a parallel configuration results in a robotic arm with
greater accuracy than a serial configuration. A parallel configuration usually results
in a robotic arm with a lower weight than a serial robotic arm not only because the
improved accuracy allows for the possibility of using smaller links, but in the parallel
configuration the motors can be mounted on the base. As a result, the links can be
33
designed to be smaller and thus weigh less since they do not have to support the
weight of the motors. However, a robotic arm with a parallel configuration is known
to lose stiffness in singular positions. In fact, this loss can be so severe that it causes
it to shake or even become free to move despite all the joints being locked. Therefore,
special care needs to be taken to avoid such scenarios.
A hybrid robotic arm could be very beneficial for this application because it allows for
the possibility of constructing an arm that has a large workspace and high stiffness
while keeping the weight to a minimum. However, the complexity of a hybrid results
in a substantial increase in the cost of production of the arm and the control of the
system is more difficult.
Based on the above, a serial layout was selected due to its inherent simplicity in terms
of control and construction.
2.1.2
Serial Arm Configuration
There are numerous combinations of six joints that could be used to achieve 6-DOF
motion. However, a kinematically-simple manipulator layout is ideal because it incorporates a spherical group of joints at the wrist connected to the end of a main-arm
comprised of three successfully parallel or perpendicular joints with no unnecessary
offsets or link lengths between joints [38]. As a result, it is easier to develop the
kinematics of the arm for this type of layout opposed to one that is not kinematically
simple. The use of a kinematically-simple 6-DOF robotic arm reduces the difficulty
involved in controlling and operating the arm. Additionally, the simplicity of the
layout makes it easier to design and manufacture. Since unnecessary offsets and link
lengths are avoided for this type of layout, the amount of material for the robotic
arm is kept to a minimum. Consequently, this helps to improve the payload-to-weight
ratio.
34
There are four groups of kinematically-simple main-arm layouts, those with three
revolute joints, two revolute joints and one prismatic joint, two prismatic joints and
one revolute joint, and three prismatic joints. Prismatic joints add significant weight
to the robot over revolute joints and thus to design a robotic arm with the highest
payload-to-weight ratio, a layout with three revolute joints is to be utilized. There
are five different kinematically-simple main-arm layouts that are comprised of three
revolute joints. The R R k R manipulator layout as shown in Figure 2.2 was
chosen.
The R R k R layout offers the largest spherical workspace of the five layouts with
three revolute joints. Since the robotic arm that is being designed is to operate on large
surfaces with highly complex curvature, this layout allows the end-effector to reach
the greatest amount of surface area without moving the entire robotic manipulator.
In order for a robotic arm with one of the other four joint main-arm layouts to be
able to reach an equivalent amount of surface area, the links must be longer.
The R R k R layout is also preferred because the two main links (links A and B
in Figure 2.2) of the robotic arm work in the same plane. This is beneficial since it
helps to keep the torsion to a minimum. It has been shown that for a SCARA robot
the amount of deflection experienced by the end-effector is greatest when there is a
combination of both bending and torsion [39]. Consequently, this feature helps to
minimize the end-effector deflection.
Since the R R k R layout has the second and third joints parallel, it offers the
possibility of the motors of the main arm being mounted as close as possible to the
base by utilizing a simple drive-train for movement of the third joint. By reducing
the amount of weight that is located between the two main links both the amount of
stress and displacement in the links is reduced and consequently high accuracy and
repeatability can be achieved while using less material.
35
Figure 2.2: 6-DOF Serial Configuration of the Arm
2.2
Analysis of the Cross-Section of the Two Main

Links
2.2.1
Cross-Section Testing Procedure
As defined by the specifications, the industrial robotic arm is to have a total reach of
5 [m]. The two main links are the links that occur after the second and third joint as
illustrated in Figure 2.2. Since the two main links supply the majority of the reach
of the kinematically-simple 6-DOF robotic arm, for the purpose of this preliminary
study the two main links are defined as having a total length of 5 [m]. The analysis
was performed under the worst possible scenario, fully extended at its maximum
reach. In this position the two main links of the arm are parallel and thus they can
be analyzed as a single straight member with a constant cross-sectional geometry.
For the purposes of testing the different cross-sectional geometries, the strength and
stiffness of the member representing the two main links was examined.
In keeping with achieving a reduction in the overall weight of the arm, it is imperative
that the links be designed such that the amount of material is kept to a minimum. In
order to do so, the two main links of the robotic arm should be designed as thin-walled
36
members. To satisfy the thin-walled beam criterion, the thickness must be at least 10
times smaller than any characteristic dimension of the cross-section, in this case the
radius. The radius must also not be more than one tenth the size of the length [40].
Therefore, a thickness of 10 [mm] and a radius of 127 [mm] (5 [in]) were chosen. Since
the reach of the full-scale robotic arm is 5 [m], then a length of 2.5 [m] for each of the
main links is long enough that a radius of 127 [mm] (5 [in]) satisfies the rules for a
thin-walled beam.
For this preliminary analysis, only the effect of using different cross-sectional geometries on the stiffness of the link is of concern. Therefore, the aforementioned dimensions for the thickness of the walls, diameter of the exterior profile, and the length of
the links are held constant.
The links of the robotic arm are constructed of high-modulus carbon fiber-epoxy
composite with a 60% fiber volume. It was chosen because it offers the greatest
stiffness due to its high tensile modulus and low ultimate tensile strain as indicated in
Table 1.2 of Section 1.4.3.2. However, for this analyis, it was modeled as an isotropic
material with a density of 1,700 [kg/m3 ], modulus of elasticity of 220 [GPa], and
Poissons ratio of 0.25 [5]. Since the purpose of this study was solely to compare
different cross-sectional geometries, modeling it as a simplified isotropic material still
provides sufficient information on how varying the cross-sectional shape reduces the
amount of deflection the end-effector experiences.
In order to properly analyze the stiffness and strength of the links with various crosssections, a force that represented the payload of the robotic arm, as well as the weight
of the arm itself was applied. A distributive force acting downwards that represented
the weight of the two main links was applied evenly along the length of the single
straight member. Additionally, a force F measured in Newtons (N) acting in the
same direction was applied at the end of the arm. It was calculated using:
37
F = (mpayload + mwrist )g
(2.1)
where mpayload is the mass of the payload, mwrist is the mass of the end-effector, links,
and motors that make up the group of spherical joints at the wrist (both measured
in [kg]), and g is the acceleration due to gravity, 9.81 [m/s2 ]. By using the specified
payload of 50 [kg] and 30 [kg] as the assumed weight of the wrist components of the
robotic arm, the force that was applied at the end of the member was found to be
784.8 [N].
This force was applied on the last 200 [mm] of the upper portion of the arm as
demonstrated in Figure 2.3. The force was applied on this area as it sufficiently
represents the area of the link that would be supporting the wrist components and
thus the load that is being held by the gripper. The other end of the arm was fixed in
all directions such that it offered 0-DOF so as to represent the arm being locked in the
fully extended position in order to determine the effectiveness of the cross-sectional
geometry in terms of stiffness and strength when the bending force is at its maximum.
Figure 2.3: Force Applied on Single Straight Member Representing Two Main Links
38
For the purpose of analyzing the various cross-sectional geometries, FEA was conducted using UGS NX 5.0 with NX Nastran as the solver by applying the aforementioned forces. The thin-walled beam representing the robotic arm fully extended
was analyzed using the Structural analysis type. Finally, the solution type chosen
for the FEA procedure was SESTATIC 101 - single constraint. The robotic arm was
represented as a mesh of 3D 10-noded tetrahedral elements in order to perform the
FEA.
2.2.2
Results of Cross-Section Analysis for the Main Links
Different families of cross-sections for the two main links were tested using the procedure described in the previous section. The different families of cross-sections and the
stiffness and strength of the single straight member which represents the two main
links extended at their maximum length are discussed below.
2.2.2.1
Basic Shaped Hollow Tubes
The initial types of link geometry that were tested were hollow. Three simple basic
shapes were chosen for the cross-section: circular, square, and rectangular (as shown
in Figure 2.4). To effectively compare the three different cross-sectional geometries,
the dimensions of the tubes were chosen such that their perimeters were nearly equal,
which results in a similar volume and mass. As a result, using a constant wall thickness
of 10 [mm], the two main links of the robotic arm have a similar volume and thus
mass for the different geometries of this first family of cross-sections. As previously
mentioned, the diameter of the circular tube representing the two main links fully
extended was chosen to be 254 [mm] (10 [in]). This equates to a perimeter of 797.96
[mm]. Therefore, the dimensions of the other two tubes were chosen such that their
perimeters were equal to 798 [mm]. Consequently, the square tube that was analyzed
has four sides that are 199.5 [mm] long. Since the serial robot layout that was chosen
39
ensures that the main two links always work in a vertical frame, using links with a
larger height than width ensures that it provides greater support for the payload that
is acting downwards due to the force of gravity. In order to provide more support, the
tube with the rectangular cross-section was designed such that its height was equal
to two times its width. Therefore, the height and width of the rectangular tube were
266 [mm] and 133 [mm], respectively.
Figure 2.4: Cross-Sections of Family of Basic Shaped Hollow Tubes

The procedure that is described in the previous section was used to determine the
maximum amount of displacement and stress in the member representing the two
main links when fully extended for the first family of cross-sectional geometries. It
was determined that the maximum displacement and stress in the hollow tube with
the circular cross-section were 3.333 [mm] and 13.72 [MPa]. For the square hollow
tube, the same test indicated that the maximum displacement and stress experienced
in the member were 4.169 [mm] and 14.31 [MPa]. As expected, the hollow tube with
the rectangular cross-section showed improvements over the square tube having a
maximum displacement of 2.775 [mm] and a maximum stress of 12.18 [MPa]. These
results indicated that the rectangular cross-section is favoured over square and circular
as it produces the stiffest and strongest links for these loading conditions. However,
due to the curvature of the circular tube it would present fewer problems during link
manufacture if using a filament winding machine. Therefore, alterations of both the
rectangular tube and circular tube were investigated.
40
2.2.2.2
Circular Tubes with Support Ribs
The hollow circular tube was modified by adding support ribs that initiated in the
middle of the tube propagating outwards until they connected to the inner wall of
the tube. Both an X-support tube and Tri-support tube were tested. The X-support
tube has four equally spaced ribs (that form an X) and was tested in two different
orientations, while the Tri-support tube has three support ribs and was tested in its
three primary orientations as shown in Figure 2.5.
Figure 2.5: Cross-Sections of Family of Circular Tubes with Support Ribs

For the X-supported tube it was found that the maximum displacement and stress
were 3.222 [mm] and 15.02 [MPa] for when the tube was in the orientation illustrated
in Design 4 of Figure 2.5 such that the force of the payload was acting at an equal angle
to all the ribs. The tests indicated that the results were the same for the member
when it was oriented such that the force of the payload was acting in a direction
parallel to two ribs as demonstrated in Design 5.
As for the Tri-supported circular tube, the maximum displacement and stress in the
member while oriented such that the force of the payload was acting in a direction
perpendicular to one of the ribs as shown in Design 6 were determined to be 3.254 [mm]
and 13.40 [MPa]. However, with the Tri-supported circular tube in an orientation such
that the force of the payload was acting in a direction parallel to one of the ribs (in
either of the orientations illustrated in Design 7 and 8) the maximum displacement
and stress were found to be 3.303 [mm] and 15.14 [MPa].
Despite the fact that the addition of four ribs (X-supported) rather than three (Tri41
supported) increases the weight of the member, it underwent a lesser amount of displacement as illustrated in Figure 2.9. Additionally, the percentage decrease in the
displacement versus the percentage increase in the mass of the two main links in comparison to the hollow circular tube was greater for the X-supported circular tube than
the Tri-supported one. This indicates that the use of four support ribs is favoured over
three as its results in a reduction in the displacement that is greater in comparison
to the increase in the total mass. However, even for the X-supported circular tube
the decrease in the maximum displacement was small in comparison to the increase
in the mass of the two main links. This small decrease in the maximum displacement
does not justify the significant increase in the mass (the percentage increase in the
mass of the two main links was more than 10 times the percentage decrease in the
deflection). Additionally, the support ribs further complicate the production as a
different process needs to be used to produce the support structure since it cannot be
produced simultaneously with the outer tube using the filament winding process. All
in all, the slight improvements in the stiffness that the addition of the support ribs
provide does not justify their use with an outer circular cross-section.
2.2.2.3
Circular Tube with Different Shaped Inner Tubes
The cross-section of the circular tube was further modified by adding an inner tube to
the support system such that the support ribs connect the exterior of the inner tube
to the interior of the exterior tube. The straight member representing the two main
links fully extended was tested with a circular, rectangular, square, and diamond inner
tube as illustrated in Figure 2.6. The inner circular tube that was tested had an inner
diameter of 152.4 [mm] (6 [in]). The other shapes of the inner tubes (rectangular,
square and diamond) were designed with a perimeter equal to that of the 152.4 [mm]
diameter inner circular tube. The weight of the link varies significantly for the different
inner tubes (as can be seen in Figure 2.9) because the ribs required to connect the
42
inner tube with the outer circular tube vary significantly in length for the differently
shaped inner tubes.
Figure 2.6: Cross-Sections of Family of Circular Tube with Different Shaped Inner
Tubes
For the circular inner tube with three support ribs, the maximum displacement and
stress in the member was found to be 2.900 [mm] and 12.74 [MPa]. As expected, the
use of four support ribs for the inner circle was found to be an improvement over three
ribs, resulted in a maximum displacement and stress of 2.876 [mm] and 12.39 [MPa].
Therefore, the other shaped inner tubes were connected to the outer circular tube by
four supporting ribs.
For the inner square tube cross-section, the maximum displacement and stress were
found to be 2.984 [mm] and 12.71 [MPa]. For the inner diamond tube as illustrated in
Design 12, the maximum displacement and stress were determined to be 2.933 [mm]
and 12.99 [MPa]. Similar to the case for the basic shapes, the rectangular inner tube
was found to reduce the displacement the most. In fact, the addition of the rectangular
inner tube reduced the displacement from 3.333 [mm] for the hollow circular tube to
2.701 [mm], as well it reduced the maximum stress from 13.72 [MPa] to 11.75 [MPa].
However, this design nearly doubled the mass of the two main links, increasing it
from 65.16 [kg] to 126.55 [kg]. This means that the percentage increase in mass was
around 5 times the percentage decrease in displacement. Therefore, the addition of
the inner tube to the support system results in an improvement of the displacement
with respect to the increase in mass by more than two times over the family of circular
tubes with support ribs discussed above. However, it makes the production process
43
considerably more complicated. The filament winding process needs to be performed

twice, and the support system needs to be connected to both the inner and outer
tubes (either by hand lay-up or some mechanical process; either way this drastically
increases the time and cost of production).
2.2.2.4
Annulus with Different Sized Inner Tubes
Another modification to the circular tube that was investigated was an annulus (a
circular tube within a circular tube) with a varying diameter for the inner tube. As
was the case for the other investigations, this design was tested using both three and
four support ribs. The annulus was tested with inner diameters of 63.5 [mm] (2.5
[in]), 101.6 [mm] (4 [in]), 152.4 [mm] (6 [in]), and 203.2 [mm] (8 [in]) as illustrated in
Figure 2.7 (shown with four support ribs).
Figure 2.7: Cross-Sections of Family of Annulus with Different Sized Inner Tubes
For an inner diameter of 63.5 [mm], the Tri-support Annulus had a maximum displacement and stress of 3.358 [mm] and 14.38 [MPa], whereas the X-support Annulus
was found to lower the maximum displacement and stress to 3.266 [mm] and 13.96
[MPa]. For the 101.6 [mm] inner diameter annulus, the maximum displacement and
stress for the Tri-support and X-support Annulus were 3.234 [mm], 3.181 [mm], 14.17
[MPa], and 13.68 [MPa], respectively. The results for the annulus with an inner diameter of 152.4 [mm] for both three and four support ribs are mentioned above in
Section 2.2.2.1. For an annulus with an inner diameter of 203.2 [mm], the maximum
displacement and stress for the Tri-support and X-support Annulus were determined
44
to be 2.405 [mm], 2.387 [mm], 10.78 [MPa], and 10.62 [MPa], respectively.
As was the case for the previous tests, the X-support proved to provide better stiffness
and strength than the Tri-support. However, as the diameter of the inner tube was
increased the improvements that the Tri-support Annulus provides grew closer to
those of the X-support. This makes sense because as the inner tube grows the ribs
become shorter and thus the amount of support that the ribs provide is lowered.
Associated with this, is the fact that the mass of the member for the Tri-supported
Annulus also gets closer to that of the X-supported Annulus as the size of the inner
diameter is increased (as demonstrated in Figure 2.9).
The results of this study indicated that as the size of the inner tube increased, the
strength and stiffness of the member were also improved. In fact, the degree of
decrease in displacement with respect to increase in mass increased significantly as
the inner diameter grew as illustrated in Figure 2.9. It continues to increase until the
inner diameter has grown so much that it is now a hollow tube with a thickness of 20
[mm] (double that of the original case). As previously mentioned, the production of
the two main links with an annulus design is more complex and thus it is not worth
using when even better results can be achieved with a simple hollow tube.
2.2.2.5
Rectangular Tubes with Support Ribs
The rectangular tube was modified by adding support ribs as illustrated in Figure
2.8. For Design 18, a vertical support in the middle of the rectangle was used. For
this I-beam supported rectangular tube, the maximum displacement and stress in the
member were determined to be 2.526 [mm] and 11.67 [MPa]. The next support that
was tested was a T-support as shown in Design 19 of Figure 2.8. The rectangular
tube with this type of support had a maximum displacement and stress of 2.601
[mm] and 12.14 [MPa]. The I-beam support proved to provide more stiffening and
strengthening than the T-support because the addition of the horizontal support rib
45
provided minimal restraint to the force of the payload acting downwards in the vertical
direction. As a result, the addition of the mass of the horizontal rib increased the
displacement and stress within the member.
Figure 2.8: Cross-Sections of Family of Rectangular Tubes with Support Ribs

The final support tested for the rectangular tube was an X-support (Design 20). This
support system resulted in a maximum displacement and stress of 2.346 [mm] and
10.67 [MPa]. These results indicate that the I-beam support and X-support designs
are considerably better than the T-support. Figure 2.9 shows that these two designs
offer a percentage reduction of the displacement that is between a third and a quarter
of the percentage increase in the mass. Further inspection of Figure 2.9 indicates
that the percentage decrease in the displacement in comparison to the percentage
increase in the mass that the I-beam support provides is better than that of the
X-support. The X-support provides a greater amount of torsional rigidity than the
I-beam support and even though the main links operate in a vertical plane, the center
of gravity of the end-effector may be offset horizontally thus resulting in torsion in the
two main links in which the X-supported rectangular tube is better suited for. The
slight improvements in the stiffness and strength that an inner support tube provides
does not warrant their use since it makes the production of the links more difficult.
46
Figure 2.9: Displacement Versus Mass for Different Cross-Sections
2.2.3
Analysis of Tapering the Main Links
Observation of Figure 2.9 indicates that the I-beam supported and X-supported rectangular tubes (Designs 18 and 20) provide the greatest amount of stiffness relative to
the mass of the two main links making them the optimal cross-section geometries for
these loading conditions. Therefore, these two link designs were tested by tapering
the link in the direction of its length. The dimensions of the cross-section at the end
of the member representing the two main links were held constant but the dimensions
of the cross-section at the beginning of the member were larger according to the taper
angle used as demonstrated in Figure 2.10. For this analysis, the member representing
the two main links was tapered with an angle between zero and one degree. Since
the links are considerably long, a taper angle of just one degree results in the larger
end having a height of 440.55 [mm] and width of 220.27 [mm] compared to 266 [mm]
and 133 [mm] for the end of the member. This results in an increase in the mass by
47
over one third of its original weight and thus using any taper angle beyond this is
excessive.
Figure 2.10: Tapered Member Representing the Two Main Links

The member with the two different cross-sectional geometries was analyzed using
taper angles of 0.05 , 0.10 , 0.15 , 0.25 , 0.50 , and 1.00 . Table 2.1 lists the results
for the maximum displacement and stress within the member at the different angles.
The amount of displacement and stress decreases as the taper angle is increased.
Table 2.1: Effect of Taper Angle on the Displacement and Stress
Taper Angle
0.00
0.05
0.10
0.15
0.25
0.50
1.00
I-beam Supported
Displacement
2.526 mm
2.343 mm
2.176 mm
2.030 mm
1.773 mm
1.305 mm
0.782 mm
Rectangular Tube
Stress
11.67 MPa
10.60 MPa
9.959 MPa
9.409 MPa
8.736 MPa
6.697 MPa
4.419 MPa
X-Supported
Displacement
2.346 mm
2.173 mm
2.018 mm
1.878 mm
1.638 mm
1.201 mm
0.717 mm
Rectangular Tube
Stress
10.67 MPa
9.971 MPa
9.685 MPa
9.107 MPa
8.197 MPa
6.379 MPa
4.195 MPa
Figure 2.11 shows how the taper angle affects both the maximum displacement and the
mass of the two main links. The results indicate that the amount of reduction in the
displacement is considerable when compared to the increase in mass. It shows that as
the taper angle is increased, the percentage decrease in the displacement relative to the
48
percentage increase in the mass is reduced. Therefore, the stiffening and strengthening
advantages of the taper angle is better realized at smaller taper angles. Lastly, the
rate of reduction in the displacement in comparison to the increase in the mass is
very similar for both the I-beam supported and X-supported rectangular tubes. As
a result, neither cross-section can be chosen over the other without further analysis
being performed on them that includes a torsional load due to the end-effector.
Figure 2.11: End-Effector Displacement Versus Mass for Various Taper Angles
2.3
Optimization of the Cross-Section
The accuracy of the robotic arm that is to be created is of great concern. As previously
mentioned, current industrial-sized robotic arms are made of steel or aluminum to
ensure high stiffness with the end result of providing good accuracy and repeatability.
Therefore, in creating a robotic arm constructed of polymer composites, it is essential
49
to ensure that the stiffness-to-weight ratio is as high as possible. Taking this into
consideration, along with the fact that the length of the robotic arm has already
been specified and a manipulator layout has already been chosen that provides a
large workspace, the stiffness of the robotic arm will be minimized with the use of an
optimization procedure. The best way to achieve this is by altering the cross-section
dimensions.
2.3.1
Determination of the Objective Function
In order to ensure that the optimization methodology arrives at a suitable solution for
the dimensions of both the I-beam and X-supported rectangular cross-sections, the
robotic arm needs to be analyzed under the worst possible loading scenario. Therefore,
the objective function and constraints for this optimization problem were developed
for the robotic arm with its two main links parallel, extending outwards fully, and the
wrist configuration extended out fully to either side such that it was perpendicular
to the main links as demonstrated in Figure 2.12. Additionally, the payload, W , that
is to be supported by the end-effector is applied at the end of the wrist such that in
one direction it is the length of the two main links, L, away from the support, and
in the other it is the length of the wrist configuration, Lwrist , away from the support.
Consequently, the inclusion of both torsion and bending must be accounted for in
developing the optimization methodology.
Figure 2.12: Configuration of Robotic Arm to be Analyzed for Cross-Section Optimization

50
In order to determine the objective function to minimize the deflection, the equation
for the deflection experienced by the end-effector must be developed. With the robotic
arm being analyzed in the position demonstrated in Figure 2.12, deflection occurs due
to both bending and torsion. In fact, there is deflection associated with the bending
of the two main links, bending of the wrist configuration, and twisting of the two main
links. As a result, the objective function has to be defined in terms of these three
components of deflection. Consequently, the concept of superposition was utilized to
create the objective function.
In order to successfully analyze all the forces acting on the arm, it is first necessary
to provide a full overview of the robotic arm. Since the arm being analyzed has two
main links, there will be a motor controlling the second main link that is located
between the two main links. Consequently, the weight of this motor will contribute to
the deflection of the robotic arm. Furthermore, it is assumed that the two main links
are of equal length and thus this force is acting in the middle of the total length of
the two main links. Additionally, the weight of the links themselves also contributes
to the deflection. Lastly, the payload that is applied at the end-effector (at the end
of the two main links) also contributes to the amount of deflection as is illustrated in
Figure 2.13.
Figure 2.13: Side View of Free Body Diagram of the Analyzed Robotic Arm
51
The deflection of the robotic arm due solely to the bending of the two main links is
first analyzed. From observing the free body diagram of the robotic arm as shown in
Figure 2.13, the situation of bending of the two main links is demonstrated in Figure
2.14. Figure 2.14 illustrates that the distributive load of the weight of the links, and
concentrated load of the weight of the motor and the payload all contribute to the
deflection of the arm. As a result, the deflection due to the bending of the two main
links, Ymainlinks , is a result of the sum of the three forces acting on the two main
links and is denoted by [41]:
Ymain links =
gwlink L4 5gmmotor L3 gW L3
+
+
8EI
48EI
3EI
(2.2)
where E is the elastic modulus, I is the moment of inertia, L is the total length of the
two links, g is the acceleration due to gravity, W is the weight of the payload, mmotor
is the weight of the motor and wlink is the weight of the link as denoted by:
wlink =
Ac L
L
(2.3)
Where Ac is the cross-sectional area and is the density of the composite material.
Figure 2.14: Free Body Diagram for Bending of the Two Main Links
The deflection of the robotic arm due to the bending of the wrist configuration,
Ywrist , is shown in Figure 2.15. It was assumed that the length of the wrist configuration was only one tenth of the length of the two main links. Consequently, the
52
weight of the links of the wrist configuration was extremely small relative to the weight
of the payload and thus it was assumed to be negligible. Therefore, the payload is
the only force acting on the wrist configuration and thus the only force contributing
to the bending of the wrist. Thus, the deflection due to the bending of the wrist is:
Ywrist =
gW L3wrist
3EI
(2.4)
Where Lwrist is the length of the wrist configuration.
Figure 2.15: Free Body Diagram for Bending of the Wrist Configuration
The deflection of the robotic arm due to the twisting of the main links, Ytwist ,
is a result of the torque that is generated due to the payload being applied at a
perpendicular distance to the main links. This is due to the fact that the wrist
configuration is oriented such that it is stretched outwards to the side, perpendicular
to the main links. The resulting torque acting on the main links is represented in
Figure 2.16. The applied torque results in the main links twisting by an angle as
demonstrated in Figure 2.17. Using basic trigonometry, the assumption of a small
angle , and the diagram shown in Figure 2.18, the equation for the displacement due
to the twisting of the two main links is determined as:
Ytwist = Lwrist sin() cos()

where is the angle of twist.
53
(2.5)
Figure 2.16: Free Body Diagram for Torque Applied on the Main Links
Figure 2.17: Resulting Deflection of the Wrist Configuration due to Twisting of the
Main Links
Figure 2.18: Geometric Representation of Deflection due to Twisting of the Main

Links
54
The total deflection of the robotic arm is thus equal to the sum of the deflections that
result due to bending of the main links and the wrist configuration, as well as twisting
of the main links. Therefore the objective function is the total deflection () of the
robotic arm:
= Ymain links + Ytwist + Ywrist
2.3.2
(2.6)
Design of the Proof-of-Concept Prototype
The optimization methodology is carried out on a proof-of-concept prototype of the

industrial-sized robotic arm that is to be designed. As of this point, there are too many
variables concerning the design of the industrial-sized robotic arm (mostly surrounding
the motors), and thus the optimization methodology was initially developed for the
proof-of-concept prototype.
The prototype is to be powered using PowerCube rotary motors by Schunk [42]. A
PR110 motor moves the first main link and the PR090 motor moves the second main
link. Using a mass of 3.2 [kg] for the PR090 and a maximum output torque of 206
[Nm] and 425 [Nm] for the PR090 and the PR110, respectively, the ideal total length
of the two main links was determined. It was assumed that the connection device for
attaching the PR090 motor to the two main links has a mass of 1.5 [kg]. Using an
efficiency of 90% for the maximum output torque of the motors and the free body
diagram of the main links with the torque of the motors and the force of the masses
being applied on the links as illustrated in Figure 2.19, the maximum payload was
determined.
It was determined that the maximum output torque of the PR110 was the limiting
factor over that of the PR090 when the two main links are of equal length and thus
the maximum payload, W , is determined using the maximum output torque of the
PR110 motor and is calculated as:
55
W =
ef f Mmax
Lgmmotor
2
L2 gAc
2
Lg
(2.7)
where Mmax is the maximum output torque of the PR110 motor and ef f is the
efficiency at which the motor operates at. In determining W , the X-supported rectangular tube cross-section was used since for an equal perimeter it is heavier than the
I-beam supported rectangular tube. Additionally, the dimensions that were used in
performing the FEA study in Section 2.2 were used.
Figure 2.19: Free Body Diagram for Maximum Output Torque of PR110 Motor Acting
on Main Links
The maximum payload was calculated at various lengths for the two main links using
dimensions for the cross-section that were proportional to the length of the two main
links. The maximum payload was compared to the payload of the robotic arm that
was proportional to the ratio of the size of the prototype to the industrial-sized arm
and it was determined that the ideal length for the two main links was 1.75 [m], 35%
the size of the industrial sized robotic arm, as shown in Table 2.2.
For a prototype that is 35% the size of the industrial sized robotic arm, the total
length of the main links and the payload are 1.75 [m] and 17.5 [kg], respectively.
Additionally, the length of the wrist was 0.175 [m] using the assumption that the
wrist configuration is one tenth the length of the main links. Therefore, with the
knowledge that the mass of the motor is equal to the mass of the PR090 motor and
56
the connection device, 4.7 [kg], it is clear that the objective function for deflection,
Equation (2.6), is in terms of the area (Ac ), moment of inertia (I), and angle of
twist (). Since the I-beam and X-supported rectangular tubes have different crosssectional geometries, it is necessary to determine the area and moment of inertia for
both cross-sections.
Table 2.2: Comparison of Length of Two Main Links to Prototype Payload
Length of Two Max Payload at 90% Max Payload as % of
Main Links
Max Output Torque
Intended Payload
1.5
21.67
1.44
1.55
20.77
1.34
1.6
19.92
1.24
1.65
19.11
1.16
1.7
18.35
1.08
1.75
17.63
1.01
1.8
16.95
0.94
1.85
16.30
0.88
1.9
15.68
0.83
1.95
15.09
0.77
2.0
14.52
0.73
2.3.3
Formulation of the Equations for the Cross-Sectional

Area
The equations for the cross-sectional area for both geometries are determined using
the concept of combination of composite parts. The I-beam supported rectangular
tube can be considered as a large rectangle minus two smaller rectangles of equal
dimension as shown in Figure 2.20. Using this approach the equation for the area for
the I-beam supported rectangular tube is determined as:
Ac = 2wt + 3ht 6t2

where h, w, and t represent the height, width, and wall thickness of the link.
57
(2.8)
Figure 2.20: Cross-Sectional Area of the I-Beam Supported Rectangular Tube

The same procedure was utilized for determining the cross-sectional area for the Xsupported rectangular tube. Since the thickness of all the walls and interior support
features is defined to be the same thickness (t), the area of the different components
comprising the X-supported rectangular tube involve some calculations. Figure 2.21
illustrates the cross-section and the distances that need to be calculated before the
dimensions of the basic components can be determined.
Figure 2.22 shows a close up view of the areas of concern. Using basic trigonometry,
the equations for the distances of xh , yh , xv , and yv were determined as:
xh =
t
2 cos tan1
w
h
(2.9)

t cos tan1 wh

yh =
cos tan1 wh tan1
w
h
(2.10)

t sin tan1 wh

xv =
cos tan1 wh tan1
w
h
(2.11)
yv =
t
2 cos tan1
58
h
w
(2.12)
Figure 2.21: Cross-Section of the X-Supported Rectangular Tube
Figure 2.22: Middle and Corner of Cross-Section for the X-Supported Rectangular
Tube
59
The X-supported rectangular tube can be considered as a rectangle minus two horizontal triangles minus two vertical triangles as demonstrated in Figure 2.23. Therefore
the cross-sectional area for the X-supported rectangular tube is the area of the rectangle minus two times the area of the horizontal triangle minus two times the area of the
vertical triangle. Thus the equation for the cross-sectional area for the X-supported
rectangular tube is:
w
h
Ac = wh (w 2t 2xh )( t yh ) (h 2t 2yv )( t xv )
2
2
(2.13)
Figure 2.23: Cross-Sectional Area of the X-Supported Rectangular Tube
2.3.4
Formulation of the Equations for Moment of Inertia
The equations for the moment of inertia for the two different cross-sectional geometries
can also be determined with the use of the combination of composite parts. By reasons
of symmetry, the neutral axis for the bending of the main links about the X-axis is
located at the midheight of the links cross-section for both cross-sectional geometries
as demonstrated in Figure 2.24. By using the equations for the area of the different
base components of the cross-section that were determined in the previous section,
the moment of inertia for those base components was determined.
60
Figure 2.24: Neutral Axis for the Cross-Sectional Geometries of Concern

For the I-beam supported rectangular link, the moment of inertia for the cross-section
is determined by subtracting two times the value of the moment of inertia of the
small rectangle from the moment of inertia of the large rectangle as is shown in
Figure 2.25. As a result of the symmetry of the base components of the I-beam
supported rectangular cross-section, the neutral axis of the large and small rectangles
is coincident with the neutral axis of the entire cross-section. Consequently, the
moment of inertia for this cross-sectional geometry is:
1
1
I = wh3
12
6
w 3t
2
(h 2t)3
(2.14)
Figure 2.25: Moment of Inertia for the I-beam Supported Rectangle Cross-Sectional
Geometry
61
The same procedure that was utilized to determine the moment of inertia for the Ibeam supported rectangular link was used to determine the moment of inertia for the
X-supported rectangular link. The equations for the dimensions of the different base
components that make up the X-supported rectangular link were used to determine
the equation for its moment of inertia. As demonstrated in Figure 2.26, the moment
of inertia for this cross-sectional geometry is equal to the moment of inertia of the
rectangle minus two times the moment of inertia of the horizontal triangle minus two
times the moment of inertia of the vertical triangle about the neutral axis of the
cross-section.
The neutral axis of the horizontal triangle is not coincident with the neutral axis of
the structure and thus the parallel axis theorem must be employed. For the vertical
triangle, since it is perpendicular to the horizontal triangle, its dimension for height is
in the same direction as the width of the structure. As a result, its y-axis is coincident
with the neutral axis of the structure and thus the moment of inertia for the vertical
triangle is computed about its y-axis as [41, 43]:
1
1
brect h3rect = 12
wh3
Ix = 12

cos tan1 w/h
1 w 2 + 1 t h 1 +

t
2
36
cos tan1 w/h

cos tan1 h/w tan1 w/h

1
cos tan1 w/h

1
h

2
t 2 1 +

t
+ 2 w 2 + cos tan1 w/

h

cos tan1 w/h

1
h
h
1 +

t t

2
3
2

w
1
sin tan
/h
1
1
t w 1 +

2 48
h 2 +

t
2
h
h
cos tan1
/w
cos tan1
/w tan1 w/h
(2.15)
62
Figure 2.26: Moment of Inertia for the X-Supported Rectangle Cross-Sectional Geometry
2.3.5
Formulation of the Equations for Angle of Twist
The final unknown in the objective function is the angle , which denotes the angle
of twist of the main links. For a thin-walled tube with a closed cross section and
constant thickness the angle of twist is defined as [41]:
H
T L ds
=
4GA2m t
where T is the internal torque acting on the links at the wrist location,
(2.16)
ds is the
line integral around the entire boundary of the links cross-sectional area, G is the
shear modulus of the composite material, and Am is the mean area enclosed within
the boundary of the centerline of the links thickness.
However, this equation is for closed cross-sections that do not include any interior
support structure. Therefore, in order to utilize this equation it is necessary to alter
the I-beam and X-supported rectangular links into an equivalent hollow rectangular
link. As is demonstrated in Figure 2.27, the equivalent hollow rectangle has the same
width and height as that of the I-beam and X-supported rectangular links. Therefore
the thickness of the equivalent rectangle is increased to account for improvement in
the torsional rigidity that is provided by the I-beam or X-support structures. This
equivalent thickness is calculated by setting the cross-sectional area of the equivalent
63
rectangle to that of either the I-beam or X-supported rectangular link (whichever is

being optimized) and rearranging and solving for the equivalent thickness, tequiv . The
equation for the resulting equivalent thickness is:
tequiv =
2(w + h)
p
(2(w + h))2 16Ac
8
(2.17)
Figure 2.27: Equivalent Hollow Rectangle for the Ideal Cross-Sectional Geometries
The mean area, Am , for the equivalent rectangle is shown in Figure 2.28. Observation
of the figure indicates that the mean area is defined as:
Am = (w tequiv )(h tequiv )

For a rectangular cross section the line integral,
centerline boundary of the thickness:
64
(2.18)
ds, is simply the perimeter of the
I
ds = 2(w tequiv ) + 2(h tequiv )
(2.19)
Figure 2.28: Mean Area Enclosed within Boundary of Centerline for Equivalent Hollow Rectangle
2.3.6
Determination of Constraints
With the use of the equations derived in the previous section, the equation for the
total deflection in the robotic arm () as illustrated in Equation (2.6) is in terms of
the cross-sectional dimensions: height (h), width (w), and thickness (t). Considering
the equation for the total amount of deflection is the objective function, the variables
for the optimization problem are h, w, and t.
The constraints for this optimization problem still need to be defined. By analyzing
the geometry, a few constraints can be created. Considering that the thickness of the
X-support structure is the same as that of the outer rectangle, it is necessary that
the width is greater than or equal to four times the thickness. If this is not the case,
then it allows the possibility of the different components overlapping. As a result,
one would end up with either a solid part or one in which it is far too difficult to
differentiate the X-support from the outer rectangular tubing.
For the case of the I-beam supported rectangular structure, as long as the width is
65
more than three times that of the thickness then the part will not be solid. However, for convenience sake the same constraint for the X-supported rectangular link is
utilized.
From the results shown in Section 2.2.2, it was determined that a rectangular shape
in which the height is greater than the width is favoured over a square cross section.
Consequently, to minimize the total deflection of the robotic arm the height should
be greater than or equal to the width. The equations for these two linear inequality
constraints are:
w 4t
4t w 0
(2.20)
hw
wh0
(2.21)
Equations (2.20) and (2.21) can be put in the form Ax b:

h
0
1 1 0
0
0 1 4
t
(2.22)
In addition to the aforementioned linear constraints, the maximum output torque of

the motors can be used to create a constraint. As previously mentioned, the maximum
output torque of the PR110 motor which is connected to the first main link is the
limiting value. As a result, the maximum output torque of the PR110 is only needed
for developing the constraint that ensures that the torque that is generated by the
motor is sufficient to keep the robotic arm stable.
The main links were analyzed under the situation in which the maximum output
torque provided by the PR110 motor was being applied. Additionally, since this
66
situation involves analysis of forces only resulting in bending of the main links it is
analyzed with the payload acting directly at the end of the main links, not at an offset
equal to the length of the wrist. Furthermore, efficiency was included for the maximum
output torque of the motors to ensure that even with the motor not operating at full
power, the robotic arm will remain stable. This is the same loading scenario that
is illustrated in Figure 2.19. Using this free body diagram of the situation and the
consideration that the efficiency multiplied by the maximum output torque of the
PR110 motor must be greater than or equal to the sum of the torques that result due
to the forces acting on the links, the following non-linear constraint is developed:
Ac +
2ef f M
2 mmotor
max
+W
0
2
L
2
L g
(2.23)
In Equation (2.23), all the terms but the cross-sectional area are known. Since the
cross-sectional area of both the I-beam and X-supported rectangular links are in terms
of the height, width, and thickness, these are the variables for the constraint.
By looking at the equations for cross-sectional area for both the I-beam and Xsupported rectangular geometries (Equations (2.8) and (2.13), respectively), it is easy
to see that at least one of the three variables is raised to a power greater than one.
As a result, the constraint defined by Equation (2.23) is non-linear.
By observing the objective function, Equation (2.6), for this optimization problem it
is easy to see that it is a non-linear multivariable function. Additionally, there are
both linear and non-linear multivariable constraints (Equations (2.22) and (2.23)).
The MATLAB fmincon solver from the Optimization Toolbox [44] was used to find
the optimal solution for h, w, and t.
67
2.3.7
Determination of Bounds for the Design Variables
The fmincon solver can also utilize upper and lower bounds on the variables to achieve
the best results possible. Considering that the PR090 motor measures 90 [mm] by 90
[mm] on each of its faces, the upper bound for the height and width of the link was set
at 140 [mm] such that the link is not considerably larger than the motor. Similarly,
the lower bound used was 45 [mm], half the dimension of the PR090 motor.
As for the thickness of the link, the lower bound was chosen as 2 [mm] because prior
studies indicated that for best results no fewer than 10 layers of composite should be
used. Additionally, the thinnest laminates that are available and that are not difficult
to work with are 0.18 [mm] thick [45]. As for the upper bound of the thickness, it
should not be more than a quarter the lower bound of the width in order to ensure
that the linear constraint shown in Equation (2.20) is satisfied. Therefore, an upper
bound of 10 [mm] was selected for the thickness.
2.3.8
Material Specifications
The optimization methodology was performed assuming that the links were constructed of high modulus carbon fiber-epoxy matrix with a 60% fiber volume. It
was modeled as an isotropic material with the same properties as those used in the
preliminary FEA of Section 2.2. Thus a density () of 1,700 [kg/m3 ], elastic modulus
(E) of 220 [GPa], and Poissons ratio () of 0.25 were used.
Due to the highly longitudinal loading of the link, the shear modulus (G) was determined using the shear modulus equation for isotropic materials [5]:
G=
E
2 (1 + )
(2.24)
Using the aforementioned values, the shear modulus for the polymer composite material was calculated to be 88 [GPa].
68
2.3.9
Solutions Found from Optimization of Cross-Section
2.3.9.1
I-beam Supported Rectangular Link
Using the constants that have been declared and the equations derived throughout
this section, the optimization solution was achieved (please see Appendix A to view
the complete MATLAB coding). For the I-beam supported rectangular link, the
optimization methodology was used at a variety of different starting points for the
height, width, and thickness ranging anywhere from above the upper bounds to below
the lower bounds. In all cases, an identical solution was determined in a different
number of iterations. The solution point that was found was 140 [mm] for both
the height and width and 2.3829 [mm] for the thickness. This equates to a minimal
end-effector displacement of 0.37146 [mm].
The values from the solution point for the height, width, and thickness were substituted into the linear constraints (Equation (2.20) and (2.21)) as well as the non-linear
constraint (Equation (2.23)). All three equations were satisfied. This indicated that
the solution point that was returned was indeed a valid solution.
The non-linear constraint returned a value that was zero to at least 11 decimal places.
Inspection of Equation (2.23) reveals that a result for this equation that is essentially
zero is a result of the cross-sectional area being set as the maximum allowable amount.
This seems quite natural since increasing the size of the link and amount of material
used to produce it leads to an increase in the stiffness and consequently a decrease in
the deflection.
The solution point for both the height and width are at the upper bounds, whereas
the thickness is just slightly greater than the lower bound. This indicates that the
ideal solution for minimal displacement is to increase both the height and width while
decreasing the thickness. This makes sense since objects with thin sections like Ibeams and other popular structural geometries have a very high moment of inertia.
The thickness is greater than the lower bound because the upper bounds of both the
69
height and width make it impossible to have a cross-section that is the maximum
allowable amount with the minimum thickness.
2.3.9.2
X-Supported Rectangular Link
As was the case for the I-beam supported rectangular link, the optimization methodology for the X-supported rectangular link was carried out for a variety of starting
points that ranged from anywhere above the upper bounds to below the lower bounds.
An identical solution of 140 [mm] for the height, 105.24 [mm] for the width, and 2.0
[mm] for the thickness was obtained. This equates to a minimal deflection for the endeffector of 0.44736 [mm]. The solution point found satisfied the linear and non-linear
constraints indicating that it is a valid solution.
The solution point for the X-supported rectangular link also has the maximum permissible cross-sectional area. The solution for the height was found to be equal to
the upper bound, whereas the width was found to be smaller than the height. Additionally, the thickness at the solution point was equal to the lower bound value.
Therefore, this indicates that the thickness of the X-supported rectangular link is
minimized and then the height is increased relative to the width. This agrees with
the results found from the preliminary FEA in Section 2.2 in that it also determined
that a rectangular cross section was found to have less deflection than a square cross
section. Considering the majority of the loading on the robotic arm is longitudinal
for this application, it is reasonable that a link with a higher ratio for the height to
the width is preferred.
2.3.9.3
Study Without Upper Bounds
For both cross-sections, the solution point was restricted to the upper bounds. Therefore, to gain a better understanding of the optimal dimensions for the cross section the
optimization methodology was carried out without these limiting upper bounds. It
70
was tested at a variety of different starting points that returned the same solution for
the most part. Alternative solutions were ignored either because they did not satisfy
the linear and non-linear constraints or the corresponding deflection was greater than
that found for the true optimal solution.
The minimal deflection for the I-beam supported rectangular link was found to be
0.20192 [mm] for a height of 233.50 [mm], width of 64.24 [mm], and thickness of 2.0
[mm]. The thickness was minimized to the lower bound of 2.0 [mm] and the height
was increased relative to the width such that it was around 3.63 times as large.
The minimal deflection for the X-supported rectangular link was determined to be
0.34535 [mm]. The solution point has a height of 184.79 [mm], width of 49.93 [mm],
and a thickness of 2.0 [mm]. Similar to the I-beam supported rectangular link, the
thickness was minimized and the height was increased to be 3.70 times the width of
the link.
2.3.9.4
Recommendations from Optimization Results
Upon carrying out the optimization methodology described, it was determined that
with a combination of both torsion and bending, the I-beam supported rectangular
link results in less deflection compared to the X-supported rectangular link. For the
proposed application, the amount of bending is considerably larger than the amount
of torsion and thus it seems sensible that the I-beam interior support structure results
in a lower amount of deflection.
The ratio of the height to the width was greater for the X-supported rectangular
link since it has greater torsional rigidity as a result of its structure. Consequently,
its height can be increased relative to the width to further increase its longitudinal
stiffness. As for the I-beam supported rectangular link, the presence of the I-beam
provides less torsional rigidity and in order to provide a significant amount of torsional rigidity, the height-to-width ratio needs to be less than that of the X-supported
71
rectangular link. However, an X-support requires more material than an I-beam.

Therefore, the perimeter of the rectangle must be smaller for the X-supported rectangular link in order that the constraints are met. This results in a lower moment of
inertia and thus the bending stiffness is sacrificed.
For both structures the solutions from the optimization methodology indicated that
in order to improve the stiffness and thus reduce the amount of deflection experienced
in the robotic arm the following modifications should be made to the cross-section
in the sequence described. First off, the cross-sectional area should be maximized.
Secondly, the thickness should be minimized. Lastly, the height should be increased
as the width is decreased in order to arrive at a link that provides considerably greater
bending stiffness than torsional rigidity.
2.4
Summary
A serial main configuration with a R R k R main-arm layout was chosen for the
6-DOF robotic arm. The cross-section for the links was determined to be rectangular
with an I-beam support in the middle. FEA revealed that it is optimal to include a
negative taper angle of less than one degree. A 35% size proof-of-concept prototype
was determined to have minimal deflection for height, width, and wall thickness of
233.50 [mm], 64.24 [mm], and 2.0 [mm], respectively.
72
Chapter 3
Prototype Link Construction
In this chapter, the construction of the prototype link is described in detail. A novel
multi-segment link design and method of production was created. A suitable layering
pattern was chosen based on the requirements of the application for the robotic arm.
The creation of joints for mounting the links to the PR110 and PR090 motor is also
discussed. In the end, a 2-DOF main-arm proof-of-concept prototype was created 1 .
3.1
Geometry
In order to test the capabilities and performance of a polymer composite robotic

arm, a prototype was constructed. As mentioned in Section 2.1.2, the two main links
provide the majority of the reach for the desired 6-DOF robot. Therefore, for testing
purposes a 2-DOF robot main-arm representing these links was constructed.
From the results of the FEA and optimization outlined in Chapter 2, the links for the
prototype were constructed to have an I-beam supported rectangular cross-section.
The two links were both designed to be 87.5 [cm] in length. As determined from the
optimization procedure, the height, width, and thickness of the prototype links were
designed to be 233.5 [mm], 64.24 [mm], and 2.0 [mm], respectively. The links were
1
Please see [46] for paper on this topic
73
constructed with a negative taper angle of 0.5 . However, for ease of production, the
prototype links were designed to only be tapered in one direction. The taper angle
was relative to the longitudinal axis such that the height of the link was tapered.
Therefore, the height at the end of the first link and beginning of the second link was
designed to be 233.5 [mm]. Consequently, the height at the beginning of the first link
and at the end of the second link were designed to be 248.75 [mm] and 218.25 [mm],
respectively.
3.2
Modular Construction Process
The construction of the prototype links was done using a modular process. This novel
process involves creating a series of identical hollow modular base units using the same
mold. The modular base units are then assembled together in the proper orientation
to create the desired interior support structure of the component. The exterior profile
is then constructed directly around the modular base units. Therefore, the assembly
of the base units is used as the mold for the construction of the exterior profile as is
illustrated in Figure 3.1.
This technique is favourable because it requires only one simple mold for producing
the hollow base units. It is considerably less expensive than a complicated mold
that would be required to produce the link in one step. Additionally, this technique
produces a stronger bond between the base units and the exterior profile. As the epoxy
hardens and creates the solid exterior profile, it cures directly to the base units.
The I-beam supported rectangular link consists of two rectangular base units. Therefore, for the production of the prototype links one rectangular link half was created
on a rectangular mold. The mold was removed and the second rectangular half was
constructed. The two halves were then placed together and the exterior profile was
constructed around them.
74
Figure 3.1: Schematic of Modular Construction Process
3.3
3.3.1
Mold Construction
Phase 1: Wooden Mold
In order to produce the rectangular halves for the I-beam supported rectangular links,
two solid wooden molds, one for each link, were constructed for reasons of economics
and ease of production. The molds were constructed in order to produce two prototype
links with the dimensions as stated above. The wooden molds were constructed out
of medium density fiberboard (MDF). MDF was chosen because it is less expensive
than natural wood. It was selected over particle board and plywood because it is less
likely to chip apart when being cut and shaped.
The MDF was cut to size for producing both prototype links. The edges of the
wood were routed to aid in the removal of the mold. Additionally, the rounded edges
produce a link that is preferred over sharp corners that act as stress concentrators.
Numerous layers of Duratec primer and sealer were applied to the molds. After each
layer had been applied and dried, the molds were sanded. This ensured that there
were no bumps that would make removal more difficult. The wooden mold for the
construction of one of the links is pictured in Figure 3.2.
75
Figure 3.2: One Piece Solid Wooden Mold

Despite all the provisions taken to aid in the removal of the mold, the cured composite
rectangular part could not be separated from the wooden mold. Failure of removal
for a second time led to a revision of the mold.
3.3.2
Phase 2: Four Piece Fully-Detachable Aluminum Mold
In order to save both time and money, only one mold was created for the production
of both prototype links. To this end, the prototype links were designed to be the exact
same size. Also, the four piece mold was not designed to the same dimensions as the
wooden mold. The rectangular base unit that was created using the wooden mold
had a thickness of approximately 1.7 [mm]. This thickness represents half the total
thickness and thus a complete link would have a thickness of around 3.4 [mm]. This
is considerably larger than the prescribed 2.0 [mm]. Therefore, if it were to be made
with the prescribed height and width then the composite prototype links would be
too heavy. Furthermore, the relatively low torsional rigidity of the rectangular base
unit seemed to contribute to the difficulty in removing the link. As a result, the ratio
of the height to the width was decreased to improve its torsional strength. Lastly,
in order to ease assembly, the width of the link was designed to be slightly smaller
than prescribed to ensure that it fit between the mounting holes of the PR090 motor.
As a result, the dimensions of the mold were such that it created a link that has a
mid-height and width of 120 [mm] and 60 [mm], respectively.
76
Since the removal of a solid mold with only a -0.5 taper was not successful, a fullydetachable mold was created. Four flat aluminum plates that are combined to produce
the rectangular mold were used. The plates were kept together using screws and nuts
and bolts that can be taken apart to allow for disassembly.
This mold relied on compliance to enable the removal of the parts of the mold. Therefore, the mold was constructed out of aluminum because of its high flexibility and
relatively low cost. Since it is very flexible, a wooden skeleton was placed inside the
combination of the aluminum plates as illustrated in Figure 3.3. The skeleton helped
to keep the shape of the mold and prevent the aluminum plates from collapsing in on
themselves and thus altering the shape of the link being constructed.
Figure 3.3: Four Piece Fully-Detachable Mold

Aluminum is non-porous and therefore it does not allow the epoxy to permeate the
mold which makes removal more difficult. Therefore, it was not necessary to apply
primer and sealer to the mold.
Upon creation of the composite rectangular base unit, the four piece mold was successfully detached from the cured part. However, the resulting composite part was
of poor quality. The part had many variations and warpages in its rectangular profile. The top and bottom aluminum plates did not hold their shape very well as the
vacuum force pulled them towards it, creating a link with severe bowing on the top
and bottom surfaces. Therefore, the mold had to be modified so that it was flexible
77
enough to enable removal of the part but rigid enough to keeps its shape during the
composite link production.
3.3.3
Phase 3: Two Piece Fully-Detachable Aluminum Mold
In order to eliminate the possibility of the top and bottom pieces flexing as the part
was cured, they were made rigid. Therefore, in order to still offer full disassembly of
the mold, two identical L-shaped halves were used for the mold in place of the four
aluminum plates as illustrated in Figure 3.4. The aluminum plate was bent such that
the side of the mold and the top of the mold were one solid piece. Similarly, the other
side wall and the bottom of the mold were one continuous piece. The same wooden
skeleton that was used for the four piece mold was used to ensure that the mold held
its shape. The end-result was a composite part with a relatively smooth profile that
was capable of being successfully removed from the mold.
Figure 3.4: Two L-Shaped Piece Fully-Detachable Mold
3.4
3.4.1
Prototype Link Construction

Technique
Based on the fact that a modular construction process as described in Section 3.2 was
to be utilized, the links were constructed as a series of hollow members. As a result,
78
the filament winding process is well-suited for the production of the links. Therefore
it is envisioned that the industrial-grade robotic arm will be constructed using filament winding. However, the 2-DOF robot main-arm proof-of-concept prototype was
developed using hand lay-up for reasons of economics.
The production of the polymer composite link was achieved using vacuum bag molding. This was used in conjunction with hand lay-up because it helps to reduce the
amount of epoxy over hand lay-up alone. The vacuum draws away the excess epoxy
from the fibers. This results in a final composite product that has a greater percentage of fibers. This is ideal because they provide the reinforcing properties to the
composite whereas the epoxy only holds the fibers in place.
Unlike the polymer composite robots discussed in Section 1.5.1, the prototype links
were not created with the use of an autoclave. The prototype links were too large
to safely fit in a conventional oven, and it was too expensive to purchase a suitable
autoclave. For these reasons, the composite links were cured at room temperature.
3.4.2
Materials Used for Prototype
Considering the hand lay-up procedure was used for the production of the prototype
links, a uni-directional twill manufactured by BGF Industries, Inc. was used. Additionally, the selected twill was uni-directional because it permits control over the
mechanical properties of the composite by having complete control over the fiber orientation. Furthermore, the twill material used was T300 carbon fiber because, as
discussed in Section 1.4.3.2 and outlined in Section 2.2.1, carbon fibers are preferred
over aramid and glass fibers because they have the highest stiffness.
Epoxy was selected as the matrix material because it has better structural properties
than other alternatives. PR2032, an epoxy manufactured by Aeropoxy, was used for
the production because it cures at room temperature. In addition to this, PH3660, a
methyl ethyl ketone peroxide (MEKP) catalyst also manufactured by Aeropoxy, was
79
used to improve the stiffness of the polymeric composite.
3.4.3
Ply Lay-up Pattern for Prototype
As mentioned in Section 2.3.9.4, the majority of loading for this application is bending.
Therefore to provide improved strength for this loading, the majority of fibers were
oriented parallel to the longitudinal axis (0 ) [25]. In addition to this, some layers
should be oriented at an angle of 45 to the longitudinal axis so as to promote strength
against both torsional and shear loads [29]. Additionally, a small percentage of 90
layers should be present to couple with the 0 layers and improve the axial strength.
Also, as indicated in [26], they help to improve the strength at the joint by reducing
the hoop stress under vertical loads.
The percentage of 45 layers is determined based on the l/R ratio as outlined in [31].
Therefore for the I-beam supported rectangular prototype link, the Arm Ratio was
determined as:
Arm Ratio =
L
1/ havg
2
(3.1)
where havg is the average height of the link. Therefore, for a link length of 87.5 [cm]
and average height of 233.5 [mm], the arm ratio is approximately 7.5. By observing
Figure 1.8, the percentage of 45 layers for maximum fundamental frequency is 25%.
In order to meet the aforementioned specifications, 12 layers were used. Therefore,
three of the layers were oriented at 45 to the longitudinal axis. Also, the link was
designed so that no more than 10% of 90 layers were present. Therefore, only one
90 layer was used. The remaining eight layers used were 0 layers in order to provide
sufficient bending stiffness. The order of fiber layers was chosen to be repetitive to
provide an even distribution of the types of layers. The inner layer was 0 , followed by
45 , 0 , 0 , 45 , 0 , 0 , 90 , 0 , 0 , 45 , and finally concluding with another 0 layer.
80
3.4.4
Manufacturing Procedure
As mentioned in Section 3.2, the prototype links were constructed by producing

two rectangular base units first and then wrapping the exterior profile around them.
Therefore, the rectangular base unit was created out of the first 6 plies of the unidirectional carbon fiber twill.
The mold was coated with six layers of mold release wax produced by T.R. Industries
to ensure the release of the mold. Additionally, the edges where the L-shaped mold
halves meet were sealed with tape to prevent the epoxy from entering this joint and
bonding the halves together. The carbon fiber twill was then cut to the proper
dimensions. Since the mold is tapered, the carbon fiber cannot be wrapped around
the entire mold, and instead was cut into individual pieces for each side of the mold.
The epoxy and stiffener combination was then applied to the twill layers with the use
of a paint brush. The layers were then placed onto the mold one at a time. A firm
roller was then used to roll out air bubbles and help ensure that the part produced
did not have any voids in it.
Once all the layers had been placed on the mold, the vacuum bag apparatus was setup
as illustrated in Figure 3.5 (this figure shows a cut away version with all components
visible). Peel ply was wrapped around the carbon fiber twill followed by release film
and breather fabric. The peel ply aided in the removal of the release film and the
breather fabric that helped absorb the excess epoxy. The part was then surrounded
by a vacuum bag that was fully sealed and connected to a vacuum pump. The vacuum
pump was then turned on, sucking up the excess epoxy as it cured creating the hardened composite base unit. The vacuum bag apparatus was removed approximately
six hours later to allow the part to fully cure.
Once cured, cotton filling was placed inside the mold and liquid nitrogen was poured
inside. This decreased the temperature of the aluminum mold drastically, causing it
to contract. The mold was then pushed out of the composite part, one half at a time.
81
Figure 3.5: Vacuum Apparatus Used to Produce the Composite Link

The epoxy and release wax residue was cleaned off of the mold.
The procedure was repeated, creating the second composite rectangular base unit.
Epoxy was then applied to the two base units bonding them together. The final six
plies were cut, soaked in epoxy, and placed on the assembled base units. The vacuum
bag apparatus was setup again and the part was cured, creating a completed link.
This entire procedure was then repeated, creating a second identical link.
3.5
Link Attachment Bracket
Once the composite prototype links were successfully produced, a bracket for attaching
the link to the motors needed to be constructed. The brackets were made out of
stainless steel because it has a very high modulus of elasticity. This helps to ensure
that any deflection of the attachment device is negligible in comparison to that of
the links. Additionally, carbon fiber is quite cathodic and can thus cause metals to
corrode so a material that is not that corrosive was used [29].
Figure 3.6 shows the design of the bracket. It consists of six plates that enclose the
exterior and interior support walls of the link. It also includes a series of spacers
that are placed between the plates in the hollow portion of the link to keep the plates
82
butted up against the walls of the link. All the plates have four equally spaced holes in
them. These holes were used to enable the connection of the bracket to the composite
link with four bolts. The diameter of the holes in the plates and the link were made
larger than the diameter of the fastening bolts so that stress concentrations were not
created from the link resting on the bolts in order to prevent the link from being
damaged. Consequently, the connection between the attachment device and the link
is fully frictional.
Figure 3.6: Bracket for Attaching Composite Link to the Joint Motor
All of the plates of the bracket were clamped together and four holes were drilled.
Then, one of the plates was placed on the link and the centers of the holes were
marked off. Four holes were then drilled through the entire width of the composite
link. Another set of four holes were drilled through the base plate at the corners.
Two of the plates that attach to the exterior walls of the link were welded to the base
plate to create a rigid connection. The spacers were glued to the plates to keep them
in place while PR2032 epoxy was applied on the other side of the plates. The plates
were then inserted into the hollow part of the link and lined up with the holes in
the composite link. The fastening bolts were then inserted through the holes in the
link and bracket and tightened to securely fasten the link to the attachment bracket.
Finally, the motor was lined up with the base plate of the bracket and was fastened
83
to it with bolts.
The final assembly of the motor with the composite link via the attachment bracket
is illustrated in Figure 3.7. Another three brackets were created and fastened to the
links and the motors to create a fully functional 2-DOF robot main-arm prototype.
The motor that powers the first link of the 2-DOF prototype was securely mounted
to the ground as is illustrated in Figure 3.8. The complete 2-DOF robot main-arm
proof-of-concept prototype is shown in Figure 3.9.
Figure 3.7: Composite Link Connected to the Motor with Attachment Bracket
Figure 3.8: 2-DOF Main-Arm Proof-of-Concept Prototype Mounted to the Floor
84
Figure 3.9: 2-DOF Main-Arm Proof-of-Concept Prototype
85
3.6
Summary
Two identical carbon fiber-epoxy composite links with an I-beam supported rectangular cross-section were created. A novel manufacturing procedure which included producing simple hollow modular base units, assembling them, and wrapping around the
entire assembly was utilized to simplify the production of the links. Steel attachment
brackets were also manufactured. A 2-DOF main-arm proof-of-concept prototype was
created by fastening the composite links to the motors with the attachment brackets.
86
Chapter 4
Prototype Link Testing
Once the composite links were created, testing was performed on the links. A variety of tests were performed to test both the static and dynamic capabilities of the
carbon fiber-epoxy composite link. The procedure is described and the results are
indicated and analyzed. Additionally, a virtual model of the prototype was created
using Abaqus CAE. FEA was performed on the virtual model in order to test the
results obtained for the physical prototype. The results and analysis of the tests are
discussed in this chapter.
4.1
4.1.1
Physical Prototype Testing

Single Link Static Testing
Static testing was performed on a single link to determine the deflection of the link
under a constant load. By using only one link, it permits a measurement of stiffness
for the link itself that does not include any displacements that result due to the
attachment brackets that connect the two composite links, nor due to the joint.
87
4.1.1.1
Weight Support Component
For the static testing, the composite link was studied as a cantilever beam. In doing
so, a force was applied on the end of the link by adding weights to the link. Therefore,
a component was created that could support the weight that was to be added on to
the link. The weight that was placed on the end of the link were standard weighted
discs used for weight-lifting. Therefore the weight support component included a
hollow cylindrical tube that could be inserted through the inner hole of the discs.
This cylindrical tube was connected to a base plate that has four holes that match
the base plate of the attachment bracket. The design of the weight support component
is illustrated below in Figure 4.1.
Figure 4.1: Device for Supporting Weight on the Composite Link

The center of the four mounting holes on the base plate of the attachment bracket
were marked on a 9 [cm] by 9 [cm] stainless steel plate. The holes were then drilled. A
hollow cylindrical stainless steel tube with a diameter of 25 [mm] and wall thickness
of 3 [mm] was cut to a length of 12 [cm]. The tube was placed in the middle of the
plate and was welded to it to create a rigid joint. The weight support component
was then fastened to the attachment bracket with nuts and bolts. The dimensions of
the component were chosen such that they matched up with the attachment bracket.
The total weight of the two components combined was 3.85 [kg]. This weight made
it possible to easily add weighted discs to achieve a total payload of 17.5 [kg].
88
4.1.1.2
Apparatus
As illustrated in Figure 4.2 the larger end of the composite link was rigidly attached
to a cement wall with the use of the attachment bracket (see Section 3.5). This was
achieved by drilling four holes in the wall and using cement anchor bolts to mount the
bracket to the wall. By mounting the link to the wall rather than a motor, it eliminated
any displacements that are a result of the motor, including any inadvertent rotation
due to loading of the link. This enabled the measurement of the displacement that
occurs due to the deflection experienced by the composite link only.
Figure 4.2: Link Mounted to Wall

Figure 4.3 shows a schematic of the experimental setup used for the investigation
of the static loading. As shown, at the far end of the composite link a laser was
mounted such that it was 30 [cm] below the bottom of the link and perpendicular
to the ground. The laser was shone near the end on the underside of the link and
measured the vertical distance between the laser and the link.
The displacement was measured at four intervals. First, the link was tested with just
the weight attachment device mounted on the end of the link. Measurements were
then taken by incrementally adding the 4.55 [kg] (10 [lb]) weighted discs to the link.
The measurements were taken 10 minutes after the weight had been added to allow
the link to settle.
89
Figure 4.3: Schematic of Experimental Setup

4.1.1.3
Results
The average values of the data collected for the vertical displacement from a series
of five tests using the laser are shown in Table 4.1. The difference between each
subsequent measurement is close to the same value; in fact the maximum percentage
difference was less than 5%. This indicates a primarily linear displacement.
Table 4.1: Vertical Displacement of the Link as Measured Using a Laser
Weight of Load (kg)
3.86
8.41
12.95
17.5
Laser Measurement (mm) 300.000 299.885 299.766 299.646

Displacement (mm)
0.115
0.119
0.120
The total vertical displacement that the composite link underwent was measured as
0.354 [mm]. However since the attachment bracket was bonded to the link with epoxy,
the displacement due to adding the first 3.85 [kg] could not be measured and thus the
aforementioned value was only due to three 10 [lb] weighted discs; a payload of 13.65
[kg]. Therefore the total deflection, d1 , that the link undergoes due to a payload of
90
17.5 [kg] was interpolated from the data shown in Table 4.1 using:
d1 = w1
d
w
(4.1)
where w1 is the weight of the attachment bracket and weight support, 3.85 [kg], d
is the total displacement measured, 0.354 [mm], due to a total weight, w, of 13.65
[kg]. Therefore, d1 was determined to be approximately 0.100 [mm]. Therefore, the
total displacement that the link undergoes due to the maximum payload of 17.5 [kg]
for the prototype was assumed to be 0.454 [mm].
4.1.1.4
Analysis
The percentage error between the values obtained from the various tests using the
same loading was less than 4%. Despite the fact that this indicates that the testing
procedure is fairly consistent, there are a few sources of error present. First, the
composite link underwent bending and thus had components of displacement in both
the vertical and horizontal direction. The measurements taken by the laser only
indicate the vertical component of displacement. Another source of error was the
fact that the laser was not positioned at the very end of the link, only near it. The
absolute end of the composite link would actually experience a displacement that is
greater than 0.454 [mm]. Another source of error was the settling of the link. The
laser data continued to vary as the link continued to creep. In order to eliminate this
from affecting the results, the measurements were taken at the same time intervals
for all 5 series of tests. This however still presented an issue as upon taking the load
off of the link, the link did not return back to exactly 30 [cm] above the laser.
4.1.2
Repeatability Testing of the 2-DOF Prototype
As mentioned in Section 3.5, the 2-DOF main-arm prototype was constructed with a
PR110 motor, a PR090 motor, and two identical carbon fiber-epoxy composite links.
91
The PR110 motor was bolted to the ground to secure the prototype. The same laser
that was used in the previous test was secured to the ground. It was positioned such
that the laser would shine near the end on the underside of the second link. Due
to limitations of the motors with respect to their torque capabilities, the tests were
performed with only a payload of the attachment bracket and weight support (3.85
[kg]).
4.1.2.1
Motor Control
The motors were connected in series to a PC using Computer Area Network (CAN)
protocol to communicate between them and the PC. A simple program was used to
move the two links in joint space using the PowerCubes built-in controllers.
4.1.2.2
Results
As illustrated in Figure 4.4, the first link was positioned 45 from vertical and the
second link was positioned 45 from the first such that the longitudinal axis of the
second link was parallel to the ground. Both motors were rotated at a speed of 10
per second, moving the links to 0 from vertical such that the arm was extended fully
and pointing straight up. The links were then moved at the same speed back to the
original position of (45 , 45 ), and the laser measured the vertical distance to the link.
This was repeated another four times to measure the repeatability.
The repeatability of the 2-DOF main-arm prototype was tested at a second location.
The first link was positioned 55 from vertical and the second link was positioned
35 from the first as illustrated in Figure 4.5. As was the case with the previous
position, the longitudinal axis of the second link was parallel with the ground. The
same procedure was used to measure the vertical displacement of the link in order to
test the repeatability. The results for both tests are shown in Table 4.2.
The results obtained from the repeatability tests shown in Table 4.2, indicate a high
92
Figure 4.4: First Repeatability Test Position (45 , 45 )
Figure 4.5: Second Repeatability Test Position (55 , 35 )

level of repeatability for the proof-of-concept prototype. When the repeatability was
tested at a location with both joint angles at 45 , the maximum variation between any
two of the five measurements was 0.031 [mm]. For the case of the proof-of-concept
prototype positioned with the PR110 motor and PR090 motor having joint angles
(55 , 35 ), the maximum variation between any two of the five measurements was
0.049 [mm]. This indicates that the proof-of-concept prototype has a repeatability of
approximately 0.05 [mm].
93
Table 4.2: Laser Measurements For Repeatability Test

Trial # Measurement (mm) Measurement (mm)
@ (45 , 45 )
@ (55 , 35 )
1
35.460
45.868
2
35.491
45.828
3
35.481
45.850
4
35.470
45.843
5
35.472
45.819
4.1.3
Destructive Tests
Polymer composite materials are known to be considerably weak in the transverse

direction and thus are prone to failure due to compressive forces. Therefore, two
tests were performed on the carbon fiber-epoxy composite link to ensure that catastrophic failure does not occur when a compressive force is applied. The tests were
performed using the LS100Plus materials testing machine by Lloyd Instruments to
apply a continually increasing force to ensure that if the motion of the link were to
ever be obstructed it would not lead to catastrophic failure.
4.1.3.1
Compression Test
A purely compressive test was performed on the composite link. Due to size limitations
of the LS100Plus, a 65 [cm] long portion of the composite link was tested. An 11.25
[cm] section of the link was cut off on both ends using a vertical drop saw.
The compressive plates on the LS100Plus have a circular shape with a diameter of 10
[cm]. Consequently, when placed next to the cross-section of the link, the link extends
beyond the circular plate. Therefore, in order to apply the compressive force evenly
across the links cross-section an adapter plate was created.
As shown in Figure 4.6, the adapter plate consists of a circular profile that matches
up with the compressive plate and a rectangular profile that matches up with the
links cross-section. A hollow steel cylinder with an inner diameter of approximately
94
10.5 [cm] was cut to a length of 5 [cm]. A 7.5 [mm] thick steel plate was cut into a
rectangle that was 10 [cm] by 15 [cm]. The hollow cylinder was then centered on the
plate and welded to it to create a rigid connection. Four 0.25 [mm] thick 10 [cm] by
2 [cm] strips of steel were then cut. The four strips were then welded on the opposite
side of the rectangular plate such that they would surround the exterior profile of the
prototype link. Strips of rubber were then glued on the inner side of the plates in
order to help reduce the creation of stress concentrations as the compressive force was
applied.
Figure 4.6: Adapter Plate Used for Compression Test

The section of composite link was loaded into the LS100Plus. The two adapter plates
were placed between the link and the compressive plates as illustrated in Figure 4.7.
The plate was lowered so that it was in slight contact with the adapter plate. The door
was closed and the upper compressive plate was lowered at a constant speed applying
an increasing force on the link. Once the force reached 98,412 [N], the link failed. The
displacement of the top plate at this point was 4.5888 [mm]. The compressive plate
continued to be lowered, compressing the link even further. However, the compressive
force required was lower as indicated by the outliers in the graph shown in Figure
4.8. Therefore, this indicated that once the composite link did fail, the strength of
the link was decreased but catastrophic failure did not occur.
95
Figure 4.7: Apparatus for Compressive Test

The section of the link after the compression test is illustrated in Figure 4.9. As
can be seen, the link split at its edge at the location of buckling. This makes sense
considering the way the link was produced. As mentioned in Section 3.4.4, the twill
was not wrapped around the exterior profile but rather was cut into individual pieces
for the four sides of the rectangular link. Therefore, the fibers are not continuous
around the edges and thus a seam exists. As previously mentioned, the full-scale arm
is designed to be produced using filament winding and thus this would not be an
issue. Despite the fact that the prototype link was weakened due to this seam, it was
still capable of withstanding a compressive force of 98,412 [N], which is equivalent to
more than 10 metric tonnes.
96
Figure 4.8: Compressive Force Versus Displacement of Composite Link
Figure 4.9: Section of Link After Compressive Test
97
4.1.3.2
Three Point Bending Test
Since bending is the primary force that will be acting on the manipulator, it is also
necessary to analyze the failure of the composite link due to bending. This was
achieved using the LS100Plus and the three point bending apparatus. An even amount
was cut off on both ends of the composite link leaving a 45 [cm] long section. The three
point bending apparatus was placed in the middle of the machine and the contacts
were spread out to their maximum span of 32 [cm]. The section of the composite link
was centered on the contacts vertically as shown in Figure 4.10. The third contact
plate which was mounted to the driving actuator was lowered such that it was just
touching the link.
Figure 4.10: Apparatus for Three Point Bend Test

The drive actuator was lowered at a constant speed applying a force in the middle of
the link. The link fractured after a deflection of 30.707 [mm] due to a force of 7,518
[N]. As was the case with the previous test, after the link had fractured it remained
intact. The force required to deflect the link afterwards, however, was decreased as
illustrated by the outliers in the graph shown in Figure 4.11.
98
Figure 4.11: Bending Force Versus Displacement of Composite Link

The force being applied in the middle of the section of link being tested caused it
to bow down such that a tensile force was applied on the bottom portion and a
compressive force on the top of the link. Figure 4.12, illustrates that the link fractured
at the top. This makes sense since composite materials are considerably stronger in
tension than in compression.
Beyond just fracturing, the composite link also delaminated. In the middle of the
I-beam support, the two modular base units split apart from one another as they
bowed in opposite directions during the bending test. The splitting of the two halves,
however, only occurred near the area where the force was applied. Additionally, the
top layer of exterior profile that surrounds the modular base units also delaminated
near the force application area. This indicates that although the seam along which
the two modular base units are in contact is an area of weakness, the epoxy bond
between them is sufficient.
99
Figure 4.12: Section of Link After Three Point Bend Test
4.2
4.2.1
Finite Element Analysis

Procedure
The link was modeled as an assembly of parts. It consisted of two smaller rectangular
parts that represented the modular base unit and one larger rectangular part that
represented the exterior profile of the link as represented in Figures 4.13 and 4.14.
Due to the modeling limitations of Abaqus pertaining to tapering, the parts were
made in SolidWorks. The parts were then saved as an IGS file and imported into
Abaqus.
These parts were designed to the same dimensions as the physical prototype. Additionally, the virtual parts were constructed as surfaces in order to accurately model
the composite prototype. Therefore, the surfaces were defined to have the same dimensions as the exterior dimensions of the physical link.
100
Figure 4.13: Modular Base Unit for Virtual Prototype
Figure 4.14: Exterior Profile for Virtual Prototype
4.2.2
Material Properties
The link was modeled as an elastic material of type lamina so that layering information of the link could be included. The mechanical properties of the material were
defined to be the same as those used by Stanley in [47] for a uni-directional graphite
fiber-epoxy composite. Therefore the longitudinal and transverse modulus of elasticity, El and Et , were defined as 135 [GPa] and 13 [GPa], respectively. The shear
modulus, G12 , G13 , and G23 (where the 1-direction is along the fibers, the 2-direction
is transverse to the fibers in the laminate, and the 3-direction is normal to the laminate) were defined as 6.4 [GPa], 6.4 [GPa], and 4.3 [GPa]. The Poissons ratio, 12 ,
was defined as 0.38.
101
The virtual prototype was defined to have a density that was equal to that of the
physical prototype. Since both the mass and volume of the prototype link are known,
the density, , is determined accordingly:
m
V
(4.2)
where m is the mass and V is the volume of the prototype link. Therefore, for a
mass of 2.21 [kg] and a volume of 0.00137326 [m3 ], the density is approximately 1,609
[kg/m3 ].
4.2.3
Modeling
The virtual prototype was created using conventional shell elements. In this, a layering
pattern was defined for the virtual prototype that was identical to the layering pattern
that was used for the production of the physical prototype. Therefore, the smaller
rectangular surface that represents the modular base unit was defined as having a
layering pattern of 0 , 45 , 0 , 0 , 45 , and 0 . The layering pattern for the larger
rectangular surface that represents the exterior profile of the link was defined as 0 ,
90 , 0 , 0 , 45 , 0 . Each layer was defined as having a thickness of 0.2833 [mm] such
that the total part thickness was 3.4 [mm].
The surfaces shown in Figure 4.13 and 4.14 were assembled to create the completed
virtual link. Two of the smaller rectangular surfaces were aligned such that they were
parallel to each other and placed side by side with their side walls contacting each
other. A tie constraint was utilized to bound the two touching faces together. The
larger rectangular surface was also aligned parallel to the two smaller rectangular surfaces. It was positioned such that it surrounded the two smaller rectangular surfaces
and was an equal distance away from all the exterior faces around its entire profile.
Another four tie constraints were used to bind the four walls of the large rectangular
surface to the parallel walls of the two smaller rectangular surfaces. This created a
102
single component by completely bounding the three individual surfaces together.

In order to perform the FEA, a mesh was created for the virtual prototype. All the
surfaces were constructed using the same mesh properties. The mesh was created
using the free technique and the advancing front algorithm. The shape of the mesh
elements were defined as being quad-dominated. The size of the elements was chosen
to be 8 since this equated to the largest number of nodes that was less than the
permitted 20,000 for the software.
4.2.4
Loading Conditions
In order to successfully model the prototype link, the faces on the virtual prototype
link were partitioned. A rectangle was partitioned on all surfaces at both ends of
the virtual link that had the same dimensions as the plate of the attachment bracket
discussed in Section 3.5. An encastre boundary condition was created on this area on
all six faces at the wider end of the link as shown in Figure 4.15. This represented a
completely rigid joint between the attachment bracket and the link that permits no
movement of the link relative to the attachment bracket.
Figure 4.15: Bounded End of Virtual Prototype Composite Link

103
Two loads were applied to the part. The first was a gravity load that represented the
weight of the link itself. The second load, which was applied in the same direction of
the gravity load, represented the weight of the payload. It was created as a surface
traction load. As was the case with the boundary condition, it was applied on the area
where the attachment bracket and the link were in contact as illustrated in Figure
4.16. The surface traction, PST was determined as:
PST =
Wg
6Aplate
(4.3)
where Aplate is the area of one of the plates on the attachment bracket. Therefore, for
a payload of 17.5 [kg] acting under the gravitational acceleration of 9.81 [m/s2 ], the
pressure acting on each 0.0090 [m2 ] area that represents the plate was 3179.17 [Pa].
Figure 4.16: Surface Traction on Virtual Prototype Composite Link
4.2.5
Results
The virtual prototype was solved using the loading scenario described above. An
exaggerated view of the link during loading is shown in Figure 4.17. It was determined
that under a payload of 17.5 [kg], the carbon fiber-epoxy composite link deflected a
maximum of 0.09195 [mm] at the end of the link where the load was applied.
104
Figure 4.17: Deflection of Virtual Prototype Link

In order to assess how well suited carbon fiber-epoxy composite is for this application,
an aluminum and steel virtual prototype link were also tested. The links were constructed as a single member that has the same dimensions as the physical prototype.
The identical loading scenario used for the composite link was utilized here. The
steel link was defined as plain carbon steel alloy A36 with a density of 7,850 [kg/m3 ],
modulus of elasticity of 207 [GPa], and Poissons ratio of 0.3. The aluminum link
was defined as aluminum alloy 6061 which has a density of 2,700 [kg/m3 ], modulus of
elasticity of 69 [GPa], and Poissons ratio of 0.33 [5].
The results for the links constructed of the three different materials is shown in Table
4.3. The FEA results indicated that the steel link deflected the least. The composite
link was second, deflecting about twice as much. The aluminum link was the least
favourable, deflecting nearly three times as much as the steel link. However, the steel
link weighed nearly five times more than the composite link. Therefore the composite
link is about two and a half times better than the steel in terms of deflection relative
to its weight.
105
Table 4.3: Comparison of Prototype Link Constructed of Different Materials

Material
Composite
Steel
Aluminum
4.3
Maximum Deflection
0.09195 mm
0.04664 mm
0.1225 mm
Weight
2.21 kg
10.78 kg
3.71 kg
Comparison of Physical and Virtual Prototype
The results from the single link static testing for the physical prototype were compared
with those found for the virtual prototype. The deflection for the physical prototype
was determined to be 0.454 [mm] for a payload of 17.5 [kg]. Meanwhile, the deflection
for the virtual prototype was found to be 0.09195 [mm].
Despite the discrepancy between the results obtained for the physical and virtual
prototype, the solution obtained for the virtual prototype is acceptable. The results
for both tests are roughly the same considering that they are on the same order of
magnitude. Additionally, the composite link was constructed by hand as mentioned in
Section 3.4.1. As a result, the constructed links have various flaws and imperfections.
Also, the strength of the link was compromised due to the production method utilized.
The carbon fiber material was not a continuous piece; rather it was cut into sections
for each face. As a result, a seam existed on the corners. As mentioned in Section
4.1.3, this seam was the point of failure for the link. In addition, the prototype links
were not cured in an autoclave which reduced its strength. It is expected that links
manufactured using filament winding and cured in an autoclave would have deflections
closer to the FEA results.
4.4
Summary
Static testing was performed on the carbon fiber-epoxy composite link. The results
indicated that the composite link effectively limited the amount of deflection expe106
rienced at the end of the link. Repeatability testing that was performed on the
2-DOF main-arm proof-of-concept prototype illustrated the damping capabilities of
composite materials by providing a repeatability of approximately 0.05 [mm]. Two
destructive tests showed that catastrophic failure did not occur and that carbon fiberepoxy composites weakness in compression is not an issue for this application. The
results from the FEA on the virtual prototype indicated that the theoretical deflection was less than that found for the physical prototype link. However, the agreement
between the two values was enough to indicate that if the links were to be constructed
using filament winding and cured in an autoclave the composite links could provide
considerably high stiffness.
107
Chapter 5
Design of the Full-Scale Robotic
Arm
In this chapter the design of the full-scale arm is discussed. Suitable motors and
gearheads were selected for the operation of joints two and three. Static testing was
performed on the part in order to determine the deflection when the links are extended
straight out. Links of varying dimensions were tested. The optimal dimensions and
layering pattern for minimal weight for a link that has stiffness equivalent to industry
standards was determined.
5.1
Motor Selection
In order to accurately test the static deflection of the two main links when extended
outwards, the motor that will operate the second main link (i.e., joint 3) must be
identified. The speed of the motor was selected by observing speeds of motors used
in other large-scale robotic arms. Table 1.1 indicates that the Kuka KR 100-2P and
Motoman EPH100, the two largest arms, have a speed of 95 [ /s] and 110 [ /s],
respectively. However, the reach of these arms is only 3.5 [m] and 3.0 [m]. Since the
robotic arm to be designed has a longer reach, the speed of the motor should be less.
108
Therefore, the motor speed, , was selected to be 60 [ /s].

The selected motor must be capable of permitting the robotic arm to meet the specifications outlined in Section 1.2. Consequently, it must enable the robotic arm with a
reach of 5 [m] and payload of 50 [kg] to move the tool at a linear speed of at least 0.15
[m/s]. Assuming that the motor that operates the first main link has a speed of at
least 30 [ /s], half the speed of the motor that operates the second link, then the tool
has a linear speed of at least 1.25 [m/s]. This scenario is demonstrated in Figure 5.1.
It illustrates the operation of the motors for two seconds. Using basic trigonometry,
X is determined to be 1.25 [m]. Therefore, the distance between the start of the first
link and end of the second link is 2.5 [m]. Consequently, the arm undergoes a linear
displacement of 2.5 [m] during a two second interval and thus it has a linear speed of
1.25 [m/s].
Figure 5.1: Scenario for Motion of Full-Scale Robotic Arm

The power necessary to move the second link and full payload at the desired speed
must be calculated in order to select an appropriate motor. The free body diagram
for the loading scenario is illustrated in Figure 5.2. In order for static equilibrium,
the torque provided by the motor, Mmotor , is:
Mmotor = WLink2
LLink2
+ WEE LLink2
2
(5.1)
where LLink2 is the length of the second main link, WLink2 is the weight of the second
109
main link, and WEE is the weight of the end-effector. It was assumed that LLink2 was
2.5 [m] since the two main links provide the majority of the 5 [m] reach and are the
same length, the mass of the link was 250 [kg], and the mass of the end-effector was
80 [kg] (50 [kg] payload plus 30 [kg] wrist). Therefore, the required moment of the
motor is determined to be 5,027.6 [Nm].
Figure 5.2: Free Body Diagram of Second Main Link

The required power, P , of the motor is defined as:
P = Mmotor
2
360
(5.2)
Therefore, in order to provide the necessary moment of 5,027.6 [Nm] at a desired

speed of 60 [ /s], the motor is required to have a power of 5,264.9 [W].
Rotary servomotors from Danaher Motion were investigated in order to determine a
suitable motor. In doing so, a safety factor of two was included such that the motor
was required to have a power of 10,529.8 [W]. Therefore, Model B-804-A which is
capable of 10,800 [W] was chosen (See Appendix B.1) [48].
A suitable gearhead for operation with the motor was also selected. The required
gearhead ratio, G.R., is calculated as:
G.R. =
RP Mmotor
60s rot
min
360
110
(5.3)
where RP Mmotor is the speed of the motor in rotations per minute. Since the motor
runs at 1,500 [rpm], a gearhead ratio of 150:1 is required to slow the speed to the
desired 60 [ /s]. Therefore, gearhead DTR90-150 was selected because it has a ratio
of 150:1 and its continuous rated torque at 3,000 [rpm] for 10,000 hour life of 81 [Nm]
is slightly greater than the continuous stall torque of the B-804-A motor of 78.6 [Nm]
operating at full speed of 1,500 [rpm] (See Appendix B) [48].
5.2
5.2.1
Dimensional Analysis of Links

Purpose
The full-scale robotic arm is to be designed such that it has a stiffness that is equivalent
or better than other robotic arms currently available. In a study performed by Airbus,
four robotic arms were fully loaded and tested in their theoretical worst position for
deflection. The results of the static flexibility determined from the test is shown
in Table 5.1 [49]. The Kuka KR 60-3 has the lowest static flexibility of 2.0 [mm].
However, this arm only has a payload and reach of 60 [kg] and 2,033 [mm]. Since
the full-scale robotic arm is intended to have a payload of 50 [kg] and reach of 5
[m], the robotic arms rated for larger payloads provide more comparable data. The
Kuka KR 240-2 and ABB IRB 7600 have a static flexibility of 3.4 [mm] and 3.5 [mm],
respectively. Therefore, the full-scale arm should have a static flexibility that is no
worse than 3.4 [mm].
FEA was performed on the two main links of the full-scale robotic arm since they
provide the majority of the reach for the robot. Therefore, the two main links were
to be designed such that they met the static flexibility requirement mentioned above.
Since only the two main links are being examined, the static flexibility of the base,
motors, and wrist configuration were not included. Therefore, the static flexibility of
the two main links was designed to be less than 3.4 [mm]. Since the two main links do
111
provide the majority of the reach to the robot, their contribution to the total static
deflection is significant. Thus, designing the two main links to have a static deflection
of 2.0 [mm] should ensure that the total static flexibility is no more than 3.4 [mm]
considering that the motors are very compact and undergo very little deflection and
that the hysteresis of the motors is insignificant [49].
Table 5.1: Static Flexibility of Four Commercial Robotic Arms
Payload (kg) Max Reach (mm) Static Flexibility (mm)
Kuka KR 60-3
60
2033
2.0
Staubli RX170
65
1835
7.0
Kuka KR 240-2
240
2701
3.4
ABB IRB 7600
400
2550
3.5
Considering that the full-scale robotic arm will be constructed using filament winding,
a virtual model in Abaqus should provide useful results. FEA was used to determine
the static flexibility of the two main links. It was used to test a variety of dimensions
for the links in order to achieve the desired static flexibility of 2.0 [mm]. In doing
so, the dimensions of the links were chosen such that the weight of the links was
minimized.
5.2.2
Testing Procedure
The static flexibility was determined for the worst position for deflection, the arm
extended straight out horizontally. Therefore, the two main links were modeled as
a single component. The attachment to the base link was represented by bounding
all faces of the initial 20 [cm] of the part as illustrated in Figure 5.3. The encastre
boundary condition was used to fully restrict this portion of the part.
As was the case for the FEA performed on the prototype, the payload of 80 [kg] was
represented by a surface traction load that was acting on the six vertical faces of the
link directed in the same direction of gravity. This force was applied on the final 20
[cm] section of the part. A gravity load was also applied to represent the weight of the
112
links. Since the links are to be made using filament winding, the density was assumed
to be 1,700 [kg/m3 ], equal to high modulus carbon fiber-epoxy composite [5].
A load was also used to represent the weight of the motor and gearhead for joint
three. The motor and gearhead have a mass of 50.6 [kg] and 5.5 [kg], respectively [48].
Therefore, a mass of 60 [kg] was used to represent their weight and the attachment
devices to the two main links. Additionally, this force was applied in the middle of
the length of the model since it is assumed that the two main links have an equal
length. The weight was presented as a pressure. The force was applied on an area
that represented the size of the motor. As shown in Appendix B.1, the base of the
motor is approximately 45 [cm] by 20 [cm]. Therefore, the pressure was applied on
this same area.
Figure 5.3: Loading Scenario for Model of Full-Scale Main Links
5.2.3
Results
5.2.3.1
Links with Height Twice the Width
The first design of component that represents the two main links (Design A) was
created with height, width, and wall thickness of 400 [mm], 200 [mm], and 3.4 [mm],
respectively. Based on the results from Section 2.2.3, it was designed with a constant
113
wall thickness and a negative 0.5 taper for the height and width. The aforementioned
values for height and width are those at the broad end of the part. The part was
designed with the same layering pattern that was used for the prototype; 0 , 45 , 0 ,
0 , 45 , 0 , 0 , 90 , 0 , 0 , 45 , 0 . Additionally, the edges of the component were
rounded with a radius of 10 [mm]. This was included because it helps reduce stress
concentrations resulting in a stronger link. The radius also helps aid in the removal
of the mold during production.
The results from the FEA for the main links with the aforementioned dimensions
revealed that the maximum static deflection was 4.804 [mm]. This is considerably
larger than the desired value of 2.0 [mm]. Therefore, the link was redesigned.
Since Design A had a static deflection that was more than two times as much as
desired, the link was redesigned to be as large as possible for Design B. As mentioned
in Section 1.2, the links of the robotic arm must be capable of fitting through a
0.76 [m] diameter opening. Therefore, the model representing the two main links
was designed to be as large as possible while still satisfying this requirement. The
broad end was designed to span 0.74 [m] between opposite corners such that 1 [cm]
of clearance is provided at all corners as illustrated in Figure 5.4. Using the same
design constraint that the height is twice the width, basic trigonometry reveals that
the angle is equal to tan1 2. Therefore, the maximum height and width that can
fit through the opening are 0.662 [m] and 0.331 [m], respectively.
The results from the FEA in Abaqus for Design B revealed that the links were overengineered, resulting in a maximum static deflection of 1.167 [mm]. Therefore, in
order to minimize the weight, the size of the links were designed to be smaller.
The height and width of the links for Design C were modified to be smaller than Design
B and larger than Design A. Considering the static deflection for Design B was closer
to the desired value than Design A, the link was redesigned to have dimensions closer
to the Design B than Design A. Therefore, the height and width at the broad end
114
Figure 5.4: Diagram of Largest Link That Fits Through 0.76 [m] Diameter Opening
of the part representing the two main links was defined as 600 [mm] and 300 [mm],
respectively. FEA determined that the static deflection was 1.499 [mm]. Therefore,
it too was over-engineered.
In order to determine the dimensions for the links that result in a static deflection of
2.0 [mm], the results for Designs A, B, and C were plotted as illustrated in Figure 5.5.
The static deflection of the link was plotted against its average cross-sectional area.
The results revealed that a non-linear relationship exists between the two variables.
According to the trend line mapping the points, a link with a cross-sectional area
of 6,500 [mm2 ] has a static deflection of 2.0 [mm] for the defined loading scenario.
Therefore, the link was redesigned such that Design D has a cross-sectional area of
6,500 [mm2 ]. Assuming a negative taper angle of 0.5 and that the height is twice the
width, Equation (2.8) for the cross-sectional area of the I-beam supported rectangular
structure is rearranged for the width, w, as:
w=
2Ac, avg + 436.35t + 12t2

16t
(5.4)
Therefore, assuming a wall thickness of 3.4 [mm], the width of the link should be
115
approximately 269 [mm] for a link with an average cross-sectional area of 6,500 [mm2 ].
Consequently, the height should be 538 [mm].
Figure 5.5: Average Cross-Sectional Area Versus Static Deflection for Links with
Width-to-Height Ratio of 2.0
The FEA procedure revealed that Design D had a static deflection of 2.054 [mm].
This is sufficiently close to the desired value of 2.0 [mm].
5.2.3.2
Relationship of Thickness and Static Deflection
The four previous links that were tested all had a wall thickness of 3.4 [mm]. Therefore,
a link with a different wall thickness was tested to determine the relationship between
the thickness and the static deflection.
The modified link, Design E, was defined to have a wall thickness of 5.1 [mm], one
and a half times larger than that of the previous links. In order to effectively test how
the variance of the wall thickness affected the static deflection, the other variables
were held constant. Therefore, the link was designed to have a cross-sectional area
116
of 6,500 [mm2 ] so that it is the same as Design D. Using Equation (5.4), the width
at the broad end of the part was determined to be 190.5 [mm] in order for the link
with 5.1 [mm] thick walls to have a cross-section equal to Design D. Consequently,
the required height of the link was 381 [mm].
The thicker link, Design E, was tested using the same procedure for FEA and it was
determined that it has a static deflection of 4.163 [mm]. Its static deflection was
considerably larger than the link with thinner walls. In fact, a 50% increase in the
wall thickness led to an increase in the static deflection of more than 100%. This
agrees with the results found for the optimization of the prototype link as discussed
in Section 2.3.9.4.
5.2.3.3
Links with Optimized Height-to-Width Ratio
According to the results found in Section 2.3.9.4, the static deflection of the I-beam
supported rectangular links was minimized when the height was 3.63 times as large
as the width. Therefore, the model that represents the two main links was designed
to this constraint.
For a link with a negative taper angle of 0.5 and a height-to-width ratio of 3.63, the
width is defined as:
w=
2Ac, avg + 436.35t + 12t2

25.78t
(5.5)
In order to determine how the variation of the height-to-width ratio affects the static
deflection, the model was designed to have a cross-sectional area and wall thickness
equal to that of the model with a height twice its width (Design D). Therefore, for a
cross-sectional area of 6,500 [mm2 ] and wall thickness of 3.4 [mm], the width at the
broad end of the link was 167 [mm]. Consequently, for a height-to-width ratio of 3.63,
the height of the link was 606 [mm].
The same FEA procedure was performed on Design F of the links and it was de117
termined that it had a static deflection of 1.859 [mm]. This is a lower value than
that found for the link with a height-to-width ratio of 2. Considering that the two
components were designed to have the same average cross-sectional area, links with
a height-to-width ratio of 3.63 do indeed have a higher stiffness-to-weight ratio than
those with a with a height-to-width ratio of 2.
The static deflection determined for Design F was however smaller than the desired
2.0 [mm], indicating that it was over-engineered. Therefore, in order to reduce the
weight of the links, the dimensions of the links were reduced for Design G.
The height at the broad end of the link was defined to be 538 [mm], such that it was
equal to that for the link with a height-to-width ratio of 2. For a width-to-height
ratio of 3.63, the width of the link was defined to be 148 [mm]. The results from the
FEA for Design G revealed that it had a static deflection that was too high at 2.650
[mm].
In order to achieve a static deflection of 2.0 [mm], the link was redesigned to be larger
than the Design G in order to improve its stiffness. The height and width of the link
were chosen to be slightly smaller than that of the original link that had a height-towidth ratio of 3.63 (Design F). Therefore, the height and width were defined to be 581
[mm] and 160 [mm] for Design H. It was determined that it had a static deflection of
2.106 [mm], slightly higher than the desired 2.0 [mm].
5.2.3.4
Different Layering Patterns
One of the major advantages of composite materials is that since their mechanical
properties are directional, they vary widely. Therefore, the stiffness of the links can
be improved by varying the layering pattern. Since the deflection of Design H for the
model representing the two main links is slightly above the desired value, alternate
layering patterns were investigated to reduce the static deflection.
The arm ratio as defined by Equation (3.1), is approximately 9.3 for Design H. Ac-
118
cording to Figure 1.8, approximately 19% of the layers should be oriented at 45 to

the longitudinal axis. Therefore, 2 of the 12 layers should be oriented in this direction.
Also, it was mentioned in Section 3.4.3 that one 90 layer should be included to help
reduce the hoop stress that exists at the joints. Therefore, the remaining 9 layers
were designed to be 0 .
Numerous different layering patterns with a repetitive order were investigated for the
links. The results of eight such layering patterns are shown in Table 5.2. The results
indicate that including one +45 layer and one -45 layer helps to improve both the
strength and stiffness of the links. Also, the results indicated that winding the layers
at a -45 angle before winding them at a +45 angle leads to a reduction in the stress.
Additionally, it was determined that it is best to have the 90 layer in between the
positive and negative 45 layers. The optimal fiber layer for reducing both the stress
and deflection was determined to be -45 , 0 , 0 , 0 , 0 , 0 , 90 , 0 , 0 , 0 , 0 , 45 .
Table 5.2: Stress and Deflection in Link for Different Layering Patterns
Layering Pattern
Von Mises Stress (MPa) Deflection (mm)
0, 0, 0, 45, 0, 0, 90, 0, 0, 45, 0, 0
12.59
1.913
0, 0, 0, -45, 0, 0, 90, 0, 0, -45, 0, 0
12.59
1.912
0, 0, 0, 45, 0, 0, 90, 0, 0, -45, 0, 0
11.28
1.892
0, 0, 0, -45, 0, 0, 90, 0, 0, 45, 0, 0
11.23
1.892
0, 0, 0, -45, 0, 0, 45, 0, 0, 90, 0, 0
11.30
1.893
0, 0, 0, 90, 0, 0, -45, 0, 0, 45, 0, 0
11.58
1.916
0, 0, 0, 0, 0, -45, 90, 45, 0, 0, 0, 0
12.61
1.913
-45, 0, 0, 0, 0, 0, 90, 0, 0, 0, 0, 45
10.87
1.892
Altering the layering pattern resulted in a reduction of the static deflection of 0.214
[mm]. This resulted in a link that was over-engineered with a static deflection of 1.892
[mm]. Therefore, in order to achieve a link that has the optimal layering pattern and
a static deflection of approximately 2.0 [mm], the link was designed to have a static
deflection of 2.214 [mm] for the original layering pattern.
The cross-sectional area was plotted against the static deflection for the three models
119
with a height-to-width ratio of 3.63 and original layering pattern (Designs F, G, and
H). As was the case for the links with a height-to-width ratio of 2, the cross-sectional
area is non-linearly related to the static deflection. Figure 5.6 illustrates that a link
with an average cross-sectional area of 6,075 [mm2 ] has a static deflection of 2.214
[mm].
Figure 5.6: Average Cross-Sectional Area Versus Static Deflection for Links with
Width-to-Height Ratio of 3.63
The model was redesigned to have an average cross-sectional area of 6,075 [mm2 ] for
Design I. Therefore, using Equation (5.5) the width was determined to be 157 [mm].
Therefore, the height was 570 [mm] such that the height-to-width ratio is 3.63. The
results from the FEA determined that the link with the original layering pattern had
a static deflection of 2.227 [mm]. However, for a layering pattern of -45 , 0 , 0 , 0 ,
0 , 0 , 90 , 0 , 0 , 0 , 0 , 45 , the static deflection was 2.000 [mm].
120
5.3
Discussion
The links were designed to have a static deflection of 2.0 [mm] while minimizing the
average cross-sectional area to 6,075 [mm2 ] by designing the links to have a layering
pattern of -45 , 0 , 0 , 0 , 0 , 0 , 90 , 0 , 0 , 0 , 0 , 45 , a constant wall thickness of
3.4 [mm], a negative taper angle of 0.5 , and a height and width at the broad end of
the first link of 570 [mm] and 157 [mm].
The mass of the two links can be calculated by rearranging Equation (4.2). The
volume of the member representing the links is defined as:
V = Ac, avg L
(5.6)
Therefore, for a length of 5 [m], an average cross-sectional area of 6,075 [mm2 ], and
a density of 1,700 [kg/m3 ], the mass of the main links, mmainlinks , is only 51.64 [kg].
Therefore, each link weighs only 25.82 [kg]. Using Equation (5.1) and (5.2), the torque
and required power of the motor at joint three was re-calculated as 2,278.6 [Nm], and
2,386.2 [W], respectively. In order to provide a safety factor of two, the motor should
be capable of providing 4,772.3 [W]. Therefore, Model B-404-C which is capable of
5,400 [W] was selected (See Appendix B.1).
Since the motor operates at a speed of 5,000 [rpm], in order to slow the speed down
to 60 [ /s] the gearhead ratio that is required was determined to be 500:1 using
Equation (5.3). Gearhead DTR60-500 was chosen because it has a ratio of 500:1 and
its continuous rated torque at 5,000 [rpm] for 10,000 hour life is 18 [Nm], whereas
the continuous stall torque of the B-404-C motor is 13.1 [Nm] while operating at full
speed of 5,000 [rpm] (See Appendix B).
The free body diagram for the loading scenario for the motor that operates the first
main link (joint 2) is illustrated in Figure 5.7. In order for static equilibrium, the
torque provided by the second motor, Mmotor2 , is:
121

Mmotor2 = LLink
WLink1
3WLink2
+ WJoint3 +
+ 2WEE
2
2

(5.7)
where LLink is the length of the first main link (which is also the length of the second
main link), WLink1 is the weight of the first main link, and WJoint3 is the weight of
the motor and gearhead at joint 3. The mass of joint three is assumed to be 20 [kg]
since the motor is 12.5 [kg] and gearhead is 2.7 [kg] and an attachment bracket is
also necessary. Therefore, with LLink equal to 2.5 [m] and WLink1 and WLink2 both
25.82 [kg], the required torque of motor two is determined to be 5,681 [Nm]. Using
Equation (5.2) and assuming a speed of 60 [ /s], the required power of the motor is
5,949.1 [W]. Therefore, including a factor of safety of two, the motor must be capable
of 11,898.2 [W]. Therefore, motor B-804-B which weighs 50.6 [kg] was selected since
it can provide 13,900 [W] of power. Since it operates at 2,000 [rpm] and a continuous
stall torque of 78.6 [Nm], gearhead DTR90-200 which weighs 5.5 [kg] and has a gear
ratio of 200:1 was chosen since it can slow the motor down to 60 [ /s]. Furthermore, its
continuous rated torque at 3,000 [rpm] for 10,000 hour life is 88 [Nm] (See Appendix
B).
Figure 5.7: Free Body Diagram of Second Main Link

The total mass, mtotal , of the robotic arm is defined as:
mtotal = mbase + mmain links + m2 + m3 + mwrist

122
(5.8)
where mbase is the mass of the base joint and first motor, m2 is the mass of joint
two, m3 is the mass of joint three, and mwrist is the mass of the wrist configuration.
Therefore, with m3 weighing 60 [kg] assuming that the weight of the attachment
bracket is small, m2 weighing 20 [kg], mmainlinks equal to 51.64 [kg], and mwrist equal
to 30 [kg], the mass of the robotic arm is 484.92 [kg] assuming that mbase represents
two thirds of the total mass of the robotic arm.
By comparing the calculated mass of the full-scale robotic arm to those listed in
Table 1.1, it is clear that the use of carbon fiber-epoxy composite provides a dramatic
reduction in weight. The calculated mass is similar to the mass for robotic arms that
have a payload of around 50 [kg] and reach of 2 [m]. However, the full-scale composite
arm has the same payload but a reach of 5 [m]. This indicates that the robotic arm
was successfully designed for high stiffness-to-weight ratio in order to minimize the
total weight.
5.4
Summary
In this chapter the full-scale arm was designed with the use of Abaqus. It was determined that constructing the links with a height-to-width ratio of 3.63 improves the
stiffness-to-weight ratio over those with a height-to-width ratio of 2.0. Additionally,
it was determined that minimizing the thickness does improve the stiffness-to-weight
ratio. Therefore the height and width at the broad end was chosen to be 570 [mm],
157 [mm], respectively. Additionally, the link was chosen to include a negative taper
angle of 0.5 and constant wall thickness of 3.4 [mm]. Using a layering pattern of
-45 , 0 , 0 , 0 , 0 , 0 , 90 , 0 , 0 , 0 , 0 , 45 , a static deflection of 2.000 [mm] was
achieved while minimizing the weight of the two main links to 51.64 [kg].
123
Chapter 6
Conclusions and Recommendations
for Future Work
6.1
Conclusions
This thesis outlines the design of a novel 6-DOF robotic arm made of carbon fiberepoxy composite material. This is envisioned to present a significant contribution to
the robotic industry because even though the use of composite materials for robotic
applications has been investigated over the last 20 years, a 6-DOF lightweight, longreach robotic arm that is constructed primarily of a polymer composite material has
not been designed as of yet.
A 6-DOF robotic arm with a payload of 50 [kg] and reach of 5 [m] was designed to be
made of a polymeric composite material because it can help alleviate the limitations
in terms of weight, speed, and repeatability of large-scale steel and aluminum robotic
arms. Carbon fiber was selected as the support material because it is favourable
over other alternatives in terms of stiffness due to its high tensile modulus and low
ultimate tensile strain. The arm was designed to be constructed using long-strand
carbon fibers because they provide further improvements as discussed in Chapter 1.
124
The robotic arm was designed to have a serial configuration with a kinematicallysimple layout. This design provides the best opportunity for creating a lightweight
robotic arm since it does not have any unnecessary offsets or link lengths between
joints [38]. This is best-suited for designing the links with carbon fiber-epoxy composite since it is an expensive material. The main-arm configuration that was selected for
the 6-DOF robotic arm was R R k R, where the first revolute joint is perpendicular
to the second revolute joint that is parallel to the third revolute joint. This layout
is beneficial because revolute joints require links with less material than prismatic
joints. Additionally, this layout has a very large workspace and minimal torsional
forces acting on it which make it possible to construct a robotic arm to the required
specifications while keeping the weight to a minimum.
The use of an internal support structure in the design of the links for the robotic
arm can offer improvements in the stiffness-to-weight ratio. Therefore, the links were
designed to have an I-beam supported rectangular cross-section in order to produce
a lightweight robotic arm with the desired stiffness.
A 35% size prototype of the main-arm of the 6-DOF robotic arm was constructed.
The design focused on the second and third links of the arm because they provide
the majority of the reach for the robotic arm with the selected manipulator layout.
Optimization was performed on these two I-beam supported rectangular linkages of
the prototype. It was determined that for minimal deflection the thickness should be
minimized and a height-to-weight ratio of 3.63 should be used.
The 35% size main-arm proof-of-concept prototype was constructed for testing purposes. One of the biggest contributions of this thesis work was the creation of a novel
production process. This entails producing a series of hollow modular base units,
ideally by filament winding (but for reasons of economics by hand lay-up for the
construction of the prototype), and then winding around the entire series to create
the part with an internal support structure. Using this method, two identical small
125
rectangular base units were first constructed. They were then bonded together and
exterior layers of carbon fiber were wrapped around them. This technique proved to
be very efficient.
Various tests were performed on the carbon fiber-epoxy composite prototype links.
Destructive testing indicated that under normal operation the links could easily withstand the applied forces. When the applied forces were increased such that fracture of
the link occurred it was determined that the composite material still maintains a portion of its strength and the links are able to retain their form to prevent catastrophic
failure.
A 2-DOF arm was created using the prototype links that were constructed. The arm
was operated and it was determined that the high damping capabilities of carbon
fiber-epoxy composite enabled a high repeatability.
The static deflection of the prototype composite link was tested. It was determined
that under full load, the link underwent a maximum deflection of 0.454 [mm]. The
results were verified by creating a virtual model of the link and performing FEA using
Abaqus software. The analysis determined that the maximum deflection was 0.09195
[mm]. The difference in deflections is due to problems encountered during the hand
lay-up production of the links.
The full-scale robotic arm is intended to be produced using filament winding. Since
this is an automated process it reduces the amount of human error that exists. Therefore, the links for the full-scale arm were designed using Abaqus software. Links of
varying dimensions were tested in order to provide a total static deflection of the links
of 2 [mm] while minimizing the weight. It was determined that for minimal weight,
the links should be constructed with a constant wall thickness of 3.4 [mm], a negative
taper angle of 0.5 , and a height and width at the broad end of the first main link of
570 [mm] and 157 [mm], respectively, such that the height-to-width ratio is 3.63. Additionally, it was determined that for maximum strength and stiffness the link should
126
be constructed using a layering pattern of -45 , 0 , 0 , 0 , 0 , 0 , 90 , 0 , 0 , 0 , 0 ,

45 .
The combined weight of the two main links for minimal weight was determined to be
51.64 [kg]. Therefore, constructing the 6-DOF robotic arm out of carbon fiber-epoxy
composite leads to a robotic arm with a total weight of approximately 485 [kg]. This
weight is comparative to commercial robotic arms that have a payload of 50 [kg] and
a reach of 2 [m]. Therefore, the use of carbon fiber-epoxy composite material enables
an increase in the reach of 3 [m] without increasing the total weight when compared
to current metal arms.
6.2
Recommendations for Future Work
The 6-DOF robotic arm is intended to be used for operation on airplanes and other
large structures. Therefore, a method for moving the entire robotic arm needs to
be investigated. For this reason, it would be beneficial to investigate the feasibility
of using carbon fiber-epoxy composite for the production of the base link and wrist
configuration since the work outlined in this thesis indicated that using this composite material for the production of the two main links resulted in dramatic weight
reductions.
The motor selection that was performed was used only as a guideline so that the
full-scale robotic arm could be modeled in Abaqus. In-depth investigation of motors
needs to be performed in order to select appropriate motors for operation of all the
links. Simultaneously, FEA needs to be performed on the virtual model of the robotic
arm in order to design the links such that the robotic arm has the desired stiffness
under these new loading conditions.
One of the major obstacles to the use of polymer composite materials in robotic applications is the difficulty in joining the motors to the links. Therefore, research needs
127
to be performed in joining techniques. Since drilling composite material leads to a reduction in their strength, non-destructive joining techniques need to be investigated.
One such method is using a carbon fiber-epoxy composite joint between the link and
the motor. This idea shows great potential because it permits the possibility of creating the link around the motor such that is essentially one unified part. However,
further research needs to be done pertaining to the design and production procedure.
Finally, after the completion of the above, a full-scale prototype of the arm needs to
be built. Extensive testing of the full-scale prototype must be conducted to ensure
the design is safe and robust for industrial applications.
128
References
[1] KUKA,
Industrial Robots - Products, On the WWW, 2009,
URL
http://www.kuka-robotics.com/usa/en/products/.
[2] FANUC
Industrial
Robotics
Robots
America
A-Z,
Inc.,
On
Products
the
WWW,
Robots
2009,
URL
http://www.fanucrobotics.com/Products/Robots/Atoz.aspx.
[3] ABB
Robots,
Robotics,
On
the
WWW,
2009,
URL
http://www.abb.com/product/us/9AAC100735.aspx.
[4] Motoman,
2009,
Material
URL
Handling
Robots,
On
the
WWW,
http://www.motoman.com/products/robots/application-
specific/materialhandling.php.
[5] W.D. Callister, Materials Science and Engineering; an Introduction, John Wiley
and Sons, 6th edition, 2003.
[6] A.K. Kaw, Mechanics of Composite Materials, CRC Press, 2nd edition, 2006.
[7] S.N. Bhattacharya, R.K. Gupta, and M.R. Kamal, Polymeric Nanocomposites;
Theory and Practice, Carl Hanser Publishers, 2008.
[8] J.H. Koo, Polymer Nanocomposites; Processing, Characterization, and Applications, McGraw-Hill, 2006.
129
[9] H.D. Boggs, Method of making a reinforced hollow cylindrical article, United
States Patent, 1957, US Patent No. 2,785,442.
[10] E.C. Young, Method of making filament wound structural columns, United
[11] P. Bourrieres, Apparatus for forming a filament-wound structure on a cylindrical
pole, United States Patent, 1989, US Patent No. 4,878,984.
[12] G.H. Holtom, Manufacturing of boat hulls and other hollow articles, United
[13] R.J. Gardiner, Method of making a curved composite beam, United States
Patent, 1985, US Patent No. 4,512,835.
[14] D.K. Buckmiller, D.W. Erickson, D.G. Olsen, and R.J. Philpot, Vehicle fork,
United States Patent, 1997, US Patent No. 5,609,349.
[15] Y. Adachi, Composite beam having a hollow cross section, United States
[16] C. Ashton, Filament wound structural columns for light poles, United States
[17] M.B. Lloyd, Reinforced tubing, United States Patent, 1915, US Patent No.
1,141,067.
[18] H.H. Derr and B.G. Gentry, Panel with triangular cross-section foam core
elements, United States Patent, 1967, US Patent No. 3,339,326.
[19] L.H. Corzine, Cellular foam core assembly, United States Patent, 1970, US
Patent No. 3,544,417.
130
[20] C. Dumlao et al., Modular polymer matrix composite support structure and
methods of constructing same, United States Patent, 2000, US Patent No.
6,081,955.
[21] L.J. Ashton and D.P. Abildskov, Light weight contoured load carrying structure, United States Patent, 1973, US Patent No. 3,779,487.
[22] J. Brogan, Truss core panels, United States Patent, 1980, US Patent No.
4,223,053.
[23] W.C. Chapman Jr., Tubular members integrated to form a structure, United
[24] K.B. Bowman, Modular fiber reinforced plastic enclosed bridge, United States
Patent, 2000, US Patent No. H1,872.
[25] S.B. Choi, B.S. Thompson, and M.V. Gandhi, Modeling and Control of a
Single-Link Flexible Manipulator Featuring a Graphite-Epoxy Composite Arm,
in IEEE CH2876-1, 1990, pp. 1450 1455.
[26] C.S. Lee, J.H. Oh, D.G. Lee, and J.H. Choi, A Composite Cantilever Arm for
Guding a Moving Wire in an Electrical Discharge Wire Cutting Machine, in
Journal of Materials Processing Technology, 2001, vol. 113, pp. 172 177.
[27] J.H. Oh, D.G. Lee, and H.S. Kim, Composite Robot End Effector for Manipulating Large LCD Glass Panels, in Journal of Composite Structures, 1999,
vol. 47, pp. 497 506.
[28] C.S. Lee and D.G. Lee, Manufacturing of Composite Sandwich Robot Structures
Uisng the Co-Cure Bonding Method, in Journal of Composite Structures, 2004,
vol. 65, pp. 307 318.
131
[29] L.B. Lessard, V. Hayward, and M.M. Roy,
Design and Fabrication of a
Graphite/Epoxy Link for a Lightweight Robotic Manipulator, in European Conference on Composite Materials, 1998, vol. 8, pp. 535 542.
[30] K. Krishnamurthy, K. Chandrashekhara, and S. Roy, A Study of Single-Link
Robots Fabricated from Orthotropic Composite Materials, in Computers and
Structures, 1990, vol. 36, pp. 139 146.
[31] G. Caprino and A. Langella, Optimization of Robotic Arms Made of Composite
Materials for Maximizing Fundamental Frequency, in Journal of Composite
[32] S. Oral and S.K. Ider, Coupled Rigid-Elastic Motion of Filament-Wound Composite Robotic Arms, in Computer Methods in Applied Mechanics and Engineering, 1997, vol. 147, pp. 117 123.
[33] A Ghazavi and F. Gordaninejad, A Comparison of the Control of a Flexible
Robot Arm Constructed from Graphite/Epoxy Versus Aluminum, in Computers
and Structures, 1995, vol. 54, pp. 621 632.
[34] G. Musch, C. Nageli, and K.H. Sprenger, Fibre Composite Robot Arms, in
Kunsttoffe, 2007, pp. 62 66.
[35] CBW Automation, Fully Automated Systems, On the WWW, 2009, URL
http://www.cbwautomation.com/index.php.
[36] Azo Materials, Worlds First Carbon Fiber Composite Robot - The KUKA
Robotics Palletizing Robot, On the WWW, 2005, URL http://www.kukarobotics.com/en/products/industrial robots/high/kr180 2 pa/start.htm.
[37] D. Willis, R. Pop-Iliev, and S. Nokleby,
Design Concepts for Developing
Composite-based Long-reach Robotic Arms, in SPE, ANTEC, 2008, pp. 683

687.
132
[38] S. Nokleby and R. Podhorodeski, A Complete Family of Kinematically-Simple

Joint Layouts: Layout Models, Associated Displacement Problem Solutions and
Applications,, in Journal of Industrial Robotics: Theory Modeling and Control,
2006, pp. 237 263.
[39] D.A. Saravanos and J.S. Lamancusa, Optimum Structural Design of Robotic
Manipulators with Fiber Reinforced Composite Materials, in Computers and
[40] L. Librescu and O. Song, Thin-Walled Composite Beams, Springer, 2006.
[41] R.C. Hibbeler, Mechanics of Materials, Prentice Hall, 6th edition, 2005.
[42] SCHUNK,
Rotary Module Products, On the WWW, 2009,
URL
http://www.schunk.com/schunk files/attachments/RotaryModules Part2 0408

190-299 EN.pdf.
[43] Efunda: Engineering Fundamentals, Area and Perimeter of a Triangle, On the
WWW, 2009, URL http://www.efunda.com/math/areas/triangleGen.cfm.
[44] The Mathworks,
Optimization Toolbox, On the WWW, 2009,
URL
http://www.mathworks.com.
[45] Graphite Store,
Carbon Fiber Laminates, On the WWW, 2009,
URL
http://www.graphitestore.com.
[46] D.A. Willis, S.B. Nokleby, and R. Pop-Iliev, Development of a Composite-Based
Long Reach Robotic Arm, in CCToMM, Symposium on Mechanisms, Machines,
and Mechatronics, 2009.
[47] G.M. Stanley, Continuum-Based Shell Elements, PhD Dissertation, Department
of Mechanical Engineering, Stanford University, 1985.
133
[48] Danaher Motion International, Product overview, On the WWW, 2008, URL
http://www.danahermotion.com/website/com/eng/products.php.
[49] M. Summers, Robot Capability Test and Development of Industrial Robot
Positioning System for the Aerospace Industry, Tech. Rep. 2005-01-3336, Airbus
UK, 2005.
134
Appendix A
MATLAB Code
A.1
Declaration of Constants
%Program: constants.m
%Declares constants that are necessary to perform calculations needed in determining
the optimal cross-section for minimal deflection.
function [L, g, rho, m motor, W, M max, eff, v, E] = constants
L=1.75; %Total length (m) of the two main links.
g=9.81; %Gravitational constant (m/s2).
rho=1700; %Density (kg/m3) of material that links are made of.
m motor=4.7; %Mass (kg) of motor to operate 2nd main link.
W=17.5; %Payload (kg) to be supported by robotic arm.
M max=425; %Maximum torque (Nm) of motor operating 1st main link.
eff=0.90; %Efficiency () of motor operating 1st main link.
v=0.25; %Poissons ratio () of material that links are made of.
E=220e9; %Elastic modulus (Pa) of material that links are made of.
135
A.2
Objective Function
%Program: objfun.m
%Determines the end-effector deflection of a rectangular cross-section R R k R
robotic arm.
function deflection = objfun(x)
%User inputs. Height, width, and thickness of rectangular cross-section.
h=x(1);
w=x(2);
t=x(3);
%Input: m.file for determining cross-sectional area and moment of inertia for specific
cross-sectional geometry.
%Output: cross-sectional area and moment of inertia of the links.
[Ac] = A I beam sup(h, w, t);
[I] = I I beam sup(h, w, t);
%Input: m.file containing constants of robotic arm under examination.
%Output: values of constants necessary to perform the following calculations.
[L, g, rho, m motor, W, M max, eff, v, E] = constants;
L wrist=L/10; %Length (m) of wrist of robotic arm.
G=E/(2*(1+v)); %Shear modulus (Pa) of material that links are made of.
T=g*W*L wrist; %Internal torque (Nm) acting on links at wrist location.
%Thickness (m) of a hollow rectangular cross-sectional with an equal cross-sectional
area to that of the link being examined.
t equiv=(2*(w+h)-sqrt((2*(w+h))2-16*Ac))/8;
%Angle (radians) of twist of the two main links.
136
phi=(T*L*(2*(w-t equiv)+2*(h-t equiv)))/

(4*G*((w-t equiv)*(h-t equiv))2*t equiv);
%Displacement as a result of twisting of main links due to torsional loading.
dy=L wrist*sin(phi)*cos(phi);
%Displacement due to bending of the two main links.
v mainlinks=((g*rho*Ac*L4)/(8*E*I))+((5*g*m motor*L3)/(48*E*I))
+((g*W*L3)/(3*E*I));
%Displacement due to bending of wrist.
v wrist=((g*W*L wrist3)/(3*E*I));
%Total deflection experienced by the end-effector.
deflection=v mainlinks + v wrist + dy;
A.3
Non Linear Constraint
%Program: nonlincon.m
%Declares the non-linear inequality and equality constraints.
function [c, ceq] = nonlincon(x)
%User inputs. Height, width, and thickness of rectangular cross-section.
h=x(1);
w=x(2);
t=x(3);
%Input: m.file for determining cross-sectional area for specific geometry.
%Output: cross-sectional area of the links.
[Ac] = A I beam sup(h, w, t);
137
%Input: m.file containing constants of robotic arm under examination.

%Output: values of constants necessary to perform the following calculations.
[L, g, rho, m motor, W, M max, eff, v, E] = constants;
%Non-linear inequality constraint based on ensuring that the moment of the arm
doesnt exceed the amount that the motor can safely withstand.
c = [Ac+((2*((m motor/2) + W))/(L*rho))-((2*M max*eff)/(g*rho*L2))];
%Non-linear equality constraints.
ceq = [];
A.4
Area of I-beam Supported Rectangular Link
%Program: A I beam sup.m

%Determines the cross-sectional area (m2) for a rectangular cross-section with an
I-beam support structure. Using the user inputs of height, width, and thickness it
outputs a value for the variable Ac.
function Ac = A I beam sup(h, w, t)
Ac = 2*w*t + 3*(h-2*t)*t;
A.5
Inertia of I-beam Supported Rectangular Link
%Program: I I beam sup.m

%Determines the moment of inertia (m4) for a rectangular cross-section with an Ibeam support structure. Using the user inputs of height, width, and thickness it
outputs a value for the variable I.
function I = I I beam sup(h, w, t)
138
I = ((w*h3)/12) - ((((w-3*t)/2)*(h-2*t)3)/6);
A.6
Area of X-Supported Rectangular Link
%Program: A X sup.m
%Determines the cross-sectional area (m2) for a rectangular cross-section with an X
support structure. Using the user inputs of weight, width, and thickness it outputs a
value for the variable Ac.
function Ac = A X sup(h, w, t)
Ac = w*h - ((w-(2+(1/(cosd(atand(w/h)))))*t)*
(h/2-(1+((cosd(atand(w/h)))/(cosd(atand(h/w)-atand(w/h)))))*t))
- ((h-(2+(1/(cosd(atand(h/w)))))*t)*
(w/2-(1+((sind(atand(w/h)))/(cosd(atand(h/w)-atand(w/h)))))*t));
A.7
Inertia of X-Supported Rectangular Link
%Program: I X sup.m
%Determines the moment of inertia (m4) for a rectangular cross-section with an X
support structure. Using the user inputs of height, width, and thickness it outputs a
value for the variable I.
function I = I X sup(h, w, t)
I=((w*h3)/12) - (2*((((w-(2+(1/(cosd(atand(w/h)))))*t)*
(h/2-(1+((cosd(atand(w/h)))/(cosd(atand(h/w)-atand(w/h)))))*t)3)/36)
(2+(1/(cosd(atand(w/h)))))*t)*
(h/2-(1+((cosd(atand(w/h)))/(cosd(atand(h/w)-atand(w/h)))))*t)*
(h/2-(((h/2-(1+((cosd(atand(w/h)))/
139
+ ((w-
(cosd(atand(h/w)-atand(w/h)))))*t)/3)+t))2)/2))
- (2*(((h-(2+(1/(cosd(atand(h/w)))))*t)3*
(w/2-(1+((sind(atand(w/h)))/(cosd(atand(h/w)-atand(w/h)))))*t))/48));
A.8
Linear Constraints
%Program: A lincon.m
%The matrix A developed from the left hand side of the equations from the two linear
inequality constraints.
function A = A lincon
A = [0, 1, -4, 1, 1, 1, 1, 1; 1, -1, 0, 1, 1, 1, 1, 1]
%Program: b lincon.m
%The vector b developed from the right hand side of the equations from the two linear
inequality constraints.
function b = b lincon
b = [0; 0]
A.9
Output for I-beam Supported Rectangular Link
>> x0 = [0.7, 0.5, 0.003]; % Make a starting guess at the solution

options = optimset(Largescale,off);
[x,fval,exitflag] = ...
fmincon(@objfun,x0,[0,-1,4;-1,1,0],[0;0],[],[],[0.045, 0.045, 0.002],[0.14, 0.14,0.01],
@nonlincon,options)
140
Optimization terminated: first-order optimality measure less than options.TolFun and

maximum constraint violation is less than options.TolCon.
Active inequalities (to within options.TolCon = 1e-006):
lower
upper
1
ineqlin
ineqnonlin
2
x=
0.1400
0.1400
0.0024
fval =
3.7146e-004
exitflag =
1
>> x(2)-x(1)
ans =
0
>> 4*x(3)-x(2)
ans =
-0.1305
>> [c] = nonlincon(x)
c=
141
-7.0758e-011
A.10
Output for X-Supported Rectangular Link

fmincon(@objfun,x0,[0,-1,4;-1,1,0],[0;0],[],[],[0.045, 0.045, 0.002],[0.14, 0.14,0.01],
@nonlincon,options)
Optimization terminated: first-order optimality measure less than options.TolFun and
maximum constraint violation is less than options.TolCon.
lower
3
upper
ineqlin
ineqnonlin
x=
0.1400
0.1052
0.0020
fval =
4.4736e-004
exitflag =
1
>> x(2)-x(1)
ans =
142
-0.0348
>> 4*x(3)-x(2)
ans =
-0.0972
c=
-4.1702e-010
A.11
Output for I-beam Supported Rectangular

Link Without Upper Bounds

fmincon(@objfun,x0,[0,-1,4;-1,1,0],[0;0],[],[],[0.045, 0.045, 0.002],[1, 1,0.01],
@nonlincon,options)
Optimization terminated: magnitude of directional derivative in search direction less
than 2*options.TolFun and maximum constraint violation is less than options.TolCon.
lower
3
upper
ineqlin
ineqnonlin
x=
143
0.23350
0.064240
0.0020
fval =
2.0192e-004
exitflag =
5
>> x(2)-x(1)
ans =
-0.16926
>> 4*x(3)-x(2)
ans =
-0.05624
c=
-7.2891e-006
A.12
Output for X-Supported Rectangular Link

Without Upper Bounds

144
fmincon(@objfun,x0,[0,-1,4;-1,1,0],[0;0],[],[],[0.045, 0.045, 0.002],[1, 1,0.01],
@nonlincon,options)
Optimization terminated: magnitude of directional derivative in search direction less
than 2*options.TolFun and maximum constraint violation is less than options.TolCon.
lower
upper
ineqlin
ineqnonlin
x=
0.18479
0.049934
0.0020
fval =
3.4535e-004
exitflag =
5
>> x(2)-x(1)
ans =
-0.13486
>> 4*x(3)-x(2)
ans =
-0.04193
145

c=
-1.9566e-007
146
Appendix B
Danaher Data Sheets
B.1
Motors
147
S E R V O
S Y S T E M S
PARAMETER
SYMBOL
UNITS
Horsepower
Kilowatts
Speed at Rated Power
Max Operating Speed
Cont. Torque (Stall) at 40C
Cont. Line Current
Peak Torque
Peak Line Current
Max Theoretical Acceleration
(B, BE, EB-40/60x)
(M, ME-40/60x)
Torque Sensitivity (Stall) 10%
Back EMF (Line to Line) 10%
Max Line-to-Line volts
DC Res at 25C
(Line-to-Line) 10%
Inductance (Line-to-Line) 30%
Rotor (B, BE, EB-40/60x)
Inertia (M, ME-40/60x)
Weight (B, BE-40/60x)
(M, ME-40/60x)
(EB-40/60x)
Static Friction
Thermal Time Constant
(B, BE, M, ME-40/60x)-(EB-40/60x)
HPRATED
kWRATED
NRATED
NMAX
Tcs
Tcs
Ics
Tps
Ips
Z
hp
kW
rpm
rpm
N-m (lb-ft)
N-m (lb-ft)
ARMS
N-m (lb-ft)
ARMS
rad/sec2
Kt
Kb
VMAX
N-m (lb-ft)/ARMS
VRMS / krpm
VRMS
Rm
Lm
Jm
ohms
mH
kg-m2
lb-ft-sec2
kg (lb)
B/M 60X & 80X SERIES MOTORS
Viscous Damping Z Source

Motor Constant at 25C
Thermal Resistance at Stall
Number of Poles
B-404-D
BE-404-D
M-404-D
ME-404-D
EB-404-D
B-406-A
BE-406-A
M-407-A
ME-407-A
EB-406-A
B-406-B
BE-406-B
M-407-B
ME-407-B
EB-406-B
B-406-C
BE-406-C
M-407-C
ME-407-C
EB-406-C
B-602-A
BE-602-A
M-603-A
ME-603-A
EB-602-A
B-602-B
BE-602-B
M-603-B
ME-603-B
EB-602-B
2.7
2.0
1500
1500
13.0 (9.6)
13.8 (10.2)
6.0
35.9 (26.5)
16.4
4.5
3.4
2500
2500
13.3 (9.8)
14.1 (10.4)
9.9
36.6 (27.0)
28.8
7.3
5.4
5000
5000
13.1 (9.7)
13.9 (10.3)
19.8
35.3 (26)
55.9
6.0
4.5
3700
3700
13.1 (9.7)
14 (10.3)
15.0
37.5 (27.6)
45
3.9
2.9
1700
1700
17.6 (13.0)
18.7 (13.8)
9.5
48.5 (35.8)
27.3
7.4
5.5
3200
3200
18.6 (13.7)
19.7 (14.5)
19.1
49.5 (36.5)
53.3
9.6
7.2
5000
5000
17 (12.5)
18 (13.3)
27.2
48.3 (35.6)
81.4
4.4
3.3
2000
2000
17.6 (13.0)
18.7 (13.8)
10.0
51.2 (37.7)
30.5
7.7
5.7
4000
4000
17.4 (12.8)
18.4 (13.6)
20.0
49.8 (36.7)
61.4
54800
8150
2.31 (1.70)
139
250
55800
8310
1.34 (0.99)
81.2
250
53700
8000
0.66 (0.49)
40.2
250
57000
8490
0.877 (0.647)
53
250
52300
7380
1.87 (1.38)
113
250
53300
7530
0.98 (0.72)
58.8
250
52000
7340
0.63 (0.46)
37.7
250
49700
6590
1.77 (1.30)
107
250
48400
6420
0.85 (0.63)
51.6
250
4.1
102
0.000656-0.000484
0.00441-0.00325
12.5 (27.5)
15.5 (34)
12.5 (27)
0.26 (0.19)
1.32
33.5
0.000656-0.000484
0.00441-0.00325
12.5 (27.5)
15.5 (34)
12.5 (27)
0.26 (0.19)
0.34
8.4
0.000656-0.000484
0.00441-0.00325
12.5 (27.5)
15.5 (34)
12.5 (27)
0.26 (0.19)
0.63
15.0
0.000656-0.000484
0.00441-0.003250
12.5 (27.5)
15.5 (34)
12.5 (27)
0.26 (0.19)
1.7
42
0.000929-0.000685
0.00657-0.00485
15.9 (35)
20 (44)
21.5 (47.6)
0.287 (0.212)
0.44
12
0.000929-0.000685
0.00657-0.00485
15.9 (35)
20 (44)
21.5 (47.6)
0.287 (0.212)
0.20
4.8
0.000929-0.000685
0.00657-0.00485
15.9 (35.0)
20 (44)
21.5 (47.6)
0.287 (0.212)
1.55
0.382
32
9
0.001028-0.000758 0.001028-0.000758
0.00775-0.00572
0.00775-0.00572
16.8 (37)
16.8 (37)
20 (44)
20 (44)
16.8 (37)
16.8 (37)
0.49 (0.36)
0.49 (0.36)
N-m (lb-ft)
minutes
25-9
25-9
25-9
25-9
28-12
28-12
28-12
28-12
28-12
Fi
Km
Rth
N-m (lb-ft)/krpm
N-m (lb-ft)/ watts
0.018 (0.013)
0.926 (0.683)
0.35
4
0.018 (0.013)
1.01 (0.746)
0.40
4
0.018 (0.013)
0.982 (0.724)
0.38
4
0.018 (0.013)
0.952 (0.702)
0.36
4
0.020 (0.015)
1.23 (.907)
0.33
4
0.020 (0.015)
1.27 (0.937)
0.32
4
0.020 (0.015)
1.21 (0.892)
0.35
4
0.072 (0.053)
1.22 (0.903)
0.33
6
0.072 (0.053)
1.22 (0.898)
0.33
6
B-602-C
BE-602-C
M-603-C
ME-603-C
EB-602-C
B-604-A
BE-604-A
M-605-A
ME-605-A
EB-604-A
B-604-B
BE-604-B
M-605-B
ME-605-B
EB-604-B
B-604-C
BE-604-C
M-605-C
ME-605-C
EB-604-C
B-606-A
BE-606-A
M-607-A
ME-607-A
EB-606-A
B-606-B
BE-606-B
M-607-B
ME-607-B
EB-606-B
B-606-C
BE-606-C
M-607-C
ME-607-C
EB-606-C
B-606-D
BE-606-D
M-607-D
ME-607-D
EB-606-D
B-802-A
BE-802-A
M-803-D
ME-803-D
EB-802-D
6.1
4.6
3000
3000
17.4 (12.8)
18.4 (13.6)
15.0
49.5 (36.5)
45
8.0
6.0
2150
2150
30.1 (22.2)
31.9 (23.5)
19.0
86.4 (63.7)
57.4
10.4
7.7
3150
3150
30.1 (22.2)
31.8 (23.5)
27.7
87.7 (64.7)
84.8
11.5
8.6
4300
4300
31.2 (23)
33.6 (24.4)
39.4
86.4 (63.7)
114.8
8.8
6.6
1550
1550
44.8 (33.0)
47.5 (35.0)
20.0
132 (97.2)
62
15
11.2
3050
3050
42 (31)
44.6 (32.9)
37.5
126 (93)
118.6
14.2
10.6
4150
4160
44.8 (33)
47.5 (35)
54.8
124 (91.6)
160
11
8.2
2300
2300
42 (31)
44.6 (32.9)
28
123 (90.4)
86.2
10.6
7.9
2000
2000
42 (31)
44.6 (32.9)
24.9
130 (96)
81
48400
6420
1.15 (0.85)
70
250
42500
6140
1.59 (1.17)
95.8
250
43100
6240
1.09 (0.80)
65.9
250
42500
6140
0.79 (0.58)
47.9
250
43400
6220
2.24 (1.65)
135.0
250
41500
5960
1.12 (0.83)
67.7
250
40900
5870
0.82 (0.60)
49.5
250
40400
5790
1.2 (1.1)
90.5
250
26700
3710
1.69 (1.25)
102
250
674
14
0.001028-0.000758
0.00775-0.00572
16.8 (37)
20 (44)
16.8 (37)
0.49 (0.36)
0.508
13.3
0.002034-0.001500
0.01406-0.01037
23.1 (51)
28.7 (63)
23.1 (51)
0.52 (0.38)
0.234
6.3
0.002034-0.001500
0.01406-0.01037
23.1 (51)
28.7 (63)
23.1 (51)
0.52 (0.38)
0.126
3.3
0.0020342-0.001500
0.01406-0.01037
23.1 (51)
28.7 (63)
23.1 (51)
0.52 (0.38)
0.500
15.6
0.00304-0.00224
0.02117-0.01561
29.9 (66)
37.8 (83)
29.9 (66)
0.941 (0.694)
0.140
3.8
0.00304-0.00224
0.02117-0.01561
29.9 (66)
37.8 (83)
29.9 (66)
0.941 (0.694)
0.076
2.1
0.00304-0.00224
0.02117-0.01561
29.9 (66)
37.8 (83)
29.9 (66)
0.941 (0.694)
0.26
7.0
0.00304-0.00224
0.02117-0.01561
29.9 (66)
37.8 (83)
29.9 (66)
0.941 (0.694)
0.361
16.3
0.00488-0.00360
0.0352-0.0259
36 (79)
43.7 (96)
36 (79)
0.64 (0.47)
C/watt
SYMBOL
UNITS
Horsepower
Kilowatts
Max Operating Speed
Cont. Line Current
Peak Torque
Peak Line Current
(B, BE, EB-60/80x)
(M, ME-60/80x)
DC Res at 25C
(Line-to-Line) 10%
Rotor (B, BE, EB-60/80x)
Inertia (M, ME-60/80x)
Weight (B, BE-60/80x)
(M, ME-60/80x)
(EB-60/80x)
Static Friction
(B, BE, M, ME-60/80x)-(EB-60/80x)
HPRATED
kWRATED
NRATED
NMAX
Tcs
Tcs
Ics
Tps
Ips
Z
hp
kW
rpm
rpm
N-m (lb-ft)
N-m (lb-ft)
ARMS
N-m (lb-ft)
ARMS
rad/sec2
Kt
Kb
VMAX
N-m (lb-ft)/ARMS
VRMS / krpm
VRMS
Rm
Lm
Jm
ohms
mH
kg-m2
lb-ft-sec2
kg (lb)
Wt
Tf
TCT
N-m (lb-ft)
minutes
28-12
33-14
33-14
33-14
38-16
38-16
38-16
Fi
Km
Rth
N-m (lb-ft)/krpm
N-m (lb-ft)/ watts
0.072 (0.053)
1.22 (0.902)
0.34
6
0.109 (0.080)
1.92 (1.42)
0.28
6
0.109 (0.080)
1.94 (1.43)
0.28
6
0.109 (0.080)
1.93 (1.42)
0.26
6
0.147 (0.108)
2.74 (2.02)
0.26
6
0.147 (0.108)
2.59 (1.91)
0.26
6
0.147 (0.108)
2.57 (1.89)
0.22
6
B-802-B
BE-802-B
M-803-B
ME-803-B
EB-802-B
B-802-C
BE-802-C
M-803-C
ME-803-C
EB-802-C
B-804-A
BE-804-A
M-805-A
ME-805-A
EB-804-A
B-804-B
BE-804-B
M-805-B
ME-805-B
EB-804-B
B-804-C
BE-804-C
M-805-C
ME-805-C
EB-804-C
B-806-A
BE-806-A
M-807-A
ME-807-A
EB-806-A
B-806-B
BE-806-B
M-807-B
ME-807-B
EB-806-B
13.6
10.1
2750
2750
40.7 (30)
43.1 (31.8)
32.4
129 (95.3)
108
8.2
6.1
1600
1600
40.7 (30)
43.1 (31.8)
18.9
129 (95.3)
60
14.5
10.8
1500
1500
78.6 (50)
83.4 (61.5)
35
232 (171)
109
18.6
13.9
2000
2000
78.6 (50)
83.4 (61.5)
48
230 (170)
147
21.0
15.7
3000
3000
78.6 (50)
83.4 (61.5)
70
232 (171)
217
20.5
15.3
1550
1550
109 (80)
115 (84.8)
49.1
323 (238)
154
26500
3680
1.26 (0.93)
76
250
26500
3680
2.16 (1.59)
130
250
27600
3870
2.25 (1.66)
136
250
27400
3850
1.65 (1.21)
99.6
250
25600
3630
1.13 (0.83)
68.1
250
0.200
9.4
0.00488-0.00360
0.0352-0.0259
36 (79)
43.7 (96)
36 (79)
0.64 (0.47)
0.568
25.7
0.00488-0.00360
0.0352-0.0259
36 (79)
43.7 (96)
36 (79)
0.64 (0.47)
0.230
13
0.00840-0.00620
0.05990-0.04415
50.6 (112)
62.8 (138)
50.6 (112)
0.91 (0.67)
0.129
7.2
0.00840-0.00620
0.05990-0.04415
50.6 (112)
62.8 (138)
50.6 (112)
0.91 (0.67)
0.058
3.2
0.00840-0.00620
0.05990-0.04415
50.6 (112)
62.8 (138)
50.6 (112)
0.91 (0.67)
C/watt
PARAMETER
SYMBOL
UNITS
Horsepower
Kilowatts
Max Operating Speed
Cont. Line Current
Peak Torque
Peak Line Current
(B, BE, EB-80x)
(M, ME-80x)
DC Res at 25C
(Line-to-Line) 10%
Rotor (B, BE, EB-80x)
Inertia (M, ME-80x)
Weight (B, BE-80x)
(M, ME-80x)
(EB-80x)
Static Friction
(B, BE, M, ME-80x)-(EB-80x)
HPRATED
kWRATED
NRATED
NMAX
Tcs
Tcs
Ics
Tps
Ips
Z
hp
kW
rpm
rpm
N-m (lb-ft)
N-m (lb-ft)
ARMS
N-m (lb-ft)
ARMS
rad/sec2
Kt
Kb
VMAX
N-m (lb-ft)/ARMS
VRMS / krpm
VRMS
Rm
Lm
Jm
ohms
mH
kg-m2
lb-ft-sec2
kg (lb)

Number of Poles
B-404-C
BE-404-C
M-405-C
ME-405-C
EB-404-C
Tf
TCT
PARAMETER

Number of Poles
B/M 80X SERIES MOTORS
Wt
B-404-B
BE-404-B
M-405-B
ME-405-B
EB-404-B
Wt
40-0.60
0.237-0.175
2.43-1.79
0.23
6
B-806-C
BE-806-C
M-807-C
ME-807-C
EB-806-C
B-808-C
21.7
16.2
3000
3000
111 (82)
117.9 (86.9)
94
327 (241)
291
12.2
9.1
900
900
109 (80)
115 (84.8)
30
362 (267)
100
20.3
15.1
1000
1000
149 (110)
158 (117)
47.4
422 (311)
134
25900
3680
2.21 (1.63)
134
250
25900
3680
1.18 (0.87)
71.6
250
28700
4080
3.62 (2.67)
219
250
2510
n/a
3.15 (2.32)
191
250
0.130
8.0
0.0126-0.0093
0.0888-0.0655
67 (147)
86.5 (190)
91 (200)
1.38 (1.02)
0.034
2.1
0.0126-0.0093
0.0888-0.0655
67 (147)
86.5 (190)
91 (200)
1.38 (1.02)
0.340
20
0.0126-0.0093
0.0888-0.0655
67 (147)
86.5 (190)
91 (200)
1.38 (1.02)
0.168
8.05
0.0168-0.0124
n/a-n/a
82 (180)
n/a-n/a
n/a-n/a
1.76 (1.30)
Tf
TCT
N-m (lb-ft)
minutes
40-0.60
40-0.60
48-0.70
48-0.70
48-0.70
55-0.80
55-0.80
Fi
Km
Rth
N-m (lb-ft)/krpm
N-m (lb-ft)/ watts
0.237 (.175)
2.43 (1.79)
0.24
6
0.237 (0.175)
2.47 (1.82)
0.25
6
0.300 (0.221)
4.06 (3.00)
0.18
6
0.300 (0.221)
3.96 (2.92)
0.17
6
.300 (0.221)
4.05 (2.98)
0.18
6
0.362 (0.267)
5.30 (3.91)
0.16
6
0.362 (0.267)
5.55 (4.10)
0.17
6
C/watt
38-16
0.147 (0.108)
2.55 (1.88)
0.25
6
55-0.80
0.362 (0.267)
5.37 (3.96)
0.17
6
60-n/a
0.358-0.264
6.64-4.90
0.14
6
www.DanaherMotion.com 815-226-2222
B/M 40X & 60X SERIES MOTORS
B/M MOTORS
B-404-A
BE-404-A
M-405-A
ME-405-A
EB-404-A
41
Kollmorgen GOLDLINE
DIMENSIONS
B, M-80x-x-A3
14.00 (.551) THRU
4 PL. EQ. SPACED ON
A 215.00 (8.464) B.C.
A .10 TIR
(.004)
58.50
57.50
(2.303)
(2.264)
See page 41 for connector details

Resolver & thermostat
Receptacle
122.5
118.5
(4.82)
(4.67)
192.4
190.0
(7.57)
(7.48)
19.0 (.75)
4.0 (.16)
180.014
179.989
(7.0871)
(7.0862)
A .10 TIR
(.004)
10.000 (.3937)
9.964 (.3923)
.050
(.002)
14.000 (.5512)
13.957 (.5495)
42.50 (1.673)
42.30 (1.665)
B-802
32.018
31.998
(1.2606)
(1.2598)
B-806
M-80X
Notes:
1. BE and ME outline and dimension data and connector information
is available by contacting the Kollmorgen Customer Support Network.
2. Dimensions in mm (inches)
Tolerances,unless otherwise specified:
metric: X decimal place .4, XX decimal places .13
inches: XX decimal places .015, XXX decimal places .005
40
(.3932)
9.987
9.962 (.3922)
-A-
SHAFT DETAIL
27.00 (1.063)
26.80 (1.055)
17.5 (.69)
"B" 2.0 (.08)
39.00
(1.535)
MIN
Motor receptacle
196.0
(7.72)
MAX.
WITH OPTIONAL
MATING PLUGS
"A" MAX.
48.018
47.998
(1.8905)
(1.8897)
B-802
B-804
B-806
B-808
M-803
M-805
M-807
A
360.4 (14.19)
449.9 (17.71)
539.4 (21.24)
628.9 (24.76)
449.9 (17.71)
539.4 (21.24)
648.7 (25.54)
B
300.5 (11.83)
390.0 (15.35)
479.5 (18.88)
569.0 (22.40)
390.0 (15.35)
479.5 (18.88)
588.8 (23.18)
KOLLMORGEN Radford, Virginia 1-800-77 SERVO
B.2
Gearheads
150
DuraTRUE 90 Size 90
Right Angle Gearheads
138
(5.44)
Metric
40
(1.57)
93
(3.67)
12
(.47)
6
(.236)
80 h7
(3.1496/3.1484)
5
(.197)
20 f6
(.7866/.7861)
22.5
(.886)
90 SQ.
(3.54)
20
(1.10)
L
K
6.5 (.256) THRU HOLES

4 PL. EQUALLY SPACED ON A 100
(3.937) BOLT CIRCLE)
3
(.118)
90 SQ.
(3.54)
4
(.158)
AD*
Ratio1
5:1 to 50:1
60:1 to 500:1
Dimension K Dimension L Backlash

mm [in]
mm [in]
[arc-min]
101 [3.99]
124 [4.89]
146 [5.76]
169 [6.66]
9 max
9 max
Weight
Efficiency
4,8 [10]
5,5 [12]
93%
88%
kg [lb]
All dimensions are: mm (inch)

AD** = Adapter length
Adapter length will vary depending on motor.
Efficiency is calculated at 100% of the rated torque.
Performance Specifications
10000 Hour Life
Part
Number
DTR90-005
DTR90-006
DTR90-009
DTR90-010
DTR90-012
DTR90-015
DTR90-020
DTR90-025
DTR90-030
DTR90-040
DTR90-050
DTR90-060
DTR90-075
DTR90-090
DTR90-100
DTR90-120
DTR90-125
DTR90-150
DTR90-200
DTR90-250
DTR90-300
DTR90-400
DTR90-500
Ratio
5:1
6:1
9:1
10:1
12:1
15:1
20:1
25:1
30:1
40:1
50:1
60:1
75:1
90:1
100:1
120:1
1251
150:1
200:1
250:1
300:1
400:1
500:1
20000 Hour Life
Tr
Tr
Tr
(1000 rpm) (3000 rpm) (5000 rpm)
Nm [in-lb]
Nm [in-lb]
Nm [in-lb]
79 [702]
82 [722]
75 [665]
79 [703]
57 [509]
82 [723]
83 [737]
71 [624]
64 [568]
67 [590]
69 [608]
106 [941]
98 [870]
98 [866]
103 [916]
75 [663]
85 [751]
106 [942]
109 [960]
92 [813]
84 [740]
87 [769]
90 [792]
58 [515]
61 [544]
69 [614]
68 [599]
57 [509]
75 [664]
77 [681]
71 [624]
55 [488]
58 [509]
59 [525]
80 [709]
90 [799]
90 [800]
88 [780]
75 [663]
85 [751]
98 [865]
100 [887]
92 [813]
72 [636]
75 [663]
77 [684]
50 [442]
53 [467]
60 [527]
58 [514]
57 [509]
66 [580]
72 [633]
71 [624]
51 [452]
53 [472]
55 [488]
69 [608]
79 [698]
78 [687]
76 [670]
75 [663]
85 [751]
85 [756]
93 [825]
92 [813]
67 [589]
69 [615]
72 [636]
Ratios are exact, higher ratios and other custom options are also
available, consult factory.
Tr = Rated output torque at rated speed for specific hours of life.
32
Tpeak
Nm [in-lb]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
144 [1272]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
144 [1272]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
167 [1479]
Tr
Tr
Tr
(1000 rpm) (3000 rpm) (5000 rpm)
Nm [in-lb]
Nm [in-lb]
Nm [in-lb]
66 [582]
69 [614]
75 [665]
76 [676]
57 [509]
80 [707]
81 [720]
71 [624]
59 [526]
62 [546]
63 [562]
90 [800]
96 [851]
98 [866]
100 [881]
75 [663]
85 [751]
98 [866]
108 [938]
92 [813]
77 [685]
80 [711]
83 [732]
47 [418]
50 [442]
56 [499]
55 [486]
57 [509]
62 [549]
68 [599]
71 [624]
51 [452]
53 [471]
55 [486]
65 [576]
75 [661]
73 [650]
72 [633]
75 [663]
85 [751]
81 [715]
88 [780]
92 [813]
67 [589]
69 [614]
72 [633]
41 [359]
43 [379]
48 [428]
47 [417]
53 [467]
53 [471]
58 [514]
62 [549]
47 [418]
49 [437]
51 [452]
56 [494]
64 [567]
63 [558]
61 [543]
69 [608]
75 [661]
69 [614]
76 [670]
81 [715]
62 [545]
64 [569]
67 [589]
kg-cm
[in-lb-sec2x10-4]
2
1,64 [14.5]
1,63 [14.4]
1,39 [12.3]
1,62 [14.3]
1,29 [11.4]
1,26 [11.1]
1,28 [11.3]
1,26 [11.1]
1,38 [12.2]
1,28 [11.3]
1,25 [11.1]
1,63 [14.4]
1,39 [12.3]
1,39 [12.3]
1,62 [14.3]
1,29 [11.4]
1,26 [11.1]
1,39 [12.3]
1,61 [14.2]
1,26 [11.1]
1,38 [12.2]
1,28 [11.3]
1,20 [10.6]
Torsional
Stiffness
Nm/arc-min
[in-lb/arc-min]
7,8 [69.2]
5,5 [48.2]
5,5 [48.8]
5,5 [48.4]
5,5 [49.0]
5,5 [49.0]
5,5 [48.7]
5,5 [48.7]
4,0 [35.6]
4,0 [35.6]
4,0 [35.6]
5,5 [48.6]
5,5 [48.4]
5,4 [48.0]
5,4 [47.8]
5,5 [48.5]
5,8 [51.6]
5,5 [48.3]
5,5 [48.5]
5,8 [51.5]
4,0 [35.5]
4,0 [35.5]
4,0 [35.6]
Tpeak = Allowable momentary peak torque for emergency stop or heavy shock loading.
J = Mass moment of inertia reflected to the input shaft (including pinion assembly).
www.danahermotion.com
DuraTRUE 90 Size 60
Right Angle Gearheads
98.5
(3.88)
50 h7
(1.9685/1.9675)
25
(.98)
68
(2.68)
10
(.39)
Metric
5
(.197)
3
(.118)
16 j6
(.6302/.6298)
18
(.709)
60 SQ.
(2.36)
16
(.63)
L
K
5 (.197) THRU HOLES 4 PL.

EQUALLY SPACED ON A 70 (2.756)
BOLT CIRCLE)
2.5
(.098)
61 SQ.
(2.40)
3.5
(.138)
AD*
Ratio1
5:1 to 50:1
60:1 to 500:1
Dimension K Dimension L Backlash

mm [in]
mm [in]
[arc-min]
79 [3.11]
96 [3.79]
109,5 [4.31]
127 [4.99]
9 max
9 max
Weight
Efficiency
2,5 [5.5]
2,7 [6]
93%
88%
kg [lb]
All dimensions are: mm (inch)

AD** = Adapter length
Adapter length will vary depending on motor.
Efficiency is calculated at 100% of the rated torque.
Performance Specifications
10000 Hour Life
Part
Number
DTR60-005
DTR60-006
DTR60-009
DTR60-010
DTR60-012
DTR60-015
DTR60-020
DTR60-025
DTR60-030
DTR60-040
DTR60-050
DTR60-060
DTR60-075
DTR60-090
DTR60-100
DTR60-120
DTR60-125
DTR60-150
DTR60-200
DTR60-250
DTR60-300
DTR60-400
DTR60-500
Ratio
5:1
6:1
9:1
10:1
12:1
15:1
20:1
25:1
30:1
40:1
50:1
60:1
75:1
90:1
100:1
120:1
1251
150:1
200:1
250:1
300:1
400:1
500:1
20000 Hour Life
Tr
Tr
Tr
(1000 rpm) (3000 rpm) (5000 rpm)
Nm [in-lb]
Nm [in-lb]
Nm [in-lb]
18 [157]
19 [166]
21 [187]
21 [182]
18 [161]
23 [206]
24 [211]
24 [214]
18 [155]
18 [161]
19 [165]
28 [244]
28 [248]
28 [244]
27 [237]
24 [210]
29 [257]
30 [268]
31 [275]
32 [279]
23 [202]
24 [210]
24 [215]
13 [113]
13 [119]
15 [134]
15 [131]
17 [147]
17 [148]
18 [161]
19 [172]
15 [134]
16 [139]
16 [143]
18 [155]
20 [178]
20 [175]
19 [171]
22 [192]
23 [207]
22 [193]
24 [210]
25 [224]
20 [175]
20 [181]
21 [186]
11 [97]
12 [102]
13 [115]
13 [112]
14 [126]
14 [127]
16 [138]
17 [148]
14 [124]
15 [130]
15 [134]
15 [133]
17 [153]
17 [150]
16 [146]
19 [164]
20 [178]
19 [165]
20 [180]
22 [193]
18 [162]
19 [169]
20 [175]
Ratios are exact, higher ratios and other custom options are also
available, consult factory.
Tr = Rated output torque at rated speed for specific hours of life.
30
Tpeak
Nm [in-lb]
52 [460]
52 [460]
52 [460]
49 [435]
45 [402]
50 [445]
51 [450]
51 [455]
49 [430]
49 [430]
49 [430]
52 [460]
50 [445]
52 [460]
49 [435]
45 [402]
51 [455]
50 [445]
51 [450]
51 [455]
49 [430]
49 [430]
49 [430]
Tr
Tr
Tr
(1000 rpm) (3000 rpm) (5000 rpm)
Nm [in-lb]
Nm [in-lb]
Nm [in-lb]
14 [127]
15 [134]
17 [152]
17 [148]
18 [161]
19 [167]
21 [182]
22 [195]
16 [143]
17 [149]
17 [153]
20 [175]
23 [201]
22 [198]
22 [193]
24 [210]
27 [235]
22 [198]
27 [237]
29 [254]
21 [186]
22 [194]
23 [199]
10 [92]
11 [97]
12 [109]
12 [106]
13 [119]
14 [120]
15 [131]
16 [140]
14 [124]
15 [129]
15 [133]
14 [126]
16 [144]
16 [142]
16 [138]
18 [155]
19 [168]
18 [156]
19 [171]
21 [182]
18 [162]
19 [168]
20 [173]
9 [79]
9 [83]
11 [94]
10 [91]
12 [102]
12 [103]
13 [112]
14 [120]
13 [115]
14 [120]
14 [124]
12 [108]
14 [124]
14 [122]
13 [119]
15 [133]
16 [144]
15 [134]
16 [146]
18 [156]
17 [150]
18 [156]
18 [162]
kg-cm
[in-lb-sec2x10-4]
2
0,42 [3.7]
0,42 [3.7]
0,36 [3.1]
0,41 [3.5]
0,33 [3.0]
0,32 [2.9]
0,33 [2.8]
0,32 [2.7]
0,35 [2.8]
0,33 [2.7]
0,32 [2.6]
0,42 [3.7]
0,35 [3.0]
0,36 [3.1]
0,41 [3.5]
0,33 [3.0]
0,32 [2.8]
0,35 [2.9]
0,41 [3.4]
0,32 [2.7]
0,35 [2.8]
0,33 [2.7]
0,30 [2.7]
Torsional
Stiffness
Nm/arc-min
[in-lb/arc-min]
2,3 [20.1]
2,3 [20.3]
2,3 [20.2]
2,3 [20.2]
2,3 [20.4]
2,3 [20.4]
2,3 [20.3]
2,3 [20.3]
1,7 [14.8]
1,7 [14.8]
1,7 [14.8]
2,3 [20.2]
2,3 [20.2]
2,3 [20.0]
2,2 [19.9]
2,3 [20.2]
2,4 [21.5]
2,3 [20.1]
2,3 [20.2]
2,4 [21.5]
1,7 [14.8]
1,7 [14.8]
1,7 [14.8]
Tpeak = Allowable momentary peak torque for emergency stop or heavy shock loading.
J = Mass moment of inertia reflected to the input shaft (including pinion assembly).
www.danahermotion.com

Darrin Willis Masters Thesis Corrected Final

Uploaded by

Copyright:

Available Formats

Darrin Willis Masters Thesis Corrected Final

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Darrin Willis Masters Thesis Corrected Final

Uploaded by

Copyright:

Available Formats

DESIGN AND DEVELOPMENT OF A NOVEL

LIGHTWEIGHT LONG-REACH COMPOSITE

Scott Nokleby, Faculty of Engineering & Applied Science

THE UNIVERSITY OF ONTARIO INSTITUTE OF TECHNOLOGY

DESIGN AND DEVELOPMENT OF A NOVEL LIGHTWEIGHT

Problem Definition and Goal of Research . . . . . . . . . . . . . . . .

Typical Properties of Traditional Commercial Robotic Arms . . . . .

Review of Polymer Composites . . . . . . . . . . . . . . . . . . . . .

Fiber-Polymer Composite Production Processes . . . . . . . .

Methods of Strengthening Hollow Composites . . . . . . . . .

Review of Literature on Composite Robots . . . . . . . . . . . . . . .

Composite Arms Used in Industry . . . . . . . . . . . . . . . .

Manipulator Layout Selection . . . . . . . . . . . . . . . . . . . . . .

Comparison of Various Types of Robotic Arms . . . . . . . . .

Serial Arm Configuration . . . . . . . . . . . . . . . . . . . . .

Analysis of the Cross-Section of the Two Main Links . . . . . . . . .

Cross-Section Testing Procedure . . . . . . . . . . . . . . . . .

Results of Cross-Section Analysis for the Main Links . . . . .

Basic Shaped Hollow Tubes . . . . . . . . . . . . . .

Circular Tubes with Support Ribs . . . . . . . . . .

Circular Tube with Different Shaped Inner Tubes . .

Annulus with Different Sized Inner Tubes . . . . . .

Rectangular Tubes with Support Ribs . . . . . . . .

Analysis of Tapering the Main Links . . . . . . . . . . . . . .

Optimization of the Cross-Section . . . . . . . . . . . . . . . . . . . .

Determination of the Objective Function . . . . . . . . . . . .

Design of the Proof-of-Concept Prototype . . . . . . . . . . .

Formulation of the Equations for the Cross-Sectional Area . .

Formulation of the Equations for Moment of Inertia . . . . . .

Formulation of the Equations for Angle of Twist . . . . . . . .

Determination of Bounds for the Design Variables . . . . . . .

Solutions Found from Optimization of Cross-Section . . . . . .

I-beam Supported Rectangular Link . . . . . . . . .

X-Supported Rectangular Link . . . . . . . . . . . .

Study Without Upper Bounds . . . . . . . . . . . . .

Recommendations from Optimization Results . . . .

3 Prototype Link Construction

Modular Construction Process . . . . . . . . . . . . . . . . . . . . . .

Phase 1: Wooden Mold . . . . . . . . . . . . . . . . . . . . . .

Phase 2: Four Piece Fully-Detachable Aluminum Mold . . . .

Phase 3: Two Piece Fully-Detachable Aluminum Mold . . . .

Prototype Link Construction . . . . . . . . . . . . . . . . . . . . . . .

Materials Used for Prototype . . . . . . . . . . . . . . . . . .

Ply Lay-up Pattern for Prototype . . . . . . . . . . . . . . . .

Link Attachment Bracket . . . . . . . . . . . . . . . . . . . . . . . . .

4 Prototype Link Testing

Physical Prototype Testing . . . . . . . . . . . . . . . . . . . . . . . .

Single Link Static Testing . . . . . . . . . . . . . . . . . . . .

Weight Support Component . . . . . . . . . . . . . .

Repeatability Testing of the 2-DOF Prototype . . . . . . . . .

Three Point Bending Test . . . . . . . . . . . . . . .

Finite Element Analysis . . . . . . . . . . . . . . . . . . . . . . . . .

Comparison of Physical and Virtual Prototype . . . . . . . . . . . . .

5 Design of the Full-Scale Robotic Arm

Dimensional Analysis of Links . . . . . . . . . . . . . . . . . . . . . .

Links with Height Twice the Width . . . . . . . . . .

Relationship of Thickness and Static Deflection . . .

Links with Optimized Height-to-Width Ratio . . . .

Different Layering Patterns . . . . . . . . . . . . . .